diff --git a/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost.ll b/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost.ll
index eabc2330ddc6..8c94d315f355 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost.ll
@@ -1,15 +1,16 @@
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+sse2 -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=SSE2
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+sse4.1 -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=SSE41
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx2 -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX2
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+xop,+avx -cost-model -analyze | FileCheck %s --check-prefix=CHECK -check-prefix=XOP --check-prefix=XOPAVX
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+xop,+avx2 -cost-model -analyze | FileCheck %s --check-prefix=CHECK -check-prefix=XOP --check-prefix=XOPAVX2
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512f -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512 --check-prefix=AVX512F
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512f,+avx512dq -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512 --check-prefix=AVX512F
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512f,+avx512bw -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512 --check-prefix=AVX512BW
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512f,+avx512vl -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512 --check-prefix=AVX512F
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512f,+avx512dq,+avx512vl -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512 --check-prefix=AVX512F
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512f,+avx512bw,+avx512vl -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512 --check-prefix=AVX512BW
+; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+sse2 | FileCheck %s --check-prefixes=CHECK,SSE,SSE2
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+sse4.1 | FileCheck %s --check-prefixes=CHECK,SSE,SSE41
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX2
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+xop,+avx | FileCheck %s --check-prefixes=CHECK,XOP,XOPAVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+xop,+avx2 | FileCheck %s --check-prefixes=CHECK,XOP,XOPAVX2
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx512f | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512F
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx512f,+avx512dq | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512F
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512BW
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx512f,+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512F
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx512f,+avx512dq,+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512VL
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx512f,+avx512bw,+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512BWVL
 
 ; Verify the cost of vector arithmetic shift right instructions.
 
@@ -18,153 +19,398 @@
 ;
 
 define <2 x i64> @var_shift_v2i64(<2 x i64> %a, <2 x i64> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v2i64':
-; SSE2: Found an estimated cost of 12 for instruction:   %shift
-; SSE41: Found an estimated cost of 12 for instruction:   %shift
-; AVX: Found an estimated cost of 12 for instruction:   %shift
-; AVX2: Found an estimated cost of 4 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'var_shift_v2i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <2 x i64> %a, %b
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; AVX1-LABEL: 'var_shift_v2i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <2 x i64> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; AVX2-LABEL: 'var_shift_v2i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <2 x i64> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; XOP-LABEL: 'var_shift_v2i64'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <2 x i64> %a, %b
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; AVX512-LABEL: 'var_shift_v2i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <2 x i64> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
   %shift = ashr <2 x i64> %a, %b
   ret <2 x i64> %shift
 }
 
 define <4 x i64> @var_shift_v4i64(<4 x i64> %a, <4 x i64> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v4i64':
-; SSE2: Found an estimated cost of 24 for instruction:   %shift
-; SSE41: Found an estimated cost of 24 for instruction:   %shift
-; AVX: Found an estimated cost of 26 for instruction:   %shift
-; AVX2: Found an estimated cost of 4 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 6 for instruction:   %shift
+; SSE-LABEL: 'var_shift_v4i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <4 x i64> %a, %b
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX1-LABEL: 'var_shift_v4i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %shift = ashr <4 x i64> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX2-LABEL: 'var_shift_v4i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <4 x i64> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOP-LABEL: 'var_shift_v4i64'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <4 x i64> %a, %b
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX512-LABEL: 'var_shift_v4i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <4 x i64> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
   %shift = ashr <4 x i64> %a, %b
   ret <4 x i64> %shift
 }
 
 define <8 x i64> @var_shift_v8i64(<8 x i64> %a, <8 x i64> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v8i64':
-; SSE2: Found an estimated cost of 48 for instruction:   %shift
-; SSE41: Found an estimated cost of 48 for instruction:   %shift
-; AVX: Found an estimated cost of 52 for instruction:   %shift
-; AVX2: Found an estimated cost of 8 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 12 for instruction:   %shift
+; SSE-LABEL: 'var_shift_v8i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = ashr <8 x i64> %a, %b
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX1-LABEL: 'var_shift_v8i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %shift = ashr <8 x i64> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX2-LABEL: 'var_shift_v8i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <8 x i64> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOP-LABEL: 'var_shift_v8i64'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <8 x i64> %a, %b
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX512-LABEL: 'var_shift_v8i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i64> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
   %shift = ashr <8 x i64> %a, %b
   ret <8 x i64> %shift
 }
 
 define <4 x i32> @var_shift_v4i32(<4 x i32> %a, <4 x i32> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v4i32':
-; SSE2: Found an estimated cost of 16 for instruction:   %shift
-; SSE41: Found an estimated cost of 12 for instruction:   %shift
-; AVX: Found an estimated cost of 12 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 2 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v4i32'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %shift = ashr <4 x i32> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; SSE41-LABEL: 'var_shift_v4i32'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <4 x i32> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; AVX1-LABEL: 'var_shift_v4i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <4 x i32> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; AVX2-LABEL: 'var_shift_v4i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <4 x i32> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; XOPAVX1-LABEL: 'var_shift_v4i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <4 x i32> %a, %b
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; XOPAVX2-LABEL: 'var_shift_v4i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <4 x i32> %a, %b
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; AVX512-LABEL: 'var_shift_v4i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <4 x i32> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; XOPAVX-LABEL: 'var_shift_v4i32'
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <4 x i32> %a, %b
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
   %shift = ashr <4 x i32> %a, %b
   ret <4 x i32> %shift
 }
 
 define <8 x i32> @var_shift_v8i32(<8 x i32> %a, <8 x i32> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v8i32':
-; SSE2: Found an estimated cost of 32 for instruction:   %shift
-; SSE41: Found an estimated cost of 24 for instruction:   %shift
-; AVX: Found an estimated cost of 26 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 6 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v8i32'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %shift = ashr <8 x i32> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; SSE41-LABEL: 'var_shift_v8i32'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <8 x i32> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX1-LABEL: 'var_shift_v8i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %shift = ashr <8 x i32> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX2-LABEL: 'var_shift_v8i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i32> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX1-LABEL: 'var_shift_v8i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <8 x i32> %a, %b
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX2-LABEL: 'var_shift_v8i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i32> %a, %b
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX512-LABEL: 'var_shift_v8i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i32> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX-LABEL: 'var_shift_v8i32'
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <8 x i32> %a, %b
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
   %shift = ashr <8 x i32> %a, %b
   ret <8 x i32> %shift
 }
 
 define <16 x i32> @var_shift_v16i32(<16 x i32> %a, <16 x i32> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v16i32':
-; SSE2: Found an estimated cost of 64 for instruction:   %shift
-; SSE41: Found an estimated cost of 48 for instruction:   %shift
-; AVX: Found an estimated cost of 52 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 12 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v16i32'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %shift = ashr <16 x i32> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; SSE41-LABEL: 'var_shift_v16i32'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = ashr <16 x i32> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX1-LABEL: 'var_shift_v16i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %shift = ashr <16 x i32> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX2-LABEL: 'var_shift_v16i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <16 x i32> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX1-LABEL: 'var_shift_v16i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <16 x i32> %a, %b
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX2-LABEL: 'var_shift_v16i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <16 x i32> %a, %b
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX512-LABEL: 'var_shift_v16i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i32> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX-LABEL: 'var_shift_v16i32'
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <16 x i32> %a, %b
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
   %shift = ashr <16 x i32> %a, %b
   ret <16 x i32> %shift
 }
 
 define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v8i16':
-; SSE2: Found an estimated cost of 32 for instruction:   %shift
-; SSE41: Found an estimated cost of 14 for instruction:   %shift
-; AVX: Found an estimated cost of 14 for instruction:   %shift
-; AVX2: Found an estimated cost of 14 for instruction:   %shift
-; AVX512F: Found an estimated cost of 14 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 2 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v8i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %shift = ashr <8 x i16> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; SSE41-LABEL: 'var_shift_v8i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %shift = ashr <8 x i16> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX-LABEL: 'var_shift_v8i16'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %shift = ashr <8 x i16> %a, %b
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; XOP-LABEL: 'var_shift_v8i16'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <8 x i16> %a, %b
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512F-LABEL: 'var_shift_v8i16'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %shift = ashr <8 x i16> %a, %b
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512BW-LABEL: 'var_shift_v8i16'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %b
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512VL-LABEL: 'var_shift_v8i16'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %shift = ashr <8 x i16> %a, %b
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512BWVL-LABEL: 'var_shift_v8i16'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %b
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
   %shift = ashr <8 x i16> %a, %b
   ret <8 x i16> %shift
 }
 
 define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v16i16':
-; SSE2: Found an estimated cost of 64 for instruction:   %shift
-; SSE41: Found an estimated cost of 28 for instruction:   %shift
-; AVX: Found an estimated cost of 30 for instruction:   %shift
-; AVX2: Found an estimated cost of 10 for instruction:   %shift
-; AVX512F: Found an estimated cost of 10 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 6 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v16i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %shift = ashr <16 x i16> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; SSE41-LABEL: 'var_shift_v16i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %shift = ashr <16 x i16> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX1-LABEL: 'var_shift_v16i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <16 x i16> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX2-LABEL: 'var_shift_v16i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <16 x i16> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; XOP-LABEL: 'var_shift_v16i16'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <16 x i16> %a, %b
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512F-LABEL: 'var_shift_v16i16'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <16 x i16> %a, %b
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512BW-LABEL: 'var_shift_v16i16'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %b
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512VL-LABEL: 'var_shift_v16i16'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <16 x i16> %a, %b
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512BWVL-LABEL: 'var_shift_v16i16'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %b
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
   %shift = ashr <16 x i16> %a, %b
   ret <16 x i16> %shift
 }
 
 define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v32i16':
-; SSE2: Found an estimated cost of 128 for instruction:   %shift
-; SSE41: Found an estimated cost of 56 for instruction:   %shift
-; AVX: Found an estimated cost of 60 for instruction:   %shift
-; AVX2: Found an estimated cost of 20 for instruction:   %shift
-; AVX512F: Found an estimated cost of 20 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 12 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v32i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 128 for instruction: %shift = ashr <32 x i16> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; SSE41-LABEL: 'var_shift_v32i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %shift = ashr <32 x i16> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX1-LABEL: 'var_shift_v32i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %shift = ashr <32 x i16> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX2-LABEL: 'var_shift_v32i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = ashr <32 x i16> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; XOP-LABEL: 'var_shift_v32i16'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <32 x i16> %a, %b
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512F-LABEL: 'var_shift_v32i16'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = ashr <32 x i16> %a, %b
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BW-LABEL: 'var_shift_v32i16'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %b
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512VL-LABEL: 'var_shift_v32i16'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = ashr <32 x i16> %a, %b
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BWVL-LABEL: 'var_shift_v32i16'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %b
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
   %shift = ashr <32 x i16> %a, %b
   ret <32 x i16> %shift
 }
 
 define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v16i8':
-; SSE2: Found an estimated cost of 54 for instruction:   %shift
-; SSE41: Found an estimated cost of 24 for instruction:   %shift
-; AVX: Found an estimated cost of 24 for instruction:   %shift
-; AVX2: Found an estimated cost of 24 for instruction:   %shift
-; AVX512: Found an estimated cost of 24 for instruction:   %shift
-; XOP: Found an estimated cost of 2 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v16i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 54 for instruction: %shift = ashr <16 x i8> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; SSE41-LABEL: 'var_shift_v16i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <16 x i8> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX-LABEL: 'var_shift_v16i8'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <16 x i8> %a, %b
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; XOP-LABEL: 'var_shift_v16i8'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <16 x i8> %a, %b
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512-LABEL: 'var_shift_v16i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <16 x i8> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
   %shift = ashr <16 x i8> %a, %b
   ret <16 x i8> %shift
 }
 
 define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v32i8':
-; SSE2: Found an estimated cost of 108 for instruction:   %shift
-; SSE41: Found an estimated cost of 48 for instruction:   %shift
-; AVX: Found an estimated cost of 50 for instruction:   %shift
-; AVX2: Found an estimated cost of 24 for instruction:   %shift
-; AVX512F: Found an estimated cost of 24 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 24 for instruction:   %shift
-; XOP: Found an estimated cost of 6 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v32i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 108 for instruction: %shift = ashr <32 x i8> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; SSE41-LABEL: 'var_shift_v32i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = ashr <32 x i8> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX1-LABEL: 'var_shift_v32i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %shift = ashr <32 x i8> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX2-LABEL: 'var_shift_v32i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <32 x i8> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; XOP-LABEL: 'var_shift_v32i8'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <32 x i8> %a, %b
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512-LABEL: 'var_shift_v32i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <32 x i8> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
   %shift = ashr <32 x i8> %a, %b
   ret <32 x i8> %shift
 }
 
 define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v64i8':
-; SSE2: Found an estimated cost of 216 for instruction:   %shift
-; SSE41: Found an estimated cost of 96 for instruction:   %shift
-; AVX: Found an estimated cost of 100 for instruction:   %shift
-; AVX2: Found an estimated cost of 48 for instruction:   %shift
-; AVX512F: Found an estimated cost of 48 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 24 for instruction:   %shift
-; XOP: Found an estimated cost of 12 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v64i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 216 for instruction: %shift = ashr <64 x i8> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; SSE41-LABEL: 'var_shift_v64i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 96 for instruction: %shift = ashr <64 x i8> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX1-LABEL: 'var_shift_v64i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 100 for instruction: %shift = ashr <64 x i8> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX2-LABEL: 'var_shift_v64i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = ashr <64 x i8> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; XOP-LABEL: 'var_shift_v64i8'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <64 x i8> %a, %b
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512F-LABEL: 'var_shift_v64i8'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = ashr <64 x i8> %a, %b
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BW-LABEL: 'var_shift_v64i8'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <64 x i8> %a, %b
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512VL-LABEL: 'var_shift_v64i8'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = ashr <64 x i8> %a, %b
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BWVL-LABEL: 'var_shift_v64i8'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <64 x i8> %a, %b
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
   %shift = ashr <64 x i8> %a, %b
   ret <64 x i8> %shift
 }
@@ -174,13 +420,30 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ;
 
 define <2 x i64> @splatvar_shift_v2i64(<2 x i64> %a, i64 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v2i64':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 4 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'splatvar_shift_v2i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <2 x i64> undef, i64 %b, i32 0
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <2 x i64> %insert, <2 x i64> undef, <2 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <2 x i64> %a, %splat
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; AVX-LABEL: 'splatvar_shift_v2i64'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <2 x i64> undef, i64 %b, i32 0
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <2 x i64> %insert, <2 x i64> undef, <2 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <2 x i64> %a, %splat
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; XOP-LABEL: 'splatvar_shift_v2i64'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <2 x i64> undef, i64 %b, i32 0
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <2 x i64> %insert, <2 x i64> undef, <2 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <2 x i64> %a, %splat
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v2i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <2 x i64> undef, i64 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <2 x i64> %insert, <2 x i64> undef, <2 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <2 x i64> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
   %insert = insertelement <2 x i64> undef, i64 %b, i32 0
   %splat = shufflevector <2 x i64> %insert, <2 x i64> undef, <2 x i32> zeroinitializer
   %shift = ashr <2 x i64> %a, %splat
@@ -188,13 +451,47 @@ define <2 x i64> @splatvar_shift_v2i64(<2 x i64> %a, i64 %b) {
 }
 
 define <4 x i64> @splatvar_shift_v4i64(<4 x i64> %a, i64 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v4i64':
-; SSE2: Found an estimated cost of 8 for instruction:   %shift
-; SSE41: Found an estimated cost of 8 for instruction:   %shift
-; AVX: Found an estimated cost of 10 for instruction:   %shift
-; AVX2: Found an estimated cost of 4 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 6 for instruction:   %shift
+; SSE-LABEL: 'splatvar_shift_v4i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <4 x i64> %a, %splat
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v4i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <4 x i64> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v4i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <4 x i64> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v4i64'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <4 x i64> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v4i64'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <4 x i64> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v4i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <4 x i64> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOPAVX-LABEL: 'splatvar_shift_v4i64'
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <4 x i64> %a, %splat
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
   %insert = insertelement <4 x i64> undef, i64 %b, i32 0
   %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
   %shift = ashr <4 x i64> %a, %splat
@@ -202,13 +499,47 @@ define <4 x i64> @splatvar_shift_v4i64(<4 x i64> %a, i64 %b) {
 }
 
 define <8 x i64> @splatvar_shift_v8i64(<8 x i64> %a, i64 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v8i64':
-; SSE2: Found an estimated cost of 16 for instruction:   %shift
-; SSE41: Found an estimated cost of 16 for instruction:   %shift
-; AVX: Found an estimated cost of 20 for instruction:   %shift
-; AVX2: Found an estimated cost of 8 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 12 for instruction:   %shift
+; SSE-LABEL: 'splatvar_shift_v8i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %shift = ashr <8 x i64> %a, %splat
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v8i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = ashr <8 x i64> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v8i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <8 x i64> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v8i64'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <8 x i64> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v8i64'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <8 x i64> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v8i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i64> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOPAVX-LABEL: 'splatvar_shift_v8i64'
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <8 x i64> %a, %splat
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
   %insert = insertelement <8 x i64> undef, i64 %b, i32 0
   %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
   %shift = ashr <8 x i64> %a, %splat
@@ -216,13 +547,12 @@ define <8 x i64> @splatvar_shift_v8i64(<8 x i64> %a, i64 %b) {
 }
 
 define <4 x i32> @splatvar_shift_v4i32(<4 x i32> %a, i32 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v4i32':
-; SSE2: Found an estimated cost of 1 for instruction:   %shift
-; SSE41: Found an estimated cost of 1 for instruction:   %shift
-; AVX: Found an estimated cost of 1 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 1 for instruction:   %shift
+; CHECK-LABEL: 'splatvar_shift_v4i32'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i32> undef, i32 %b, i32 0
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i32> %insert, <4 x i32> undef, <4 x i32> zeroinitializer
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <4 x i32> %a, %splat
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
   %insert = insertelement <4 x i32> undef, i32 %b, i32 0
   %splat = shufflevector <4 x i32> %insert, <4 x i32> undef, <4 x i32> zeroinitializer
   %shift = ashr <4 x i32> %a, %splat
@@ -230,14 +560,47 @@ define <4 x i32> @splatvar_shift_v4i32(<4 x i32> %a, i32 %b) {
 }
 
 define <8 x i32> @splatvar_shift_v8i32(<8 x i32> %a, i32 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v8i32':
-; SSE2: Found an estimated cost of 2 for instruction:   %shift
-; SSE41: Found an estimated cost of 2 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 6 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE-LABEL: 'splatvar_shift_v8i32'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <8 x i32> %a, %splat
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v8i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <8 x i32> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v8i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i32> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v8i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <8 x i32> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v8i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i32> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v8i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i32> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX-LABEL: 'splatvar_shift_v8i32'
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <8 x i32> %a, %splat
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
   %insert = insertelement <8 x i32> undef, i32 %b, i32 0
   %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
   %shift = ashr <8 x i32> %a, %splat
@@ -245,14 +608,47 @@ define <8 x i32> @splatvar_shift_v8i32(<8 x i32> %a, i32 %b) {
 }
 
 define <16 x i32> @splatvar_shift_v16i32(<16 x i32> %a, i32 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v16i32':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 8 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 12 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'splatvar_shift_v16i32'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i32> %a, %splat
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v16i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <16 x i32> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v16i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <16 x i32> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v16i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <16 x i32> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v16i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <16 x i32> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v16i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i32> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX-LABEL: 'splatvar_shift_v16i32'
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <16 x i32> %a, %splat
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
   %insert = insertelement <16 x i32> undef, i32 %b, i32 0
   %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
   %shift = ashr <16 x i32> %a, %splat
@@ -260,13 +656,36 @@ define <16 x i32> @splatvar_shift_v16i32(<16 x i32> %a, i32 %b) {
 }
 
 define <8 x i16> @splatvar_shift_v8i16(<8 x i16> %a, i16 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v8i16':
-; SSE2: Found an estimated cost of 1 for instruction:   %shift
-; SSE41: Found an estimated cost of 1 for instruction:   %shift
-; AVX: Found an estimated cost of 1 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 1 for instruction:   %shift
+; SSE2-LABEL: 'splatvar_shift_v8i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %splat
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; SSE41-LABEL: 'splatvar_shift_v8i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %splat
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX-LABEL: 'splatvar_shift_v8i16'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %splat
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; XOP-LABEL: 'splatvar_shift_v8i16'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %splat
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v8i16'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
   %insert = insertelement <8 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
   %shift = ashr <8 x i16> %a, %splat
@@ -274,14 +693,53 @@ define <8 x i16> @splatvar_shift_v8i16(<8 x i16> %a, i16 %b) {
 }
 
 define <16 x i16> @splatvar_shift_v16i16(<16 x i16> %a, i16 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v16i16':
-; SSE2: Found an estimated cost of 2 for instruction:   %shift
-; SSE41: Found an estimated cost of 2 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 6 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE2-LABEL: 'splatvar_shift_v16i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <16 x i16> %a, %splat
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; SSE41-LABEL: 'splatvar_shift_v16i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <16 x i16> %a, %splat
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v16i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i16> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v16i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v16i16'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <16 x i16> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v16i16'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v16i16'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; XOPAVX-LABEL: 'splatvar_shift_v16i16'
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <16 x i16> %a, %splat
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
   %insert = insertelement <16 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
   %shift = ashr <16 x i16> %a, %splat
@@ -289,15 +747,71 @@ define <16 x i16> @splatvar_shift_v16i16(<16 x i16> %a, i16 %b) {
 }
 
 define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v32i16':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 8 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512F: Found an estimated cost of 2 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 12 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE2-LABEL: 'splatvar_shift_v32i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <32 x i16> %a, %splat
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; SSE41-LABEL: 'splatvar_shift_v32i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <32 x i16> %a, %splat
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v32i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <32 x i16> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v32i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <32 x i16> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v32i16'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <32 x i16> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v32i16'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <32 x i16> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512F-LABEL: 'splatvar_shift_v32i16'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <32 x i16> %a, %splat
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BW-LABEL: 'splatvar_shift_v32i16'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %splat
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512VL-LABEL: 'splatvar_shift_v32i16'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <32 x i16> %a, %splat
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BWVL-LABEL: 'splatvar_shift_v32i16'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %splat
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; XOPAVX-LABEL: 'splatvar_shift_v32i16'
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <32 x i16> %a, %splat
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
   %shift = ashr <32 x i16> %a, %splat
@@ -305,13 +819,36 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 }
 
 define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v16i8':
-; SSE2: Found an estimated cost of 54 for instruction:   %shift
-; SSE41: Found an estimated cost of 24 for instruction:   %shift
-; AVX: Found an estimated cost of 24 for instruction:   %shift
-; AVX2: Found an estimated cost of 24 for instruction:   %shift
-; AVX512: Found an estimated cost of 24 for instruction:   %shift
-; XOP: Found an estimated cost of 2 for instruction:   %shift
+; SSE2-LABEL: 'splatvar_shift_v16i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 54 for instruction: %shift = ashr <16 x i8> %a, %splat
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; SSE41-LABEL: 'splatvar_shift_v16i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <16 x i8> %a, %splat
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX-LABEL: 'splatvar_shift_v16i8'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <16 x i8> %a, %splat
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; XOP-LABEL: 'splatvar_shift_v16i8'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <16 x i8> %a, %splat
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v16i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <16 x i8> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
   %insert = insertelement <16 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
   %shift = ashr <16 x i8> %a, %splat
@@ -319,13 +856,53 @@ define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
 }
 
 define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v32i8':
-; SSE2: Found an estimated cost of 108 for instruction:   %shift
-; SSE41: Found an estimated cost of 48 for instruction:   %shift
-; AVX: Found an estimated cost of 50 for instruction:   %shift
-; AVX2: Found an estimated cost of 24 for instruction:   %shift
-; AVX512: Found an estimated cost of 24 for instruction:   %shift
-; XOP: Found an estimated cost of 6 for instruction:   %shift
+; SSE2-LABEL: 'splatvar_shift_v32i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 108 for instruction: %shift = ashr <32 x i8> %a, %splat
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; SSE41-LABEL: 'splatvar_shift_v32i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = ashr <32 x i8> %a, %splat
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v32i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %shift = ashr <32 x i8> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v32i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <32 x i8> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v32i8'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <32 x i8> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v32i8'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <32 x i8> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v32i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <32 x i8> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; XOPAVX-LABEL: 'splatvar_shift_v32i8'
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <32 x i8> %a, %splat
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
   %insert = insertelement <32 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
   %shift = ashr <32 x i8> %a, %splat
@@ -333,14 +910,71 @@ define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 }
 
 define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v64i8':
-; SSE2: Found an estimated cost of 216 for instruction:   %shift
-; SSE41: Found an estimated cost of 96 for instruction:   %shift
-; AVX: Found an estimated cost of 100 for instruction:   %shift
-; AVX2: Found an estimated cost of 48 for instruction:   %shift
-; AVX512F: Found an estimated cost of 48 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 24 for instruction:   %shift
-; XOP: Found an estimated cost of 12 for instruction:   %shift
+; SSE2-LABEL: 'splatvar_shift_v64i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 216 for instruction: %shift = ashr <64 x i8> %a, %splat
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; SSE41-LABEL: 'splatvar_shift_v64i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 96 for instruction: %shift = ashr <64 x i8> %a, %splat
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v64i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 100 for instruction: %shift = ashr <64 x i8> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v64i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = ashr <64 x i8> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v64i8'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <64 x i8> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v64i8'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <64 x i8> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512F-LABEL: 'splatvar_shift_v64i8'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = ashr <64 x i8> %a, %splat
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BW-LABEL: 'splatvar_shift_v64i8'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <64 x i8> %a, %splat
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512VL-LABEL: 'splatvar_shift_v64i8'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = ashr <64 x i8> %a, %splat
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BWVL-LABEL: 'splatvar_shift_v64i8'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <64 x i8> %a, %splat
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; XOPAVX-LABEL: 'splatvar_shift_v64i8'
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <64 x i8> %a, %splat
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
   %shift = ashr <64 x i8> %a, %splat
@@ -352,152 +986,398 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ;
 
 define <2 x i64> @constant_shift_v2i64(<2 x i64> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v2i64':
-; SSE2: Found an estimated cost of 12 for instruction:   %shift
-; SSE41: Found an estimated cost of 12 for instruction:   %shift
-; AVX: Found an estimated cost of 12 for instruction:   %shift
-; AVX2: Found an estimated cost of 4 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'constant_shift_v2i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <2 x i64> %a, <i64 1, i64 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; AVX1-LABEL: 'constant_shift_v2i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <2 x i64> %a, <i64 1, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; AVX2-LABEL: 'constant_shift_v2i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <2 x i64> %a, <i64 1, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; XOP-LABEL: 'constant_shift_v2i64'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <2 x i64> %a, <i64 1, i64 7>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; AVX512-LABEL: 'constant_shift_v2i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <2 x i64> %a, <i64 1, i64 7>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
   %shift = ashr <2 x i64> %a, <i64 1, i64 7>
   ret <2 x i64> %shift
 }
 
 define <4 x i64> @constant_shift_v4i64(<4 x i64> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v4i64':
-; SSE2: Found an estimated cost of 24 for instruction:   %shift
-; SSE41: Found an estimated cost of 24 for instruction:   %shift
-; AVX: Found an estimated cost of 26 for instruction:   %shift
-; AVX2: Found an estimated cost of 4 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 6 for instruction:   %shift
+; SSE-LABEL: 'constant_shift_v4i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <4 x i64> %a, <i64 1, i64 7, i64 15, i64 31>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX1-LABEL: 'constant_shift_v4i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %shift = ashr <4 x i64> %a, <i64 1, i64 7, i64 15, i64 31>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX2-LABEL: 'constant_shift_v4i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <4 x i64> %a, <i64 1, i64 7, i64 15, i64 31>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOP-LABEL: 'constant_shift_v4i64'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <4 x i64> %a, <i64 1, i64 7, i64 15, i64 31>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX512-LABEL: 'constant_shift_v4i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <4 x i64> %a, <i64 1, i64 7, i64 15, i64 31>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
   %shift = ashr <4 x i64> %a, <i64 1, i64 7, i64 15, i64 31>
   ret <4 x i64> %shift
 }
 
 define <8 x i64> @constant_shift_v8i64(<8 x i64> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v8i64':
-; SSE2: Found an estimated cost of 48 for instruction:   %shift
-; SSE41: Found an estimated cost of 48 for instruction:   %shift
-; AVX: Found an estimated cost of 52 for instruction:   %shift
-; AVX2: Found an estimated cost of 8 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 12 for instruction:   %shift
+; SSE-LABEL: 'constant_shift_v8i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = ashr <8 x i64> %a, <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX1-LABEL: 'constant_shift_v8i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %shift = ashr <8 x i64> %a, <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX2-LABEL: 'constant_shift_v8i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <8 x i64> %a, <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOP-LABEL: 'constant_shift_v8i64'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <8 x i64> %a, <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX512-LABEL: 'constant_shift_v8i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i64> %a, <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
   %shift = ashr <8 x i64> %a, <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>
   ret <8 x i64> %shift
 }
 
 define <4 x i32> @constant_shift_v4i32(<4 x i32> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v4i32':
-; SSE2: Found an estimated cost of 16 for instruction:   %shift
-; SSE41: Found an estimated cost of 12 for instruction:   %shift
-; AVX: Found an estimated cost of 12 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 2 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v4i32'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %shift = ashr <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; SSE41-LABEL: 'constant_shift_v4i32'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; AVX1-LABEL: 'constant_shift_v4i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; AVX2-LABEL: 'constant_shift_v4i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; XOPAVX1-LABEL: 'constant_shift_v4i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; XOPAVX2-LABEL: 'constant_shift_v4i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; AVX512-LABEL: 'constant_shift_v4i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; XOPAVX-LABEL: 'constant_shift_v4i32'
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
   %shift = ashr <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
   ret <4 x i32> %shift
 }
 
 define <8 x i32> @constant_shift_v8i32(<8 x i32> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v8i32':
-; SSE2: Found an estimated cost of 32 for instruction:   %shift
-; SSE41: Found an estimated cost of 24 for instruction:   %shift
-; AVX: Found an estimated cost of 26 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 6 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v8i32'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %shift = ashr <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; SSE41-LABEL: 'constant_shift_v8i32'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX1-LABEL: 'constant_shift_v8i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %shift = ashr <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX2-LABEL: 'constant_shift_v8i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX1-LABEL: 'constant_shift_v8i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX2-LABEL: 'constant_shift_v8i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX512-LABEL: 'constant_shift_v8i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX-LABEL: 'constant_shift_v8i32'
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
   %shift = ashr <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
   ret <8 x i32> %shift
 }
 
 define <16 x i32> @constant_shift_v16i32(<16 x i32> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v16i32':
-; SSE2: Found an estimated cost of 64 for instruction:   %shift
-; SSE41: Found an estimated cost of 48 for instruction:   %shift
-; AVX: Found an estimated cost of 52 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 12 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v16i32'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %shift = ashr <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; SSE41-LABEL: 'constant_shift_v16i32'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = ashr <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX1-LABEL: 'constant_shift_v16i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %shift = ashr <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX2-LABEL: 'constant_shift_v16i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX1-LABEL: 'constant_shift_v16i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX2-LABEL: 'constant_shift_v16i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX512-LABEL: 'constant_shift_v16i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX-LABEL: 'constant_shift_v16i32'
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
   %shift = ashr <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
   ret <16 x i32> %shift
 }
 
 define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v8i16':
-; SSE2: Found an estimated cost of 32 for instruction:   %shift
-; SSE41: Found an estimated cost of 14 for instruction:   %shift
-; AVX: Found an estimated cost of 14 for instruction:   %shift
-; AVX2: Found an estimated cost of 14 for instruction:   %shift
-; AVX512F: Found an estimated cost of 14 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 2 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v8i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; SSE41-LABEL: 'constant_shift_v8i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX-LABEL: 'constant_shift_v8i16'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; XOP-LABEL: 'constant_shift_v8i16'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512F-LABEL: 'constant_shift_v8i16'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512BW-LABEL: 'constant_shift_v8i16'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512VL-LABEL: 'constant_shift_v8i16'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512BWVL-LABEL: 'constant_shift_v8i16'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
   %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
 }
 
 define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v16i16':
-; SSE2: Found an estimated cost of 64 for instruction:   %shift
-; SSE41: Found an estimated cost of 28 for instruction:   %shift
-; AVX: Found an estimated cost of 30 for instruction:   %shift
-; AVX2: Found an estimated cost of 10 for instruction:   %shift
-; AVX512F: Found an estimated cost of 10 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 6 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v16i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; SSE41-LABEL: 'constant_shift_v16i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX1-LABEL: 'constant_shift_v16i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX2-LABEL: 'constant_shift_v16i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; XOP-LABEL: 'constant_shift_v16i16'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512F-LABEL: 'constant_shift_v16i16'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512BW-LABEL: 'constant_shift_v16i16'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512VL-LABEL: 'constant_shift_v16i16'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512BWVL-LABEL: 'constant_shift_v16i16'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
   %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
 }
 
 define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v32i16':
-; SSE2: Found an estimated cost of 128 for instruction:   %shift
-; SSE41: Found an estimated cost of 56 for instruction:   %shift
-; AVX: Found an estimated cost of 60 for instruction:   %shift
-; AVX2: Found an estimated cost of 20 for instruction:   %shift
-; AVX512F: Found an estimated cost of 20 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 12 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v32i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 128 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; SSE41-LABEL: 'constant_shift_v32i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX1-LABEL: 'constant_shift_v32i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX2-LABEL: 'constant_shift_v32i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; XOP-LABEL: 'constant_shift_v32i16'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512F-LABEL: 'constant_shift_v32i16'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BW-LABEL: 'constant_shift_v32i16'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512VL-LABEL: 'constant_shift_v32i16'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BWVL-LABEL: 'constant_shift_v32i16'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
   %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
 }
 
 define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v16i8':
-; SSE2: Found an estimated cost of 54 for instruction:   %shift
-; SSE41: Found an estimated cost of 24 for instruction:   %shift
-; AVX: Found an estimated cost of 24 for instruction:   %shift
-; AVX2: Found an estimated cost of 24 for instruction:   %shift
-; AVX512: Found an estimated cost of 24 for instruction:   %shift
-; XOP: Found an estimated cost of 2 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v16i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 54 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; SSE41-LABEL: 'constant_shift_v16i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX-LABEL: 'constant_shift_v16i8'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; XOP-LABEL: 'constant_shift_v16i8'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512-LABEL: 'constant_shift_v16i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
   %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
 }
 
 define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v32i8':
-; SSE2: Found an estimated cost of 108 for instruction:   %shift
-; SSE41: Found an estimated cost of 48 for instruction:   %shift
-; AVX: Found an estimated cost of 50 for instruction:   %shift
-; AVX2: Found an estimated cost of 24 for instruction:   %shift
-; AVX512: Found an estimated cost of 24 for instruction:   %shift
-; XOP: Found an estimated cost of 6 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v32i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 108 for instruction: %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; SSE41-LABEL: 'constant_shift_v32i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX1-LABEL: 'constant_shift_v32i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX2-LABEL: 'constant_shift_v32i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; XOP-LABEL: 'constant_shift_v32i8'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512-LABEL: 'constant_shift_v32i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
   %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <32 x i8> %shift
 }
 
 define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v64i8':
-; SSE2: Found an estimated cost of 216 for instruction:   %shift
-; SSE41: Found an estimated cost of 96 for instruction:   %shift
-; AVX: Found an estimated cost of 100 for instruction:   %shift
-; AVX2: Found an estimated cost of 48 for instruction:   %shift
-; AVX512F: Found an estimated cost of 48 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 24 for instruction:   %shift
-; XOP: Found an estimated cost of 12 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v64i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 216 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; SSE41-LABEL: 'constant_shift_v64i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 96 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX1-LABEL: 'constant_shift_v64i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 100 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX2-LABEL: 'constant_shift_v64i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; XOP-LABEL: 'constant_shift_v64i8'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512F-LABEL: 'constant_shift_v64i8'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BW-LABEL: 'constant_shift_v64i8'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512VL-LABEL: 'constant_shift_v64i8'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BWVL-LABEL: 'constant_shift_v64i8'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
   %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
 }
@@ -507,155 +1387,341 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ;
 
 define <2 x i64> @splatconstant_shift_v2i64(<2 x i64> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v2i64':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 4 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v2i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <2 x i64> %a, <i64 7, i64 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; AVX-LABEL: 'splatconstant_shift_v2i64'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <2 x i64> %a, <i64 7, i64 7>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; XOP-LABEL: 'splatconstant_shift_v2i64'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <2 x i64> %a, <i64 7, i64 7>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; AVX512-LABEL: 'splatconstant_shift_v2i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <2 x i64> %a, <i64 7, i64 7>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
   %shift = ashr <2 x i64> %a, <i64 7, i64 7>
   ret <2 x i64> %shift
 }
 
 define <4 x i64> @splatconstant_shift_v4i64(<4 x i64> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v4i64':
-; SSE2: Found an estimated cost of 8 for instruction:   %shift
-; SSE41: Found an estimated cost of 8 for instruction:   %shift
-; AVX: Found an estimated cost of 10 for instruction:   %shift
-; AVX2: Found an estimated cost of 4 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 6 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 4 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v4i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <4 x i64> %a, <i64 7, i64 7, i64 7, i64 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v4i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <4 x i64> %a, <i64 7, i64 7, i64 7, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v4i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <4 x i64> %a, <i64 7, i64 7, i64 7, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v4i64'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <4 x i64> %a, <i64 7, i64 7, i64 7, i64 7>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v4i64'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <4 x i64> %a, <i64 7, i64 7, i64 7, i64 7>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX512-LABEL: 'splatconstant_shift_v4i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <4 x i64> %a, <i64 7, i64 7, i64 7, i64 7>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOPAVX-LABEL: 'splatconstant_shift_v4i64'
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <4 x i64> %a, <i64 7, i64 7, i64 7, i64 7>
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
   %shift = ashr <4 x i64> %a, <i64 7, i64 7, i64 7, i64 7>
   ret <4 x i64> %shift
 }
 
 define <8 x i64> @splatconstant_shift_v8i64(<8 x i64> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v8i64':
-; SSE2: Found an estimated cost of 16 for instruction:   %shift
-; SSE41: Found an estimated cost of 16 for instruction:   %shift
-; AVX: Found an estimated cost of 20 for instruction:   %shift
-; AVX2: Found an estimated cost of 8 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 12 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 8 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v8i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %shift = ashr <8 x i64> %a, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v8i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = ashr <8 x i64> %a, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v8i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <8 x i64> %a, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v8i64'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <8 x i64> %a, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v8i64'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <8 x i64> %a, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX512-LABEL: 'splatconstant_shift_v8i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i64> %a, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOPAVX-LABEL: 'splatconstant_shift_v8i64'
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <8 x i64> %a, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
   %shift = ashr <8 x i64> %a, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
   ret <8 x i64> %shift
 }
 
 define <4 x i32> @splatconstant_shift_v4i32(<4 x i32> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v4i32':
-; SSE2: Found an estimated cost of 1 for instruction:   %shift
-; SSE41: Found an estimated cost of 1 for instruction:   %shift
-; AVX: Found an estimated cost of 1 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 1 for instruction:   %shift
+; CHECK-LABEL: 'splatconstant_shift_v4i32'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <4 x i32> %a, <i32 5, i32 5, i32 5, i32 5>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
   %shift = ashr <4 x i32> %a, <i32 5, i32 5, i32 5, i32 5>
   ret <4 x i32> %shift
 }
 
 define <8 x i32> @splatconstant_shift_v8i32(<8 x i32> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v8i32':
-; SSE2: Found an estimated cost of 2 for instruction:   %shift
-; SSE41: Found an estimated cost of 2 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 6 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v8i32'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <8 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v8i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <8 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v8i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v8i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <8 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v8i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX512-LABEL: 'splatconstant_shift_v8i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX-LABEL: 'splatconstant_shift_v8i32'
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <8 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
   %shift = ashr <8 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
   ret <8 x i32> %shift
 }
 
 define <16 x i32> @splatconstant_shift_v16i32(<16 x i32> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v16i32':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 8 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 12 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v16i32'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v16i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <16 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v16i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <16 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v16i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <16 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v16i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <16 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX512-LABEL: 'splatconstant_shift_v16i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX-LABEL: 'splatconstant_shift_v16i32'
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <16 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
   %shift = ashr <16 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
   ret <16 x i32> %shift
 }
 
 define <8 x i16> @splatconstant_shift_v8i16(<8 x i16> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v8i16':
-; SSE2: Found an estimated cost of 1 for instruction:   %shift
-; SSE41: Found an estimated cost of 1 for instruction:   %shift
-; AVX: Found an estimated cost of 1 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 1 for instruction:   %shift
+; CHECK-LABEL: 'splatconstant_shift_v8i16'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
   %shift = ashr <8 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <8 x i16> %shift
 }
 
 define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v16i16':
-; SSE2: Found an estimated cost of 2 for instruction:   %shift
-; SSE41: Found an estimated cost of 2 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 6 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v16i16'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v16i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v16i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v16i16'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v16i16'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512-LABEL: 'splatconstant_shift_v16i16'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; XOPAVX-LABEL: 'splatconstant_shift_v16i16'
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
   %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
 }
 
 define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v32i16':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 8 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512F: Found an estimated cost of 2 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 12 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v32i16'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v32i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v32i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v32i16'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v32i16'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512F-LABEL: 'splatconstant_shift_v32i16'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BW-LABEL: 'splatconstant_shift_v32i16'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512VL-LABEL: 'splatconstant_shift_v32i16'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BWVL-LABEL: 'splatconstant_shift_v32i16'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; XOPAVX-LABEL: 'splatconstant_shift_v32i16'
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
   %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
 }
 
 define <16 x i8> @splatconstant_shift_v16i8(<16 x i8> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v16i8':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 4 for instruction:   %shift
-; AVX512: Found an estimated cost of 4 for instruction:   %shift
-; XOP: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v16i8'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX-LABEL: 'splatconstant_shift_v16i8'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; XOP-LABEL: 'splatconstant_shift_v16i8'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512-LABEL: 'splatconstant_shift_v16i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
   %shift = ashr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <16 x i8> %shift
 }
 
 define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v32i8':
-; SSE2: Found an estimated cost of 8 for instruction:   %shift
-; SSE41: Found an estimated cost of 8 for instruction:   %shift
-; AVX: Found an estimated cost of 10 for instruction:   %shift
-; AVX2: Found an estimated cost of 4 for instruction:   %shift
-; AVX512: Found an estimated cost of 4 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 6 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 4 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v32i8'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v32i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v32i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v32i8'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v32i8'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512-LABEL: 'splatconstant_shift_v32i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; XOPAVX-LABEL: 'splatconstant_shift_v32i8'
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
   %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <32 x i8> %shift
 }
 
 define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v64i8':
-; SSE2: Found an estimated cost of 16 for instruction:   %shift
-; SSE41: Found an estimated cost of 16 for instruction:   %shift
-; AVX: Found an estimated cost of 20 for instruction:   %shift
-; AVX2: Found an estimated cost of 8 for instruction:   %shift
-; AVX512F: Found an estimated cost of 8 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 4 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 12 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 8 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v64i8'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v64i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v64i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v64i8'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v64i8'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512F-LABEL: 'splatconstant_shift_v64i8'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BW-LABEL: 'splatconstant_shift_v64i8'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512VL-LABEL: 'splatconstant_shift_v64i8'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BWVL-LABEL: 'splatconstant_shift_v64i8'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; XOPAVX-LABEL: 'splatconstant_shift_v64i8'
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; XOPAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
   %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <64 x i8> %shift
 }
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost.ll b/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost.ll
index 6e890369d677..c60a8dcf741c 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost.ll
@@ -1,15 +1,16 @@
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+sse2 -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=SSE2
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+sse4.1 -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=SSE41
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx2 -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX2
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+xop,+avx -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=XOP --check-prefix=XOPAVX
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+xop,+avx2 -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=XOP --check-prefix=XOPAVX2
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512f -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512 --check-prefix=AVX512F
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512f,+avx512dq -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512 --check-prefix=AVX512F
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512f,+avx512bw -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512 --check-prefix=AVX512BW
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512f,+avx512vl -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512 --check-prefix=AVX512F
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512f,+avx512dq,+avx512vl -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512 --check-prefix=AVX512F
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512f,+avx512bw,+avx512vl -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512 --check-prefix=AVX512BW
+; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+sse2 | FileCheck %s --check-prefixes=CHECK,SSE,SSE2
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+sse4.1 | FileCheck %s --check-prefixes=CHECK,SSE,SSE41
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX2
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+xop,+avx | FileCheck %s --check-prefixes=CHECK,XOP,XOPAVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+xop,+avx2 | FileCheck %s --check-prefixes=CHECK,XOP,XOPAVX2
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx512f | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512F
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx512f,+avx512dq | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512F
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512BW
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx512f,+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512F
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx512f,+avx512dq,+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512VL
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx512f,+avx512bw,+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512BWVL
 
 ; Verify the cost of vector logical shift right instructions.
 
@@ -18,155 +19,401 @@
 ;
 
 define <2 x i64> @var_shift_v2i64(<2 x i64> %a, <2 x i64> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v2i64':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 2 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE-LABEL: 'var_shift_v2i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <2 x i64> %a, %b
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; AVX1-LABEL: 'var_shift_v2i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <2 x i64> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; AVX2-LABEL: 'var_shift_v2i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <2 x i64> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; XOPAVX1-LABEL: 'var_shift_v2i64'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <2 x i64> %a, %b
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; XOPAVX2-LABEL: 'var_shift_v2i64'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <2 x i64> %a, %b
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; AVX512-LABEL: 'var_shift_v2i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <2 x i64> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
   %shift = lshr <2 x i64> %a, %b
   ret <2 x i64> %shift
 }
 
 define <4 x i64> @var_shift_v4i64(<4 x i64> %a, <4 x i64> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v4i64':
-; SSE2: Found an estimated cost of 8 for instruction:   %shift
-; SSE41: Found an estimated cost of 8 for instruction:   %shift
-; AVX: Found an estimated cost of 10 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 6 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE-LABEL: 'var_shift_v4i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <4 x i64> %a, %b
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX1-LABEL: 'var_shift_v4i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = lshr <4 x i64> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX2-LABEL: 'var_shift_v4i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <4 x i64> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOPAVX1-LABEL: 'var_shift_v4i64'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = lshr <4 x i64> %a, %b
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOPAVX2-LABEL: 'var_shift_v4i64'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <4 x i64> %a, %b
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX512-LABEL: 'var_shift_v4i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <4 x i64> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
   %shift = lshr <4 x i64> %a, %b
   ret <4 x i64> %shift
 }
 
 define <8 x i64> @var_shift_v8i64(<8 x i64> %a, <8 x i64> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v8i64':
-; SSE2: Found an estimated cost of 16 for instruction:   %shift
-; SSE41: Found an estimated cost of 16 for instruction:   %shift
-; AVX: Found an estimated cost of 20 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 12 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'var_shift_v8i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %shift = lshr <8 x i64> %a, %b
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX1-LABEL: 'var_shift_v8i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = lshr <8 x i64> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX2-LABEL: 'var_shift_v8i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <8 x i64> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOPAVX1-LABEL: 'var_shift_v8i64'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <8 x i64> %a, %b
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOPAVX2-LABEL: 'var_shift_v8i64'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <8 x i64> %a, %b
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX512-LABEL: 'var_shift_v8i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i64> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
   %shift = lshr <8 x i64> %a, %b
   ret <8 x i64> %shift
 }
 
 define <4 x i32> @var_shift_v4i32(<4 x i32> %a, <4 x i32> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v4i32':
-; SSE2: Found an estimated cost of 16 for instruction:   %shift
-; SSE41: Found an estimated cost of 11 for instruction:   %shift
-; AVX: Found an estimated cost of 11 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 2 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v4i32'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %shift = lshr <4 x i32> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; SSE41-LABEL: 'var_shift_v4i32'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = lshr <4 x i32> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; AVX1-LABEL: 'var_shift_v4i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = lshr <4 x i32> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; AVX2-LABEL: 'var_shift_v4i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <4 x i32> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; XOPAVX1-LABEL: 'var_shift_v4i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <4 x i32> %a, %b
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; XOPAVX2-LABEL: 'var_shift_v4i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <4 x i32> %a, %b
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; AVX512-LABEL: 'var_shift_v4i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <4 x i32> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
   %shift = lshr <4 x i32> %a, %b
   ret <4 x i32> %shift
 }
 
 define <8 x i32> @var_shift_v8i32(<8 x i32> %a, <8 x i32> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v8i32':
-; SSE2: Found an estimated cost of 32 for instruction:   %shift
-; SSE41: Found an estimated cost of 22 for instruction:   %shift
-; AVX: Found an estimated cost of 24 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 6 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v8i32'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %shift = lshr <8 x i32> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; SSE41-LABEL: 'var_shift_v8i32'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = lshr <8 x i32> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX1-LABEL: 'var_shift_v8i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = lshr <8 x i32> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX2-LABEL: 'var_shift_v8i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i32> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX1-LABEL: 'var_shift_v8i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = lshr <8 x i32> %a, %b
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX2-LABEL: 'var_shift_v8i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i32> %a, %b
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX512-LABEL: 'var_shift_v8i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i32> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
   %shift = lshr <8 x i32> %a, %b
   ret <8 x i32> %shift
 }
 
 define <16 x i32> @var_shift_v16i32(<16 x i32> %a, <16 x i32> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v16i32':
-; SSE2: Found an estimated cost of 64 for instruction:   %shift
-; SSE41: Found an estimated cost of 44 for instruction:   %shift
-; AVX: Found an estimated cost of 48 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 12 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v16i32'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %shift = lshr <16 x i32> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; SSE41-LABEL: 'var_shift_v16i32'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %shift = lshr <16 x i32> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX1-LABEL: 'var_shift_v16i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = lshr <16 x i32> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX2-LABEL: 'var_shift_v16i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <16 x i32> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX1-LABEL: 'var_shift_v16i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <16 x i32> %a, %b
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX2-LABEL: 'var_shift_v16i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <16 x i32> %a, %b
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX512-LABEL: 'var_shift_v16i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i32> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
   %shift = lshr <16 x i32> %a, %b
   ret <16 x i32> %shift
 }
 
 define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v8i16':
-; SSE2: Found an estimated cost of 32 for instruction:   %shift
-; SSE41: Found an estimated cost of 14 for instruction:   %shift
-; AVX: Found an estimated cost of 14 for instruction:   %shift
-; AVX2: Found an estimated cost of 14 for instruction:   %shift
-; AVX512F: Found an estimated cost of 14 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 2 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v8i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %shift = lshr <8 x i16> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; SSE41-LABEL: 'var_shift_v8i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %shift = lshr <8 x i16> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX-LABEL: 'var_shift_v8i16'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %shift = lshr <8 x i16> %a, %b
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; XOP-LABEL: 'var_shift_v8i16'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <8 x i16> %a, %b
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512F-LABEL: 'var_shift_v8i16'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %shift = lshr <8 x i16> %a, %b
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512BW-LABEL: 'var_shift_v8i16'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %b
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512VL-LABEL: 'var_shift_v8i16'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %shift = lshr <8 x i16> %a, %b
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512BWVL-LABEL: 'var_shift_v8i16'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %b
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
   %shift = lshr <8 x i16> %a, %b
   ret <8 x i16> %shift
 }
 
 define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v16i16':
-; SSE2: Found an estimated cost of 64 for instruction:   %shift
-; SSE41: Found an estimated cost of 28 for instruction:   %shift
-; AVX: Found an estimated cost of 30 for instruction:   %shift
-; AVX2: Found an estimated cost of 10 for instruction:   %shift
-; AVX512F: Found an estimated cost of 10 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 6 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v16i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %shift = lshr <16 x i16> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; SSE41-LABEL: 'var_shift_v16i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %shift = lshr <16 x i16> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX1-LABEL: 'var_shift_v16i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = lshr <16 x i16> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX2-LABEL: 'var_shift_v16i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = lshr <16 x i16> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; XOP-LABEL: 'var_shift_v16i16'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = lshr <16 x i16> %a, %b
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512F-LABEL: 'var_shift_v16i16'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = lshr <16 x i16> %a, %b
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512BW-LABEL: 'var_shift_v16i16'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %b
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512VL-LABEL: 'var_shift_v16i16'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = lshr <16 x i16> %a, %b
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512BWVL-LABEL: 'var_shift_v16i16'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %b
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
   %shift = lshr <16 x i16> %a, %b
   ret <16 x i16> %shift
 }
 
 define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v32i16':
-; SSE2: Found an estimated cost of 128 for instruction:   %shift
-; SSE41: Found an estimated cost of 56 for instruction:   %shift
-; AVX: Found an estimated cost of 60 for instruction:   %shift
-; AVX2: Found an estimated cost of 20 for instruction:   %shift
-; AVX512F: Found an estimated cost of 20 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 12 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v32i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 128 for instruction: %shift = lshr <32 x i16> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; SSE41-LABEL: 'var_shift_v32i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %shift = lshr <32 x i16> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX1-LABEL: 'var_shift_v32i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %shift = lshr <32 x i16> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX2-LABEL: 'var_shift_v32i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = lshr <32 x i16> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; XOP-LABEL: 'var_shift_v32i16'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <32 x i16> %a, %b
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512F-LABEL: 'var_shift_v32i16'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = lshr <32 x i16> %a, %b
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BW-LABEL: 'var_shift_v32i16'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %b
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512VL-LABEL: 'var_shift_v32i16'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = lshr <32 x i16> %a, %b
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BWVL-LABEL: 'var_shift_v32i16'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %b
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
   %shift = lshr <32 x i16> %a, %b
   ret <32 x i16> %shift
 }
 
 define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v16i8':
-; SSE2: Found an estimated cost of 26 for instruction:   %shift
-; SSE41: Found an estimated cost of 12 for instruction:   %shift
-; AVX: Found an estimated cost of 12 for instruction:   %shift
-; AVX2: Found an estimated cost of 12 for instruction:   %shift
-; AVX512: Found an estimated cost of 12 for instruction:   %shift
-; XOP: Found an estimated cost of 2 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v16i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %shift = lshr <16 x i8> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; SSE41-LABEL: 'var_shift_v16i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <16 x i8> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX-LABEL: 'var_shift_v16i8'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <16 x i8> %a, %b
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; XOP-LABEL: 'var_shift_v16i8'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <16 x i8> %a, %b
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512-LABEL: 'var_shift_v16i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <16 x i8> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
   %shift = lshr <16 x i8> %a, %b
   ret <16 x i8> %shift
 }
 
 define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v32i8':
-; SSE2: Found an estimated cost of 52 for instruction:   %shift
-; SSE41: Found an estimated cost of 24 for instruction:   %shift
-; AVX: Found an estimated cost of 26 for instruction:   %shift
-; AVX2: Found an estimated cost of 11 for instruction:   %shift
-; AVX512: Found an estimated cost of 11 for instruction:   %shift
-; XOP: Found an estimated cost of 6 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v32i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %shift = lshr <32 x i8> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; SSE41-LABEL: 'var_shift_v32i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = lshr <32 x i8> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX1-LABEL: 'var_shift_v32i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %shift = lshr <32 x i8> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX2-LABEL: 'var_shift_v32i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = lshr <32 x i8> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; XOP-LABEL: 'var_shift_v32i8'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = lshr <32 x i8> %a, %b
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512-LABEL: 'var_shift_v32i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = lshr <32 x i8> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
   %shift = lshr <32 x i8> %a, %b
   ret <32 x i8> %shift
 }
 
 define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v64i8':
-; SSE2: Found an estimated cost of 104 for instruction:   %shift
-; SSE41: Found an estimated cost of 48 for instruction:   %shift
-; AVX: Found an estimated cost of 52 for instruction:   %shift
-; AVX2: Found an estimated cost of 22 for instruction:   %shift
-; AVX512F: Found an estimated cost of 22 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 11 for instruction:   %shift
-; XOP: Found an estimated cost of 12 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v64i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 104 for instruction: %shift = lshr <64 x i8> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; SSE41-LABEL: 'var_shift_v64i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = lshr <64 x i8> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX1-LABEL: 'var_shift_v64i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %shift = lshr <64 x i8> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX2-LABEL: 'var_shift_v64i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = lshr <64 x i8> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; XOP-LABEL: 'var_shift_v64i8'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <64 x i8> %a, %b
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512F-LABEL: 'var_shift_v64i8'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = lshr <64 x i8> %a, %b
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BW-LABEL: 'var_shift_v64i8'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = lshr <64 x i8> %a, %b
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512VL-LABEL: 'var_shift_v64i8'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = lshr <64 x i8> %a, %b
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BWVL-LABEL: 'var_shift_v64i8'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = lshr <64 x i8> %a, %b
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
   %shift = lshr <64 x i8> %a, %b
   ret <64 x i8> %shift
 }
@@ -176,13 +423,12 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ;
 
 define <2 x i64> @splatvar_shift_v2i64(<2 x i64> %a, i64 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v2i64':
-; SSE2: Found an estimated cost of 1 for instruction:   %shift
-; SSE41: Found an estimated cost of 1 for instruction:   %shift
-; AVX: Found an estimated cost of 1 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 1 for instruction:   %shift
+; CHECK-LABEL: 'splatvar_shift_v2i64'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <2 x i64> undef, i64 %b, i32 0
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <2 x i64> %insert, <2 x i64> undef, <2 x i32> zeroinitializer
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <2 x i64> %a, %splat
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
   %insert = insertelement <2 x i64> undef, i64 %b, i32 0
   %splat = shufflevector <2 x i64> %insert, <2 x i64> undef, <2 x i32> zeroinitializer
   %shift = lshr <2 x i64> %a, %splat
@@ -190,14 +436,42 @@ define <2 x i64> @splatvar_shift_v2i64(<2 x i64> %a, i64 %b) {
 }
 
 define <4 x i64> @splatvar_shift_v4i64(<4 x i64> %a, i64 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v4i64':
-; SSE2: Found an estimated cost of 2 for instruction:   %shift
-; SSE41: Found an estimated cost of 2 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 6 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE-LABEL: 'splatvar_shift_v4i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <4 x i64> %a, %splat
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v4i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <4 x i64> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v4i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <4 x i64> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v4i64'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = lshr <4 x i64> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v4i64'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <4 x i64> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v4i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <4 x i64> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
   %insert = insertelement <4 x i64> undef, i64 %b, i32 0
   %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
   %shift = lshr <4 x i64> %a, %splat
@@ -205,14 +479,42 @@ define <4 x i64> @splatvar_shift_v4i64(<4 x i64> %a, i64 %b) {
 }
 
 define <8 x i64> @splatvar_shift_v8i64(<8 x i64> %a, i64 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v8i64':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 8 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 12 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'splatvar_shift_v8i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <8 x i64> %a, %splat
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v8i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <8 x i64> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v8i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <8 x i64> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v8i64'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <8 x i64> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v8i64'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <8 x i64> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v8i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i64> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
   %insert = insertelement <8 x i64> undef, i64 %b, i32 0
   %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
   %shift = lshr <8 x i64> %a, %splat
@@ -220,13 +522,12 @@ define <8 x i64> @splatvar_shift_v8i64(<8 x i64> %a, i64 %b) {
 }
 
 define <4 x i32> @splatvar_shift_v4i32(<4 x i32> %a, i32 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v4i32':
-; SSE2: Found an estimated cost of 1 for instruction:   %shift
-; SSE41: Found an estimated cost of 1 for instruction:   %shift
-; AVX: Found an estimated cost of 1 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 1 for instruction:   %shift
+; CHECK-LABEL: 'splatvar_shift_v4i32'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i32> undef, i32 %b, i32 0
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i32> %insert, <4 x i32> undef, <4 x i32> zeroinitializer
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <4 x i32> %a, %splat
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
   %insert = insertelement <4 x i32> undef, i32 %b, i32 0
   %splat = shufflevector <4 x i32> %insert, <4 x i32> undef, <4 x i32> zeroinitializer
   %shift = lshr <4 x i32> %a, %splat
@@ -234,14 +535,42 @@ define <4 x i32> @splatvar_shift_v4i32(<4 x i32> %a, i32 %b) {
 }
 
 define <8 x i32> @splatvar_shift_v8i32(<8 x i32> %a, i32 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v8i32':
-; SSE2: Found an estimated cost of 2 for instruction:   %shift
-; SSE41: Found an estimated cost of 2 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 6 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE-LABEL: 'splatvar_shift_v8i32'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <8 x i32> %a, %splat
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v8i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <8 x i32> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v8i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i32> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v8i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = lshr <8 x i32> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v8i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i32> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v8i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i32> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
   %insert = insertelement <8 x i32> undef, i32 %b, i32 0
   %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
   %shift = lshr <8 x i32> %a, %splat
@@ -249,14 +578,42 @@ define <8 x i32> @splatvar_shift_v8i32(<8 x i32> %a, i32 %b) {
 }
 
 define <16 x i32> @splatvar_shift_v16i32(<16 x i32> %a, i32 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v16i32':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 8 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 12 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'splatvar_shift_v16i32'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i32> %a, %splat
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v16i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <16 x i32> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v16i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <16 x i32> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v16i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <16 x i32> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v16i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <16 x i32> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v16i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i32> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
   %insert = insertelement <16 x i32> undef, i32 %b, i32 0
   %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
   %shift = lshr <16 x i32> %a, %splat
@@ -264,13 +621,36 @@ define <16 x i32> @splatvar_shift_v16i32(<16 x i32> %a, i32 %b) {
 }
 
 define <8 x i16> @splatvar_shift_v8i16(<8 x i16> %a, i16 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v8i16':
-; SSE2: Found an estimated cost of 1 for instruction:   %shift
-; SSE41: Found an estimated cost of 1 for instruction:   %shift
-; AVX: Found an estimated cost of 1 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 1 for instruction:   %shift
+; SSE2-LABEL: 'splatvar_shift_v8i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %splat
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; SSE41-LABEL: 'splatvar_shift_v8i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %splat
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX-LABEL: 'splatvar_shift_v8i16'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %splat
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; XOP-LABEL: 'splatvar_shift_v8i16'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %splat
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v8i16'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
   %insert = insertelement <8 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
   %shift = lshr <8 x i16> %a, %splat
@@ -278,14 +658,48 @@ define <8 x i16> @splatvar_shift_v8i16(<8 x i16> %a, i16 %b) {
 }
 
 define <16 x i16> @splatvar_shift_v16i16(<16 x i16> %a, i16 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v16i16':
-; SSE2: Found an estimated cost of 2 for instruction:   %shift
-; SSE41: Found an estimated cost of 2 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 6 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE2-LABEL: 'splatvar_shift_v16i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <16 x i16> %a, %splat
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; SSE41-LABEL: 'splatvar_shift_v16i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <16 x i16> %a, %splat
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v16i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i16> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v16i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v16i16'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = lshr <16 x i16> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v16i16'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v16i16'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
   %insert = insertelement <16 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
   %shift = lshr <16 x i16> %a, %splat
@@ -293,15 +707,66 @@ define <16 x i16> @splatvar_shift_v16i16(<16 x i16> %a, i16 %b) {
 }
 
 define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v32i16':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 8 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512F: Found an estimated cost of 2 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 12 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE2-LABEL: 'splatvar_shift_v32i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i16> %a, %splat
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; SSE41-LABEL: 'splatvar_shift_v32i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i16> %a, %splat
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v32i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <32 x i16> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v32i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <32 x i16> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v32i16'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <32 x i16> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v32i16'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <32 x i16> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512F-LABEL: 'splatvar_shift_v32i16'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <32 x i16> %a, %splat
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BW-LABEL: 'splatvar_shift_v32i16'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %splat
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512VL-LABEL: 'splatvar_shift_v32i16'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <32 x i16> %a, %splat
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BWVL-LABEL: 'splatvar_shift_v32i16'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %splat
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
   %shift = lshr <32 x i16> %a, %splat
@@ -309,13 +774,36 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 }
 
 define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v16i8':
-; SSE2: Found an estimated cost of 26 for instruction:   %shift
-; SSE41: Found an estimated cost of 12 for instruction:   %shift
-; AVX: Found an estimated cost of 12 for instruction:   %shift
-; AVX2: Found an estimated cost of 12 for instruction:   %shift
-; AVX512: Found an estimated cost of 12 for instruction:   %shift
-; XOP: Found an estimated cost of 2 for instruction:   %shift
+; SSE2-LABEL: 'splatvar_shift_v16i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %shift = lshr <16 x i8> %a, %splat
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; SSE41-LABEL: 'splatvar_shift_v16i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <16 x i8> %a, %splat
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX-LABEL: 'splatvar_shift_v16i8'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <16 x i8> %a, %splat
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; XOP-LABEL: 'splatvar_shift_v16i8'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <16 x i8> %a, %splat
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v16i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <16 x i8> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
   %insert = insertelement <16 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
   %shift = lshr <16 x i8> %a, %splat
@@ -323,13 +811,48 @@ define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
 }
 
 define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v32i8':
-; SSE2: Found an estimated cost of 52 for instruction:   %shift
-; SSE41: Found an estimated cost of 24 for instruction:   %shift
-; AVX: Found an estimated cost of 26 for instruction:   %shift
-; AVX2: Found an estimated cost of 11 for instruction:   %shift
-; AVX512: Found an estimated cost of 11 for instruction:   %shift
-; XOP: Found an estimated cost of 6 for instruction:   %shift
+; SSE2-LABEL: 'splatvar_shift_v32i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %shift = lshr <32 x i8> %a, %splat
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; SSE41-LABEL: 'splatvar_shift_v32i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = lshr <32 x i8> %a, %splat
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v32i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %shift = lshr <32 x i8> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v32i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = lshr <32 x i8> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v32i8'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = lshr <32 x i8> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v32i8'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = lshr <32 x i8> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v32i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = lshr <32 x i8> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
   %insert = insertelement <32 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
   %shift = lshr <32 x i8> %a, %splat
@@ -337,14 +860,66 @@ define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 }
 
 define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v64i8':
-; SSE2: Found an estimated cost of 104 for instruction:   %shift
-; SSE41: Found an estimated cost of 48 for instruction:   %shift
-; AVX: Found an estimated cost of 52 for instruction:   %shift
-; AVX2: Found an estimated cost of 22 for instruction:   %shift
-; AVX512F: Found an estimated cost of 22 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 11 for instruction:   %shift
-; XOP: Found an estimated cost of 12 for instruction:   %shift
+; SSE2-LABEL: 'splatvar_shift_v64i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 104 for instruction: %shift = lshr <64 x i8> %a, %splat
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; SSE41-LABEL: 'splatvar_shift_v64i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = lshr <64 x i8> %a, %splat
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v64i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %shift = lshr <64 x i8> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v64i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = lshr <64 x i8> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v64i8'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <64 x i8> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v64i8'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <64 x i8> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512F-LABEL: 'splatvar_shift_v64i8'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = lshr <64 x i8> %a, %splat
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BW-LABEL: 'splatvar_shift_v64i8'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = lshr <64 x i8> %a, %splat
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512VL-LABEL: 'splatvar_shift_v64i8'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = lshr <64 x i8> %a, %splat
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BWVL-LABEL: 'splatvar_shift_v64i8'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = lshr <64 x i8> %a, %splat
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
   %shift = lshr <64 x i8> %a, %splat
@@ -356,155 +931,401 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ;
 
 define <2 x i64> @constant_shift_v2i64(<2 x i64> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v2i64':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 2 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE-LABEL: 'constant_shift_v2i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <2 x i64> %a, <i64 1, i64 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; AVX1-LABEL: 'constant_shift_v2i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <2 x i64> %a, <i64 1, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; AVX2-LABEL: 'constant_shift_v2i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <2 x i64> %a, <i64 1, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; XOPAVX1-LABEL: 'constant_shift_v2i64'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <2 x i64> %a, <i64 1, i64 7>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; XOPAVX2-LABEL: 'constant_shift_v2i64'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <2 x i64> %a, <i64 1, i64 7>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; AVX512-LABEL: 'constant_shift_v2i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <2 x i64> %a, <i64 1, i64 7>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
   %shift = lshr <2 x i64> %a, <i64 1, i64 7>
   ret <2 x i64> %shift
 }
 
 define <4 x i64> @constant_shift_v4i64(<4 x i64> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v4i64':
-; SSE2: Found an estimated cost of 8 for instruction:   %shift
-; SSE41: Found an estimated cost of 8 for instruction:   %shift
-; AVX: Found an estimated cost of 10 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 6 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE-LABEL: 'constant_shift_v4i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <4 x i64> %a, <i64 1, i64 7, i64 15, i64 31>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX1-LABEL: 'constant_shift_v4i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = lshr <4 x i64> %a, <i64 1, i64 7, i64 15, i64 31>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX2-LABEL: 'constant_shift_v4i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <4 x i64> %a, <i64 1, i64 7, i64 15, i64 31>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOPAVX1-LABEL: 'constant_shift_v4i64'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = lshr <4 x i64> %a, <i64 1, i64 7, i64 15, i64 31>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOPAVX2-LABEL: 'constant_shift_v4i64'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <4 x i64> %a, <i64 1, i64 7, i64 15, i64 31>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX512-LABEL: 'constant_shift_v4i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <4 x i64> %a, <i64 1, i64 7, i64 15, i64 31>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
   %shift = lshr <4 x i64> %a, <i64 1, i64 7, i64 15, i64 31>
   ret <4 x i64> %shift
 }
 
 define <8 x i64> @constant_shift_v8i64(<8 x i64> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v8i64':
-; SSE2: Found an estimated cost of 16 for instruction:   %shift
-; SSE41: Found an estimated cost of 16 for instruction:   %shift
-; AVX: Found an estimated cost of 20 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 12 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'constant_shift_v8i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %shift = lshr <8 x i64> %a, <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX1-LABEL: 'constant_shift_v8i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = lshr <8 x i64> %a, <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX2-LABEL: 'constant_shift_v8i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <8 x i64> %a, <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOPAVX1-LABEL: 'constant_shift_v8i64'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <8 x i64> %a, <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOPAVX2-LABEL: 'constant_shift_v8i64'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <8 x i64> %a, <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX512-LABEL: 'constant_shift_v8i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i64> %a, <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
   %shift = lshr <8 x i64> %a, <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>
   ret <8 x i64> %shift
 }
 
 define <4 x i32> @constant_shift_v4i32(<4 x i32> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v4i32':
-; SSE2: Found an estimated cost of 16 for instruction:   %shift
-; SSE41: Found an estimated cost of 11 for instruction:   %shift
-; AVX: Found an estimated cost of 11 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 2 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v4i32'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %shift = lshr <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; SSE41-LABEL: 'constant_shift_v4i32'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = lshr <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; AVX1-LABEL: 'constant_shift_v4i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = lshr <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; AVX2-LABEL: 'constant_shift_v4i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; XOPAVX1-LABEL: 'constant_shift_v4i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; XOPAVX2-LABEL: 'constant_shift_v4i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; AVX512-LABEL: 'constant_shift_v4i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
   %shift = lshr <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
   ret <4 x i32> %shift
 }
 
 define <8 x i32> @constant_shift_v8i32(<8 x i32> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v8i32':
-; SSE2: Found an estimated cost of 32 for instruction:   %shift
-; SSE41: Found an estimated cost of 22 for instruction:   %shift
-; AVX: Found an estimated cost of 24 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 6 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v8i32'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %shift = lshr <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; SSE41-LABEL: 'constant_shift_v8i32'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = lshr <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX1-LABEL: 'constant_shift_v8i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = lshr <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX2-LABEL: 'constant_shift_v8i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX1-LABEL: 'constant_shift_v8i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = lshr <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX2-LABEL: 'constant_shift_v8i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX512-LABEL: 'constant_shift_v8i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
   %shift = lshr <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
   ret <8 x i32> %shift
 }
 
 define <16 x i32> @constant_shift_v16i32(<16 x i32> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v16i32':
-; SSE2: Found an estimated cost of 64 for instruction:   %shift
-; SSE41: Found an estimated cost of 44 for instruction:   %shift
-; AVX: Found an estimated cost of 48 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 12 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v16i32'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %shift = lshr <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; SSE41-LABEL: 'constant_shift_v16i32'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %shift = lshr <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX1-LABEL: 'constant_shift_v16i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = lshr <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX2-LABEL: 'constant_shift_v16i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX1-LABEL: 'constant_shift_v16i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX2-LABEL: 'constant_shift_v16i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX512-LABEL: 'constant_shift_v16i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
   %shift = lshr <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
   ret <16 x i32> %shift
 }
 
 define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v8i16':
-; SSE2: Found an estimated cost of 32 for instruction:   %shift
-; SSE41: Found an estimated cost of 14 for instruction:   %shift
-; AVX: Found an estimated cost of 14 for instruction:   %shift
-; AVX2: Found an estimated cost of 14 for instruction:   %shift
-; AVX512F: Found an estimated cost of 14 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 2 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v8i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; SSE41-LABEL: 'constant_shift_v8i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX-LABEL: 'constant_shift_v8i16'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; XOP-LABEL: 'constant_shift_v8i16'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512F-LABEL: 'constant_shift_v8i16'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512BW-LABEL: 'constant_shift_v8i16'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512VL-LABEL: 'constant_shift_v8i16'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512BWVL-LABEL: 'constant_shift_v8i16'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
   %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
 }
 
 define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v16i16':
-; SSE2: Found an estimated cost of 64 for instruction:   %shift
-; SSE41: Found an estimated cost of 28 for instruction:   %shift
-; AVX: Found an estimated cost of 30 for instruction:   %shift
-; AVX2: Found an estimated cost of 10 for instruction:   %shift
-; AVX512F: Found an estimated cost of 10 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 6 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v16i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; SSE41-LABEL: 'constant_shift_v16i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX1-LABEL: 'constant_shift_v16i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX2-LABEL: 'constant_shift_v16i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; XOP-LABEL: 'constant_shift_v16i16'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512F-LABEL: 'constant_shift_v16i16'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512BW-LABEL: 'constant_shift_v16i16'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512VL-LABEL: 'constant_shift_v16i16'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512BWVL-LABEL: 'constant_shift_v16i16'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
   %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
 }
 
 define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v32i16':
-; SSE2: Found an estimated cost of 128 for instruction:   %shift
-; SSE41: Found an estimated cost of 56 for instruction:   %shift
-; AVX: Found an estimated cost of 60 for instruction:   %shift
-; AVX2: Found an estimated cost of 20 for instruction:   %shift
-; AVX512F: Found an estimated cost of 20 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 12 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v32i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 128 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; SSE41-LABEL: 'constant_shift_v32i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX1-LABEL: 'constant_shift_v32i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX2-LABEL: 'constant_shift_v32i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; XOP-LABEL: 'constant_shift_v32i16'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512F-LABEL: 'constant_shift_v32i16'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BW-LABEL: 'constant_shift_v32i16'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512VL-LABEL: 'constant_shift_v32i16'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BWVL-LABEL: 'constant_shift_v32i16'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
   %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
 }
 
 define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v16i8':
-; SSE2: Found an estimated cost of 26 for instruction:   %shift
-; SSE41: Found an estimated cost of 12 for instruction:   %shift
-; AVX: Found an estimated cost of 12 for instruction:   %shift
-; AVX2: Found an estimated cost of 12 for instruction:   %shift
-; AVX512: Found an estimated cost of 12 for instruction:   %shift
-; XOP: Found an estimated cost of 2 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v16i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; SSE41-LABEL: 'constant_shift_v16i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX-LABEL: 'constant_shift_v16i8'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; XOP-LABEL: 'constant_shift_v16i8'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512-LABEL: 'constant_shift_v16i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
   %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
 }
 
 define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v32i8':
-; SSE2: Found an estimated cost of 52 for instruction:   %shift
-; SSE41: Found an estimated cost of 24 for instruction:   %shift
-; AVX: Found an estimated cost of 26 for instruction:   %shift
-; AVX2: Found an estimated cost of 11 for instruction:   %shift
-; AVX512: Found an estimated cost of 11 for instruction:   %shift
-; XOP: Found an estimated cost of 6 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v32i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; SSE41-LABEL: 'constant_shift_v32i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX1-LABEL: 'constant_shift_v32i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX2-LABEL: 'constant_shift_v32i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; XOP-LABEL: 'constant_shift_v32i8'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512-LABEL: 'constant_shift_v32i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
   %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <32 x i8> %shift
 }
 
 define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v64i8':
-; SSE2: Found an estimated cost of 104 for instruction:   %shift
-; SSE41: Found an estimated cost of 48 for instruction:   %shift
-; AVX: Found an estimated cost of 52 for instruction:   %shift
-; AVX2: Found an estimated cost of 22 for instruction:   %shift
-; AVX512F: Found an estimated cost of 22 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 11 for instruction:   %shift
-; XOP: Found an estimated cost of 12 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v64i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 104 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; SSE41-LABEL: 'constant_shift_v64i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX1-LABEL: 'constant_shift_v64i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX2-LABEL: 'constant_shift_v64i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; XOP-LABEL: 'constant_shift_v64i8'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512F-LABEL: 'constant_shift_v64i8'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BW-LABEL: 'constant_shift_v64i8'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512VL-LABEL: 'constant_shift_v64i8'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BWVL-LABEL: 'constant_shift_v64i8'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
   %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
 }
@@ -514,155 +1335,293 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ;
 
 define <2 x i64> @splatconstant_shift_v2i64(<2 x i64> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v2i64':
-; SSE2: Found an estimated cost of 1 for instruction:   %shift
-; SSE41: Found an estimated cost of 1 for instruction:   %shift
-; AVX: Found an estimated cost of 1 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 1 for instruction:   %shift
+; CHECK-LABEL: 'splatconstant_shift_v2i64'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <2 x i64> %a, <i64 7, i64 7>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
   %shift = lshr <2 x i64> %a, <i64 7, i64 7>
   ret <2 x i64> %shift
 }
 
 define <4 x i64> @splatconstant_shift_v4i64(<4 x i64> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v4i64':
-; SSE2: Found an estimated cost of 2 for instruction:   %shift
-; SSE41: Found an estimated cost of 2 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 6 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v4i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <4 x i64> %a, <i64 7, i64 7, i64 7, i64 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v4i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <4 x i64> %a, <i64 7, i64 7, i64 7, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v4i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <4 x i64> %a, <i64 7, i64 7, i64 7, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v4i64'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = lshr <4 x i64> %a, <i64 7, i64 7, i64 7, i64 7>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v4i64'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <4 x i64> %a, <i64 7, i64 7, i64 7, i64 7>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX512-LABEL: 'splatconstant_shift_v4i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <4 x i64> %a, <i64 7, i64 7, i64 7, i64 7>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
   %shift = lshr <4 x i64> %a, <i64 7, i64 7, i64 7, i64 7>
   ret <4 x i64> %shift
 }
 
 define <8 x i64> @splatconstant_shift_v8i64(<8 x i64> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v8i64':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 8 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 12 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v8i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <8 x i64> %a, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v8i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <8 x i64> %a, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v8i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <8 x i64> %a, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v8i64'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <8 x i64> %a, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v8i64'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <8 x i64> %a, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX512-LABEL: 'splatconstant_shift_v8i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i64> %a, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
   %shift = lshr <8 x i64> %a, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
   ret <8 x i64> %shift
 }
 
 define <4 x i32> @splatconstant_shift_v4i32(<4 x i32> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v4i32':
-; SSE2: Found an estimated cost of 1 for instruction:   %shift
-; SSE41: Found an estimated cost of 1 for instruction:   %shift
-; AVX: Found an estimated cost of 1 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 1 for instruction:   %shift
+; CHECK-LABEL: 'splatconstant_shift_v4i32'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <4 x i32> %a, <i32 5, i32 5, i32 5, i32 5>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
   %shift = lshr <4 x i32> %a, <i32 5, i32 5, i32 5, i32 5>
   ret <4 x i32> %shift
 }
 
 define <8 x i32> @splatconstant_shift_v8i32(<8 x i32> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v8i32':
-; SSE2: Found an estimated cost of 2 for instruction:   %shift
-; SSE41: Found an estimated cost of 2 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 6 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v8i32'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <8 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v8i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <8 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v8i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v8i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = lshr <8 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v8i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX512-LABEL: 'splatconstant_shift_v8i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
   %shift = lshr <8 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
   ret <8 x i32> %shift
 }
 
 define <16 x i32> @splatconstant_shift_v16i32(<16 x i32> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v16i32':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 8 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 12 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v16i32'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v16i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <16 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v16i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <16 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v16i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <16 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v16i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <16 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX512-LABEL: 'splatconstant_shift_v16i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
   %shift = lshr <16 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
   ret <16 x i32> %shift
 }
 
 define <8 x i16> @splatconstant_shift_v8i16(<8 x i16> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v8i16':
-; SSE2: Found an estimated cost of 1 for instruction:   %shift
-; SSE41: Found an estimated cost of 1 for instruction:   %shift
-; AVX: Found an estimated cost of 1 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 1 for instruction:   %shift
+; CHECK-LABEL: 'splatconstant_shift_v8i16'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
   %shift = lshr <8 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <8 x i16> %shift
 }
 
 define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v16i16':
-; SSE2: Found an estimated cost of 2 for instruction:   %shift
-; SSE41: Found an estimated cost of 2 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 6 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v16i16'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v16i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v16i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v16i16'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v16i16'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512-LABEL: 'splatconstant_shift_v16i16'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
   %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
 }
 
 define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v32i16':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 8 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512F: Found an estimated cost of 2 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 12 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v32i16'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v32i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v32i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v32i16'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v32i16'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512F-LABEL: 'splatconstant_shift_v32i16'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BW-LABEL: 'splatconstant_shift_v32i16'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512VL-LABEL: 'splatconstant_shift_v32i16'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BWVL-LABEL: 'splatconstant_shift_v32i16'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
   %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
 }
 
 define <16 x i8> @splatconstant_shift_v16i8(<16 x i8> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v16i8':
-; SSE2: Found an estimated cost of 2 for instruction:   %shift
-; SSE41: Found an estimated cost of 2 for instruction:   %shift
-; AVX: Found an estimated cost of 2 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 2 for instruction:   %shift
-; XOP: Found an estimated cost of 2 for instruction:   %shift
+; CHECK-LABEL: 'splatconstant_shift_v16i8'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
   %shift = lshr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <16 x i8> %shift
 }
 
 define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v32i8':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 6 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 2 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 6 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v32i8'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v32i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = lshr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v32i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v32i8'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = lshr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v32i8'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512-LABEL: 'splatconstant_shift_v32i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
   %shift = lshr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <32 x i8> %shift
 }
 
 define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v64i8':
-; SSE2: Found an estimated cost of 8 for instruction:   %shift
-; SSE41: Found an estimated cost of 8 for instruction:   %shift
-; AVX: Found an estimated cost of 12 for instruction:   %shift
-; AVX2: Found an estimated cost of 4 for instruction:   %shift
-; AVX512F: Found an estimated cost of 4 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 2 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 12 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 4 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v64i8'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v64i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v64i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v64i8'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v64i8'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512F-LABEL: 'splatconstant_shift_v64i8'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BW-LABEL: 'splatconstant_shift_v64i8'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512VL-LABEL: 'splatconstant_shift_v64i8'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BWVL-LABEL: 'splatconstant_shift_v64i8'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
   %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <64 x i8> %shift
 }
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-shl-cost.ll b/llvm/test/Analysis/CostModel/X86/vshift-shl-cost.ll
index dd123c4d1320..51575f91be9d 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-shl-cost.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-shl-cost.ll
@@ -1,173 +1,411 @@
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+sse2 -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=SSE2
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+sse4.1 -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=SSE41
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx2 -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX2
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+xop,+avx -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=XOP --check-prefix=XOPAVX
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+xop,+avx2 -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=XOP --check-prefix=XOPAVX2
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512f -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512 --check-prefix=AVX512F
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512f,+avx512dq -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512 --check-prefix=AVX512F
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512f,+avx512bw -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512 --check-prefix=AVX512BW
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512f,+avx512vl -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512 --check-prefix=AVX512F
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512f,+avx512dq,+avx512vl -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512 --check-prefix=AVX512F
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512f,+avx512bw,+avx512vl -cost-model -analyze | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512 --check-prefix=AVX512BW
+; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+sse2 | FileCheck %s --check-prefixes=CHECK,SSE,SSE2
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+sse4.1 | FileCheck %s --check-prefixes=CHECK,SSE,SSE41
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX2
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+xop,+avx | FileCheck %s --check-prefixes=CHECK,XOP,XOPAVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+xop,+avx2 | FileCheck %s --check-prefixes=CHECK,XOP,XOPAVX2
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx512f | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512F
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx512f,+avx512dq | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512F
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512BW
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx512f,+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512F
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx512f,+avx512dq,+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512VL
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx512f,+avx512bw,+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512BWVL
 
 ; Verify the cost of vector shift left instructions.
 
-;
 ;
 ; Variable Shifts
 ;
 
 define <2 x i64> @var_shift_v2i64(<2 x i64> %a, <2 x i64> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v2i64':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE-LABEL: 'var_shift_v2i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <2 x i64> %a, %b
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; AVX1-LABEL: 'var_shift_v2i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <2 x i64> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; AVX2-LABEL: 'var_shift_v2i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <2 x i64> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; XOP-LABEL: 'var_shift_v2i64'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <2 x i64> %a, %b
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; AVX512-LABEL: 'var_shift_v2i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <2 x i64> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
   %shift = shl <2 x i64> %a, %b
   ret <2 x i64> %shift
 }
 
 define <4 x i64> @var_shift_v4i64(<4 x i64> %a, <4 x i64> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v4i64':
-; SSE2: Found an estimated cost of 8 for instruction:   %shift
-; SSE41: Found an estimated cost of 8 for instruction:   %shift
-; AVX: Found an estimated cost of 10 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 4 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE-LABEL: 'var_shift_v4i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <4 x i64> %a, %b
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX1-LABEL: 'var_shift_v4i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = shl <4 x i64> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX2-LABEL: 'var_shift_v4i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i64> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOPAVX1-LABEL: 'var_shift_v4i64'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <4 x i64> %a, %b
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOPAVX2-LABEL: 'var_shift_v4i64'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i64> %a, %b
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX512-LABEL: 'var_shift_v4i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i64> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
   %shift = shl <4 x i64> %a, %b
   ret <4 x i64> %shift
 }
 
 define <8 x i64> @var_shift_v8i64(<8 x i64> %a, <8 x i64> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v8i64':
-; SSE2: Found an estimated cost of 16 for instruction:   %shift
-; SSE41: Found an estimated cost of 16 for instruction:   %shift
-; AVX: Found an estimated cost of 20 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 8 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'var_shift_v8i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %shift = shl <8 x i64> %a, %b
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX1-LABEL: 'var_shift_v8i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = shl <8 x i64> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX2-LABEL: 'var_shift_v8i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <8 x i64> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOPAVX1-LABEL: 'var_shift_v8i64'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <8 x i64> %a, %b
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOPAVX2-LABEL: 'var_shift_v8i64'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <8 x i64> %a, %b
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX512-LABEL: 'var_shift_v8i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i64> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
   %shift = shl <8 x i64> %a, %b
   ret <8 x i64> %shift
 }
 
 define <4 x i32> @var_shift_v4i32(<4 x i32> %a, <4 x i32> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v4i32':
-; SSE2: Found an estimated cost of 10 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v4i32'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = shl <4 x i32> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; SSE41-LABEL: 'var_shift_v4i32'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <4 x i32> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; AVX1-LABEL: 'var_shift_v4i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <4 x i32> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; AVX2-LABEL: 'var_shift_v4i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i32> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; XOP-LABEL: 'var_shift_v4i32'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i32> %a, %b
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; AVX512-LABEL: 'var_shift_v4i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i32> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
   %shift = shl <4 x i32> %a, %b
   ret <4 x i32> %shift
 }
 
 define <8 x i32> @var_shift_v8i32(<8 x i32> %a, <8 x i32> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v8i32':
-; SSE2: Found an estimated cost of 20 for instruction:   %shift
-; SSE41: Found an estimated cost of 8 for instruction:   %shift
-; AVX: Found an estimated cost of 10 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 4 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v8i32'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = shl <8 x i32> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; SSE41-LABEL: 'var_shift_v8i32'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <8 x i32> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX1-LABEL: 'var_shift_v8i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = shl <8 x i32> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX2-LABEL: 'var_shift_v8i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i32> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX1-LABEL: 'var_shift_v8i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <8 x i32> %a, %b
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX2-LABEL: 'var_shift_v8i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i32> %a, %b
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX512-LABEL: 'var_shift_v8i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i32> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
   %shift = shl <8 x i32> %a, %b
   ret <8 x i32> %shift
 }
 
 define <16 x i32> @var_shift_v16i32(<16 x i32> %a, <16 x i32> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v16i32':
-; SSE2: Found an estimated cost of 40 for instruction:   %shift
-; SSE41: Found an estimated cost of 16 for instruction:   %shift
-; AVX: Found an estimated cost of 20 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 8 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v16i32'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %shift = shl <16 x i32> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; SSE41-LABEL: 'var_shift_v16i32'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %shift = shl <16 x i32> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX1-LABEL: 'var_shift_v16i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = shl <16 x i32> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX2-LABEL: 'var_shift_v16i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i32> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX1-LABEL: 'var_shift_v16i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <16 x i32> %a, %b
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX2-LABEL: 'var_shift_v16i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i32> %a, %b
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX512-LABEL: 'var_shift_v16i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i32> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
   %shift = shl <16 x i32> %a, %b
   ret <16 x i32> %shift
 }
 
 define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v8i16':
-; SSE2: Found an estimated cost of 32 for instruction:   %shift
-; SSE41: Found an estimated cost of 14 for instruction:   %shift
-; AVX: Found an estimated cost of 14 for instruction:   %shift
-; AVX2: Found an estimated cost of 14 for instruction:   %shift
-; AVX512F: Found an estimated cost of 14 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 1 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v8i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %shift = shl <8 x i16> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; SSE41-LABEL: 'var_shift_v8i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %shift = shl <8 x i16> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX-LABEL: 'var_shift_v8i16'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %shift = shl <8 x i16> %a, %b
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; XOP-LABEL: 'var_shift_v8i16'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %b
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512F-LABEL: 'var_shift_v8i16'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %shift = shl <8 x i16> %a, %b
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512BW-LABEL: 'var_shift_v8i16'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %b
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512VL-LABEL: 'var_shift_v8i16'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %shift = shl <8 x i16> %a, %b
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512BWVL-LABEL: 'var_shift_v8i16'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %b
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
   %shift = shl <8 x i16> %a, %b
   ret <8 x i16> %shift
 }
 
 define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v16i16':
-; SSE2: Found an estimated cost of 64 for instruction:   %shift
-; SSE41: Found an estimated cost of 28 for instruction:   %shift
-; AVX: Found an estimated cost of 30 for instruction:   %shift
-; AVX2: Found an estimated cost of 10 for instruction:   %shift
-; AVX512F: Found an estimated cost of 10 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 4 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v16i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %shift = shl <16 x i16> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; SSE41-LABEL: 'var_shift_v16i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %shift = shl <16 x i16> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX1-LABEL: 'var_shift_v16i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = shl <16 x i16> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX2-LABEL: 'var_shift_v16i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = shl <16 x i16> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; XOP-LABEL: 'var_shift_v16i16'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i16> %a, %b
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512F-LABEL: 'var_shift_v16i16'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = shl <16 x i16> %a, %b
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512BW-LABEL: 'var_shift_v16i16'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %b
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512VL-LABEL: 'var_shift_v16i16'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = shl <16 x i16> %a, %b
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512BWVL-LABEL: 'var_shift_v16i16'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %b
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
   %shift = shl <16 x i16> %a, %b
   ret <16 x i16> %shift
 }
 
 define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v32i16':
-; SSE2: Found an estimated cost of 128 for instruction:   %shift
-; SSE41: Found an estimated cost of 56 for instruction:   %shift
-; AVX: Found an estimated cost of 60 for instruction:   %shift
-; AVX2: Found an estimated cost of 20 for instruction:   %shift
-; AVX512F: Found an estimated cost of 20 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 8 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v32i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 128 for instruction: %shift = shl <32 x i16> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; SSE41-LABEL: 'var_shift_v32i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %shift = shl <32 x i16> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX1-LABEL: 'var_shift_v32i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %shift = shl <32 x i16> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX2-LABEL: 'var_shift_v32i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = shl <32 x i16> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; XOP-LABEL: 'var_shift_v32i16'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <32 x i16> %a, %b
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512F-LABEL: 'var_shift_v32i16'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = shl <32 x i16> %a, %b
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BW-LABEL: 'var_shift_v32i16'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, %b
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512VL-LABEL: 'var_shift_v32i16'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = shl <32 x i16> %a, %b
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BWVL-LABEL: 'var_shift_v32i16'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, %b
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
   %shift = shl <32 x i16> %a, %b
   ret <32 x i16> %shift
 }
 
 define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v16i8':
-; SSE2: Found an estimated cost of 26 for instruction:   %shift
-; SSE41: Found an estimated cost of 11 for instruction:   %shift
-; AVX: Found an estimated cost of 11 for instruction:   %shift
-; AVX2: Found an estimated cost of 11 for instruction:   %shift
-; AVX512: Found an estimated cost of 11 for instruction:   %shift
-; XOP: Found an estimated cost of 1 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v16i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %shift = shl <16 x i8> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; SSE41-LABEL: 'var_shift_v16i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <16 x i8> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX-LABEL: 'var_shift_v16i8'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <16 x i8> %a, %b
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; XOP-LABEL: 'var_shift_v16i8'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i8> %a, %b
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512-LABEL: 'var_shift_v16i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <16 x i8> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
   %shift = shl <16 x i8> %a, %b
   ret <16 x i8> %shift
 }
 
 define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v32i8':
-; SSE2: Found an estimated cost of 52 for instruction:   %shift
-; SSE41: Found an estimated cost of 22 for instruction:   %shift
-; AVX: Found an estimated cost of 24 for instruction:   %shift
-; AVX2: Found an estimated cost of 11 for instruction:   %shift
-; AVX512: Found an estimated cost of 11 for instruction:   %shift
-; XOP: Found an estimated cost of 4 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v32i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %shift = shl <32 x i8> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; SSE41-LABEL: 'var_shift_v32i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <32 x i8> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX1-LABEL: 'var_shift_v32i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = shl <32 x i8> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX2-LABEL: 'var_shift_v32i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <32 x i8> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; XOP-LABEL: 'var_shift_v32i8'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i8> %a, %b
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512-LABEL: 'var_shift_v32i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <32 x i8> %a, %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
   %shift = shl <32 x i8> %a, %b
   ret <32 x i8> %shift
 }
 
 define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
-; CHECK: 'Cost Model Analysis' for function 'var_shift_v64i8':
-; SSE2: Found an estimated cost of 104 for instruction:   %shift
-; SSE41: Found an estimated cost of 44 for instruction:   %shift
-; AVX: Found an estimated cost of 48 for instruction:   %shift
-; AVX2: Found an estimated cost of 22 for instruction:   %shift
-; AVX512F: Found an estimated cost of 22 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 11 for instruction:   %shift
-; XOP: Found an estimated cost of 8 for instruction:   %shift
+; SSE2-LABEL: 'var_shift_v64i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 104 for instruction: %shift = shl <64 x i8> %a, %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; SSE41-LABEL: 'var_shift_v64i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %shift = shl <64 x i8> %a, %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX1-LABEL: 'var_shift_v64i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = shl <64 x i8> %a, %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX2-LABEL: 'var_shift_v64i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <64 x i8> %a, %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; XOP-LABEL: 'var_shift_v64i8'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <64 x i8> %a, %b
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512F-LABEL: 'var_shift_v64i8'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <64 x i8> %a, %b
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BW-LABEL: 'var_shift_v64i8'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <64 x i8> %a, %b
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512VL-LABEL: 'var_shift_v64i8'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <64 x i8> %a, %b
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BWVL-LABEL: 'var_shift_v64i8'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <64 x i8> %a, %b
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
   %shift = shl <64 x i8> %a, %b
   ret <64 x i8> %shift
 }
@@ -177,13 +415,12 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ;
 
 define <2 x i64> @splatvar_shift_v2i64(<2 x i64> %a, i64 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v2i64':
-; SSE2: Found an estimated cost of 1 for instruction:   %shift
-; SSE41: Found an estimated cost of 1 for instruction:   %shift
-; AVX: Found an estimated cost of 1 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 1 for instruction:   %shift
+; CHECK-LABEL: 'splatvar_shift_v2i64'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <2 x i64> undef, i64 %b, i32 0
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <2 x i64> %insert, <2 x i64> undef, <2 x i32> zeroinitializer
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <2 x i64> %a, %splat
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
   %insert = insertelement <2 x i64> undef, i64 %b, i32 0
   %splat = shufflevector <2 x i64> %insert, <2 x i64> undef, <2 x i32> zeroinitializer
   %shift = shl <2 x i64> %a, %splat
@@ -191,14 +428,42 @@ define <2 x i64> @splatvar_shift_v2i64(<2 x i64> %a, i64 %b) {
 }
 
 define <4 x i64> @splatvar_shift_v4i64(<4 x i64> %a, i64 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v4i64':
-; SSE2: Found an estimated cost of 2 for instruction:   %shift
-; SSE41: Found an estimated cost of 2 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 4 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE-LABEL: 'splatvar_shift_v4i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <4 x i64> %a, %splat
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v4i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <4 x i64> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v4i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i64> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v4i64'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <4 x i64> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v4i64'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i64> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v4i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i64> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
   %insert = insertelement <4 x i64> undef, i64 %b, i32 0
   %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
   %shift = shl <4 x i64> %a, %splat
@@ -206,14 +471,42 @@ define <4 x i64> @splatvar_shift_v4i64(<4 x i64> %a, i64 %b) {
 }
 
 define <8 x i64> @splatvar_shift_v8i64(<8 x i64> %a, i64 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v8i64':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 8 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 8 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'splatvar_shift_v8i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <8 x i64> %a, %splat
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v8i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <8 x i64> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v8i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <8 x i64> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v8i64'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <8 x i64> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v8i64'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <8 x i64> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v8i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i64> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
   %insert = insertelement <8 x i64> undef, i64 %b, i32 0
   %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
   %shift = shl <8 x i64> %a, %splat
@@ -221,13 +514,12 @@ define <8 x i64> @splatvar_shift_v8i64(<8 x i64> %a, i64 %b) {
 }
 
 define <4 x i32> @splatvar_shift_v4i32(<4 x i32> %a, i32 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v4i32':
-; SSE2: Found an estimated cost of 1 for instruction:   %shift
-; SSE41: Found an estimated cost of 1 for instruction:   %shift
-; AVX: Found an estimated cost of 1 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 1 for instruction:   %shift
+; CHECK-LABEL: 'splatvar_shift_v4i32'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i32> undef, i32 %b, i32 0
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i32> %insert, <4 x i32> undef, <4 x i32> zeroinitializer
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i32> %a, %splat
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
   %insert = insertelement <4 x i32> undef, i32 %b, i32 0
   %splat = shufflevector <4 x i32> %insert, <4 x i32> undef, <4 x i32> zeroinitializer
   %shift = shl <4 x i32> %a, %splat
@@ -235,14 +527,42 @@ define <4 x i32> @splatvar_shift_v4i32(<4 x i32> %a, i32 %b) {
 }
 
 define <8 x i32> @splatvar_shift_v8i32(<8 x i32> %a, i32 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v8i32':
-; SSE2: Found an estimated cost of 2 for instruction:   %shift
-; SSE41: Found an estimated cost of 2 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 4 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE-LABEL: 'splatvar_shift_v8i32'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <8 x i32> %a, %splat
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v8i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <8 x i32> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v8i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i32> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v8i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <8 x i32> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v8i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i32> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v8i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i32> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
   %insert = insertelement <8 x i32> undef, i32 %b, i32 0
   %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
   %shift = shl <8 x i32> %a, %splat
@@ -250,14 +570,42 @@ define <8 x i32> @splatvar_shift_v8i32(<8 x i32> %a, i32 %b) {
 }
 
 define <16 x i32> @splatvar_shift_v16i32(<16 x i32> %a, i32 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v16i32':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 8 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 8 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'splatvar_shift_v16i32'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i32> %a, %splat
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v16i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <16 x i32> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v16i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i32> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v16i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <16 x i32> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v16i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i32> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v16i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i32> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
   %insert = insertelement <16 x i32> undef, i32 %b, i32 0
   %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
   %shift = shl <16 x i32> %a, %splat
@@ -265,13 +613,36 @@ define <16 x i32> @splatvar_shift_v16i32(<16 x i32> %a, i32 %b) {
 }
 
 define <8 x i16> @splatvar_shift_v8i16(<8 x i16> %a, i16 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v8i16':
-; SSE2: Found an estimated cost of 1 for instruction:   %shift
-; SSE41: Found an estimated cost of 1 for instruction:   %shift
-; AVX: Found an estimated cost of 1 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 1 for instruction:   %shift
+; SSE2-LABEL: 'splatvar_shift_v8i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %splat
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; SSE41-LABEL: 'splatvar_shift_v8i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %splat
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX-LABEL: 'splatvar_shift_v8i16'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %splat
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; XOP-LABEL: 'splatvar_shift_v8i16'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %splat
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v8i16'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
   %insert = insertelement <8 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
   %shift = shl <8 x i16> %a, %splat
@@ -279,14 +650,48 @@ define <8 x i16> @splatvar_shift_v8i16(<8 x i16> %a, i16 %b) {
 }
 
 define <16 x i16> @splatvar_shift_v16i16(<16 x i16> %a, i16 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v16i16':
-; SSE2: Found an estimated cost of 2 for instruction:   %shift
-; SSE41: Found an estimated cost of 2 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 4 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE2-LABEL: 'splatvar_shift_v16i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i16> %a, %splat
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; SSE41-LABEL: 'splatvar_shift_v16i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i16> %a, %splat
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v16i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i16> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v16i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v16i16'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i16> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v16i16'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v16i16'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
   %insert = insertelement <16 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
   %shift = shl <16 x i16> %a, %splat
@@ -294,15 +699,66 @@ define <16 x i16> @splatvar_shift_v16i16(<16 x i16> %a, i16 %b) {
 }
 
 define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v32i16':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 8 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512F: Found an estimated cost of 2 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 8 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE2-LABEL: 'splatvar_shift_v32i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i16> %a, %splat
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; SSE41-LABEL: 'splatvar_shift_v32i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i16> %a, %splat
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v32i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <32 x i16> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v32i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i16> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v32i16'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <32 x i16> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v32i16'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i16> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512F-LABEL: 'splatvar_shift_v32i16'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i16> %a, %splat
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BW-LABEL: 'splatvar_shift_v32i16'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, %splat
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512VL-LABEL: 'splatvar_shift_v32i16'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i16> %a, %splat
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BWVL-LABEL: 'splatvar_shift_v32i16'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, %splat
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
   %shift = shl <32 x i16> %a, %splat
@@ -310,13 +766,36 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 }
 
 define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v16i8':
-; SSE2: Found an estimated cost of 26 for instruction:   %shift
-; SSE41: Found an estimated cost of 11 for instruction:   %shift
-; AVX: Found an estimated cost of 11 for instruction:   %shift
-; AVX2: Found an estimated cost of 11 for instruction:   %shift
-; AVX512: Found an estimated cost of 11 for instruction:   %shift
-; XOP: Found an estimated cost of 1 for instruction:   %shift
+; SSE2-LABEL: 'splatvar_shift_v16i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %shift = shl <16 x i8> %a, %splat
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; SSE41-LABEL: 'splatvar_shift_v16i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <16 x i8> %a, %splat
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX-LABEL: 'splatvar_shift_v16i8'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <16 x i8> %a, %splat
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; XOP-LABEL: 'splatvar_shift_v16i8'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i8> %a, %splat
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v16i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <16 x i8> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
   %insert = insertelement <16 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
   %shift = shl <16 x i8> %a, %splat
@@ -324,13 +803,48 @@ define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
 }
 
 define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v32i8':
-; SSE2: Found an estimated cost of 52 for instruction:   %shift
-; SSE41: Found an estimated cost of 22 for instruction:   %shift
-; AVX: Found an estimated cost of 24 for instruction:   %shift
-; AVX2: Found an estimated cost of 11 for instruction:   %shift
-; AVX512: Found an estimated cost of 11 for instruction:   %shift
-; XOP: Found an estimated cost of 4 for instruction:   %shift
+; SSE2-LABEL: 'splatvar_shift_v32i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %shift = shl <32 x i8> %a, %splat
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; SSE41-LABEL: 'splatvar_shift_v32i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <32 x i8> %a, %splat
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v32i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = shl <32 x i8> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v32i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <32 x i8> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v32i8'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i8> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v32i8'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i8> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512-LABEL: 'splatvar_shift_v32i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <32 x i8> %a, %splat
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
   %insert = insertelement <32 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
   %shift = shl <32 x i8> %a, %splat
@@ -338,14 +852,66 @@ define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 }
 
 define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
-; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v64i8':
-; SSE2: Found an estimated cost of 104 for instruction:   %shift
-; SSE41: Found an estimated cost of 44 for instruction:   %shift
-; AVX: Found an estimated cost of 48 for instruction:   %shift
-; AVX2: Found an estimated cost of 22 for instruction:   %shift
-; AVX512F: Found an estimated cost of 22 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 11 for instruction:   %shift
-; XOP: Found an estimated cost of 8 for instruction:   %shift
+; SSE2-LABEL: 'splatvar_shift_v64i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 104 for instruction: %shift = shl <64 x i8> %a, %splat
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; SSE41-LABEL: 'splatvar_shift_v64i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %shift = shl <64 x i8> %a, %splat
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX1-LABEL: 'splatvar_shift_v64i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = shl <64 x i8> %a, %splat
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX2-LABEL: 'splatvar_shift_v64i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <64 x i8> %a, %splat
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; XOPAVX1-LABEL: 'splatvar_shift_v64i8'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <64 x i8> %a, %splat
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; XOPAVX2-LABEL: 'splatvar_shift_v64i8'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <64 x i8> %a, %splat
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512F-LABEL: 'splatvar_shift_v64i8'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <64 x i8> %a, %splat
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BW-LABEL: 'splatvar_shift_v64i8'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <64 x i8> %a, %splat
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512VL-LABEL: 'splatvar_shift_v64i8'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <64 x i8> %a, %splat
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BWVL-LABEL: 'splatvar_shift_v64i8'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <64 x i8> %a, %splat
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
   %shift = shl <64 x i8> %a, %splat
@@ -357,155 +923,353 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ;
 
 define <2 x i64> @constant_shift_v2i64(<2 x i64> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v2i64':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE-LABEL: 'constant_shift_v2i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <2 x i64> %a, <i64 1, i64 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; AVX1-LABEL: 'constant_shift_v2i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <2 x i64> %a, <i64 1, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; AVX2-LABEL: 'constant_shift_v2i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <2 x i64> %a, <i64 1, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; XOP-LABEL: 'constant_shift_v2i64'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <2 x i64> %a, <i64 1, i64 7>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; AVX512-LABEL: 'constant_shift_v2i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <2 x i64> %a, <i64 1, i64 7>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
   %shift = shl <2 x i64> %a, <i64 1, i64 7>
   ret <2 x i64> %shift
 }
 
 define <4 x i64> @constant_shift_v4i64(<4 x i64> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v4i64':
-; SSE2: Found an estimated cost of 8 for instruction:   %shift
-; SSE41: Found an estimated cost of 8 for instruction:   %shift
-; AVX: Found an estimated cost of 10 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 4 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE-LABEL: 'constant_shift_v4i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <4 x i64> %a, <i64 1, i64 7, i64 15, i64 31>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX1-LABEL: 'constant_shift_v4i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = shl <4 x i64> %a, <i64 1, i64 7, i64 15, i64 31>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX2-LABEL: 'constant_shift_v4i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i64> %a, <i64 1, i64 7, i64 15, i64 31>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOPAVX1-LABEL: 'constant_shift_v4i64'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <4 x i64> %a, <i64 1, i64 7, i64 15, i64 31>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOPAVX2-LABEL: 'constant_shift_v4i64'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i64> %a, <i64 1, i64 7, i64 15, i64 31>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX512-LABEL: 'constant_shift_v4i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i64> %a, <i64 1, i64 7, i64 15, i64 31>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
   %shift = shl <4 x i64> %a, <i64 1, i64 7, i64 15, i64 31>
   ret <4 x i64> %shift
 }
 
 define <8 x i64> @constant_shift_v8i64(<8 x i64> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v8i64':
-; SSE2: Found an estimated cost of 16 for instruction:   %shift
-; SSE41: Found an estimated cost of 16 for instruction:   %shift
-; AVX: Found an estimated cost of 20 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 8 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'constant_shift_v8i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %shift = shl <8 x i64> %a, <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX1-LABEL: 'constant_shift_v8i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shift = shl <8 x i64> %a, <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX2-LABEL: 'constant_shift_v8i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <8 x i64> %a, <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOPAVX1-LABEL: 'constant_shift_v8i64'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <8 x i64> %a, <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOPAVX2-LABEL: 'constant_shift_v8i64'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <8 x i64> %a, <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX512-LABEL: 'constant_shift_v8i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i64> %a, <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
   %shift = shl <8 x i64> %a, <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>
   ret <8 x i64> %shift
 }
 
 define <4 x i32> @constant_shift_v4i32(<4 x i32> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v4i32':
-; SSE2: Found an estimated cost of 6 for instruction:   %shift
-; SSE41: Found an estimated cost of 2 for instruction:   %shift
-; AVX: Found an estimated cost of 2 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v4i32'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = shl <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; SSE41-LABEL: 'constant_shift_v4i32'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; AVX1-LABEL: 'constant_shift_v4i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; AVX2-LABEL: 'constant_shift_v4i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; XOP-LABEL: 'constant_shift_v4i32'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; AVX512-LABEL: 'constant_shift_v4i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
   %shift = shl <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
   ret <4 x i32> %shift
 }
 
 define <8 x i32> @constant_shift_v8i32(<8 x i32> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v8i32':
-; SSE2: Found an estimated cost of 12 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 4 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v8i32'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = shl <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; SSE41-LABEL: 'constant_shift_v8i32'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX1-LABEL: 'constant_shift_v8i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX2-LABEL: 'constant_shift_v8i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX1-LABEL: 'constant_shift_v8i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX2-LABEL: 'constant_shift_v8i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX512-LABEL: 'constant_shift_v8i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
   %shift = shl <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
   ret <8 x i32> %shift
 }
 
 define <16 x i32> @constant_shift_v16i32(<16 x i32> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v16i32':
-; SSE2: Found an estimated cost of 24 for instruction:   %shift
-; SSE41: Found an estimated cost of 8 for instruction:   %shift
-; AVX: Found an estimated cost of 8 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 8 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v16i32'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = shl <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; SSE41-LABEL: 'constant_shift_v16i32'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX1-LABEL: 'constant_shift_v16i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX2-LABEL: 'constant_shift_v16i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX1-LABEL: 'constant_shift_v16i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX2-LABEL: 'constant_shift_v16i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX512-LABEL: 'constant_shift_v16i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
   %shift = shl <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
   ret <16 x i32> %shift
 }
 
 define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v8i16':
-; SSE2: Found an estimated cost of 1 for instruction:   %shift
-; SSE41: Found an estimated cost of 1 for instruction:   %shift
-; AVX: Found an estimated cost of 1 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 1 for instruction:   %shift
+; CHECK-LABEL: 'constant_shift_v8i16'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
   %shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
 }
 
 define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v16i16':
-; SSE2: Found an estimated cost of 2 for instruction:   %shift
-; SSE41: Found an estimated cost of 2 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 4 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE-LABEL: 'constant_shift_v16i16'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX1-LABEL: 'constant_shift_v16i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX2-LABEL: 'constant_shift_v16i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; XOPAVX1-LABEL: 'constant_shift_v16i16'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; XOPAVX2-LABEL: 'constant_shift_v16i16'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512-LABEL: 'constant_shift_v16i16'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
   %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
 }
 
 define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v32i16':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 8 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512F: Found an estimated cost of 2 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 8 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'constant_shift_v32i16'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX1-LABEL: 'constant_shift_v32i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX2-LABEL: 'constant_shift_v32i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; XOPAVX1-LABEL: 'constant_shift_v32i16'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; XOPAVX2-LABEL: 'constant_shift_v32i16'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512F-LABEL: 'constant_shift_v32i16'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BW-LABEL: 'constant_shift_v32i16'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512VL-LABEL: 'constant_shift_v32i16'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BWVL-LABEL: 'constant_shift_v32i16'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
   %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
 }
 
 define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v16i8':
-; SSE2: Found an estimated cost of 26 for instruction:   %shift
-; SSE41: Found an estimated cost of 11 for instruction:   %shift
-; AVX: Found an estimated cost of 11 for instruction:   %shift
-; AVX2: Found an estimated cost of 11 for instruction:   %shift
-; AVX512: Found an estimated cost of 11 for instruction:   %shift
-; XOP: Found an estimated cost of 1 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v16i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; SSE41-LABEL: 'constant_shift_v16i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX-LABEL: 'constant_shift_v16i8'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; XOP-LABEL: 'constant_shift_v16i8'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512-LABEL: 'constant_shift_v16i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
   %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
 }
 
 define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v32i8':
-; SSE2: Found an estimated cost of 52 for instruction:   %shift
-; SSE41: Found an estimated cost of 22 for instruction:   %shift
-; AVX: Found an estimated cost of 24 for instruction:   %shift
-; AVX2: Found an estimated cost of 11 for instruction:   %shift
-; AVX512: Found an estimated cost of 11 for instruction:   %shift
-; XOP: Found an estimated cost of 4 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v32i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; SSE41-LABEL: 'constant_shift_v32i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX1-LABEL: 'constant_shift_v32i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX2-LABEL: 'constant_shift_v32i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; XOP-LABEL: 'constant_shift_v32i8'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512-LABEL: 'constant_shift_v32i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
   %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <32 x i8> %shift
 }
 
 define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
-; CHECK: 'Cost Model Analysis' for function 'constant_shift_v64i8':
-; SSE2: Found an estimated cost of 104 for instruction:   %shift
-; SSE41: Found an estimated cost of 44 for instruction:   %shift
-; AVX: Found an estimated cost of 48 for instruction:   %shift
-; AVX2: Found an estimated cost of 22 for instruction:   %shift
-; AVX512F: Found an estimated cost of 22 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 11 for instruction:   %shift
-; XOP: Found an estimated cost of 8 for instruction:   %shift
+; SSE2-LABEL: 'constant_shift_v64i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 104 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; SSE41-LABEL: 'constant_shift_v64i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX1-LABEL: 'constant_shift_v64i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX2-LABEL: 'constant_shift_v64i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; XOP-LABEL: 'constant_shift_v64i8'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512F-LABEL: 'constant_shift_v64i8'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BW-LABEL: 'constant_shift_v64i8'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512VL-LABEL: 'constant_shift_v64i8'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BWVL-LABEL: 'constant_shift_v64i8'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
   %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
 }
@@ -515,157 +1279,305 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ;
 
 define <2 x i64> @splatconstant_shift_v2i64(<2 x i64> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v2i64':
-; SSE2: Found an estimated cost of 1 for instruction:   %shift
-; SSE41: Found an estimated cost of 1 for instruction:   %shift
-; AVX: Found an estimated cost of 1 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; CHECK-LABEL: 'splatconstant_shift_v2i64'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <2 x i64> %a, <i64 7, i64 7>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
   %shift = shl <2 x i64> %a, <i64 7, i64 7>
   ret <2 x i64> %shift
 }
 
 define <4 x i64> @splatconstant_shift_v4i64(<4 x i64> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v4i64':
-; SSE2: Found an estimated cost of 2 for instruction:   %shift
-; SSE41: Found an estimated cost of 2 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 4 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v4i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <4 x i64> %a, <i64 7, i64 7, i64 7, i64 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v4i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <4 x i64> %a, <i64 7, i64 7, i64 7, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v4i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i64> %a, <i64 7, i64 7, i64 7, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v4i64'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <4 x i64> %a, <i64 7, i64 7, i64 7, i64 7>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v4i64'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i64> %a, <i64 7, i64 7, i64 7, i64 7>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
+; AVX512-LABEL: 'splatconstant_shift_v4i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i64> %a, <i64 7, i64 7, i64 7, i64 7>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
   %shift = shl <4 x i64> %a, <i64 7, i64 7, i64 7, i64 7>
   ret <4 x i64> %shift
 }
 
 define <8 x i64> @splatconstant_shift_v8i64(<8 x i64> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v8i64':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 8 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 8 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v8i64'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <8 x i64> %a, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v8i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <8 x i64> %a, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v8i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <8 x i64> %a, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v8i64'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <8 x i64> %a, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v8i64'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <8 x i64> %a, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
+; AVX512-LABEL: 'splatconstant_shift_v8i64'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i64> %a, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
   %shift = shl <8 x i64> %a, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
   ret <8 x i64> %shift
 }
 
 define <4 x i32> @splatconstant_shift_v4i32(<4 x i32> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v4i32':
-; SSE2: Found an estimated cost of 1 for instruction:   %shift
-; SSE41: Found an estimated cost of 1 for instruction:   %shift
-; AVX: Found an estimated cost of 1 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; CHECK-LABEL: 'splatconstant_shift_v4i32'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i32> %a, <i32 5, i32 5, i32 5, i32 5>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
   %shift = shl <4 x i32> %a, <i32 5, i32 5, i32 5, i32 5>
   ret <4 x i32> %shift
 }
 
 define <8 x i32> @splatconstant_shift_v8i32(<8 x i32> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v8i32':
-; SSE2: Found an estimated cost of 2 for instruction:   %shift
-; SSE41: Found an estimated cost of 2 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 4 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v8i32'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <8 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v8i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <8 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v8i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v8i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <8 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v8i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
+; AVX512-LABEL: 'splatconstant_shift_v8i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
   %shift = shl <8 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
   ret <8 x i32> %shift
 }
 
 define <16 x i32> @splatconstant_shift_v16i32(<16 x i32> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v16i32':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 8 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 8 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v16i32'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v16i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <16 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v16i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v16i32'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <16 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v16i32'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
+; AVX512-LABEL: 'splatconstant_shift_v16i32'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
   %shift = shl <16 x i32> %a, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
   ret <16 x i32> %shift
 }
 
 define <8 x i16> @splatconstant_shift_v8i16(<8 x i16> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v8i16':
-; SSE2: Found an estimated cost of 1 for instruction:   %shift
-; SSE41: Found an estimated cost of 1 for instruction:   %shift
-; AVX: Found an estimated cost of 1 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOP: Found an estimated cost of 1 for instruction:   %shift
+; CHECK-LABEL: 'splatconstant_shift_v8i16'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
   %shift = shl <8 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <8 x i16> %shift
 }
 
 define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v16i16':
-; SSE2: Found an estimated cost of 2 for instruction:   %shift
-; SSE41: Found an estimated cost of 2 for instruction:   %shift
-; AVX: Found an estimated cost of 4 for instruction:   %shift
-; AVX2: Found an estimated cost of 1 for instruction:   %shift
-; AVX512: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 4 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v16i16'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v16i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v16i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v16i16'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v16i16'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512-LABEL: 'splatconstant_shift_v16i16'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
   %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
 }
 
 define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v32i16':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 8 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512F: Found an estimated cost of 2 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 1 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 8 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v32i16'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v32i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v32i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v32i16'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v32i16'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512F-LABEL: 'splatconstant_shift_v32i16'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BW-LABEL: 'splatconstant_shift_v32i16'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512VL-LABEL: 'splatconstant_shift_v32i16'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512BWVL-LABEL: 'splatconstant_shift_v32i16'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
   %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
 }
 
 define <16 x i8> @splatconstant_shift_v16i8(<16 x i8> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v16i8':
-; SSE2: Found an estimated cost of 2 for instruction:   %shift
-; SSE41: Found an estimated cost of 2 for instruction:   %shift
-; AVX: Found an estimated cost of 2 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 2 for instruction:   %shift
-; XOP: Found an estimated cost of 1 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v16i8'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX-LABEL: 'splatconstant_shift_v16i8'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; XOP-LABEL: 'splatconstant_shift_v16i8'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512-LABEL: 'splatconstant_shift_v16i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
   %shift = shl <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <16 x i8> %shift
 }
 
 define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v32i8':
-; SSE2: Found an estimated cost of 4 for instruction:   %shift
-; SSE41: Found an estimated cost of 4 for instruction:   %shift
-; AVX: Found an estimated cost of 6 for instruction:   %shift
-; AVX2: Found an estimated cost of 2 for instruction:   %shift
-; AVX512: Found an estimated cost of 2 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 4 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v32i8'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v32i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = shl <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v32i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v32i8'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v32i8'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512-LABEL: 'splatconstant_shift_v32i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
   %shift = shl <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <32 x i8> %shift
 }
 
 define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
-; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v64i8':
-; SSE2: Found an estimated cost of 8 for instruction:   %shift
-; SSE41: Found an estimated cost of 8 for instruction:   %shift
-; AVX: Found an estimated cost of 12 for instruction:   %shift
-; AVX2: Found an estimated cost of 4 for instruction:   %shift
-; AVX512F: Found an estimated cost of 4 for instruction:   %shift
-; AVX512BW: Found an estimated cost of 2 for instruction:   %shift
-; XOPAVX: Found an estimated cost of 8 for instruction:   %shift
-; XOPAVX2: Found an estimated cost of 4 for instruction:   %shift
+; SSE-LABEL: 'splatconstant_shift_v64i8'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX1-LABEL: 'splatconstant_shift_v64i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX2-LABEL: 'splatconstant_shift_v64i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; XOPAVX1-LABEL: 'splatconstant_shift_v64i8'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; XOPAVX2-LABEL: 'splatconstant_shift_v64i8'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512F-LABEL: 'splatconstant_shift_v64i8'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BW-LABEL: 'splatconstant_shift_v64i8'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512VL-LABEL: 'splatconstant_shift_v64i8'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512BWVL-LABEL: 'splatconstant_shift_v64i8'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
   %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <64 x i8> %shift
 }
@@ -678,20 +1590,22 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; non-uniform constant.
 
 define <8 x i16> @test1(<8 x i16> %a) {
+; CHECK-LABEL: 'test1'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+;
   %shl = shl <8 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
   ret <8 x i16> %shl
 }
-; CHECK: 'Cost Model Analysis' for function 'test1':
-; CHECK: Found an estimated cost of 1 for instruction:   %shl
-
 
 define <8 x i16> @test2(<8 x i16> %a) {
+; CHECK-LABEL: 'test2'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 0, i16 undef, i16 0, i16 0, i16 1, i16 undef, i16 -1, i16 1>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+;
   %shl = shl <8 x i16> %a, <i16 0, i16 undef, i16 0, i16 0, i16 1, i16 undef, i16 -1, i16 1>
   ret <8 x i16> %shl
 }
-; CHECK: 'Cost Model Analysis' for function 'test2':
-; CHECK: Found an estimated cost of 1 for instruction:   %shl
-
 
 ; With SSE4.1, v4i32 shifts can be lowered into a single pmulld instruction.
 ; Make sure that the estimated cost is always 2 except for the case where
@@ -699,28 +1613,62 @@ define <8 x i16> @test2(<8 x i16> %a) {
 ; v4i32 mul as a 2x shuffle, 2x pmuludq, 2x shuffle.
 
 define <4 x i32> @test3(<4 x i32> %a) {
+; SSE2-LABEL: 'test3'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shl = shl <4 x i32> %a, <i32 1, i32 -1, i32 2, i32 -3>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shl
+;
+; SSE41-LABEL: 'test3'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shl = shl <4 x i32> %a, <i32 1, i32 -1, i32 2, i32 -3>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shl
+;
+; AVX1-LABEL: 'test3'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shl = shl <4 x i32> %a, <i32 1, i32 -1, i32 2, i32 -3>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shl
+;
+; AVX2-LABEL: 'test3'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <4 x i32> %a, <i32 1, i32 -1, i32 2, i32 -3>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shl
+;
+; XOP-LABEL: 'test3'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <4 x i32> %a, <i32 1, i32 -1, i32 2, i32 -3>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shl
+;
+; AVX512-LABEL: 'test3'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <4 x i32> %a, <i32 1, i32 -1, i32 2, i32 -3>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shl
+;
   %shl = shl <4 x i32> %a, <i32 1, i32 -1, i32 2, i32 -3>
   ret <4 x i32> %shl
 }
-; CHECK: 'Cost Model Analysis' for function 'test3':
-; SSE2: Found an estimated cost of 6 for instruction:   %shl
-; SSE41: Found an estimated cost of 2 for instruction:   %shl
-; AVX: Found an estimated cost of 2 for instruction:   %shl
-; AVX2: Found an estimated cost of 1 for instruction:   %shl
-; XOP: Found an estimated cost of 1 for instruction:   %shl
-
 
 define <4 x i32> @test4(<4 x i32> %a) {
+; SSE2-LABEL: 'test4'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shl = shl <4 x i32> %a, <i32 0, i32 0, i32 1, i32 1>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shl
+;
+; SSE41-LABEL: 'test4'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shl = shl <4 x i32> %a, <i32 0, i32 0, i32 1, i32 1>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shl
+;
+; AVX1-LABEL: 'test4'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shl = shl <4 x i32> %a, <i32 0, i32 0, i32 1, i32 1>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shl
+;
+; AVX2-LABEL: 'test4'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <4 x i32> %a, <i32 0, i32 0, i32 1, i32 1>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shl
+;
+; XOP-LABEL: 'test4'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <4 x i32> %a, <i32 0, i32 0, i32 1, i32 1>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shl
+;
+; AVX512-LABEL: 'test4'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <4 x i32> %a, <i32 0, i32 0, i32 1, i32 1>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shl
+;
   %shl = shl <4 x i32> %a, <i32 0, i32 0, i32 1, i32 1>
   ret <4 x i32> %shl
 }
-; CHECK: 'Cost Model Analysis' for function 'test4':
-; SSE2: Found an estimated cost of 6 for instruction:   %shl
-; SSE41: Found an estimated cost of 2 for instruction:   %shl
-; AVX: Found an estimated cost of 2 for instruction:   %shl
-; AVX2: Found an estimated cost of 1 for instruction:   %shl
-; XOP: Found an estimated cost of 1 for instruction:   %shl
-
 
 ; On AVX2 we are able to lower the following shift into a single
 ; vpsllvq. Therefore, the expected cost is only 1.
@@ -728,16 +1676,29 @@ define <4 x i32> @test4(<4 x i32> %a) {
 ; vpsllv instructions.
 
 define <2 x i64> @test5(<2 x i64> %a) {
+; SSE-LABEL: 'test5'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <2 x i64> %a, <i64 2, i64 3>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shl
+;
+; AVX1-LABEL: 'test5'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <2 x i64> %a, <i64 2, i64 3>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shl
+;
+; AVX2-LABEL: 'test5'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <2 x i64> %a, <i64 2, i64 3>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shl
+;
+; XOP-LABEL: 'test5'
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <2 x i64> %a, <i64 2, i64 3>
+; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shl
+;
+; AVX512-LABEL: 'test5'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <2 x i64> %a, <i64 2, i64 3>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shl
+;
   %shl = shl <2 x i64> %a, <i64 2, i64 3>
   ret <2 x i64> %shl
 }
-; CHECK: 'Cost Model Analysis' for function 'test5':
-; SSE2: Found an estimated cost of 4 for instruction:   %shl
-; SSE41: Found an estimated cost of 4 for instruction:   %shl
-; AVX: Found an estimated cost of 4 for instruction:   %shl
-; AVX2: Found an estimated cost of 1 for instruction:   %shl
-; XOP: Found an estimated cost of 1 for instruction:   %shl
-
 
 ; v16i16 and v8i32 shift left by non-uniform constant are lowered into
 ; vector multiply instructions.  With AVX (but not AVX2), the vector multiply
@@ -751,36 +1712,71 @@ define <2 x i64> @test5(<2 x i64> %a) {
 ; the cost computed in the case of 'test1'. That is because the backend
 ; simply emits 2 pmullw with no extract/insert.
 
-
 define <16 x i16> @test6(<16 x i16> %a) {
+; SSE-LABEL: 'test6'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
+;
+; AVX1-LABEL: 'test6'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
+;
+; AVX2-LABEL: 'test6'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
+;
+; XOPAVX1-LABEL: 'test6'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
+;
+; XOPAVX2-LABEL: 'test6'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
+;
+; AVX512-LABEL: 'test6'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
+;
   %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
   ret <16 x i16> %shl
 }
-; CHECK: 'Cost Model Analysis' for function 'test6':
-; SSE2: Found an estimated cost of 2 for instruction:   %shl
-; SSE41: Found an estimated cost of 2 for instruction:   %shl
-; AVX: Found an estimated cost of 4 for instruction:   %shl
-; AVX2: Found an estimated cost of 1 for instruction:   %shl
-; XOPAVX: Found an estimated cost of 4 for instruction:   %shl
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shl
-
 
 ; With SSE2 and SSE4.1, the vector shift cost for 'test7' is twice
 ; the cost computed in the case of 'test3'. That is because the multiply
 ; is type-legalized into two 4i32 vector multiply.
 
 define <8 x i32> @test7(<8 x i32> %a) {
+; SSE2-LABEL: 'test7'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shl = shl <8 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shl
+;
+; SSE41-LABEL: 'test7'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <8 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shl
+;
+; AVX1-LABEL: 'test7'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <8 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shl
+;
+; AVX2-LABEL: 'test7'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shl
+;
+; XOPAVX1-LABEL: 'test7'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <8 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shl
+;
+; XOPAVX2-LABEL: 'test7'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shl
+;
+; AVX512-LABEL: 'test7'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shl
+;
   %shl = shl <8 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
   ret <8 x i32> %shl
 }
-; CHECK: 'Cost Model Analysis' for function 'test7':
-; SSE2: Found an estimated cost of 12 for instruction:   %shl
-; SSE41: Found an estimated cost of 4 for instruction:   %shl
-; AVX: Found an estimated cost of 4 for instruction:   %shl
-; AVX2: Found an estimated cost of 1 for instruction:   %shl
-; XOPAVX: Found an estimated cost of 4 for instruction:   %shl
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shl
-
 
 ; On AVX2 we are able to lower the following shift into a single
 ; vpsllvq. Therefore, the expected cost is only 1.
@@ -788,47 +1784,111 @@ define <8 x i32> @test7(<8 x i32> %a) {
 ; vpsllv instructions.
 
 define <4 x i64> @test8(<4 x i64> %a) {
+; SSE-LABEL: 'test8'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shl = shl <4 x i64> %a, <i64 1, i64 2, i64 3, i64 4>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shl
+;
+; AVX1-LABEL: 'test8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shl = shl <4 x i64> %a, <i64 1, i64 2, i64 3, i64 4>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shl
+;
+; AVX2-LABEL: 'test8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <4 x i64> %a, <i64 1, i64 2, i64 3, i64 4>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shl
+;
+; XOPAVX1-LABEL: 'test8'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <4 x i64> %a, <i64 1, i64 2, i64 3, i64 4>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shl
+;
+; XOPAVX2-LABEL: 'test8'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <4 x i64> %a, <i64 1, i64 2, i64 3, i64 4>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shl
+;
+; AVX512-LABEL: 'test8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <4 x i64> %a, <i64 1, i64 2, i64 3, i64 4>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shl
+;
   %shl = shl <4 x i64> %a, <i64 1, i64 2, i64 3, i64 4>
   ret <4 x i64> %shl
 }
-; CHECK: 'Cost Model Analysis' for function 'test8':
-; SSE2: Found an estimated cost of 8 for instruction:   %shl
-; SSE41: Found an estimated cost of 8 for instruction:   %shl
-; AVX: Found an estimated cost of 10 for instruction:   %shl
-; AVX2: Found an estimated cost of 1 for instruction:   %shl
-; XOPAVX: Found an estimated cost of 4 for instruction:   %shl
-; XOPAVX2: Found an estimated cost of 1 for instruction:   %shl
-
 
 ; Same as 'test6', with the difference that the cost is double.
 
 define <32 x i16> @test9(<32 x i16> %a) {
+; SSE-LABEL: 'test9'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
+;
+; AVX1-LABEL: 'test9'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
+;
+; AVX2-LABEL: 'test9'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
+;
+; XOPAVX1-LABEL: 'test9'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
+;
+; XOPAVX2-LABEL: 'test9'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
+;
+; AVX512F-LABEL: 'test9'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
+;
+; AVX512BW-LABEL: 'test9'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
+;
+; AVX512VL-LABEL: 'test9'
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; AVX512VL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
+;
+; AVX512BWVL-LABEL: 'test9'
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
+;
   %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
   ret <32 x i16> %shl
 }
-; CHECK: 'Cost Model Analysis' for function 'test9':
-; SSE2: Found an estimated cost of 4 for instruction:   %shl
-; SSE41: Found an estimated cost of 4 for instruction:   %shl
-; AVX: Found an estimated cost of 8 for instruction:   %shl
-; AVX2: Found an estimated cost of 2 for instruction:   %shl
-; XOPAVX: Found an estimated cost of 8 for instruction:   %shl
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shl
-
 
 ; Same as 'test7', except that now the cost is double.
 
 define <16 x i32> @test10(<16 x i32> %a) {
+; SSE2-LABEL: 'test10'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %shl = shl <16 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shl
+;
+; SSE41-LABEL: 'test10'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shl = shl <16 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shl
+;
+; AVX1-LABEL: 'test10'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shl = shl <16 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shl
+;
+; AVX2-LABEL: 'test10'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shl = shl <16 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shl
+;
+; XOPAVX1-LABEL: 'test10'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shl = shl <16 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shl
+;
+; XOPAVX2-LABEL: 'test10'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shl = shl <16 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shl
+;
+; AVX512-LABEL: 'test10'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <16 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shl
+;
   %shl = shl <16 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
   ret <16 x i32> %shl
 }
-; CHECK: 'Cost Model Analysis' for function 'test10':
-; SSE2: Found an estimated cost of 24 for instruction:   %shl
-; SSE41: Found an estimated cost of 8 for instruction:   %shl
-; AVX: Found an estimated cost of 8 for instruction:   %shl
-; AVX2: Found an estimated cost of 2 for instruction:   %shl
-; XOPAVX: Found an estimated cost of 8 for instruction:   %shl
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shl
-
 
 ; On AVX2 we are able to lower the following shift into a sequence of
 ; two vpsllvq instructions. Therefore, the expected cost is only 2.
@@ -836,13 +1896,30 @@ define <16 x i32> @test10(<16 x i32> %a) {
 ; instructions.
 
 define <8 x i64> @test11(<8 x i64> %a) {
+; SSE-LABEL: 'test11'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %shl = shl <8 x i64> %a, <i64 1, i64 1, i64 2, i64 3, i64 1, i64 1, i64 2, i64 3>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shl
+;
+; AVX1-LABEL: 'test11'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %shl = shl <8 x i64> %a, <i64 1, i64 1, i64 2, i64 3, i64 1, i64 1, i64 2, i64 3>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shl
+;
+; AVX2-LABEL: 'test11'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shl = shl <8 x i64> %a, <i64 1, i64 1, i64 2, i64 3, i64 1, i64 1, i64 2, i64 3>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shl
+;
+; XOPAVX1-LABEL: 'test11'
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shl = shl <8 x i64> %a, <i64 1, i64 1, i64 2, i64 3, i64 1, i64 1, i64 2, i64 3>
+; XOPAVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shl
+;
+; XOPAVX2-LABEL: 'test11'
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shl = shl <8 x i64> %a, <i64 1, i64 1, i64 2, i64 3, i64 1, i64 1, i64 2, i64 3>
+; XOPAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shl
+;
+; AVX512-LABEL: 'test11'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i64> %a, <i64 1, i64 1, i64 2, i64 3, i64 1, i64 1, i64 2, i64 3>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shl
+;
   %shl = shl <8 x i64> %a, <i64 1, i64 1, i64 2, i64 3, i64 1, i64 1, i64 2, i64 3>
   ret <8 x i64> %shl
 }
-; CHECK: 'Cost Model Analysis' for function 'test11':
-; SSE2: Found an estimated cost of 16 for instruction:   %shl
-; SSE41: Found an estimated cost of 16 for instruction:   %shl
-; AVX: Found an estimated cost of 20 for instruction:   %shl
-; AVX2: Found an estimated cost of 2 for instruction:   %shl
-; XOPAVX: Found an estimated cost of 8 for instruction:   %shl
-; XOPAVX2: Found an estimated cost of 2 for instruction:   %shl