aboutsummaryrefslogtreecommitdiff
path: root/test/CodeGen/X86/avx512-calling-conv.ll
diff options
context:
space:
mode:
Diffstat (limited to 'test/CodeGen/X86/avx512-calling-conv.ll')
-rw-r--r--test/CodeGen/X86/avx512-calling-conv.ll103
1 files changed, 62 insertions, 41 deletions
diff --git a/test/CodeGen/X86/avx512-calling-conv.ll b/test/CodeGen/X86/avx512-calling-conv.ll
index a61aeba5aff9..fce592a5318b 100644
--- a/test/CodeGen/X86/avx512-calling-conv.ll
+++ b/test/CodeGen/X86/avx512-calling-conv.ll
@@ -1,13 +1,18 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl | FileCheck %s --check-prefix=ALL_X64 --check-prefix=KNL
-; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=skx | FileCheck %s --check-prefix=ALL_X64 --check-prefix=SKX
-; RUN: llc < %s -mtriple=i686-apple-darwin -mcpu=knl | FileCheck %s --check-prefix=KNL_X32
+; RUN: llc < %s -mtriple=x86_64-apple-darwin9 -mcpu=knl | FileCheck %s --check-prefix=ALL_X64 --check-prefix=KNL
+; RUN: llc < %s -mtriple=x86_64-apple-darwin9 -mcpu=skx | FileCheck %s --check-prefix=ALL_X64 --check-prefix=SKX
+; RUN: llc < %s -mtriple=i686-apple-darwin9 -mcpu=knl | FileCheck %s --check-prefix=KNL_X32
define <16 x i1> @test1() {
-; ALL_X64-LABEL: test1:
-; ALL_X64: ## BB#0:
-; ALL_X64-NEXT: vxorps %xmm0, %xmm0, %xmm0
-; ALL_X64-NEXT: retq
+; KNL-LABEL: test1:
+; KNL: ## BB#0:
+; KNL-NEXT: vxorps %xmm0, %xmm0, %xmm0
+; KNL-NEXT: retq
+;
+; SKX-LABEL: test1:
+; SKX: ## BB#0:
+; SKX-NEXT: vpxord %xmm0, %xmm0, %xmm0
+; SKX-NEXT: retq
;
; KNL_X32-LABEL: test1:
; KNL_X32: ## BB#0:
@@ -25,7 +30,8 @@ define <16 x i1> @test2(<16 x i1>%a, <16 x i1>%b) {
; KNL-NEXT: vpslld $31, %zmm0, %zmm0
; KNL-NEXT: vptestmd %zmm0, %zmm0, %k1
; KNL-NEXT: vptestmd %zmm1, %zmm1, %k1 {%k1}
-; KNL-NEXT: vpbroadcastd {{.*}}(%rip), %zmm0 {%k1} {z}
+; KNL-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0
+; KNL-NEXT: vmovdqa32 %zmm0, %zmm0 {%k1} {z}
; KNL-NEXT: vpmovdb %zmm0, %xmm0
; KNL-NEXT: retq
;
@@ -47,7 +53,8 @@ define <16 x i1> @test2(<16 x i1>%a, <16 x i1>%b) {
; KNL_X32-NEXT: vpslld $31, %zmm0, %zmm0
; KNL_X32-NEXT: vptestmd %zmm0, %zmm0, %k1
; KNL_X32-NEXT: vptestmd %zmm1, %zmm1, %k1 {%k1}
-; KNL_X32-NEXT: vpbroadcastd LCPI1_0, %zmm0 {%k1} {z}
+; KNL_X32-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0
+; KNL_X32-NEXT: vmovdqa32 %zmm0, %zmm0 {%k1} {z}
; KNL_X32-NEXT: vpmovdb %zmm0, %xmm0
; KNL_X32-NEXT: retl
%c = and <16 x i1>%a, %b
@@ -63,7 +70,8 @@ define <8 x i1> @test3(<8 x i1>%a, <8 x i1>%b) {
; KNL-NEXT: vpsllq $63, %zmm0, %zmm0
; KNL-NEXT: vptestmq %zmm0, %zmm0, %k1
; KNL-NEXT: vptestmq %zmm1, %zmm1, %k1 {%k1}
-; KNL-NEXT: vpbroadcastq {{.*}}(%rip), %zmm0 {%k1} {z}
+; KNL-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0
+; KNL-NEXT: vmovdqa64 %zmm0, %zmm0 {%k1} {z}
; KNL-NEXT: vpmovqw %zmm0, %xmm0
; KNL-NEXT: retq
;
@@ -86,8 +94,8 @@ define <8 x i1> @test3(<8 x i1>%a, <8 x i1>%b) {
; KNL_X32-NEXT: vpsllvq %zmm2, %zmm0, %zmm0
; KNL_X32-NEXT: vptestmq %zmm0, %zmm0, %k1
; KNL_X32-NEXT: vptestmq %zmm1, %zmm1, %k1 {%k1}
-; KNL_X32-NEXT: vpbroadcastd LCPI2_1, %zmm0
-; KNL_X32-NEXT: vmovdqu64 %zmm0, %zmm0 {%k1} {z}
+; KNL_X32-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0
+; KNL_X32-NEXT: vmovdqa64 %zmm0, %zmm0 {%k1} {z}
; KNL_X32-NEXT: vpmovqw %zmm0, %xmm0
; KNL_X32-NEXT: retl
%c = and <8 x i1>%a, %b
@@ -102,11 +110,10 @@ define <4 x i1> @test4(<4 x i1>%a, <4 x i1>%b) {
;
; SKX-LABEL: test4:
; SKX: ## BB#0:
+; SKX-NEXT: vpslld $31, %xmm1, %xmm1
; SKX-NEXT: vpslld $31, %xmm0, %xmm0
-; SKX-NEXT: vpmovd2m %xmm0, %k0
-; SKX-NEXT: vpslld $31, %xmm1, %xmm0
-; SKX-NEXT: vpmovd2m %xmm0, %k1
-; SKX-NEXT: kandw %k1, %k0, %k0
+; SKX-NEXT: vptestmd %xmm0, %xmm0, %k1
+; SKX-NEXT: vptestmd %xmm1, %xmm1, %k0 {%k1}
; SKX-NEXT: vpmovm2d %k0, %xmm0
; SKX-NEXT: retq
;
@@ -128,6 +135,7 @@ define <8 x i32> @test5(<8 x i32>%a, <8 x i32>%b) {
; KNL-NEXT: .cfi_def_cfa_offset 16
; KNL-NEXT: vpcmpgtd %ymm1, %ymm0, %ymm0
; KNL-NEXT: vpmovdw %zmm0, %ymm0
+; KNL-NEXT: ## kill: %XMM0<def> %XMM0<kill> %YMM0<kill>
; KNL-NEXT: callq _func8xi1
; KNL-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; KNL-NEXT: vpslld $31, %ymm0, %ymm0
@@ -143,7 +151,7 @@ define <8 x i32> @test5(<8 x i32>%a, <8 x i32>%b) {
; SKX-NEXT: vpcmpgtd %ymm1, %ymm0, %k0
; SKX-NEXT: vpmovm2w %k0, %xmm0
; SKX-NEXT: callq _func8xi1
-; SKX-NEXT: vpmovzxwd %xmm0, %ymm0
+; SKX-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; SKX-NEXT: vpslld $31, %ymm0, %ymm0
; SKX-NEXT: vpsrad $31, %ymm0, %ymm0
; SKX-NEXT: popq %rax
@@ -156,7 +164,8 @@ define <8 x i32> @test5(<8 x i32>%a, <8 x i32>%b) {
; KNL_X32-NEXT: .cfi_def_cfa_offset 16
; KNL_X32-NEXT: vpcmpgtd %ymm1, %ymm0, %ymm0
; KNL_X32-NEXT: vpmovdw %zmm0, %ymm0
-; KNL_X32-NEXT: calll L_func8xi1$stub
+; KNL_X32-NEXT: ## kill: %XMM0<def> %XMM0<kill> %YMM0<kill>
+; KNL_X32-NEXT: calll _func8xi1
; KNL_X32-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; KNL_X32-NEXT: vpslld $31, %ymm0, %ymm0
; KNL_X32-NEXT: vpsrad $31, %ymm0, %ymm0
@@ -177,10 +186,11 @@ define <16 x i32> @test6(<16 x i32>%a, <16 x i32>%b) {
; KNL-NEXT: Ltmp1:
; KNL-NEXT: .cfi_def_cfa_offset 16
; KNL-NEXT: vpcmpgtd %zmm1, %zmm0, %k1
-; KNL-NEXT: vpbroadcastd {{.*}}(%rip), %zmm0 {%k1} {z}
+; KNL-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0
+; KNL-NEXT: vmovdqa32 %zmm0, %zmm0 {%k1} {z}
; KNL-NEXT: vpmovdb %zmm0, %xmm0
; KNL-NEXT: callq _func16xi1
-; KNL-NEXT: vpmovzxbd %xmm0, %zmm0
+; KNL-NEXT: vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
; KNL-NEXT: vpslld $31, %zmm0, %zmm0
; KNL-NEXT: vpsrad $31, %zmm0, %zmm0
; KNL-NEXT: popq %rax
@@ -194,7 +204,7 @@ define <16 x i32> @test6(<16 x i32>%a, <16 x i32>%b) {
; SKX-NEXT: vpcmpgtd %zmm1, %zmm0, %k0
; SKX-NEXT: vpmovm2b %k0, %xmm0
; SKX-NEXT: callq _func16xi1
-; SKX-NEXT: vpmovzxbd %xmm0, %zmm0
+; SKX-NEXT: vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
; SKX-NEXT: vpslld $31, %zmm0, %zmm0
; SKX-NEXT: vpsrad $31, %zmm0, %zmm0
; SKX-NEXT: popq %rax
@@ -206,10 +216,11 @@ define <16 x i32> @test6(<16 x i32>%a, <16 x i32>%b) {
; KNL_X32-NEXT: Ltmp1:
; KNL_X32-NEXT: .cfi_def_cfa_offset 16
; KNL_X32-NEXT: vpcmpgtd %zmm1, %zmm0, %k1
-; KNL_X32-NEXT: vpbroadcastd LCPI5_0, %zmm0 {%k1} {z}
+; KNL_X32-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0
+; KNL_X32-NEXT: vmovdqa32 %zmm0, %zmm0 {%k1} {z}
; KNL_X32-NEXT: vpmovdb %zmm0, %xmm0
-; KNL_X32-NEXT: calll L_func16xi1$stub
-; KNL_X32-NEXT: vpmovzxbd %xmm0, %zmm0
+; KNL_X32-NEXT: calll _func16xi1
+; KNL_X32-NEXT: vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
; KNL_X32-NEXT: vpslld $31, %zmm0, %zmm0
; KNL_X32-NEXT: vpsrad $31, %zmm0, %zmm0
; KNL_X32-NEXT: addl $12, %esp
@@ -254,7 +265,7 @@ define <4 x i32> @test7(<4 x i32>%a, <4 x i32>%b) {
; KNL_X32-NEXT: Ltmp2:
; KNL_X32-NEXT: .cfi_def_cfa_offset 16
; KNL_X32-NEXT: vpcmpgtd %xmm1, %xmm0, %xmm0
-; KNL_X32-NEXT: calll L_func4xi1$stub
+; KNL_X32-NEXT: calll _func4xi1
; KNL_X32-NEXT: vpslld $31, %xmm0, %xmm0
; KNL_X32-NEXT: vpsrad $31, %xmm0, %xmm0
; KNL_X32-NEXT: addl $12, %esp
@@ -273,14 +284,15 @@ define <8 x i1> @test7a(<8 x i32>%a, <8 x i32>%b) {
; KNL-NEXT: .cfi_def_cfa_offset 16
; KNL-NEXT: vpcmpgtd %ymm1, %ymm0, %ymm0
; KNL-NEXT: vpmovdw %zmm0, %ymm0
+; KNL-NEXT: ## kill: %XMM0<def> %XMM0<kill> %YMM0<kill>
; KNL-NEXT: callq _func8xi1
; KNL-NEXT: vpmovsxwq %xmm0, %zmm0
; KNL-NEXT: vpsllq $63, %zmm0, %zmm0
; KNL-NEXT: movb $85, %al
-; KNL-NEXT: movzbl %al, %eax
; KNL-NEXT: kmovw %eax, %k1
; KNL-NEXT: vptestmq %zmm0, %zmm0, %k1 {%k1}
-; KNL-NEXT: vpbroadcastq {{.*}}(%rip), %zmm0 {%k1} {z}
+; KNL-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0
+; KNL-NEXT: vmovdqa64 %zmm0, %zmm0 {%k1} {z}
; KNL-NEXT: vpmovqw %zmm0, %xmm0
; KNL-NEXT: popq %rax
; KNL-NEXT: retq
@@ -309,15 +321,15 @@ define <8 x i1> @test7a(<8 x i32>%a, <8 x i32>%b) {
; KNL_X32-NEXT: .cfi_def_cfa_offset 16
; KNL_X32-NEXT: vpcmpgtd %ymm1, %ymm0, %ymm0
; KNL_X32-NEXT: vpmovdw %zmm0, %ymm0
-; KNL_X32-NEXT: calll L_func8xi1$stub
+; KNL_X32-NEXT: ## kill: %XMM0<def> %XMM0<kill> %YMM0<kill>
+; KNL_X32-NEXT: calll _func8xi1
; KNL_X32-NEXT: vpmovsxwq %xmm0, %zmm0
; KNL_X32-NEXT: vpsllvq LCPI7_0, %zmm0, %zmm0
; KNL_X32-NEXT: movb $85, %al
-; KNL_X32-NEXT: movzbl %al, %eax
; KNL_X32-NEXT: kmovw %eax, %k1
; KNL_X32-NEXT: vptestmq %zmm0, %zmm0, %k1 {%k1}
-; KNL_X32-NEXT: vpbroadcastd LCPI7_1, %zmm0
-; KNL_X32-NEXT: vmovdqu64 %zmm0, %zmm0 {%k1} {z}
+; KNL_X32-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0
+; KNL_X32-NEXT: vmovdqa64 %zmm0, %zmm0 {%k1} {z}
; KNL_X32-NEXT: vpmovqw %zmm0, %xmm0
; KNL_X32-NEXT: addl $12, %esp
; KNL_X32-NEXT: retl
@@ -328,14 +340,23 @@ define <8 x i1> @test7a(<8 x i32>%a, <8 x i32>%b) {
}
define <16 x i8> @test8(<16 x i8> %a1, <16 x i8> %a2, i1 %cond) {
-; ALL_X64-LABEL: test8:
-; ALL_X64: ## BB#0:
-; ALL_X64-NEXT: testb $1, %dil
-; ALL_X64-NEXT: jne LBB8_2
-; ALL_X64-NEXT: ## BB#1:
-; ALL_X64-NEXT: vmovaps %zmm1, %zmm0
-; ALL_X64-NEXT: LBB8_2:
-; ALL_X64-NEXT: retq
+; KNL-LABEL: test8:
+; KNL: ## BB#0:
+; KNL-NEXT: testb $1, %dil
+; KNL-NEXT: jne LBB8_2
+; KNL-NEXT: ## BB#1:
+; KNL-NEXT: vmovaps %zmm1, %zmm0
+; KNL-NEXT: LBB8_2:
+; KNL-NEXT: retq
+;
+; SKX-LABEL: test8:
+; SKX: ## BB#0:
+; SKX-NEXT: testb $1, %dil
+; SKX-NEXT: jne LBB8_2
+; SKX-NEXT: ## BB#1:
+; SKX-NEXT: vmovaps %xmm1, %xmm0
+; SKX-NEXT: LBB8_2:
+; SKX-NEXT: retq
;
; KNL_X32-LABEL: test8:
; KNL_X32: ## BB#0:
@@ -358,7 +379,7 @@ define i1 @test9(double %a, double %b) {
;
; KNL_X32-LABEL: test9:
; KNL_X32: ## BB#0:
-; KNL_X32-NEXT: vmovsd {{[0-9]+}}(%esp), %xmm0
+; KNL_X32-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; KNL_X32-NEXT: vucomisd {{[0-9]+}}(%esp), %xmm0
; KNL_X32-NEXT: setb %al
; KNL_X32-NEXT: retl
@@ -464,7 +485,7 @@ define i32 @test12(i32 %a1, i32 %a2, i32 %b1) {
; KNL_X32-NEXT: movl %eax, {{[0-9]+}}(%esp)
; KNL_X32-NEXT: movl %edi, (%esp)
; KNL_X32-NEXT: calll _test11
-; KNL_X32-NEXT: movb %al, %bl
+; KNL_X32-NEXT: movl %eax, %ebx
; KNL_X32-NEXT: movzbl %bl, %eax
; KNL_X32-NEXT: movl %eax, {{[0-9]+}}(%esp)
; KNL_X32-NEXT: movl %esi, {{[0-9]+}}(%esp)
generated by cgit v1.2.3 (git 2.25.1) at 2025-10-28 18:12:36 +0000