3 files changed, 82 insertions, 0 deletions
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx2.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx2.ll
index 8324c5810f9..9c56d9c5e1d 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx2.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx2.ll
@@ -183,3 +183,21 @@ define <4 x double> @combine_permd_as_vpbroadcastsd256(<2 x double> %a) {
   %4 = bitcast <8 x float> %3 to <4 x double>
   ret <4 x double> %4
 }
+
+define <32 x i8> @combine_pshufb_as_pslldq(<32 x i8> %a0) {
+; CHECK-LABEL: combine_pshufb_as_pslldq:
+; CHECK:       # BB#0:
+; CHECK-NEXT:    vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[0,1,2,3,4,5],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,17,18,19,20,21]
+; CHECK-NEXT:    retq
+  %res0 = call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %a0, <32 x i8> <i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5>)
+  ret <32 x i8> %res0
+}
+
+define <32 x i8> @combine_pshufb_as_psrldq(<32 x i8> %a0) {
+; CHECK-LABEL: combine_pshufb_as_psrldq:
+; CHECK:       # BB#0:
+; CHECK-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; CHECK-NEXT:    retq
+  %res0 = call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %a0, <32 x i8> <i8 15, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 15, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128>)
+  ret <32 x i8> %res0
+}
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bw.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bw.ll
index 751ee526c27..c3d3f9199b0 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bw.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bw.ll
@@ -399,3 +399,39 @@ define <8 x i64> @combine_permvar_as_vpbroadcastq512(<8 x i64> %x0) {
   %1 = call <8 x i64> @llvm.x86.avx512.mask.permvar.di.512(<8 x i64> %x0, <8 x i64> zeroinitializer, <8 x i64> undef, i8 -1)
   ret <8 x i64> %1
 }
+
+define <64 x i8> @combine_pshufb_as_pslldq(<64 x i8> %a0) {
+; CHECK-LABEL: combine_pshufb_as_pslldq:
+; CHECK:       # BB#0:
+; CHECK-NEXT:    vpshufb {{.*#+}} zmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zmm0[0,1,2,3,4,5],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zmm0[16,17,18,19,20,21],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zmm0[32,33,34,35,36,37],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zmm0[48,49,50,51,52,53]
+; CHECK-NEXT:    retq
+  %res0 = call <64 x i8> @llvm.x86.avx512.mask.pshuf.b.512(<64 x i8> %a0, <64 x i8> <i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5>, <64 x i8> undef, i64 -1)
+  ret <64 x i8> %res0
+}
+define <64 x i8> @combine_pshufb_as_pslldq_mask(<64 x i8> %a0, i64 %m) {
+; CHECK-LABEL: combine_pshufb_as_pslldq_mask:
+; CHECK:       # BB#0:
+; CHECK-NEXT:    kmovq %rdi, %k1
+; CHECK-NEXT:    vpshufb {{.*#+}} zmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zmm0[0,1,2,3,4,5],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zmm0[16,17,18,19,20,21],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zmm0[32,33,34,35,36,37],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zmm0[48,49,50,51,52,53]
+; CHECK-NEXT:    retq
+  %res0 = call <64 x i8> @llvm.x86.avx512.mask.pshuf.b.512(<64 x i8> %a0, <64 x i8> <i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5>, <64 x i8> zeroinitializer, i64 %m)
+  ret <64 x i8> %res0
+}
+
+define <64 x i8> @combine_pshufb_as_psrldq(<64 x i8> %a0) {
+; CHECK-LABEL: combine_pshufb_as_psrldq:
+; CHECK:       # BB#0:
+; CHECK-NEXT:    vpshufb {{.*#+}} zmm0 = zmm0[15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zmm0[31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zmm0[47],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zmm0[63],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; CHECK-NEXT:    retq
+  %res0 = call <64 x i8> @llvm.x86.avx512.mask.pshuf.b.512(<64 x i8> %a0, <64 x i8> <i8 15, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 15, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 15, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 15, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128>, <64 x i8> undef, i64 -1)
+  ret <64 x i8> %res0
+}
+define <64 x i8> @combine_pshufb_as_psrldq_mask(<64 x i8> %a0, i64 %m) {
+; CHECK-LABEL: combine_pshufb_as_psrldq_mask:
+; CHECK:       # BB#0:
+; CHECK-NEXT:    kmovq %rdi, %k1
+; CHECK-NEXT:    vpshufb {{.*#+}} zmm0 = zmm0[15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zmm0[31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zmm0[47],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zmm0[63],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; CHECK-NEXT:    retq
+  %res0 = call <64 x i8> @llvm.x86.avx512.mask.pshuf.b.512(<64 x i8> %a0, <64 x i8> <i8 15, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 15, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 15, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 15, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128>, <64 x i8> zeroinitializer, i64 %m)
+  ret <64 x i8> %res0
+}
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-ssse3.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-ssse3.ll
index 4667cb45768..8ce96a4fdc7 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-ssse3.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-ssse3.ll
@@ -136,6 +136,34 @@ define <16 x i8> @combine_pshufb_psrldq(<16 x i8> %a0) {
   ret <16 x i8> %2
 }
 
+define <16 x i8> @combine_pshufb_as_pslldq(<16 x i8> %a0) {
+; SSE-LABEL: combine_pshufb_as_pslldq:
+; SSE:       # BB#0:
+; SSE-NEXT:    pshufb {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: combine_pshufb_as_pslldq:
+; AVX:       # BB#0:
+; AVX-NEXT:    vpshufb {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
+; AVX-NEXT:    retq
+  %res0 = call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %a0, <16 x i8> <i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5>)
+  ret <16 x i8> %res0
+}
+
+define <16 x i8> @combine_pshufb_as_psrldq(<16 x i8> %a0) {
+; SSE-LABEL: combine_pshufb_as_psrldq:
+; SSE:       # BB#0:
+; SSE-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: combine_pshufb_as_psrldq:
+; AVX:       # BB#0:
+; AVX-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX-NEXT:    retq
+  %res0 = call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %a0, <16 x i8> <i8 15, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128>)
+  ret <16 x i8> %res0
+}
+
 define <16 x i8> @combine_pshufb_as_unary_unpcklbw(<16 x i8> %a0) {
 ; SSE-LABEL: combine_pshufb_as_unary_unpcklbw:
 ; SSE:       # BB#0: