aboutsummaryrefslogtreecommitdiff
path: root/test/CodeGen/X86/haddsub-2.ll
diff options
context:
space:
mode:
Diffstat (limited to 'test/CodeGen/X86/haddsub-2.ll')
-rw-r--r--test/CodeGen/X86/haddsub-2.ll1017
1 files changed, 842 insertions, 175 deletions
diff --git a/test/CodeGen/X86/haddsub-2.ll b/test/CodeGen/X86/haddsub-2.ll
index c6bac5858807..517a663bc815 100644
--- a/test/CodeGen/X86/haddsub-2.ll
+++ b/test/CodeGen/X86/haddsub-2.ll
@@ -1,11 +1,19 @@
-; RUN: llc < %s -march=x86-64 -mattr=+sse2,+sse3 | FileCheck %s -check-prefix=CHECK -check-prefix=SSE3
-; RUN: llc < %s -march=x86-64 -mattr=+sse2,+sse3,+ssse3 | FileCheck %s -check-prefix=CHECK -check-prefix=SSSE3
-; RUN: llc < %s -march=x86-64 -mattr=+avx | FileCheck %s -check-prefix=CHECK -check-prefix=AVX
-; RUN: llc < %s -march=x86-64 -mattr=+avx2 | FileCheck %s -check-prefix=CHECK -check-prefix=AVX2
-
-
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+sse2,+sse3 | FileCheck %s --check-prefix=SSE --check-prefix=SSE3
+; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+sse2,+sse3,+ssse3 | FileCheck %s --check-prefix=SSE --check-prefix=SSSE3
+; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx | FileCheck %s --check-prefix=AVX --check-prefix=AVX1
+; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx2 | FileCheck %s --check-prefix=AVX --check-prefix=AVX2
define <4 x float> @hadd_ps_test1(<4 x float> %A, <4 x float> %B) {
+; SSE-LABEL: hadd_ps_test1:
+; SSE: # BB#0:
+; SSE-NEXT: haddps %xmm1, %xmm0
+; SSE-NEXT: retq
+;
+; AVX-LABEL: hadd_ps_test1:
+; AVX: # BB#0:
+; AVX-NEXT: vhaddps %xmm1, %xmm0, %xmm0
+; AVX-NEXT: retq
%vecext = extractelement <4 x float> %A, i32 0
%vecext1 = extractelement <4 x float> %A, i32 1
%add = fadd float %vecext, %vecext1
@@ -24,12 +32,17 @@ define <4 x float> @hadd_ps_test1(<4 x float> %A, <4 x float> %B) {
%vecinit13 = insertelement <4 x float> %vecinit9, float %add12, i32 3
ret <4 x float> %vecinit13
}
-; CHECK-LABEL: hadd_ps_test1
-; CHECK: haddps
-; CHECK-NEXT: ret
-
define <4 x float> @hadd_ps_test2(<4 x float> %A, <4 x float> %B) {
+; SSE-LABEL: hadd_ps_test2:
+; SSE: # BB#0:
+; SSE-NEXT: haddps %xmm1, %xmm0
+; SSE-NEXT: retq
+;
+; AVX-LABEL: hadd_ps_test2:
+; AVX: # BB#0:
+; AVX-NEXT: vhaddps %xmm1, %xmm0, %xmm0
+; AVX-NEXT: retq
%vecext = extractelement <4 x float> %A, i32 2
%vecext1 = extractelement <4 x float> %A, i32 3
%add = fadd float %vecext, %vecext1
@@ -48,12 +61,17 @@ define <4 x float> @hadd_ps_test2(<4 x float> %A, <4 x float> %B) {
%vecinit13 = insertelement <4 x float> %vecinit9, float %add12, i32 2
ret <4 x float> %vecinit13
}
-; CHECK-LABEL: hadd_ps_test2
-; CHECK: haddps
-; CHECK-NEXT: ret
-
define <4 x float> @hsub_ps_test1(<4 x float> %A, <4 x float> %B) {
+; SSE-LABEL: hsub_ps_test1:
+; SSE: # BB#0:
+; SSE-NEXT: hsubps %xmm1, %xmm0
+; SSE-NEXT: retq
+;
+; AVX-LABEL: hsub_ps_test1:
+; AVX: # BB#0:
+; AVX-NEXT: vhsubps %xmm1, %xmm0, %xmm0
+; AVX-NEXT: retq
%vecext = extractelement <4 x float> %A, i32 0
%vecext1 = extractelement <4 x float> %A, i32 1
%sub = fsub float %vecext, %vecext1
@@ -72,12 +90,17 @@ define <4 x float> @hsub_ps_test1(<4 x float> %A, <4 x float> %B) {
%vecinit13 = insertelement <4 x float> %vecinit9, float %sub12, i32 3
ret <4 x float> %vecinit13
}
-; CHECK-LABEL: hsub_ps_test1
-; CHECK: hsubps
-; CHECK-NEXT: ret
-
define <4 x float> @hsub_ps_test2(<4 x float> %A, <4 x float> %B) {
+; SSE-LABEL: hsub_ps_test2:
+; SSE: # BB#0:
+; SSE-NEXT: hsubps %xmm1, %xmm0
+; SSE-NEXT: retq
+;
+; AVX-LABEL: hsub_ps_test2:
+; AVX: # BB#0:
+; AVX-NEXT: vhsubps %xmm1, %xmm0, %xmm0
+; AVX-NEXT: retq
%vecext = extractelement <4 x float> %A, i32 2
%vecext1 = extractelement <4 x float> %A, i32 3
%sub = fsub float %vecext, %vecext1
@@ -96,12 +119,46 @@ define <4 x float> @hsub_ps_test2(<4 x float> %A, <4 x float> %B) {
%vecinit13 = insertelement <4 x float> %vecinit9, float %sub12, i32 2
ret <4 x float> %vecinit13
}
-; CHECK-LABEL: hsub_ps_test2
-; CHECK: hsubps
-; CHECK-NEXT: ret
-
define <4 x i32> @phadd_d_test1(<4 x i32> %A, <4 x i32> %B) {
+; SSE3-LABEL: phadd_d_test1:
+; SSE3: # BB#0:
+; SSE3-NEXT: movd %xmm0, %eax
+; SSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,2,3]
+; SSE3-NEXT: movd %xmm2, %ecx
+; SSE3-NEXT: addl %eax, %ecx
+; SSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm0[2,3,0,1]
+; SSE3-NEXT: movd %xmm2, %eax
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
+; SSE3-NEXT: movd %xmm0, %edx
+; SSE3-NEXT: addl %eax, %edx
+; SSE3-NEXT: movd %xmm1, %eax
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
+; SSE3-NEXT: movd %xmm0, %esi
+; SSE3-NEXT: addl %eax, %esi
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
+; SSE3-NEXT: movd %xmm0, %eax
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm1[3,1,2,3]
+; SSE3-NEXT: movd %xmm0, %edi
+; SSE3-NEXT: addl %eax, %edi
+; SSE3-NEXT: movd %edi, %xmm0
+; SSE3-NEXT: movd %edx, %xmm1
+; SSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT: movd %esi, %xmm2
+; SSE3-NEXT: movd %ecx, %xmm0
+; SSE3-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE3-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: phadd_d_test1:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: phaddd %xmm1, %xmm0
+; SSSE3-NEXT: retq
+;
+; AVX-LABEL: phadd_d_test1:
+; AVX: # BB#0:
+; AVX-NEXT: vphaddd %xmm1, %xmm0, %xmm0
+; AVX-NEXT: retq
%vecext = extractelement <4 x i32> %A, i32 0
%vecext1 = extractelement <4 x i32> %A, i32 1
%add = add i32 %vecext, %vecext1
@@ -120,15 +177,46 @@ define <4 x i32> @phadd_d_test1(<4 x i32> %A, <4 x i32> %B) {
%vecinit13 = insertelement <4 x i32> %vecinit9, i32 %add12, i32 3
ret <4 x i32> %vecinit13
}
-; CHECK-LABEL: phadd_d_test1
-; SSE3-NOT: phaddd
-; SSSE3: phaddd
-; AVX: vphaddd
-; AVX2 vphaddd
-; CHECK: ret
-
define <4 x i32> @phadd_d_test2(<4 x i32> %A, <4 x i32> %B) {
+; SSE3-LABEL: phadd_d_test2:
+; SSE3: # BB#0:
+; SSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm0[2,3,0,1]
+; SSE3-NEXT: movd %xmm2, %eax
+; SSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm0[3,1,2,3]
+; SSE3-NEXT: movd %xmm2, %ecx
+; SSE3-NEXT: addl %eax, %ecx
+; SSE3-NEXT: movd %xmm0, %eax
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
+; SSE3-NEXT: movd %xmm0, %edx
+; SSE3-NEXT: addl %eax, %edx
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm1[3,1,2,3]
+; SSE3-NEXT: movd %xmm0, %eax
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
+; SSE3-NEXT: movd %xmm0, %esi
+; SSE3-NEXT: addl %eax, %esi
+; SSE3-NEXT: movd %esi, %xmm0
+; SSE3-NEXT: movd %ecx, %xmm2
+; SSE3-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
+; SSE3-NEXT: movd %xmm0, %eax
+; SSE3-NEXT: movd %xmm1, %ecx
+; SSE3-NEXT: addl %eax, %ecx
+; SSE3-NEXT: movd %ecx, %xmm1
+; SSE3-NEXT: movd %edx, %xmm0
+; SSE3-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE3-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: phadd_d_test2:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: phaddd %xmm1, %xmm0
+; SSSE3-NEXT: retq
+;
+; AVX-LABEL: phadd_d_test2:
+; AVX: # BB#0:
+; AVX-NEXT: vphaddd %xmm1, %xmm0, %xmm0
+; AVX-NEXT: retq
%vecext = extractelement <4 x i32> %A, i32 2
%vecext1 = extractelement <4 x i32> %A, i32 3
%add = add i32 %vecext, %vecext1
@@ -147,15 +235,46 @@ define <4 x i32> @phadd_d_test2(<4 x i32> %A, <4 x i32> %B) {
%vecinit13 = insertelement <4 x i32> %vecinit9, i32 %add12, i32 2
ret <4 x i32> %vecinit13
}
-; CHECK-LABEL: phadd_d_test2
-; SSE3-NOT: phaddd
-; SSSE3: phaddd
-; AVX: vphaddd
-; AVX2 vphaddd
-; CHECK: ret
-
define <4 x i32> @phsub_d_test1(<4 x i32> %A, <4 x i32> %B) {
+; SSE3-LABEL: phsub_d_test1:
+; SSE3: # BB#0:
+; SSE3-NEXT: movd %xmm0, %eax
+; SSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,2,3]
+; SSE3-NEXT: movd %xmm2, %ecx
+; SSE3-NEXT: subl %ecx, %eax
+; SSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm0[2,3,0,1]
+; SSE3-NEXT: movd %xmm2, %ecx
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
+; SSE3-NEXT: movd %xmm0, %edx
+; SSE3-NEXT: subl %edx, %ecx
+; SSE3-NEXT: movd %xmm1, %edx
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
+; SSE3-NEXT: movd %xmm0, %esi
+; SSE3-NEXT: subl %esi, %edx
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
+; SSE3-NEXT: movd %xmm0, %esi
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm1[3,1,2,3]
+; SSE3-NEXT: movd %xmm0, %edi
+; SSE3-NEXT: subl %edi, %esi
+; SSE3-NEXT: movd %esi, %xmm0
+; SSE3-NEXT: movd %ecx, %xmm1
+; SSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT: movd %edx, %xmm2
+; SSE3-NEXT: movd %eax, %xmm0
+; SSE3-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE3-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: phsub_d_test1:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: phsubd %xmm1, %xmm0
+; SSSE3-NEXT: retq
+;
+; AVX-LABEL: phsub_d_test1:
+; AVX: # BB#0:
+; AVX-NEXT: vphsubd %xmm1, %xmm0, %xmm0
+; AVX-NEXT: retq
%vecext = extractelement <4 x i32> %A, i32 0
%vecext1 = extractelement <4 x i32> %A, i32 1
%sub = sub i32 %vecext, %vecext1
@@ -174,15 +293,46 @@ define <4 x i32> @phsub_d_test1(<4 x i32> %A, <4 x i32> %B) {
%vecinit13 = insertelement <4 x i32> %vecinit9, i32 %sub12, i32 3
ret <4 x i32> %vecinit13
}
-; CHECK-LABEL: phsub_d_test1
-; SSE3-NOT: phsubd
-; SSSE3: phsubd
-; AVX: vphsubd
-; AVX2 vphsubd
-; CHECK: ret
-
define <4 x i32> @phsub_d_test2(<4 x i32> %A, <4 x i32> %B) {
+; SSE3-LABEL: phsub_d_test2:
+; SSE3: # BB#0:
+; SSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm0[2,3,0,1]
+; SSE3-NEXT: movd %xmm2, %eax
+; SSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm0[3,1,2,3]
+; SSE3-NEXT: movd %xmm2, %ecx
+; SSE3-NEXT: subl %ecx, %eax
+; SSE3-NEXT: movd %xmm0, %ecx
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
+; SSE3-NEXT: movd %xmm0, %edx
+; SSE3-NEXT: subl %edx, %ecx
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
+; SSE3-NEXT: movd %xmm0, %edx
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm1[3,1,2,3]
+; SSE3-NEXT: movd %xmm0, %esi
+; SSE3-NEXT: subl %esi, %edx
+; SSE3-NEXT: movd %edx, %xmm0
+; SSE3-NEXT: movd %eax, %xmm2
+; SSE3-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE3-NEXT: movd %xmm1, %eax
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
+; SSE3-NEXT: movd %xmm0, %edx
+; SSE3-NEXT: subl %edx, %eax
+; SSE3-NEXT: movd %eax, %xmm1
+; SSE3-NEXT: movd %ecx, %xmm0
+; SSE3-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE3-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: phsub_d_test2:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: phsubd %xmm1, %xmm0
+; SSSE3-NEXT: retq
+;
+; AVX-LABEL: phsub_d_test2:
+; AVX: # BB#0:
+; AVX-NEXT: vphsubd %xmm1, %xmm0, %xmm0
+; AVX-NEXT: retq
%vecext = extractelement <4 x i32> %A, i32 2
%vecext1 = extractelement <4 x i32> %A, i32 3
%sub = sub i32 %vecext, %vecext1
@@ -201,15 +351,17 @@ define <4 x i32> @phsub_d_test2(<4 x i32> %A, <4 x i32> %B) {
%vecinit13 = insertelement <4 x i32> %vecinit9, i32 %sub12, i32 2
ret <4 x i32> %vecinit13
}
-; CHECK-LABEL: phsub_d_test2
-; SSE3-NOT: phsubd
-; SSSE3: phsubd
-; AVX: vphsubd
-; AVX2 vphsubd
-; CHECK: ret
-
define <2 x double> @hadd_pd_test1(<2 x double> %A, <2 x double> %B) {
+; SSE-LABEL: hadd_pd_test1:
+; SSE: # BB#0:
+; SSE-NEXT: haddpd %xmm1, %xmm0
+; SSE-NEXT: retq
+;
+; AVX-LABEL: hadd_pd_test1:
+; AVX: # BB#0:
+; AVX-NEXT: vhaddpd %xmm1, %xmm0, %xmm0
+; AVX-NEXT: retq
%vecext = extractelement <2 x double> %A, i32 0
%vecext1 = extractelement <2 x double> %A, i32 1
%add = fadd double %vecext, %vecext1
@@ -220,12 +372,17 @@ define <2 x double> @hadd_pd_test1(<2 x double> %A, <2 x double> %B) {
%vecinit2 = insertelement <2 x double> %vecinit, double %add2, i32 1
ret <2 x double> %vecinit2
}
-; CHECK-LABEL: hadd_pd_test1
-; CHECK: haddpd
-; CHECK-NEXT: ret
-
define <2 x double> @hadd_pd_test2(<2 x double> %A, <2 x double> %B) {
+; SSE-LABEL: hadd_pd_test2:
+; SSE: # BB#0:
+; SSE-NEXT: haddpd %xmm1, %xmm0
+; SSE-NEXT: retq
+;
+; AVX-LABEL: hadd_pd_test2:
+; AVX: # BB#0:
+; AVX-NEXT: vhaddpd %xmm1, %xmm0, %xmm0
+; AVX-NEXT: retq
%vecext = extractelement <2 x double> %A, i32 1
%vecext1 = extractelement <2 x double> %A, i32 0
%add = fadd double %vecext, %vecext1
@@ -236,12 +393,17 @@ define <2 x double> @hadd_pd_test2(<2 x double> %A, <2 x double> %B) {
%vecinit2 = insertelement <2 x double> %vecinit, double %add2, i32 1
ret <2 x double> %vecinit2
}
-; CHECK-LABEL: hadd_pd_test2
-; CHECK: haddpd
-; CHECK-NEXT: ret
-
define <2 x double> @hsub_pd_test1(<2 x double> %A, <2 x double> %B) {
+; SSE-LABEL: hsub_pd_test1:
+; SSE: # BB#0:
+; SSE-NEXT: hsubpd %xmm1, %xmm0
+; SSE-NEXT: retq
+;
+; AVX-LABEL: hsub_pd_test1:
+; AVX: # BB#0:
+; AVX-NEXT: vhsubpd %xmm1, %xmm0, %xmm0
+; AVX-NEXT: retq
%vecext = extractelement <2 x double> %A, i32 0
%vecext1 = extractelement <2 x double> %A, i32 1
%sub = fsub double %vecext, %vecext1
@@ -252,12 +414,17 @@ define <2 x double> @hsub_pd_test1(<2 x double> %A, <2 x double> %B) {
%vecinit2 = insertelement <2 x double> %vecinit, double %sub2, i32 1
ret <2 x double> %vecinit2
}
-; CHECK-LABEL: hsub_pd_test1
-; CHECK: hsubpd
-; CHECK-NEXT: ret
-
define <2 x double> @hsub_pd_test2(<2 x double> %A, <2 x double> %B) {
+; SSE-LABEL: hsub_pd_test2:
+; SSE: # BB#0:
+; SSE-NEXT: hsubpd %xmm1, %xmm0
+; SSE-NEXT: retq
+;
+; AVX-LABEL: hsub_pd_test2:
+; AVX: # BB#0:
+; AVX-NEXT: vhsubpd %xmm1, %xmm0, %xmm0
+; AVX-NEXT: retq
%vecext = extractelement <2 x double> %B, i32 0
%vecext1 = extractelement <2 x double> %B, i32 1
%sub = fsub double %vecext, %vecext1
@@ -268,12 +435,23 @@ define <2 x double> @hsub_pd_test2(<2 x double> %A, <2 x double> %B) {
%vecinit2 = insertelement <2 x double> %vecinit, double %sub2, i32 0
ret <2 x double> %vecinit2
}
-; CHECK-LABEL: hsub_pd_test2
-; CHECK: hsubpd
-; CHECK-NEXT: ret
-
define <4 x double> @avx_vhadd_pd_test(<4 x double> %A, <4 x double> %B) {
+; SSE-LABEL: avx_vhadd_pd_test:
+; SSE: # BB#0:
+; SSE-NEXT: haddpd %xmm1, %xmm0
+; SSE-NEXT: haddpd %xmm3, %xmm2
+; SSE-NEXT: movapd %xmm2, %xmm1
+; SSE-NEXT: retq
+;
+; AVX-LABEL: avx_vhadd_pd_test:
+; AVX: # BB#0:
+; AVX-NEXT: vextractf128 $1, %ymm1, %xmm2
+; AVX-NEXT: vhaddpd %xmm2, %xmm1, %xmm1
+; AVX-NEXT: vextractf128 $1, %ymm0, %xmm2
+; AVX-NEXT: vhaddpd %xmm2, %xmm0, %xmm0
+; AVX-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX-NEXT: retq
%vecext = extractelement <4 x double> %A, i32 0
%vecext1 = extractelement <4 x double> %A, i32 1
%add = fadd double %vecext, %vecext1
@@ -292,19 +470,23 @@ define <4 x double> @avx_vhadd_pd_test(<4 x double> %A, <4 x double> %B) {
%vecinit13 = insertelement <4 x double> %vecinit9, double %add12, i32 3
ret <4 x double> %vecinit13
}
-; CHECK-LABEL: avx_vhadd_pd_test
-; SSE3: haddpd
-; SSE3-NEXT: haddpd
-; SSSE3: haddpd
-; SSSE3: haddpd
-; AVX: vhaddpd
-; AVX: vhaddpd
-; AVX2: vhaddpd
-; AVX2: vhaddpd
-; CHECK: ret
-
define <4 x double> @avx_vhsub_pd_test(<4 x double> %A, <4 x double> %B) {
+; SSE-LABEL: avx_vhsub_pd_test:
+; SSE: # BB#0:
+; SSE-NEXT: hsubpd %xmm1, %xmm0
+; SSE-NEXT: hsubpd %xmm3, %xmm2
+; SSE-NEXT: movapd %xmm2, %xmm1
+; SSE-NEXT: retq
+;
+; AVX-LABEL: avx_vhsub_pd_test:
+; AVX: # BB#0:
+; AVX-NEXT: vextractf128 $1, %ymm1, %xmm2
+; AVX-NEXT: vhsubpd %xmm2, %xmm1, %xmm1
+; AVX-NEXT: vextractf128 $1, %ymm0, %xmm2
+; AVX-NEXT: vhsubpd %xmm2, %xmm0, %xmm0
+; AVX-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX-NEXT: retq
%vecext = extractelement <4 x double> %A, i32 0
%vecext1 = extractelement <4 x double> %A, i32 1
%sub = fsub double %vecext, %vecext1
@@ -323,19 +505,86 @@ define <4 x double> @avx_vhsub_pd_test(<4 x double> %A, <4 x double> %B) {
%vecinit13 = insertelement <4 x double> %vecinit9, double %sub12, i32 3
ret <4 x double> %vecinit13
}
-; CHECK-LABEL: avx_vhsub_pd_test
-; SSE3: hsubpd
-; SSE3-NEXT: hsubpd
-; SSSE3: hsubpd
-; SSSE3-NEXT: hsubpd
-; AVX: vhsubpd
-; AVX: vhsubpd
-; AVX2: vhsubpd
-; AVX2: vhsubpd
-; CHECK: ret
-
define <8 x i32> @avx2_vphadd_d_test(<8 x i32> %A, <8 x i32> %B) {
+; SSE3-LABEL: avx2_vphadd_d_test:
+; SSE3: # BB#0:
+; SSE3-NEXT: movd %xmm0, %ecx
+; SSE3-NEXT: pshufd {{.*#+}} xmm4 = xmm0[1,1,2,3]
+; SSE3-NEXT: movd %xmm4, %r8d
+; SSE3-NEXT: addl %ecx, %r8d
+; SSE3-NEXT: pshufd {{.*#+}} xmm4 = xmm0[2,3,0,1]
+; SSE3-NEXT: movd %xmm4, %edx
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
+; SSE3-NEXT: movd %xmm0, %r9d
+; SSE3-NEXT: addl %edx, %r9d
+; SSE3-NEXT: movd %xmm1, %esi
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
+; SSE3-NEXT: movd %xmm0, %r10d
+; SSE3-NEXT: addl %esi, %r10d
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
+; SSE3-NEXT: movd %xmm0, %esi
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm1[3,1,2,3]
+; SSE3-NEXT: movd %xmm0, %edi
+; SSE3-NEXT: addl %esi, %edi
+; SSE3-NEXT: movd %xmm2, %eax
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm2[1,1,2,3]
+; SSE3-NEXT: movd %xmm0, %r11d
+; SSE3-NEXT: addl %eax, %r11d
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,3,0,1]
+; SSE3-NEXT: movd %xmm0, %eax
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm2[3,1,2,3]
+; SSE3-NEXT: movd %xmm0, %ecx
+; SSE3-NEXT: addl %eax, %ecx
+; SSE3-NEXT: movd %xmm3, %eax
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,1,2,3]
+; SSE3-NEXT: movd %xmm0, %edx
+; SSE3-NEXT: addl %eax, %edx
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm3[2,3,0,1]
+; SSE3-NEXT: movd %xmm0, %eax
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm3[3,1,2,3]
+; SSE3-NEXT: movd %xmm0, %esi
+; SSE3-NEXT: addl %eax, %esi
+; SSE3-NEXT: movd %edi, %xmm0
+; SSE3-NEXT: movd %r9d, %xmm1
+; SSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT: movd %r10d, %xmm2
+; SSE3-NEXT: movd %r8d, %xmm0
+; SSE3-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE3-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE3-NEXT: movd %esi, %xmm1
+; SSE3-NEXT: movd %ecx, %xmm2
+; SSE3-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE3-NEXT: movd %edx, %xmm3
+; SSE3-NEXT: movd %r11d, %xmm1
+; SSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; SSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: avx2_vphadd_d_test:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: phaddd %xmm1, %xmm0
+; SSSE3-NEXT: phaddd %xmm3, %xmm2
+; SSSE3-NEXT: movdqa %xmm2, %xmm1
+; SSSE3-NEXT: retq
+;
+; AVX1-LABEL: avx2_vphadd_d_test:
+; AVX1: # BB#0:
+; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT: vphaddd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
+; AVX1-NEXT: vphaddd %xmm2, %xmm0, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: avx2_vphadd_d_test:
+; AVX2: # BB#0:
+; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2
+; AVX2-NEXT: vphaddd %xmm2, %xmm1, %xmm1
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm2
+; AVX2-NEXT: vphaddd %xmm2, %xmm0, %xmm0
+; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT: retq
%vecext = extractelement <8 x i32> %A, i32 0
%vecext1 = extractelement <8 x i32> %A, i32 1
%add = add i32 %vecext, %vecext1
@@ -370,17 +619,154 @@ define <8 x i32> @avx2_vphadd_d_test(<8 x i32> %A, <8 x i32> %B) {
%vecinit29 = insertelement <8 x i32> %vecinit25, i32 %add28, i32 7
ret <8 x i32> %vecinit29
}
-; CHECK-LABEL: avx2_vphadd_d_test
-; SSE3-NOT: phaddd
-; SSSE3: phaddd
-; SSSE3-NEXT: phaddd
-; AVX: vphaddd
-; AVX: vphaddd
-; AVX2: vphaddd
-; AVX2: vphaddd
-; CHECK: ret
define <16 x i16> @avx2_vphadd_w_test(<16 x i16> %a, <16 x i16> %b) {
+; SSE3-LABEL: avx2_vphadd_w_test:
+; SSE3: # BB#0:
+; SSE3-NEXT: pushq %rbp
+; SSE3-NEXT: .Ltmp0:
+; SSE3-NEXT: .cfi_def_cfa_offset 16
+; SSE3-NEXT: pushq %r15
+; SSE3-NEXT: .Ltmp1:
+; SSE3-NEXT: .cfi_def_cfa_offset 24
+; SSE3-NEXT: pushq %r14
+; SSE3-NEXT: .Ltmp2:
+; SSE3-NEXT: .cfi_def_cfa_offset 32
+; SSE3-NEXT: pushq %r13
+; SSE3-NEXT: .Ltmp3:
+; SSE3-NEXT: .cfi_def_cfa_offset 40
+; SSE3-NEXT: pushq %r12
+; SSE3-NEXT: .Ltmp4:
+; SSE3-NEXT: .cfi_def_cfa_offset 48
+; SSE3-NEXT: pushq %rbx
+; SSE3-NEXT: .Ltmp5:
+; SSE3-NEXT: .cfi_def_cfa_offset 56
+; SSE3-NEXT: .Ltmp6:
+; SSE3-NEXT: .cfi_offset %rbx, -56
+; SSE3-NEXT: .Ltmp7:
+; SSE3-NEXT: .cfi_offset %r12, -48
+; SSE3-NEXT: .Ltmp8:
+; SSE3-NEXT: .cfi_offset %r13, -40
+; SSE3-NEXT: .Ltmp9:
+; SSE3-NEXT: .cfi_offset %r14, -32
+; SSE3-NEXT: .Ltmp10:
+; SSE3-NEXT: .cfi_offset %r15, -24
+; SSE3-NEXT: .Ltmp11:
+; SSE3-NEXT: .cfi_offset %rbp, -16
+; SSE3-NEXT: movd %xmm0, %eax
+; SSE3-NEXT: pextrw $1, %xmm0, %ecx
+; SSE3-NEXT: addl %eax, %ecx
+; SSE3-NEXT: movl %ecx, -{{[0-9]+}}(%rsp) # 4-byte Spill
+; SSE3-NEXT: pextrw $2, %xmm0, %eax
+; SSE3-NEXT: pextrw $3, %xmm0, %r11d
+; SSE3-NEXT: addl %eax, %r11d
+; SSE3-NEXT: pextrw $4, %xmm0, %eax
+; SSE3-NEXT: pextrw $5, %xmm0, %r10d
+; SSE3-NEXT: addl %eax, %r10d
+; SSE3-NEXT: pextrw $6, %xmm0, %eax
+; SSE3-NEXT: pextrw $7, %xmm0, %r13d
+; SSE3-NEXT: addl %eax, %r13d
+; SSE3-NEXT: movd %xmm1, %eax
+; SSE3-NEXT: pextrw $1, %xmm1, %r14d
+; SSE3-NEXT: addl %eax, %r14d
+; SSE3-NEXT: pextrw $2, %xmm1, %eax
+; SSE3-NEXT: pextrw $3, %xmm1, %ebp
+; SSE3-NEXT: addl %eax, %ebp
+; SSE3-NEXT: pextrw $4, %xmm1, %eax
+; SSE3-NEXT: pextrw $5, %xmm1, %ebx
+; SSE3-NEXT: addl %eax, %ebx
+; SSE3-NEXT: pextrw $6, %xmm1, %eax
+; SSE3-NEXT: pextrw $7, %xmm1, %edx
+; SSE3-NEXT: addl %eax, %edx
+; SSE3-NEXT: movd %xmm2, %eax
+; SSE3-NEXT: pextrw $1, %xmm2, %ecx
+; SSE3-NEXT: addl %eax, %ecx
+; SSE3-NEXT: movl %ecx, -{{[0-9]+}}(%rsp) # 4-byte Spill
+; SSE3-NEXT: pextrw $2, %xmm2, %eax
+; SSE3-NEXT: pextrw $3, %xmm2, %r12d
+; SSE3-NEXT: addl %eax, %r12d
+; SSE3-NEXT: pextrw $4, %xmm2, %eax
+; SSE3-NEXT: pextrw $5, %xmm2, %r15d
+; SSE3-NEXT: addl %eax, %r15d
+; SSE3-NEXT: pextrw $6, %xmm2, %eax
+; SSE3-NEXT: pextrw $7, %xmm2, %r8d
+; SSE3-NEXT: addl %eax, %r8d
+; SSE3-NEXT: movd %xmm3, %eax
+; SSE3-NEXT: pextrw $1, %xmm3, %r9d
+; SSE3-NEXT: addl %eax, %r9d
+; SSE3-NEXT: pextrw $2, %xmm3, %eax
+; SSE3-NEXT: pextrw $3, %xmm3, %esi
+; SSE3-NEXT: addl %eax, %esi
+; SSE3-NEXT: pextrw $4, %xmm3, %eax
+; SSE3-NEXT: pextrw $5, %xmm3, %edi
+; SSE3-NEXT: addl %eax, %edi
+; SSE3-NEXT: pextrw $6, %xmm3, %ecx
+; SSE3-NEXT: pextrw $7, %xmm3, %eax
+; SSE3-NEXT: addl %ecx, %eax
+; SSE3-NEXT: movd %edx, %xmm8
+; SSE3-NEXT: movd %r13d, %xmm3
+; SSE3-NEXT: movd %ebp, %xmm9
+; SSE3-NEXT: movd %r11d, %xmm4
+; SSE3-NEXT: movd %ebx, %xmm10
+; SSE3-NEXT: movd %r10d, %xmm7
+; SSE3-NEXT: movd %r14d, %xmm11
+; SSE3-NEXT: movd -{{[0-9]+}}(%rsp), %xmm0 # 4-byte Folded Reload
+; SSE3-NEXT: # xmm0 = mem[0],zero,zero,zero
+; SSE3-NEXT: movd %eax, %xmm12
+; SSE3-NEXT: movd %r8d, %xmm6
+; SSE3-NEXT: movd %esi, %xmm13
+; SSE3-NEXT: movd %r12d, %xmm5
+; SSE3-NEXT: movd %edi, %xmm14
+; SSE3-NEXT: movd %r15d, %xmm2
+; SSE3-NEXT: movd %r9d, %xmm15
+; SSE3-NEXT: movd -{{[0-9]+}}(%rsp), %xmm1 # 4-byte Folded Reload
+; SSE3-NEXT: # xmm1 = mem[0],zero,zero,zero
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm10[0],xmm7[1],xmm10[1],xmm7[2],xmm10[2],xmm7[3],xmm10[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1],xmm0[2],xmm11[2],xmm0[3],xmm11[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1],xmm0[2],xmm7[2],xmm0[3],xmm7[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm12[0],xmm6[1],xmm12[1],xmm6[2],xmm12[2],xmm6[3],xmm12[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm13[0],xmm5[1],xmm13[1],xmm5[2],xmm13[2],xmm5[3],xmm13[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm14[0],xmm2[1],xmm14[1],xmm2[2],xmm14[2],xmm2[3],xmm14[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1],xmm1[2],xmm15[2],xmm1[3],xmm15[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; SSE3-NEXT: popq %rbx
+; SSE3-NEXT: popq %r12
+; SSE3-NEXT: popq %r13
+; SSE3-NEXT: popq %r14
+; SSE3-NEXT: popq %r15
+; SSE3-NEXT: popq %rbp
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: avx2_vphadd_w_test:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: phaddw %xmm1, %xmm0
+; SSSE3-NEXT: phaddw %xmm3, %xmm2
+; SSSE3-NEXT: movdqa %xmm2, %xmm1
+; SSSE3-NEXT: retq
+;
+; AVX1-LABEL: avx2_vphadd_w_test:
+; AVX1: # BB#0:
+; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT: vphaddw %xmm2, %xmm1, %xmm1
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
+; AVX1-NEXT: vphaddw %xmm2, %xmm0, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: avx2_vphadd_w_test:
+; AVX2: # BB#0:
+; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2
+; AVX2-NEXT: vphaddw %xmm2, %xmm1, %xmm1
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm2
+; AVX2-NEXT: vphaddw %xmm2, %xmm0, %xmm0
+; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT: retq
%vecext = extractelement <16 x i16> %a, i32 0
%vecext1 = extractelement <16 x i16> %a, i32 1
%add = add i16 %vecext, %vecext1
@@ -447,20 +833,58 @@ define <16 x i16> @avx2_vphadd_w_test(<16 x i16> %a, <16 x i16> %b) {
%vecinit108 = insertelement <16 x i16> %vecinit101, i16 %add106, i32 15
ret <16 x i16> %vecinit108
}
-; CHECK-LABEL: avx2_vphadd_w_test
-; SSE3-NOT: phaddw
-; SSSE3: phaddw
-; SSSE3-NEXT: phaddw
-; AVX: vphaddw
-; AVX: vphaddw
-; AVX2: vphaddw
-; AVX2: vphaddw
-; CHECK: ret
-
; Verify that we don't select horizontal subs in the following functions.
define <4 x i32> @not_a_hsub_1(<4 x i32> %A, <4 x i32> %B) {
+; SSE-LABEL: not_a_hsub_1:
+; SSE: # BB#0:
+; SSE-NEXT: movd %xmm0, %eax
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,2,3]
+; SSE-NEXT: movd %xmm2, %ecx
+; SSE-NEXT: subl %ecx, %eax
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[2,3,0,1]
+; SSE-NEXT: movd %xmm2, %ecx
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
+; SSE-NEXT: movd %xmm0, %edx
+; SSE-NEXT: subl %edx, %ecx
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
+; SSE-NEXT: movd %xmm0, %edx
+; SSE-NEXT: movd %xmm1, %esi
+; SSE-NEXT: subl %esi, %edx
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[3,1,2,3]
+; SSE-NEXT: movd %xmm0, %esi
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
+; SSE-NEXT: movd %xmm0, %edi
+; SSE-NEXT: subl %edi, %esi
+; SSE-NEXT: movd %esi, %xmm0
+; SSE-NEXT: movd %ecx, %xmm1
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: movd %edx, %xmm2
+; SSE-NEXT: movd %eax, %xmm0
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: retq
+;
+; AVX-LABEL: not_a_hsub_1:
+; AVX: # BB#0:
+; AVX-NEXT: vmovd %xmm0, %eax
+; AVX-NEXT: vpextrd $1, %xmm0, %ecx
+; AVX-NEXT: subl %ecx, %eax
+; AVX-NEXT: vpextrd $2, %xmm0, %ecx
+; AVX-NEXT: vpextrd $3, %xmm0, %edx
+; AVX-NEXT: subl %edx, %ecx
+; AVX-NEXT: vpextrd $1, %xmm1, %edx
+; AVX-NEXT: vmovd %xmm1, %esi
+; AVX-NEXT: subl %esi, %edx
+; AVX-NEXT: vpextrd $3, %xmm1, %esi
+; AVX-NEXT: vpextrd $2, %xmm1, %edi
+; AVX-NEXT: subl %edi, %esi
+; AVX-NEXT: vmovd %eax, %xmm0
+; AVX-NEXT: vpinsrd $1, %ecx, %xmm0, %xmm0
+; AVX-NEXT: vpinsrd $2, %edx, %xmm0, %xmm0
+; AVX-NEXT: vpinsrd $3, %esi, %xmm0, %xmm0
+; AVX-NEXT: retq
%vecext = extractelement <4 x i32> %A, i32 0
%vecext1 = extractelement <4 x i32> %A, i32 1
%sub = sub i32 %vecext, %vecext1
@@ -479,12 +903,45 @@ define <4 x i32> @not_a_hsub_1(<4 x i32> %A, <4 x i32> %B) {
%vecinit13 = insertelement <4 x i32> %vecinit9, i32 %sub12, i32 3
ret <4 x i32> %vecinit13
}
-; CHECK-LABEL: not_a_hsub_1
-; CHECK-NOT: phsubd
-; CHECK: ret
-
define <4 x float> @not_a_hsub_2(<4 x float> %A, <4 x float> %B) {
+; SSE-LABEL: not_a_hsub_2:
+; SSE: # BB#0:
+; SSE-NEXT: movapd %xmm0, %xmm2
+; SSE-NEXT: shufpd {{.*#+}} xmm2 = xmm2[1,0]
+; SSE-NEXT: movapd %xmm0, %xmm3
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[3,1,2,3]
+; SSE-NEXT: subss %xmm3, %xmm2
+; SSE-NEXT: movshdup {{.*#+}} xmm3 = xmm0[1,1,3,3]
+; SSE-NEXT: subss %xmm3, %xmm0
+; SSE-NEXT: movaps %xmm1, %xmm3
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[3,1,2,3]
+; SSE-NEXT: movaps %xmm1, %xmm4
+; SSE-NEXT: shufpd {{.*#+}} xmm4 = xmm4[1,0]
+; SSE-NEXT: subss %xmm4, %xmm3
+; SSE-NEXT: unpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; SSE-NEXT: movshdup {{.*#+}} xmm3 = xmm1[1,1,3,3]
+; SSE-NEXT: subss %xmm3, %xmm1
+; SSE-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT: retq
+;
+; AVX-LABEL: not_a_hsub_2:
+; AVX: # BB#0:
+; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-NEXT: vpermilps {{.*#+}} xmm3 = xmm0[3,1,2,3]
+; AVX-NEXT: vsubss %xmm3, %xmm2, %xmm2
+; AVX-NEXT: vmovshdup {{.*#+}} xmm3 = xmm0[1,1,3,3]
+; AVX-NEXT: vsubss %xmm3, %xmm0, %xmm0
+; AVX-NEXT: vpermilps {{.*#+}} xmm3 = xmm1[3,1,2,3]
+; AVX-NEXT: vpermilpd {{.*#+}} xmm4 = xmm1[1,0]
+; AVX-NEXT: vsubss %xmm4, %xmm3, %xmm3
+; AVX-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[2,3]
+; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
+; AVX-NEXT: vsubss %xmm2, %xmm1, %xmm1
+; AVX-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0],xmm0[3]
+; AVX-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],xmm3[0]
+; AVX-NEXT: retq
%vecext = extractelement <4 x float> %A, i32 2
%vecext1 = extractelement <4 x float> %A, i32 3
%sub = fsub float %vecext, %vecext1
@@ -503,12 +960,28 @@ define <4 x float> @not_a_hsub_2(<4 x float> %A, <4 x float> %B) {
%vecinit13 = insertelement <4 x float> %vecinit9, float %sub12, i32 2
ret <4 x float> %vecinit13
}
-; CHECK-LABEL: not_a_hsub_2
-; CHECK-NOT: hsubps
-; CHECK: ret
-
define <2 x double> @not_a_hsub_3(<2 x double> %A, <2 x double> %B) {
+; SSE-LABEL: not_a_hsub_3:
+; SSE: # BB#0:
+; SSE-NEXT: movapd %xmm1, %xmm2
+; SSE-NEXT: shufpd {{.*#+}} xmm2 = xmm2[1,0]
+; SSE-NEXT: subsd %xmm2, %xmm1
+; SSE-NEXT: movapd %xmm0, %xmm2
+; SSE-NEXT: shufpd {{.*#+}} xmm2 = xmm2[1,0]
+; SSE-NEXT: subsd %xmm0, %xmm2
+; SSE-NEXT: unpcklpd {{.*#+}} xmm2 = xmm2[0],xmm1[0]
+; SSE-NEXT: movapd %xmm2, %xmm0
+; SSE-NEXT: retq
+;
+; AVX-LABEL: not_a_hsub_3:
+; AVX: # BB#0:
+; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vsubsd %xmm2, %xmm1, %xmm1
+; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-NEXT: vsubsd %xmm0, %xmm2, %xmm0
+; AVX-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX-NEXT: retq
%vecext = extractelement <2 x double> %B, i32 0
%vecext1 = extractelement <2 x double> %B, i32 1
%sub = fsub double %vecext, %vecext1
@@ -519,15 +992,21 @@ define <2 x double> @not_a_hsub_3(<2 x double> %A, <2 x double> %B) {
%vecinit2 = insertelement <2 x double> %vecinit, double %sub2, i32 0
ret <2 x double> %vecinit2
}
-; CHECK-LABEL: not_a_hsub_3
-; CHECK-NOT: hsubpd
-; CHECK: ret
-
; Test AVX horizontal add/sub of packed single/double precision
; floating point values from 256-bit vectors.
define <8 x float> @avx_vhadd_ps(<8 x float> %a, <8 x float> %b) {
+; SSE-LABEL: avx_vhadd_ps:
+; SSE: # BB#0:
+; SSE-NEXT: haddps %xmm2, %xmm0
+; SSE-NEXT: haddps %xmm3, %xmm1
+; SSE-NEXT: retq
+;
+; AVX-LABEL: avx_vhadd_ps:
+; AVX: # BB#0:
+; AVX-NEXT: vhaddps %ymm1, %ymm0, %ymm0
+; AVX-NEXT: retq
%vecext = extractelement <8 x float> %a, i32 0
%vecext1 = extractelement <8 x float> %a, i32 1
%add = fadd float %vecext, %vecext1
@@ -562,17 +1041,18 @@ define <8 x float> @avx_vhadd_ps(<8 x float> %a, <8 x float> %b) {
%vecinit29 = insertelement <8 x float> %vecinit25, float %add28, i32 7
ret <8 x float> %vecinit29
}
-; CHECK-LABEL: avx_vhadd_ps
-; SSE3: haddps
-; SSE3-NEXT: haddps
-; SSSE3: haddps
-; SSSE3-NEXT: haddps
-; AVX: vhaddps
-; AVX2: vhaddps
-; CHECK: ret
-
define <8 x float> @avx_vhsub_ps(<8 x float> %a, <8 x float> %b) {
+; SSE-LABEL: avx_vhsub_ps:
+; SSE: # BB#0:
+; SSE-NEXT: hsubps %xmm2, %xmm0
+; SSE-NEXT: hsubps %xmm3, %xmm1
+; SSE-NEXT: retq
+;
+; AVX-LABEL: avx_vhsub_ps:
+; AVX: # BB#0:
+; AVX-NEXT: vhsubps %ymm1, %ymm0, %ymm0
+; AVX-NEXT: retq
%vecext = extractelement <8 x float> %a, i32 0
%vecext1 = extractelement <8 x float> %a, i32 1
%sub = fsub float %vecext, %vecext1
@@ -607,17 +1087,18 @@ define <8 x float> @avx_vhsub_ps(<8 x float> %a, <8 x float> %b) {
%vecinit29 = insertelement <8 x float> %vecinit25, float %sub28, i32 7
ret <8 x float> %vecinit29
}
-; CHECK-LABEL: avx_vhsub_ps
-; SSE3: hsubps
-; SSE3-NEXT: hsubps
-; SSSE3: hsubps
-; SSSE3-NEXT: hsubps
-; AVX: vhsubps
-; AVX2: vhsubps
-; CHECK: ret
-
define <4 x double> @avx_hadd_pd(<4 x double> %a, <4 x double> %b) {
+; SSE-LABEL: avx_hadd_pd:
+; SSE: # BB#0:
+; SSE-NEXT: haddpd %xmm2, %xmm0
+; SSE-NEXT: haddpd %xmm3, %xmm1
+; SSE-NEXT: retq
+;
+; AVX-LABEL: avx_hadd_pd:
+; AVX: # BB#0:
+; AVX-NEXT: vhaddpd %ymm1, %ymm0, %ymm0
+; AVX-NEXT: retq
%vecext = extractelement <4 x double> %a, i32 0
%vecext1 = extractelement <4 x double> %a, i32 1
%add = fadd double %vecext, %vecext1
@@ -636,17 +1117,18 @@ define <4 x double> @avx_hadd_pd(<4 x double> %a, <4 x double> %b) {
%vecinit13 = insertelement <4 x double> %vecinit9, double %add12, i32 3
ret <4 x double> %vecinit13
}
-; CHECK-LABEL: avx_hadd_pd
-; SSE3: haddpd
-; SSE3-NEXT: haddpd
-; SSSE3: haddpd
-; SSSE3-NEXT: haddpd
-; AVX: vhaddpd
-; AVX2: vhaddpd
-; CHECK: ret
-
define <4 x double> @avx_hsub_pd(<4 x double> %a, <4 x double> %b) {
+; SSE-LABEL: avx_hsub_pd:
+; SSE: # BB#0:
+; SSE-NEXT: hsubpd %xmm2, %xmm0
+; SSE-NEXT: hsubpd %xmm3, %xmm1
+; SSE-NEXT: retq
+;
+; AVX-LABEL: avx_hsub_pd:
+; AVX: # BB#0:
+; AVX-NEXT: vhsubpd %ymm1, %ymm0, %ymm0
+; AVX-NEXT: retq
%vecext = extractelement <4 x double> %a, i32 0
%vecext1 = extractelement <4 x double> %a, i32 1
%sub = fsub double %vecext, %vecext1
@@ -665,19 +1147,83 @@ define <4 x double> @avx_hsub_pd(<4 x double> %a, <4 x double> %b) {
%vecinit13 = insertelement <4 x double> %vecinit9, double %sub12, i32 3
ret <4 x double> %vecinit13
}
-; CHECK-LABEL: avx_hsub_pd
-; SSE3: hsubpd
-; SSE3-NEXT: hsubpd
-; SSSE3: hsubpd
-; SSSE3-NEXT: hsubpd
-; AVX: vhsubpd
-; AVX2: vhsubpd
-; CHECK: ret
-
; Test AVX2 horizontal add of packed integer values from 256-bit vectors.
define <8 x i32> @avx2_hadd_d(<8 x i32> %a, <8 x i32> %b) {
+; SSE3-LABEL: avx2_hadd_d:
+; SSE3: # BB#0:
+; SSE3-NEXT: movd %xmm0, %ecx
+; SSE3-NEXT: pshufd {{.*#+}} xmm4 = xmm0[1,1,2,3]
+; SSE3-NEXT: movd %xmm4, %r8d
+; SSE3-NEXT: addl %ecx, %r8d
+; SSE3-NEXT: pshufd {{.*#+}} xmm4 = xmm0[2,3,0,1]
+; SSE3-NEXT: movd %xmm4, %edx
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
+; SSE3-NEXT: movd %xmm0, %r9d
+; SSE3-NEXT: addl %edx, %r9d
+; SSE3-NEXT: movd %xmm2, %esi
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm2[1,1,2,3]
+; SSE3-NEXT: movd %xmm0, %r10d
+; SSE3-NEXT: addl %esi, %r10d
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,3,0,1]
+; SSE3-NEXT: movd %xmm0, %esi
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm2[3,1,2,3]
+; SSE3-NEXT: movd %xmm0, %edi
+; SSE3-NEXT: addl %esi, %edi
+; SSE3-NEXT: movd %xmm1, %eax
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
+; SSE3-NEXT: movd %xmm0, %r11d
+; SSE3-NEXT: addl %eax, %r11d
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
+; SSE3-NEXT: movd %xmm0, %eax
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm1[3,1,2,3]
+; SSE3-NEXT: movd %xmm0, %ecx
+; SSE3-NEXT: addl %eax, %ecx
+; SSE3-NEXT: movd %xmm3, %eax
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,1,2,3]
+; SSE3-NEXT: movd %xmm0, %edx
+; SSE3-NEXT: addl %eax, %edx
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm3[2,3,0,1]
+; SSE3-NEXT: movd %xmm0, %eax
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm3[3,1,2,3]
+; SSE3-NEXT: movd %xmm0, %esi
+; SSE3-NEXT: addl %eax, %esi
+; SSE3-NEXT: movd %edi, %xmm0
+; SSE3-NEXT: movd %r9d, %xmm1
+; SSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT: movd %r10d, %xmm2
+; SSE3-NEXT: movd %r8d, %xmm0
+; SSE3-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE3-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE3-NEXT: movd %esi, %xmm1
+; SSE3-NEXT: movd %ecx, %xmm2
+; SSE3-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE3-NEXT: movd %edx, %xmm3
+; SSE3-NEXT: movd %r11d, %xmm1
+; SSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; SSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: avx2_hadd_d:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: phaddd %xmm2, %xmm0
+; SSSE3-NEXT: phaddd %xmm3, %xmm1
+; SSSE3-NEXT: retq
+;
+; AVX1-LABEL: avx2_hadd_d:
+; AVX1: # BB#0:
+; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
+; AVX1-NEXT: vphaddd %xmm2, %xmm3, %xmm2
+; AVX1-NEXT: vphaddd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: avx2_hadd_d:
+; AVX2: # BB#0:
+; AVX2-NEXT: vphaddd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT: retq
%vecext = extractelement <8 x i32> %a, i32 0
%vecext1 = extractelement <8 x i32> %a, i32 1
%add = add i32 %vecext, %vecext1
@@ -712,18 +1258,149 @@ define <8 x i32> @avx2_hadd_d(<8 x i32> %a, <8 x i32> %b) {
%vecinit29 = insertelement <8 x i32> %vecinit25, i32 %add28, i32 7
ret <8 x i32> %vecinit29
}
-; CHECK-LABEL: avx2_hadd_d
-; SSE3-NOT: phaddd
-; SSSE3: phaddd
-; SSSE3-NEXT: phaddd
-; AVX: vphaddd
-; AVX: vphaddd
-; AVX2: vphaddd
-; AVX2-NOT: vphaddd
-; CHECK: ret
-
define <16 x i16> @avx2_hadd_w(<16 x i16> %a, <16 x i16> %b) {
+; SSE3-LABEL: avx2_hadd_w:
+; SSE3: # BB#0:
+; SSE3-NEXT: pushq %rbp
+; SSE3-NEXT: .Ltmp12:
+; SSE3-NEXT: .cfi_def_cfa_offset 16
+; SSE3-NEXT: pushq %r15
+; SSE3-NEXT: .Ltmp13:
+; SSE3-NEXT: .cfi_def_cfa_offset 24
+; SSE3-NEXT: pushq %r14
+; SSE3-NEXT: .Ltmp14:
+; SSE3-NEXT: .cfi_def_cfa_offset 32
+; SSE3-NEXT: pushq %r13
+; SSE3-NEXT: .Ltmp15:
+; SSE3-NEXT: .cfi_def_cfa_offset 40
+; SSE3-NEXT: pushq %r12
+; SSE3-NEXT: .Ltmp16:
+; SSE3-NEXT: .cfi_def_cfa_offset 48
+; SSE3-NEXT: pushq %rbx
+; SSE3-NEXT: .Ltmp17:
+; SSE3-NEXT: .cfi_def_cfa_offset 56
+; SSE3-NEXT: .Ltmp18:
+; SSE3-NEXT: .cfi_offset %rbx, -56
+; SSE3-NEXT: .Ltmp19:
+; SSE3-NEXT: .cfi_offset %r12, -48
+; SSE3-NEXT: .Ltmp20:
+; SSE3-NEXT: .cfi_offset %r13, -40
+; SSE3-NEXT: .Ltmp21:
+; SSE3-NEXT: .cfi_offset %r14, -32
+; SSE3-NEXT: .Ltmp22:
+; SSE3-NEXT: .cfi_offset %r15, -24
+; SSE3-NEXT: .Ltmp23:
+; SSE3-NEXT: .cfi_offset %rbp, -16
+; SSE3-NEXT: movd %xmm0, %eax
+; SSE3-NEXT: pextrw $1, %xmm0, %ecx
+; SSE3-NEXT: addl %eax, %ecx
+; SSE3-NEXT: movl %ecx, -{{[0-9]+}}(%rsp) # 4-byte Spill
+; SSE3-NEXT: pextrw $2, %xmm0, %eax
+; SSE3-NEXT: pextrw $3, %xmm0, %r15d
+; SSE3-NEXT: addl %eax, %r15d
+; SSE3-NEXT: pextrw $4, %xmm0, %eax
+; SSE3-NEXT: pextrw $5, %xmm0, %r14d
+; SSE3-NEXT: addl %eax, %r14d
+; SSE3-NEXT: pextrw $6, %xmm0, %eax
+; SSE3-NEXT: pextrw $7, %xmm0, %r13d
+; SSE3-NEXT: addl %eax, %r13d
+; SSE3-NEXT: movd %xmm1, %eax
+; SSE3-NEXT: pextrw $1, %xmm1, %ecx
+; SSE3-NEXT: addl %eax, %ecx
+; SSE3-NEXT: movl %ecx, -{{[0-9]+}}(%rsp) # 4-byte Spill
+; SSE3-NEXT: pextrw $2, %xmm1, %eax
+; SSE3-NEXT: pextrw $3, %xmm1, %r11d
+; SSE3-NEXT: addl %eax, %r11d
+; SSE3-NEXT: pextrw $4, %xmm1, %eax
+; SSE3-NEXT: pextrw $5, %xmm1, %r10d
+; SSE3-NEXT: addl %eax, %r10d
+; SSE3-NEXT: pextrw $6, %xmm1, %eax
+; SSE3-NEXT: pextrw $7, %xmm1, %r12d
+; SSE3-NEXT: addl %eax, %r12d
+; SSE3-NEXT: movd %xmm2, %eax
+; SSE3-NEXT: pextrw $1, %xmm2, %ebx
+; SSE3-NEXT: addl %eax, %ebx
+; SSE3-NEXT: pextrw $2, %xmm2, %eax
+; SSE3-NEXT: pextrw $3, %xmm2, %ecx
+; SSE3-NEXT: addl %eax, %ecx
+; SSE3-NEXT: pextrw $4, %xmm2, %esi
+; SSE3-NEXT: pextrw $5, %xmm2, %r8d
+; SSE3-NEXT: addl %esi, %r8d
+; SSE3-NEXT: pextrw $6, %xmm2, %esi
+; SSE3-NEXT: pextrw $7, %xmm2, %edx
+; SSE3-NEXT: addl %esi, %edx
+; SSE3-NEXT: movd %xmm3, %edi
+; SSE3-NEXT: pextrw $1, %xmm3, %r9d
+; SSE3-NEXT: addl %edi, %r9d
+; SSE3-NEXT: pextrw $2, %xmm3, %ebp
+; SSE3-NEXT: pextrw $3, %xmm3, %edi
+; SSE3-NEXT: addl %ebp, %edi
+; SSE3-NEXT: pextrw $4, %xmm3, %eax
+; SSE3-NEXT: pextrw $5, %xmm3, %ebp
+; SSE3-NEXT: addl %eax, %ebp
+; SSE3-NEXT: pextrw $6, %xmm3, %esi
+; SSE3-NEXT: pextrw $7, %xmm3, %eax
+; SSE3-NEXT: addl %esi, %eax
+; SSE3-NEXT: movd %edx, %xmm8
+; SSE3-NEXT: movd %r13d, %xmm3
+; SSE3-NEXT: movd %ecx, %xmm9
+; SSE3-NEXT: movd %r15d, %xmm4
+; SSE3-NEXT: movd %r8d, %xmm10
+; SSE3-NEXT: movd %r14d, %xmm7
+; SSE3-NEXT: movd %ebx, %xmm11
+; SSE3-NEXT: movd -{{[0-9]+}}(%rsp), %xmm0 # 4-byte Folded Reload
+; SSE3-NEXT: # xmm0 = mem[0],zero,zero,zero
+; SSE3-NEXT: movd %eax, %xmm12
+; SSE3-NEXT: movd %r12d, %xmm6
+; SSE3-NEXT: movd %edi, %xmm13
+; SSE3-NEXT: movd %r11d, %xmm5
+; SSE3-NEXT: movd %ebp, %xmm14
+; SSE3-NEXT: movd %r10d, %xmm2
+; SSE3-NEXT: movd %r9d, %xmm15
+; SSE3-NEXT: movd -{{[0-9]+}}(%rsp), %xmm1 # 4-byte Folded Reload
+; SSE3-NEXT: # xmm1 = mem[0],zero,zero,zero
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm10[0],xmm7[1],xmm10[1],xmm7[2],xmm10[2],xmm7[3],xmm10[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1],xmm0[2],xmm11[2],xmm0[3],xmm11[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1],xmm0[2],xmm7[2],xmm0[3],xmm7[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm12[0],xmm6[1],xmm12[1],xmm6[2],xmm12[2],xmm6[3],xmm12[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm13[0],xmm5[1],xmm13[1],xmm5[2],xmm13[2],xmm5[3],xmm13[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm14[0],xmm2[1],xmm14[1],xmm2[2],xmm14[2],xmm2[3],xmm14[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1],xmm1[2],xmm15[2],xmm1[3],xmm15[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; SSE3-NEXT: popq %rbx
+; SSE3-NEXT: popq %r12
+; SSE3-NEXT: popq %r13
+; SSE3-NEXT: popq %r14
+; SSE3-NEXT: popq %r15
+; SSE3-NEXT: popq %rbp
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: avx2_hadd_w:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: phaddw %xmm2, %xmm0
+; SSSE3-NEXT: phaddw %xmm3, %xmm1
+; SSSE3-NEXT: retq
+;
+; AVX1-LABEL: avx2_hadd_w:
+; AVX1: # BB#0:
+; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
+; AVX1-NEXT: vphaddw %xmm2, %xmm3, %xmm2
+; AVX1-NEXT: vphaddw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: avx2_hadd_w:
+; AVX2: # BB#0:
+; AVX2-NEXT: vphaddw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT: retq
%vecext = extractelement <16 x i16> %a, i32 0
%vecext1 = extractelement <16 x i16> %a, i32 1
%add = add i16 %vecext, %vecext1
@@ -790,13 +1467,3 @@ define <16 x i16> @avx2_hadd_w(<16 x i16> %a, <16 x i16> %b) {
%vecinit108 = insertelement <16 x i16> %vecinit101, i16 %add106, i32 15
ret <16 x i16> %vecinit108
}
-; CHECK-LABEL: avx2_hadd_w
-; SSE3-NOT: phaddw
-; SSSE3: phaddw
-; SSSE3-NEXT: phaddw
-; AVX: vphaddw
-; AVX: vphaddw
-; AVX2: vphaddw
-; AVX2-NOT: vphaddw
-; CHECK: ret
-
generated by cgit v1.2.3 (git 2.25.1) at 2025-10-31 19:58:35 +0000