From 17eace47cb3ee7c819a9141c123e3dbdf9d7a6b4 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Wed, 9 Jan 2019 12:34:10 +0000
Subject: [PATCH] [X86] Add extra test coverage for combining shuffles to
 PACKSS/PACKUS

llvm-svn: 350707
---
 .../X86/vector-shuffle-combining-avx2.ll      | 38 +++++++++++++
 .../X86/vector-shuffle-combining-avx512bw.ll  | 56 +++++++++++++++++++
 2 files changed, 94 insertions(+)

diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx2.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx2.ll
index 2ea0f1ab3e71..dda9e140d459 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx2.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx2.ll
@@ -611,6 +611,44 @@ define <32 x i8> @shuffle_combine_packuswb_pshufb(<16 x i16> %a0, <16 x i16> %a1
 }
 declare <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16>, <16 x i16>) nounwind readnone
 
+define <32 x i8> @combine_pshufb_as_packsswb(<16 x i16> %a0, <16 x i16> %a1) nounwind {
+; CHECK-LABEL: combine_pshufb_as_packsswb:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpsraw $11, %ymm0, %ymm0
+; CHECK-NEXT:    vpsraw $11, %ymm1, %ymm1
+; CHECK-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,2,4,6,8,10,12,14],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,18,20,22,24,26,28,30],zero,zero,zero,zero,zero,zero,zero,zero
+; CHECK-NEXT:    vpshufb {{.*#+}} ymm1 = zero,zero,zero,zero,zero,zero,zero,zero,ymm1[0,2,4,6,8,10,12,14],zero,zero,zero,zero,zero,zero,zero,zero,ymm1[16,18,20,22,24,26,28,30]
+; CHECK-NEXT:    vpor %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    ret{{[l|q]}}
+  %1 = ashr <16 x i16> %a0, <i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11>
+  %2 = ashr <16 x i16> %a1, <i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11>
+  %3 = bitcast <16 x i16> %1 to <32 x i8>
+  %4 = bitcast <16 x i16> %2 to <32 x i8>
+  %5 = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %3, <32 x i8> <i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>)
+  %6 = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %4, <32 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14>)
+  %7 = or <32 x i8> %5, %6
+  ret <32 x i8> %7
+}
+
+define <32 x i8> @combine_pshufb_as_packuswb(<16 x i16> %a0, <16 x i16> %a1) nounwind {
+; CHECK-LABEL: combine_pshufb_as_packuswb:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpsrlw $11, %ymm0, %ymm0
+; CHECK-NEXT:    vpsrlw $11, %ymm1, %ymm1
+; CHECK-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,2,4,6,8,10,12,14],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,18,20,22,24,26,28,30],zero,zero,zero,zero,zero,zero,zero,zero
+; CHECK-NEXT:    vpshufb {{.*#+}} ymm1 = zero,zero,zero,zero,zero,zero,zero,zero,ymm1[0,2,4,6,8,10,12,14],zero,zero,zero,zero,zero,zero,zero,zero,ymm1[16,18,20,22,24,26,28,30]
+; CHECK-NEXT:    vpor %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    ret{{[l|q]}}
+  %1 = lshr <16 x i16> %a0, <i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11>
+  %2 = lshr <16 x i16> %a1, <i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11>
+  %3 = bitcast <16 x i16> %1 to <32 x i8>
+  %4 = bitcast <16 x i16> %2 to <32 x i8>
+  %5 = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %3, <32 x i8> <i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>)
+  %6 = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %4, <32 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14>)
+  %7 = or <32 x i8> %5, %6
+  ret <32 x i8> %7
+}
+
 define <16 x i8> @combine_pshufb_insertion_as_broadcast_v2i64(i64 %a0) {
 ; X86-LABEL: combine_pshufb_insertion_as_broadcast_v2i64:
 ; X86:       # %bb.0:
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bw.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bw.ll
index b8d3824f1caf..5c40520ac1ca 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bw.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bw.ll
@@ -824,6 +824,62 @@ define <32 x i16> @combine_pshufb_as_pshufhw(<32 x i16> %a0) {
   ret <32 x i16> %1
 }
 
+define <64 x i8> @combine_pshufb_as_packsswb(<32 x i16> %a0, <32 x i16> %a1) nounwind {
+; X32-LABEL: combine_pshufb_as_packsswb:
+; X32:       # %bb.0:
+; X32-NEXT:    vpsraw $11, %zmm0, %zmm0
+; X32-NEXT:    vpsraw $11, %zmm1, %zmm1
+; X32-NEXT:    vpshufb {{.*#+}} zmm0 = zmm0[0,2,4,6,8,10,12,14],zero,zero,zero,zero,zero,zero,zero,zero,zmm0[16,18,20,22,24,26,28,30],zero,zero,zero,zero,zero,zero,zero,zero,zmm0[32,34,36,38,40,42,44,46],zero,zero,zero,zero,zero,zero,zero,zero,zmm0[48,50,52,54,56,58,60,62],zero,zero,zero,zero,zero,zero,zero,zero
+; X32-NEXT:    vpshufb {{.*#+}} zmm1 = zero,zero,zero,zero,zero,zero,zero,zero,zmm1[0,2,4,6,8,10,12,14],zero,zero,zero,zero,zero,zero,zero,zero,zmm1[16,18,20,22,24,26,28,30],zero,zero,zero,zero,zero,zero,zero,zero,zmm1[32,34,36,38,40,42,44,46],zero,zero,zero,zero,zero,zero,zero,zero,zmm1[48,50,52,54,56,58,60,62]
+; X32-NEXT:    vporq %zmm1, %zmm0, %zmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: combine_pshufb_as_packsswb:
+; X64:       # %bb.0:
+; X64-NEXT:    vpsraw $11, %zmm0, %zmm0
+; X64-NEXT:    vpsraw $11, %zmm1, %zmm1
+; X64-NEXT:    vpshufb {{.*#+}} zmm0 = zmm0[0,2,4,6,8,10,12,14],zero,zero,zero,zero,zero,zero,zero,zero,zmm0[16,18,20,22,24,26,28,30],zero,zero,zero,zero,zero,zero,zero,zero,zmm0[32,34,36,38,40,42,44,46],zero,zero,zero,zero,zero,zero,zero,zero,zmm0[48,50,52,54,56,58,60,62],zero,zero,zero,zero,zero,zero,zero,zero
+; X64-NEXT:    vpshufb {{.*#+}} zmm1 = zero,zero,zero,zero,zero,zero,zero,zero,zmm1[0,2,4,6,8,10,12,14],zero,zero,zero,zero,zero,zero,zero,zero,zmm1[16,18,20,22,24,26,28,30],zero,zero,zero,zero,zero,zero,zero,zero,zmm1[32,34,36,38,40,42,44,46],zero,zero,zero,zero,zero,zero,zero,zero,zmm1[48,50,52,54,56,58,60,62]
+; X64-NEXT:    vporq %zmm1, %zmm0, %zmm0
+; X64-NEXT:    retq
+  %1 = ashr <32 x i16> %a0, <i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11>
+  %2 = ashr <32 x i16> %a1, <i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11>
+  %3 = bitcast <32 x i16> %1 to <64 x i8>
+  %4 = bitcast <32 x i16> %2 to <64 x i8>
+  %5 = tail call <64 x i8> @llvm.x86.avx512.mask.pshuf.b.512(<64 x i8> %3, <64 x i8> <i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <64 x i8> undef, i64 -1)
+  %6 = tail call <64 x i8> @llvm.x86.avx512.mask.pshuf.b.512(<64 x i8> %4, <64 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14>, <64 x i8> undef, i64 -1)
+  %7 = or <64 x i8> %5, %6
+  ret <64 x i8> %7
+}
+
+define <64 x i8> @combine_pshufb_as_packuswb(<32 x i16> %a0, <32 x i16> %a1) nounwind {
+; X32-LABEL: combine_pshufb_as_packuswb:
+; X32:       # %bb.0:
+; X32-NEXT:    vpsrlw $11, %zmm0, %zmm0
+; X32-NEXT:    vpsrlw $11, %zmm1, %zmm1
+; X32-NEXT:    vpshufb {{.*#+}} zmm0 = zmm0[0,2,4,6,8,10,12,14],zero,zero,zero,zero,zero,zero,zero,zero,zmm0[16,18,20,22,24,26,28,30],zero,zero,zero,zero,zero,zero,zero,zero,zmm0[32,34,36,38,40,42,44,46],zero,zero,zero,zero,zero,zero,zero,zero,zmm0[48,50,52,54,56,58,60,62],zero,zero,zero,zero,zero,zero,zero,zero
+; X32-NEXT:    vpshufb {{.*#+}} zmm1 = zero,zero,zero,zero,zero,zero,zero,zero,zmm1[0,2,4,6,8,10,12,14],zero,zero,zero,zero,zero,zero,zero,zero,zmm1[16,18,20,22,24,26,28,30],zero,zero,zero,zero,zero,zero,zero,zero,zmm1[32,34,36,38,40,42,44,46],zero,zero,zero,zero,zero,zero,zero,zero,zmm1[48,50,52,54,56,58,60,62]
+; X32-NEXT:    vporq %zmm1, %zmm0, %zmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: combine_pshufb_as_packuswb:
+; X64:       # %bb.0:
+; X64-NEXT:    vpsrlw $11, %zmm0, %zmm0
+; X64-NEXT:    vpsrlw $11, %zmm1, %zmm1
+; X64-NEXT:    vpshufb {{.*#+}} zmm0 = zmm0[0,2,4,6,8,10,12,14],zero,zero,zero,zero,zero,zero,zero,zero,zmm0[16,18,20,22,24,26,28,30],zero,zero,zero,zero,zero,zero,zero,zero,zmm0[32,34,36,38,40,42,44,46],zero,zero,zero,zero,zero,zero,zero,zero,zmm0[48,50,52,54,56,58,60,62],zero,zero,zero,zero,zero,zero,zero,zero
+; X64-NEXT:    vpshufb {{.*#+}} zmm1 = zero,zero,zero,zero,zero,zero,zero,zero,zmm1[0,2,4,6,8,10,12,14],zero,zero,zero,zero,zero,zero,zero,zero,zmm1[16,18,20,22,24,26,28,30],zero,zero,zero,zero,zero,zero,zero,zero,zmm1[32,34,36,38,40,42,44,46],zero,zero,zero,zero,zero,zero,zero,zero,zmm1[48,50,52,54,56,58,60,62]
+; X64-NEXT:    vporq %zmm1, %zmm0, %zmm0
+; X64-NEXT:    retq
+  %1 = lshr <32 x i16> %a0, <i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11>
+  %2 = lshr <32 x i16> %a1, <i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11>
+  %3 = bitcast <32 x i16> %1 to <64 x i8>
+  %4 = bitcast <32 x i16> %2 to <64 x i8>
+  %5 = tail call <64 x i8> @llvm.x86.avx512.mask.pshuf.b.512(<64 x i8> %3, <64 x i8> <i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <64 x i8> undef, i64 -1)
+  %6 = tail call <64 x i8> @llvm.x86.avx512.mask.pshuf.b.512(<64 x i8> %4, <64 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 2, i8 4, i8 6, i8 8, i8 10, i8 12, i8 14>, <64 x i8> undef, i64 -1)
+  %7 = or <64 x i8> %5, %6
+  ret <64 x i8> %7
+}
+
 define <32 x i16> @combine_vpermi2var_32i16_as_pshufb(<32 x i16> %a0) {
 ; X32-LABEL: combine_vpermi2var_32i16_as_pshufb:
 ; X32:       # %bb.0: