Updated and extended the information about each instruction in HSW and SNB to include the following data:

•static latency
•number of uOps from which the instructions consists
•all ports used by the instruction

Reviewers: 
 RKSimon 
 zvi  
aymanmus  
m_zuckerman 

Differential Revision: https://reviews.llvm.org/D33897
 

llvm-svn: 306414

llvm/test/CodeGen/X86/avx2-intrinsics-x86.ll

@@ -1619,10 +1619,10 @@ define <8 x float>  @test_gather_mask(<8 x float> %a0, float* %a, <8 x i32> %idx
 ;
 ; AVX512VL-LABEL: test_gather_mask:
 ; AVX512VL:       ## BB#0:
-; AVX512VL-NEXT:    movl {{[0-9]+}}(%esp), %eax ## encoding: [0x8b,0x44,0x24,0x04]
-; AVX512VL-NEXT:    vmovaps %ymm2, %ymm3 ## EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xda]
-; AVX512VL-NEXT:    vgatherdps %ymm3, (%eax,%ymm1,4), %ymm0 ## encoding: [0xc4,0xe2,0x65,0x92,0x04,0x88]
 ; AVX512VL-NEXT:    movl {{[0-9]+}}(%esp), %eax ## encoding: [0x8b,0x44,0x24,0x08]
+; AVX512VL-NEXT:    movl {{[0-9]+}}(%esp), %ecx ## encoding: [0x8b,0x4c,0x24,0x04]
+; AVX512VL-NEXT:    vmovaps %ymm2, %ymm3 ## EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xda]
+; AVX512VL-NEXT:    vgatherdps %ymm3, (%ecx,%ymm1,4), %ymm0 ## encoding: [0xc4,0xe2,0x65,0x92,0x04,0x89]
 ; AVX512VL-NEXT:    vmovups %ymm2, (%eax) ## EVEX TO VEX Compression encoding: [0xc5,0xfc,0x11,0x10]
 ; AVX512VL-NEXT:    retl ## encoding: [0xc3]
   %a_i8 = bitcast float* %a to i8*

llvm/test/CodeGen/X86/avx2-schedule.ll

@@ -9,7 +9,7 @@ define <32 x i8> @test_pabsb(<32 x i8> %a0, <32 x i8> *%a1) {
 ; HASWELL-NEXT:    vpabsb %ymm0, %ymm0 # sched: [1:0.50]
 ; HASWELL-NEXT:    vpabsb (%rdi), %ymm1 # sched: [5:0.50]
 ; HASWELL-NEXT:    vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; ZNVER1-LABEL: test_pabsb:
 ; ZNVER1:       # BB#0:
@@ -29,9 +29,9 @@ define <8 x i32> @test_pabsd(<8 x i32> %a0, <8 x i32> *%a1) {
 ; HASWELL-LABEL: test_pabsd:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpabsd %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT:    vpabsd (%rdi), %ymm1 # sched: [5:0.50]
+; HASWELL-NEXT:    vpabsd (%rdi), %ymm1 # sched: [1:0.50]
 ; HASWELL-NEXT:    vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; ZNVER1-LABEL: test_pabsd:
 ; ZNVER1:       # BB#0:
@@ -51,9 +51,9 @@ define <16 x i16> @test_pabsw(<16 x i16> %a0, <16 x i16> *%a1) {
 ; HASWELL-LABEL: test_pabsw:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpabsw %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT:    vpabsw (%rdi), %ymm1 # sched: [5:0.50]
+; HASWELL-NEXT:    vpabsw (%rdi), %ymm1 # sched: [1:0.50]
 ; HASWELL-NEXT:    vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; ZNVER1-LABEL: test_pabsw:
 ; ZNVER1:       # BB#0:
@@ -74,7 +74,7 @@ define <32 x i8> @test_paddb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; HASWELL-NEXT:    vpaddb (%rdi), %ymm0, %ymm0 # sched: [5:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; ZNVER1-LABEL: test_paddb:
 ; ZNVER1:       # BB#0:
@@ -92,7 +92,7 @@ define <8 x i32> @test_paddd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; HASWELL-NEXT:    vpaddd (%rdi), %ymm0, %ymm0 # sched: [5:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; ZNVER1-LABEL: test_paddd:
 ; ZNVER1:       # BB#0:
@@ -109,8 +109,8 @@ define <4 x i64> @test_paddq(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
 ; HASWELL-LABEL: test_paddq:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT:    vpaddq (%rdi), %ymm0, %ymm0 # sched: [5:0.50]
+; HASWELL-NEXT:    vpaddq (%rdi), %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; ZNVER1-LABEL: test_paddq:
 ; ZNVER1:       # BB#0:
@@ -128,7 +128,7 @@ define <16 x i16> @test_paddw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; HASWELL-NEXT:    vpaddw (%rdi), %ymm0, %ymm0 # sched: [5:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; ZNVER1-LABEL: test_paddw:
 ; ZNVER1:       # BB#0:
@@ -145,9 +145,9 @@ define <4 x i64> @test_pand(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
 ; HASWELL-LABEL: test_pand:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpand %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; HASWELL-NEXT:    vpand (%rdi), %ymm0, %ymm0 # sched: [5:0.50]
+; HASWELL-NEXT:    vpand (%rdi), %ymm0, %ymm0 # sched: [1:0.50]
 ; HASWELL-NEXT:    vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; ZNVER1-LABEL: test_pand:
 ; ZNVER1:       # BB#0:
@@ -166,9 +166,9 @@ define <4 x i64> @test_pandn(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
 ; HASWELL-LABEL: test_pandn:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpandn %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; HASWELL-NEXT:    vpandn (%rdi), %ymm0, %ymm1 # sched: [5:0.50]
+; HASWELL-NEXT:    vpandn (%rdi), %ymm0, %ymm1 # sched: [1:0.50]
 ; HASWELL-NEXT:    vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; ZNVER1-LABEL: test_pandn:
 ; ZNVER1:       # BB#0:
@@ -190,7 +190,7 @@ define <8 x i32> @test_pmulld(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpmulld %ymm1, %ymm0, %ymm0 # sched: [10:2.00]
 ; HASWELL-NEXT:    vpmulld (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; ZNVER1-LABEL: test_pmulld:
 ; ZNVER1:       # BB#0:
@@ -207,8 +207,8 @@ define <16 x i16> @test_pmullw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2)
 ; HASWELL-LABEL: test_pmullw:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpmullw %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; HASWELL-NEXT:    vpmullw (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
+; HASWELL-NEXT:    vpmullw (%rdi), %ymm0, %ymm0 # sched: [5:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; ZNVER1-LABEL: test_pmullw:
 ; ZNVER1:       # BB#0:
@@ -225,9 +225,9 @@ define <4 x i64> @test_por(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
 ; HASWELL-LABEL: test_por:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; HASWELL-NEXT:    vpor (%rdi), %ymm0, %ymm0 # sched: [5:0.50]
+; HASWELL-NEXT:    vpor (%rdi), %ymm0, %ymm0 # sched: [1:0.50]
 ; HASWELL-NEXT:    vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; ZNVER1-LABEL: test_por:
 ; ZNVER1:       # BB#0:
@@ -246,8 +246,8 @@ define <32 x i8> @test_psubb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
 ; HASWELL-LABEL: test_psubb:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpsubb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT:    vpsubb (%rdi), %ymm0, %ymm0 # sched: [5:0.50]
+; HASWELL-NEXT:    vpsubb (%rdi), %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; ZNVER1-LABEL: test_psubb:
 ; ZNVER1:       # BB#0:
@@ -264,8 +264,8 @@ define <8 x i32> @test_psubd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
 ; HASWELL-LABEL: test_psubd:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT:    vpsubd (%rdi), %ymm0, %ymm0 # sched: [5:0.50]
+; HASWELL-NEXT:    vpsubd (%rdi), %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; ZNVER1-LABEL: test_psubd:
 ; ZNVER1:       # BB#0:
@@ -282,8 +282,8 @@ define <4 x i64> @test_psubq(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
 ; HASWELL-LABEL: test_psubq:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpsubq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT:    vpsubq (%rdi), %ymm0, %ymm0 # sched: [5:0.50]
+; HASWELL-NEXT:    vpsubq (%rdi), %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; ZNVER1-LABEL: test_psubq:
 ; ZNVER1:       # BB#0:
@@ -300,8 +300,8 @@ define <16 x i16> @test_psubw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
 ; HASWELL-LABEL: test_psubw:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpsubw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT:    vpsubw (%rdi), %ymm0, %ymm0 # sched: [5:0.50]
+; HASWELL-NEXT:    vpsubw (%rdi), %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; ZNVER1-LABEL: test_psubw:
 ; ZNVER1:       # BB#0:
@@ -318,9 +318,9 @@ define <4 x i64> @test_pxor(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
 ; HASWELL-LABEL: test_pxor:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpxor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; HASWELL-NEXT:    vpxor (%rdi), %ymm0, %ymm0 # sched: [5:0.50]
+; HASWELL-NEXT:    vpxor (%rdi), %ymm0, %ymm0 # sched: [1:0.50]
 ; HASWELL-NEXT:    vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; ZNVER1-LABEL: test_pxor:
 ; ZNVER1:       # BB#0:

llvm/test/CodeGen/X86/avx2-vector-shifts.ll

@@ -381,6 +381,7 @@ define <4 x i32> @srl_trunc_and_v4i64(<4 x i32> %x, <4 x i64> %y) nounwind {
 ; X32-NEXT:    vpsrlvd %xmm1, %xmm0, %xmm0
 ; X32-NEXT:    vzeroupper
 ; X32-NEXT:    retl
+; X32-NEXT:    ## -- End function
 ;
 ; X64-LABEL: srl_trunc_and_v4i64:
 ; X64:       ## BB#0:
@@ -391,6 +392,7 @@ define <4 x i32> @srl_trunc_and_v4i64(<4 x i32> %x, <4 x i64> %y) nounwind {
 ; X64-NEXT:    vpsrlvd %xmm1, %xmm0, %xmm0
 ; X64-NEXT:    vzeroupper
 ; X64-NEXT:    retq
+; X64-NEXT:    ## -- End function
   %and = and <4 x i64> %y, <i64 8, i64 8, i64 8, i64 8>
   %trunc = trunc <4 x i64> %and to <4 x i32>
   %sra = lshr <4 x i32> %x, %trunc
@@ -412,6 +414,7 @@ define <8 x i16> @shl_8i16(<8 x i16> %r, <8 x i16> %a) nounwind {
 ; X32-NEXT:    ## kill: %XMM0<def> %XMM0<kill> %YMM0<kill>
 ; X32-NEXT:    vzeroupper
 ; X32-NEXT:    retl
+; X32-NEXT:    ## -- End function
 ;
 ; X64-LABEL: shl_8i16:
 ; X64:       ## BB#0:
@@ -423,6 +426,7 @@ define <8 x i16> @shl_8i16(<8 x i16> %r, <8 x i16> %a) nounwind {
 ; X64-NEXT:    ## kill: %XMM0<def> %XMM0<kill> %YMM0<kill>
 ; X64-NEXT:    vzeroupper
 ; X64-NEXT:    retq
+; X64-NEXT:    ## -- End function
   %shl = shl <8 x i16> %r, %a
   ret <8 x i16> %shl
 }
@@ -434,13 +438,14 @@ define <16 x i16> @shl_16i16(<16 x i16> %r, <16 x i16> %a) nounwind {
 ; X32-NEXT:    vpunpckhwd {{.*#+}} ymm3 = ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15]
 ; X32-NEXT:    vpunpckhwd {{.*#+}} ymm4 = ymm2[4],ymm0[4],ymm2[5],ymm0[5],ymm2[6],ymm0[6],ymm2[7],ymm0[7],ymm2[12],ymm0[12],ymm2[13],ymm0[13],ymm2[14],ymm0[14],ymm2[15],ymm0[15]
 ; X32-NEXT:    vpsllvd %ymm3, %ymm4, %ymm3
-; X32-NEXT:    vpsrld $16, %ymm3, %ymm3
 ; X32-NEXT:    vpunpcklwd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[8],ymm2[8],ymm1[9],ymm2[9],ymm1[10],ymm2[10],ymm1[11],ymm2[11]
 ; X32-NEXT:    vpunpcklwd {{.*#+}} ymm0 = ymm2[0],ymm0[0],ymm2[1],ymm0[1],ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[8],ymm0[8],ymm2[9],ymm0[9],ymm2[10],ymm0[10],ymm2[11],ymm0[11]
 ; X32-NEXT:    vpsllvd %ymm1, %ymm0, %ymm0
+; X32-NEXT:    vpsrld $16, %ymm3, %ymm1
 ; X32-NEXT:    vpsrld $16, %ymm0, %ymm0
-; X32-NEXT:    vpackusdw %ymm3, %ymm0, %ymm0
+; X32-NEXT:    vpackusdw %ymm1, %ymm0, %ymm0
 ; X32-NEXT:    retl
+; X32-NEXT:    ## -- End function
 ;
 ; X64-LABEL: shl_16i16:
 ; X64:       ## BB#0:
@@ -448,13 +453,14 @@ define <16 x i16> @shl_16i16(<16 x i16> %r, <16 x i16> %a) nounwind {
 ; X64-NEXT:    vpunpckhwd {{.*#+}} ymm3 = ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15]
 ; X64-NEXT:    vpunpckhwd {{.*#+}} ymm4 = ymm2[4],ymm0[4],ymm2[5],ymm0[5],ymm2[6],ymm0[6],ymm2[7],ymm0[7],ymm2[12],ymm0[12],ymm2[13],ymm0[13],ymm2[14],ymm0[14],ymm2[15],ymm0[15]
 ; X64-NEXT:    vpsllvd %ymm3, %ymm4, %ymm3
-; X64-NEXT:    vpsrld $16, %ymm3, %ymm3
 ; X64-NEXT:    vpunpcklwd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[8],ymm2[8],ymm1[9],ymm2[9],ymm1[10],ymm2[10],ymm1[11],ymm2[11]
 ; X64-NEXT:    vpunpcklwd {{.*#+}} ymm0 = ymm2[0],ymm0[0],ymm2[1],ymm0[1],ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[8],ymm0[8],ymm2[9],ymm0[9],ymm2[10],ymm0[10],ymm2[11],ymm0[11]
 ; X64-NEXT:    vpsllvd %ymm1, %ymm0, %ymm0
+; X64-NEXT:    vpsrld $16, %ymm3, %ymm1
 ; X64-NEXT:    vpsrld $16, %ymm0, %ymm0
-; X64-NEXT:    vpackusdw %ymm3, %ymm0, %ymm0
+; X64-NEXT:    vpackusdw %ymm1, %ymm0, %ymm0
 ; X64-NEXT:    retq
+; X64-NEXT:    ## -- End function
   %shl = shl <16 x i16> %r, %a
   ret <16 x i16> %shl
 }
@@ -474,6 +480,7 @@ define <32 x i8> @shl_32i8(<32 x i8> %r, <32 x i8> %a) nounwind {
 ; X32-NEXT:    vpaddb %ymm1, %ymm1, %ymm1
 ; X32-NEXT:    vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
 ; X32-NEXT:    retl
+; X32-NEXT:    ## -- End function
 ;
 ; X64-LABEL: shl_32i8:
 ; X64:       ## BB#0:
@@ -489,6 +496,7 @@ define <32 x i8> @shl_32i8(<32 x i8> %r, <32 x i8> %a) nounwind {
 ; X64-NEXT:    vpaddb %ymm1, %ymm1, %ymm1
 ; X64-NEXT:    vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
 ; X64-NEXT:    retq
+; X64-NEXT:    ## -- End function
   %shl = shl <32 x i8> %r, %a
   ret <32 x i8> %shl
 }
@@ -504,6 +512,7 @@ define <8 x i16> @ashr_8i16(<8 x i16> %r, <8 x i16> %a) nounwind {
 ; X32-NEXT:    ## kill: %XMM0<def> %XMM0<kill> %YMM0<kill>
 ; X32-NEXT:    vzeroupper
 ; X32-NEXT:    retl
+; X32-NEXT:    ## -- End function
 ;
 ; X64-LABEL: ashr_8i16:
 ; X64:       ## BB#0:
@@ -515,6 +524,7 @@ define <8 x i16> @ashr_8i16(<8 x i16> %r, <8 x i16> %a) nounwind {
 ; X64-NEXT:    ## kill: %XMM0<def> %XMM0<kill> %YMM0<kill>
 ; X64-NEXT:    vzeroupper
 ; X64-NEXT:    retq
+; X64-NEXT:    ## -- End function
   %ashr = ashr <8 x i16> %r, %a
   ret <8 x i16> %ashr
 }
@@ -526,13 +536,14 @@ define <16 x i16> @ashr_16i16(<16 x i16> %r, <16 x i16> %a) nounwind {
 ; X32-NEXT:    vpunpckhwd {{.*#+}} ymm3 = ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15]
 ; X32-NEXT:    vpunpckhwd {{.*#+}} ymm4 = ymm2[4],ymm0[4],ymm2[5],ymm0[5],ymm2[6],ymm0[6],ymm2[7],ymm0[7],ymm2[12],ymm0[12],ymm2[13],ymm0[13],ymm2[14],ymm0[14],ymm2[15],ymm0[15]
 ; X32-NEXT:    vpsravd %ymm3, %ymm4, %ymm3
-; X32-NEXT:    vpsrld $16, %ymm3, %ymm3
 ; X32-NEXT:    vpunpcklwd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[8],ymm2[8],ymm1[9],ymm2[9],ymm1[10],ymm2[10],ymm1[11],ymm2[11]
 ; X32-NEXT:    vpunpcklwd {{.*#+}} ymm0 = ymm2[0],ymm0[0],ymm2[1],ymm0[1],ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[8],ymm0[8],ymm2[9],ymm0[9],ymm2[10],ymm0[10],ymm2[11],ymm0[11]
 ; X32-NEXT:    vpsravd %ymm1, %ymm0, %ymm0
+; X32-NEXT:    vpsrld $16, %ymm3, %ymm1
 ; X32-NEXT:    vpsrld $16, %ymm0, %ymm0
-; X32-NEXT:    vpackusdw %ymm3, %ymm0, %ymm0
+; X32-NEXT:    vpackusdw %ymm1, %ymm0, %ymm0
 ; X32-NEXT:    retl
+; X32-NEXT:    ## -- End function
 ;
 ; X64-LABEL: ashr_16i16:
 ; X64:       ## BB#0:
@@ -540,13 +551,14 @@ define <16 x i16> @ashr_16i16(<16 x i16> %r, <16 x i16> %a) nounwind {
 ; X64-NEXT:    vpunpckhwd {{.*#+}} ymm3 = ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15]
 ; X64-NEXT:    vpunpckhwd {{.*#+}} ymm4 = ymm2[4],ymm0[4],ymm2[5],ymm0[5],ymm2[6],ymm0[6],ymm2[7],ymm0[7],ymm2[12],ymm0[12],ymm2[13],ymm0[13],ymm2[14],ymm0[14],ymm2[15],ymm0[15]
 ; X64-NEXT:    vpsravd %ymm3, %ymm4, %ymm3
-; X64-NEXT:    vpsrld $16, %ymm3, %ymm3
 ; X64-NEXT:    vpunpcklwd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[8],ymm2[8],ymm1[9],ymm2[9],ymm1[10],ymm2[10],ymm1[11],ymm2[11]
 ; X64-NEXT:    vpunpcklwd {{.*#+}} ymm0 = ymm2[0],ymm0[0],ymm2[1],ymm0[1],ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[8],ymm0[8],ymm2[9],ymm0[9],ymm2[10],ymm0[10],ymm2[11],ymm0[11]
 ; X64-NEXT:    vpsravd %ymm1, %ymm0, %ymm0
+; X64-NEXT:    vpsrld $16, %ymm3, %ymm1
 ; X64-NEXT:    vpsrld $16, %ymm0, %ymm0
-; X64-NEXT:    vpackusdw %ymm3, %ymm0, %ymm0
+; X64-NEXT:    vpackusdw %ymm1, %ymm0, %ymm0
 ; X64-NEXT:    retq
+; X64-NEXT:    ## -- End function
   %ashr = ashr <16 x i16> %r, %a
   ret <16 x i16> %ashr
 }
@@ -579,6 +591,7 @@ define <32 x i8> @ashr_32i8(<32 x i8> %r, <32 x i8> %a) nounwind {
 ; X32-NEXT:    vpsrlw $8, %ymm0, %ymm0
 ; X32-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
 ; X32-NEXT:    retl
+; X32-NEXT:    ## -- End function
 ;
 ; X64-LABEL: ashr_32i8:
 ; X64:       ## BB#0:
@@ -607,6 +620,7 @@ define <32 x i8> @ashr_32i8(<32 x i8> %r, <32 x i8> %a) nounwind {
 ; X64-NEXT:    vpsrlw $8, %ymm0, %ymm0
 ; X64-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
 ; X64-NEXT:    retq
+; X64-NEXT:    ## -- End function
   %ashr = ashr <32 x i8> %r, %a
   ret <32 x i8> %ashr
 }
@@ -622,6 +636,7 @@ define <8 x i16> @lshr_8i16(<8 x i16> %r, <8 x i16> %a) nounwind {
 ; X32-NEXT:    ## kill: %XMM0<def> %XMM0<kill> %YMM0<kill>
 ; X32-NEXT:    vzeroupper
 ; X32-NEXT:    retl
+; X32-NEXT:    ## -- End function
 ;
 ; X64-LABEL: lshr_8i16:
 ; X64:       ## BB#0:
@@ -633,6 +648,7 @@ define <8 x i16> @lshr_8i16(<8 x i16> %r, <8 x i16> %a) nounwind {
 ; X64-NEXT:    ## kill: %XMM0<def> %XMM0<kill> %YMM0<kill>
 ; X64-NEXT:    vzeroupper
 ; X64-NEXT:    retq
+; X64-NEXT:    ## -- End function
   %lshr = lshr <8 x i16> %r, %a
   ret <8 x i16> %lshr
 }
@@ -644,13 +660,14 @@ define <16 x i16> @lshr_16i16(<16 x i16> %r, <16 x i16> %a) nounwind {
 ; X32-NEXT:    vpunpckhwd {{.*#+}} ymm3 = ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15]
 ; X32-NEXT:    vpunpckhwd {{.*#+}} ymm4 = ymm2[4],ymm0[4],ymm2[5],ymm0[5],ymm2[6],ymm0[6],ymm2[7],ymm0[7],ymm2[12],ymm0[12],ymm2[13],ymm0[13],ymm2[14],ymm0[14],ymm2[15],ymm0[15]
 ; X32-NEXT:    vpsrlvd %ymm3, %ymm4, %ymm3
-; X32-NEXT:    vpsrld $16, %ymm3, %ymm3
 ; X32-NEXT:    vpunpcklwd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[8],ymm2[8],ymm1[9],ymm2[9],ymm1[10],ymm2[10],ymm1[11],ymm2[11]
 ; X32-NEXT:    vpunpcklwd {{.*#+}} ymm0 = ymm2[0],ymm0[0],ymm2[1],ymm0[1],ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[8],ymm0[8],ymm2[9],ymm0[9],ymm2[10],ymm0[10],ymm2[11],ymm0[11]
 ; X32-NEXT:    vpsrlvd %ymm1, %ymm0, %ymm0
+; X32-NEXT:    vpsrld $16, %ymm3, %ymm1
 ; X32-NEXT:    vpsrld $16, %ymm0, %ymm0
-; X32-NEXT:    vpackusdw %ymm3, %ymm0, %ymm0
+; X32-NEXT:    vpackusdw %ymm1, %ymm0, %ymm0
 ; X32-NEXT:    retl
+; X32-NEXT:    ## -- End function
 ;
 ; X64-LABEL: lshr_16i16:
 ; X64:       ## BB#0:
@@ -658,13 +675,14 @@ define <16 x i16> @lshr_16i16(<16 x i16> %r, <16 x i16> %a) nounwind {
 ; X64-NEXT:    vpunpckhwd {{.*#+}} ymm3 = ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15]
 ; X64-NEXT:    vpunpckhwd {{.*#+}} ymm4 = ymm2[4],ymm0[4],ymm2[5],ymm0[5],ymm2[6],ymm0[6],ymm2[7],ymm0[7],ymm2[12],ymm0[12],ymm2[13],ymm0[13],ymm2[14],ymm0[14],ymm2[15],ymm0[15]
 ; X64-NEXT:    vpsrlvd %ymm3, %ymm4, %ymm3
-; X64-NEXT:    vpsrld $16, %ymm3, %ymm3
 ; X64-NEXT:    vpunpcklwd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[8],ymm2[8],ymm1[9],ymm2[9],ymm1[10],ymm2[10],ymm1[11],ymm2[11]
 ; X64-NEXT:    vpunpcklwd {{.*#+}} ymm0 = ymm2[0],ymm0[0],ymm2[1],ymm0[1],ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[8],ymm0[8],ymm2[9],ymm0[9],ymm2[10],ymm0[10],ymm2[11],ymm0[11]
 ; X64-NEXT:    vpsrlvd %ymm1, %ymm0, %ymm0
+; X64-NEXT:    vpsrld $16, %ymm3, %ymm1
 ; X64-NEXT:    vpsrld $16, %ymm0, %ymm0
-; X64-NEXT:    vpackusdw %ymm3, %ymm0, %ymm0
+; X64-NEXT:    vpackusdw %ymm1, %ymm0, %ymm0
 ; X64-NEXT:    retq
+; X64-NEXT:    ## -- End function
   %lshr = lshr <16 x i16> %r, %a
   ret <16 x i16> %lshr
 }
@@ -685,6 +703,7 @@ define <32 x i8> @lshr_32i8(<32 x i8> %r, <32 x i8> %a) nounwind {
 ; X32-NEXT:    vpaddb %ymm1, %ymm1, %ymm1
 ; X32-NEXT:    vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
 ; X32-NEXT:    retl
+; X32-NEXT:    ## -- End function
 ;
 ; X64-LABEL: lshr_32i8:
 ; X64:       ## BB#0:
@@ -701,6 +720,7 @@ define <32 x i8> @lshr_32i8(<32 x i8> %r, <32 x i8> %a) nounwind {
 ; X64-NEXT:    vpaddb %ymm1, %ymm1, %ymm1
 ; X64-NEXT:    vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
 ; X64-NEXT:    retq
+; X64-NEXT:    ## -- End function
   %lshr = lshr <32 x i8> %r, %a
   ret <32 x i8> %lshr
 }

llvm/test/CodeGen/X86/avx512-cmp.ll

@@ -14,6 +14,7 @@ define double @test1(double %a, double %b) nounwind {
 ; ALL-NEXT:  LBB0_2: ## %l2
 ; ALL-NEXT:    vaddsd %xmm1, %xmm0, %xmm0
 ; ALL-NEXT:    retq
+; ALL-NEXT:    ## -- End function
   %tobool = fcmp une double %a, %b
   br i1 %tobool, label %l1, label %l2
 
@@ -36,6 +37,7 @@ define float @test2(float %a, float %b) nounwind {
 ; ALL-NEXT:  LBB1_2: ## %l2
 ; ALL-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; ALL-NEXT:    retq
+; ALL-NEXT:    ## -- End function
   %tobool = fcmp olt float %a, %b
   br i1 %tobool, label %l1, label %l2
 
@@ -124,11 +126,11 @@ entry:
 define i32 @test8(i32 %a1, i32 %a2, i32 %a3) {
 ; ALL-LABEL: test8:
 ; ALL:       ## BB#0:
-; ALL-NEXT:    notl %edi
 ; ALL-NEXT:    xorl $-2147483648, %esi ## imm = 0x80000000
 ; ALL-NEXT:    testl %edx, %edx
 ; ALL-NEXT:    movl $1, %eax
 ; ALL-NEXT:    cmovel %eax, %edx
+; ALL-NEXT:    notl %edi
 ; ALL-NEXT:    orl %edi, %esi
 ; ALL-NEXT:    cmovnel %edx, %eax
 ; ALL-NEXT:    retq

llvm/test/CodeGen/X86/avx512-cvt.ll

@@ -1545,19 +1545,19 @@ define <4 x double> @uitofp_4i1_double(<4 x i32> %a) {
 }
 
 define <2 x float> @uitofp_2i1_float(<2 x i32> %a) {
-; NOVL-LABEL: uitofp_2i1_float:
+; KNL-LABEL: uitofp_2i1_float:
-; NOVL:       # BB#0:
+; KNL:       # BB#0:
-; NOVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; KNL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; NOVL-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; KNL-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
-; NOVL-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; KNL-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
-; NOVL-NEXT:    vpextrb $8, %xmm0, %eax
+; KNL-NEXT:    vpextrb $8, %xmm0, %eax
-; NOVL-NEXT:    andl $1, %eax
+; KNL-NEXT:    andl $1, %eax
-; NOVL-NEXT:    vcvtsi2ssl %eax, %xmm2, %xmm1
+; KNL-NEXT:    vpextrb $0, %xmm0, %ecx
-; NOVL-NEXT:    vpextrb $0, %xmm0, %eax
+; KNL-NEXT:    vcvtsi2ssl %eax, %xmm2, %xmm0
-; NOVL-NEXT:    andl $1, %eax
+; KNL-NEXT:    andl $1, %ecx
-; NOVL-NEXT:    vcvtsi2ssl %eax, %xmm2, %xmm0
+; KNL-NEXT:    vcvtsi2ssl %ecx, %xmm2, %xmm1
-; NOVL-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[2,3]
+; KNL-NEXT:    vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3]
-; NOVL-NEXT:    retq
+; KNL-NEXT:    retq
 ;
 ; VL-LABEL: uitofp_2i1_float:
 ; VL:       # BB#0:
@@ -1567,6 +1567,34 @@ define <2 x float> @uitofp_2i1_float(<2 x i32> %a) {
 ; VL-NEXT:    vpbroadcastd {{.*}}(%rip), %xmm0 {%k1} {z}
 ; VL-NEXT:    vcvtudq2ps %xmm0, %xmm0
 ; VL-NEXT:    retq
+;
+; AVX512DQ-LABEL: uitofp_2i1_float:
+; AVX512DQ:       # BB#0:
+; AVX512DQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; AVX512DQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512DQ-NEXT:    vpextrb $8, %xmm0, %eax
+; AVX512DQ-NEXT:    andl $1, %eax
+; AVX512DQ-NEXT:    vcvtsi2ssl %eax, %xmm2, %xmm1
+; AVX512DQ-NEXT:    vpextrb $0, %xmm0, %eax
+; AVX512DQ-NEXT:    andl $1, %eax
+; AVX512DQ-NEXT:    vcvtsi2ssl %eax, %xmm2, %xmm0
+; AVX512DQ-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[2,3]
+; AVX512DQ-NEXT:    retq
+;
+; AVX512BW-LABEL: uitofp_2i1_float:
+; AVX512BW:       # BB#0:
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; AVX512BW-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512BW-NEXT:    vpextrb $8, %xmm0, %eax
+; AVX512BW-NEXT:    andl $1, %eax
+; AVX512BW-NEXT:    vcvtsi2ssl %eax, %xmm2, %xmm1
+; AVX512BW-NEXT:    vpextrb $0, %xmm0, %eax
+; AVX512BW-NEXT:    andl $1, %eax
+; AVX512BW-NEXT:    vcvtsi2ssl %eax, %xmm2, %xmm0
+; AVX512BW-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[2,3]
+; AVX512BW-NEXT:    retq
   %mask = icmp ult <2 x i32> %a, zeroinitializer
   %1 = uitofp <2 x i1> %mask to <2 x float>
   ret <2 x float> %1

llvm/test/CodeGen/X86/avx512-insert-extract.ll

@@ -12,6 +12,7 @@ define <16 x float> @test1(<16 x float> %x, float* %br, float %y) nounwind {
 ; KNL-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0],xmm0[3]
 ; KNL-NEXT:    vinsertf32x4 $3, %xmm0, %zmm2, %zmm0
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test1:
 ; SKX:       ## BB#0:
@@ -21,6 +22,7 @@ define <16 x float> @test1(<16 x float> %x, float* %br, float %y) nounwind {
 ; SKX-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0],xmm0[3]
 ; SKX-NEXT:    vinsertf32x4 $3, %xmm0, %zmm2, %zmm0
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
   %rrr = load float, float* %br
   %rrr2 = insertelement <16 x float> %x, float %rrr, i32 1
   %rrr3 = insertelement <16 x float> %rrr2, float %y, i32 14
@@ -36,6 +38,7 @@ define <8 x double> @test2(<8 x double> %x, double* %br, double %y) nounwind {
 ; KNL-NEXT:    vmovsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
 ; KNL-NEXT:    vinsertf32x4 $3, %xmm0, %zmm2, %zmm0
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test2:
 ; SKX:       ## BB#0:
@@ -45,6 +48,7 @@ define <8 x double> @test2(<8 x double> %x, double* %br, double %y) nounwind {
 ; SKX-NEXT:    vmovsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
 ; SKX-NEXT:    vinsertf64x2 $3, %xmm0, %zmm2, %zmm0
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
   %rrr = load double, double* %br
   %rrr2 = insertelement <8 x double> %x, double %rrr, i32 1
   %rrr3 = insertelement <8 x double> %rrr2, double %y, i32 6
@@ -58,6 +62,7 @@ define <16 x float> @test3(<16 x float> %x) nounwind {
 ; KNL-NEXT:    vinsertps {{.*#+}} xmm1 = xmm0[0],xmm1[0],xmm0[2,3]
 ; KNL-NEXT:    vinsertf32x4 $0, %xmm1, %zmm0, %zmm0
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test3:
 ; SKX:       ## BB#0:
@@ -65,6 +70,7 @@ define <16 x float> @test3(<16 x float> %x) nounwind {
 ; SKX-NEXT:    vinsertps {{.*#+}} xmm1 = xmm0[0],xmm1[0],xmm0[2,3]
 ; SKX-NEXT:    vinsertf32x4 $0, %xmm1, %zmm0, %zmm0
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
   %eee = extractelement <16 x float> %x, i32 4
   %rrr2 = insertelement <16 x float> %x, float %eee, i32 1
   ret <16 x float> %rrr2
@@ -78,6 +84,7 @@ define <8 x i64> @test4(<8 x i64> %x) nounwind {
 ; KNL-NEXT:    vpinsrq $1, %rax, %xmm0, %xmm1
 ; KNL-NEXT:    vinserti32x4 $0, %xmm1, %zmm0, %zmm0
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test4:
 ; SKX:       ## BB#0:
@@ -86,6 +93,7 @@ define <8 x i64> @test4(<8 x i64> %x) nounwind {
 ; SKX-NEXT:    vpinsrq $1, %rax, %xmm0, %xmm1
 ; SKX-NEXT:    vinserti64x2 $0, %xmm1, %zmm0, %zmm0
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
   %eee = extractelement <8 x i64> %x, i32 4
   %rrr2 = insertelement <8 x i64> %x, i64 %eee, i32 1
   ret <8 x i64> %rrr2
@@ -96,11 +104,13 @@ define i32 @test5(<4 x float> %x) nounwind {
 ; KNL:       ## BB#0:
 ; KNL-NEXT:    vextractps $3, %xmm0, %eax
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test5:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vextractps $3, %xmm0, %eax
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
   %ef = extractelement <4 x float> %x, i32 3
   %ei = bitcast float %ef to i32
   ret i32 %ei
@@ -111,11 +121,13 @@ define void @test6(<4 x float> %x, float* %out) nounwind {
 ; KNL:       ## BB#0:
 ; KNL-NEXT:    vextractps $3, %xmm0, (%rdi)
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test6:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vextractps $3, %xmm0, (%rdi)
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
   %ef = extractelement <4 x float> %x, i32 3
   store float %ef, float* %out, align 4
   ret void
@@ -135,6 +147,7 @@ define float @test7(<16 x float> %x, i32 %ind) nounwind {
 ; KNL-NEXT:    movq %rbp, %rsp
 ; KNL-NEXT:    popq %rbp
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test7:
 ; SKX:       ## BB#0:
@@ -150,6 +163,7 @@ define float @test7(<16 x float> %x, i32 %ind) nounwind {
 ; SKX-NEXT:    popq %rbp
 ; SKX-NEXT:    vzeroupper
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
   %e = extractelement <16 x float> %x, i32 %ind
   ret float %e
 }
@@ -168,6 +182,7 @@ define double @test8(<8 x double> %x, i32 %ind) nounwind {
 ; KNL-NEXT:    movq %rbp, %rsp
 ; KNL-NEXT:    popq %rbp
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test8:
 ; SKX:       ## BB#0:
@@ -183,6 +198,7 @@ define double @test8(<8 x double> %x, i32 %ind) nounwind {
 ; SKX-NEXT:    popq %rbp
 ; SKX-NEXT:    vzeroupper
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
   %e = extractelement <8 x double> %x, i32 %ind
   ret double %e
 }
@@ -201,6 +217,7 @@ define float @test9(<8 x float> %x, i32 %ind) nounwind {
 ; KNL-NEXT:    movq %rbp, %rsp
 ; KNL-NEXT:    popq %rbp
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test9:
 ; SKX:       ## BB#0:
@@ -216,6 +233,7 @@ define float @test9(<8 x float> %x, i32 %ind) nounwind {
 ; SKX-NEXT:    popq %rbp
 ; SKX-NEXT:    vzeroupper
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
   %e = extractelement <8 x float> %x, i32 %ind
   ret float %e
 }
@@ -234,6 +252,7 @@ define i32 @test10(<16 x i32> %x, i32 %ind) nounwind {
 ; KNL-NEXT:    movq %rbp, %rsp
 ; KNL-NEXT:    popq %rbp
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test10:
 ; SKX:       ## BB#0:
@@ -249,6 +268,7 @@ define i32 @test10(<16 x i32> %x, i32 %ind) nounwind {
 ; SKX-NEXT:    popq %rbp
 ; SKX-NEXT:    vzeroupper
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
   %e = extractelement <16 x i32> %x, i32 %ind
   ret i32 %e
 }
@@ -1114,137 +1134,137 @@ define i32 @test_insertelement_v32i1(i32 %a, i32 %b, <32 x i32> %x , <32 x i32>
 ; KNL-NEXT:    .cfi_def_cfa_register %rbp
 ; KNL-NEXT:    andq $-32, %rsp
 ; KNL-NEXT:    subq $32, %rsp
-; KNL-NEXT:    xorl %eax, %eax
-; KNL-NEXT:    cmpl %esi, %edi
-; KNL-NEXT:    setb %al
 ; KNL-NEXT:    vpcmpltud %zmm3, %zmm1, %k0
 ; KNL-NEXT:    kshiftlw $14, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
 ; KNL-NEXT:    kshiftlw $15, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %edx
+; KNL-NEXT:    kmovw %k1, %ecx
-; KNL-NEXT:    vmovd %edx, %xmm1
+; KNL-NEXT:    vmovd %ecx, %xmm1
-; KNL-NEXT:    vpinsrb $1, %ecx, %xmm1, %xmm1
+; KNL-NEXT:    vpinsrb $1, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $13, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $2, %ecx, %xmm1, %xmm1
+; KNL-NEXT:    vpinsrb $2, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $12, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $3, %ecx, %xmm1, %xmm1
+; KNL-NEXT:    vpinsrb $3, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $11, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $4, %ecx, %xmm1, %xmm1
+; KNL-NEXT:    vpinsrb $4, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $10, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $5, %ecx, %xmm1, %xmm1
+; KNL-NEXT:    vpinsrb $5, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $9, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $6, %ecx, %xmm1, %xmm1
+; KNL-NEXT:    vpinsrb $6, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $8, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $7, %ecx, %xmm1, %xmm1
+; KNL-NEXT:    vpinsrb $7, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $7, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $8, %ecx, %xmm1, %xmm1
+; KNL-NEXT:    vpinsrb $8, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $6, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $9, %ecx, %xmm1, %xmm1
+; KNL-NEXT:    vpinsrb $9, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $5, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $10, %ecx, %xmm1, %xmm1
+; KNL-NEXT:    vpinsrb $10, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $4, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $11, %ecx, %xmm1, %xmm1
+; KNL-NEXT:    vpinsrb $11, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $3, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $12, %ecx, %xmm1, %xmm1
+; KNL-NEXT:    vpinsrb $12, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $2, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $13, %ecx, %xmm1, %xmm1
+; KNL-NEXT:    vpinsrb $13, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $1, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $14, %ecx, %xmm1, %xmm1
+; KNL-NEXT:    vpinsrb $14, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %ecx
+; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    vpinsrb $15, %ecx, %xmm1, %xmm1
+; KNL-NEXT:    vpinsrb $15, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    vpcmpltud %zmm2, %zmm0, %k0
+; KNL-NEXT:    kshiftlw $15, %k0, %k1
+; KNL-NEXT:    kshiftrw $15, %k1, %k1
+; KNL-NEXT:    kmovw %k1, %eax
 ; KNL-NEXT:    kshiftlw $14, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    vmovd %eax, %xmm0
-; KNL-NEXT:    kshiftlw $15, %k0, %k1
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
+; KNL-NEXT:    vpinsrb $1, %eax, %xmm0, %xmm0
-; KNL-NEXT:    kmovw %k1, %edx
-; KNL-NEXT:    vmovd %edx, %xmm0
-; KNL-NEXT:    vpinsrb $1, %ecx, %xmm0, %xmm0
 ; KNL-NEXT:    kshiftlw $13, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $2, %ecx, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0
 ; KNL-NEXT:    kshiftlw $12, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $3, %ecx, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $3, %eax, %xmm0, %xmm0
 ; KNL-NEXT:    kshiftlw $11, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $4, %ecx, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0
 ; KNL-NEXT:    kshiftlw $10, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $5, %ecx, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $5, %eax, %xmm0, %xmm0
 ; KNL-NEXT:    kshiftlw $9, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $6, %ecx, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0
 ; KNL-NEXT:    kshiftlw $8, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $7, %ecx, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0
 ; KNL-NEXT:    kshiftlw $7, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $8, %ecx, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0
 ; KNL-NEXT:    kshiftlw $6, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $9, %ecx, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $9, %eax, %xmm0, %xmm0
 ; KNL-NEXT:    kshiftlw $5, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $10, %ecx, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0
 ; KNL-NEXT:    kshiftlw $4, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $11, %ecx, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $11, %eax, %xmm0, %xmm0
 ; KNL-NEXT:    kshiftlw $3, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $12, %ecx, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0
 ; KNL-NEXT:    kshiftlw $2, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $13, %ecx, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $13, %eax, %xmm0, %xmm0
 ; KNL-NEXT:    kshiftlw $1, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $14, %ecx, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0
+; KNL-NEXT:    xorl %eax, %eax
+; KNL-NEXT:    cmpl %esi, %edi
 ; KNL-NEXT:    kshiftrw $15, %k0, %k0
 ; KNL-NEXT:    kmovw %k0, %ecx
 ; KNL-NEXT:    vpinsrb $15, %ecx, %xmm0, %xmm0
+; KNL-NEXT:    setb %al
 ; KNL-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; KNL-NEXT:    vpsllw $7, %ymm0, %ymm0
 ; KNL-NEXT:    vpand {{.*}}(%rip), %ymm0, %ymm0
@@ -1299,8 +1319,8 @@ define i8 @test_iinsertelement_v4i1(i32 %a, i32 %b, <4 x i32> %x , <4 x i32> %y)
 ; KNL-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
 ; KNL-NEXT:    vpextrb $4, %xmm0, %ecx
 ; KNL-NEXT:    kmovw %ecx, %k1
-; KNL-NEXT:    vpternlogq $255, %zmm1, %zmm1, %zmm1 {%k1} {z}
 ; KNL-NEXT:    vpextrb $0, %xmm0, %ecx
+; KNL-NEXT:    vpternlogq $255, %zmm1, %zmm1, %zmm1 {%k1} {z}
 ; KNL-NEXT:    kmovw %ecx, %k1
 ; KNL-NEXT:    vpternlogq $255, %zmm2, %zmm2, %zmm2 {%k1} {z}
 ; KNL-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,8,2,3,4,5,6,7]
@@ -2124,8 +2144,8 @@ define i16 @test_extractelement_variable_v32i16(<32 x i16> %t1, i32 %index) {
 define i8 @test_extractelement_variable_v16i8(<16 x i8> %t1, i32 %index) {
 ; KNL-LABEL: test_extractelement_variable_v16i8:
 ; KNL:       ## BB#0:
-; KNL-NEXT:    ## kill: %EDI<def> %EDI<kill> %RDI<def>
 ; KNL-NEXT:    vmovaps %xmm0, -{{[0-9]+}}(%rsp)
+; KNL-NEXT:    ## kill: %EDI<def> %EDI<kill> %RDI<def>
 ; KNL-NEXT:    andl $15, %edi
 ; KNL-NEXT:    leaq -{{[0-9]+}}(%rsp), %rax
 ; KNL-NEXT:    movb (%rdi,%rax), %al
@@ -2156,8 +2176,8 @@ define i8 @test_extractelement_variable_v32i8(<32 x i8> %t1, i32 %index) {
 ; KNL-NEXT:    .cfi_def_cfa_register %rbp
 ; KNL-NEXT:    andq $-32, %rsp
 ; KNL-NEXT:    subq $64, %rsp
-; KNL-NEXT:    ## kill: %EDI<def> %EDI<kill> %RDI<def>
 ; KNL-NEXT:    vmovaps %ymm0, (%rsp)
+; KNL-NEXT:    ## kill: %EDI<def> %EDI<kill> %RDI<def>
 ; KNL-NEXT:    andl $31, %edi
 ; KNL-NEXT:    movq %rsp, %rax
 ; KNL-NEXT:    movb (%rdi,%rax), %al
@@ -2204,9 +2224,9 @@ define i8 @test_extractelement_variable_v64i8(<64 x i8> %t1, i32 %index) {
 ; KNL-NEXT:    .cfi_def_cfa_register %rbp
 ; KNL-NEXT:    andq $-64, %rsp
 ; KNL-NEXT:    subq $128, %rsp
-; KNL-NEXT:    ## kill: %EDI<def> %EDI<kill> %RDI<def>
 ; KNL-NEXT:    vmovaps %ymm1, {{[0-9]+}}(%rsp)
 ; KNL-NEXT:    vmovaps %ymm0, (%rsp)
+; KNL-NEXT:    ## kill: %EDI<def> %EDI<kill> %RDI<def>
 ; KNL-NEXT:    andl $63, %edi
 ; KNL-NEXT:    movq %rsp, %rax
 ; KNL-NEXT:    movb (%rdi,%rax), %al
@@ -2295,12 +2315,12 @@ define i8 @test_extractelement_variable_v64i8_indexi8(<64 x i8> %t1, i8 %index)
 define zeroext i8 @test_extractelement_varible_v2i1(<2 x i64> %a, <2 x i64> %b, i32 %index) {
 ; KNL-LABEL: test_extractelement_varible_v2i1:
 ; KNL:       ## BB#0:
-; KNL-NEXT:    ## kill: %EDI<def> %EDI<kill> %RDI<def>
 ; KNL-NEXT:    vmovdqa {{.*#+}} xmm2 = [9223372036854775808,9223372036854775808]
 ; KNL-NEXT:    vpxor %xmm2, %xmm1, %xmm1
 ; KNL-NEXT:    vpxor %xmm2, %xmm0, %xmm0
 ; KNL-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
 ; KNL-NEXT:    vmovdqa %xmm0, -{{[0-9]+}}(%rsp)
+; KNL-NEXT:    ## kill: %EDI<def> %EDI<kill> %RDI<def>
 ; KNL-NEXT:    andl $1, %edi
 ; KNL-NEXT:    movl -24(%rsp,%rdi,8), %eax
 ; KNL-NEXT:    andl $1, %eax
@@ -2325,12 +2345,12 @@ define zeroext i8 @test_extractelement_varible_v2i1(<2 x i64> %a, <2 x i64> %b,
 define zeroext i8 @test_extractelement_varible_v4i1(<4 x i32> %a, <4 x i32> %b, i32 %index) {
 ; KNL-LABEL: test_extractelement_varible_v4i1:
 ; KNL:       ## BB#0:
-; KNL-NEXT:    ## kill: %EDI<def> %EDI<kill> %RDI<def>
 ; KNL-NEXT:    vpbroadcastd {{.*}}(%rip), %xmm2
 ; KNL-NEXT:    vpxor %xmm2, %xmm1, %xmm1
 ; KNL-NEXT:    vpxor %xmm2, %xmm0, %xmm0
 ; KNL-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
 ; KNL-NEXT:    vmovdqa %xmm0, -{{[0-9]+}}(%rsp)
+; KNL-NEXT:    ## kill: %EDI<def> %EDI<kill> %RDI<def>
 ; KNL-NEXT:    andl $3, %edi
 ; KNL-NEXT:    movl -24(%rsp,%rdi,4), %eax
 ; KNL-NEXT:    andl $1, %eax

llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll

@@ -2880,7 +2880,6 @@ declare <8 x i64> @llvm.x86.avx512.mask.pmulu.dq.512(<16 x i32>, <16 x i32>, <8
 define <4 x float> @test_mask_vextractf32x4(<4 x float> %b, <16 x float> %a, i8 %mask) {
 ; CHECK-LABEL: test_mask_vextractf32x4:
 ; CHECK:       ## BB#0:
-; CHECK-NEXT:    vextractf32x4 $2, %zmm1, %xmm1
 ; CHECK-NEXT:    kmovw %edi, %k0
 ; CHECK-NEXT:    kshiftlw $12, %k0, %k1
 ; CHECK-NEXT:    kshiftrw $15, %k1, %k1
@@ -2898,6 +2897,7 @@ define <4 x float> @test_mask_vextractf32x4(<4 x float> %b, <16 x float> %a, i8
 ; CHECK-NEXT:    vpinsrb $8, %eax, %xmm2, %xmm2
 ; CHECK-NEXT:    kmovw %k1, %eax
 ; CHECK-NEXT:    vpinsrb $12, %eax, %xmm2, %xmm2
+; CHECK-NEXT:    vextractf32x4 $2, %zmm1, %xmm1
 ; CHECK-NEXT:    vpslld $31, %xmm2, %xmm2
 ; CHECK-NEXT:    vblendvps %xmm2, %xmm1, %xmm0, %xmm0
 ; CHECK-NEXT:    retq
@@ -2941,7 +2941,6 @@ declare <4 x i64> @llvm.x86.avx512.mask.vextracti64x4.512(<8 x i64>, i32, <4 x i
 define <4 x i32> @test_maskz_vextracti32x4(<16 x i32> %a, i8 %mask) {
 ; CHECK-LABEL: test_maskz_vextracti32x4:
 ; CHECK:       ## BB#0:
-; CHECK-NEXT:    vextracti32x4 $2, %zmm0, %xmm0
 ; CHECK-NEXT:    kmovw %edi, %k0
 ; CHECK-NEXT:    kshiftlw $12, %k0, %k1
 ; CHECK-NEXT:    kshiftrw $15, %k1, %k1
@@ -2959,6 +2958,7 @@ define <4 x i32> @test_maskz_vextracti32x4(<16 x i32> %a, i8 %mask) {
 ; CHECK-NEXT:    vpinsrb $8, %eax, %xmm1, %xmm1
 ; CHECK-NEXT:    kmovw %k1, %eax
 ; CHECK-NEXT:    vpinsrb $12, %eax, %xmm1, %xmm1
+; CHECK-NEXT:    vextracti32x4 $2, %zmm0, %xmm0
 ; CHECK-NEXT:    vpslld $31, %xmm1, %xmm1
 ; CHECK-NEXT:    vpsrad $31, %xmm1, %xmm1
 ; CHECK-NEXT:    vpand %xmm0, %xmm1, %xmm0

llvm/test/CodeGen/X86/avx512-mask-op.ll

@@ -1837,73 +1837,8 @@ define void @ktest_2(<32 x float> %in, float * %base) {
 ; KNL-NEXT:    .cfi_def_cfa_register %rbp
 ; KNL-NEXT:    andq $-32, %rsp
 ; KNL-NEXT:    subq $32, %rsp
-; KNL-NEXT:    vmovups (%rdi), %zmm2
+; KNL-NEXT:    vmovups 64(%rdi), %zmm2
-; KNL-NEXT:    vmovups 64(%rdi), %zmm3
+; KNL-NEXT:    vcmpltps %zmm1, %zmm2, %k2
-; KNL-NEXT:    vcmpltps %zmm1, %zmm3, %k1
-; KNL-NEXT:    kshiftlw $14, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    kshiftlw $15, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %ecx
-; KNL-NEXT:    vmovd %ecx, %xmm3
-; KNL-NEXT:    vpinsrb $1, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $13, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    vpinsrb $2, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $12, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    vpinsrb $3, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $11, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    vpinsrb $4, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $10, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    vpinsrb $5, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $9, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    vpinsrb $6, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $8, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    vpinsrb $7, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $7, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    vpinsrb $8, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $6, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    vpinsrb $9, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $5, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    vpinsrb $10, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $4, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    vpinsrb $11, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $3, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    vpinsrb $12, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $2, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    vpinsrb $13, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $1, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    vpinsrb $14, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftrw $15, %k1, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    vpinsrb $15, %eax, %xmm3, %xmm3
-; KNL-NEXT:    vcmpltps %zmm0, %zmm2, %k2
 ; KNL-NEXT:    kshiftlw $14, %k2, %k0
 ; KNL-NEXT:    kshiftrw $15, %k0, %k0
 ; KNL-NEXT:    kmovw %k0, %eax
@@ -1967,10 +1902,139 @@ define void @ktest_2(<32 x float> %in, float * %base) {
 ; KNL-NEXT:    kshiftrw $15, %k2, %k0
 ; KNL-NEXT:    kmovw %k0, %eax
 ; KNL-NEXT:    vpinsrb $15, %eax, %xmm2, %xmm2
-; KNL-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; KNL-NEXT:    vmovups (%rdi), %zmm3
-; KNL-NEXT:    vmovups 4(%rdi), %zmm3 {%k2} {z}
+; KNL-NEXT:    vcmpltps %zmm0, %zmm3, %k1
-; KNL-NEXT:    vmovups 68(%rdi), %zmm4 {%k1} {z}
+; KNL-NEXT:    kshiftlw $14, %k1, %k0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    kshiftlw $15, %k1, %k0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %ecx
+; KNL-NEXT:    vmovd %ecx, %xmm3
+; KNL-NEXT:    vpinsrb $1, %eax, %xmm3, %xmm3
+; KNL-NEXT:    kshiftlw $13, %k1, %k0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    vpinsrb $2, %eax, %xmm3, %xmm3
+; KNL-NEXT:    kshiftlw $12, %k1, %k0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    vpinsrb $3, %eax, %xmm3, %xmm3
+; KNL-NEXT:    kshiftlw $11, %k1, %k0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    vpinsrb $4, %eax, %xmm3, %xmm3
+; KNL-NEXT:    kshiftlw $10, %k1, %k0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    vpinsrb $5, %eax, %xmm3, %xmm3
+; KNL-NEXT:    kshiftlw $9, %k1, %k0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    vpinsrb $6, %eax, %xmm3, %xmm3
+; KNL-NEXT:    kshiftlw $8, %k1, %k0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    vpinsrb $7, %eax, %xmm3, %xmm3
+; KNL-NEXT:    kshiftlw $7, %k1, %k0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    vpinsrb $8, %eax, %xmm3, %xmm3
+; KNL-NEXT:    kshiftlw $6, %k1, %k0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    vpinsrb $9, %eax, %xmm3, %xmm3
+; KNL-NEXT:    kshiftlw $5, %k1, %k0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    vpinsrb $10, %eax, %xmm3, %xmm3
+; KNL-NEXT:    kshiftlw $4, %k1, %k0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    vpinsrb $11, %eax, %xmm3, %xmm3
+; KNL-NEXT:    kshiftlw $3, %k1, %k0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    vpinsrb $12, %eax, %xmm3, %xmm3
+; KNL-NEXT:    kshiftlw $2, %k1, %k0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    vpinsrb $13, %eax, %xmm3, %xmm3
+; KNL-NEXT:    kshiftlw $1, %k1, %k0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    vpinsrb $14, %eax, %xmm3, %xmm3
+; KNL-NEXT:    kshiftrw $15, %k1, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    vpinsrb $15, %eax, %xmm3, %xmm3
+; KNL-NEXT:    vmovups 68(%rdi), %zmm4 {%k2} {z}
 ; KNL-NEXT:    vcmpltps %zmm4, %zmm1, %k0
+; KNL-NEXT:    kshiftlw $14, %k0, %k2
+; KNL-NEXT:    kshiftrw $15, %k2, %k2
+; KNL-NEXT:    kmovw %k2, %eax
+; KNL-NEXT:    kshiftlw $15, %k0, %k2
+; KNL-NEXT:    kshiftrw $15, %k2, %k2
+; KNL-NEXT:    kmovw %k2, %ecx
+; KNL-NEXT:    vmovd %ecx, %xmm4
+; KNL-NEXT:    vpinsrb $1, %eax, %xmm4, %xmm4
+; KNL-NEXT:    kshiftlw $13, %k0, %k2
+; KNL-NEXT:    kshiftrw $15, %k2, %k2
+; KNL-NEXT:    kmovw %k2, %eax
+; KNL-NEXT:    vpinsrb $2, %eax, %xmm4, %xmm4
+; KNL-NEXT:    kshiftlw $12, %k0, %k2
+; KNL-NEXT:    kshiftrw $15, %k2, %k2
+; KNL-NEXT:    kmovw %k2, %eax
+; KNL-NEXT:    vpinsrb $3, %eax, %xmm4, %xmm4
+; KNL-NEXT:    kshiftlw $11, %k0, %k2
+; KNL-NEXT:    kshiftrw $15, %k2, %k2
+; KNL-NEXT:    kmovw %k2, %eax
+; KNL-NEXT:    vpinsrb $4, %eax, %xmm4, %xmm4
+; KNL-NEXT:    kshiftlw $10, %k0, %k2
+; KNL-NEXT:    kshiftrw $15, %k2, %k2
+; KNL-NEXT:    kmovw %k2, %eax
+; KNL-NEXT:    vpinsrb $5, %eax, %xmm4, %xmm4
+; KNL-NEXT:    kshiftlw $9, %k0, %k2
+; KNL-NEXT:    kshiftrw $15, %k2, %k2
+; KNL-NEXT:    kmovw %k2, %eax
+; KNL-NEXT:    vpinsrb $6, %eax, %xmm4, %xmm4
+; KNL-NEXT:    kshiftlw $8, %k0, %k2
+; KNL-NEXT:    kshiftrw $15, %k2, %k2
+; KNL-NEXT:    kmovw %k2, %eax
+; KNL-NEXT:    vpinsrb $7, %eax, %xmm4, %xmm4
+; KNL-NEXT:    kshiftlw $7, %k0, %k2
+; KNL-NEXT:    kshiftrw $15, %k2, %k2
+; KNL-NEXT:    kmovw %k2, %eax
+; KNL-NEXT:    vpinsrb $8, %eax, %xmm4, %xmm4
+; KNL-NEXT:    kshiftlw $6, %k0, %k2
+; KNL-NEXT:    kshiftrw $15, %k2, %k2
+; KNL-NEXT:    kmovw %k2, %eax
+; KNL-NEXT:    vpinsrb $9, %eax, %xmm4, %xmm4
+; KNL-NEXT:    kshiftlw $5, %k0, %k2
+; KNL-NEXT:    kshiftrw $15, %k2, %k2
+; KNL-NEXT:    kmovw %k2, %eax
+; KNL-NEXT:    vpinsrb $10, %eax, %xmm4, %xmm4
+; KNL-NEXT:    kshiftlw $4, %k0, %k2
+; KNL-NEXT:    kshiftrw $15, %k2, %k2
+; KNL-NEXT:    kmovw %k2, %eax
+; KNL-NEXT:    vpinsrb $11, %eax, %xmm4, %xmm4
+; KNL-NEXT:    kshiftlw $3, %k0, %k2
+; KNL-NEXT:    kshiftrw $15, %k2, %k2
+; KNL-NEXT:    kmovw %k2, %eax
+; KNL-NEXT:    vpinsrb $12, %eax, %xmm4, %xmm4
+; KNL-NEXT:    kshiftlw $2, %k0, %k2
+; KNL-NEXT:    kshiftrw $15, %k2, %k2
+; KNL-NEXT:    kmovw %k2, %eax
+; KNL-NEXT:    vpinsrb $13, %eax, %xmm4, %xmm4
+; KNL-NEXT:    kshiftlw $1, %k0, %k2
+; KNL-NEXT:    kshiftrw $15, %k2, %k2
+; KNL-NEXT:    kmovw %k2, %eax
+; KNL-NEXT:    vpinsrb $14, %eax, %xmm4, %xmm4
+; KNL-NEXT:    vinserti128 $1, %xmm2, %ymm3, %ymm2
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    vpinsrb $15, %eax, %xmm4, %xmm3
+; KNL-NEXT:    vmovups 4(%rdi), %zmm4 {%k1} {z}
+; KNL-NEXT:    vcmpltps %zmm4, %zmm0, %k0
 ; KNL-NEXT:    kshiftlw $14, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
 ; KNL-NEXT:    kmovw %k1, %eax
@@ -2034,71 +2098,7 @@ define void @ktest_2(<32 x float> %in, float * %base) {
 ; KNL-NEXT:    kshiftrw $15, %k0, %k0
 ; KNL-NEXT:    kmovw %k0, %eax
 ; KNL-NEXT:    vpinsrb $15, %eax, %xmm4, %xmm4
-; KNL-NEXT:    vcmpltps %zmm3, %zmm0, %k0
+; KNL-NEXT:    vinserti128 $1, %xmm3, %ymm4, %ymm3
-; KNL-NEXT:    kshiftlw $14, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    kshiftlw $15, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
-; KNL-NEXT:    vmovd %ecx, %xmm3
-; KNL-NEXT:    vpinsrb $1, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $13, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $2, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $12, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $3, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $11, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $4, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $10, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $5, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $9, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $6, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $8, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $7, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $7, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $8, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $6, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $9, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $5, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $10, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $4, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $11, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $3, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $12, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $2, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $13, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftlw $1, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $14, %eax, %xmm3, %xmm3
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    vpinsrb $15, %eax, %xmm3, %xmm3
-; KNL-NEXT:    vinserti128 $1, %xmm4, %ymm3, %ymm3
 ; KNL-NEXT:    vpor %ymm3, %ymm2, %ymm2
 ; KNL-NEXT:    vextracti128 $1, %ymm2, %xmm3
 ; KNL-NEXT:    vpmovsxbd %xmm3, %zmm3
@@ -2943,36 +2943,6 @@ define void @store_64i1(<64 x i1>* %a, <64 x i1> %v) {
 ;
 ; KNL-LABEL: store_64i1:
 ; KNL:       ## BB#0:
-; KNL-NEXT:    pushq %rbp
-; KNL-NEXT:  Lcfi9:
-; KNL-NEXT:    .cfi_def_cfa_offset 16
-; KNL-NEXT:    pushq %r15
-; KNL-NEXT:  Lcfi10:
-; KNL-NEXT:    .cfi_def_cfa_offset 24
-; KNL-NEXT:    pushq %r14
-; KNL-NEXT:  Lcfi11:
-; KNL-NEXT:    .cfi_def_cfa_offset 32
-; KNL-NEXT:    pushq %r13
-; KNL-NEXT:  Lcfi12:
-; KNL-NEXT:    .cfi_def_cfa_offset 40
-; KNL-NEXT:    pushq %r12
-; KNL-NEXT:  Lcfi13:
-; KNL-NEXT:    .cfi_def_cfa_offset 48
-; KNL-NEXT:    pushq %rbx
-; KNL-NEXT:  Lcfi14:
-; KNL-NEXT:    .cfi_def_cfa_offset 56
-; KNL-NEXT:  Lcfi15:
-; KNL-NEXT:    .cfi_offset %rbx, -56
-; KNL-NEXT:  Lcfi16:
-; KNL-NEXT:    .cfi_offset %r12, -48
-; KNL-NEXT:  Lcfi17:
-; KNL-NEXT:    .cfi_offset %r13, -40
-; KNL-NEXT:  Lcfi18:
-; KNL-NEXT:    .cfi_offset %r14, -32
-; KNL-NEXT:  Lcfi19:
-; KNL-NEXT:    .cfi_offset %r15, -24
-; KNL-NEXT:  Lcfi20:
-; KNL-NEXT:    .cfi_offset %rbp, -16
 ; KNL-NEXT:    vpmovsxbd %xmm0, %zmm0
 ; KNL-NEXT:    vpslld $31, %zmm0, %zmm0
 ; KNL-NEXT:    vpmovsxbd %xmm1, %zmm1
@@ -2984,281 +2954,275 @@ define void @store_64i1(<64 x i1>* %a, <64 x i1> %v) {
 ; KNL-NEXT:    vptestmd %zmm3, %zmm3, %k0
 ; KNL-NEXT:    kshiftlw $14, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r8d
+; KNL-NEXT:    kmovw %k1, %eax
 ; KNL-NEXT:    kshiftlw $15, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r9d
+; KNL-NEXT:    kmovw %k1, %ecx
 ; KNL-NEXT:    kshiftlw $13, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r10d
+; KNL-NEXT:    vmovd %ecx, %xmm3
+; KNL-NEXT:    kmovw %k1, %ecx
 ; KNL-NEXT:    kshiftlw $12, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r11d
+; KNL-NEXT:    vpinsrb $1, %eax, %xmm3, %xmm3
+; KNL-NEXT:    kmovw %k1, %eax
 ; KNL-NEXT:    kshiftlw $11, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r14d
+; KNL-NEXT:    vpinsrb $2, %ecx, %xmm3, %xmm3
+; KNL-NEXT:    kmovw %k1, %ecx
 ; KNL-NEXT:    kshiftlw $10, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r15d
+; KNL-NEXT:    vpinsrb $3, %eax, %xmm3, %xmm3
+; KNL-NEXT:    kmovw %k1, %eax
 ; KNL-NEXT:    kshiftlw $9, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r12d
+; KNL-NEXT:    vpinsrb $4, %ecx, %xmm3, %xmm3
+; KNL-NEXT:    kmovw %k1, %ecx
 ; KNL-NEXT:    kshiftlw $8, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r13d
+; KNL-NEXT:    vpinsrb $5, %eax, %xmm3, %xmm3
+; KNL-NEXT:    kmovw %k1, %eax
 ; KNL-NEXT:    kshiftlw $7, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ebx
+; KNL-NEXT:    vpinsrb $6, %ecx, %xmm3, %xmm3
+; KNL-NEXT:    kmovw %k1, %ecx
 ; KNL-NEXT:    kshiftlw $6, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ebp
+; KNL-NEXT:    vpinsrb $7, %eax, %xmm3, %xmm3
+; KNL-NEXT:    kmovw %k1, %eax
 ; KNL-NEXT:    kshiftlw $5, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %eax
+; KNL-NEXT:    vpinsrb $8, %ecx, %xmm3, %xmm3
+; KNL-NEXT:    kmovw %k1, %ecx
 ; KNL-NEXT:    kshiftlw $4, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    vpinsrb $9, %eax, %xmm3, %xmm3
+; KNL-NEXT:    kmovw %k1, %eax
 ; KNL-NEXT:    kshiftlw $3, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %edx
+; KNL-NEXT:    vpinsrb $10, %ecx, %xmm3, %xmm3
+; KNL-NEXT:    kmovw %k1, %ecx
 ; KNL-NEXT:    kshiftlw $2, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %esi
+; KNL-NEXT:    vpinsrb $11, %eax, %xmm3, %xmm3
+; KNL-NEXT:    kmovw %k1, %eax
 ; KNL-NEXT:    kshiftlw $1, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    vmovd %r9d, %xmm3
+; KNL-NEXT:    vpinsrb $12, %ecx, %xmm3, %xmm3
-; KNL-NEXT:    kmovw %k1, %r9d
+; KNL-NEXT:    kmovw %k1, %ecx
-; KNL-NEXT:    vptestmd %zmm2, %zmm2, %k2
+; KNL-NEXT:    vptestmd %zmm2, %zmm2, %k1
 ; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $1, %r8d, %xmm3, %xmm2
+; KNL-NEXT:    vpinsrb $13, %eax, %xmm3, %xmm2
-; KNL-NEXT:    vpinsrb $2, %r10d, %xmm2, %xmm2
-; KNL-NEXT:    vpinsrb $3, %r11d, %xmm2, %xmm2
-; KNL-NEXT:    vpinsrb $4, %r14d, %xmm2, %xmm2
-; KNL-NEXT:    vpinsrb $5, %r15d, %xmm2, %xmm2
-; KNL-NEXT:    vpinsrb $6, %r12d, %xmm2, %xmm2
-; KNL-NEXT:    vpinsrb $7, %r13d, %xmm2, %xmm2
-; KNL-NEXT:    vpinsrb $8, %ebx, %xmm2, %xmm2
-; KNL-NEXT:    vpinsrb $9, %ebp, %xmm2, %xmm2
-; KNL-NEXT:    vpinsrb $10, %eax, %xmm2, %xmm2
-; KNL-NEXT:    vpinsrb $11, %ecx, %xmm2, %xmm2
-; KNL-NEXT:    vpinsrb $12, %edx, %xmm2, %xmm2
-; KNL-NEXT:    vpinsrb $13, %esi, %xmm2, %xmm2
-; KNL-NEXT:    vpinsrb $14, %r9d, %xmm2, %xmm2
 ; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    vpinsrb $15, %eax, %xmm2, %xmm2
-; KNL-NEXT:    vpmovsxbd %xmm2, %zmm2
-; KNL-NEXT:    vpslld $31, %zmm2, %zmm2
-; KNL-NEXT:    vptestmd %zmm2, %zmm2, %k0
-; KNL-NEXT:    kmovw %k0, 6(%rdi)
-; KNL-NEXT:    kshiftlw $14, %k2, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %r8d
-; KNL-NEXT:    kshiftlw $15, %k2, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %r10d
-; KNL-NEXT:    kshiftlw $13, %k2, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %r9d
-; KNL-NEXT:    kshiftlw $12, %k2, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %r11d
-; KNL-NEXT:    kshiftlw $11, %k2, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %r14d
-; KNL-NEXT:    kshiftlw $10, %k2, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %r15d
-; KNL-NEXT:    kshiftlw $9, %k2, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %r12d
-; KNL-NEXT:    kshiftlw $8, %k2, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %r13d
-; KNL-NEXT:    kshiftlw $7, %k2, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %ecx
-; KNL-NEXT:    kshiftlw $6, %k2, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %esi
-; KNL-NEXT:    kshiftlw $5, %k2, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %ebp
-; KNL-NEXT:    kshiftlw $4, %k2, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %ebx
-; KNL-NEXT:    kshiftlw $3, %k2, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    kshiftlw $2, %k2, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %edx
-; KNL-NEXT:    kshiftlw $1, %k2, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vmovd %r10d, %xmm2
-; KNL-NEXT:    kmovw %k0, %r10d
-; KNL-NEXT:    vptestmd %zmm1, %zmm1, %k1
-; KNL-NEXT:    kshiftrw $15, %k2, %k0
-; KNL-NEXT:    vpinsrb $1, %r8d, %xmm2, %xmm1
-; KNL-NEXT:    vpinsrb $2, %r9d, %xmm1, %xmm1
-; KNL-NEXT:    vpinsrb $3, %r11d, %xmm1, %xmm1
-; KNL-NEXT:    vpinsrb $4, %r14d, %xmm1, %xmm1
-; KNL-NEXT:    vpinsrb $5, %r15d, %xmm1, %xmm1
-; KNL-NEXT:    vpinsrb $6, %r12d, %xmm1, %xmm1
-; KNL-NEXT:    vpinsrb $7, %r13d, %xmm1, %xmm1
-; KNL-NEXT:    vpinsrb $8, %ecx, %xmm1, %xmm1
-; KNL-NEXT:    vpinsrb $9, %esi, %xmm1, %xmm1
-; KNL-NEXT:    vpinsrb $10, %ebp, %xmm1, %xmm1
-; KNL-NEXT:    vpinsrb $11, %ebx, %xmm1, %xmm1
-; KNL-NEXT:    vpinsrb $12, %eax, %xmm1, %xmm1
-; KNL-NEXT:    vpinsrb $13, %edx, %xmm1, %xmm1
-; KNL-NEXT:    vpinsrb $14, %r10d, %xmm1, %xmm1
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    vpinsrb $15, %eax, %xmm1, %xmm1
-; KNL-NEXT:    vpmovsxbd %xmm1, %zmm1
-; KNL-NEXT:    vpslld $31, %zmm1, %zmm1
-; KNL-NEXT:    vptestmd %zmm1, %zmm1, %k0
-; KNL-NEXT:    kmovw %k0, 4(%rdi)
 ; KNL-NEXT:    kshiftlw $14, %k1, %k0
 ; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %r8d
+; KNL-NEXT:    vpinsrb $14, %ecx, %xmm2, %xmm2
+; KNL-NEXT:    kmovw %k0, %ecx
 ; KNL-NEXT:    kshiftlw $15, %k1, %k0
 ; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %r10d
+; KNL-NEXT:    vpinsrb $15, %eax, %xmm2, %xmm2
+; KNL-NEXT:    kmovw %k0, %eax
 ; KNL-NEXT:    kshiftlw $13, %k1, %k0
 ; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %r9d
+; KNL-NEXT:    vmovd %eax, %xmm3
+; KNL-NEXT:    kmovw %k0, %eax
 ; KNL-NEXT:    kshiftlw $12, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    vpinsrb $1, %ecx, %xmm3, %xmm3
-; KNL-NEXT:    kmovw %k0, %r11d
-; KNL-NEXT:    kshiftlw $11, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %r14d
-; KNL-NEXT:    kshiftlw $10, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %r15d
-; KNL-NEXT:    kshiftlw $9, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %r12d
-; KNL-NEXT:    kshiftlw $8, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %r13d
-; KNL-NEXT:    kshiftlw $7, %k1, %k0
 ; KNL-NEXT:    kshiftrw $15, %k0, %k0
 ; KNL-NEXT:    kmovw %k0, %ecx
+; KNL-NEXT:    kshiftlw $11, %k1, %k0
+; KNL-NEXT:    vpinsrb $2, %eax, %xmm3, %xmm3
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    kshiftlw $10, %k1, %k0
+; KNL-NEXT:    vpinsrb $3, %ecx, %xmm3, %xmm3
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %ecx
+; KNL-NEXT:    kshiftlw $9, %k1, %k0
+; KNL-NEXT:    vpinsrb $4, %eax, %xmm3, %xmm3
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    kshiftlw $8, %k1, %k0
+; KNL-NEXT:    vpinsrb $5, %ecx, %xmm3, %xmm3
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %ecx
+; KNL-NEXT:    kshiftlw $7, %k1, %k0
+; KNL-NEXT:    vpinsrb $6, %eax, %xmm3, %xmm3
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
 ; KNL-NEXT:    kshiftlw $6, %k1, %k0
+; KNL-NEXT:    vpinsrb $7, %ecx, %xmm3, %xmm3
 ; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %esi
+; KNL-NEXT:    kmovw %k0, %ecx
 ; KNL-NEXT:    kshiftlw $5, %k1, %k0
+; KNL-NEXT:    vpinsrb $8, %eax, %xmm3, %xmm3
 ; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %ebp
+; KNL-NEXT:    kmovw %k0, %eax
 ; KNL-NEXT:    kshiftlw $4, %k1, %k0
+; KNL-NEXT:    vpinsrb $9, %ecx, %xmm3, %xmm3
 ; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %ebx
+; KNL-NEXT:    kmovw %k0, %ecx
 ; KNL-NEXT:    kshiftlw $3, %k1, %k0
+; KNL-NEXT:    vpinsrb $10, %eax, %xmm3, %xmm3
 ; KNL-NEXT:    kshiftrw $15, %k0, %k0
 ; KNL-NEXT:    kmovw %k0, %eax
 ; KNL-NEXT:    kshiftlw $2, %k1, %k0
+; KNL-NEXT:    vpinsrb $11, %ecx, %xmm3, %xmm3
 ; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %edx
+; KNL-NEXT:    kmovw %k0, %ecx
 ; KNL-NEXT:    kshiftlw $1, %k1, %k0
+; KNL-NEXT:    vpinsrb $12, %eax, %xmm3, %xmm3
 ; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vmovd %r10d, %xmm1
+; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    kmovw %k0, %r10d
+; KNL-NEXT:    vptestmd %zmm1, %zmm1, %k0
-; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k0
+; KNL-NEXT:    kshiftrw $15, %k1, %k1
+; KNL-NEXT:    vpmovsxbd %xmm2, %zmm1
+; KNL-NEXT:    vpslld $31, %zmm1, %zmm1
+; KNL-NEXT:    vpinsrb $13, %ecx, %xmm3, %xmm2
+; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    vptestmd %zmm1, %zmm1, %k1
+; KNL-NEXT:    vpinsrb $14, %eax, %xmm2, %xmm1
+; KNL-NEXT:    vpinsrb $15, %ecx, %xmm1, %xmm1
+; KNL-NEXT:    kmovw %k1, 6(%rdi)
+; KNL-NEXT:    vpmovsxbd %xmm1, %zmm1
+; KNL-NEXT:    vpslld $31, %zmm1, %zmm1
+; KNL-NEXT:    vptestmd %zmm1, %zmm1, %k1
+; KNL-NEXT:    kmovw %k1, 4(%rdi)
+; KNL-NEXT:    kshiftlw $14, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    vpinsrb $1, %r8d, %xmm1, %xmm0
-; KNL-NEXT:    vpinsrb $2, %r9d, %xmm0, %xmm0
-; KNL-NEXT:    vpinsrb $3, %r11d, %xmm0, %xmm0
-; KNL-NEXT:    vpinsrb $4, %r14d, %xmm0, %xmm0
-; KNL-NEXT:    vpinsrb $5, %r15d, %xmm0, %xmm0
-; KNL-NEXT:    vpinsrb $6, %r12d, %xmm0, %xmm0
-; KNL-NEXT:    vpinsrb $7, %r13d, %xmm0, %xmm0
-; KNL-NEXT:    vpinsrb $8, %ecx, %xmm0, %xmm0
-; KNL-NEXT:    vpinsrb $9, %esi, %xmm0, %xmm0
-; KNL-NEXT:    vpinsrb $10, %ebp, %xmm0, %xmm0
-; KNL-NEXT:    vpinsrb $11, %ebx, %xmm0, %xmm0
-; KNL-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0
-; KNL-NEXT:    vpinsrb $13, %edx, %xmm0, %xmm0
-; KNL-NEXT:    vpinsrb $14, %r10d, %xmm0, %xmm0
 ; KNL-NEXT:    kmovw %k1, %eax
+; KNL-NEXT:    kshiftlw $15, %k0, %k1
+; KNL-NEXT:    kshiftrw $15, %k1, %k1
+; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kshiftlw $13, %k0, %k1
+; KNL-NEXT:    kshiftrw $15, %k1, %k1
+; KNL-NEXT:    vmovd %ecx, %xmm1
+; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kshiftlw $12, %k0, %k1
+; KNL-NEXT:    kshiftrw $15, %k1, %k1
+; KNL-NEXT:    vpinsrb $1, %eax, %xmm1, %xmm1
+; KNL-NEXT:    kmovw %k1, %eax
+; KNL-NEXT:    kshiftlw $11, %k0, %k1
+; KNL-NEXT:    kshiftrw $15, %k1, %k1
+; KNL-NEXT:    vpinsrb $2, %ecx, %xmm1, %xmm1
+; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kshiftlw $10, %k0, %k1
+; KNL-NEXT:    kshiftrw $15, %k1, %k1
+; KNL-NEXT:    vpinsrb $3, %eax, %xmm1, %xmm1
+; KNL-NEXT:    kmovw %k1, %eax
+; KNL-NEXT:    kshiftlw $9, %k0, %k1
+; KNL-NEXT:    kshiftrw $15, %k1, %k1
+; KNL-NEXT:    vpinsrb $4, %ecx, %xmm1, %xmm1
+; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kshiftlw $8, %k0, %k1
+; KNL-NEXT:    kshiftrw $15, %k1, %k1
+; KNL-NEXT:    vpinsrb $5, %eax, %xmm1, %xmm1
+; KNL-NEXT:    kmovw %k1, %eax
+; KNL-NEXT:    kshiftlw $7, %k0, %k1
+; KNL-NEXT:    kshiftrw $15, %k1, %k1
+; KNL-NEXT:    vpinsrb $6, %ecx, %xmm1, %xmm1
+; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kshiftlw $6, %k0, %k1
+; KNL-NEXT:    kshiftrw $15, %k1, %k1
+; KNL-NEXT:    vpinsrb $7, %eax, %xmm1, %xmm1
+; KNL-NEXT:    kmovw %k1, %eax
+; KNL-NEXT:    kshiftlw $5, %k0, %k1
+; KNL-NEXT:    kshiftrw $15, %k1, %k1
+; KNL-NEXT:    vpinsrb $8, %ecx, %xmm1, %xmm1
+; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kshiftlw $4, %k0, %k1
+; KNL-NEXT:    kshiftrw $15, %k1, %k1
+; KNL-NEXT:    vpinsrb $9, %eax, %xmm1, %xmm1
+; KNL-NEXT:    kmovw %k1, %eax
+; KNL-NEXT:    kshiftlw $3, %k0, %k1
+; KNL-NEXT:    kshiftrw $15, %k1, %k1
+; KNL-NEXT:    vpinsrb $10, %ecx, %xmm1, %xmm1
+; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    kshiftlw $2, %k0, %k1
+; KNL-NEXT:    kshiftrw $15, %k1, %k1
+; KNL-NEXT:    vpinsrb $11, %eax, %xmm1, %xmm1
+; KNL-NEXT:    kmovw %k1, %eax
+; KNL-NEXT:    kshiftlw $1, %k0, %k1
+; KNL-NEXT:    kshiftrw $15, %k1, %k1
+; KNL-NEXT:    vpinsrb $12, %ecx, %xmm1, %xmm1
+; KNL-NEXT:    kmovw %k1, %ecx
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k1
+; KNL-NEXT:    vpinsrb $13, %eax, %xmm1, %xmm0
+; KNL-NEXT:    vpinsrb $14, %ecx, %xmm0, %xmm0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
 ; KNL-NEXT:    vpinsrb $15, %eax, %xmm0, %xmm0
 ; KNL-NEXT:    vpmovsxbd %xmm0, %zmm0
 ; KNL-NEXT:    vpslld $31, %zmm0, %zmm0
-; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k1
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k0
-; KNL-NEXT:    kmovw %k1, 2(%rdi)
+; KNL-NEXT:    kmovw %k0, 2(%rdi)
-; KNL-NEXT:    kshiftlw $14, %k0, %k1
+; KNL-NEXT:    kshiftlw $14, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r8d
-; KNL-NEXT:    kshiftlw $15, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r9d
-; KNL-NEXT:    kshiftlw $13, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r10d
-; KNL-NEXT:    kshiftlw $12, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r11d
-; KNL-NEXT:    kshiftlw $11, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r14d
-; KNL-NEXT:    kshiftlw $10, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r15d
-; KNL-NEXT:    kshiftlw $9, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r12d
-; KNL-NEXT:    kshiftlw $8, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r13d
-; KNL-NEXT:    kshiftlw $7, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %edx
-; KNL-NEXT:    kshiftlw $6, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %esi
-; KNL-NEXT:    kshiftlw $5, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ebp
-; KNL-NEXT:    kshiftlw $4, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ebx
-; KNL-NEXT:    kshiftlw $3, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    kshiftlw $2, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
-; KNL-NEXT:    kshiftlw $1, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    vmovd %r9d, %xmm0
-; KNL-NEXT:    kmovw %k1, %r9d
-; KNL-NEXT:    vpinsrb $1, %r8d, %xmm0, %xmm0
-; KNL-NEXT:    vpinsrb $2, %r10d, %xmm0, %xmm0
-; KNL-NEXT:    vpinsrb $3, %r11d, %xmm0, %xmm0
-; KNL-NEXT:    vpinsrb $4, %r14d, %xmm0, %xmm0
-; KNL-NEXT:    vpinsrb $5, %r15d, %xmm0, %xmm0
-; KNL-NEXT:    vpinsrb $6, %r12d, %xmm0, %xmm0
-; KNL-NEXT:    vpinsrb $7, %r13d, %xmm0, %xmm0
-; KNL-NEXT:    vpinsrb $8, %edx, %xmm0, %xmm0
-; KNL-NEXT:    vpinsrb $9, %esi, %xmm0, %xmm0
-; KNL-NEXT:    vpinsrb $10, %ebp, %xmm0, %xmm0
-; KNL-NEXT:    vpinsrb $11, %ebx, %xmm0, %xmm0
-; KNL-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0
 ; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $13, %ecx, %xmm0, %xmm0
+; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    vpinsrb $14, %r9d, %xmm0, %xmm0
+; KNL-NEXT:    kshiftlw $15, %k1, %k0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %ecx
+; KNL-NEXT:    kshiftlw $13, %k1, %k0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    vmovd %ecx, %xmm0
+; KNL-NEXT:    kmovw %k0, %ecx
+; KNL-NEXT:    kshiftlw $12, %k1, %k0
+; KNL-NEXT:    vpinsrb $1, %eax, %xmm0, %xmm0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    kshiftlw $11, %k1, %k0
+; KNL-NEXT:    vpinsrb $2, %ecx, %xmm0, %xmm0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %ecx
+; KNL-NEXT:    kshiftlw $10, %k1, %k0
+; KNL-NEXT:    vpinsrb $3, %eax, %xmm0, %xmm0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    kshiftlw $9, %k1, %k0
+; KNL-NEXT:    vpinsrb $4, %ecx, %xmm0, %xmm0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %ecx
+; KNL-NEXT:    kshiftlw $8, %k1, %k0
+; KNL-NEXT:    vpinsrb $5, %eax, %xmm0, %xmm0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    kshiftlw $7, %k1, %k0
+; KNL-NEXT:    vpinsrb $6, %ecx, %xmm0, %xmm0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %ecx
+; KNL-NEXT:    kshiftlw $6, %k1, %k0
+; KNL-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    kshiftlw $5, %k1, %k0
+; KNL-NEXT:    vpinsrb $8, %ecx, %xmm0, %xmm0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %ecx
+; KNL-NEXT:    kshiftlw $4, %k1, %k0
+; KNL-NEXT:    vpinsrb $9, %eax, %xmm0, %xmm0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    kshiftlw $3, %k1, %k0
+; KNL-NEXT:    vpinsrb $10, %ecx, %xmm0, %xmm0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %ecx
+; KNL-NEXT:    kshiftlw $2, %k1, %k0
+; KNL-NEXT:    vpinsrb $11, %eax, %xmm0, %xmm0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    kshiftlw $1, %k1, %k0
+; KNL-NEXT:    vpinsrb $12, %ecx, %xmm0, %xmm0
+; KNL-NEXT:    kshiftrw $15, %k0, %k0
+; KNL-NEXT:    kmovw %k0, %ecx
+; KNL-NEXT:    kshiftrw $15, %k1, %k0
+; KNL-NEXT:    vpinsrb $13, %eax, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $14, %ecx, %xmm0, %xmm0
 ; KNL-NEXT:    kmovw %k0, %eax
 ; KNL-NEXT:    vpinsrb $15, %eax, %xmm0, %xmm0
 ; KNL-NEXT:    vpmovsxbd %xmm0, %zmm0
 ; KNL-NEXT:    vpslld $31, %zmm0, %zmm0
 ; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k0
 ; KNL-NEXT:    kmovw %k0, (%rdi)
-; KNL-NEXT:    popq %rbx
-; KNL-NEXT:    popq %r12
-; KNL-NEXT:    popq %r13
-; KNL-NEXT:    popq %r14
-; KNL-NEXT:    popq %r15
-; KNL-NEXT:    popq %rbp
 ; KNL-NEXT:    retq
 ;
 ; SKX-LABEL: store_64i1:

llvm/test/CodeGen/X86/avx512-vec-cmp.ll

@@ -8,6 +8,7 @@ define <16 x float> @test1(<16 x float> %x, <16 x float> %y) nounwind {
 ; CHECK-NEXT:    vcmpleps %zmm1, %zmm0, %k1
 ; CHECK-NEXT:    vblendmps %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
+; CHECK-NEXT:    ## -- End function
   %mask = fcmp ole <16 x float> %x, %y
   %max = select <16 x i1> %mask, <16 x float> %x, <16 x float> %y
   ret <16 x float> %max
@@ -19,6 +20,7 @@ define <8 x double> @test2(<8 x double> %x, <8 x double> %y) nounwind {
 ; CHECK-NEXT:    vcmplepd %zmm1, %zmm0, %k1
 ; CHECK-NEXT:    vblendmpd %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
+; CHECK-NEXT:    ## -- End function
   %mask = fcmp ole <8 x double> %x, %y
   %max = select <8 x i1> %mask, <8 x double> %x, <8 x double> %y
   ret <8 x double> %max
@@ -30,6 +32,7 @@ define <16 x i32> @test3(<16 x i32> %x, <16 x i32> %x1, <16 x i32>* %yp) nounwin
 ; CHECK-NEXT:    vpcmpeqd (%rdi), %zmm0, %k1
 ; CHECK-NEXT:    vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
+; CHECK-NEXT:    ## -- End function
   %y = load <16 x i32>, <16 x i32>* %yp, align 4
   %mask = icmp eq <16 x i32> %x, %y
   %max = select <16 x i1> %mask, <16 x i32> %x, <16 x i32> %x1
@@ -42,6 +45,7 @@ define <16 x i32> @test4_unsigned(<16 x i32> %x, <16 x i32> %y, <16 x i32> %x1)
 ; CHECK-NEXT:    vpcmpnltud %zmm1, %zmm0, %k1
 ; CHECK-NEXT:    vpblendmd %zmm2, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
+; CHECK-NEXT:    ## -- End function
   %mask = icmp uge <16 x i32> %x, %y
   %max = select <16 x i1> %mask, <16 x i32> %x1, <16 x i32> %y
   ret <16 x i32> %max
@@ -53,6 +57,7 @@ define <8 x i64> @test5(<8 x i64> %x, <8 x i64> %y) nounwind {
 ; CHECK-NEXT:    vpcmpeqq %zmm1, %zmm0, %k1
 ; CHECK-NEXT:    vpblendmq %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
+; CHECK-NEXT:    ## -- End function
   %mask = icmp eq <8 x i64> %x, %y
   %max = select <8 x i1> %mask, <8 x i64> %x, <8 x i64> %y
   ret <8 x i64> %max
@@ -64,6 +69,7 @@ define <8 x i64> @test6_unsigned(<8 x i64> %x, <8 x i64> %y, <8 x i64> %x1) noun
 ; CHECK-NEXT:    vpcmpnleuq %zmm1, %zmm0, %k1
 ; CHECK-NEXT:    vpblendmq %zmm2, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
+; CHECK-NEXT:    ## -- End function
   %mask = icmp ugt <8 x i64> %x, %y
   %max = select <8 x i1> %mask, <8 x i64> %x1, <8 x i64> %y
   ret <8 x i64> %max
@@ -117,12 +123,14 @@ define <8 x i32> @test9(<8 x i32> %x, <8 x i32> %y) nounwind {
 ; KNL-NEXT:    vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
 ; KNL-NEXT:    ## kill: %YMM0<def> %YMM0<kill> %ZMM0<kill>
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test9:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vpcmpeqd %ymm1, %ymm0, %k1
 ; SKX-NEXT:    vpblendmd %ymm0, %ymm1, %ymm0 {%k1}
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
   %mask = icmp eq <8 x i32> %x, %y
   %max = select <8 x i1> %mask, <8 x i32> %x, <8 x i32> %y
   ret <8 x i32> %max
@@ -137,12 +145,14 @@ define <8 x float> @test10(<8 x float> %x, <8 x float> %y) nounwind {
 ; KNL-NEXT:    vblendmps %zmm0, %zmm1, %zmm0 {%k1}
 ; KNL-NEXT:    ## kill: %YMM0<def> %YMM0<kill> %ZMM0<kill>
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test10:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpeqps %ymm1, %ymm0, %k1
 ; SKX-NEXT:    vblendmps %ymm0, %ymm1, %ymm0 {%k1}
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
 
   %mask = fcmp oeq <8 x float> %x, %y
   %max = select <8 x i1> %mask, <8 x float> %x, <8 x float> %y
@@ -154,6 +164,7 @@ define <8 x i32> @test11_unsigned(<8 x i32> %x, <8 x i32> %y) nounwind {
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpmaxud %ymm1, %ymm0, %ymm0
 ; CHECK-NEXT:    retq
+; CHECK-NEXT:    ## -- End function
   %mask = icmp ugt <8 x i32> %x, %y
   %max = select <8 x i1> %mask, <8 x i32> %x, <8 x i32> %y
   ret <8 x i32> %max
@@ -168,6 +179,7 @@ define i16 @test12(<16 x i64> %a, <16 x i64> %b) nounwind {
 ; KNL-NEXT:    kmovw %k0, %eax
 ; KNL-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test12:
 ; SKX:       ## BB#0:
@@ -178,6 +190,7 @@ define i16 @test12(<16 x i64> %a, <16 x i64> %b) nounwind {
 ; SKX-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
 ; SKX-NEXT:    vzeroupper
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
   %res = icmp eq <16 x i64> %a, %b
   %res1 = bitcast <16 x i1> %res to i16
   ret i16 %res1
@@ -330,6 +343,7 @@ define i32 @test12_v32i32(<32 x i32> %a, <32 x i32> %b) nounwind {
 ; KNL-NEXT:    movq %rbp, %rsp
 ; KNL-NEXT:    popq %rbp
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test12_v32i32:
 ; SKX:       ## BB#0:
@@ -339,6 +353,7 @@ define i32 @test12_v32i32(<32 x i32> %a, <32 x i32> %b) nounwind {
 ; SKX-NEXT:    kmovd %k0, %eax
 ; SKX-NEXT:    vzeroupper
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
   %res = icmp eq <32 x i32> %a, %b
   %res1 = bitcast <32 x i1> %res to i32
   ret i32 %res1
@@ -562,72 +577,72 @@ define i64 @test12_v64i16(<64 x i16> %a, <64 x i16> %b) nounwind {
 ; KNL-NEXT:    vpinsrb $15, %eax, %xmm0, %xmm0
 ; KNL-NEXT:    vpmovsxbd %xmm0, %zmm0
 ; KNL-NEXT:    vpslld $31, %zmm0, %zmm0
-; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k0
+; KNL-NEXT:    vpcmpeqw %ymm6, %ymm2, %ymm1
-; KNL-NEXT:    kmovw %k0, {{[0-9]+}}(%rsp)
+; KNL-NEXT:    vpmovsxwd %ymm1, %zmm1
-; KNL-NEXT:    vpcmpeqw %ymm6, %ymm2, %ymm0
+; KNL-NEXT:    vpslld $31, %zmm1, %zmm1
-; KNL-NEXT:    vpmovsxwd %ymm0, %zmm0
+; KNL-NEXT:    vptestmd %zmm1, %zmm1, %k0
-; KNL-NEXT:    vpslld $31, %zmm0, %zmm0
-; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k0
 ; KNL-NEXT:    kshiftlw $14, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
 ; KNL-NEXT:    kmovw %k1, %eax
 ; KNL-NEXT:    kshiftlw $15, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
 ; KNL-NEXT:    kmovw %k1, %ecx
-; KNL-NEXT:    vmovd %ecx, %xmm0
+; KNL-NEXT:    vmovd %ecx, %xmm1
-; KNL-NEXT:    vpinsrb $1, %eax, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $1, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $13, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
 ; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $2, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $12, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
 ; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $3, %eax, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $3, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $11, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
 ; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $4, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $10, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
 ; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $5, %eax, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $5, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $9, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
 ; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $6, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $8, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
 ; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $7, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $7, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
 ; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $8, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $6, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
 ; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $9, %eax, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $9, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $5, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
 ; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $10, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $4, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
 ; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $11, %eax, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $11, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $3, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
 ; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0
+; KNL-NEXT:    vpinsrb $12, %eax, %xmm1, %xmm1
 ; KNL-NEXT:    kshiftlw $2, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
 ; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    vpinsrb $13, %eax, %xmm0, %xmm0
 ; KNL-NEXT:    kshiftlw $1, %k0, %k1
 ; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %eax
+; KNL-NEXT:    kmovw %k1, %ecx
-; KNL-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k1
+; KNL-NEXT:    vpinsrb $13, %eax, %xmm1, %xmm0
+; KNL-NEXT:    vpinsrb $14, %ecx, %xmm0, %xmm0
+; KNL-NEXT:    kmovw %k1, {{[0-9]+}}(%rsp)
 ; KNL-NEXT:    kshiftrw $15, %k0, %k0
 ; KNL-NEXT:    kmovw %k0, %eax
 ; KNL-NEXT:    vpinsrb $15, %eax, %xmm0, %xmm0
@@ -642,6 +657,7 @@ define i64 @test12_v64i16(<64 x i16> %a, <64 x i16> %b) nounwind {
 ; KNL-NEXT:    movq %rbp, %rsp
 ; KNL-NEXT:    popq %rbp
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test12_v64i16:
 ; SKX:       ## BB#0:
@@ -651,6 +667,7 @@ define i64 @test12_v64i16(<64 x i16> %a, <64 x i16> %b) nounwind {
 ; SKX-NEXT:    kmovq %k0, %rax
 ; SKX-NEXT:    vzeroupper
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
   %res = icmp eq <64 x i16> %a, %b
   %res1 = bitcast <64 x i1> %res to i64
   ret i64 %res1
@@ -704,6 +721,7 @@ define <16 x i32> @test16(<16 x i32> %x, <16 x i32> %y, <16 x i32> %x1) nounwind
 ; CHECK-NEXT:    vpcmpled %zmm0, %zmm1, %k1
 ; CHECK-NEXT:    vpblendmd %zmm2, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
+; CHECK-NEXT:    ## -- End function
   %mask = icmp sge <16 x i32> %x, %y
   %max = select <16 x i1> %mask, <16 x i32> %x1, <16 x i32> %y
   ret <16 x i32> %max
@@ -715,6 +733,7 @@ define <16 x i32> @test17(<16 x i32> %x, <16 x i32> %x1, <16 x i32>* %y.ptr) nou
 ; CHECK-NEXT:    vpcmpgtd (%rdi), %zmm0, %k1
 ; CHECK-NEXT:    vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
+; CHECK-NEXT:    ## -- End function
   %y = load <16 x i32>, <16 x i32>* %y.ptr, align 4
   %mask = icmp sgt <16 x i32> %x, %y
   %max = select <16 x i1> %mask, <16 x i32> %x, <16 x i32> %x1
@@ -727,6 +746,7 @@ define <16 x i32> @test18(<16 x i32> %x, <16 x i32> %x1, <16 x i32>* %y.ptr) nou
 ; CHECK-NEXT:    vpcmpled (%rdi), %zmm0, %k1
 ; CHECK-NEXT:    vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
+; CHECK-NEXT:    ## -- End function
   %y = load <16 x i32>, <16 x i32>* %y.ptr, align 4
   %mask = icmp sle <16 x i32> %x, %y
   %max = select <16 x i1> %mask, <16 x i32> %x, <16 x i32> %x1
@@ -739,6 +759,7 @@ define <16 x i32> @test19(<16 x i32> %x, <16 x i32> %x1, <16 x i32>* %y.ptr) nou
 ; CHECK-NEXT:    vpcmpleud (%rdi), %zmm0, %k1
 ; CHECK-NEXT:    vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
+; CHECK-NEXT:    ## -- End function
   %y = load <16 x i32>, <16 x i32>* %y.ptr, align 4
   %mask = icmp ule <16 x i32> %x, %y
   %max = select <16 x i1> %mask, <16 x i32> %x, <16 x i32> %x1
@@ -752,6 +773,7 @@ define <16 x i32> @test20(<16 x i32> %x, <16 x i32> %y, <16 x i32> %x1, <16 x i3
 ; CHECK-NEXT:    vpcmpeqd %zmm3, %zmm2, %k1 {%k1}
 ; CHECK-NEXT:    vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
+; CHECK-NEXT:    ## -- End function
   %mask1 = icmp eq <16 x i32> %x1, %y1
   %mask0 = icmp eq <16 x i32> %x, %y
   %mask = select <16 x i1> %mask0, <16 x i1> %mask1, <16 x i1> zeroinitializer
@@ -766,6 +788,7 @@ define <8 x i64> @test21(<8 x i64> %x, <8 x i64> %y, <8 x i64> %x1, <8 x i64> %y
 ; CHECK-NEXT:    vpcmpleq %zmm2, %zmm3, %k1 {%k1}
 ; CHECK-NEXT:    vpblendmq %zmm0, %zmm2, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
+; CHECK-NEXT:    ## -- End function
   %mask1 = icmp sge <8 x i64> %x1, %y1
   %mask0 = icmp sle <8 x i64> %x, %y
   %mask = select <8 x i1> %mask0, <8 x i1> %mask1, <8 x i1> zeroinitializer
@@ -780,6 +803,7 @@ define <8 x i64> @test22(<8 x i64> %x, <8 x i64>* %y.ptr, <8 x i64> %x1, <8 x i6
 ; CHECK-NEXT:    vpcmpgtq (%rdi), %zmm0, %k1 {%k1}
 ; CHECK-NEXT:    vpblendmq %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
+; CHECK-NEXT:    ## -- End function
   %mask1 = icmp sgt <8 x i64> %x1, %y1
   %y = load <8 x i64>, <8 x i64>* %y.ptr, align 4
   %mask0 = icmp sgt <8 x i64> %x, %y
@@ -795,6 +819,7 @@ define <16 x i32> @test23(<16 x i32> %x, <16 x i32>* %y.ptr, <16 x i32> %x1, <16
 ; CHECK-NEXT:    vpcmpleud (%rdi), %zmm0, %k1 {%k1}
 ; CHECK-NEXT:    vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
+; CHECK-NEXT:    ## -- End function
   %mask1 = icmp sge <16 x i32> %x1, %y1
   %y = load <16 x i32>, <16 x i32>* %y.ptr, align 4
   %mask0 = icmp ule <16 x i32> %x, %y
@@ -809,6 +834,7 @@ define <8 x i64> @test24(<8 x i64> %x, <8 x i64> %x1, i64* %yb.ptr) nounwind {
 ; CHECK-NEXT:    vpcmpeqq (%rdi){1to8}, %zmm0, %k1
 ; CHECK-NEXT:    vpblendmq %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
+; CHECK-NEXT:    ## -- End function
   %yb = load i64, i64* %yb.ptr, align 4
   %y.0 = insertelement <8 x i64> undef, i64 %yb, i32 0
   %y = shufflevector <8 x i64> %y.0, <8 x i64> undef, <8 x i32> zeroinitializer
@@ -823,6 +849,7 @@ define <16 x i32> @test25(<16 x i32> %x, i32* %yb.ptr, <16 x i32> %x1) nounwind
 ; CHECK-NEXT:    vpcmpled (%rdi){1to16}, %zmm0, %k1
 ; CHECK-NEXT:    vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
+; CHECK-NEXT:    ## -- End function
   %yb = load i32, i32* %yb.ptr, align 4
   %y.0 = insertelement <16 x i32> undef, i32 %yb, i32 0
   %y = shufflevector <16 x i32> %y.0, <16 x i32> undef, <16 x i32> zeroinitializer
@@ -838,6 +865,7 @@ define <16 x i32> @test26(<16 x i32> %x, i32* %yb.ptr, <16 x i32> %x1, <16 x i32
 ; CHECK-NEXT:    vpcmpgtd (%rdi){1to16}, %zmm0, %k1 {%k1}
 ; CHECK-NEXT:    vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
+; CHECK-NEXT:    ## -- End function
   %mask1 = icmp sge <16 x i32> %x1, %y1
   %yb = load i32, i32* %yb.ptr, align 4
   %y.0 = insertelement <16 x i32> undef, i32 %yb, i32 0
@@ -855,6 +883,7 @@ define <8 x i64> @test27(<8 x i64> %x, i64* %yb.ptr, <8 x i64> %x1, <8 x i64> %y
 ; CHECK-NEXT:    vpcmpleq (%rdi){1to8}, %zmm0, %k1 {%k1}
 ; CHECK-NEXT:    vpblendmq %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
+; CHECK-NEXT:    ## -- End function
   %mask1 = icmp sge <8 x i64> %x1, %y1
   %yb = load i64, i64* %yb.ptr, align 4
   %y.0 = insertelement <8 x i64> undef, i64 %yb, i32 0
@@ -920,12 +949,14 @@ define <4 x double> @test30(<4 x double> %x, <4 x double> %y) nounwind {
 ; KNL-NEXT:    vcmpeqpd %ymm1, %ymm0, %ymm2
 ; KNL-NEXT:    vblendvpd %ymm2, %ymm0, %ymm1, %ymm0
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test30:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpeqpd %ymm1, %ymm0, %k1
 ; SKX-NEXT:    vblendmpd %ymm0, %ymm1, %ymm0 {%k1}
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
 
   %mask = fcmp oeq <4 x double> %x, %y
   %max = select <4 x i1> %mask, <4 x double> %x, <4 x double> %y
@@ -938,12 +969,14 @@ define <2 x double> @test31(<2 x double> %x, <2 x double> %x1, <2 x double>* %yp
 ; KNL-NEXT:    vcmpltpd (%rdi), %xmm0, %xmm2
 ; KNL-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test31:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltpd (%rdi), %xmm0, %k1
 ; SKX-NEXT:    vblendmpd %xmm0, %xmm1, %xmm0 {%k1}
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
 
   %y = load <2 x double>, <2 x double>* %yp, align 4
   %mask = fcmp olt <2 x double> %x, %y
@@ -957,12 +990,14 @@ define <4 x double> @test32(<4 x double> %x, <4 x double> %x1, <4 x double>* %yp
 ; KNL-NEXT:    vcmpltpd (%rdi), %ymm0, %ymm2
 ; KNL-NEXT:    vblendvpd %ymm2, %ymm0, %ymm1, %ymm0
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test32:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltpd (%rdi), %ymm0, %k1
 ; SKX-NEXT:    vblendmpd %ymm0, %ymm1, %ymm0 {%k1}
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
 
   %y = load <4 x double>, <4 x double>* %yp, align 4
   %mask = fcmp ogt <4 x double> %y, %x
@@ -976,6 +1011,7 @@ define <8 x double> @test33(<8 x double> %x, <8 x double> %x1, <8 x double>* %yp
 ; CHECK-NEXT:    vcmpltpd (%rdi), %zmm0, %k1
 ; CHECK-NEXT:    vblendmpd %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
+; CHECK-NEXT:    ## -- End function
   %y = load <8 x double>, <8 x double>* %yp, align 4
   %mask = fcmp olt <8 x double> %x, %y
   %max = select <8 x i1> %mask, <8 x double> %x, <8 x double> %x1
@@ -988,12 +1024,14 @@ define <4 x float> @test34(<4 x float> %x, <4 x float> %x1, <4 x float>* %yp) no
 ; KNL-NEXT:    vcmpltps (%rdi), %xmm0, %xmm2
 ; KNL-NEXT:    vblendvps %xmm2, %xmm0, %xmm1, %xmm0
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test34:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltps (%rdi), %xmm0, %k1
 ; SKX-NEXT:    vblendmps %xmm0, %xmm1, %xmm0 {%k1}
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
   %y = load <4 x float>, <4 x float>* %yp, align 4
   %mask = fcmp olt <4 x float> %x, %y
   %max = select <4 x i1> %mask, <4 x float> %x, <4 x float> %x1
@@ -1010,12 +1048,14 @@ define <8 x float> @test35(<8 x float> %x, <8 x float> %x1, <8 x float>* %yp) no
 ; KNL-NEXT:    vblendmps %zmm0, %zmm1, %zmm0 {%k1}
 ; KNL-NEXT:    ## kill: %YMM0<def> %YMM0<kill> %ZMM0<kill>
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test35:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltps (%rdi), %ymm0, %k1
 ; SKX-NEXT:    vblendmps %ymm0, %ymm1, %ymm0 {%k1}
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
 
   %y = load <8 x float>, <8 x float>* %yp, align 4
   %mask = fcmp ogt <8 x float> %y, %x
@@ -1029,6 +1069,7 @@ define <16 x float> @test36(<16 x float> %x, <16 x float> %x1, <16 x float>* %yp
 ; CHECK-NEXT:    vcmpltps (%rdi), %zmm0, %k1
 ; CHECK-NEXT:    vblendmps %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
+; CHECK-NEXT:    ## -- End function
   %y = load <16 x float>, <16 x float>* %yp, align 4
   %mask = fcmp olt <16 x float> %x, %y
   %max = select <16 x i1> %mask, <16 x float> %x, <16 x float> %x1
@@ -1041,6 +1082,7 @@ define <8 x double> @test37(<8 x double> %x, <8 x double> %x1, double* %ptr) nou
 ; CHECK-NEXT:    vcmpltpd (%rdi){1to8}, %zmm0, %k1
 ; CHECK-NEXT:    vblendmpd %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
+; CHECK-NEXT:    ## -- End function
 
   %a = load double, double* %ptr
   %v = insertelement <8 x double> undef, double %a, i32 0
@@ -1058,12 +1100,14 @@ define <4 x double> @test38(<4 x double> %x, <4 x double> %x1, double* %ptr) nou
 ; KNL-NEXT:    vcmpltpd %ymm2, %ymm0, %ymm2
 ; KNL-NEXT:    vblendvpd %ymm2, %ymm0, %ymm1, %ymm0
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test38:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltpd (%rdi){1to4}, %ymm0, %k1
 ; SKX-NEXT:    vblendmpd %ymm0, %ymm1, %ymm0 {%k1}
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
 
   %a = load double, double* %ptr
   %v = insertelement <4 x double> undef, double %a, i32 0
@@ -1081,12 +1125,14 @@ define <2 x double> @test39(<2 x double> %x, <2 x double> %x1, double* %ptr) nou
 ; KNL-NEXT:    vcmpltpd %xmm2, %xmm0, %xmm2
 ; KNL-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test39:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltpd (%rdi){1to2}, %xmm0, %k1
 ; SKX-NEXT:    vblendmpd %xmm0, %xmm1, %xmm0 {%k1}
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
 
   %a = load double, double* %ptr
   %v = insertelement <2 x double> undef, double %a, i32 0
@@ -1104,6 +1150,7 @@ define <16  x float> @test40(<16  x float> %x, <16  x float> %x1, float* %ptr) n
 ; CHECK-NEXT:    vcmpltps (%rdi){1to16}, %zmm0, %k1
 ; CHECK-NEXT:    vblendmps %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
+; CHECK-NEXT:    ## -- End function
 
   %a = load float, float* %ptr
   %v = insertelement <16  x float> undef, float %a, i32 0
@@ -1124,12 +1171,14 @@ define <8  x float> @test41(<8  x float> %x, <8  x float> %x1, float* %ptr) noun
 ; KNL-NEXT:    vblendmps %zmm0, %zmm1, %zmm0 {%k1}
 ; KNL-NEXT:    ## kill: %YMM0<def> %YMM0<kill> %ZMM0<kill>
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test41:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltps (%rdi){1to8}, %ymm0, %k1
 ; SKX-NEXT:    vblendmps %ymm0, %ymm1, %ymm0 {%k1}
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
 
   %a = load float, float* %ptr
   %v = insertelement <8  x float> undef, float %a, i32 0
@@ -1147,12 +1196,14 @@ define <4  x float> @test42(<4  x float> %x, <4  x float> %x1, float* %ptr) noun
 ; KNL-NEXT:    vcmpltps %xmm2, %xmm0, %xmm2
 ; KNL-NEXT:    vblendvps %xmm2, %xmm0, %xmm1, %xmm0
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test42:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltps (%rdi){1to4}, %xmm0, %k1
 ; SKX-NEXT:    vblendmps %xmm0, %xmm1, %xmm0 {%k1}
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
 
   %a = load float, float* %ptr
   %v = insertelement <4  x float> undef, float %a, i32 0
@@ -1172,6 +1223,7 @@ define <8 x double> @test43(<8 x double> %x, <8 x double> %x1, double* %ptr,<8 x
 ; KNL-NEXT:    vcmpltpd (%rdi){1to8}, %zmm0, %k1 {%k1}
 ; KNL-NEXT:    vblendmpd %zmm0, %zmm1, %zmm0 {%k1}
 ; KNL-NEXT:    retq
+; KNL-NEXT:    ## -- End function
 ;
 ; SKX-LABEL: test43:
 ; SKX:       ## BB#0:
@@ -1180,6 +1232,7 @@ define <8 x double> @test43(<8 x double> %x, <8 x double> %x1, double* %ptr,<8 x
 ; SKX-NEXT:    vcmpltpd (%rdi){1to8}, %zmm0, %k1 {%k1}
 ; SKX-NEXT:    vblendmpd %zmm0, %zmm1, %zmm0 {%k1}
 ; SKX-NEXT:    retq
+; SKX-NEXT:    ## -- End function
 
   %a = load double, double* %ptr
   %v = insertelement <8 x double> undef, double %a, i32 0

llvm/test/CodeGen/X86/avx512bwvl-intrinsics-upgrade.ll

@@ -2695,32 +2695,32 @@ declare <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16>, <16 x i16>, <32
 define <8 x i32> @test_cmp_b_256(<32 x i8> %a0, <32 x i8> %a1) {
 ; CHECK-LABEL: test_cmp_b_256:
 ; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x04]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpcmpleb %ymm0, %ymm1, %k0 ## encoding: [0x62,0xf3,0x75,0x28,0x3f,0xc0,0x02]
+; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
+; CHECK-NEXT:    vpcmpgtb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x28,0x64,0xc1]
+; CHECK-NEXT:    kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0]
+; CHECK-NEXT:    vmovd %eax, %xmm2 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd0]
+; CHECK-NEXT:    vpinsrd $1, %ecx, %xmm2, %xmm2 ## encoding: [0xc4,0xe3,0x69,0x22,0xd1,0x01]
+; CHECK-NEXT:    vpinsrd $2, %edx, %xmm2, %xmm2 ## encoding: [0xc4,0xe3,0x69,0x22,0xd2,0x02]
+; CHECK-NEXT:    kxnord %k0, %k0, %k0 ## encoding: [0xc4,0xe1,0xfd,0x46,0xc0]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpinsrd $3, %eax, %xmm2, %xmm2 ## encoding: [0xc4,0xe3,0x69,0x22,0xd0,0x03]
 ; CHECK-NEXT:    vpcmpeqb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
-; CHECK-NEXT:    kmovd %k0, %r8d ## encoding: [0xc5,0x7b,0x93,0xc0]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
 ; CHECK-NEXT:    vpcmpgtb %ymm0, %ymm1, %k0 ## encoding: [0x62,0xf1,0x75,0x28,0x64,0xc0]
 ; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
 ; CHECK-NEXT:    vpcmpleb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x02]
 ; CHECK-NEXT:    kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0]
-; CHECK-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x04]
+; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0]
+; CHECK-NEXT:    vmovd %eax, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc8]
-; CHECK-NEXT:    vpcmpleb %ymm0, %ymm1, %k0 ## encoding: [0x62,0xf3,0x75,0x28,0x3f,0xc0,0x02]
+; CHECK-NEXT:    vpunpckldq %xmm0, %xmm1, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf1,0x62,0xc0]
-; CHECK-NEXT:    kmovd %k0, %edi ## encoding: [0xc5,0xfb,0x93,0xf8]
+; CHECK-NEXT:    ## xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; CHECK-NEXT:    vpcmpgtb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x28,0x64,0xc1]
+; CHECK-NEXT:    vmovd %edx, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xca]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpunpcklqdq %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6c,0xc1]
-; CHECK-NEXT:    vmovd %esi, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
+; CHECK-NEXT:    ## xmm0 = xmm0[0],xmm1[0]
-; CHECK-NEXT:    vpinsrd $1, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x01]
+; CHECK-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x38,0xc2,0x01]
-; CHECK-NEXT:    vpinsrd $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x02]
-; CHECK-NEXT:    kxnord %k0, %k0, %k0 ## encoding: [0xc4,0xe1,0xfd,0x46,0xc0]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    vpinsrd $3, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x03]
-; CHECK-NEXT:    vmovd %ecx, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
-; CHECK-NEXT:    vmovd %r8d, %xmm2 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd0]
-; CHECK-NEXT:    vpunpckldq %xmm1, %xmm2, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
-; CHECK-NEXT:    ## xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; CHECK-NEXT:    vmovd %edx, %xmm2 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd2]
-; CHECK-NEXT:    vpunpcklqdq %xmm2, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
-; CHECK-NEXT:    ## xmm1 = xmm1[0],xmm2[0]
-; CHECK-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 -1)
   %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0
@@ -2750,23 +2750,23 @@ define <8 x i32> @test_mask_cmp_b_256(<32 x i8> %a0, <32 x i8> %a1, i32 %mask) {
 ; CHECK-NEXT:    vpcmpgtb %ymm0, %ymm1, %k0 {%k1} ## encoding: [0x62,0xf1,0x75,0x29,0x64,0xc0]
 ; CHECK-NEXT:    kmovd %k0, %r9d ## encoding: [0xc5,0x7b,0x93,0xc8]
 ; CHECK-NEXT:    vpcmpleb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x02]
-; CHECK-NEXT:    kmovd %k0, %r10d ## encoding: [0xc5,0x7b,0x93,0xd0]
-; CHECK-NEXT:    kxord %k0, %k0, %k0 ## encoding: [0xc4,0xe1,0xfd,0x47,0xc0]
-; CHECK-NEXT:    kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0]
-; CHECK-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    vpcmpleb %ymm0, %ymm1, %k0 {%k1} ## encoding: [0x62,0xf3,0x75,0x29,0x3f,0xc0,0x02]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vpcmpgtb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x64,0xc1]
 ; CHECK-NEXT:    kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0]
-; CHECK-NEXT:    vmovd %eax, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc0]
+; CHECK-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
-; CHECK-NEXT:    vpinsrd $1, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc1,0x01]
+; CHECK-NEXT:    kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0]
-; CHECK-NEXT:    vpinsrd $2, %edx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc2,0x02]
+; CHECK-NEXT:    vpcmpleb %ymm0, %ymm1, %k0 {%k1} ## encoding: [0x62,0xf3,0x75,0x29,0x3f,0xc0,0x02]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpcmpgtb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x64,0xc1]
+; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
+; CHECK-NEXT:    vmovd %esi, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
+; CHECK-NEXT:    vpinsrd $1, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x01]
+; CHECK-NEXT:    vpinsrd $2, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc1,0x02]
 ; CHECK-NEXT:    vpinsrd $3, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x03]
 ; CHECK-NEXT:    vmovd %r8d, %xmm1 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xc8]
 ; CHECK-NEXT:    vpinsrd $1, %r9d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xc9,0x01]
-; CHECK-NEXT:    vpinsrd $2, %r10d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xca,0x02]
+; CHECK-NEXT:    vpinsrd $2, %edx, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xca,0x02]
-; CHECK-NEXT:    vpinsrd $3, %esi, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xce,0x03]
+; CHECK-NEXT:    kxord %k0, %k0, %k0 ## encoding: [0xc4,0xe1,0xfd,0x47,0xc0]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpinsrd $3, %eax, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xc8,0x03]
 ; CHECK-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 %mask)
@@ -2793,32 +2793,32 @@ declare i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8>, <32 x i8>, i32, i32) noun
 define <8 x i32> @test_ucmp_b_256(<32 x i8> %a0, <32 x i8> %a1) {
 ; CHECK-LABEL: test_ucmp_b_256:
 ; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x04]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpcmpnltub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x05]
+; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
+; CHECK-NEXT:    vpcmpnleub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x06]
+; CHECK-NEXT:    kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0]
+; CHECK-NEXT:    vmovd %eax, %xmm2 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd0]
+; CHECK-NEXT:    vpinsrd $1, %ecx, %xmm2, %xmm2 ## encoding: [0xc4,0xe3,0x69,0x22,0xd1,0x01]
+; CHECK-NEXT:    vpinsrd $2, %edx, %xmm2, %xmm2 ## encoding: [0xc4,0xe3,0x69,0x22,0xd2,0x02]
+; CHECK-NEXT:    kxnord %k0, %k0, %k0 ## encoding: [0xc4,0xe1,0xfd,0x46,0xc0]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpinsrd $3, %eax, %xmm2, %xmm2 ## encoding: [0xc4,0xe3,0x69,0x22,0xd0,0x03]
 ; CHECK-NEXT:    vpcmpeqb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
-; CHECK-NEXT:    kmovd %k0, %r8d ## encoding: [0xc5,0x7b,0x93,0xc0]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
 ; CHECK-NEXT:    vpcmpltub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x01]
 ; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
 ; CHECK-NEXT:    vpcmpleub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x02]
 ; CHECK-NEXT:    kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0]
-; CHECK-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x04]
+; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0]
+; CHECK-NEXT:    vmovd %eax, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc8]
-; CHECK-NEXT:    vpcmpnltub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x05]
+; CHECK-NEXT:    vpunpckldq %xmm0, %xmm1, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf1,0x62,0xc0]
-; CHECK-NEXT:    kmovd %k0, %edi ## encoding: [0xc5,0xfb,0x93,0xf8]
+; CHECK-NEXT:    ## xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; CHECK-NEXT:    vpcmpnleub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x06]
+; CHECK-NEXT:    vmovd %edx, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xca]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpunpcklqdq %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6c,0xc1]
-; CHECK-NEXT:    vmovd %esi, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
+; CHECK-NEXT:    ## xmm0 = xmm0[0],xmm1[0]
-; CHECK-NEXT:    vpinsrd $1, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x01]
+; CHECK-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x38,0xc2,0x01]
-; CHECK-NEXT:    vpinsrd $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x02]
-; CHECK-NEXT:    kxnord %k0, %k0, %k0 ## encoding: [0xc4,0xe1,0xfd,0x46,0xc0]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    vpinsrd $3, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x03]
-; CHECK-NEXT:    vmovd %ecx, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
-; CHECK-NEXT:    vmovd %r8d, %xmm2 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd0]
-; CHECK-NEXT:    vpunpckldq %xmm1, %xmm2, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
-; CHECK-NEXT:    ## xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; CHECK-NEXT:    vmovd %edx, %xmm2 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd2]
-; CHECK-NEXT:    vpunpcklqdq %xmm2, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
-; CHECK-NEXT:    ## xmm1 = xmm1[0],xmm2[0]
-; CHECK-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 -1)
   %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0
@@ -2848,23 +2848,23 @@ define <8 x i32> @test_mask_ucmp_b_256(<32 x i8> %a0, <32 x i8> %a1, i32 %mask)
 ; CHECK-NEXT:    vpcmpltub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x01]
 ; CHECK-NEXT:    kmovd %k0, %r9d ## encoding: [0xc5,0x7b,0x93,0xc8]
 ; CHECK-NEXT:    vpcmpleub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x02]
-; CHECK-NEXT:    kmovd %k0, %r10d ## encoding: [0xc5,0x7b,0x93,0xd0]
-; CHECK-NEXT:    kxord %k0, %k0, %k0 ## encoding: [0xc4,0xe1,0xfd,0x47,0xc0]
-; CHECK-NEXT:    kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0]
-; CHECK-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    vpcmpnltub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x05]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vpcmpnleub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x06]
 ; CHECK-NEXT:    kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0]
-; CHECK-NEXT:    vmovd %eax, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc0]
+; CHECK-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
-; CHECK-NEXT:    vpinsrd $1, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc1,0x01]
+; CHECK-NEXT:    kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0]
-; CHECK-NEXT:    vpinsrd $2, %edx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc2,0x02]
+; CHECK-NEXT:    vpcmpnltub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x05]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpcmpnleub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x06]
+; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
+; CHECK-NEXT:    vmovd %esi, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
+; CHECK-NEXT:    vpinsrd $1, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x01]
+; CHECK-NEXT:    vpinsrd $2, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc1,0x02]
 ; CHECK-NEXT:    vpinsrd $3, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x03]
 ; CHECK-NEXT:    vmovd %r8d, %xmm1 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xc8]
 ; CHECK-NEXT:    vpinsrd $1, %r9d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xc9,0x01]
-; CHECK-NEXT:    vpinsrd $2, %r10d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xca,0x02]
+; CHECK-NEXT:    vpinsrd $2, %edx, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xca,0x02]
-; CHECK-NEXT:    vpinsrd $3, %esi, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xce,0x03]
+; CHECK-NEXT:    kxord %k0, %k0, %k0 ## encoding: [0xc4,0xe1,0xfd,0x47,0xc0]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpinsrd $3, %eax, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xc8,0x03]
 ; CHECK-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 %mask)

llvm/test/CodeGen/X86/bitcast-and-setcc-256.ll

@@ -453,10 +453,10 @@ define i32 @v32i8(<32 x i8> %a, <32 x i8> %b, <32 x i8> %c, <32 x i8> %d) {
 ; SSE2-SSSE3-NEXT:    pcmpgtb %xmm2, %xmm0
 ; SSE2-SSSE3-NEXT:    pcmpgtb %xmm3, %xmm1
 ; SSE2-SSSE3-NEXT:    pcmpgtb %xmm6, %xmm4
-; SSE2-SSSE3-NEXT:    pand %xmm0, %xmm4
 ; SSE2-SSSE3-NEXT:    pcmpgtb %xmm7, %xmm5
 ; SSE2-SSSE3-NEXT:    pand %xmm1, %xmm5
 ; SSE2-SSSE3-NEXT:    movdqa %xmm5, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    pand %xmm0, %xmm4
 ; SSE2-SSSE3-NEXT:    movdqa %xmm4, -{{[0-9]+}}(%rsp)
 ; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
 ; SSE2-SSSE3-NEXT:    andb $1, %al

llvm/test/CodeGen/X86/extractelement-legalization-store-ordering.ll

@@ -1,3 +1,4 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple i386-apple-darwin -mcpu=yonah | FileCheck %s
 
 target datalayout = "e-m:o-p:32:32-f64:32:64-f80:128-n8:16:32-S128"
@@ -6,31 +7,32 @@ target datalayout = "e-m:o-p:32:32-f64:32:64-f80:128-n8:16:32-S128"
 ; into loads, off the stack or a previous store.
 ; Be very explicit about the ordering/stack offsets.
 
-; CHECK-LABEL: test_extractelement_legalization_storereuse:
-; CHECK:      # BB#0
-; CHECK-NEXT: pushl %ebx
-; CHECK-NEXT: pushl %edi
-; CHECK-NEXT: pushl %esi
-; CHECK-NEXT: movl 16(%esp), %eax
-; CHECK-NEXT: movl 24(%esp), %ecx
-; CHECK-NEXT: movl 20(%esp), %edx
-; CHECK-NEXT: paddd (%edx), %xmm0
-; CHECK-NEXT: movdqa %xmm0, (%edx)
-; CHECK-NEXT:	movl	(%edx), %esi
-; CHECK-NEXT:	movl	4(%edx), %edi
-; CHECK-NEXT:	shll	$4, %ecx
-; CHECK-NEXT:	movl	8(%edx), %ebx
-; CHECK-NEXT:	movl	12(%edx), %edx
-; CHECK-NEXT: movl %esi, 12(%eax,%ecx)
-; CHECK-NEXT: movl %edi, (%eax,%ecx)
-; CHECK-NEXT: movl %ebx, 8(%eax,%ecx)
-; CHECK-NEXT: movl %edx, 4(%eax,%ecx)
-; CHECK-NEXT: popl %esi
-; CHECK-NEXT: popl %edi
-; CHECK-NEXT: popl %ebx
-; CHECK-NEXT: retl
 
 define void @test_extractelement_legalization_storereuse(<4 x i32> %a, i32* nocapture %x, i32* nocapture readonly %y, i32 %i) #0 {
+; CHECK-LABEL:   _test_extractelement_legalization_storereuse: ## @test_extractelement_legalization_storereuse 
+; CHECK:       ## BB#0: ## %entry
+; CHECK-NEXT:    pushl %ebx
+; CHECK-NEXT:    pushl %edi
+; CHECK-NEXT:    pushl %esi
+; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; CHECK-NEXT:    paddd (%ecx), %xmm0
+; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; CHECK-NEXT:    movdqa %xmm0, (%ecx)
+; CHECK-NEXT:    movl (%ecx), %esi
+; CHECK-NEXT:    movl 4(%ecx), %edi
+; CHECK-NEXT:    shll $4, %edx
+; CHECK-NEXT:    movl 8(%ecx), %ebx
+; CHECK-NEXT:    movl 12(%ecx), %ecx
+; CHECK-NEXT:    movl %esi, 12(%eax,%edx)
+; CHECK-NEXT:    movl %edi, (%eax,%edx)
+; CHECK-NEXT:    movl %ebx, 8(%eax,%edx)
+; CHECK-NEXT:    movl %ecx, 4(%eax,%edx)
+; CHECK-NEXT:    popl %esi
+; CHECK-NEXT:    popl %edi
+; CHECK-NEXT:    popl %ebx
+; CHECK-NEXT:    retl
+; CHECK-NEXT:    ## -- End function
 entry:
   %0 = bitcast i32* %y to <4 x i32>*
   %1 = load <4 x i32>, <4 x i32>* %0, align 16

llvm/test/CodeGen/X86/fp128-i128.ll

@@ -50,8 +50,8 @@ define void @TestUnionLD1(fp128 %s, i64 %n) #0 {
 ; CHECK-NEXT:    andq %rdi, %rcx
 ; CHECK-NEXT:    movabsq $-281474976710656, %rdx # imm = 0xFFFF000000000000
 ; CHECK-NEXT:    andq -{{[0-9]+}}(%rsp), %rdx
-; CHECK-NEXT:    orq %rcx, %rdx
 ; CHECK-NEXT:    movq %rax, -{{[0-9]+}}(%rsp)
+; CHECK-NEXT:    orq %rcx, %rdx
 ; CHECK-NEXT:    movq %rdx, -{{[0-9]+}}(%rsp)
 ; CHECK-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
 ; CHECK-NEXT:    jmp foo # TAILCALL

llvm/test/CodeGen/X86/gather-addresses.ll

@@ -16,11 +16,10 @@
 ; LIN: sarq    $32, %r[[REG2]]
 ; LIN: movslq	%e[[REG4]], %r[[REG3:.+]]
 ; LIN: sarq    $32, %r[[REG4]]
-; LIN: movsd	(%rdi,%r[[REG1]],8), %xmm0
+; LIN: movsd	(%rdi,%rsi,8), %xmm1
-; LIN: movhpd	(%rdi,%r[[REG2]],8), %xmm0
+; LIN: movhpd   (%rdi,%rax,8), %xmm1
-; LIN: movsd	(%rdi,%r[[REG3]],8), %xmm1
+; LIN: movdqa   (%rsi), %xmm0 
-; LIN: movhpd	(%rdi,%r[[REG4]],8), %xmm1
+; LIN: movq     %rdi, %xmm1  
-
 ; WIN: movdqa	(%rdx), %xmm0
 ; WIN: pand 	(%r8), %xmm0
 ; WIN: pextrq	$1, %xmm0, %r[[REG4:.+]]
@@ -29,10 +28,10 @@
 ; WIN: sarq    $32, %r[[REG2]]
 ; WIN: movslq	%e[[REG4]], %r[[REG3:.+]]
 ; WIN: sarq    $32, %r[[REG4]]
-; WIN: movsd	(%rcx,%r[[REG1]],8), %xmm0
+; WIN: movsd    (%rcx,%r9,8), %xmm1
-; WIN: movhpd	(%rcx,%r[[REG2]],8), %xmm0
+; WIN: movhpd   (%rcx,%rax,8), %xmm1
-; WIN: movsd	(%rcx,%r[[REG3]],8), %xmm1
+; WIN: movdqa   (%rdx), %xmm0
-; WIN: movhpd	(%rcx,%r[[REG4]],8), %xmm1
+; WIN: movq     %rdx, %xmm1 
 
 define <4 x double> @foo(double* %p, <4 x i32>* %i, <4 x i32>* %h) nounwind {
   %a = load <4 x i32>, <4 x i32>* %i

llvm/test/CodeGen/X86/illegal-bitfield-loadstore.ll

@@ -112,23 +112,23 @@ define void @i56_and_or(i56* %a) {
 define void @i56_insert_bit(i56* %a, i1 zeroext %bit) {
 ; CHECK-LABEL: i56_insert_bit:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    movzbl %sil, %eax
+; CHECK-NEXT:    movzwl 4(%rdi), %eax
-; CHECK-NEXT:    movzwl 4(%rdi), %ecx
+; CHECK-NEXT:    movzbl 6(%rdi), %ecx
-; CHECK-NEXT:    movzbl 6(%rdi), %edx
+; CHECK-NEXT:    movl (%rdi), %edx
-; CHECK-NEXT:    movl (%rdi), %esi
+; CHECK-NEXT:    movb %cl, 6(%rdi)
-; CHECK-NEXT:    movb %dl, 6(%rdi)
+; CHECK-NEXT:    movzbl %sil, %esi
-; CHECK-NEXT:    # kill: %EDX<def> %EDX<kill> %RDX<kill> %RDX<def>
+; CHECK-NEXT:    # kill: %ECX<def> %ECX<kill> %RCX<kill> %RCX<def>
-; CHECK-NEXT:    shll $16, %edx
+; CHECK-NEXT:    shll $16, %ecx
-; CHECK-NEXT:    orl %ecx, %edx
+; CHECK-NEXT:    orl %eax, %ecx
-; CHECK-NEXT:    shlq $32, %rdx
+; CHECK-NEXT:    shlq $32, %rcx
-; CHECK-NEXT:    orq %rdx, %rsi
+; CHECK-NEXT:    orq %rcx, %rdx
-; CHECK-NEXT:    shlq $13, %rax
+; CHECK-NEXT:    shlq $13, %rsi
-; CHECK-NEXT:    movabsq $72057594037919743, %rcx # imm = 0xFFFFFFFFFFDFFF
+; CHECK-NEXT:    movabsq $72057594037919743, %rax # imm = 0xFFFFFFFFFFDFFF
-; CHECK-NEXT:    andq %rsi, %rcx
+; CHECK-NEXT:    andq %rdx, %rax
-; CHECK-NEXT:    orq %rax, %rcx
+; CHECK-NEXT:    orq %rsi, %rax
-; CHECK-NEXT:    movl %ecx, (%rdi)
+; CHECK-NEXT:    movl %eax, (%rdi)
-; CHECK-NEXT:    shrq $32, %rcx
+; CHECK-NEXT:    shrq $32, %rax
-; CHECK-NEXT:    movw %cx, 4(%rdi)
+; CHECK-NEXT:    movw %ax, 4(%rdi)
 ; CHECK-NEXT:    retq
   %extbit = zext i1 %bit to i56
   %b = load i56, i56* %a, align 1

llvm/test/CodeGen/X86/mul-constant-i32.ll

@@ -17,7 +17,7 @@ define i32 @test_mul_by_1(i32 %x) {
 ; X64-HSW-LABEL: test_mul_by_1:
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    movl %edi, %eax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_1:
 ; X64-JAG:       # BB#0:
@@ -32,7 +32,7 @@ define i32 @test_mul_by_1(i32 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_1:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    movl %edi, %eax # sched: [1:0.25]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_1:
 ; JAG-NOOPT:       # BB#0:
@@ -63,7 +63,7 @@ define i32 @test_mul_by_2(i32 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; X64-HSW-NEXT:    leal (%rdi,%rdi), %eax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_2:
 ; X64-JAG:       # BB#0:
@@ -81,7 +81,7 @@ define i32 @test_mul_by_2(i32 %x) {
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; HSW-NOOPT-NEXT:    leal (%rdi,%rdi), %eax # sched: [1:0.50]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_2:
 ; JAG-NOOPT:       # BB#0:
@@ -114,7 +114,7 @@ define i32 @test_mul_by_3(i32 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; X64-HSW-NEXT:    leal (%rdi,%rdi,2), %eax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_3:
 ; X64-JAG:       # BB#0:
@@ -131,7 +131,7 @@ define i32 @test_mul_by_3(i32 %x) {
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; HSW-NOOPT-NEXT:    leal (%rdi,%rdi,2), %eax # sched: [1:0.50]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_3:
 ; JAG-NOOPT:       # BB#0:
@@ -165,7 +165,7 @@ define i32 @test_mul_by_4(i32 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; X64-HSW-NEXT:    leal (,%rdi,4), %eax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_4:
 ; X64-JAG:       # BB#0:
@@ -183,7 +183,7 @@ define i32 @test_mul_by_4(i32 %x) {
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; HSW-NOOPT-NEXT:    leal (,%rdi,4), %eax # sched: [1:0.50]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_4:
 ; JAG-NOOPT:       # BB#0:
@@ -216,7 +216,7 @@ define i32 @test_mul_by_5(i32 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; X64-HSW-NEXT:    leal (%rdi,%rdi,4), %eax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_5:
 ; X64-JAG:       # BB#0:
@@ -233,7 +233,7 @@ define i32 @test_mul_by_5(i32 %x) {
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; HSW-NOOPT-NEXT:    leal (%rdi,%rdi,4), %eax # sched: [1:0.50]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_5:
 ; JAG-NOOPT:       # BB#0:
@@ -269,7 +269,7 @@ define i32 @test_mul_by_6(i32 %x) {
 ; X64-HSW-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; X64-HSW-NEXT:    addl %edi, %edi # sched: [1:0.25]
 ; X64-HSW-NEXT:    leal (%rdi,%rdi,2), %eax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_6:
 ; X64-JAG:       # BB#0:
@@ -285,8 +285,8 @@ define i32 @test_mul_by_6(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_6:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    imull $6, %edi, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull $6, %edi, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_6:
 ; JAG-NOOPT:       # BB#0:
@@ -321,7 +321,7 @@ define i32 @test_mul_by_7(i32 %x) {
 ; X64-HSW-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; X64-HSW-NEXT:    leal (,%rdi,8), %eax # sched: [1:0.50]
 ; X64-HSW-NEXT:    subl %edi, %eax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_7:
 ; X64-JAG:       # BB#0:
@@ -337,8 +337,8 @@ define i32 @test_mul_by_7(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_7:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    imull $7, %edi, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull $7, %edi, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_7:
 ; JAG-NOOPT:       # BB#0:
@@ -371,7 +371,7 @@ define i32 @test_mul_by_8(i32 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; X64-HSW-NEXT:    leal (,%rdi,8), %eax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_8:
 ; X64-JAG:       # BB#0:
@@ -389,7 +389,7 @@ define i32 @test_mul_by_8(i32 %x) {
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; HSW-NOOPT-NEXT:    leal (,%rdi,8), %eax # sched: [1:0.50]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_8:
 ; JAG-NOOPT:       # BB#0:
@@ -422,7 +422,7 @@ define i32 @test_mul_by_9(i32 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; X64-HSW-NEXT:    leal (%rdi,%rdi,8), %eax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_9:
 ; X64-JAG:       # BB#0:
@@ -439,7 +439,7 @@ define i32 @test_mul_by_9(i32 %x) {
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; HSW-NOOPT-NEXT:    leal (%rdi,%rdi,8), %eax # sched: [1:0.50]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_9:
 ; JAG-NOOPT:       # BB#0:
@@ -475,7 +475,7 @@ define i32 @test_mul_by_10(i32 %x) {
 ; X64-HSW-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; X64-HSW-NEXT:    addl %edi, %edi # sched: [1:0.25]
 ; X64-HSW-NEXT:    leal (%rdi,%rdi,4), %eax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_10:
 ; X64-JAG:       # BB#0:
@@ -491,8 +491,8 @@ define i32 @test_mul_by_10(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_10:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    imull $10, %edi, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull $10, %edi, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_10:
 ; JAG-NOOPT:       # BB#0:
@@ -527,7 +527,7 @@ define i32 @test_mul_by_11(i32 %x) {
 ; X64-HSW-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; X64-HSW-NEXT:    leal (%rdi,%rdi,4), %eax # sched: [1:0.50]
 ; X64-HSW-NEXT:    leal (%rdi,%rax,2), %eax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_11:
 ; X64-JAG:       # BB#0:
@@ -543,8 +543,8 @@ define i32 @test_mul_by_11(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_11:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    imull $11, %edi, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull $11, %edi, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_11:
 ; JAG-NOOPT:       # BB#0:
@@ -575,9 +575,9 @@ define i32 @test_mul_by_12(i32 %x) {
 ; X64-HSW-LABEL: test_mul_by_12:
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
-; X64-HSW-NEXT:    shll $2, %edi # sched: [1:0.50]
+; X64-HSW-NEXT:    shll $2, %edi # sched: [1:1.00]
 ; X64-HSW-NEXT:    leal (%rdi,%rdi,2), %eax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_12:
 ; X64-JAG:       # BB#0:
@@ -593,8 +593,8 @@ define i32 @test_mul_by_12(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_12:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    imull $12, %edi, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull $12, %edi, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_12:
 ; JAG-NOOPT:       # BB#0:
@@ -629,7 +629,7 @@ define i32 @test_mul_by_13(i32 %x) {
 ; X64-HSW-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; X64-HSW-NEXT:    leal (%rdi,%rdi,2), %eax # sched: [1:0.50]
 ; X64-HSW-NEXT:    leal (%rdi,%rax,4), %eax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_13:
 ; X64-JAG:       # BB#0:
@@ -645,8 +645,8 @@ define i32 @test_mul_by_13(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_13:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    imull $13, %edi, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull $13, %edi, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_13:
 ; JAG-NOOPT:       # BB#0:
@@ -681,7 +681,7 @@ define i32 @test_mul_by_14(i32 %x) {
 ; X64-HSW-NEXT:    leal (%rdi,%rdi,2), %eax # sched: [1:0.50]
 ; X64-HSW-NEXT:    leal (%rdi,%rax,4), %eax # sched: [1:0.50]
 ; X64-HSW-NEXT:    addl %edi, %eax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_14:
 ; X64-JAG:       # BB#0:
@@ -698,8 +698,8 @@ define i32 @test_mul_by_14(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_14:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    imull $14, %edi, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull $14, %edi, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_14:
 ; JAG-NOOPT:       # BB#0:
@@ -732,7 +732,7 @@ define i32 @test_mul_by_15(i32 %x) {
 ; X64-HSW-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; X64-HSW-NEXT:    leal (%rdi,%rdi,4), %eax # sched: [1:0.50]
 ; X64-HSW-NEXT:    leal (%rax,%rax,2), %eax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_15:
 ; X64-JAG:       # BB#0:
@@ -748,8 +748,8 @@ define i32 @test_mul_by_15(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_15:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    imull $15, %edi, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull $15, %edi, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_15:
 ; JAG-NOOPT:       # BB#0:
@@ -780,9 +780,9 @@ define i32 @test_mul_by_16(i32 %x) {
 ;
 ; X64-HSW-LABEL: test_mul_by_16:
 ; X64-HSW:       # BB#0:
-; X64-HSW-NEXT:    shll $4, %edi # sched: [1:0.50]
+; X64-HSW-NEXT:    shll $4, %edi # sched: [1:1.00]
 ; X64-HSW-NEXT:    movl %edi, %eax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_16:
 ; X64-JAG:       # BB#0:
@@ -798,9 +798,9 @@ define i32 @test_mul_by_16(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_16:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    shll $4, %edi # sched: [1:0.50]
+; HSW-NOOPT-NEXT:    shll $4, %edi # sched: [1:1.00]
 ; HSW-NOOPT-NEXT:    movl %edi, %eax # sched: [1:0.25]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_16:
 ; JAG-NOOPT:       # BB#0:
@@ -836,9 +836,9 @@ define i32 @test_mul_by_17(i32 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; X64-HSW-NEXT:    movl %edi, %eax # sched: [1:0.25]
-; X64-HSW-NEXT:    shll $4, %eax # sched: [1:0.50]
+; X64-HSW-NEXT:    shll $4, %eax # sched: [1:1.00]
 ; X64-HSW-NEXT:    leal (%rax,%rdi), %eax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_17:
 ; X64-JAG:       # BB#0:
@@ -855,8 +855,8 @@ define i32 @test_mul_by_17(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_17:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    imull $17, %edi, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull $17, %edi, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_17:
 ; JAG-NOOPT:       # BB#0:
@@ -892,7 +892,7 @@ define i32 @test_mul_by_18(i32 %x) {
 ; X64-HSW-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; X64-HSW-NEXT:    addl %edi, %edi # sched: [1:0.25]
 ; X64-HSW-NEXT:    leal (%rdi,%rdi,8), %eax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_18:
 ; X64-JAG:       # BB#0:
@@ -908,8 +908,8 @@ define i32 @test_mul_by_18(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_18:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    imull $18, %edi, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull $18, %edi, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_18:
 ; JAG-NOOPT:       # BB#0:
@@ -944,9 +944,9 @@ define i32 @test_mul_by_19(i32 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; X64-HSW-NEXT:    leal (%rdi,%rdi,4), %eax # sched: [1:0.50]
-; X64-HSW-NEXT:    shll $2, %eax # sched: [1:0.50]
+; X64-HSW-NEXT:    shll $2, %eax # sched: [1:1.00]
 ; X64-HSW-NEXT:    subl %edi, %eax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_19:
 ; X64-JAG:       # BB#0:
@@ -963,8 +963,8 @@ define i32 @test_mul_by_19(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_19:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    imull $19, %edi, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull $19, %edi, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_19:
 ; JAG-NOOPT:       # BB#0:
@@ -995,9 +995,9 @@ define i32 @test_mul_by_20(i32 %x) {
 ; X64-HSW-LABEL: test_mul_by_20:
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
-; X64-HSW-NEXT:    shll $2, %edi # sched: [1:0.50]
+; X64-HSW-NEXT:    shll $2, %edi # sched: [1:1.00]
 ; X64-HSW-NEXT:    leal (%rdi,%rdi,4), %eax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_20:
 ; X64-JAG:       # BB#0:
@@ -1013,8 +1013,8 @@ define i32 @test_mul_by_20(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_20:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    imull $20, %edi, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull $20, %edi, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_20:
 ; JAG-NOOPT:       # BB#0:
@@ -1049,7 +1049,7 @@ define i32 @test_mul_by_21(i32 %x) {
 ; X64-HSW-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; X64-HSW-NEXT:    leal (%rdi,%rdi,4), %eax # sched: [1:0.50]
 ; X64-HSW-NEXT:    leal (%rdi,%rax,4), %eax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_21:
 ; X64-JAG:       # BB#0:
@@ -1065,8 +1065,8 @@ define i32 @test_mul_by_21(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_21:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    imull $21, %edi, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull $21, %edi, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_21:
 ; JAG-NOOPT:       # BB#0:
@@ -1101,7 +1101,7 @@ define i32 @test_mul_by_22(i32 %x) {
 ; X64-HSW-NEXT:    leal (%rdi,%rdi,4), %eax # sched: [1:0.50]
 ; X64-HSW-NEXT:    leal (%rdi,%rax,4), %eax # sched: [1:0.50]
 ; X64-HSW-NEXT:    addl %edi, %eax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_22:
 ; X64-JAG:       # BB#0:
@@ -1118,8 +1118,8 @@ define i32 @test_mul_by_22(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_22:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    imull $22, %edi, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull $22, %edi, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_22:
 ; JAG-NOOPT:       # BB#0:
@@ -1152,9 +1152,9 @@ define i32 @test_mul_by_23(i32 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; X64-HSW-NEXT:    leal (%rdi,%rdi,2), %eax # sched: [1:0.50]
-; X64-HSW-NEXT:    shll $3, %eax # sched: [1:0.50]
+; X64-HSW-NEXT:    shll $3, %eax # sched: [1:1.00]
 ; X64-HSW-NEXT:    subl %edi, %eax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_23:
 ; X64-JAG:       # BB#0:
@@ -1171,8 +1171,8 @@ define i32 @test_mul_by_23(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_23:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    imull $23, %edi, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull $23, %edi, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_23:
 ; JAG-NOOPT:       # BB#0:
@@ -1203,9 +1203,9 @@ define i32 @test_mul_by_24(i32 %x) {
 ; X64-HSW-LABEL: test_mul_by_24:
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
-; X64-HSW-NEXT:    shll $3, %edi # sched: [1:0.50]
+; X64-HSW-NEXT:    shll $3, %edi # sched: [1:1.00]
 ; X64-HSW-NEXT:    leal (%rdi,%rdi,2), %eax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_24:
 ; X64-JAG:       # BB#0:
@@ -1221,8 +1221,8 @@ define i32 @test_mul_by_24(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_24:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    imull $24, %edi, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull $24, %edi, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_24:
 ; JAG-NOOPT:       # BB#0:
@@ -1257,7 +1257,7 @@ define i32 @test_mul_by_25(i32 %x) {
 ; X64-HSW-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; X64-HSW-NEXT:    leal (%rdi,%rdi,4), %eax # sched: [1:0.50]
 ; X64-HSW-NEXT:    leal (%rax,%rax,4), %eax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_25:
 ; X64-JAG:       # BB#0:
@@ -1273,8 +1273,8 @@ define i32 @test_mul_by_25(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_25:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    imull $25, %edi, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull $25, %edi, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_25:
 ; JAG-NOOPT:       # BB#0:
@@ -1311,7 +1311,7 @@ define i32 @test_mul_by_26(i32 %x) {
 ; X64-HSW-NEXT:    leal (%rdi,%rdi,8), %eax # sched: [1:0.50]
 ; X64-HSW-NEXT:    leal (%rax,%rax,2), %eax # sched: [1:0.50]
 ; X64-HSW-NEXT:    subl %edi, %eax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_26:
 ; X64-JAG:       # BB#0:
@@ -1328,8 +1328,8 @@ define i32 @test_mul_by_26(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_26:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    imull $26, %edi, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull $26, %edi, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_26:
 ; JAG-NOOPT:       # BB#0:
@@ -1362,7 +1362,7 @@ define i32 @test_mul_by_27(i32 %x) {
 ; X64-HSW-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; X64-HSW-NEXT:    leal (%rdi,%rdi,8), %eax # sched: [1:0.50]
 ; X64-HSW-NEXT:    leal (%rax,%rax,2), %eax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_27:
 ; X64-JAG:       # BB#0:
@@ -1378,8 +1378,8 @@ define i32 @test_mul_by_27(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_27:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    imull $27, %edi, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull $27, %edi, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_27:
 ; JAG-NOOPT:       # BB#0:
@@ -1416,7 +1416,7 @@ define i32 @test_mul_by_28(i32 %x) {
 ; X64-HSW-NEXT:    leal (%rdi,%rdi,8), %eax # sched: [1:0.50]
 ; X64-HSW-NEXT:    leal (%rax,%rax,2), %eax # sched: [1:0.50]
 ; X64-HSW-NEXT:    addl %edi, %eax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_28:
 ; X64-JAG:       # BB#0:
@@ -1433,8 +1433,8 @@ define i32 @test_mul_by_28(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_28:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    imull $28, %edi, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull $28, %edi, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_28:
 ; JAG-NOOPT:       # BB#0:
@@ -1471,7 +1471,7 @@ define i32 @test_mul_by_29(i32 %x) {
 ; X64-HSW-NEXT:    leal (%rax,%rax,2), %eax # sched: [1:0.50]
 ; X64-HSW-NEXT:    addl %edi, %eax # sched: [1:0.25]
 ; X64-HSW-NEXT:    addl %edi, %eax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_29:
 ; X64-JAG:       # BB#0:
@@ -1489,8 +1489,8 @@ define i32 @test_mul_by_29(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_29:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    imull $29, %edi, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull $29, %edi, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_29:
 ; JAG-NOOPT:       # BB#0:
@@ -1523,10 +1523,10 @@ define i32 @test_mul_by_30(i32 %x) {
 ; X64-HSW-LABEL: test_mul_by_30:
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    movl %edi, %eax # sched: [1:0.25]
-; X64-HSW-NEXT:    shll $5, %eax # sched: [1:0.50]
+; X64-HSW-NEXT:    shll $5, %eax # sched: [1:1.00]
 ; X64-HSW-NEXT:    subl %edi, %eax # sched: [1:0.25]
 ; X64-HSW-NEXT:    subl %edi, %eax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_30:
 ; X64-JAG:       # BB#0:
@@ -1543,8 +1543,8 @@ define i32 @test_mul_by_30(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_30:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    imull $30, %edi, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull $30, %edi, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_30:
 ; JAG-NOOPT:       # BB#0:
@@ -1576,9 +1576,9 @@ define i32 @test_mul_by_31(i32 %x) {
 ; X64-HSW-LABEL: test_mul_by_31:
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    movl %edi, %eax # sched: [1:0.25]
-; X64-HSW-NEXT:    shll $5, %eax # sched: [1:0.50]
+; X64-HSW-NEXT:    shll $5, %eax # sched: [1:1.00]
 ; X64-HSW-NEXT:    subl %edi, %eax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_31:
 ; X64-JAG:       # BB#0:
@@ -1594,8 +1594,8 @@ define i32 @test_mul_by_31(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_31:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    imull $31, %edi, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull $31, %edi, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_31:
 ; JAG-NOOPT:       # BB#0:
@@ -1626,9 +1626,9 @@ define i32 @test_mul_by_32(i32 %x) {
 ;
 ; X64-HSW-LABEL: test_mul_by_32:
 ; X64-HSW:       # BB#0:
-; X64-HSW-NEXT:    shll $5, %edi # sched: [1:0.50]
+; X64-HSW-NEXT:    shll $5, %edi # sched: [1:1.00]
 ; X64-HSW-NEXT:    movl %edi, %eax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_32:
 ; X64-JAG:       # BB#0:
@@ -1644,9 +1644,9 @@ define i32 @test_mul_by_32(i32 %x) {
 ;
 ; HSW-NOOPT-LABEL: test_mul_by_32:
 ; HSW-NOOPT:       # BB#0:
-; HSW-NOOPT-NEXT:    shll $5, %edi # sched: [1:0.50]
+; HSW-NOOPT-NEXT:    shll $5, %edi # sched: [1:1.00]
 ; HSW-NOOPT-NEXT:    movl %edi, %eax # sched: [1:0.25]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_32:
 ; JAG-NOOPT:       # BB#0:
@@ -1686,8 +1686,8 @@ define i32 @test_mul_spec(i32 %x) nounwind {
 ; X64-HSW-NEXT:    addl $42, %ecx # sched: [1:0.25]
 ; X64-HSW-NEXT:    leal (%rdi,%rdi,4), %eax # sched: [1:0.50]
 ; X64-HSW-NEXT:    addl $2, %eax # sched: [1:0.25]
-; X64-HSW-NEXT:    imull %ecx, %eax # sched: [4:1.00]
+; X64-HSW-NEXT:    imull %ecx, %eax # sched: [3:1.00]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_spec:
 ; X64-JAG:       # BB#0:
@@ -1712,8 +1712,8 @@ define i32 @test_mul_spec(i32 %x) nounwind {
 ; HSW-NOOPT-NEXT:    addl $42, %ecx # sched: [1:0.25]
 ; HSW-NOOPT-NEXT:    leal (%rdi,%rdi,4), %eax # sched: [1:0.50]
 ; HSW-NOOPT-NEXT:    addl $2, %eax # sched: [1:0.25]
-; HSW-NOOPT-NEXT:    imull %ecx, %eax # sched: [4:1.00]
+; HSW-NOOPT-NEXT:    imull %ecx, %eax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_spec:
 ; JAG-NOOPT:       # BB#0:

llvm/test/CodeGen/X86/mul-constant-i64.ll

@@ -18,7 +18,7 @@ define i64 @test_mul_by_1(i64 %x) nounwind {
 ; X64-HSW-LABEL: test_mul_by_1:
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    movq %rdi, %rax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_1:
 ; X64-JAG:       # BB#0:
@@ -34,7 +34,7 @@ define i64 @test_mul_by_1(i64 %x) nounwind {
 ; HSW-NOOPT-LABEL: test_mul_by_1:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    movq %rdi, %rax # sched: [1:0.25]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_1:
 ; JAG-NOOPT:       # BB#0:
@@ -66,7 +66,7 @@ define i64 @test_mul_by_2(i64 %x) {
 ; X64-HSW-LABEL: test_mul_by_2:
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    leaq (%rdi,%rdi), %rax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_2:
 ; X64-JAG:       # BB#0:
@@ -84,7 +84,7 @@ define i64 @test_mul_by_2(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_2:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    leaq (%rdi,%rdi), %rax # sched: [1:0.50]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_2:
 ; JAG-NOOPT:       # BB#0:
@@ -116,7 +116,7 @@ define i64 @test_mul_by_3(i64 %x) {
 ; X64-HSW-LABEL: test_mul_by_3:
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    leaq (%rdi,%rdi,2), %rax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_3:
 ; X64-JAG:       # BB#0:
@@ -134,7 +134,7 @@ define i64 @test_mul_by_3(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_3:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    leaq (%rdi,%rdi,2), %rax # sched: [1:0.50]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_3:
 ; JAG-NOOPT:       # BB#0:
@@ -166,7 +166,7 @@ define i64 @test_mul_by_4(i64 %x) {
 ; X64-HSW-LABEL: test_mul_by_4:
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    leaq (,%rdi,4), %rax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_4:
 ; X64-JAG:       # BB#0:
@@ -184,7 +184,7 @@ define i64 @test_mul_by_4(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_4:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    leaq (,%rdi,4), %rax # sched: [1:0.50]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_4:
 ; JAG-NOOPT:       # BB#0:
@@ -216,7 +216,7 @@ define i64 @test_mul_by_5(i64 %x) {
 ; X64-HSW-LABEL: test_mul_by_5:
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    leaq (%rdi,%rdi,4), %rax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_5:
 ; X64-JAG:       # BB#0:
@@ -234,7 +234,7 @@ define i64 @test_mul_by_5(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_5:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    leaq (%rdi,%rdi,4), %rax # sched: [1:0.50]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_5:
 ; JAG-NOOPT:       # BB#0:
@@ -268,7 +268,7 @@ define i64 @test_mul_by_6(i64 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    addq %rdi, %rdi # sched: [1:0.25]
 ; X64-HSW-NEXT:    leaq (%rdi,%rdi,2), %rax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_6:
 ; X64-JAG:       # BB#0:
@@ -287,7 +287,7 @@ define i64 @test_mul_by_6(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_6:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    imulq $6, %rdi, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_6:
 ; JAG-NOOPT:       # BB#0:
@@ -323,7 +323,7 @@ define i64 @test_mul_by_7(i64 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    leaq (,%rdi,8), %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    subq %rdi, %rax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_7:
 ; X64-JAG:       # BB#0:
@@ -342,7 +342,7 @@ define i64 @test_mul_by_7(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_7:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    imulq $7, %rdi, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_7:
 ; JAG-NOOPT:       # BB#0:
@@ -375,7 +375,7 @@ define i64 @test_mul_by_8(i64 %x) {
 ; X64-HSW-LABEL: test_mul_by_8:
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    leaq (,%rdi,8), %rax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_8:
 ; X64-JAG:       # BB#0:
@@ -393,7 +393,7 @@ define i64 @test_mul_by_8(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_8:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    leaq (,%rdi,8), %rax # sched: [1:0.50]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_8:
 ; JAG-NOOPT:       # BB#0:
@@ -425,7 +425,7 @@ define i64 @test_mul_by_9(i64 %x) {
 ; X64-HSW-LABEL: test_mul_by_9:
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    leaq (%rdi,%rdi,8), %rax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_9:
 ; X64-JAG:       # BB#0:
@@ -443,7 +443,7 @@ define i64 @test_mul_by_9(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_9:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    leaq (%rdi,%rdi,8), %rax # sched: [1:0.50]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_9:
 ; JAG-NOOPT:       # BB#0:
@@ -477,7 +477,7 @@ define i64 @test_mul_by_10(i64 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    addq %rdi, %rdi # sched: [1:0.25]
 ; X64-HSW-NEXT:    leaq (%rdi,%rdi,4), %rax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_10:
 ; X64-JAG:       # BB#0:
@@ -496,7 +496,7 @@ define i64 @test_mul_by_10(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_10:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    imulq $10, %rdi, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_10:
 ; JAG-NOOPT:       # BB#0:
@@ -532,7 +532,7 @@ define i64 @test_mul_by_11(i64 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    leaq (%rdi,%rdi,4), %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    leaq (%rdi,%rax,2), %rax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_11:
 ; X64-JAG:       # BB#0:
@@ -551,7 +551,7 @@ define i64 @test_mul_by_11(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_11:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    imulq $11, %rdi, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_11:
 ; JAG-NOOPT:       # BB#0:
@@ -585,7 +585,7 @@ define i64 @test_mul_by_12(i64 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    shlq $2, %rdi # sched: [1:0.50]
 ; X64-HSW-NEXT:    leaq (%rdi,%rdi,2), %rax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_12:
 ; X64-JAG:       # BB#0:
@@ -604,7 +604,7 @@ define i64 @test_mul_by_12(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_12:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    imulq $12, %rdi, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_12:
 ; JAG-NOOPT:       # BB#0:
@@ -640,7 +640,7 @@ define i64 @test_mul_by_13(i64 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    leaq (%rdi,%rdi,2), %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    leaq (%rdi,%rax,4), %rax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_13:
 ; X64-JAG:       # BB#0:
@@ -659,7 +659,7 @@ define i64 @test_mul_by_13(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_13:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    imulq $13, %rdi, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_13:
 ; JAG-NOOPT:       # BB#0:
@@ -696,7 +696,7 @@ define i64 @test_mul_by_14(i64 %x) {
 ; X64-HSW-NEXT:    leaq (%rdi,%rdi,2), %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    leaq (%rdi,%rax,4), %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    addq %rdi, %rax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_14:
 ; X64-JAG:       # BB#0:
@@ -716,7 +716,7 @@ define i64 @test_mul_by_14(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_14:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    imulq $14, %rdi, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_14:
 ; JAG-NOOPT:       # BB#0:
@@ -751,7 +751,7 @@ define i64 @test_mul_by_15(i64 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    leaq (%rdi,%rdi,4), %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    leaq (%rax,%rax,2), %rax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_15:
 ; X64-JAG:       # BB#0:
@@ -770,7 +770,7 @@ define i64 @test_mul_by_15(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_15:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    imulq $15, %rdi, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_15:
 ; JAG-NOOPT:       # BB#0:
@@ -804,7 +804,7 @@ define i64 @test_mul_by_16(i64 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    shlq $4, %rdi # sched: [1:0.50]
 ; X64-HSW-NEXT:    movq %rdi, %rax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_16:
 ; X64-JAG:       # BB#0:
@@ -824,7 +824,7 @@ define i64 @test_mul_by_16(i64 %x) {
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    shlq $4, %rdi # sched: [1:0.50]
 ; HSW-NOOPT-NEXT:    movq %rdi, %rax # sched: [1:0.25]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_16:
 ; JAG-NOOPT:       # BB#0:
@@ -864,7 +864,7 @@ define i64 @test_mul_by_17(i64 %x) {
 ; X64-HSW-NEXT:    movq %rdi, %rax # sched: [1:0.25]
 ; X64-HSW-NEXT:    shlq $4, %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    leaq (%rax,%rdi), %rax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_17:
 ; X64-JAG:       # BB#0:
@@ -884,7 +884,7 @@ define i64 @test_mul_by_17(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_17:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    imulq $17, %rdi, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_17:
 ; JAG-NOOPT:       # BB#0:
@@ -920,7 +920,7 @@ define i64 @test_mul_by_18(i64 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    addq %rdi, %rdi # sched: [1:0.25]
 ; X64-HSW-NEXT:    leaq (%rdi,%rdi,8), %rax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_18:
 ; X64-JAG:       # BB#0:
@@ -939,7 +939,7 @@ define i64 @test_mul_by_18(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_18:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    imulq $18, %rdi, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_18:
 ; JAG-NOOPT:       # BB#0:
@@ -977,7 +977,7 @@ define i64 @test_mul_by_19(i64 %x) {
 ; X64-HSW-NEXT:    leaq (%rdi,%rdi,4), %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    shlq $2, %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    subq %rdi, %rax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_19:
 ; X64-JAG:       # BB#0:
@@ -997,7 +997,7 @@ define i64 @test_mul_by_19(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_19:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    imulq $19, %rdi, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_19:
 ; JAG-NOOPT:       # BB#0:
@@ -1031,7 +1031,7 @@ define i64 @test_mul_by_20(i64 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    shlq $2, %rdi # sched: [1:0.50]
 ; X64-HSW-NEXT:    leaq (%rdi,%rdi,4), %rax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_20:
 ; X64-JAG:       # BB#0:
@@ -1050,7 +1050,7 @@ define i64 @test_mul_by_20(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_20:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    imulq $20, %rdi, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_20:
 ; JAG-NOOPT:       # BB#0:
@@ -1086,7 +1086,7 @@ define i64 @test_mul_by_21(i64 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    leaq (%rdi,%rdi,4), %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    leaq (%rdi,%rax,4), %rax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_21:
 ; X64-JAG:       # BB#0:
@@ -1105,7 +1105,7 @@ define i64 @test_mul_by_21(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_21:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    imulq $21, %rdi, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_21:
 ; JAG-NOOPT:       # BB#0:
@@ -1142,7 +1142,7 @@ define i64 @test_mul_by_22(i64 %x) {
 ; X64-HSW-NEXT:    leaq (%rdi,%rdi,4), %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    leaq (%rdi,%rax,4), %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    addq %rdi, %rax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_22:
 ; X64-JAG:       # BB#0:
@@ -1162,7 +1162,7 @@ define i64 @test_mul_by_22(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_22:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    imulq $22, %rdi, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_22:
 ; JAG-NOOPT:       # BB#0:
@@ -1199,7 +1199,7 @@ define i64 @test_mul_by_23(i64 %x) {
 ; X64-HSW-NEXT:    leaq (%rdi,%rdi,2), %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    shlq $3, %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    subq %rdi, %rax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_23:
 ; X64-JAG:       # BB#0:
@@ -1219,7 +1219,7 @@ define i64 @test_mul_by_23(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_23:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    imulq $23, %rdi, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_23:
 ; JAG-NOOPT:       # BB#0:
@@ -1253,7 +1253,7 @@ define i64 @test_mul_by_24(i64 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    shlq $3, %rdi # sched: [1:0.50]
 ; X64-HSW-NEXT:    leaq (%rdi,%rdi,2), %rax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_24:
 ; X64-JAG:       # BB#0:
@@ -1272,7 +1272,7 @@ define i64 @test_mul_by_24(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_24:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    imulq $24, %rdi, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_24:
 ; JAG-NOOPT:       # BB#0:
@@ -1308,7 +1308,7 @@ define i64 @test_mul_by_25(i64 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    leaq (%rdi,%rdi,4), %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    leaq (%rax,%rax,4), %rax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_25:
 ; X64-JAG:       # BB#0:
@@ -1327,7 +1327,7 @@ define i64 @test_mul_by_25(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_25:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    imulq $25, %rdi, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_25:
 ; JAG-NOOPT:       # BB#0:
@@ -1365,7 +1365,7 @@ define i64 @test_mul_by_26(i64 %x) {
 ; X64-HSW-NEXT:    leaq (%rdi,%rdi,8), %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    leaq (%rax,%rax,2), %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    subq %rdi, %rax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_26:
 ; X64-JAG:       # BB#0:
@@ -1385,7 +1385,7 @@ define i64 @test_mul_by_26(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_26:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    imulq $26, %rdi, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_26:
 ; JAG-NOOPT:       # BB#0:
@@ -1420,7 +1420,7 @@ define i64 @test_mul_by_27(i64 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    leaq (%rdi,%rdi,8), %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    leaq (%rax,%rax,2), %rax # sched: [1:0.50]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_27:
 ; X64-JAG:       # BB#0:
@@ -1439,7 +1439,7 @@ define i64 @test_mul_by_27(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_27:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    imulq $27, %rdi, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_27:
 ; JAG-NOOPT:       # BB#0:
@@ -1477,7 +1477,7 @@ define i64 @test_mul_by_28(i64 %x) {
 ; X64-HSW-NEXT:    leaq (%rdi,%rdi,8), %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    leaq (%rax,%rax,2), %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    addq %rdi, %rax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_28:
 ; X64-JAG:       # BB#0:
@@ -1497,7 +1497,7 @@ define i64 @test_mul_by_28(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_28:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    imulq $28, %rdi, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_28:
 ; JAG-NOOPT:       # BB#0:
@@ -1536,7 +1536,7 @@ define i64 @test_mul_by_29(i64 %x) {
 ; X64-HSW-NEXT:    leaq (%rax,%rax,2), %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    addq %rdi, %rax # sched: [1:0.25]
 ; X64-HSW-NEXT:    addq %rdi, %rax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_29:
 ; X64-JAG:       # BB#0:
@@ -1557,7 +1557,7 @@ define i64 @test_mul_by_29(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_29:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    imulq $29, %rdi, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_29:
 ; JAG-NOOPT:       # BB#0:
@@ -1596,7 +1596,7 @@ define i64 @test_mul_by_30(i64 %x) {
 ; X64-HSW-NEXT:    shlq $5, %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    subq %rdi, %rax # sched: [1:0.25]
 ; X64-HSW-NEXT:    subq %rdi, %rax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_30:
 ; X64-JAG:       # BB#0:
@@ -1617,7 +1617,7 @@ define i64 @test_mul_by_30(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_30:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    imulq $30, %rdi, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_30:
 ; JAG-NOOPT:       # BB#0:
@@ -1654,7 +1654,7 @@ define i64 @test_mul_by_31(i64 %x) {
 ; X64-HSW-NEXT:    movq %rdi, %rax # sched: [1:0.25]
 ; X64-HSW-NEXT:    shlq $5, %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    subq %rdi, %rax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_31:
 ; X64-JAG:       # BB#0:
@@ -1674,7 +1674,7 @@ define i64 @test_mul_by_31(i64 %x) {
 ; HSW-NOOPT-LABEL: test_mul_by_31:
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    imulq $31, %rdi, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_31:
 ; JAG-NOOPT:       # BB#0:
@@ -1709,7 +1709,7 @@ define i64 @test_mul_by_32(i64 %x) {
 ; X64-HSW:       # BB#0:
 ; X64-HSW-NEXT:    shlq $5, %rdi # sched: [1:0.50]
 ; X64-HSW-NEXT:    movq %rdi, %rax # sched: [1:0.25]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_by_32:
 ; X64-JAG:       # BB#0:
@@ -1729,7 +1729,7 @@ define i64 @test_mul_by_32(i64 %x) {
 ; HSW-NOOPT:       # BB#0:
 ; HSW-NOOPT-NEXT:    shlq $5, %rdi # sched: [1:0.50]
 ; HSW-NOOPT-NEXT:    movq %rdi, %rax # sched: [1:0.25]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_by_32:
 ; JAG-NOOPT:       # BB#0:
@@ -1793,7 +1793,7 @@ define i64 @test_mul_spec(i64 %x) nounwind {
 ; X64-HSW-NEXT:    leaq (%rdi,%rdi,4), %rax # sched: [1:0.50]
 ; X64-HSW-NEXT:    addq $2, %rax # sched: [1:0.25]
 ; X64-HSW-NEXT:    imulq %rcx, %rax # sched: [3:1.00]
-; X64-HSW-NEXT:    retq # sched: [1:1.00]
+; X64-HSW-NEXT:    retq # sched: [2:1.00]
 ;
 ; X64-JAG-LABEL: test_mul_spec:
 ; X64-JAG:       # BB#0:
@@ -1841,7 +1841,7 @@ define i64 @test_mul_spec(i64 %x) nounwind {
 ; HSW-NOOPT-NEXT:    leaq (%rdi,%rdi,4), %rax # sched: [1:0.50]
 ; HSW-NOOPT-NEXT:    addq $2, %rax # sched: [1:0.25]
 ; HSW-NOOPT-NEXT:    imulq %rcx, %rax # sched: [3:1.00]
-; HSW-NOOPT-NEXT:    retq # sched: [1:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [2:1.00]
 ;
 ; JAG-NOOPT-LABEL: test_mul_spec:
 ; JAG-NOOPT:       # BB#0:

llvm/test/CodeGen/X86/pr32329.ll

@@ -59,8 +59,8 @@ define void @foo() local_unnamed_addr {
 ; X86-NEXT:    cmovnel %ecx, %esi
 ; X86-NEXT:    cmpl %edx, %edi
 ; X86-NEXT:    movl %ebp, var_50+4
-; X86-NEXT:    movl %esi, var_50
 ; X86-NEXT:    setge var_205
+; X86-NEXT:    movl %esi, var_50
 ; X86-NEXT:    imull %eax, %ebx
 ; X86-NEXT:    movb %bl, var_218
 ; X86-NEXT:    popl %esi

llvm/test/CodeGen/X86/recip-fastmath.ll

@@ -45,15 +45,15 @@ define float @f32_no_estimate(float %x) #0 {
 ;
 ; SANDY-LABEL: f32_no_estimate:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [4:0.50]
+; SANDY-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [6:0.50]
-; SANDY-NEXT:    vdivss %xmm0, %xmm1, %xmm0 # sched: [12:1.00]
+; SANDY-NEXT:    vdivss %xmm0, %xmm1, %xmm0 # sched: [14:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: f32_no_estimate:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [4:0.50]
+; HASWELL-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [?:5.000000e-01]
-; HASWELL-NEXT:    vdivss %xmm0, %xmm1, %xmm0 # sched: [12:1.00]
+; HASWELL-NEXT:    vdivss %xmm0, %xmm1, %xmm0 # sched: [13:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: f32_no_estimate:
 ; HASWELL-NO-FMA:       # BB#0:
@@ -63,9 +63,9 @@ define float @f32_no_estimate(float %x) #0 {
 ;
 ; AVX512-LABEL: f32_no_estimate:
 ; AVX512:       # BB#0:
-; AVX512-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [4:0.50]
+; AVX512-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [?:5.000000e-01]
-; AVX512-NEXT:    vdivss %xmm0, %xmm1, %xmm0 # sched: [12:1.00]
+; AVX512-NEXT:    vdivss %xmm0, %xmm1, %xmm0 # sched: [13:1.00]
-; AVX512-NEXT:    retq # sched: [1:1.00]
+; AVX512-NEXT:    retq # sched: [2:1.00]
   %div = fdiv fast float 1.0, %x
   ret float %div
 }
@@ -113,18 +113,18 @@ define float @f32_one_step(float %x) #1 {
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
 ; SANDY-NEXT:    vmulss %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
-; SANDY-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [4:0.50]
+; SANDY-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [6:0.50]
 ; SANDY-NEXT:    vsubss %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: f32_one_step:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
-; HASWELL-NEXT:    vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0
+; HASWELL-NEXT:    vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vfmadd132ss %xmm1, %xmm1, %xmm0
+; HASWELL-NEXT:    vfmadd132ss %xmm1, %xmm1, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: f32_one_step:
 ; HASWELL-NO-FMA:       # BB#0:
@@ -139,9 +139,9 @@ define float @f32_one_step(float %x) #1 {
 ; AVX512-LABEL: f32_one_step:
 ; AVX512:       # BB#0:
 ; AVX512-NEXT:    vrcp14ss %xmm0, %xmm0, %xmm1
-; AVX512-NEXT:    vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0
+; AVX512-NEXT:    vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0 # sched: [5:0.50]
-; AVX512-NEXT:    vfmadd132ss %xmm1, %xmm1, %xmm0
+; AVX512-NEXT:    vfmadd132ss %xmm1, %xmm1, %xmm0 # sched: [5:0.50]
-; AVX512-NEXT:    retq # sched: [1:1.00]
+; AVX512-NEXT:    retq # sched: [2:1.00]
   %div = fdiv fast float 1.0, %x
   ret float %div
 }
@@ -207,7 +207,7 @@ define float @f32_two_step(float %x) #2 {
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
 ; SANDY-NEXT:    vmulss %xmm1, %xmm0, %xmm2 # sched: [5:1.00]
-; SANDY-NEXT:    vmovss {{.*#+}} xmm3 = mem[0],zero,zero,zero sched: [4:0.50]
+; SANDY-NEXT:    vmovss {{.*#+}} xmm3 = mem[0],zero,zero,zero sched: [6:0.50]
 ; SANDY-NEXT:    vsubss %xmm2, %xmm3, %xmm2 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulss %xmm2, %xmm1, %xmm2 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddss %xmm2, %xmm1, %xmm1 # sched: [3:1.00]
@@ -215,18 +215,18 @@ define float @f32_two_step(float %x) #2 {
 ; SANDY-NEXT:    vsubss %xmm0, %xmm3, %xmm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: f32_two_step:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
-; HASWELL-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [4:0.50]
+; HASWELL-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [?:5.000000e-01]
 ; HASWELL-NEXT:    vmovaps %xmm1, %xmm3 # sched: [1:1.00]
-; HASWELL-NEXT:    vfnmadd213ss %xmm2, %xmm0, %xmm3
+; HASWELL-NEXT:    vfnmadd213ss %xmm2, %xmm0, %xmm3 # sched: [5:0.50]
-; HASWELL-NEXT:    vfmadd132ss %xmm1, %xmm1, %xmm3
+; HASWELL-NEXT:    vfmadd132ss %xmm1, %xmm1, %xmm3 # sched: [5:0.50]
-; HASWELL-NEXT:    vfnmadd213ss %xmm2, %xmm3, %xmm0
+; HASWELL-NEXT:    vfnmadd213ss %xmm2, %xmm3, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vfmadd132ss %xmm3, %xmm3, %xmm0
+; HASWELL-NEXT:    vfmadd132ss %xmm3, %xmm3, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: f32_two_step:
 ; HASWELL-NO-FMA:       # BB#0:
@@ -245,13 +245,13 @@ define float @f32_two_step(float %x) #2 {
 ; AVX512-LABEL: f32_two_step:
 ; AVX512:       # BB#0:
 ; AVX512-NEXT:    vrcp14ss %xmm0, %xmm0, %xmm1
-; AVX512-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [4:0.50]
+; AVX512-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [?:5.000000e-01]
 ; AVX512-NEXT:    vmovaps %xmm1, %xmm3 # sched: [1:1.00]
-; AVX512-NEXT:    vfnmadd213ss %xmm2, %xmm0, %xmm3
+; AVX512-NEXT:    vfnmadd213ss %xmm2, %xmm0, %xmm3 # sched: [5:0.50]
-; AVX512-NEXT:    vfmadd132ss %xmm1, %xmm1, %xmm3
+; AVX512-NEXT:    vfmadd132ss %xmm1, %xmm1, %xmm3 # sched: [5:0.50]
-; AVX512-NEXT:    vfnmadd213ss %xmm2, %xmm3, %xmm0
+; AVX512-NEXT:    vfnmadd213ss %xmm2, %xmm3, %xmm0 # sched: [5:0.50]
-; AVX512-NEXT:    vfmadd132ss %xmm3, %xmm3, %xmm0
+; AVX512-NEXT:    vfmadd132ss %xmm3, %xmm3, %xmm0 # sched: [5:0.50]
-; AVX512-NEXT:    retq # sched: [1:1.00]
+; AVX512-NEXT:    retq # sched: [2:1.00]
   %div = fdiv fast float 1.0, %x
   ret float %div
 }
@@ -284,15 +284,15 @@ define <4 x float> @v4f32_no_estimate(<4 x float> %x) #0 {
 ;
 ; SANDY-LABEL: v4f32_no_estimate:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vmovaps {{.*#+}} xmm1 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50]
+; SANDY-NEXT:    vmovaps {{.*#+}} xmm1 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [6:0.50]
-; SANDY-NEXT:    vdivps %xmm0, %xmm1, %xmm0 # sched: [12:1.00]
+; SANDY-NEXT:    vdivps %xmm0, %xmm1, %xmm0 # sched: [14:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: v4f32_no_estimate:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm1 # sched: [4:0.50]
+; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm1 # sched: [?:5.000000e-01]
-; HASWELL-NEXT:    vdivps %xmm0, %xmm1, %xmm0 # sched: [12:1.00]
+; HASWELL-NEXT:    vdivps %xmm0, %xmm1, %xmm0 # sched: [13:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v4f32_no_estimate:
 ; HASWELL-NO-FMA:       # BB#0:
@@ -302,9 +302,9 @@ define <4 x float> @v4f32_no_estimate(<4 x float> %x) #0 {
 ;
 ; AVX512-LABEL: v4f32_no_estimate:
 ; AVX512:       # BB#0:
-; AVX512-NEXT:    vbroadcastss {{.*}}(%rip), %xmm1 # sched: [4:0.50]
+; AVX512-NEXT:    vbroadcastss {{.*}}(%rip), %xmm1 # sched: [?:5.000000e-01]
-; AVX512-NEXT:    vdivps %xmm0, %xmm1, %xmm0 # sched: [12:1.00]
+; AVX512-NEXT:    vdivps %xmm0, %xmm1, %xmm0 # sched: [13:1.00]
-; AVX512-NEXT:    retq # sched: [1:1.00]
+; AVX512-NEXT:    retq # sched: [2:1.00]
   %div = fdiv fast <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>, %x
   ret <4 x float> %div
 }
@@ -350,21 +350,21 @@ define <4 x float> @v4f32_one_step(<4 x float> %x) #1 {
 ;
 ; SANDY-LABEL: v4f32_one_step:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
+; SANDY-NEXT:    vrcpps %xmm0, %xmm1 # sched: [7:3.00]
 ; SANDY-NEXT:    vmulps %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
-; SANDY-NEXT:    vmovaps {{.*#+}} xmm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50]
+; SANDY-NEXT:    vmovaps {{.*#+}} xmm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [6:0.50]
 ; SANDY-NEXT:    vsubps %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: v4f32_one_step:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
-; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01]
-; HASWELL-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm0
+; HASWELL-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0
+; HASWELL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v4f32_one_step:
 ; HASWELL-NO-FMA:       # BB#0:
@@ -379,17 +379,17 @@ define <4 x float> @v4f32_one_step(<4 x float> %x) #1 {
 ; KNL-LABEL: v4f32_one_step:
 ; KNL:       # BB#0:
 ; KNL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
-; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01]
-; KNL-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm0
+; KNL-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm0 # sched: [5:0.50]
-; KNL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0
+; KNL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0 # sched: [5:0.50]
-; KNL-NEXT:    retq # sched: [1:1.00]
+; KNL-NEXT:    retq # sched: [2:1.00]
 ;
 ; SKX-LABEL: v4f32_one_step:
 ; SKX:       # BB#0:
 ; SKX-NEXT:    vrcp14ps %xmm0, %xmm1
 ; SKX-NEXT:    vfnmadd213ps {{.*}}(%rip){1to4}, %xmm1, %xmm0
-; SKX-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0
+; SKX-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0 # sched: [5:0.50]
-; SKX-NEXT:    retq # sched: [1:1.00]
+; SKX-NEXT:    retq # sched: [2:1.00]
   %div = fdiv fast <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>, %x
   ret <4 x float> %div
 }
@@ -453,9 +453,9 @@ define <4 x float> @v4f32_two_step(<4 x float> %x) #2 {
 ;
 ; SANDY-LABEL: v4f32_two_step:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
+; SANDY-NEXT:    vrcpps %xmm0, %xmm1 # sched: [7:3.00]
 ; SANDY-NEXT:    vmulps %xmm1, %xmm0, %xmm2 # sched: [5:1.00]
-; SANDY-NEXT:    vmovaps {{.*#+}} xmm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50]
+; SANDY-NEXT:    vmovaps {{.*#+}} xmm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [6:0.50]
 ; SANDY-NEXT:    vsubps %xmm2, %xmm3, %xmm2 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps %xmm2, %xmm1, %xmm2 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddps %xmm2, %xmm1, %xmm1 # sched: [3:1.00]
@@ -463,18 +463,18 @@ define <4 x float> @v4f32_two_step(<4 x float> %x) #2 {
 ; SANDY-NEXT:    vsubps %xmm0, %xmm3, %xmm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: v4f32_two_step:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
-; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01]
 ; HASWELL-NEXT:    vmovaps %xmm1, %xmm3 # sched: [1:1.00]
-; HASWELL-NEXT:    vfnmadd213ps %xmm2, %xmm0, %xmm3
+; HASWELL-NEXT:    vfnmadd213ps %xmm2, %xmm0, %xmm3 # sched: [5:0.50]
-; HASWELL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm3
+; HASWELL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm3 # sched: [5:0.50]
-; HASWELL-NEXT:    vfnmadd213ps %xmm2, %xmm3, %xmm0
+; HASWELL-NEXT:    vfnmadd213ps %xmm2, %xmm3, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vfmadd132ps %xmm3, %xmm3, %xmm0
+; HASWELL-NEXT:    vfmadd132ps %xmm3, %xmm3, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v4f32_two_step:
 ; HASWELL-NO-FMA:       # BB#0:
@@ -493,24 +493,24 @@ define <4 x float> @v4f32_two_step(<4 x float> %x) #2 {
 ; KNL-LABEL: v4f32_two_step:
 ; KNL:       # BB#0:
 ; KNL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
-; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01]
 ; KNL-NEXT:    vmovaps %xmm1, %xmm3 # sched: [1:1.00]
-; KNL-NEXT:    vfnmadd213ps %xmm2, %xmm0, %xmm3
+; KNL-NEXT:    vfnmadd213ps %xmm2, %xmm0, %xmm3 # sched: [5:0.50]
-; KNL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm3
+; KNL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm3 # sched: [5:0.50]
-; KNL-NEXT:    vfnmadd213ps %xmm2, %xmm3, %xmm0
+; KNL-NEXT:    vfnmadd213ps %xmm2, %xmm3, %xmm0 # sched: [5:0.50]
-; KNL-NEXT:    vfmadd132ps %xmm3, %xmm3, %xmm0
+; KNL-NEXT:    vfmadd132ps %xmm3, %xmm3, %xmm0 # sched: [5:0.50]
-; KNL-NEXT:    retq # sched: [1:1.00]
+; KNL-NEXT:    retq # sched: [2:1.00]
 ;
 ; SKX-LABEL: v4f32_two_step:
 ; SKX:       # BB#0:
 ; SKX-NEXT:    vrcp14ps %xmm0, %xmm1
-; SKX-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; SKX-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01]
 ; SKX-NEXT:    vmovaps %xmm1, %xmm3 # sched: [1:1.00]
-; SKX-NEXT:    vfnmadd213ps %xmm2, %xmm0, %xmm3
+; SKX-NEXT:    vfnmadd213ps %xmm2, %xmm0, %xmm3 # sched: [5:0.50]
-; SKX-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm3
+; SKX-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm3 # sched: [5:0.50]
-; SKX-NEXT:    vfnmadd213ps %xmm2, %xmm3, %xmm0
+; SKX-NEXT:    vfnmadd213ps %xmm2, %xmm3, %xmm0 # sched: [5:0.50]
-; SKX-NEXT:    vfmadd132ps %xmm3, %xmm3, %xmm0
+; SKX-NEXT:    vfmadd132ps %xmm3, %xmm3, %xmm0 # sched: [5:0.50]
-; SKX-NEXT:    retq # sched: [1:1.00]
+; SKX-NEXT:    retq # sched: [2:1.00]
   %div = fdiv fast <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>, %x
   ret <4 x float> %div
 }
@@ -546,15 +546,15 @@ define <8 x float> @v8f32_no_estimate(<8 x float> %x) #0 {
 ;
 ; SANDY-LABEL: v8f32_no_estimate:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vmovaps {{.*#+}} ymm1 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50]
+; SANDY-NEXT:    vmovaps {{.*#+}} ymm1 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50]
-; SANDY-NEXT:    vdivps %ymm0, %ymm1, %ymm0 # sched: [12:1.00]
+; SANDY-NEXT:    vdivps %ymm0, %ymm1, %ymm0 # sched: [29:3.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: v8f32_no_estimate:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %ymm1 # sched: [5:1.00]
-; HASWELL-NEXT:    vdivps %ymm0, %ymm1, %ymm0 # sched: [19:2.00]
+; HASWELL-NEXT:    vdivps %ymm0, %ymm1, %ymm0 # sched: [21:2.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v8f32_no_estimate:
 ; HASWELL-NO-FMA:       # BB#0:
@@ -565,8 +565,8 @@ define <8 x float> @v8f32_no_estimate(<8 x float> %x) #0 {
 ; AVX512-LABEL: v8f32_no_estimate:
 ; AVX512:       # BB#0:
 ; AVX512-NEXT:    vbroadcastss {{.*}}(%rip), %ymm1 # sched: [5:1.00]
-; AVX512-NEXT:    vdivps %ymm0, %ymm1, %ymm0 # sched: [19:2.00]
+; AVX512-NEXT:    vdivps %ymm0, %ymm1, %ymm0 # sched: [21:2.00]
-; AVX512-NEXT:    retq # sched: [1:1.00]
+; AVX512-NEXT:    retq # sched: [2:1.00]
   %div = fdiv fast <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>, %x
   ret <8 x float> %div
 }
@@ -621,19 +621,19 @@ define <8 x float> @v8f32_one_step(<8 x float> %x) #1 {
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vrcpps %ymm0, %ymm1 # sched: [5:1.00]
 ; SANDY-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; SANDY-NEXT:    vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50]
+; SANDY-NEXT:    vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50]
 ; SANDY-NEXT:    vsubps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: v8f32_one_step:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
+; HASWELL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [11:2.00]
 ; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
-; HASWELL-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm0
+; HASWELL-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0
+; HASWELL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0 # sched: [5:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v8f32_one_step:
 ; HASWELL-NO-FMA:       # BB#0:
@@ -647,18 +647,18 @@ define <8 x float> @v8f32_one_step(<8 x float> %x) #1 {
 ;
 ; KNL-LABEL: v8f32_one_step:
 ; KNL:       # BB#0:
-; KNL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
+; KNL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [11:2.00]
 ; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
-; KNL-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm0
+; KNL-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm0 # sched: [5:0.50]
-; KNL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0
+; KNL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0 # sched: [5:0.50]
-; KNL-NEXT:    retq # sched: [1:1.00]
+; KNL-NEXT:    retq # sched: [2:1.00]
 ;
 ; SKX-LABEL: v8f32_one_step:
 ; SKX:       # BB#0:
 ; SKX-NEXT:    vrcp14ps %ymm0, %ymm1
 ; SKX-NEXT:    vfnmadd213ps {{.*}}(%rip){1to8}, %ymm1, %ymm0
-; SKX-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0
+; SKX-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0 # sched: [5:0.50]
-; SKX-NEXT:    retq # sched: [1:1.00]
+; SKX-NEXT:    retq # sched: [2:1.00]
   %div = fdiv fast <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>, %x
   ret <8 x float> %div
 }
@@ -737,7 +737,7 @@ define <8 x float> @v8f32_two_step(<8 x float> %x) #2 {
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vrcpps %ymm0, %ymm1 # sched: [5:1.00]
 ; SANDY-NEXT:    vmulps %ymm1, %ymm0, %ymm2 # sched: [5:1.00]
-; SANDY-NEXT:    vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50]
+; SANDY-NEXT:    vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50]
 ; SANDY-NEXT:    vsubps %ymm2, %ymm3, %ymm2 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps %ymm2, %ymm1, %ymm2 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddps %ymm2, %ymm1, %ymm1 # sched: [3:1.00]
@@ -745,18 +745,18 @@ define <8 x float> @v8f32_two_step(<8 x float> %x) #2 {
 ; SANDY-NEXT:    vsubps %ymm0, %ymm3, %ymm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: v8f32_two_step:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
+; HASWELL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [11:2.00]
 ; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
 ; HASWELL-NEXT:    vmovaps %ymm1, %ymm3 # sched: [1:1.00]
-; HASWELL-NEXT:    vfnmadd213ps %ymm2, %ymm0, %ymm3
+; HASWELL-NEXT:    vfnmadd213ps %ymm2, %ymm0, %ymm3 # sched: [5:0.50]
-; HASWELL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm3
+; HASWELL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm3 # sched: [5:0.50]
-; HASWELL-NEXT:    vfnmadd213ps %ymm2, %ymm3, %ymm0
+; HASWELL-NEXT:    vfnmadd213ps %ymm2, %ymm3, %ymm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vfmadd132ps %ymm3, %ymm3, %ymm0
+; HASWELL-NEXT:    vfmadd132ps %ymm3, %ymm3, %ymm0 # sched: [5:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v8f32_two_step:
 ; HASWELL-NO-FMA:       # BB#0:
@@ -774,25 +774,25 @@ define <8 x float> @v8f32_two_step(<8 x float> %x) #2 {
 ;
 ; KNL-LABEL: v8f32_two_step:
 ; KNL:       # BB#0:
-; KNL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
+; KNL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [11:2.00]
 ; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
 ; KNL-NEXT:    vmovaps %ymm1, %ymm3 # sched: [1:1.00]
-; KNL-NEXT:    vfnmadd213ps %ymm2, %ymm0, %ymm3
+; KNL-NEXT:    vfnmadd213ps %ymm2, %ymm0, %ymm3 # sched: [5:0.50]
-; KNL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm3
+; KNL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm3 # sched: [5:0.50]
-; KNL-NEXT:    vfnmadd213ps %ymm2, %ymm3, %ymm0
+; KNL-NEXT:    vfnmadd213ps %ymm2, %ymm3, %ymm0 # sched: [5:0.50]
-; KNL-NEXT:    vfmadd132ps %ymm3, %ymm3, %ymm0
+; KNL-NEXT:    vfmadd132ps %ymm3, %ymm3, %ymm0 # sched: [5:0.50]
-; KNL-NEXT:    retq # sched: [1:1.00]
+; KNL-NEXT:    retq # sched: [2:1.00]
 ;
 ; SKX-LABEL: v8f32_two_step:
 ; SKX:       # BB#0:
 ; SKX-NEXT:    vrcp14ps %ymm0, %ymm1
 ; SKX-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
 ; SKX-NEXT:    vmovaps %ymm1, %ymm3 # sched: [1:1.00]
-; SKX-NEXT:    vfnmadd213ps %ymm2, %ymm0, %ymm3
+; SKX-NEXT:    vfnmadd213ps %ymm2, %ymm0, %ymm3 # sched: [5:0.50]
-; SKX-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm3
+; SKX-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm3 # sched: [5:0.50]
-; SKX-NEXT:    vfnmadd213ps %ymm2, %ymm3, %ymm0
+; SKX-NEXT:    vfnmadd213ps %ymm2, %ymm3, %ymm0 # sched: [5:0.50]
-; SKX-NEXT:    vfmadd132ps %ymm3, %ymm3, %ymm0
+; SKX-NEXT:    vfmadd132ps %ymm3, %ymm3, %ymm0 # sched: [5:0.50]
-; SKX-NEXT:    retq # sched: [1:1.00]
+; SKX-NEXT:    retq # sched: [2:1.00]
   %div = fdiv fast <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>, %x
   ret <8 x float> %div
 }

llvm/test/CodeGen/X86/recip-fastmath2.ll

@@ -39,26 +39,26 @@ define float @f32_no_step_2(float %x) #3 {
 ; SANDY-LABEL: f32_no_step_2:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vrcpss %xmm0, %xmm0, %xmm0 # sched: [5:1.00]
-; SANDY-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:1.00]
+; SANDY-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [11:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: f32_no_step_2:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vrcpss %xmm0, %xmm0, %xmm0 # sched: [5:1.00]
-; HASWELL-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
+; HASWELL-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: f32_no_step_2:
 ; HASWELL-NO-FMA:       # BB#0:
 ; HASWELL-NO-FMA-NEXT:    vrcpss %xmm0, %xmm0, %xmm0 # sched: [5:1.00]
-; HASWELL-NO-FMA-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
+; HASWELL-NO-FMA-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50]
-; HASWELL-NO-FMA-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NO-FMA-NEXT:    retq # sched: [2:1.00]
 ;
 ; AVX512-LABEL: f32_no_step_2:
 ; AVX512:       # BB#0:
 ; AVX512-NEXT:    vrcp14ss %xmm0, %xmm0, %xmm0
-; AVX512-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
+; AVX512-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50]
-; AVX512-NEXT:    retq # sched: [1:1.00]
+; AVX512-NEXT:    retq # sched: [2:1.00]
   %div = fdiv fast float 1234.0, %x
   ret float %div
 }
@@ -110,39 +110,39 @@ define float @f32_one_step_2(float %x) #1 {
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
 ; SANDY-NEXT:    vmulss %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
-; SANDY-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [4:0.50]
+; SANDY-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [6:0.50]
 ; SANDY-NEXT:    vsubss %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:1.00]
+; SANDY-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [11:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: f32_one_step_2:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
-; HASWELL-NEXT:    vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0
+; HASWELL-NEXT:    vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vfmadd132ss %xmm1, %xmm1, %xmm0
+; HASWELL-NEXT:    vfmadd132ss %xmm1, %xmm1, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
+; HASWELL-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: f32_one_step_2:
 ; HASWELL-NO-FMA:       # BB#0:
 ; HASWELL-NO-FMA-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulss %xmm1, %xmm0, %xmm0 # sched: [5:0.50]
-; HASWELL-NO-FMA-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [4:0.50]
+; HASWELL-NO-FMA-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [?:5.000000e-01]
 ; HASWELL-NO-FMA-NEXT:    vsubss %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; HASWELL-NO-FMA-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
+; HASWELL-NO-FMA-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50]
-; HASWELL-NO-FMA-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NO-FMA-NEXT:    retq # sched: [2:1.00]
 ;
 ; AVX512-LABEL: f32_one_step_2:
 ; AVX512:       # BB#0:
 ; AVX512-NEXT:    vrcp14ss %xmm0, %xmm0, %xmm1
-; AVX512-NEXT:    vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0
+; AVX512-NEXT:    vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0 # sched: [5:0.50]
-; AVX512-NEXT:    vfmadd132ss %xmm1, %xmm1, %xmm0
+; AVX512-NEXT:    vfmadd132ss %xmm1, %xmm1, %xmm0 # sched: [5:0.50]
-; AVX512-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
+; AVX512-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50]
-; AVX512-NEXT:    retq # sched: [1:1.00]
+; AVX512-NEXT:    retq # sched: [2:1.00]
   %div = fdiv fast float 3456.0, %x
   ret float %div
 }
@@ -198,43 +198,43 @@ define float @f32_one_step_2_divs(float %x) #1 {
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
 ; SANDY-NEXT:    vmulss %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
-; SANDY-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [4:0.50]
+; SANDY-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [6:0.50]
 ; SANDY-NEXT:    vsubss %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm1 # sched: [9:1.00]
+; SANDY-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm1 # sched: [11:1.00]
 ; SANDY-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: f32_one_step_2_divs:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
-; HASWELL-NEXT:    vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0
+; HASWELL-NEXT:    vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vfmadd132ss %xmm1, %xmm1, %xmm0
+; HASWELL-NEXT:    vfmadd132ss %xmm1, %xmm1, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm1 # sched: [9:0.50]
+; HASWELL-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm1 # sched: [5:0.50]
 ; HASWELL-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: f32_one_step_2_divs:
 ; HASWELL-NO-FMA:       # BB#0:
 ; HASWELL-NO-FMA-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulss %xmm1, %xmm0, %xmm0 # sched: [5:0.50]
-; HASWELL-NO-FMA-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [4:0.50]
+; HASWELL-NO-FMA-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [?:5.000000e-01]
 ; HASWELL-NO-FMA-NEXT:    vsubss %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; HASWELL-NO-FMA-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm1 # sched: [9:0.50]
+; HASWELL-NO-FMA-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm1 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
-; HASWELL-NO-FMA-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NO-FMA-NEXT:    retq # sched: [2:1.00]
 ;
 ; AVX512-LABEL: f32_one_step_2_divs:
 ; AVX512:       # BB#0:
 ; AVX512-NEXT:    vrcp14ss %xmm0, %xmm0, %xmm1
-; AVX512-NEXT:    vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0
+; AVX512-NEXT:    vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0 # sched: [5:0.50]
-; AVX512-NEXT:    vfmadd132ss %xmm1, %xmm1, %xmm0
+; AVX512-NEXT:    vfmadd132ss %xmm1, %xmm1, %xmm0 # sched: [5:0.50]
-; AVX512-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm1 # sched: [9:0.50]
+; AVX512-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm1 # sched: [5:0.50]
 ; AVX512-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
-; AVX512-NEXT:    retq # sched: [1:1.00]
+; AVX512-NEXT:    retq # sched: [2:1.00]
   %div = fdiv fast float 3456.0, %x
   %div2 = fdiv fast float %div, %x
   ret float %div2
@@ -305,7 +305,7 @@ define float @f32_two_step_2(float %x) #2 {
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
 ; SANDY-NEXT:    vmulss %xmm1, %xmm0, %xmm2 # sched: [5:1.00]
-; SANDY-NEXT:    vmovss {{.*#+}} xmm3 = mem[0],zero,zero,zero sched: [4:0.50]
+; SANDY-NEXT:    vmovss {{.*#+}} xmm3 = mem[0],zero,zero,zero sched: [6:0.50]
 ; SANDY-NEXT:    vsubss %xmm2, %xmm3, %xmm2 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulss %xmm2, %xmm1, %xmm2 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddss %xmm2, %xmm1, %xmm1 # sched: [3:1.00]
@@ -313,26 +313,26 @@ define float @f32_two_step_2(float %x) #2 {
 ; SANDY-NEXT:    vsubss %xmm0, %xmm3, %xmm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:1.00]
+; SANDY-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [11:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: f32_two_step_2:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
-; HASWELL-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [4:0.50]
+; HASWELL-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [?:5.000000e-01]
 ; HASWELL-NEXT:    vmovaps %xmm1, %xmm3 # sched: [1:1.00]
-; HASWELL-NEXT:    vfnmadd213ss %xmm2, %xmm0, %xmm3
+; HASWELL-NEXT:    vfnmadd213ss %xmm2, %xmm0, %xmm3 # sched: [5:0.50]
-; HASWELL-NEXT:    vfmadd132ss %xmm1, %xmm1, %xmm3
+; HASWELL-NEXT:    vfmadd132ss %xmm1, %xmm1, %xmm3 # sched: [5:0.50]
-; HASWELL-NEXT:    vfnmadd213ss %xmm2, %xmm3, %xmm0
+; HASWELL-NEXT:    vfnmadd213ss %xmm2, %xmm3, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vfmadd132ss %xmm3, %xmm3, %xmm0
+; HASWELL-NEXT:    vfmadd132ss %xmm3, %xmm3, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
+; HASWELL-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: f32_two_step_2:
 ; HASWELL-NO-FMA:       # BB#0:
 ; HASWELL-NO-FMA-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulss %xmm1, %xmm0, %xmm2 # sched: [5:0.50]
-; HASWELL-NO-FMA-NEXT:    vmovss {{.*#+}} xmm3 = mem[0],zero,zero,zero sched: [4:0.50]
+; HASWELL-NO-FMA-NEXT:    vmovss {{.*#+}} xmm3 = mem[0],zero,zero,zero sched: [?:5.000000e-01]
 ; HASWELL-NO-FMA-NEXT:    vsubss %xmm2, %xmm3, %xmm2 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulss %xmm2, %xmm1, %xmm2 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    vaddss %xmm2, %xmm1, %xmm1 # sched: [3:1.00]
@@ -340,20 +340,20 @@ define float @f32_two_step_2(float %x) #2 {
 ; HASWELL-NO-FMA-NEXT:    vsubss %xmm0, %xmm3, %xmm0 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; HASWELL-NO-FMA-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
+; HASWELL-NO-FMA-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50]
-; HASWELL-NO-FMA-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NO-FMA-NEXT:    retq # sched: [2:1.00]
 ;
 ; AVX512-LABEL: f32_two_step_2:
 ; AVX512:       # BB#0:
 ; AVX512-NEXT:    vrcp14ss %xmm0, %xmm0, %xmm1
-; AVX512-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [4:0.50]
+; AVX512-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [?:5.000000e-01]
 ; AVX512-NEXT:    vmovaps %xmm1, %xmm3 # sched: [1:1.00]
-; AVX512-NEXT:    vfnmadd213ss %xmm2, %xmm0, %xmm3
+; AVX512-NEXT:    vfnmadd213ss %xmm2, %xmm0, %xmm3 # sched: [5:0.50]
-; AVX512-NEXT:    vfmadd132ss %xmm1, %xmm1, %xmm3
+; AVX512-NEXT:    vfmadd132ss %xmm1, %xmm1, %xmm3 # sched: [5:0.50]
-; AVX512-NEXT:    vfnmadd213ss %xmm2, %xmm3, %xmm0
+; AVX512-NEXT:    vfnmadd213ss %xmm2, %xmm3, %xmm0 # sched: [5:0.50]
-; AVX512-NEXT:    vfmadd132ss %xmm3, %xmm3, %xmm0
+; AVX512-NEXT:    vfmadd132ss %xmm3, %xmm3, %xmm0 # sched: [5:0.50]
-; AVX512-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
+; AVX512-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50]
-; AVX512-NEXT:    retq # sched: [1:1.00]
+; AVX512-NEXT:    retq # sched: [2:1.00]
   %div = fdiv fast float 6789.0, %x
   ret float %div
 }
@@ -403,51 +403,51 @@ define <4 x float> @v4f32_one_step2(<4 x float> %x) #1 {
 ;
 ; SANDY-LABEL: v4f32_one_step2:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
+; SANDY-NEXT:    vrcpps %xmm0, %xmm1 # sched: [7:3.00]
 ; SANDY-NEXT:    vmulps %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
-; SANDY-NEXT:    vmovaps {{.*#+}} xmm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50]
+; SANDY-NEXT:    vmovaps {{.*#+}} xmm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [6:0.50]
 ; SANDY-NEXT:    vsubps %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:1.00]
+; SANDY-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [11:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: v4f32_one_step2:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
-; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01]
-; HASWELL-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm0
+; HASWELL-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0
+; HASWELL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
+; HASWELL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v4f32_one_step2:
 ; HASWELL-NO-FMA:       # BB#0:
 ; HASWELL-NO-FMA-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %xmm1, %xmm0, %xmm0 # sched: [5:0.50]
-; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01]
 ; HASWELL-NO-FMA-NEXT:    vsubps %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
+; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50]
-; HASWELL-NO-FMA-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NO-FMA-NEXT:    retq # sched: [2:1.00]
 ;
 ; KNL-LABEL: v4f32_one_step2:
 ; KNL:       # BB#0:
 ; KNL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
-; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01]
-; KNL-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm0
+; KNL-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm0 # sched: [5:0.50]
-; KNL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0
+; KNL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0 # sched: [5:0.50]
-; KNL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
+; KNL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50]
-; KNL-NEXT:    retq # sched: [1:1.00]
+; KNL-NEXT:    retq # sched: [2:1.00]
 ;
 ; SKX-LABEL: v4f32_one_step2:
 ; SKX:       # BB#0:
 ; SKX-NEXT:    vrcp14ps %xmm0, %xmm1
 ; SKX-NEXT:    vfnmadd213ps {{.*}}(%rip){1to4}, %xmm1, %xmm0
-; SKX-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0
+; SKX-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0 # sched: [5:0.50]
-; SKX-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
+; SKX-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50]
-; SKX-NEXT:    retq # sched: [1:1.00]
+; SKX-NEXT:    retq # sched: [2:1.00]
   %div = fdiv fast <4 x float> <float 1.0, float 2.0, float 3.0, float 4.0>, %x
   ret <4 x float> %div
 }
@@ -501,56 +501,56 @@ define <4 x float> @v4f32_one_step_2_divs(<4 x float> %x) #1 {
 ;
 ; SANDY-LABEL: v4f32_one_step_2_divs:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
+; SANDY-NEXT:    vrcpps %xmm0, %xmm1 # sched: [7:3.00]
 ; SANDY-NEXT:    vmulps %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
-; SANDY-NEXT:    vmovaps {{.*#+}} xmm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50]
+; SANDY-NEXT:    vmovaps {{.*#+}} xmm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [6:0.50]
 ; SANDY-NEXT:    vsubps %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [9:1.00]
+; SANDY-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [11:1.00]
 ; SANDY-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: v4f32_one_step_2_divs:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
-; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01]
-; HASWELL-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm0
+; HASWELL-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0
+; HASWELL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [9:0.50]
+; HASWELL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [5:0.50]
 ; HASWELL-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v4f32_one_step_2_divs:
 ; HASWELL-NO-FMA:       # BB#0:
 ; HASWELL-NO-FMA-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %xmm1, %xmm0, %xmm0 # sched: [5:0.50]
-; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01]
 ; HASWELL-NO-FMA-NEXT:    vsubps %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [9:0.50]
+; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
-; HASWELL-NO-FMA-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NO-FMA-NEXT:    retq # sched: [2:1.00]
 ;
 ; KNL-LABEL: v4f32_one_step_2_divs:
 ; KNL:       # BB#0:
 ; KNL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
-; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01]
-; KNL-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm0
+; KNL-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm0 # sched: [5:0.50]
-; KNL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0
+; KNL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0 # sched: [5:0.50]
-; KNL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [9:0.50]
+; KNL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [5:0.50]
 ; KNL-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
-; KNL-NEXT:    retq # sched: [1:1.00]
+; KNL-NEXT:    retq # sched: [2:1.00]
 ;
 ; SKX-LABEL: v4f32_one_step_2_divs:
 ; SKX:       # BB#0:
 ; SKX-NEXT:    vrcp14ps %xmm0, %xmm1
 ; SKX-NEXT:    vfnmadd213ps {{.*}}(%rip){1to4}, %xmm1, %xmm0
-; SKX-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0
+; SKX-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0 # sched: [5:0.50]
-; SKX-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [9:0.50]
+; SKX-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [5:0.50]
 ; SKX-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
-; SKX-NEXT:    retq # sched: [1:1.00]
+; SKX-NEXT:    retq # sched: [2:1.00]
   %div = fdiv fast <4 x float> <float 1.0, float 2.0, float 3.0, float 4.0>, %x
   %div2 = fdiv fast <4 x float> %div, %x
   ret <4 x float> %div2
@@ -619,9 +619,9 @@ define <4 x float> @v4f32_two_step2(<4 x float> %x) #2 {
 ;
 ; SANDY-LABEL: v4f32_two_step2:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
+; SANDY-NEXT:    vrcpps %xmm0, %xmm1 # sched: [7:3.00]
 ; SANDY-NEXT:    vmulps %xmm1, %xmm0, %xmm2 # sched: [5:1.00]
-; SANDY-NEXT:    vmovaps {{.*#+}} xmm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50]
+; SANDY-NEXT:    vmovaps {{.*#+}} xmm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [6:0.50]
 ; SANDY-NEXT:    vsubps %xmm2, %xmm3, %xmm2 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps %xmm2, %xmm1, %xmm2 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddps %xmm2, %xmm1, %xmm1 # sched: [3:1.00]
@@ -629,26 +629,26 @@ define <4 x float> @v4f32_two_step2(<4 x float> %x) #2 {
 ; SANDY-NEXT:    vsubps %xmm0, %xmm3, %xmm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:1.00]
+; SANDY-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [11:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: v4f32_two_step2:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
-; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01]
 ; HASWELL-NEXT:    vmovaps %xmm1, %xmm3 # sched: [1:1.00]
-; HASWELL-NEXT:    vfnmadd213ps %xmm2, %xmm0, %xmm3
+; HASWELL-NEXT:    vfnmadd213ps %xmm2, %xmm0, %xmm3 # sched: [5:0.50]
-; HASWELL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm3
+; HASWELL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm3 # sched: [5:0.50]
-; HASWELL-NEXT:    vfnmadd213ps %xmm2, %xmm3, %xmm0
+; HASWELL-NEXT:    vfnmadd213ps %xmm2, %xmm3, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vfmadd132ps %xmm3, %xmm3, %xmm0
+; HASWELL-NEXT:    vfmadd132ps %xmm3, %xmm3, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
+; HASWELL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v4f32_two_step2:
 ; HASWELL-NO-FMA:       # BB#0:
 ; HASWELL-NO-FMA-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %xmm1, %xmm0, %xmm2 # sched: [5:0.50]
-; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*}}(%rip), %xmm3 # sched: [4:0.50]
+; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*}}(%rip), %xmm3 # sched: [?:5.000000e-01]
 ; HASWELL-NO-FMA-NEXT:    vsubps %xmm2, %xmm3, %xmm2 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %xmm2, %xmm1, %xmm2 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    vaddps %xmm2, %xmm1, %xmm1 # sched: [3:1.00]
@@ -656,32 +656,32 @@ define <4 x float> @v4f32_two_step2(<4 x float> %x) #2 {
 ; HASWELL-NO-FMA-NEXT:    vsubps %xmm0, %xmm3, %xmm0 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
+; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50]
-; HASWELL-NO-FMA-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NO-FMA-NEXT:    retq # sched: [2:1.00]
 ;
 ; KNL-LABEL: v4f32_two_step2:
 ; KNL:       # BB#0:
 ; KNL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
-; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01]
 ; KNL-NEXT:    vmovaps %xmm1, %xmm3 # sched: [1:1.00]
-; KNL-NEXT:    vfnmadd213ps %xmm2, %xmm0, %xmm3
+; KNL-NEXT:    vfnmadd213ps %xmm2, %xmm0, %xmm3 # sched: [5:0.50]
-; KNL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm3
+; KNL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm3 # sched: [5:0.50]
-; KNL-NEXT:    vfnmadd213ps %xmm2, %xmm3, %xmm0
+; KNL-NEXT:    vfnmadd213ps %xmm2, %xmm3, %xmm0 # sched: [5:0.50]
-; KNL-NEXT:    vfmadd132ps %xmm3, %xmm3, %xmm0
+; KNL-NEXT:    vfmadd132ps %xmm3, %xmm3, %xmm0 # sched: [5:0.50]
-; KNL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
+; KNL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50]
-; KNL-NEXT:    retq # sched: [1:1.00]
+; KNL-NEXT:    retq # sched: [2:1.00]
 ;
 ; SKX-LABEL: v4f32_two_step2:
 ; SKX:       # BB#0:
 ; SKX-NEXT:    vrcp14ps %xmm0, %xmm1
-; SKX-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; SKX-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01]
 ; SKX-NEXT:    vmovaps %xmm1, %xmm3 # sched: [1:1.00]
-; SKX-NEXT:    vfnmadd213ps %xmm2, %xmm0, %xmm3
+; SKX-NEXT:    vfnmadd213ps %xmm2, %xmm0, %xmm3 # sched: [5:0.50]
-; SKX-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm3
+; SKX-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm3 # sched: [5:0.50]
-; SKX-NEXT:    vfnmadd213ps %xmm2, %xmm3, %xmm0
+; SKX-NEXT:    vfnmadd213ps %xmm2, %xmm3, %xmm0 # sched: [5:0.50]
-; SKX-NEXT:    vfmadd132ps %xmm3, %xmm3, %xmm0
+; SKX-NEXT:    vfmadd132ps %xmm3, %xmm3, %xmm0 # sched: [5:0.50]
-; SKX-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
+; SKX-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50]
-; SKX-NEXT:    retq # sched: [1:1.00]
+; SKX-NEXT:    retq # sched: [2:1.00]
   %div = fdiv fast <4 x float> <float 1.0, float 2.0, float 3.0, float 4.0>, %x
   ret <4 x float> %div
 }
@@ -741,49 +741,49 @@ define <8 x float> @v8f32_one_step2(<8 x float> %x) #1 {
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vrcpps %ymm0, %ymm1 # sched: [5:1.00]
 ; SANDY-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; SANDY-NEXT:    vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50]
+; SANDY-NEXT:    vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50]
 ; SANDY-NEXT:    vsubps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
+; SANDY-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:2.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: v8f32_one_step2:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
+; HASWELL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [11:2.00]
 ; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
-; HASWELL-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm0
+; HASWELL-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0
+; HASWELL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
+; HASWELL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v8f32_one_step2:
 ; HASWELL-NO-FMA:       # BB#0:
-; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
+; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm1 # sched: [11:2.00]
-; HASWELL-NO-FMA-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
+; HASWELL-NO-FMA-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
 ; HASWELL-NO-FMA-NEXT:    vsubps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
-; HASWELL-NO-FMA-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
+; HASWELL-NO-FMA-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
+; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50]
-; HASWELL-NO-FMA-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NO-FMA-NEXT:    retq # sched: [2:1.00]
 ;
 ; KNL-LABEL: v8f32_one_step2:
 ; KNL:       # BB#0:
-; KNL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
+; KNL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [11:2.00]
 ; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
-; KNL-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm0
+; KNL-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm0 # sched: [5:0.50]
-; KNL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0
+; KNL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0 # sched: [5:0.50]
-; KNL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
+; KNL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50]
-; KNL-NEXT:    retq # sched: [1:1.00]
+; KNL-NEXT:    retq # sched: [2:1.00]
 ;
 ; SKX-LABEL: v8f32_one_step2:
 ; SKX:       # BB#0:
 ; SKX-NEXT:    vrcp14ps %ymm0, %ymm1
 ; SKX-NEXT:    vfnmadd213ps {{.*}}(%rip){1to8}, %ymm1, %ymm0
-; SKX-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0
+; SKX-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0 # sched: [5:0.50]
-; SKX-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
+; SKX-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50]
-; SKX-NEXT:    retq # sched: [1:1.00]
+; SKX-NEXT:    retq # sched: [2:1.00]
   %div = fdiv fast <8 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0>, %x
   ret <8 x float> %div
 }
@@ -848,54 +848,54 @@ define <8 x float> @v8f32_one_step_2_divs(<8 x float> %x) #1 {
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vrcpps %ymm0, %ymm1 # sched: [5:1.00]
 ; SANDY-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; SANDY-NEXT:    vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50]
+; SANDY-NEXT:    vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50]
 ; SANDY-NEXT:    vsubps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [9:1.00]
+; SANDY-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [12:2.00]
 ; SANDY-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: v8f32_one_step_2_divs:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
+; HASWELL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [11:2.00]
 ; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
-; HASWELL-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm0
+; HASWELL-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0
+; HASWELL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [9:1.00]
+; HASWELL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [5:0.50]
-; HASWELL-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
+; HASWELL-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v8f32_one_step_2_divs:
 ; HASWELL-NO-FMA:       # BB#0:
-; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
+; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm1 # sched: [11:2.00]
-; HASWELL-NO-FMA-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
+; HASWELL-NO-FMA-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
 ; HASWELL-NO-FMA-NEXT:    vsubps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
-; HASWELL-NO-FMA-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
+; HASWELL-NO-FMA-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [9:1.00]
+; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [5:0.50]
-; HASWELL-NO-FMA-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
+; HASWELL-NO-FMA-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:0.50]
-; HASWELL-NO-FMA-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NO-FMA-NEXT:    retq # sched: [2:1.00]
 ;
 ; KNL-LABEL: v8f32_one_step_2_divs:
 ; KNL:       # BB#0:
-; KNL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
+; KNL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [11:2.00]
 ; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
-; KNL-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm0
+; KNL-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm0 # sched: [5:0.50]
-; KNL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0
+; KNL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0 # sched: [5:0.50]
-; KNL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [9:1.00]
+; KNL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [5:0.50]
-; KNL-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
+; KNL-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:0.50]
-; KNL-NEXT:    retq # sched: [1:1.00]
+; KNL-NEXT:    retq # sched: [2:1.00]
 ;
 ; SKX-LABEL: v8f32_one_step_2_divs:
 ; SKX:       # BB#0:
 ; SKX-NEXT:    vrcp14ps %ymm0, %ymm1
 ; SKX-NEXT:    vfnmadd213ps {{.*}}(%rip){1to8}, %ymm1, %ymm0
-; SKX-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0
+; SKX-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0 # sched: [5:0.50]
-; SKX-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [9:1.00]
+; SKX-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [5:0.50]
-; SKX-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
+; SKX-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:0.50]
-; SKX-NEXT:    retq # sched: [1:1.00]
+; SKX-NEXT:    retq # sched: [2:1.00]
   %div = fdiv fast <8 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0>, %x
   %div2 = fdiv fast <8 x float> %div, %x
   ret <8 x float> %div2
@@ -980,7 +980,7 @@ define <8 x float> @v8f32_two_step2(<8 x float> %x) #2 {
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vrcpps %ymm0, %ymm1 # sched: [5:1.00]
 ; SANDY-NEXT:    vmulps %ymm1, %ymm0, %ymm2 # sched: [5:1.00]
-; SANDY-NEXT:    vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50]
+; SANDY-NEXT:    vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50]
 ; SANDY-NEXT:    vsubps %ymm2, %ymm3, %ymm2 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps %ymm2, %ymm1, %ymm2 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddps %ymm2, %ymm1, %ymm1 # sched: [3:1.00]
@@ -988,59 +988,59 @@ define <8 x float> @v8f32_two_step2(<8 x float> %x) #2 {
 ; SANDY-NEXT:    vsubps %ymm0, %ymm3, %ymm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
+; SANDY-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:2.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: v8f32_two_step2:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
+; HASWELL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [11:2.00]
 ; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
 ; HASWELL-NEXT:    vmovaps %ymm1, %ymm3 # sched: [1:1.00]
-; HASWELL-NEXT:    vfnmadd213ps %ymm2, %ymm0, %ymm3
+; HASWELL-NEXT:    vfnmadd213ps %ymm2, %ymm0, %ymm3 # sched: [5:0.50]
-; HASWELL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm3
+; HASWELL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm3 # sched: [5:0.50]
-; HASWELL-NEXT:    vfnmadd213ps %ymm2, %ymm3, %ymm0
+; HASWELL-NEXT:    vfnmadd213ps %ymm2, %ymm3, %ymm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vfmadd132ps %ymm3, %ymm3, %ymm0
+; HASWELL-NEXT:    vfmadd132ps %ymm3, %ymm3, %ymm0 # sched: [5:0.50]
-; HASWELL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
+; HASWELL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v8f32_two_step2:
 ; HASWELL-NO-FMA:       # BB#0:
-; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
+; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm1 # sched: [11:2.00]
-; HASWELL-NO-FMA-NEXT:    vmulps %ymm1, %ymm0, %ymm2 # sched: [5:1.00]
+; HASWELL-NO-FMA-NEXT:    vmulps %ymm1, %ymm0, %ymm2 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*}}(%rip), %ymm3 # sched: [5:1.00]
 ; HASWELL-NO-FMA-NEXT:    vsubps %ymm2, %ymm3, %ymm2 # sched: [3:1.00]
-; HASWELL-NO-FMA-NEXT:    vmulps %ymm2, %ymm1, %ymm2 # sched: [5:1.00]
+; HASWELL-NO-FMA-NEXT:    vmulps %ymm2, %ymm1, %ymm2 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    vaddps %ymm2, %ymm1, %ymm1 # sched: [3:1.00]
-; HASWELL-NO-FMA-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
+; HASWELL-NO-FMA-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    vsubps %ymm0, %ymm3, %ymm0 # sched: [3:1.00]
-; HASWELL-NO-FMA-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
+; HASWELL-NO-FMA-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
+; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50]
-; HASWELL-NO-FMA-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NO-FMA-NEXT:    retq # sched: [2:1.00]
 ;
 ; KNL-LABEL: v8f32_two_step2:
 ; KNL:       # BB#0:
-; KNL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
+; KNL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [11:2.00]
 ; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
 ; KNL-NEXT:    vmovaps %ymm1, %ymm3 # sched: [1:1.00]
-; KNL-NEXT:    vfnmadd213ps %ymm2, %ymm0, %ymm3
+; KNL-NEXT:    vfnmadd213ps %ymm2, %ymm0, %ymm3 # sched: [5:0.50]
-; KNL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm3
+; KNL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm3 # sched: [5:0.50]
-; KNL-NEXT:    vfnmadd213ps %ymm2, %ymm3, %ymm0
+; KNL-NEXT:    vfnmadd213ps %ymm2, %ymm3, %ymm0 # sched: [5:0.50]
-; KNL-NEXT:    vfmadd132ps %ymm3, %ymm3, %ymm0
+; KNL-NEXT:    vfmadd132ps %ymm3, %ymm3, %ymm0 # sched: [5:0.50]
-; KNL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
+; KNL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50]
-; KNL-NEXT:    retq # sched: [1:1.00]
+; KNL-NEXT:    retq # sched: [2:1.00]
 ;
 ; SKX-LABEL: v8f32_two_step2:
 ; SKX:       # BB#0:
 ; SKX-NEXT:    vrcp14ps %ymm0, %ymm1
 ; SKX-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
 ; SKX-NEXT:    vmovaps %ymm1, %ymm3 # sched: [1:1.00]
-; SKX-NEXT:    vfnmadd213ps %ymm2, %ymm0, %ymm3
+; SKX-NEXT:    vfnmadd213ps %ymm2, %ymm0, %ymm3 # sched: [5:0.50]
-; SKX-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm3
+; SKX-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm3 # sched: [5:0.50]
-; SKX-NEXT:    vfnmadd213ps %ymm2, %ymm3, %ymm0
+; SKX-NEXT:    vfnmadd213ps %ymm2, %ymm3, %ymm0 # sched: [5:0.50]
-; SKX-NEXT:    vfmadd132ps %ymm3, %ymm3, %ymm0
+; SKX-NEXT:    vfmadd132ps %ymm3, %ymm3, %ymm0 # sched: [5:0.50]
-; SKX-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
+; SKX-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50]
-; SKX-NEXT:    retq # sched: [1:1.00]
+; SKX-NEXT:    retq # sched: [2:1.00]
   %div = fdiv fast <8 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0>, %x
   ret <8 x float> %div
 }
@@ -1070,27 +1070,27 @@ define <8 x float> @v8f32_no_step(<8 x float> %x) #3 {
 ; SANDY-LABEL: v8f32_no_step:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vrcpps %ymm0, %ymm0 # sched: [5:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: v8f32_no_step:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vrcpps %ymm0, %ymm0 # sched: [7:2.00]
+; HASWELL-NEXT:    vrcpps %ymm0, %ymm0 # sched: [11:2.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v8f32_no_step:
 ; HASWELL-NO-FMA:       # BB#0:
-; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm0 # sched: [7:2.00]
+; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm0 # sched: [11:2.00]
-; HASWELL-NO-FMA-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NO-FMA-NEXT:    retq # sched: [2:1.00]
 ;
 ; KNL-LABEL: v8f32_no_step:
 ; KNL:       # BB#0:
-; KNL-NEXT:    vrcpps %ymm0, %ymm0 # sched: [7:2.00]
+; KNL-NEXT:    vrcpps %ymm0, %ymm0 # sched: [11:2.00]
-; KNL-NEXT:    retq # sched: [1:1.00]
+; KNL-NEXT:    retq # sched: [2:1.00]
 ;
 ; SKX-LABEL: v8f32_no_step:
 ; SKX:       # BB#0:
 ; SKX-NEXT:    vrcp14ps %ymm0, %ymm0
-; SKX-NEXT:    retq # sched: [1:1.00]
+; SKX-NEXT:    retq # sched: [2:1.00]
   %div = fdiv fast <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>, %x
   ret <8 x float> %div
 }
@@ -1125,32 +1125,32 @@ define <8 x float> @v8f32_no_step2(<8 x float> %x) #3 {
 ; SANDY-LABEL: v8f32_no_step2:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vrcpps %ymm0, %ymm0 # sched: [5:1.00]
-; SANDY-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
+; SANDY-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:2.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: v8f32_no_step2:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vrcpps %ymm0, %ymm0 # sched: [7:2.00]
+; HASWELL-NEXT:    vrcpps %ymm0, %ymm0 # sched: [11:2.00]
-; HASWELL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
+; HASWELL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v8f32_no_step2:
 ; HASWELL-NO-FMA:       # BB#0:
-; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm0 # sched: [7:2.00]
+; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm0 # sched: [11:2.00]
-; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
+; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50]
-; HASWELL-NO-FMA-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NO-FMA-NEXT:    retq # sched: [2:1.00]
 ;
 ; KNL-LABEL: v8f32_no_step2:
 ; KNL:       # BB#0:
-; KNL-NEXT:    vrcpps %ymm0, %ymm0 # sched: [7:2.00]
+; KNL-NEXT:    vrcpps %ymm0, %ymm0 # sched: [11:2.00]
-; KNL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
+; KNL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50]
-; KNL-NEXT:    retq # sched: [1:1.00]
+; KNL-NEXT:    retq # sched: [2:1.00]
 ;
 ; SKX-LABEL: v8f32_no_step2:
 ; SKX:       # BB#0:
 ; SKX-NEXT:    vrcp14ps %ymm0, %ymm0
-; SKX-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
+; SKX-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50]
-; SKX-NEXT:    retq # sched: [1:1.00]
+; SKX-NEXT:    retq # sched: [2:1.00]
   %div = fdiv fast <8 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0>, %x
   ret <8 x float> %div
 }

llvm/test/CodeGen/X86/sse3-schedule.ll

@@ -31,14 +31,14 @@ define <2 x double> @test_addsubpd(<2 x double> %a0, <2 x double> %a1, <2 x doub
 ; SANDY-LABEL: test_addsubpd:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vaddsubpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    vaddsubpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
+; SANDY-NEXT:    vaddsubpd (%rdi), %xmm0, %xmm0 # sched: [9:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_addsubpd:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vaddsubpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; HASWELL-NEXT:    vaddsubpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
+; HASWELL-NEXT:    vaddsubpd (%rdi), %xmm0, %xmm0 # sched: [3:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_addsubpd:
 ; BTVER2:       # BB#0:
@@ -74,14 +74,14 @@ define <4 x float> @test_addsubps(<4 x float> %a0, <4 x float> %a1, <4 x float>
 ; SANDY-LABEL: test_addsubps:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vaddsubps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    vaddsubps (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
+; SANDY-NEXT:    vaddsubps (%rdi), %xmm0, %xmm0 # sched: [9:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_addsubps:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vaddsubps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; HASWELL-NEXT:    vaddsubps (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
+; HASWELL-NEXT:    vaddsubps (%rdi), %xmm0, %xmm0 # sched: [3:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_addsubps:
 ; BTVER2:       # BB#0:
@@ -116,15 +116,15 @@ define <2 x double> @test_haddpd(<2 x double> %a0, <2 x double> %a1, <2 x double
 ;
 ; SANDY-LABEL: test_haddpd:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vhaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
+; SANDY-NEXT:    vhaddpd %xmm1, %xmm0, %xmm0 # sched: [5:2.00]
-; SANDY-NEXT:    vhaddpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
+; SANDY-NEXT:    vhaddpd (%rdi), %xmm0, %xmm0 # sched: [11:2.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_haddpd:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vhaddpd %xmm1, %xmm0, %xmm0 # sched: [5:2.00]
-; HASWELL-NEXT:    vhaddpd (%rdi), %xmm0, %xmm0 # sched: [9:2.00]
+; HASWELL-NEXT:    vhaddpd (%rdi), %xmm0, %xmm0 # sched: [5:2.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_haddpd:
 ; BTVER2:       # BB#0:
@@ -159,15 +159,15 @@ define <4 x float> @test_haddps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%
 ;
 ; SANDY-LABEL: test_haddps:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vhaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
+; SANDY-NEXT:    vhaddps %xmm1, %xmm0, %xmm0 # sched: [5:2.00]
-; SANDY-NEXT:    vhaddps (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
+; SANDY-NEXT:    vhaddps (%rdi), %xmm0, %xmm0 # sched: [11:2.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_haddps:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vhaddps %xmm1, %xmm0, %xmm0 # sched: [5:2.00]
-; HASWELL-NEXT:    vhaddps (%rdi), %xmm0, %xmm0 # sched: [9:2.00]
+; HASWELL-NEXT:    vhaddps (%rdi), %xmm0, %xmm0 # sched: [5:2.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_haddps:
 ; BTVER2:       # BB#0:
@@ -202,15 +202,15 @@ define <2 x double> @test_hsubpd(<2 x double> %a0, <2 x double> %a1, <2 x double
 ;
 ; SANDY-LABEL: test_hsubpd:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vhsubpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
+; SANDY-NEXT:    vhsubpd %xmm1, %xmm0, %xmm0 # sched: [5:2.00]
-; SANDY-NEXT:    vhsubpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
+; SANDY-NEXT:    vhsubpd (%rdi), %xmm0, %xmm0 # sched: [11:2.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_hsubpd:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vhsubpd %xmm1, %xmm0, %xmm0 # sched: [5:2.00]
-; HASWELL-NEXT:    vhsubpd (%rdi), %xmm0, %xmm0 # sched: [9:2.00]
+; HASWELL-NEXT:    vhsubpd (%rdi), %xmm0, %xmm0 # sched: [5:2.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_hsubpd:
 ; BTVER2:       # BB#0:
@@ -245,15 +245,15 @@ define <4 x float> @test_hsubps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%
 ;
 ; SANDY-LABEL: test_hsubps:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vhsubps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
+; SANDY-NEXT:    vhsubps %xmm1, %xmm0, %xmm0 # sched: [5:2.00]
-; SANDY-NEXT:    vhsubps (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
+; SANDY-NEXT:    vhsubps (%rdi), %xmm0, %xmm0 # sched: [11:2.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_hsubps:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vhsubps %xmm1, %xmm0, %xmm0 # sched: [5:2.00]
-; HASWELL-NEXT:    vhsubps (%rdi), %xmm0, %xmm0 # sched: [9:2.00]
+; HASWELL-NEXT:    vhsubps (%rdi), %xmm0, %xmm0 # sched: [5:2.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_hsubps:
 ; BTVER2:       # BB#0:
@@ -287,13 +287,13 @@ define <16 x i8> @test_lddqu(i8* %a0) {
 ;
 ; SANDY-LABEL: test_lddqu:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vlddqu (%rdi), %xmm0 # sched: [4:0.50]
+; SANDY-NEXT:    vlddqu (%rdi), %xmm0 # sched: [6:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_lddqu:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vlddqu (%rdi), %xmm0 # sched: [4:0.50]
+; HASWELL-NEXT:    vlddqu (%rdi), %xmm0 # sched: [?:5.000000e-01]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_lddqu:
 ; BTVER2:       # BB#0:
@@ -330,16 +330,16 @@ define <2 x double> @test_movddup(<2 x double> %a0, <2 x double> *%a1) {
 ; SANDY-LABEL: test_movddup:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vmovddup {{.*#+}} xmm0 = xmm0[0,0] sched: [1:1.00]
-; SANDY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0] sched: [4:0.50]
+; SANDY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0] sched: [6:0.50]
 ; SANDY-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_movddup:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vmovddup {{.*#+}} xmm0 = xmm0[0,0] sched: [1:1.00]
-; HASWELL-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0] sched: [4:0.50]
+; HASWELL-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0] sched: [?:5.000000e-01]
 ; HASWELL-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_movddup:
 ; BTVER2:       # BB#0:
@@ -380,16 +380,16 @@ define <4 x float> @test_movshdup(<4 x float> %a0, <4 x float> *%a1) {
 ; SANDY-LABEL: test_movshdup:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vmovshdup {{.*#+}} xmm0 = xmm0[1,1,3,3] sched: [1:1.00]
-; SANDY-NEXT:    vmovshdup {{.*#+}} xmm1 = mem[1,1,3,3] sched: [4:0.50]
+; SANDY-NEXT:    vmovshdup {{.*#+}} xmm1 = mem[1,1,3,3] sched: [6:0.50]
 ; SANDY-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_movshdup:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vmovshdup {{.*#+}} xmm0 = xmm0[1,1,3,3] sched: [1:1.00]
-; HASWELL-NEXT:    vmovshdup {{.*#+}} xmm1 = mem[1,1,3,3] sched: [4:0.50]
+; HASWELL-NEXT:    vmovshdup {{.*#+}} xmm1 = mem[1,1,3,3] sched: [?:5.000000e-01]
 ; HASWELL-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_movshdup:
 ; BTVER2:       # BB#0:
@@ -430,16 +430,16 @@ define <4 x float> @test_movsldup(<4 x float> %a0, <4 x float> *%a1) {
 ; SANDY-LABEL: test_movsldup:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vmovsldup {{.*#+}} xmm0 = xmm0[0,0,2,2] sched: [1:1.00]
-; SANDY-NEXT:    vmovsldup {{.*#+}} xmm1 = mem[0,0,2,2] sched: [4:0.50]
+; SANDY-NEXT:    vmovsldup {{.*#+}} xmm1 = mem[0,0,2,2] sched: [6:0.50]
 ; SANDY-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_movsldup:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vmovsldup {{.*#+}} xmm0 = xmm0[0,0,2,2] sched: [1:1.00]
-; HASWELL-NEXT:    vmovsldup {{.*#+}} xmm1 = mem[0,0,2,2] sched: [4:0.50]
+; HASWELL-NEXT:    vmovsldup {{.*#+}} xmm1 = mem[0,0,2,2] sched: [?:5.000000e-01]
 ; HASWELL-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_movsldup:
 ; BTVER2:       # BB#0:

llvm/test/CodeGen/X86/sse41-schedule.ll

@@ -25,17 +25,17 @@ define <2 x double> @test_blendpd(<2 x double> %a0, <2 x double> %a1, <2 x doubl
 ;
 ; SANDY-LABEL: test_blendpd:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vblendpd {{.*#+}} xmm0 = xmm0[0],xmm1[1] sched: [1:0.50]
+; SANDY-NEXT:    vblendpd {{.*#+}} xmm0 = xmm0[0],xmm1[1] sched: [1:1.00]
 ; SANDY-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    vblendpd {{.*#+}} xmm0 = xmm0[0],mem[1] sched: [5:0.50]
+; SANDY-NEXT:    vblendpd {{.*#+}} xmm0 = xmm0[0],mem[1] sched: [7:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_blendpd:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vblendpd {{.*#+}} xmm0 = xmm0[0],xmm1[1] sched: [1:0.33]
 ; HASWELL-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; HASWELL-NEXT:    vblendpd {{.*#+}} xmm0 = xmm0[0],mem[1] sched: [5:0.50]
+; HASWELL-NEXT:    vblendpd {{.*#+}} xmm0 = xmm0[0],mem[1] sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_blendpd:
 ; BTVER2:       # BB#0:
@@ -65,15 +65,15 @@ define <4 x float> @test_blendps(<4 x float> %a0, <4 x float> %a1, <4 x float> *
 ;
 ; SANDY-LABEL: test_blendps:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1,2],xmm0[3] sched: [1:0.50]
+; SANDY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1,2],xmm0[3] sched: [1:1.00]
-; SANDY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],mem[1],xmm0[2,3] sched: [5:0.50]
+; SANDY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],mem[1],xmm0[2,3] sched: [7:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_blendps:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1,2],xmm0[3] sched: [1:0.33]
-; HASWELL-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],mem[1],xmm0[2,3] sched: [5:0.50]
+; HASWELL-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],mem[1],xmm0[2,3] sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_blendps:
 ; BTVER2:       # BB#0:
@@ -107,15 +107,15 @@ define <2 x double> @test_blendvpd(<2 x double> %a0, <2 x double> %a1, <2 x doub
 ;
 ; SANDY-LABEL: test_blendvpd:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vblendvpd %xmm2, %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
+; SANDY-NEXT:    vblendvpd %xmm2, %xmm1, %xmm0, %xmm0 # sched: [2:2.00]
-; SANDY-NEXT:    vblendvpd %xmm2, (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
+; SANDY-NEXT:    vblendvpd %xmm2, (%rdi), %xmm0, %xmm0 # sched: [8:2.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_blendvpd:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vblendvpd %xmm2, %xmm1, %xmm0, %xmm0 # sched: [2:2.00]
-; HASWELL-NEXT:    vblendvpd %xmm2, (%rdi), %xmm0, %xmm0 # sched: [6:2.00]
+; HASWELL-NEXT:    vblendvpd %xmm2, (%rdi), %xmm0, %xmm0 # sched: [2:2.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_blendvpd:
 ; BTVER2:       # BB#0:
@@ -150,15 +150,15 @@ define <4 x float> @test_blendvps(<4 x float> %a0, <4 x float> %a1, <4 x float>
 ;
 ; SANDY-LABEL: test_blendvps:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vblendvps %xmm2, %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
+; SANDY-NEXT:    vblendvps %xmm2, %xmm1, %xmm0, %xmm0 # sched: [2:2.00]
-; SANDY-NEXT:    vblendvps %xmm2, (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
+; SANDY-NEXT:    vblendvps %xmm2, (%rdi), %xmm0, %xmm0 # sched: [8:2.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_blendvps:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vblendvps %xmm2, %xmm1, %xmm0, %xmm0 # sched: [2:2.00]
-; HASWELL-NEXT:    vblendvps %xmm2, (%rdi), %xmm0, %xmm0 # sched: [6:2.00]
+; HASWELL-NEXT:    vblendvps %xmm2, (%rdi), %xmm0, %xmm0 # sched: [2:2.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_blendvps:
 ; BTVER2:       # BB#0:
@@ -187,15 +187,15 @@ define <2 x double> @test_dppd(<2 x double> %a0, <2 x double> %a1, <2 x double>
 ;
 ; SANDY-LABEL: test_dppd:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vdppd $7, %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
+; SANDY-NEXT:    vdppd $7, %xmm1, %xmm0, %xmm0 # sched: [9:1.00]
-; SANDY-NEXT:    vdppd $7, (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
+; SANDY-NEXT:    vdppd $7, (%rdi), %xmm0, %xmm0 # sched: [15:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_dppd:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vdppd $7, %xmm1, %xmm0, %xmm0 # sched: [9:1.00]
-; HASWELL-NEXT:    vdppd $7, (%rdi), %xmm0, %xmm0 # sched: [13:1.00]
+; HASWELL-NEXT:    vdppd $7, (%rdi), %xmm0, %xmm0 # sched: [9:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_dppd:
 ; BTVER2:       # BB#0:
@@ -224,15 +224,15 @@ define <4 x float> @test_dpps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a2
 ;
 ; SANDY-LABEL: test_dpps:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vdpps $7, %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
+; SANDY-NEXT:    vdpps $7, %xmm1, %xmm0, %xmm0 # sched: [12:2.00]
 ; SANDY-NEXT:    vdpps $7, (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_dpps:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vdpps $7, %xmm1, %xmm0, %xmm0 # sched: [14:2.00]
-; HASWELL-NEXT:    vdpps $7, (%rdi), %xmm0, %xmm0 # sched: [18:2.00]
+; HASWELL-NEXT:    vdpps $7, (%rdi), %xmm0, %xmm0 # sched: [14:2.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_dpps:
 ; BTVER2:       # BB#0:
@@ -262,14 +262,14 @@ define <4 x float> @test_insertps(<4 x float> %a0, <4 x float> %a1, float *%a2)
 ; SANDY-LABEL: test_insertps:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vinsertps {{.*#+}} xmm0 = zero,xmm1[0],xmm0[2,3] sched: [1:1.00]
-; SANDY-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0] sched: [5:1.00]
+; SANDY-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0] sched: [7:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_insertps:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vinsertps {{.*#+}} xmm0 = zero,xmm1[0],xmm0[2,3] sched: [1:1.00]
-; HASWELL-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0] sched: [5:1.00]
+; HASWELL-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0] sched: [1:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_insertps:
 ; BTVER2:       # BB#0:
@@ -296,13 +296,13 @@ define <2 x i64> @test_movntdqa(i8* %a0) {
 ;
 ; SANDY-LABEL: test_movntdqa:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vmovntdqa (%rdi), %xmm0 # sched: [4:0.50]
+; SANDY-NEXT:    vmovntdqa (%rdi), %xmm0 # sched: [6:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_movntdqa:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vmovntdqa (%rdi), %xmm0 # sched: [4:0.50]
+; HASWELL-NEXT:    vmovntdqa (%rdi), %xmm0 # sched: [?:5.000000e-01]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_movntdqa:
 ; BTVER2:       # BB#0:
@@ -328,15 +328,15 @@ define <8 x i16> @test_mpsadbw(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {
 ;
 ; SANDY-LABEL: test_mpsadbw:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vmpsadbw $7, %xmm1, %xmm0, %xmm0 # sched: [6:1.00]
+; SANDY-NEXT:    vmpsadbw $7, %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
-; SANDY-NEXT:    vmpsadbw $7, (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
+; SANDY-NEXT:    vmpsadbw $7, (%rdi), %xmm0, %xmm0 # sched: [11:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_mpsadbw:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vmpsadbw $7, %xmm1, %xmm0, %xmm0 # sched: [6:2.00]
+; HASWELL-NEXT:    vmpsadbw $7, %xmm1, %xmm0, %xmm0 # sched: [7:2.00]
-; HASWELL-NEXT:    vmpsadbw $7, (%rdi), %xmm0, %xmm0 # sched: [6:2.00]
+; HASWELL-NEXT:    vmpsadbw $7, (%rdi), %xmm0, %xmm0 # sched: [7:2.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_mpsadbw:
 ; BTVER2:       # BB#0:
@@ -367,14 +367,14 @@ define <8 x i16> @test_packusdw(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
 ; SANDY-LABEL: test_packusdw:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vpackusdw (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vpackusdw (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_packusdw:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
-; HASWELL-NEXT:    vpackusdw (%rdi), %xmm0, %xmm0 # sched: [5:1.00]
+; HASWELL-NEXT:    vpackusdw (%rdi), %xmm0, %xmm0 # sched: [1:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_packusdw:
 ; BTVER2:       # BB#0:
@@ -411,14 +411,14 @@ define <16 x i8> @test_pblendvb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> %a2, <16
 ; SANDY-LABEL: test_pblendvb:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpblendvb %xmm2, %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
-; SANDY-NEXT:    vpblendvb %xmm2, (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
+; SANDY-NEXT:    vpblendvb %xmm2, (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pblendvb:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpblendvb %xmm2, %xmm1, %xmm0, %xmm0 # sched: [2:2.00]
-; HASWELL-NEXT:    vpblendvb %xmm2, (%rdi), %xmm0, %xmm0 # sched: [6:2.00]
+; HASWELL-NEXT:    vpblendvb %xmm2, (%rdi), %xmm0, %xmm0 # sched: [2:2.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pblendvb:
 ; BTVER2:       # BB#0:
@@ -448,14 +448,14 @@ define <8 x i16> @test_pblendw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {
 ; SANDY-LABEL: test_pblendw:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4],xmm1[5],xmm0[6],xmm1[7] sched: [1:0.50]
-; SANDY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],mem[2,3],xmm0[4,5,6],mem[7] sched: [5:0.50]
+; SANDY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],mem[2,3],xmm0[4,5,6],mem[7] sched: [7:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pblendw:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4],xmm1[5],xmm0[6],xmm1[7] sched: [1:1.00]
-; HASWELL-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],mem[2,3],xmm0[4,5,6],mem[7] sched: [4:1.00]
+; HASWELL-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],mem[2,3],xmm0[4,5,6],mem[7] sched: [1:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pblendw:
 ; BTVER2:       # BB#0:
@@ -484,14 +484,14 @@ define <2 x i64> @test_pcmpeqq(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) {
 ; SANDY-LABEL: test_pcmpeqq:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vpcmpeqq (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vpcmpeqq (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pcmpeqq:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    vpcmpeqq (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; HASWELL-NEXT:    vpcmpeqq (%rdi), %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pcmpeqq:
 ; BTVER2:       # BB#0:
@@ -521,15 +521,15 @@ define i32 @test_pextrb(<16 x i8> %a0, i8 *%a1) {
 ;
 ; SANDY-LABEL: test_pextrb:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vpextrb $3, %xmm0, %eax # sched: [1:0.50]
+; SANDY-NEXT:    vpextrb $3, %xmm0, %eax # sched: [3:1.00]
 ; SANDY-NEXT:    vpextrb $1, %xmm0, (%rdi) # sched: [5:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pextrb:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vpextrb $3, %xmm0, %eax # sched: [1:1.00]
+; HASWELL-NEXT:    vpextrb $3, %xmm0, %eax # sched: [2:1.00]
-; HASWELL-NEXT:    vpextrb $1, %xmm0, (%rdi) # sched: [5:1.00]
+; HASWELL-NEXT:    vpextrb $1, %xmm0, (%rdi) # sched: [1:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pextrb:
 ; BTVER2:       # BB#0:
@@ -558,15 +558,15 @@ define i32 @test_pextrd(<4 x i32> %a0, i32 *%a1) {
 ;
 ; SANDY-LABEL: test_pextrd:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vpextrd $3, %xmm0, %eax # sched: [1:0.50]
+; SANDY-NEXT:    vpextrd $3, %xmm0, %eax # sched: [3:1.00]
 ; SANDY-NEXT:    vpextrd $1, %xmm0, (%rdi) # sched: [5:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pextrd:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vpextrd $3, %xmm0, %eax # sched: [1:1.00]
+; HASWELL-NEXT:    vpextrd $3, %xmm0, %eax # sched: [2:1.00]
-; HASWELL-NEXT:    vpextrd $1, %xmm0, (%rdi) # sched: [5:1.00]
+; HASWELL-NEXT:    vpextrd $1, %xmm0, (%rdi) # sched: [1:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pextrd:
 ; BTVER2:       # BB#0:
@@ -594,15 +594,15 @@ define i64 @test_pextrq(<2 x i64> %a0, <2 x i64> %a1, i64 *%a2) {
 ;
 ; SANDY-LABEL: test_pextrq:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vpextrq $1, %xmm0, %rax # sched: [1:0.50]
+; SANDY-NEXT:    vpextrq $1, %xmm0, %rax # sched: [3:1.00]
 ; SANDY-NEXT:    vpextrq $1, %xmm0, (%rdi) # sched: [5:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pextrq:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vpextrq $1, %xmm0, %rax # sched: [1:1.00]
+; HASWELL-NEXT:    vpextrq $1, %xmm0, %rax # sched: [2:1.00]
-; HASWELL-NEXT:    vpextrq $1, %xmm0, (%rdi) # sched: [5:1.00]
+; HASWELL-NEXT:    vpextrq $1, %xmm0, (%rdi) # sched: [1:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pextrq:
 ; BTVER2:       # BB#0:
@@ -630,15 +630,15 @@ define i32 @test_pextrw(<8 x i16> %a0, i16 *%a1) {
 ;
 ; SANDY-LABEL: test_pextrw:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vpextrw $3, %xmm0, %eax # sched: [1:0.50]
+; SANDY-NEXT:    vpextrw $3, %xmm0, %eax # sched: [3:1.00]
 ; SANDY-NEXT:    vpextrw $1, %xmm0, (%rdi) # sched: [5:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pextrw:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vpextrw $3, %xmm0, %eax # sched: [1:1.00]
+; HASWELL-NEXT:    vpextrw $3, %xmm0, %eax # sched: [2:1.00]
-; HASWELL-NEXT:    vpextrw $1, %xmm0, (%rdi) # sched: [5:1.00]
+; HASWELL-NEXT:    vpextrw $1, %xmm0, (%rdi) # sched: [1:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pextrw:
 ; BTVER2:       # BB#0:
@@ -667,15 +667,15 @@ define <8 x i16> @test_phminposuw(<8 x i16> *%a0) {
 ;
 ; SANDY-LABEL: test_phminposuw:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vphminposuw (%rdi), %xmm0 # sched: [9:1.00]
+; SANDY-NEXT:    vphminposuw (%rdi), %xmm0 # sched: [11:1.00]
 ; SANDY-NEXT:    vphminposuw %xmm0, %xmm0 # sched: [5:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_phminposuw:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vphminposuw (%rdi), %xmm0 # sched: [9:1.00]
+; HASWELL-NEXT:    vphminposuw (%rdi), %xmm0 # sched: [5:1.00]
 ; HASWELL-NEXT:    vphminposuw %xmm0, %xmm0 # sched: [5:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_phminposuw:
 ; BTVER2:       # BB#0:
@@ -704,15 +704,15 @@ define <16 x i8> @test_pinsrb(<16 x i8> %a0, i8 %a1, i8 *%a2) {
 ;
 ; SANDY-LABEL: test_pinsrb:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vpinsrb $1, %edi, %xmm0, %xmm0 # sched: [1:0.50]
+; SANDY-NEXT:    vpinsrb $1, %edi, %xmm0, %xmm0 # sched: [2:1.00]
-; SANDY-NEXT:    vpinsrb $3, (%rsi), %xmm0, %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vpinsrb $3, (%rsi), %xmm0, %xmm0 # sched: [7:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pinsrb:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vpinsrb $1, %edi, %xmm0, %xmm0 # sched: [1:1.00]
+; HASWELL-NEXT:    vpinsrb $1, %edi, %xmm0, %xmm0 # sched: [2:2.00]
-; HASWELL-NEXT:    vpinsrb $3, (%rsi), %xmm0, %xmm0 # sched: [5:1.00]
+; HASWELL-NEXT:    vpinsrb $3, (%rsi), %xmm0, %xmm0 # sched: [1:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pinsrb:
 ; BTVER2:       # BB#0:
@@ -740,15 +740,15 @@ define <4 x i32> @test_pinsrd(<4 x i32> %a0, i32 %a1, i32 *%a2) {
 ;
 ; SANDY-LABEL: test_pinsrd:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vpinsrd $1, %edi, %xmm0, %xmm0 # sched: [1:0.50]
+; SANDY-NEXT:    vpinsrd $1, %edi, %xmm0, %xmm0 # sched: [2:1.00]
-; SANDY-NEXT:    vpinsrd $3, (%rsi), %xmm0, %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vpinsrd $3, (%rsi), %xmm0, %xmm0 # sched: [7:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pinsrd:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vpinsrd $1, %edi, %xmm0, %xmm0 # sched: [1:1.00]
+; HASWELL-NEXT:    vpinsrd $1, %edi, %xmm0, %xmm0 # sched: [2:2.00]
-; HASWELL-NEXT:    vpinsrd $3, (%rsi), %xmm0, %xmm0 # sched: [5:1.00]
+; HASWELL-NEXT:    vpinsrd $3, (%rsi), %xmm0, %xmm0 # sched: [1:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pinsrd:
 ; BTVER2:       # BB#0:
@@ -778,17 +778,17 @@ define <2 x i64> @test_pinsrq(<2 x i64> %a0, <2 x i64> %a1, i64 %a2, i64 *%a3) {
 ;
 ; SANDY-LABEL: test_pinsrq:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vpinsrq $1, %rdi, %xmm0, %xmm0 # sched: [1:0.50]
+; SANDY-NEXT:    vpinsrq $1, %rdi, %xmm0, %xmm0 # sched: [2:1.00]
-; SANDY-NEXT:    vpinsrq $1, (%rsi), %xmm1, %xmm1 # sched: [5:0.50]
+; SANDY-NEXT:    vpinsrq $1, (%rsi), %xmm1, %xmm1 # sched: [7:0.50]
 ; SANDY-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pinsrq:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vpinsrq $1, %rdi, %xmm0, %xmm0 # sched: [1:1.00]
+; HASWELL-NEXT:    vpinsrq $1, %rdi, %xmm0, %xmm0 # sched: [2:2.00]
-; HASWELL-NEXT:    vpinsrq $1, (%rsi), %xmm1, %xmm1 # sched: [5:1.00]
+; HASWELL-NEXT:    vpinsrq $1, (%rsi), %xmm1, %xmm1 # sched: [1:1.00]
 ; HASWELL-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pinsrq:
 ; BTVER2:       # BB#0:
@@ -819,14 +819,14 @@ define <16 x i8> @test_pmaxsb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {
 ; SANDY-LABEL: test_pmaxsb:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpmaxsb %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vpmaxsb (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vpmaxsb (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pmaxsb:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpmaxsb %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    vpmaxsb (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; HASWELL-NEXT:    vpmaxsb (%rdi), %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pmaxsb:
 ; BTVER2:       # BB#0:
@@ -856,14 +856,14 @@ define <4 x i32> @test_pmaxsd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
 ; SANDY-LABEL: test_pmaxsd:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpmaxsd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vpmaxsd (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vpmaxsd (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pmaxsd:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpmaxsd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    vpmaxsd (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; HASWELL-NEXT:    vpmaxsd (%rdi), %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pmaxsd:
 ; BTVER2:       # BB#0:
@@ -893,14 +893,14 @@ define <4 x i32> @test_pmaxud(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
 ; SANDY-LABEL: test_pmaxud:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpmaxud %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vpmaxud (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vpmaxud (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pmaxud:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpmaxud %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    vpmaxud (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; HASWELL-NEXT:    vpmaxud (%rdi), %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pmaxud:
 ; BTVER2:       # BB#0:
@@ -930,14 +930,14 @@ define <8 x i16> @test_pmaxuw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {
 ; SANDY-LABEL: test_pmaxuw:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpmaxuw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vpmaxuw (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vpmaxuw (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pmaxuw:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpmaxuw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    vpmaxuw (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; HASWELL-NEXT:    vpmaxuw (%rdi), %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pmaxuw:
 ; BTVER2:       # BB#0:
@@ -967,14 +967,14 @@ define <16 x i8> @test_pminsb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {
 ; SANDY-LABEL: test_pminsb:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpminsb %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vpminsb (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vpminsb (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pminsb:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpminsb %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    vpminsb (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; HASWELL-NEXT:    vpminsb (%rdi), %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pminsb:
 ; BTVER2:       # BB#0:
@@ -1004,14 +1004,14 @@ define <4 x i32> @test_pminsd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
 ; SANDY-LABEL: test_pminsd:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpminsd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vpminsd (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vpminsd (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pminsd:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpminsd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    vpminsd (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; HASWELL-NEXT:    vpminsd (%rdi), %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pminsd:
 ; BTVER2:       # BB#0:
@@ -1041,14 +1041,14 @@ define <4 x i32> @test_pminud(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
 ; SANDY-LABEL: test_pminud:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpminud %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vpminud (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vpminud (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pminud:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpminud %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    vpminud (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; HASWELL-NEXT:    vpminud (%rdi), %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pminud:
 ; BTVER2:       # BB#0:
@@ -1078,14 +1078,14 @@ define <8 x i16> @test_pminuw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {
 ; SANDY-LABEL: test_pminuw:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpminuw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vpminuw (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vpminuw (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pminuw:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpminuw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    vpminuw (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; HASWELL-NEXT:    vpminuw (%rdi), %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pminuw:
 ; BTVER2:       # BB#0:
@@ -1118,16 +1118,16 @@ define <8 x i16> @test_pmovsxbw(<16 x i8> %a0, <8 x i8> *%a1) {
 ; SANDY-LABEL: test_pmovsxbw:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpmovsxbw %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vpmovsxbw (%rdi), %xmm1 # sched: [5:0.50]
+; SANDY-NEXT:    vpmovsxbw (%rdi), %xmm1 # sched: [7:0.50]
 ; SANDY-NEXT:    vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pmovsxbw:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpmovsxbw %xmm0, %xmm0 # sched: [1:1.00]
-; HASWELL-NEXT:    vpmovsxbw (%rdi), %xmm1 # sched: [5:1.00]
+; HASWELL-NEXT:    vpmovsxbw (%rdi), %xmm1 # sched: [1:1.00]
 ; HASWELL-NEXT:    vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pmovsxbw:
 ; BTVER2:       # BB#0:
@@ -1162,16 +1162,16 @@ define <4 x i32> @test_pmovsxbd(<16 x i8> %a0, <4 x i8> *%a1) {
 ; SANDY-LABEL: test_pmovsxbd:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpmovsxbd %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vpmovsxbd (%rdi), %xmm1 # sched: [5:0.50]
+; SANDY-NEXT:    vpmovsxbd (%rdi), %xmm1 # sched: [7:0.50]
 ; SANDY-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pmovsxbd:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpmovsxbd %xmm0, %xmm0 # sched: [1:1.00]
-; HASWELL-NEXT:    vpmovsxbd (%rdi), %xmm1 # sched: [5:1.00]
+; HASWELL-NEXT:    vpmovsxbd (%rdi), %xmm1 # sched: [1:1.00]
 ; HASWELL-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pmovsxbd:
 ; BTVER2:       # BB#0:
@@ -1206,16 +1206,16 @@ define <2 x i64> @test_pmovsxbq(<16 x i8> %a0, <2 x i8> *%a1) {
 ; SANDY-LABEL: test_pmovsxbq:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpmovsxbq %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vpmovsxbq (%rdi), %xmm1 # sched: [5:0.50]
+; SANDY-NEXT:    vpmovsxbq (%rdi), %xmm1 # sched: [7:0.50]
 ; SANDY-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pmovsxbq:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpmovsxbq %xmm0, %xmm0 # sched: [1:1.00]
-; HASWELL-NEXT:    vpmovsxbq (%rdi), %xmm1 # sched: [5:1.00]
+; HASWELL-NEXT:    vpmovsxbq (%rdi), %xmm1 # sched: [1:1.00]
 ; HASWELL-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pmovsxbq:
 ; BTVER2:       # BB#0:
@@ -1250,16 +1250,16 @@ define <2 x i64> @test_pmovsxdq(<4 x i32> %a0, <2 x i32> *%a1) {
 ; SANDY-LABEL: test_pmovsxdq:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpmovsxdq %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vpmovsxdq (%rdi), %xmm1 # sched: [5:0.50]
+; SANDY-NEXT:    vpmovsxdq (%rdi), %xmm1 # sched: [7:0.50]
 ; SANDY-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pmovsxdq:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpmovsxdq %xmm0, %xmm0 # sched: [1:1.00]
-; HASWELL-NEXT:    vpmovsxdq (%rdi), %xmm1 # sched: [5:1.00]
+; HASWELL-NEXT:    vpmovsxdq (%rdi), %xmm1 # sched: [1:1.00]
 ; HASWELL-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pmovsxdq:
 ; BTVER2:       # BB#0:
@@ -1294,16 +1294,16 @@ define <4 x i32> @test_pmovsxwd(<8 x i16> %a0, <4 x i16> *%a1) {
 ; SANDY-LABEL: test_pmovsxwd:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpmovsxwd %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vpmovsxwd (%rdi), %xmm1 # sched: [5:0.50]
+; SANDY-NEXT:    vpmovsxwd (%rdi), %xmm1 # sched: [7:0.50]
 ; SANDY-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pmovsxwd:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpmovsxwd %xmm0, %xmm0 # sched: [1:1.00]
-; HASWELL-NEXT:    vpmovsxwd (%rdi), %xmm1 # sched: [5:1.00]
+; HASWELL-NEXT:    vpmovsxwd (%rdi), %xmm1 # sched: [1:1.00]
 ; HASWELL-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pmovsxwd:
 ; BTVER2:       # BB#0:
@@ -1338,16 +1338,16 @@ define <2 x i64> @test_pmovsxwq(<8 x i16> %a0, <2 x i16> *%a1) {
 ; SANDY-LABEL: test_pmovsxwq:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpmovsxwq %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vpmovsxwq (%rdi), %xmm1 # sched: [5:0.50]
+; SANDY-NEXT:    vpmovsxwq (%rdi), %xmm1 # sched: [7:0.50]
 ; SANDY-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pmovsxwq:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpmovsxwq %xmm0, %xmm0 # sched: [1:1.00]
-; HASWELL-NEXT:    vpmovsxwq (%rdi), %xmm1 # sched: [5:1.00]
+; HASWELL-NEXT:    vpmovsxwq (%rdi), %xmm1 # sched: [1:1.00]
 ; HASWELL-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pmovsxwq:
 ; BTVER2:       # BB#0:
@@ -1382,16 +1382,16 @@ define <8 x i16> @test_pmovzxbw(<16 x i8> %a0, <8 x i8> *%a1) {
 ; SANDY-LABEL: test_pmovzxbw:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero sched: [1:0.50]
-; SANDY-NEXT:    vpmovzxbw {{.*#+}} xmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero sched: [5:0.50]
+; SANDY-NEXT:    vpmovzxbw {{.*#+}} xmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero sched: [7:0.50]
 ; SANDY-NEXT:    vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pmovzxbw:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero sched: [1:1.00]
-; HASWELL-NEXT:    vpmovzxbw {{.*#+}} xmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero sched: [5:1.00]
+; HASWELL-NEXT:    vpmovzxbw {{.*#+}} xmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero sched: [1:1.00]
 ; HASWELL-NEXT:    vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pmovzxbw:
 ; BTVER2:       # BB#0:
@@ -1426,16 +1426,16 @@ define <4 x i32> @test_pmovzxbd(<16 x i8> %a0, <4 x i8> *%a1) {
 ; SANDY-LABEL: test_pmovzxbd:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero sched: [1:0.50]
-; SANDY-NEXT:    vpmovzxbd {{.*#+}} xmm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero sched: [5:0.50]
+; SANDY-NEXT:    vpmovzxbd {{.*#+}} xmm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero sched: [7:0.50]
 ; SANDY-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pmovzxbd:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero sched: [1:1.00]
-; HASWELL-NEXT:    vpmovzxbd {{.*#+}} xmm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero sched: [5:1.00]
+; HASWELL-NEXT:    vpmovzxbd {{.*#+}} xmm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero sched: [1:1.00]
 ; HASWELL-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pmovzxbd:
 ; BTVER2:       # BB#0:
@@ -1470,16 +1470,16 @@ define <2 x i64> @test_pmovzxbq(<16 x i8> %a0, <2 x i8> *%a1) {
 ; SANDY-LABEL: test_pmovzxbq:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero sched: [1:0.50]
-; SANDY-NEXT:    vpmovzxbq {{.*#+}} xmm1 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero sched: [5:0.50]
+; SANDY-NEXT:    vpmovzxbq {{.*#+}} xmm1 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero sched: [7:0.50]
 ; SANDY-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pmovzxbq:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero sched: [1:1.00]
-; HASWELL-NEXT:    vpmovzxbq {{.*#+}} xmm1 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero sched: [5:1.00]
+; HASWELL-NEXT:    vpmovzxbq {{.*#+}} xmm1 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero sched: [1:1.00]
 ; HASWELL-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pmovzxbq:
 ; BTVER2:       # BB#0:
@@ -1514,16 +1514,16 @@ define <2 x i64> @test_pmovzxdq(<4 x i32> %a0, <2 x i32> *%a1) {
 ; SANDY-LABEL: test_pmovzxdq:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero sched: [1:0.50]
-; SANDY-NEXT:    vpmovzxdq {{.*#+}} xmm1 = mem[0],zero,mem[1],zero sched: [5:0.50]
+; SANDY-NEXT:    vpmovzxdq {{.*#+}} xmm1 = mem[0],zero,mem[1],zero sched: [7:0.50]
 ; SANDY-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pmovzxdq:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero sched: [1:1.00]
-; HASWELL-NEXT:    vpmovzxdq {{.*#+}} xmm1 = mem[0],zero,mem[1],zero sched: [5:1.00]
+; HASWELL-NEXT:    vpmovzxdq {{.*#+}} xmm1 = mem[0],zero,mem[1],zero sched: [1:1.00]
 ; HASWELL-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pmovzxdq:
 ; BTVER2:       # BB#0:
@@ -1558,16 +1558,16 @@ define <4 x i32> @test_pmovzxwd(<8 x i16> %a0, <4 x i16> *%a1) {
 ; SANDY-LABEL: test_pmovzxwd:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero sched: [1:0.50]
-; SANDY-NEXT:    vpmovzxwd {{.*#+}} xmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero sched: [5:0.50]
+; SANDY-NEXT:    vpmovzxwd {{.*#+}} xmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero sched: [7:0.50]
 ; SANDY-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pmovzxwd:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero sched: [1:1.00]
-; HASWELL-NEXT:    vpmovzxwd {{.*#+}} xmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero sched: [5:1.00]
+; HASWELL-NEXT:    vpmovzxwd {{.*#+}} xmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero sched: [1:1.00]
 ; HASWELL-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pmovzxwd:
 ; BTVER2:       # BB#0:
@@ -1602,16 +1602,16 @@ define <2 x i64> @test_pmovzxwq(<8 x i16> %a0, <2 x i16> *%a1) {
 ; SANDY-LABEL: test_pmovzxwq:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero sched: [1:0.50]
-; SANDY-NEXT:    vpmovzxwq {{.*#+}} xmm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero sched: [5:0.50]
+; SANDY-NEXT:    vpmovzxwq {{.*#+}} xmm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero sched: [7:0.50]
 ; SANDY-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pmovzxwq:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero sched: [1:1.00]
-; HASWELL-NEXT:    vpmovzxwq {{.*#+}} xmm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero sched: [5:1.00]
+; HASWELL-NEXT:    vpmovzxwq {{.*#+}} xmm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero sched: [1:1.00]
 ; HASWELL-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pmovzxwq:
 ; BTVER2:       # BB#0:
@@ -1642,15 +1642,15 @@ define <2 x i64> @test_pmuldq(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
 ;
 ; SANDY-LABEL: test_pmuldq:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vpmuldq %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
+; SANDY-NEXT:    vpmuldq %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vpmuldq (%rdi), %xmm0, %xmm0 # sched: [9:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pmuldq:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpmuldq %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
-; HASWELL-NEXT:    vpmuldq (%rdi), %xmm0, %xmm0 # sched: [9:1.00]
+; HASWELL-NEXT:    vpmuldq (%rdi), %xmm0, %xmm0 # sched: [5:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pmuldq:
 ; BTVER2:       # BB#0:
@@ -1680,15 +1680,15 @@ define <4 x i32> @test_pmulld(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
 ;
 ; SANDY-LABEL: test_pmulld:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vpmulld %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
+; SANDY-NEXT:    vpmulld %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vpmulld (%rdi), %xmm0, %xmm0 # sched: [9:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pmulld:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpmulld %xmm1, %xmm0, %xmm0 # sched: [10:2.00]
 ; HASWELL-NEXT:    vpmulld (%rdi), %xmm0, %xmm0 # sched: [10:2.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pmulld:
 ; BTVER2:       # BB#0:
@@ -1724,23 +1724,23 @@ define i32 @test_ptest(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) {
 ;
 ; SANDY-LABEL: test_ptest:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vptest %xmm1, %xmm0 # sched: [1:0.33]
+; SANDY-NEXT:    vptest %xmm1, %xmm0 # sched: [2:1.00]
-; SANDY-NEXT:    setb %al # sched: [1:0.33]
+; SANDY-NEXT:    setb %al # sched: [1:1.00]
-; SANDY-NEXT:    vptest (%rdi), %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vptest (%rdi), %xmm0 # sched: [8:1.00]
-; SANDY-NEXT:    setb %cl # sched: [1:0.33]
+; SANDY-NEXT:    setb %cl # sched: [1:1.00]
 ; SANDY-NEXT:    andb %al, %cl # sched: [1:0.33]
 ; SANDY-NEXT:    movzbl %cl, %eax # sched: [1:0.33]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_ptest:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vptest %xmm1, %xmm0 # sched: [2:1.00]
-; HASWELL-NEXT:    setb %al # sched: [1:0.50]
+; HASWELL-NEXT:    setb %al # sched: [1:1.00]
 ; HASWELL-NEXT:    vptest (%rdi), %xmm0 # sched: [2:1.00]
-; HASWELL-NEXT:    setb %cl # sched: [1:0.50]
+; HASWELL-NEXT:    setb %cl # sched: [1:1.00]
 ; HASWELL-NEXT:    andb %al, %cl # sched: [1:0.25]
 ; HASWELL-NEXT:    movzbl %cl, %eax # sched: [1:0.25]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_ptest:
 ; BTVER2:       # BB#0:
@@ -1778,16 +1778,16 @@ define <2 x double> @test_roundpd(<2 x double> %a0, <2 x double> *%a1) {
 ; SANDY-LABEL: test_roundpd:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vroundpd $7, %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    vroundpd $7, (%rdi), %xmm1 # sched: [7:1.00]
+; SANDY-NEXT:    vroundpd $7, (%rdi), %xmm1 # sched: [9:1.00]
 ; SANDY-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_roundpd:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vroundpd $7, %xmm0, %xmm0 # sched: [6:2.00]
+; HASWELL-NEXT:    vroundpd $7, %xmm0, %xmm0 # sched: [5:2.00]
-; HASWELL-NEXT:    vroundpd $7, (%rdi), %xmm1 # sched: [10:2.00]
+; HASWELL-NEXT:    vroundpd $7, (%rdi), %xmm1 # sched: [6:1.00]
 ; HASWELL-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_roundpd:
 ; BTVER2:       # BB#0:
@@ -1822,16 +1822,16 @@ define <4 x float> @test_roundps(<4 x float> %a0, <4 x float> *%a1) {
 ; SANDY-LABEL: test_roundps:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vroundps $7, %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    vroundps $7, (%rdi), %xmm1 # sched: [7:1.00]
+; SANDY-NEXT:    vroundps $7, (%rdi), %xmm1 # sched: [9:1.00]
 ; SANDY-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_roundps:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vroundps $7, %xmm0, %xmm0 # sched: [6:2.00]
+; HASWELL-NEXT:    vroundps $7, %xmm0, %xmm0 # sched: [5:2.00]
-; HASWELL-NEXT:    vroundps $7, (%rdi), %xmm1 # sched: [10:2.00]
+; HASWELL-NEXT:    vroundps $7, (%rdi), %xmm1 # sched: [6:1.00]
 ; HASWELL-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_roundps:
 ; BTVER2:       # BB#0:
@@ -1867,16 +1867,16 @@ define <2 x double> @test_roundsd(<2 x double> %a0, <2 x double> %a1, <2 x doubl
 ; SANDY-LABEL: test_roundsd:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vroundsd $7, %xmm1, %xmm0, %xmm1 # sched: [3:1.00]
-; SANDY-NEXT:    vroundsd $7, (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
+; SANDY-NEXT:    vroundsd $7, (%rdi), %xmm0, %xmm0 # sched: [9:1.00]
 ; SANDY-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_roundsd:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vroundsd $7, %xmm1, %xmm0, %xmm1 # sched: [6:2.00]
+; HASWELL-NEXT:    vroundsd $7, %xmm1, %xmm0, %xmm1 # sched: [5:2.00]
-; HASWELL-NEXT:    vroundsd $7, (%rdi), %xmm0, %xmm0 # sched: [10:2.00]
+; HASWELL-NEXT:    vroundsd $7, (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; HASWELL-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_roundsd:
 ; BTVER2:       # BB#0:
@@ -1912,16 +1912,16 @@ define <4 x float> @test_roundss(<4 x float> %a0, <4 x float> %a1, <4 x float> *
 ; SANDY-LABEL: test_roundss:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vroundss $7, %xmm1, %xmm0, %xmm1 # sched: [3:1.00]
-; SANDY-NEXT:    vroundss $7, (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
+; SANDY-NEXT:    vroundss $7, (%rdi), %xmm0, %xmm0 # sched: [9:1.00]
 ; SANDY-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_roundss:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vroundss $7, %xmm1, %xmm0, %xmm1 # sched: [6:2.00]
+; HASWELL-NEXT:    vroundss $7, %xmm1, %xmm0, %xmm1 # sched: [3:1.00]
-; HASWELL-NEXT:    vroundss $7, (%rdi), %xmm0, %xmm0 # sched: [10:2.00]
+; HASWELL-NEXT:    vroundss $7, (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; HASWELL-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_roundss:
 ; BTVER2:       # BB#0:

llvm/test/CodeGen/X86/sse42-schedule.ll

@@ -26,16 +26,16 @@ define i32 @crc32_32_8(i32 %a0, i8 %a1, i8 *%a2) {
 ; SANDY-LABEL: crc32_32_8:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    crc32b %sil, %edi # sched: [3:1.00]
-; SANDY-NEXT:    crc32b (%rdx), %edi # sched: [7:1.00]
+; SANDY-NEXT:    crc32b (%rdx), %edi # sched: [8:1.00]
 ; SANDY-NEXT:    movl %edi, %eax # sched: [1:0.33]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: crc32_32_8:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    crc32b %sil, %edi # sched: [3:1.00]
 ; HASWELL-NEXT:    crc32b (%rdx), %edi # sched: [7:1.00]
 ; HASWELL-NEXT:    movl %edi, %eax # sched: [1:0.25]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: crc32_32_8:
 ; BTVER2:       # BB#0:
@@ -68,16 +68,16 @@ define i32 @crc32_32_16(i32 %a0, i16 %a1, i16 *%a2) {
 ; SANDY-LABEL: crc32_32_16:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    crc32w %si, %edi # sched: [3:1.00]
-; SANDY-NEXT:    crc32w (%rdx), %edi # sched: [7:1.00]
+; SANDY-NEXT:    crc32w (%rdx), %edi # sched: [8:1.00]
 ; SANDY-NEXT:    movl %edi, %eax # sched: [1:0.33]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: crc32_32_16:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    crc32w %si, %edi # sched: [3:1.00]
 ; HASWELL-NEXT:    crc32w (%rdx), %edi # sched: [7:1.00]
 ; HASWELL-NEXT:    movl %edi, %eax # sched: [1:0.25]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: crc32_32_16:
 ; BTVER2:       # BB#0:
@@ -112,14 +112,14 @@ define i32 @crc32_32_32(i32 %a0, i32 %a1, i32 *%a2) {
 ; SANDY-NEXT:    crc32l %esi, %edi # sched: [3:1.00]
 ; SANDY-NEXT:    crc32l (%rdx), %edi # sched: [7:1.00]
 ; SANDY-NEXT:    movl %edi, %eax # sched: [1:0.33]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: crc32_32_32:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    crc32l %esi, %edi # sched: [3:1.00]
 ; HASWELL-NEXT:    crc32l (%rdx), %edi # sched: [7:1.00]
 ; HASWELL-NEXT:    movl %edi, %eax # sched: [1:0.25]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: crc32_32_32:
 ; BTVER2:       # BB#0:
@@ -152,16 +152,16 @@ define i64 @crc32_64_8(i64 %a0, i8 %a1, i8 *%a2) nounwind {
 ; SANDY-LABEL: crc32_64_8:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    crc32b %sil, %edi # sched: [3:1.00]
-; SANDY-NEXT:    crc32b (%rdx), %edi # sched: [7:1.00]
+; SANDY-NEXT:    crc32b (%rdx), %edi # sched: [8:1.00]
 ; SANDY-NEXT:    movq %rdi, %rax # sched: [1:0.33]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: crc32_64_8:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    crc32b %sil, %edi # sched: [3:1.00]
 ; HASWELL-NEXT:    crc32b (%rdx), %edi # sched: [7:1.00]
 ; HASWELL-NEXT:    movq %rdi, %rax # sched: [1:0.25]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: crc32_64_8:
 ; BTVER2:       # BB#0:
@@ -196,14 +196,14 @@ define i64 @crc32_64_64(i64 %a0, i64 %a1, i64 *%a2) {
 ; SANDY-NEXT:    crc32q %rsi, %rdi # sched: [3:1.00]
 ; SANDY-NEXT:    crc32q (%rdx), %rdi # sched: [7:1.00]
 ; SANDY-NEXT:    movq %rdi, %rax # sched: [1:0.33]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: crc32_64_64:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    crc32q %rsi, %rdi # sched: [3:1.00]
 ; HASWELL-NEXT:    crc32q (%rdx), %rdi # sched: [7:1.00]
 ; HASWELL-NEXT:    movq %rdi, %rax # sched: [1:0.25]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: crc32_64_64:
 ; BTVER2:       # BB#0:
@@ -256,20 +256,20 @@ define i32 @test_pcmpestri(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {
 ; SANDY-NEXT:    vpcmpestri $7, (%rdi), %xmm0 # sched: [4:2.33]
 ; SANDY-NEXT:    # kill: %ECX<def> %ECX<kill> %RCX<def>
 ; SANDY-NEXT:    leal (%rcx,%rsi), %eax # sched: [1:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pcmpestri:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    movl $7, %eax # sched: [1:0.25]
 ; HASWELL-NEXT:    movl $7, %edx # sched: [1:0.25]
-; HASWELL-NEXT:    vpcmpestri $7, %xmm1, %xmm0 # sched: [11:3.00]
+; HASWELL-NEXT:    vpcmpestri $7, %xmm1, %xmm0 # sched: [18:4.00]
 ; HASWELL-NEXT:    movl %ecx, %esi # sched: [1:0.25]
 ; HASWELL-NEXT:    movl $7, %eax # sched: [1:0.25]
 ; HASWELL-NEXT:    movl $7, %edx # sched: [1:0.25]
-; HASWELL-NEXT:    vpcmpestri $7, (%rdi), %xmm0 # sched: [11:3.00]
+; HASWELL-NEXT:    vpcmpestri $7, (%rdi), %xmm0 # sched: [18:4.00]
 ; HASWELL-NEXT:    # kill: %ECX<def> %ECX<kill> %RCX<def>
 ; HASWELL-NEXT:    leal (%rcx,%rsi), %eax # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pcmpestri:
 ; BTVER2:       # BB#0:
@@ -320,17 +320,17 @@ define <16 x i8> @test_pcmpestrm(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {
 ; SANDY-NEXT:    movl $7, %eax # sched: [1:0.33]
 ; SANDY-NEXT:    movl $7, %edx # sched: [1:0.33]
 ; SANDY-NEXT:    vpcmpestrm $7, (%rdi), %xmm0 # sched: [11:2.33]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pcmpestrm:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    movl $7, %eax # sched: [1:0.25]
 ; HASWELL-NEXT:    movl $7, %edx # sched: [1:0.25]
-; HASWELL-NEXT:    vpcmpestrm $7, %xmm1, %xmm0 # sched: [10:4.00]
+; HASWELL-NEXT:    vpcmpestrm $7, %xmm1, %xmm0 # sched: [19:4.00]
 ; HASWELL-NEXT:    movl $7, %eax # sched: [1:0.25]
 ; HASWELL-NEXT:    movl $7, %edx # sched: [1:0.25]
-; HASWELL-NEXT:    vpcmpestrm $7, (%rdi), %xmm0 # sched: [10:3.00]
+; HASWELL-NEXT:    vpcmpestrm $7, (%rdi), %xmm0 # sched: [19:4.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pcmpestrm:
 ; BTVER2:       # BB#0:
@@ -369,12 +369,12 @@ define i32 @test_pcmpistri(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {
 ;
 ; SANDY-LABEL: test_pcmpistri:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vpcmpistri $7, %xmm1, %xmm0 # sched: [3:1.00]
+; SANDY-NEXT:    vpcmpistri $7, %xmm1, %xmm0 # sched: [11:3.00]
 ; SANDY-NEXT:    movl %ecx, %eax # sched: [1:0.33]
-; SANDY-NEXT:    vpcmpistri $7, (%rdi), %xmm0 # sched: [3:1.00]
+; SANDY-NEXT:    vpcmpistri $7, (%rdi), %xmm0 # sched: [17:3.00]
 ; SANDY-NEXT:    # kill: %ECX<def> %ECX<kill> %RCX<def>
 ; SANDY-NEXT:    leal (%rcx,%rax), %eax # sched: [1:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pcmpistri:
 ; HASWELL:       # BB#0:
@@ -383,7 +383,7 @@ define i32 @test_pcmpistri(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {
 ; HASWELL-NEXT:    vpcmpistri $7, (%rdi), %xmm0 # sched: [11:3.00]
 ; HASWELL-NEXT:    # kill: %ECX<def> %ECX<kill> %RCX<def>
 ; HASWELL-NEXT:    leal (%rcx,%rax), %eax # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pcmpistri:
 ; BTVER2:       # BB#0:
@@ -416,15 +416,15 @@ define <16 x i8> @test_pcmpistrm(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {
 ;
 ; SANDY-LABEL: test_pcmpistrm:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vpcmpistrm $7, %xmm1, %xmm0 # sched: [11:1.00]
+; SANDY-NEXT:    vpcmpistrm $7, %xmm1, %xmm0 # sched: [11:3.00]
-; SANDY-NEXT:    vpcmpistrm $7, (%rdi), %xmm0 # sched: [11:1.00]
+; SANDY-NEXT:    vpcmpistrm $7, (%rdi), %xmm0 # sched: [17:3.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pcmpistrm:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vpcmpistrm $7, %xmm1, %xmm0 # sched: [10:3.00]
+; HASWELL-NEXT:    vpcmpistrm $7, %xmm1, %xmm0 # sched: [11:3.00]
-; HASWELL-NEXT:    vpcmpistrm $7, (%rdi), %xmm0 # sched: [10:3.00]
+; HASWELL-NEXT:    vpcmpistrm $7, (%rdi), %xmm0 # sched: [11:3.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pcmpistrm:
 ; BTVER2:       # BB#0:
@@ -453,15 +453,15 @@ define <2 x i64> @test_pcmpgtq(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) {
 ;
 ; SANDY-LABEL: test_pcmpgtq:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
+; SANDY-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
-; SANDY-NEXT:    vpcmpgtq (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vpcmpgtq (%rdi), %xmm0, %xmm0 # sched: [11:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pcmpgtq:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
 ; HASWELL-NEXT:    vpcmpgtq (%rdi), %xmm0, %xmm0 # sched: [5:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pcmpgtq:
 ; BTVER2:       # BB#0:

llvm/test/CodeGen/X86/ssse3-schedule.ll

@@ -35,16 +35,16 @@ define <16 x i8> @test_pabsb(<16 x i8> %a0, <16 x i8> *%a1) {
 ; SANDY-LABEL: test_pabsb:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpabsb %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vpabsb (%rdi), %xmm1 # sched: [5:0.50]
+; SANDY-NEXT:    vpabsb (%rdi), %xmm1 # sched: [7:0.50]
 ; SANDY-NEXT:    vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pabsb:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpabsb %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    vpabsb (%rdi), %xmm1 # sched: [5:0.50]
+; HASWELL-NEXT:    vpabsb (%rdi), %xmm1 # sched: [1:0.50]
 ; HASWELL-NEXT:    vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pabsb:
 ; BTVER2:       # BB#0:
@@ -86,16 +86,16 @@ define <4 x i32> @test_pabsd(<4 x i32> %a0, <4 x i32> *%a1) {
 ; SANDY-LABEL: test_pabsd:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpabsd %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vpabsd (%rdi), %xmm1 # sched: [5:0.50]
+; SANDY-NEXT:    vpabsd (%rdi), %xmm1 # sched: [7:0.50]
 ; SANDY-NEXT:    vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pabsd:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpabsd %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    vpabsd (%rdi), %xmm1 # sched: [5:0.50]
+; HASWELL-NEXT:    vpabsd (%rdi), %xmm1 # sched: [1:0.50]
 ; HASWELL-NEXT:    vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pabsd:
 ; BTVER2:       # BB#0:
@@ -136,12 +136,12 @@ define <8 x i16> @test_pabsw(<8 x i16> %a0, <8 x i16> *%a1) {
 ; SANDY-LABEL: test_pabsw:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpabsw %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pabsw:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpabsw %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pabsw:
 ; BTVER2:       # BB#0:
@@ -182,14 +182,14 @@ define <8 x i16> @test_palignr(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {
 ; SANDY-LABEL: test_palignr:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[6,7,8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5] sched: [1:0.50]
-; SANDY-NEXT:    vpalignr {{.*#+}} xmm0 = mem[14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13] sched: [5:0.50]
+; SANDY-NEXT:    vpalignr {{.*#+}} xmm0 = mem[14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13] sched: [7:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_palignr:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[6,7,8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5] sched: [1:1.00]
-; HASWELL-NEXT:    vpalignr {{.*#+}} xmm0 = mem[14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13] sched: [5:1.00]
+; HASWELL-NEXT:    vpalignr {{.*#+}} xmm0 = mem[14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13] sched: [1:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_palignr:
 ; BTVER2:       # BB#0:
@@ -223,15 +223,15 @@ define <4 x i32> @test_phaddd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
 ;
 ; SANDY-LABEL: test_phaddd:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vphaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
+; SANDY-NEXT:    vphaddd %xmm1, %xmm0, %xmm0 # sched: [3:1.50]
-; SANDY-NEXT:    vphaddd (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vphaddd (%rdi), %xmm0, %xmm0 # sched: [9:1.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_phaddd:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vphaddd %xmm1, %xmm0, %xmm0 # sched: [3:2.00]
-; HASWELL-NEXT:    vphaddd (%rdi), %xmm0, %xmm0 # sched: [6:2.00]
+; HASWELL-NEXT:    vphaddd (%rdi), %xmm0, %xmm0 # sched: [3:2.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_phaddd:
 ; BTVER2:       # BB#0:
@@ -274,15 +274,15 @@ define <8 x i16> @test_phaddsw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {
 ;
 ; SANDY-LABEL: test_phaddsw:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vphaddsw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
+; SANDY-NEXT:    vphaddsw %xmm1, %xmm0, %xmm0 # sched: [3:1.50]
-; SANDY-NEXT:    vphaddsw (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vphaddsw (%rdi), %xmm0, %xmm0 # sched: [9:1.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_phaddsw:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vphaddsw %xmm1, %xmm0, %xmm0 # sched: [3:2.00]
-; HASWELL-NEXT:    vphaddsw (%rdi), %xmm0, %xmm0 # sched: [6:2.00]
+; HASWELL-NEXT:    vphaddsw (%rdi), %xmm0, %xmm0 # sched: [3:2.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_phaddsw:
 ; BTVER2:       # BB#0:
@@ -317,15 +317,15 @@ define <8 x i16> @test_phaddw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {
 ;
 ; SANDY-LABEL: test_phaddw:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vphaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
+; SANDY-NEXT:    vphaddw %xmm1, %xmm0, %xmm0 # sched: [3:1.50]
-; SANDY-NEXT:    vphaddw (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vphaddw (%rdi), %xmm0, %xmm0 # sched: [9:1.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_phaddw:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vphaddw %xmm1, %xmm0, %xmm0 # sched: [3:2.00]
-; HASWELL-NEXT:    vphaddw (%rdi), %xmm0, %xmm0 # sched: [6:2.00]
+; HASWELL-NEXT:    vphaddw (%rdi), %xmm0, %xmm0 # sched: [3:2.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_phaddw:
 ; BTVER2:       # BB#0:
@@ -360,15 +360,15 @@ define <4 x i32> @test_phsubd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
 ;
 ; SANDY-LABEL: test_phsubd:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vphsubd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
+; SANDY-NEXT:    vphsubd %xmm1, %xmm0, %xmm0 # sched: [3:1.50]
-; SANDY-NEXT:    vphsubd (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vphsubd (%rdi), %xmm0, %xmm0 # sched: [9:1.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_phsubd:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vphsubd %xmm1, %xmm0, %xmm0 # sched: [3:2.00]
-; HASWELL-NEXT:    vphsubd (%rdi), %xmm0, %xmm0 # sched: [6:2.00]
+; HASWELL-NEXT:    vphsubd (%rdi), %xmm0, %xmm0 # sched: [3:2.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_phsubd:
 ; BTVER2:       # BB#0:
@@ -411,15 +411,15 @@ define <8 x i16> @test_phsubsw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {
 ;
 ; SANDY-LABEL: test_phsubsw:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vphsubsw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
+; SANDY-NEXT:    vphsubsw %xmm1, %xmm0, %xmm0 # sched: [3:1.50]
-; SANDY-NEXT:    vphsubsw (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vphsubsw (%rdi), %xmm0, %xmm0 # sched: [9:1.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_phsubsw:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vphsubsw %xmm1, %xmm0, %xmm0 # sched: [3:2.00]
-; HASWELL-NEXT:    vphsubsw (%rdi), %xmm0, %xmm0 # sched: [6:2.00]
+; HASWELL-NEXT:    vphsubsw (%rdi), %xmm0, %xmm0 # sched: [3:2.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_phsubsw:
 ; BTVER2:       # BB#0:
@@ -454,15 +454,15 @@ define <8 x i16> @test_phsubw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {
 ;
 ; SANDY-LABEL: test_phsubw:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vphsubw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
+; SANDY-NEXT:    vphsubw %xmm1, %xmm0, %xmm0 # sched: [3:1.50]
-; SANDY-NEXT:    vphsubw (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vphsubw (%rdi), %xmm0, %xmm0 # sched: [9:1.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_phsubw:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vphsubw %xmm1, %xmm0, %xmm0 # sched: [3:2.00]
-; HASWELL-NEXT:    vphsubw (%rdi), %xmm0, %xmm0 # sched: [6:2.00]
+; HASWELL-NEXT:    vphsubw (%rdi), %xmm0, %xmm0 # sched: [3:2.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_phsubw:
 ; BTVER2:       # BB#0:
@@ -497,15 +497,15 @@ define <8 x i16> @test_pmaddubsw(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {
 ;
 ; SANDY-LABEL: test_pmaddubsw:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vpmaddubsw %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
+; SANDY-NEXT:    vpmaddubsw %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vpmaddubsw (%rdi), %xmm0, %xmm0 # sched: [9:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pmaddubsw:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpmaddubsw %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
-; HASWELL-NEXT:    vpmaddubsw (%rdi), %xmm0, %xmm0 # sched: [9:1.00]
+; HASWELL-NEXT:    vpmaddubsw (%rdi), %xmm0, %xmm0 # sched: [5:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pmaddubsw:
 ; BTVER2:       # BB#0:
@@ -538,13 +538,13 @@ define <8 x i16> @test_pmulhrsw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {
 ;
 ; SANDY-LABEL: test_pmulhrsw:
 ; SANDY:       # BB#0:
-; SANDY-NEXT:    vpmulhrsw %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
+; SANDY-NEXT:    vpmulhrsw %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pmulhrsw:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpmulhrsw %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pmulhrsw:
 ; BTVER2:       # BB#0:
@@ -579,14 +579,14 @@ define <16 x i8> @test_pshufb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {
 ; SANDY-LABEL: test_pshufb:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpshufb %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vpshufb (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vpshufb (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pshufb:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpshufb %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
-; HASWELL-NEXT:    vpshufb (%rdi), %xmm0, %xmm0 # sched: [5:1.00]
+; HASWELL-NEXT:    vpshufb (%rdi), %xmm0, %xmm0 # sched: [1:1.00]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_pshufb:
 ; BTVER2:       # BB#0:
@@ -630,14 +630,14 @@ define <16 x i8> @test_psignb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {
 ; SANDY-LABEL: test_psignb:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpsignb %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vpsignb (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vpsignb (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_psignb:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpsignb %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    vpsignb (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; HASWELL-NEXT:    vpsignb (%rdi), %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_psignb:
 ; BTVER2:       # BB#0:
@@ -681,14 +681,14 @@ define <4 x i32> @test_psignd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
 ; SANDY-LABEL: test_psignd:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpsignd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vpsignd (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vpsignd (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_psignd:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpsignd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    vpsignd (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; HASWELL-NEXT:    vpsignd (%rdi), %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_psignd:
 ; BTVER2:       # BB#0:
@@ -732,14 +732,14 @@ define <8 x i16> @test_psignw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {
 ; SANDY-LABEL: test_psignw:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vpsignw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vpsignw (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; SANDY-NEXT:    vpsignw (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SANDY-NEXT:    retq # sched: [5:1.00]
+; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_psignw:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpsignw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    vpsignw (%rdi), %xmm0, %xmm0 # sched: [5:0.50]
+; HASWELL-NEXT:    vpsignw (%rdi), %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT:    retq # sched: [1:1.00]
+; HASWELL-NEXT:    retq # sched: [2:1.00]
 ;
 ; BTVER2-LABEL: test_psignw:
 ; BTVER2:       # BB#0:

llvm/test/CodeGen/X86/vector-shift-ashr-512.ll

@@ -201,14 +201,14 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, <64 x i8> %b) nounwind {
 ; AVX512DQ-NEXT:    vpsraw $2, %ymm0, %ymm5
 ; AVX512DQ-NEXT:    vpaddw %ymm2, %ymm2, %ymm8
 ; AVX512DQ-NEXT:    vpblendvb %ymm8, %ymm5, %ymm0, %ymm0
+; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} ymm5 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31]
+; AVX512DQ-NEXT:    vpsraw $4, %ymm5, %ymm9
+; AVX512DQ-NEXT:    vpblendvb %ymm3, %ymm9, %ymm5, %ymm3
 ; AVX512DQ-NEXT:    vpsraw $1, %ymm0, %ymm5
 ; AVX512DQ-NEXT:    vpaddw %ymm8, %ymm8, %ymm9
 ; AVX512DQ-NEXT:    vpblendvb %ymm9, %ymm5, %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vpsrlw $8, %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vpackuswb %ymm4, %ymm0, %ymm0
-; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} ymm4 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31]
-; AVX512DQ-NEXT:    vpsraw $4, %ymm4, %ymm5
-; AVX512DQ-NEXT:    vpblendvb %ymm3, %ymm5, %ymm4, %ymm3
 ; AVX512DQ-NEXT:    vpsraw $2, %ymm3, %ymm4
 ; AVX512DQ-NEXT:    vpblendvb %ymm6, %ymm4, %ymm3, %ymm3
 ; AVX512DQ-NEXT:    vpsraw $1, %ymm3, %ymm4
@@ -328,14 +328,14 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) nounwind {
 ; AVX512DQ-NEXT:    vpsraw $2, %ymm0, %ymm5
 ; AVX512DQ-NEXT:    vpaddw %ymm2, %ymm2, %ymm8
 ; AVX512DQ-NEXT:    vpblendvb %ymm8, %ymm5, %ymm0, %ymm0
+; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} ymm5 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31]
+; AVX512DQ-NEXT:    vpsraw $4, %ymm5, %ymm9
+; AVX512DQ-NEXT:    vpblendvb %ymm3, %ymm9, %ymm5, %ymm3
 ; AVX512DQ-NEXT:    vpsraw $1, %ymm0, %ymm5
 ; AVX512DQ-NEXT:    vpaddw %ymm8, %ymm8, %ymm9
 ; AVX512DQ-NEXT:    vpblendvb %ymm9, %ymm5, %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vpsrlw $8, %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vpackuswb %ymm4, %ymm0, %ymm0
-; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} ymm4 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31]
-; AVX512DQ-NEXT:    vpsraw $4, %ymm4, %ymm5
-; AVX512DQ-NEXT:    vpblendvb %ymm3, %ymm5, %ymm4, %ymm3
 ; AVX512DQ-NEXT:    vpsraw $2, %ymm3, %ymm4
 ; AVX512DQ-NEXT:    vpblendvb %ymm6, %ymm4, %ymm3, %ymm3
 ; AVX512DQ-NEXT:    vpsraw $1, %ymm3, %ymm4

llvm/test/CodeGen/X86/vector-shuffle-512-v32.ll

@@ -68,13 +68,13 @@ define <32 x i16> @shuffle_v32i16_0f_1f_0e_16_0d_1d_04_1e_0b_1b_0a_1a_09_19_08_1
 ; KNL-NEXT:    vpshufb {{.*#+}} xmm4 = xmm1[8,9,12,13,12,13,10,11,0,1,4,5,4,5,0,1]
 ; KNL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm2[0,1,0,3]
 ; KNL-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm1[0,3,2,2,4,5,6,7]
-; KNL-NEXT:    vinserti128 $1, %xmm2, %ymm4, %ymm1
+; KNL-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; KNL-NEXT:    vextracti128 $1, %ymm0, %xmm5
+; KNL-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
-; KNL-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm5[4],xmm0[4],xmm5[5],xmm0[5],xmm5[6],xmm0[6],xmm5[7],xmm0[7]
 ; KNL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[12,13,10,11,8,9,14,15,4,5,2,3,2,3,6,7]
-; KNL-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[6,7,2,3,4,5,6,7,2,3,2,3,0,1,14,15]
+; KNL-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[6,7,2,3,4,5,6,7,2,3,2,3,0,1,14,15]
-; KNL-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm0
+; KNL-NEXT:    vinserti128 $1, %xmm2, %ymm4, %ymm5
-; KNL-NEXT:    vpblendw {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
+; KNL-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; KNL-NEXT:    vpblendw {{.*#+}} ymm1 = ymm0[0],ymm5[1],ymm0[2],ymm5[3],ymm0[4],ymm5[5],ymm0[6],ymm5[7],ymm0[8],ymm5[9],ymm0[10],ymm5[11],ymm0[12],ymm5[13],ymm0[14],ymm5[15]
 ; KNL-NEXT:    vextracti128 $1, %ymm3, %xmm3
 ; KNL-NEXT:    vpbroadcastw %xmm3, %ymm3
 ; KNL-NEXT:    vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0]