CodeGen peephole: fold redundant phys reg copies

[oota-llvm.git] / test / CodeGen / X86 / vector-shuffle-sse4a.ll

diff --git a/test/CodeGen/X86/vector-shuffle-sse4a.ll b/test/CodeGen/X86/vector-shuffle-sse4a.ll
index 58807b9c0fdcc4ac519dd494a1c004a884ae8e9b..eec915d91bbb0640a91b7fb4fcb4b010d4feee3e 100644 (file)
--- a/test/CodeGen/X86/vector-shuffle-sse4a.ll
+++ b/test/CodeGen/X86/vector-shuffle-sse4a.ll
@@ -1,3 +1,4 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+ssse3,+sse4a | FileCheck %s --check-prefix=ALL --check-prefix=BTVER1
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+sse4a | FileCheck %s --check-prefix=ALL --check-prefix=BTVER2
 
@@ -296,5 +297,65 @@ define <8 x i16> @shuf_089uuuuu(<8 x i16> %a0, <8 x i16> %a1) {
   ret <8 x i16> %s
 }
 
+;
+; Special Cases
+;
+
+; Out of range.
+define <16 x i8> @shuffle_8_18_uuuuuuuuuuuuuu(<16 x i8> %a, <16 x i8> %b) {
+; BTVER1-LABEL: shuffle_8_18_uuuuuuuuuuuuuu:
+; BTVER1:       # BB#0:
+; BTVER1-NEXT:    psrld $16, %xmm1
+; BTVER1-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; BTVER1-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; BTVER1-NEXT:    retq
+;
+; BTVER2-LABEL: shuffle_8_18_uuuuuuuuuuuuuu:
+; BTVER2:       # BB#0:
+; BTVER2-NEXT:    vpsrld $16, %xmm1, %xmm1
+; BTVER2-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; BTVER2-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; BTVER2-NEXT:    retq
+  %1 = shufflevector <16 x i8> %a, <16 x i8> %b, <16 x i32> <i32 8, i32 18, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+  ret <16 x i8> %1
+}
+
+define <16 x i8> @shuffle_uu_0_5_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu(<16 x i8> %v) {
+; BTVER1-LABEL: shuffle_uu_0_5_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu:
+; BTVER1:       # BB#0:
+; BTVER1-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,0,5,5,4,4,5,5,4,4,5,5,6,6,7,7]
+; BTVER1-NEXT:    retq
+;
+; BTVER2-LABEL: shuffle_uu_0_5_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu:
+; BTVER2:       # BB#0:
+; BTVER2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,5,5,4,4,5,5,4,4,5,5,6,6,7,7]
+; BTVER2-NEXT:    retq
+  %1 = shufflevector <16 x i8> %v, <16 x i8> zeroinitializer, <16 x i32> <i32 undef, i32 0, i32 5, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+  ret <16 x i8> %1
+}
+
+define <16 x i8> @shuffle_uu_16_4_16_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu(<16 x i8> %v) {
+; BTVER1-LABEL: shuffle_uu_16_4_16_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu:
+; BTVER1:       # BB#0:
+; BTVER1-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[u],zero,xmm0[4],zero,xmm0[u,u,u,u,u,u,u,u,u,u,u,u]
+; BTVER1-NEXT:    retq
+;
+; BTVER2-LABEL: shuffle_uu_16_4_16_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu:
+; BTVER2:       # BB#0:
+; BTVER2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[u],zero,xmm0[4],zero,xmm0[u,u,u,u,u,u,u,u,u,u,u,u]
+; BTVER2-NEXT:    retq
+  %1 = shufflevector <16 x i8> %v, <16 x i8> zeroinitializer, <16 x i32> <i32 undef, i32 16, i32 4, i32 16, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+  ret <16 x i8> %1
+}
+
+define <16 x i8> @shuffle_uu_uu_4_16_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu(<16 x i8> %v) {
+; ALL-LABEL: shuffle_uu_uu_4_16_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu:
+; ALL:       # BB#0:
+; ALL-NEXT:    extrq {{.*#+}} xmm0 = xmm0[2,3,4],zero,zero,zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
+; ALL-NEXT:    retq
+  %1 = shufflevector <16 x i8> %v, <16 x i8> zeroinitializer, <16 x i32> <i32 undef, i32 undef, i32 4, i32 16, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+  ret <16 x i8> %1
+}
+
 declare <2 x i64> @llvm.x86.sse4a.extrqi(<2 x i64>, i8, i8) nounwind
 declare <2 x i64> @llvm.x86.sse4a.insertqi(<2 x i64>, <2 x i64>, i8, i8) nounwind