summaryrefslogtreecommitdiffstats
path: root/llvm/test/CodeGen/X86/avx512-insert-extract.ll
diff options
context:
space:
mode:
Diffstat (limited to 'llvm/test/CodeGen/X86/avx512-insert-extract.ll')
-rw-r--r--llvm/test/CodeGen/X86/avx512-insert-extract.ll441
1 files changed, 164 insertions, 277 deletions
diff --git a/llvm/test/CodeGen/X86/avx512-insert-extract.ll b/llvm/test/CodeGen/X86/avx512-insert-extract.ll
index 2bfc8b63816..4b9eb5d76b3 100644
--- a/llvm/test/CodeGen/X86/avx512-insert-extract.ll
+++ b/llvm/test/CodeGen/X86/avx512-insert-extract.ll
@@ -1,7 +1,7 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl | FileCheck --check-prefix=CHECK --check-prefix=KNL %s
-; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=skx | FileCheck --check-prefix=CHECK --check-prefix=SKX --check-prefix=SKX_ONLY %s
-; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=skx -mattr=avx512vbmi | FileCheck --check-prefix=CHECK --check-prefix=SKX --check-prefix=SKX_VBMI %s
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mattr=+avx512f | FileCheck --check-prefix=CHECK --check-prefix=KNL %s
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mattr=+avx512f,+avx512bw,+avx512vl,+avx512dq | FileCheck --check-prefix=CHECK --check-prefix=SKX --check-prefix=SKX_ONLY %s
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mattr=+avx512f,+avx512bw,+avx512vl,+avx512dq,+avx512vbmi | FileCheck --check-prefix=CHECK --check-prefix=SKX --check-prefix=SKX_VBMI %s
define <16 x float> @test1(<16 x float> %x, float* %br, float %y) nounwind {
; CHECK-LABEL: test1:
@@ -12,7 +12,6 @@ define <16 x float> @test1(<16 x float> %x, float* %br, float %y) nounwind {
; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0],xmm0[3]
; CHECK-NEXT: vinsertf32x4 $3, %xmm0, %zmm2, %zmm0
; CHECK-NEXT: retq
-; CHECK-NEXT: ## -- End function
%rrr = load float, float* %br
%rrr2 = insertelement <16 x float> %x, float %rrr, i32 1
%rrr3 = insertelement <16 x float> %rrr2, float %y, i32 14
@@ -28,7 +27,6 @@ define <8 x double> @test2(<8 x double> %x, double* %br, double %y) nounwind {
; CHECK-NEXT: vmovsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
; CHECK-NEXT: vinsertf32x4 $3, %xmm0, %zmm2, %zmm0
; CHECK-NEXT: retq
-; CHECK-NEXT: ## -- End function
%rrr = load double, double* %br
%rrr2 = insertelement <8 x double> %x, double %rrr, i32 1
%rrr3 = insertelement <8 x double> %rrr2, double %y, i32 6
@@ -42,7 +40,6 @@ define <16 x float> @test3(<16 x float> %x) nounwind {
; CHECK-NEXT: vinsertps {{.*#+}} xmm1 = xmm0[0],xmm1[0],xmm0[2,3]
; CHECK-NEXT: vinsertf32x4 $0, %xmm1, %zmm0, %zmm0
; CHECK-NEXT: retq
-; CHECK-NEXT: ## -- End function
%eee = extractelement <16 x float> %x, i32 4
%rrr2 = insertelement <16 x float> %x, float %eee, i32 1
ret <16 x float> %rrr2
@@ -56,7 +53,6 @@ define <8 x i64> @test4(<8 x i64> %x) nounwind {
; CHECK-NEXT: vpinsrq $1, %rax, %xmm0, %xmm1
; CHECK-NEXT: vinserti32x4 $0, %xmm1, %zmm0, %zmm0
; CHECK-NEXT: retq
-; CHECK-NEXT: ## -- End function
%eee = extractelement <8 x i64> %x, i32 4
%rrr2 = insertelement <8 x i64> %x, i64 %eee, i32 1
ret <8 x i64> %rrr2
@@ -67,7 +63,6 @@ define i32 @test5(<4 x float> %x) nounwind {
; CHECK: ## BB#0:
; CHECK-NEXT: vextractps $3, %xmm0, %eax
; CHECK-NEXT: retq
-; CHECK-NEXT: ## -- End function
%ef = extractelement <4 x float> %x, i32 3
%ei = bitcast float %ef to i32
ret i32 %ei
@@ -78,148 +73,83 @@ define void @test6(<4 x float> %x, float* %out) nounwind {
; CHECK: ## BB#0:
; CHECK-NEXT: vextractps $3, %xmm0, (%rdi)
; CHECK-NEXT: retq
-; CHECK-NEXT: ## -- End function
%ef = extractelement <4 x float> %x, i32 3
store float %ef, float* %out, align 4
ret void
}
define float @test7(<16 x float> %x, i32 %ind) nounwind {
-; KNL-LABEL: test7:
-; KNL: ## BB#0:
-; KNL-NEXT: pushq %rbp
-; KNL-NEXT: movq %rsp, %rbp
-; KNL-NEXT: andq $-64, %rsp
-; KNL-NEXT: subq $128, %rsp
-; KNL-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def>
-; KNL-NEXT: vmovaps %zmm0, (%rsp)
-; KNL-NEXT: andl $15, %edi
-; KNL-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; KNL-NEXT: movq %rbp, %rsp
-; KNL-NEXT: popq %rbp
-; KNL-NEXT: retq
-; KNL-NEXT: ## -- End function
-;
-; SKX-LABEL: test7:
-; SKX: ## BB#0:
-; SKX-NEXT: pushq %rbp
-; SKX-NEXT: movq %rsp, %rbp
-; SKX-NEXT: andq $-64, %rsp
-; SKX-NEXT: subq $128, %rsp
-; SKX-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def>
-; SKX-NEXT: vmovaps %zmm0, (%rsp)
-; SKX-NEXT: andl $15, %edi
-; SKX-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SKX-NEXT: movq %rbp, %rsp
-; SKX-NEXT: popq %rbp
-; SKX-NEXT: vzeroupper
-; SKX-NEXT: retq
-; SKX-NEXT: ## -- End function
+; CHECK-LABEL: test7:
+; CHECK: ## BB#0:
+; CHECK-NEXT: pushq %rbp
+; CHECK-NEXT: movq %rsp, %rbp
+; CHECK-NEXT: andq $-64, %rsp
+; CHECK-NEXT: subq $128, %rsp
+; CHECK-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def>
+; CHECK-NEXT: vmovaps %zmm0, (%rsp)
+; CHECK-NEXT: andl $15, %edi
+; CHECK-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-NEXT: movq %rbp, %rsp
+; CHECK-NEXT: popq %rbp
+; CHECK-NEXT: vzeroupper
+; CHECK-NEXT: retq
%e = extractelement <16 x float> %x, i32 %ind
ret float %e
}
define double @test8(<8 x double> %x, i32 %ind) nounwind {
-; KNL-LABEL: test8:
-; KNL: ## BB#0:
-; KNL-NEXT: pushq %rbp
-; KNL-NEXT: movq %rsp, %rbp
-; KNL-NEXT: andq $-64, %rsp
-; KNL-NEXT: subq $128, %rsp
-; KNL-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def>
-; KNL-NEXT: vmovaps %zmm0, (%rsp)
-; KNL-NEXT: andl $7, %edi
-; KNL-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
-; KNL-NEXT: movq %rbp, %rsp
-; KNL-NEXT: popq %rbp
-; KNL-NEXT: retq
-; KNL-NEXT: ## -- End function
-;
-; SKX-LABEL: test8:
-; SKX: ## BB#0:
-; SKX-NEXT: pushq %rbp
-; SKX-NEXT: movq %rsp, %rbp
-; SKX-NEXT: andq $-64, %rsp
-; SKX-NEXT: subq $128, %rsp
-; SKX-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def>
-; SKX-NEXT: vmovaps %zmm0, (%rsp)
-; SKX-NEXT: andl $7, %edi
-; SKX-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
-; SKX-NEXT: movq %rbp, %rsp
-; SKX-NEXT: popq %rbp
-; SKX-NEXT: vzeroupper
-; SKX-NEXT: retq
-; SKX-NEXT: ## -- End function
+; CHECK-LABEL: test8:
+; CHECK: ## BB#0:
+; CHECK-NEXT: pushq %rbp
+; CHECK-NEXT: movq %rsp, %rbp
+; CHECK-NEXT: andq $-64, %rsp
+; CHECK-NEXT: subq $128, %rsp
+; CHECK-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def>
+; CHECK-NEXT: vmovaps %zmm0, (%rsp)
+; CHECK-NEXT: andl $7, %edi
+; CHECK-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
+; CHECK-NEXT: movq %rbp, %rsp
+; CHECK-NEXT: popq %rbp
+; CHECK-NEXT: vzeroupper
+; CHECK-NEXT: retq
%e = extractelement <8 x double> %x, i32 %ind
ret double %e
}
define float @test9(<8 x float> %x, i32 %ind) nounwind {
-; KNL-LABEL: test9:
-; KNL: ## BB#0:
-; KNL-NEXT: pushq %rbp
-; KNL-NEXT: movq %rsp, %rbp
-; KNL-NEXT: andq $-32, %rsp
-; KNL-NEXT: subq $64, %rsp
-; KNL-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def>
-; KNL-NEXT: vmovaps %ymm0, (%rsp)
-; KNL-NEXT: andl $7, %edi
-; KNL-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; KNL-NEXT: movq %rbp, %rsp
-; KNL-NEXT: popq %rbp
-; KNL-NEXT: retq
-; KNL-NEXT: ## -- End function
-;
-; SKX-LABEL: test9:
-; SKX: ## BB#0:
-; SKX-NEXT: pushq %rbp
-; SKX-NEXT: movq %rsp, %rbp
-; SKX-NEXT: andq $-32, %rsp
-; SKX-NEXT: subq $64, %rsp
-; SKX-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def>
-; SKX-NEXT: vmovaps %ymm0, (%rsp)
-; SKX-NEXT: andl $7, %edi
-; SKX-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SKX-NEXT: movq %rbp, %rsp
-; SKX-NEXT: popq %rbp
-; SKX-NEXT: vzeroupper
-; SKX-NEXT: retq
-; SKX-NEXT: ## -- End function
+; CHECK-LABEL: test9:
+; CHECK: ## BB#0:
+; CHECK-NEXT: pushq %rbp
+; CHECK-NEXT: movq %rsp, %rbp
+; CHECK-NEXT: andq $-32, %rsp
+; CHECK-NEXT: subq $64, %rsp
+; CHECK-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def>
+; CHECK-NEXT: vmovaps %ymm0, (%rsp)
+; CHECK-NEXT: andl $7, %edi
+; CHECK-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-NEXT: movq %rbp, %rsp
+; CHECK-NEXT: popq %rbp
+; CHECK-NEXT: vzeroupper
+; CHECK-NEXT: retq
%e = extractelement <8 x float> %x, i32 %ind
ret float %e
}
define i32 @test10(<16 x i32> %x, i32 %ind) nounwind {
-; KNL-LABEL: test10:
-; KNL: ## BB#0:
-; KNL-NEXT: pushq %rbp
-; KNL-NEXT: movq %rsp, %rbp
-; KNL-NEXT: andq $-64, %rsp
-; KNL-NEXT: subq $128, %rsp
-; KNL-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def>
-; KNL-NEXT: vmovaps %zmm0, (%rsp)
-; KNL-NEXT: andl $15, %edi
-; KNL-NEXT: movl (%rsp,%rdi,4), %eax
-; KNL-NEXT: movq %rbp, %rsp
-; KNL-NEXT: popq %rbp
-; KNL-NEXT: retq
-; KNL-NEXT: ## -- End function
-;
-; SKX-LABEL: test10:
-; SKX: ## BB#0:
-; SKX-NEXT: pushq %rbp
-; SKX-NEXT: movq %rsp, %rbp
-; SKX-NEXT: andq $-64, %rsp
-; SKX-NEXT: subq $128, %rsp
-; SKX-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def>
-; SKX-NEXT: vmovaps %zmm0, (%rsp)
-; SKX-NEXT: andl $15, %edi
-; SKX-NEXT: movl (%rsp,%rdi,4), %eax
-; SKX-NEXT: movq %rbp, %rsp
-; SKX-NEXT: popq %rbp
-; SKX-NEXT: vzeroupper
-; SKX-NEXT: retq
-; SKX-NEXT: ## -- End function
+; CHECK-LABEL: test10:
+; CHECK: ## BB#0:
+; CHECK-NEXT: pushq %rbp
+; CHECK-NEXT: movq %rsp, %rbp
+; CHECK-NEXT: andq $-64, %rsp
+; CHECK-NEXT: subq $128, %rsp
+; CHECK-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def>
+; CHECK-NEXT: vmovaps %zmm0, (%rsp)
+; CHECK-NEXT: andl $15, %edi
+; CHECK-NEXT: movl (%rsp,%rdi,4), %eax
+; CHECK-NEXT: movq %rbp, %rsp
+; CHECK-NEXT: popq %rbp
+; CHECK-NEXT: vzeroupper
+; CHECK-NEXT: retq
%e = extractelement <16 x i32> %x, i32 %ind
ret i32 %e
}
@@ -274,6 +204,7 @@ define i64 @test12(<16 x i64>%a, <16 x i64>%b, i64 %a1, i64 %b1) {
; KNL-NEXT: testb $1, %al
; KNL-NEXT: cmoveq %rsi, %rdi
; KNL-NEXT: movq %rdi, %rax
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test12:
@@ -339,6 +270,7 @@ define i64 @test14(<8 x i64>%a, <8 x i64>%b, i64 %a1, i64 %b1) {
; KNL-NEXT: testb $1, %al
; KNL-NEXT: cmoveq %rsi, %rdi
; KNL-NEXT: movq %rdi, %rax
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test14:
@@ -387,6 +319,7 @@ define i16 @test16(i1 *%addr, i16 %a) {
; KNL-NEXT: vptestmd %zmm0, %zmm0, %k0
; KNL-NEXT: kmovw %k0, %eax
; KNL-NEXT: ## kill: %AX<def> %AX<kill> %EAX<kill>
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test16:
@@ -424,6 +357,7 @@ define i8 @test17(i1 *%addr, i8 %a) {
; KNL-NEXT: vptestmq %zmm0, %zmm0, %k0
; KNL-NEXT: kmovw %k0, %eax
; KNL-NEXT: ## kill: %AL<def> %AL<kill> %EAX<kill>
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test17:
@@ -448,20 +382,13 @@ define i8 @test17(i1 *%addr, i8 %a) {
}
define i64 @extract_v8i64(<8 x i64> %x, i64* %dst) {
-; KNL-LABEL: extract_v8i64:
-; KNL: ## BB#0:
-; KNL-NEXT: vpextrq $1, %xmm0, %rax
-; KNL-NEXT: vextracti128 $1, %ymm0, %xmm0
-; KNL-NEXT: vpextrq $1, %xmm0, (%rdi)
-; KNL-NEXT: retq
-;
-; SKX-LABEL: extract_v8i64:
-; SKX: ## BB#0:
-; SKX-NEXT: vpextrq $1, %xmm0, %rax
-; SKX-NEXT: vextracti128 $1, %ymm0, %xmm0
-; SKX-NEXT: vpextrq $1, %xmm0, (%rdi)
-; SKX-NEXT: vzeroupper
-; SKX-NEXT: retq
+; CHECK-LABEL: extract_v8i64:
+; CHECK: ## BB#0:
+; CHECK-NEXT: vpextrq $1, %xmm0, %rax
+; CHECK-NEXT: vextracti128 $1, %ymm0, %xmm0
+; CHECK-NEXT: vpextrq $1, %xmm0, (%rdi)
+; CHECK-NEXT: vzeroupper
+; CHECK-NEXT: retq
%r1 = extractelement <8 x i64> %x, i32 1
%r2 = extractelement <8 x i64> %x, i32 3
store i64 %r2, i64* %dst, align 1
@@ -469,20 +396,13 @@ define i64 @extract_v8i64(<8 x i64> %x, i64* %dst) {
}
define i64 @extract_v4i64(<4 x i64> %x, i64* %dst) {
-; KNL-LABEL: extract_v4i64:
-; KNL: ## BB#0:
-; KNL-NEXT: vpextrq $1, %xmm0, %rax
-; KNL-NEXT: vextracti128 $1, %ymm0, %xmm0
-; KNL-NEXT: vpextrq $1, %xmm0, (%rdi)
-; KNL-NEXT: retq
-;
-; SKX-LABEL: extract_v4i64:
-; SKX: ## BB#0:
-; SKX-NEXT: vpextrq $1, %xmm0, %rax
-; SKX-NEXT: vextracti128 $1, %ymm0, %xmm0
-; SKX-NEXT: vpextrq $1, %xmm0, (%rdi)
-; SKX-NEXT: vzeroupper
-; SKX-NEXT: retq
+; CHECK-LABEL: extract_v4i64:
+; CHECK: ## BB#0:
+; CHECK-NEXT: vpextrq $1, %xmm0, %rax
+; CHECK-NEXT: vextracti128 $1, %ymm0, %xmm0
+; CHECK-NEXT: vpextrq $1, %xmm0, (%rdi)
+; CHECK-NEXT: vzeroupper
+; CHECK-NEXT: retq
%r1 = extractelement <4 x i64> %x, i32 1
%r2 = extractelement <4 x i64> %x, i32 3
store i64 %r2, i64* %dst, align 1
@@ -502,20 +422,13 @@ define i64 @extract_v2i64(<2 x i64> %x, i64* %dst) {
}
define i32 @extract_v16i32(<16 x i32> %x, i32* %dst) {
-; KNL-LABEL: extract_v16i32:
-; KNL: ## BB#0:
-; KNL-NEXT: vpextrd $1, %xmm0, %eax
-; KNL-NEXT: vextracti128 $1, %ymm0, %xmm0
-; KNL-NEXT: vpextrd $1, %xmm0, (%rdi)
-; KNL-NEXT: retq
-;
-; SKX-LABEL: extract_v16i32:
-; SKX: ## BB#0:
-; SKX-NEXT: vpextrd $1, %xmm0, %eax
-; SKX-NEXT: vextracti128 $1, %ymm0, %xmm0
-; SKX-NEXT: vpextrd $1, %xmm0, (%rdi)
-; SKX-NEXT: vzeroupper
-; SKX-NEXT: retq
+; CHECK-LABEL: extract_v16i32:
+; CHECK: ## BB#0:
+; CHECK-NEXT: vpextrd $1, %xmm0, %eax
+; CHECK-NEXT: vextracti128 $1, %ymm0, %xmm0
+; CHECK-NEXT: vpextrd $1, %xmm0, (%rdi)
+; CHECK-NEXT: vzeroupper
+; CHECK-NEXT: retq
%r1 = extractelement <16 x i32> %x, i32 1
%r2 = extractelement <16 x i32> %x, i32 5
store i32 %r2, i32* %dst, align 1
@@ -523,20 +436,13 @@ define i32 @extract_v16i32(<16 x i32> %x, i32* %dst) {
}
define i32 @extract_v8i32(<8 x i32> %x, i32* %dst) {
-; KNL-LABEL: extract_v8i32:
-; KNL: ## BB#0:
-; KNL-NEXT: vpextrd $1, %xmm0, %eax
-; KNL-NEXT: vextracti128 $1, %ymm0, %xmm0
-; KNL-NEXT: vpextrd $1, %xmm0, (%rdi)
-; KNL-NEXT: retq
-;
-; SKX-LABEL: extract_v8i32:
-; SKX: ## BB#0:
-; SKX-NEXT: vpextrd $1, %xmm0, %eax
-; SKX-NEXT: vextracti128 $1, %ymm0, %xmm0
-; SKX-NEXT: vpextrd $1, %xmm0, (%rdi)
-; SKX-NEXT: vzeroupper
-; SKX-NEXT: retq
+; CHECK-LABEL: extract_v8i32:
+; CHECK: ## BB#0:
+; CHECK-NEXT: vpextrd $1, %xmm0, %eax
+; CHECK-NEXT: vextracti128 $1, %ymm0, %xmm0
+; CHECK-NEXT: vpextrd $1, %xmm0, (%rdi)
+; CHECK-NEXT: vzeroupper
+; CHECK-NEXT: retq
%r1 = extractelement <8 x i32> %x, i32 1
%r2 = extractelement <8 x i32> %x, i32 5
store i32 %r2, i32* %dst, align 1
@@ -556,22 +462,14 @@ define i32 @extract_v4i32(<4 x i32> %x, i32* %dst) {
}
define i16 @extract_v32i16(<32 x i16> %x, i16* %dst) {
-; KNL-LABEL: extract_v32i16:
-; KNL: ## BB#0:
-; KNL-NEXT: vpextrw $1, %xmm0, %eax
-; KNL-NEXT: vextracti128 $1, %ymm0, %xmm0
-; KNL-NEXT: vpextrw $1, %xmm0, (%rdi)
-; KNL-NEXT: ## kill: %AX<def> %AX<kill> %EAX<kill>
-; KNL-NEXT: retq
-;
-; SKX-LABEL: extract_v32i16:
-; SKX: ## BB#0:
-; SKX-NEXT: vpextrw $1, %xmm0, %eax
-; SKX-NEXT: vextracti128 $1, %ymm0, %xmm0
-; SKX-NEXT: vpextrw $1, %xmm0, (%rdi)
-; SKX-NEXT: ## kill: %AX<def> %AX<kill> %EAX<kill>
-; SKX-NEXT: vzeroupper
-; SKX-NEXT: retq
+; CHECK-LABEL: extract_v32i16:
+; CHECK: ## BB#0:
+; CHECK-NEXT: vpextrw $1, %xmm0, %eax
+; CHECK-NEXT: vextracti128 $1, %ymm0, %xmm0
+; CHECK-NEXT: vpextrw $1, %xmm0, (%rdi)
+; CHECK-NEXT: ## kill: %AX<def> %AX<kill> %EAX<kill>
+; CHECK-NEXT: vzeroupper
+; CHECK-NEXT: retq
%r1 = extractelement <32 x i16> %x, i32 1
%r2 = extractelement <32 x i16> %x, i32 9
store i16 %r2, i16* %dst, align 1
@@ -579,22 +477,14 @@ define i16 @extract_v32i16(<32 x i16> %x, i16* %dst) {
}
define i16 @extract_v16i16(<16 x i16> %x, i16* %dst) {
-; KNL-LABEL: extract_v16i16:
-; KNL: ## BB#0:
-; KNL-NEXT: vpextrw $1, %xmm0, %eax
-; KNL-NEXT: vextracti128 $1, %ymm0, %xmm0
-; KNL-NEXT: vpextrw $1, %xmm0, (%rdi)
-; KNL-NEXT: ## kill: %AX<def> %AX<kill> %EAX<kill>
-; KNL-NEXT: retq
-;
-; SKX-LABEL: extract_v16i16:
-; SKX: ## BB#0:
-; SKX-NEXT: vpextrw $1, %xmm0, %eax
-; SKX-NEXT: vextracti128 $1, %ymm0, %xmm0
-; SKX-NEXT: vpextrw $1, %xmm0, (%rdi)
-; SKX-NEXT: ## kill: %AX<def> %AX<kill> %EAX<kill>
-; SKX-NEXT: vzeroupper
-; SKX-NEXT: retq
+; CHECK-LABEL: extract_v16i16:
+; CHECK: ## BB#0:
+; CHECK-NEXT: vpextrw $1, %xmm0, %eax
+; CHECK-NEXT: vextracti128 $1, %ymm0, %xmm0
+; CHECK-NEXT: vpextrw $1, %xmm0, (%rdi)
+; CHECK-NEXT: ## kill: %AX<def> %AX<kill> %EAX<kill>
+; CHECK-NEXT: vzeroupper
+; CHECK-NEXT: retq
%r1 = extractelement <16 x i16> %x, i32 1
%r2 = extractelement <16 x i16> %x, i32 9
store i16 %r2, i16* %dst, align 1
@@ -615,22 +505,14 @@ define i16 @extract_v8i16(<8 x i16> %x, i16* %dst) {
}
define i8 @extract_v64i8(<64 x i8> %x, i8* %dst) {
-; KNL-LABEL: extract_v64i8:
-; KNL: ## BB#0:
-; KNL-NEXT: vpextrb $1, %xmm0, %eax
-; KNL-NEXT: vextracti128 $1, %ymm0, %xmm0
-; KNL-NEXT: vpextrb $1, %xmm0, (%rdi)
-; KNL-NEXT: ## kill: %AL<def> %AL<kill> %EAX<kill>
-; KNL-NEXT: retq
-;
-; SKX-LABEL: extract_v64i8:
-; SKX: ## BB#0:
-; SKX-NEXT: vpextrb $1, %xmm0, %eax
-; SKX-NEXT: vextracti128 $1, %ymm0, %xmm0
-; SKX-NEXT: vpextrb $1, %xmm0, (%rdi)
-; SKX-NEXT: ## kill: %AL<def> %AL<kill> %EAX<kill>
-; SKX-NEXT: vzeroupper
-; SKX-NEXT: retq
+; CHECK-LABEL: extract_v64i8:
+; CHECK: ## BB#0:
+; CHECK-NEXT: vpextrb $1, %xmm0, %eax
+; CHECK-NEXT: vextracti128 $1, %ymm0, %xmm0
+; CHECK-NEXT: vpextrb $1, %xmm0, (%rdi)
+; CHECK-NEXT: ## kill: %AL<def> %AL<kill> %EAX<kill>
+; CHECK-NEXT: vzeroupper
+; CHECK-NEXT: retq
%r1 = extractelement <64 x i8> %x, i32 1
%r2 = extractelement <64 x i8> %x, i32 17
store i8 %r2, i8* %dst, align 1
@@ -638,22 +520,14 @@ define i8 @extract_v64i8(<64 x i8> %x, i8* %dst) {
}
define i8 @extract_v32i8(<32 x i8> %x, i8* %dst) {
-; KNL-LABEL: extract_v32i8:
-; KNL: ## BB#0:
-; KNL-NEXT: vpextrb $1, %xmm0, %eax
-; KNL-NEXT: vextracti128 $1, %ymm0, %xmm0
-; KNL-NEXT: vpextrb $1, %xmm0, (%rdi)
-; KNL-NEXT: ## kill: %AL<def> %AL<kill> %EAX<kill>
-; KNL-NEXT: retq
-;
-; SKX-LABEL: extract_v32i8:
-; SKX: ## BB#0:
-; SKX-NEXT: vpextrb $1, %xmm0, %eax
-; SKX-NEXT: vextracti128 $1, %ymm0, %xmm0
-; SKX-NEXT: vpextrb $1, %xmm0, (%rdi)
-; SKX-NEXT: ## kill: %AL<def> %AL<kill> %EAX<kill>
-; SKX-NEXT: vzeroupper
-; SKX-NEXT: retq
+; CHECK-LABEL: extract_v32i8:
+; CHECK: ## BB#0:
+; CHECK-NEXT: vpextrb $1, %xmm0, %eax
+; CHECK-NEXT: vextracti128 $1, %ymm0, %xmm0
+; CHECK-NEXT: vpextrb $1, %xmm0, (%rdi)
+; CHECK-NEXT: ## kill: %AL<def> %AL<kill> %EAX<kill>
+; CHECK-NEXT: vzeroupper
+; CHECK-NEXT: retq
%r1 = extractelement <32 x i8> %x, i32 1
%r2 = extractelement <32 x i8> %x, i32 17
store i8 %r2, i8* %dst, align 1
@@ -936,6 +810,7 @@ define i32 @test_insertelement_v32i1(i32 %a, i32 %b, <32 x i32> %x , <32 x i32>
; KNL-NEXT: subq $32, %rsp
; KNL-NEXT: xorl %eax, %eax
; KNL-NEXT: cmpl %esi, %edi
+; KNL-NEXT: setb %al
; KNL-NEXT: vpcmpltud %zmm3, %zmm1, %k0
; KNL-NEXT: kshiftlw $14, %k0, %k1
; KNL-NEXT: kshiftrw $15, %k1, %k1
@@ -1061,7 +936,6 @@ define i32 @test_insertelement_v32i1(i32 %a, i32 %b, <32 x i32> %x , <32 x i32>
; KNL-NEXT: kshiftrw $15, %k1, %k1
; KNL-NEXT: kmovw %k1, %ecx
; KNL-NEXT: vpinsrb $14, %ecx, %xmm0, %xmm0
-; KNL-NEXT: setb %al
; KNL-NEXT: kshiftrw $15, %k0, %k0
; KNL-NEXT: kmovw %k0, %ecx
; KNL-NEXT: vpinsrb $15, %ecx, %xmm0, %xmm0
@@ -1083,6 +957,7 @@ define i32 @test_insertelement_v32i1(i32 %a, i32 %b, <32 x i32> %x , <32 x i32>
; KNL-NEXT: movl (%rsp), %eax
; KNL-NEXT: movq %rbp, %rsp
; KNL-NEXT: popq %rbp
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test_insertelement_v32i1:
@@ -1112,23 +987,23 @@ define i8 @test_iinsertelement_v4i1(i32 %a, i32 %b, <4 x i32> %x , <4 x i32> %y)
; KNL-LABEL: test_iinsertelement_v4i1:
; KNL: ## BB#0:
; KNL-NEXT: cmpl %esi, %edi
+; KNL-NEXT: setb %al
; KNL-NEXT: vpbroadcastd {{.*#+}} xmm2 = [2147483648,2147483648,2147483648,2147483648]
; KNL-NEXT: vpxor %xmm2, %xmm0, %xmm0
; KNL-NEXT: vpxor %xmm2, %xmm1, %xmm1
; KNL-NEXT: vpcmpgtd %xmm0, %xmm1, %xmm0
-; KNL-NEXT: vpextrb $4, %xmm0, %eax
-; KNL-NEXT: setb %cl
-; KNL-NEXT: kmovw %eax, %k1
+; KNL-NEXT: vpextrb $4, %xmm0, %ecx
+; KNL-NEXT: kmovw %ecx, %k1
; KNL-NEXT: vpternlogq $255, %zmm1, %zmm1, %zmm1 {%k1} {z}
-; KNL-NEXT: vpextrb $0, %xmm0, %eax
-; KNL-NEXT: kmovw %eax, %k1
+; KNL-NEXT: vpextrb $0, %xmm0, %ecx
+; KNL-NEXT: kmovw %ecx, %k1
; KNL-NEXT: vpternlogq $255, %zmm2, %zmm2, %zmm2 {%k1} {z}
; KNL-NEXT: vmovdqa64 {{.*#+}} zmm3 = [0,8,2,3,4,5,6,7]
; KNL-NEXT: vpermi2q %zmm1, %zmm2, %zmm3
; KNL-NEXT: vpsllq $63, %zmm3, %zmm1
; KNL-NEXT: vptestmq %zmm1, %zmm1, %k1
; KNL-NEXT: vpternlogq $255, %zmm1, %zmm1, %zmm1 {%k1} {z}
-; KNL-NEXT: kmovw %ecx, %k1
+; KNL-NEXT: kmovw %eax, %k1
; KNL-NEXT: vpternlogq $255, %zmm2, %zmm2, %zmm2 {%k1} {z}
; KNL-NEXT: vmovdqa64 {{.*#+}} zmm3 = [0,1,8,3,4,5,6,7]
; KNL-NEXT: vpermi2q %zmm2, %zmm1, %zmm3
@@ -1144,6 +1019,7 @@ define i8 @test_iinsertelement_v4i1(i32 %a, i32 %b, <4 x i32> %x , <4 x i32> %y)
; KNL-NEXT: vptestmq %zmm0, %zmm0, %k0
; KNL-NEXT: kmovw %k0, %eax
; KNL-NEXT: ## kill: %AL<def> %AL<kill> %EAX<kill>
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test_iinsertelement_v4i1:
@@ -1188,6 +1064,7 @@ define i8 @test_iinsertelement_v2i1(i32 %a, i32 %b, <2 x i64> %x , <2 x i64> %y)
; KNL-NEXT: vptestmq %zmm0, %zmm0, %k0
; KNL-NEXT: kmovw %k0, %eax
; KNL-NEXT: ## kill: %AL<def> %AL<kill> %EAX<kill>
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test_iinsertelement_v2i1:
@@ -1308,6 +1185,7 @@ define zeroext i8 @test_extractelement_v32i1(<32 x i8> %a, <32 x i8> %b) {
; KNL-NEXT: vpcmpgtb %ymm1, %ymm0, %ymm0
; KNL-NEXT: vpextrb $2, %xmm0, %eax
; KNL-NEXT: andl $1, %eax
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test_extractelement_v32i1:
@@ -1338,6 +1216,7 @@ define zeroext i8 @test_extractelement_v64i1(<64 x i8> %a, <64 x i8> %b) {
; KNL-NEXT: movb $4, %cl
; KNL-NEXT: subb %al, %cl
; KNL-NEXT: movzbl %cl, %eax
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test_extractelement_v64i1:
@@ -1370,6 +1249,7 @@ define zeroext i8 @extractelement_v64i1_alt(<64 x i8> %a, <64 x i8> %b) {
; KNL-NEXT: movb $4, %cl
; KNL-NEXT: subb %al, %cl
; KNL-NEXT: movzbl %cl, %eax
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: extractelement_v64i1_alt:
@@ -1421,6 +1301,7 @@ define i64 @test_extractelement_variable_v4i64(<4 x i64> %t1, i32 %index) {
; KNL-NEXT: movq (%rsp,%rdi,8), %rax
; KNL-NEXT: movq %rbp, %rsp
; KNL-NEXT: popq %rbp
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test_extractelement_variable_v4i64:
@@ -1466,6 +1347,7 @@ define i64 @test_extractelement_variable_v8i64(<8 x i64> %t1, i32 %index) {
; KNL-NEXT: movq (%rsp,%rdi,8), %rax
; KNL-NEXT: movq %rbp, %rsp
; KNL-NEXT: popq %rbp
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test_extractelement_variable_v8i64:
@@ -1523,6 +1405,7 @@ define double @test_extractelement_variable_v4f64(<4 x double> %t1, i32 %index)
; KNL-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; KNL-NEXT: movq %rbp, %rsp
; KNL-NEXT: popq %rbp
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test_extractelement_variable_v4f64:
@@ -1568,6 +1451,7 @@ define double @test_extractelement_variable_v8f64(<8 x double> %t1, i32 %index)
; KNL-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; KNL-NEXT: movq %rbp, %rsp
; KNL-NEXT: popq %rbp
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test_extractelement_variable_v8f64:
@@ -1625,6 +1509,7 @@ define i32 @test_extractelement_variable_v8i32(<8 x i32> %t1, i32 %index) {
; KNL-NEXT: movl (%rsp,%rdi,4), %eax
; KNL-NEXT: movq %rbp, %rsp
; KNL-NEXT: popq %rbp
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test_extractelement_variable_v8i32:
@@ -1670,6 +1555,7 @@ define i32 @test_extractelement_variable_v16i32(<16 x i32> %t1, i32 %index) {
; KNL-NEXT: movl (%rsp,%rdi,4), %eax
; KNL-NEXT: movq %rbp, %rsp
; KNL-NEXT: popq %rbp
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test_extractelement_variable_v16i32:
@@ -1727,6 +1613,7 @@ define float @test_extractelement_variable_v8f32(<8 x float> %t1, i32 %index) {
; KNL-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; KNL-NEXT: movq %rbp, %rsp
; KNL-NEXT: popq %rbp
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test_extractelement_variable_v8f32:
@@ -1772,6 +1659,7 @@ define float @test_extractelement_variable_v16f32(<16 x float> %t1, i32 %index)
; KNL-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; KNL-NEXT: movq %rbp, %rsp
; KNL-NEXT: popq %rbp
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test_extractelement_variable_v16f32:
@@ -1829,6 +1717,7 @@ define i16 @test_extractelement_variable_v16i16(<16 x i16> %t1, i32 %index) {
; KNL-NEXT: movzwl (%rsp,%rdi,2), %eax
; KNL-NEXT: movq %rbp, %rsp
; KNL-NEXT: popq %rbp
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test_extractelement_variable_v16i16:
@@ -1875,6 +1764,7 @@ define i16 @test_extractelement_variable_v32i16(<32 x i16> %t1, i32 %index) {
; KNL-NEXT: movzwl (%rsp,%rdi,2), %eax
; KNL-NEXT: movq %rbp, %rsp
; KNL-NEXT: popq %rbp
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test_extractelement_variable_v32i16:
@@ -1902,23 +1792,14 @@ define i16 @test_extractelement_variable_v32i16(<32 x i16> %t1, i32 %index) {
}
define i8 @test_extractelement_variable_v16i8(<16 x i8> %t1, i32 %index) {
-; KNL-LABEL: test_extractelement_variable_v16i8:
-; KNL: ## BB#0:
-; KNL-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp)
-; KNL-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def>
-; KNL-NEXT: andl $15, %edi
-; KNL-NEXT: leaq -{{[0-9]+}}(%rsp), %rax
-; KNL-NEXT: movb (%rdi,%rax), %al
-; KNL-NEXT: retq
-;
-; SKX-LABEL: test_extractelement_variable_v16i8:
-; SKX: ## BB#0:
-; SKX-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def>
-; SKX-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp)
-; SKX-NEXT: andl $15, %edi
-; SKX-NEXT: leaq -{{[0-9]+}}(%rsp), %rax
-; SKX-NEXT: movb (%rdi,%rax), %al
-; SKX-NEXT: retq
+; CHECK-LABEL: test_extractelement_variable_v16i8:
+; CHECK: ## BB#0:
+; CHECK-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def>
+; CHECK-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp)
+; CHECK-NEXT: andl $15, %edi
+; CHECK-NEXT: leaq -{{[0-9]+}}(%rsp), %rax
+; CHECK-NEXT: movb (%rdi,%rax), %al
+; CHECK-NEXT: retq
%t2 = extractelement <16 x i8> %t1, i32 %index
ret i8 %t2
}
@@ -1936,13 +1817,14 @@ define i8 @test_extractelement_variable_v32i8(<32 x i8> %t1, i32 %index) {
; KNL-NEXT: .cfi_def_cfa_register %rbp
; KNL-NEXT: andq $-32, %rsp
; KNL-NEXT: subq $64, %rsp
-; KNL-NEXT: vmovaps %ymm0, (%rsp)
; KNL-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def>
+; KNL-NEXT: vmovaps %ymm0, (%rsp)
; KNL-NEXT: andl $31, %edi
; KNL-NEXT: movq %rsp, %rax
; KNL-NEXT: movb (%rdi,%rax), %al
; KNL-NEXT: movq %rbp, %rsp
; KNL-NEXT: popq %rbp
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test_extractelement_variable_v32i8:
@@ -1984,14 +1866,15 @@ define i8 @test_extractelement_variable_v64i8(<64 x i8> %t1, i32 %index) {
; KNL-NEXT: .cfi_def_cfa_register %rbp
; KNL-NEXT: andq $-64, %rsp
; KNL-NEXT: subq $128, %rsp
+; KNL-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def>
; KNL-NEXT: vmovaps %ymm1, {{[0-9]+}}(%rsp)
; KNL-NEXT: vmovaps %ymm0, (%rsp)
-; KNL-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def>
; KNL-NEXT: andl $63, %edi
; KNL-NEXT: movq %rsp, %rax
; KNL-NEXT: movb (%rdi,%rax), %al
; KNL-NEXT: movq %rbp, %rsp
; KNL-NEXT: popq %rbp
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test_extractelement_variable_v64i8:
@@ -2042,6 +1925,7 @@ define i8 @test_extractelement_variable_v64i8_indexi8(<64 x i8> %t1, i8 %index)
; KNL-NEXT: movb (%rax,%rcx), %al
; KNL-NEXT: movq %rbp, %rsp
; KNL-NEXT: popq %rbp
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test_extractelement_variable_v64i8_indexi8:
@@ -2075,12 +1959,12 @@ define i8 @test_extractelement_variable_v64i8_indexi8(<64 x i8> %t1, i8 %index)
define zeroext i8 @test_extractelement_varible_v2i1(<2 x i64> %a, <2 x i64> %b, i32 %index) {
; KNL-LABEL: test_extractelement_varible_v2i1:
; KNL: ## BB#0:
+; KNL-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def>
; KNL-NEXT: vmovdqa {{.*#+}} xmm2 = [9223372036854775808,9223372036854775808]
; KNL-NEXT: vpxor %xmm2, %xmm1, %xmm1
; KNL-NEXT: vpxor %xmm2, %xmm0, %xmm0
; KNL-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm0
; KNL-NEXT: vmovdqa %xmm0, -{{[0-9]+}}(%rsp)
-; KNL-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def>
; KNL-NEXT: andl $1, %edi
; KNL-NEXT: movl -24(%rsp,%rdi,8), %eax
; KNL-NEXT: andl $1, %eax
@@ -2105,12 +1989,12 @@ define zeroext i8 @test_extractelement_varible_v2i1(<2 x i64> %a, <2 x i64> %b,
define zeroext i8 @test_extractelement_varible_v4i1(<4 x i32> %a, <4 x i32> %b, i32 %index) {
; KNL-LABEL: test_extractelement_varible_v4i1:
; KNL: ## BB#0:
+; KNL-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def>
; KNL-NEXT: vpbroadcastd {{.*#+}} xmm2 = [2147483648,2147483648,2147483648,2147483648]
; KNL-NEXT: vpxor %xmm2, %xmm1, %xmm1
; KNL-NEXT: vpxor %xmm2, %xmm0, %xmm0
; KNL-NEXT: vpcmpgtd %xmm1, %xmm0, %xmm0
; KNL-NEXT: vmovdqa %xmm0, -{{[0-9]+}}(%rsp)
-; KNL-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def>
; KNL-NEXT: andl $3, %edi
; KNL-NEXT: movl -24(%rsp,%rdi,4), %eax
; KNL-NEXT: andl $1, %eax
@@ -2156,6 +2040,7 @@ define zeroext i8 @test_extractelement_varible_v8i1(<8 x i32> %a, <8 x i32> %b,
; KNL-NEXT: andl $1, %eax
; KNL-NEXT: movq %rbp, %rsp
; KNL-NEXT: popq %rbp
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test_extractelement_varible_v8i1:
@@ -2209,6 +2094,7 @@ define zeroext i8 @test_extractelement_varible_v16i1(<16 x i32> %a, <16 x i32> %
; KNL-NEXT: andl $1, %eax
; KNL-NEXT: movq %rbp, %rsp
; KNL-NEXT: popq %rbp
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test_extractelement_varible_v16i1:
@@ -2265,6 +2151,7 @@ define zeroext i8 @test_extractelement_varible_v32i1(<32 x i8> %a, <32 x i8> %b,
; KNL-NEXT: andl $1, %eax
; KNL-NEXT: movq %rbp, %rsp
; KNL-NEXT: popq %rbp
+; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
; SKX-LABEL: test_extractelement_varible_v32i1:
OpenPOWER on IntegriCloud