[X86][AVX] Add 256-bit vector horizontal op redundant shuffle tests

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@337558 91177308-0d34-0410-b5e6-96231b3b80d8

1 files changed, 260 insertions, 1 deletions

diff --git a/test/CodeGen/X86/haddsub-shuf.ll b/test/CodeGen/X86/haddsub-shuf.ll
index 3b126b7b6df..8ede7035e7d 100644
--- a/test/CodeGen/X86/haddsub-shuf.ll
+++ b/test/CodeGen/X86/haddsub-shuf.ll
@@ -1,6 +1,7 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+ssse3 | FileCheck %s --check-prefix=SSSE3
-; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx | FileCheck %s --check-prefix=AVX
+; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx  | FileCheck %s --check-prefixes=AVX,AVX1
+; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=AVX,AVX2
 
 ; The next 8 tests check for matching the horizontal op and eliminating the shuffle.
 ; PR34111 - https://bugs.llvm.org/show_bug.cgi?id=34111
@@ -22,6 +23,55 @@ define <4 x float> @hadd_v4f32(<4 x float> %a) {
   ret <4 x float> %shuf
 }
 
+define <8 x float> @hadd_v8f32a(<8 x float> %a) {
+; SSSE3-LABEL: hadd_v8f32a:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movaps %xmm0, %xmm2
+; SSSE3-NEXT:    haddps %xmm1, %xmm2
+; SSSE3-NEXT:    movddup {{.*#+}} xmm0 = xmm2[0,0]
+; SSSE3-NEXT:    movaps %xmm2, %xmm1
+; SSSE3-NEXT:    retq
+;
+; AVX1-LABEL: hadd_v8f32a:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vhaddps %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovddup {{.*#+}} xmm1 = xmm0[0,0]
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: hadd_v8f32a:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-NEXT:    vhaddps %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,2,1]
+; AVX2-NEXT:    retq
+  %a0 = shufflevector <8 x float> %a, <8 x float> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+  %a1 = shufflevector <8 x float> %a, <8 x float> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+  %hop = fadd <4 x float> %a0, %a1
+  %shuf = shufflevector <4 x float> %hop, <4 x float> undef, <8 x i32> <i32 undef, i32 undef, i32 0, i32 1, i32 undef, i32 undef, i32 2, i32 3>
+  ret <8 x float> %shuf
+}
+
+define <8 x float> @hadd_v8f32b(<8 x float> %a) {
+; SSSE3-LABEL: hadd_v8f32b:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    haddps %xmm0, %xmm0
+; SSSE3-NEXT:    haddps %xmm1, %xmm1
+; SSSE3-NEXT:    retq
+;
+; AVX-LABEL: hadd_v8f32b:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vhaddps %ymm0, %ymm0, %ymm0
+; AVX-NEXT:    vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]
+; AVX-NEXT:    retq
+  %a0 = shufflevector <8 x float> %a, <8 x float> undef, <8 x i32> <i32 0, i32 2, i32 undef, i32 undef, i32 4, i32 6, i32 undef, i32 undef>
+  %a1 = shufflevector <8 x float> %a, <8 x float> undef, <8 x i32> <i32 1, i32 3, i32 undef, i32 undef, i32 5, i32 7, i32 undef, i32 undef>
+  %hop = fadd <8 x float> %a0, %a1
+  %shuf = shufflevector <8 x float> %hop, <8 x float> undef, <8 x i32> <i32 0, i32 1, i32 0, i32 1, i32 4, i32 5, i32 4, i32 5>
+  ret <8 x float> %shuf
+}
+
 define <4 x float> @hsub_v4f32(<4 x float> %a) {
 ; SSSE3-LABEL: hsub_v4f32:
 ; SSSE3:       # %bb.0:
@@ -39,6 +89,55 @@ define <4 x float> @hsub_v4f32(<4 x float> %a) {
   ret <4 x float> %shuf
 }
 
+define <8 x float> @hsub_v8f32a(<8 x float> %a) {
+; SSSE3-LABEL: hsub_v8f32a:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movaps %xmm0, %xmm2
+; SSSE3-NEXT:    hsubps %xmm1, %xmm2
+; SSSE3-NEXT:    movddup {{.*#+}} xmm0 = xmm2[0,0]
+; SSSE3-NEXT:    movaps %xmm2, %xmm1
+; SSSE3-NEXT:    retq
+;
+; AVX1-LABEL: hsub_v8f32a:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vhsubps %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovddup {{.*#+}} xmm1 = xmm0[0,0]
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: hsub_v8f32a:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-NEXT:    vhsubps %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,2,1]
+; AVX2-NEXT:    retq
+  %a0 = shufflevector <8 x float> %a, <8 x float> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+  %a1 = shufflevector <8 x float> %a, <8 x float> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+  %hop = fsub <4 x float> %a0, %a1
+  %shuf = shufflevector <4 x float> %hop, <4 x float> undef, <8 x i32> <i32 undef, i32 undef, i32 0, i32 1, i32 undef, i32 undef, i32 2, i32 3>
+  ret <8 x float> %shuf
+}
+
+define <8 x float> @hsub_v8f32b(<8 x float> %a) {
+; SSSE3-LABEL: hsub_v8f32b:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    hsubps %xmm0, %xmm0
+; SSSE3-NEXT:    hsubps %xmm1, %xmm1
+; SSSE3-NEXT:    retq
+;
+; AVX-LABEL: hsub_v8f32b:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vhsubps %ymm0, %ymm0, %ymm0
+; AVX-NEXT:    vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]
+; AVX-NEXT:    retq
+  %a0 = shufflevector <8 x float> %a, <8 x float> undef, <8 x i32> <i32 0, i32 2, i32 undef, i32 undef, i32 4, i32 6, i32 undef, i32 undef>
+  %a1 = shufflevector <8 x float> %a, <8 x float> undef, <8 x i32> <i32 1, i32 3, i32 undef, i32 undef, i32 5, i32 7, i32 undef, i32 undef>
+  %hop = fsub <8 x float> %a0, %a1
+  %shuf = shufflevector <8 x float> %hop, <8 x float> undef, <8 x i32> <i32 0, i32 1, i32 0, i32 1, i32 4, i32 5, i32 4, i32 5>
+  ret <8 x float> %shuf
+}
+
 define <2 x double> @hadd_v2f64(<2 x double> %a) {
 ; SSSE3-LABEL: hadd_v2f64:
 ; SSSE3:       # %bb.0:
@@ -56,6 +155,25 @@ define <2 x double> @hadd_v2f64(<2 x double> %a) {
   ret <2 x double> %shuf
 }
 
+define <4 x double> @hadd_v4f64(<4 x double> %a) {
+; SSSE3-LABEL: hadd_v4f64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    haddpd %xmm0, %xmm0
+; SSSE3-NEXT:    haddpd %xmm1, %xmm1
+; SSSE3-NEXT:    retq
+;
+; AVX-LABEL: hadd_v4f64:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vhaddpd %ymm0, %ymm0, %ymm0
+; AVX-NEXT:    vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]
+; AVX-NEXT:    retq
+  %a0 = shufflevector <4 x double> %a, <4 x double> undef, <4 x i32> <i32 0, i32 undef, i32 2, i32 undef>
+  %a1 = shufflevector <4 x double> %a, <4 x double> undef, <4 x i32> <i32 1, i32 undef, i32 3, i32 undef>
+  %hop = fadd <4 x double> %a0, %a1
+  %shuf = shufflevector <4 x double> %hop, <4 x double> undef, <4 x i32> <i32 0, i32 0, i32 2, i32 2>
+  ret <4 x double> %shuf
+}
+
 define <2 x double> @hsub_v2f64(<2 x double> %a) {
 ; SSSE3-LABEL: hsub_v2f64:
 ; SSSE3:       # %bb.0:
@@ -73,6 +191,25 @@ define <2 x double> @hsub_v2f64(<2 x double> %a) {
   ret <2 x double> %shuf
 }
 
+define <4 x double> @hsub_v4f64(<4 x double> %a) {
+; SSSE3-LABEL: hsub_v4f64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    hsubpd %xmm0, %xmm0
+; SSSE3-NEXT:    hsubpd %xmm1, %xmm1
+; SSSE3-NEXT:    retq
+;
+; AVX-LABEL: hsub_v4f64:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vhsubpd %ymm0, %ymm0, %ymm0
+; AVX-NEXT:    vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]
+; AVX-NEXT:    retq
+  %a0 = shufflevector <4 x double> %a, <4 x double> undef, <4 x i32> <i32 0, i32 undef, i32 2, i32 undef>
+  %a1 = shufflevector <4 x double> %a, <4 x double> undef, <4 x i32> <i32 1, i32 undef, i32 3, i32 undef>
+  %hop = fsub <4 x double> %a0, %a1
+  %shuf = shufflevector <4 x double> %hop, <4 x double> undef, <4 x i32> <i32 0, i32 0, i32 2, i32 2>
+  ret <4 x double> %shuf
+}
+
 define <4 x i32> @hadd_v4i32(<4 x i32> %a) {
 ; SSSE3-LABEL: hadd_v4i32:
 ; SSSE3:       # %bb.0:
@@ -90,6 +227,67 @@ define <4 x i32> @hadd_v4i32(<4 x i32> %a) {
   ret <4 x i32> %shuf
 }
 
+define <8 x i32> @hadd_v8i32a(<8 x i32> %a) {
+; SSSE3-LABEL: hadd_v8i32a:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    phaddd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,1,0,1]
+; SSSE3-NEXT:    movdqa %xmm2, %xmm1
+; SSSE3-NEXT:    retq
+;
+; AVX1-LABEL: hadd_v8i32a:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vphaddd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[0,1,0,1]
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: hadd_v8i32a:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT:    vphaddd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,2,1]
+; AVX2-NEXT:    retq
+  %a0 = shufflevector <8 x i32> %a, <8 x i32> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+  %a1 = shufflevector <8 x i32> %a, <8 x i32> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+  %hop = add <4 x i32> %a0, %a1
+  %shuf = shufflevector <4 x i32> %hop, <4 x i32> undef, <8 x i32> <i32 undef, i32 undef, i32 0, i32 1, i32 undef, i32 undef, i32 2, i32 3>
+  ret <8 x i32> %shuf
+}
+
+define <8 x i32> @hadd_v8i32b(<8 x i32> %a) {
+; SSSE3-LABEL: hadd_v8i32b:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    phaddd %xmm0, %xmm0
+; SSSE3-NEXT:    phaddd %xmm1, %xmm1
+; SSSE3-NEXT:    retq
+;
+; AVX1-LABEL: hadd_v8i32b:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm1 = ymm0[0,2,2,3,4,6,6,7]
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,3,2,3,5,7,6,7]
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
+; AVX1-NEXT:    vpaddd %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpaddd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT:    vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: hadd_v8i32b:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vphaddd %ymm0, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
+; AVX2-NEXT:    retq
+  %a0 = shufflevector <8 x i32> %a, <8 x i32> undef, <8 x i32> <i32 0, i32 2, i32 undef, i32 undef, i32 4, i32 6, i32 undef, i32 undef>
+  %a1 = shufflevector <8 x i32> %a, <8 x i32> undef, <8 x i32> <i32 1, i32 3, i32 undef, i32 undef, i32 5, i32 7, i32 undef, i32 undef>
+  %hop = add <8 x i32> %a0, %a1
+  %shuf = shufflevector <8 x i32> %hop, <8 x i32> undef, <8 x i32> <i32 0, i32 1, i32 0, i32 1, i32 4, i32 5, i32 4, i32 5>
+  ret <8 x i32> %shuf
+}
+
 define <4 x i32> @hsub_v4i32(<4 x i32> %a) {
 ; SSSE3-LABEL: hsub_v4i32:
 ; SSSE3:       # %bb.0:
@@ -107,6 +305,67 @@ define <4 x i32> @hsub_v4i32(<4 x i32> %a) {
   ret <4 x i32> %shuf
 }
 
+define <8 x i32> @hsub_v8i32a(<8 x i32> %a) {
+; SSSE3-LABEL: hsub_v8i32a:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    phsubd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,1,0,1]
+; SSSE3-NEXT:    movdqa %xmm2, %xmm1
+; SSSE3-NEXT:    retq
+;
+; AVX1-LABEL: hsub_v8i32a:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vphsubd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[0,1,0,1]
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: hsub_v8i32a:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT:    vphsubd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,2,1]
+; AVX2-NEXT:    retq
+  %a0 = shufflevector <8 x i32> %a, <8 x i32> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+  %a1 = shufflevector <8 x i32> %a, <8 x i32> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+  %hop = sub <4 x i32> %a0, %a1
+  %shuf = shufflevector <4 x i32> %hop, <4 x i32> undef, <8 x i32> <i32 undef, i32 undef, i32 0, i32 1, i32 undef, i32 undef, i32 2, i32 3>
+  ret <8 x i32> %shuf
+}
+
+define <8 x i32> @hsub_v8i32b(<8 x i32> %a) {
+; SSSE3-LABEL: hsub_v8i32b:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    phsubd %xmm0, %xmm0
+; SSSE3-NEXT:    phsubd %xmm1, %xmm1
+; SSSE3-NEXT:    retq
+;
+; AVX1-LABEL: hsub_v8i32b:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm1 = ymm0[0,2,2,3,4,6,6,7]
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,3,2,3,5,7,6,7]
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
+; AVX1-NEXT:    vpsubd %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsubd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT:    vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: hsub_v8i32b:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vphsubd %ymm0, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
+; AVX2-NEXT:    retq
+  %a0 = shufflevector <8 x i32> %a, <8 x i32> undef, <8 x i32> <i32 0, i32 2, i32 undef, i32 undef, i32 4, i32 6, i32 undef, i32 undef>
+  %a1 = shufflevector <8 x i32> %a, <8 x i32> undef, <8 x i32> <i32 1, i32 3, i32 undef, i32 undef, i32 5, i32 7, i32 undef, i32 undef>
+  %hop = sub <8 x i32> %a0, %a1
+  %shuf = shufflevector <8 x i32> %hop, <8 x i32> undef, <8 x i32> <i32 0, i32 1, i32 0, i32 1, i32 4, i32 5, i32 4, i32 5>
+  ret <8 x i32> %shuf
+}
+
 define <8 x i16> @hadd_v8i16(<8 x i16> %a) {
 ; SSSE3-LABEL: hadd_v8i16:
 ; SSSE3:       # %bb.0: