777 lines
40 KiB
LLVM
777 lines
40 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx512dq --show-mc-encoding | FileCheck %s --check-prefixes=CHECK,X86
|
|
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512dq --show-mc-encoding | FileCheck %s --check-prefixes=CHECK,X64
|
|
|
|
declare <2 x double> @llvm.x86.avx512.mask.vextractf64x2.512(<8 x double>, i32, <2 x double>, i8)
|
|
|
|
define <2 x double>@test_int_x86_avx512_vextractf64x2_512(<8 x double> %x0, <2 x double> %x2) {
|
|
; CHECK-LABEL: test_int_x86_avx512_vextractf64x2_512:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vextractf128 $1, %ymm0, %xmm0 # encoding: [0xc4,0xe3,0x7d,0x19,0xc0,0x01]
|
|
; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
|
|
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
|
|
%res = call <2 x double> @llvm.x86.avx512.mask.vextractf64x2.512(<8 x double> %x0,i32 1, <2 x double> zeroinitializer, i8 -1)
|
|
ret <2 x double> %res
|
|
}
|
|
|
|
define <2 x double>@test_int_x86_avx512_mask_vextractf64x2_512(<8 x double> %x0, <2 x double> %x2, i8 %x3) {
|
|
; X86-LABEL: test_int_x86_avx512_mask_vextractf64x2_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vextractf64x2 $1, %zmm0, %xmm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x19,0xc1,0x01]
|
|
; X86-NEXT: vmovapd %xmm1, %xmm0 # encoding: [0xc5,0xf9,0x28,0xc1]
|
|
; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_mask_vextractf64x2_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vextractf64x2 $1, %zmm0, %xmm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x19,0xc1,0x01]
|
|
; X64-NEXT: vmovapd %xmm1, %xmm0 # encoding: [0xc5,0xf9,0x28,0xc1]
|
|
; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
%res = call <2 x double> @llvm.x86.avx512.mask.vextractf64x2.512(<8 x double> %x0,i32 1, <2 x double> %x2, i8 %x3)
|
|
ret <2 x double> %res
|
|
}
|
|
|
|
define <2 x double>@test_int_x86_avx512_maskz_vextractf64x2_512(<8 x double> %x0, i8 %x3) {
|
|
; X86-LABEL: test_int_x86_avx512_maskz_vextractf64x2_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vextractf64x2 $1, %zmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0xc9,0x19,0xc0,0x01]
|
|
; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_maskz_vextractf64x2_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vextractf64x2 $1, %zmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0xc9,0x19,0xc0,0x01]
|
|
; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
%res = call <2 x double> @llvm.x86.avx512.mask.vextractf64x2.512(<8 x double> %x0,i32 1, <2 x double> zeroinitializer, i8 %x3)
|
|
ret <2 x double> %res
|
|
}
|
|
|
|
declare <8 x float> @llvm.x86.avx512.mask.vextractf32x8.512(<16 x float>, i32, <8 x float>, i8)
|
|
|
|
define <8 x float>@test_int_x86_avx512_vextractf32x8(<16 x float> %x0, <8 x float> %x2) {
|
|
; CHECK-LABEL: test_int_x86_avx512_vextractf32x8:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vextractf64x4 $1, %zmm0, %ymm0 # encoding: [0x62,0xf3,0xfd,0x48,0x1b,0xc0,0x01]
|
|
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
|
|
%res = call <8 x float> @llvm.x86.avx512.mask.vextractf32x8.512(<16 x float> %x0,i32 1, <8 x float> zeroinitializer, i8 -1)
|
|
ret <8 x float> %res
|
|
}
|
|
|
|
define <8 x float>@test_int_x86_avx512_mask_vextractf32x8(<16 x float> %x0, <8 x float> %x2, i8 %x3) {
|
|
; X86-LABEL: test_int_x86_avx512_mask_vextractf32x8:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vextractf32x8 $1, %zmm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x1b,0xc1,0x01]
|
|
; X86-NEXT: vmovaps %ymm1, %ymm0 # encoding: [0xc5,0xfc,0x28,0xc1]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_mask_vextractf32x8:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vextractf32x8 $1, %zmm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x1b,0xc1,0x01]
|
|
; X64-NEXT: vmovaps %ymm1, %ymm0 # encoding: [0xc5,0xfc,0x28,0xc1]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
%res = call <8 x float> @llvm.x86.avx512.mask.vextractf32x8.512(<16 x float> %x0,i32 1, <8 x float> %x2, i8 %x3)
|
|
ret <8 x float> %res
|
|
}
|
|
|
|
define <8 x float>@test_int_x86_avx512_maskz_vextractf32x8(<16 x float> %x0, i8 %x3) {
|
|
; X86-LABEL: test_int_x86_avx512_maskz_vextractf32x8:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vextractf32x8 $1, %zmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xc9,0x1b,0xc0,0x01]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_maskz_vextractf32x8:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vextractf32x8 $1, %zmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xc9,0x1b,0xc0,0x01]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
%res = call <8 x float> @llvm.x86.avx512.mask.vextractf32x8.512(<16 x float> %x0,i32 1, <8 x float> zeroinitializer, i8 %x3)
|
|
ret <8 x float> %res
|
|
}
|
|
|
|
declare <16 x float> @llvm.x86.avx512.mask.insertf32x8.512(<16 x float>, <8 x float>, i32, <16 x float>, i16)
|
|
|
|
define <16 x float>@test_int_x86_avx512_insertf32x8_512(<16 x float> %x0, <8 x float> %x1, <16 x float> %x3) {
|
|
; CHECK-LABEL: test_int_x86_avx512_insertf32x8_512:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vinsertf64x4 $1, %ymm1, %zmm0, %zmm0 # encoding: [0x62,0xf3,0xfd,0x48,0x1a,0xc1,0x01]
|
|
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
|
|
%res = call <16 x float> @llvm.x86.avx512.mask.insertf32x8.512(<16 x float> %x0, <8 x float> %x1, i32 1, <16 x float> %x3, i16 -1)
|
|
ret <16 x float> %res
|
|
}
|
|
|
|
define <16 x float>@test_int_x86_avx512_mask_insertf32x8_512(<16 x float> %x0, <8 x float> %x1, <16 x float> %x3, i16 %x4) {
|
|
; X86-LABEL: test_int_x86_avx512_mask_insertf32x8_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vinsertf32x8 $1, %ymm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x1a,0xd1,0x01]
|
|
; X86-NEXT: vmovaps %zmm2, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_mask_insertf32x8_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vinsertf32x8 $1, %ymm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x1a,0xd1,0x01]
|
|
; X64-NEXT: vmovaps %zmm2, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
%res = call <16 x float> @llvm.x86.avx512.mask.insertf32x8.512(<16 x float> %x0, <8 x float> %x1, i32 1, <16 x float> %x3, i16 %x4)
|
|
ret <16 x float> %res
|
|
}
|
|
|
|
define <16 x float>@test_int_x86_avx512_maskz_insertf32x8_512(<16 x float> %x0, <8 x float> %x1, i16 %x4) {
|
|
; X86-LABEL: test_int_x86_avx512_maskz_insertf32x8_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vinsertf32x8 $1, %ymm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xc9,0x1a,0xc1,0x01]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_maskz_insertf32x8_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vinsertf32x8 $1, %ymm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xc9,0x1a,0xc1,0x01]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
%res = call <16 x float> @llvm.x86.avx512.mask.insertf32x8.512(<16 x float> %x0, <8 x float> %x1, i32 1, <16 x float> zeroinitializer, i16 %x4)
|
|
ret <16 x float> %res
|
|
}
|
|
|
|
declare <8 x double> @llvm.x86.avx512.mask.insertf64x2.512(<8 x double>, <2 x double>, i32, <8 x double>, i8)
|
|
|
|
define <8 x double>@test_int_x86_avx512_insertf64x2_512(<8 x double> %x0, <2 x double> %x1,<8 x double> %x3) {
|
|
; CHECK-LABEL: test_int_x86_avx512_insertf64x2_512:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vinsertf32x4 $1, %xmm1, %zmm0, %zmm0 # encoding: [0x62,0xf3,0x7d,0x48,0x18,0xc1,0x01]
|
|
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
|
|
%res = call <8 x double> @llvm.x86.avx512.mask.insertf64x2.512(<8 x double> %x0, <2 x double> %x1, i32 1, <8 x double> %x3, i8 -1)
|
|
ret <8 x double> %res
|
|
}
|
|
|
|
define <8 x double>@test_int_x86_avx512_mask_insertf64x2_512(<8 x double> %x0, <2 x double> %x1,<8 x double> %x3, i8 %x4) {
|
|
; X86-LABEL: test_int_x86_avx512_mask_insertf64x2_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vinsertf64x2 $1, %xmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x18,0xd1,0x01]
|
|
; X86-NEXT: vmovapd %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x28,0xc2]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_mask_insertf64x2_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vinsertf64x2 $1, %xmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x18,0xd1,0x01]
|
|
; X64-NEXT: vmovapd %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x28,0xc2]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
%res = call <8 x double> @llvm.x86.avx512.mask.insertf64x2.512(<8 x double> %x0, <2 x double> %x1, i32 1, <8 x double> %x3, i8 %x4)
|
|
ret <8 x double> %res
|
|
}
|
|
|
|
define <8 x double>@test_int_x86_avx512_maskz_insertf64x2_512(<8 x double> %x0, <2 x double> %x1, i8 %x4) {
|
|
; X86-LABEL: test_int_x86_avx512_maskz_insertf64x2_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vinsertf64x2 $1, %xmm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0xc9,0x18,0xc1,0x01]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_maskz_insertf64x2_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vinsertf64x2 $1, %xmm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0xc9,0x18,0xc1,0x01]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
%res = call <8 x double> @llvm.x86.avx512.mask.insertf64x2.512(<8 x double> %x0, <2 x double> %x1, i32 1, <8 x double> zeroinitializer, i8 %x4)
|
|
ret <8 x double> %res
|
|
}
|
|
|
|
declare <16 x i32> @llvm.x86.avx512.mask.inserti32x8.512(<16 x i32>, <8 x i32>, i32, <16 x i32>, i16)
|
|
|
|
define <16 x i32>@test_int_x86_avx512_inserti32x8_512(<16 x i32> %x0, <8 x i32> %x1, <16 x i32> %x3) {
|
|
; CHECK-LABEL: test_int_x86_avx512_inserti32x8_512:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vinsertf64x4 $1, %ymm1, %zmm0, %zmm0 # encoding: [0x62,0xf3,0xfd,0x48,0x1a,0xc1,0x01]
|
|
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
|
|
%res = call <16 x i32> @llvm.x86.avx512.mask.inserti32x8.512(<16 x i32> %x0, <8 x i32> %x1, i32 1, <16 x i32> %x3, i16 -1)
|
|
ret <16 x i32> %res
|
|
}
|
|
|
|
define <16 x i32>@test_int_x86_avx512_mask_inserti32x8_512(<16 x i32> %x0, <8 x i32> %x1, <16 x i32> %x3, i16 %x4) {
|
|
; X86-LABEL: test_int_x86_avx512_mask_inserti32x8_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vinserti32x8 $1, %ymm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x3a,0xd1,0x01]
|
|
; X86-NEXT: vmovdqa64 %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc2]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_mask_inserti32x8_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vinserti32x8 $1, %ymm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x3a,0xd1,0x01]
|
|
; X64-NEXT: vmovdqa64 %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc2]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
%res = call <16 x i32> @llvm.x86.avx512.mask.inserti32x8.512(<16 x i32> %x0, <8 x i32> %x1, i32 1, <16 x i32> %x3, i16 %x4)
|
|
ret <16 x i32> %res
|
|
}
|
|
|
|
define <16 x i32>@test_int_x86_avx512_maskz_inserti32x8_512(<16 x i32> %x0, <8 x i32> %x1, i16 %x4) {
|
|
; X86-LABEL: test_int_x86_avx512_maskz_inserti32x8_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vinserti32x8 $1, %ymm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xc9,0x3a,0xc1,0x01]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_maskz_inserti32x8_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vinserti32x8 $1, %ymm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xc9,0x3a,0xc1,0x01]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
%res = call <16 x i32> @llvm.x86.avx512.mask.inserti32x8.512(<16 x i32> %x0, <8 x i32> %x1, i32 1, <16 x i32> zeroinitializer, i16 %x4)
|
|
ret <16 x i32> %res
|
|
}
|
|
|
|
declare <8 x i64> @llvm.x86.avx512.mask.inserti64x2.512(<8 x i64>, <2 x i64>, i32, <8 x i64>, i8)
|
|
|
|
define <8 x i64>@test_int_x86_avx512_inserti64x2_512(<8 x i64> %x0, <2 x i64> %x1, <8 x i64> %x3) {
|
|
; CHECK-LABEL: test_int_x86_avx512_inserti64x2_512:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vinsertf32x4 $1, %xmm1, %zmm0, %zmm0 # encoding: [0x62,0xf3,0x7d,0x48,0x18,0xc1,0x01]
|
|
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
|
|
%res = call <8 x i64> @llvm.x86.avx512.mask.inserti64x2.512(<8 x i64> %x0, <2 x i64> %x1, i32 1, <8 x i64> %x3, i8 -1)
|
|
ret <8 x i64> %res
|
|
}
|
|
|
|
define <8 x i64>@test_int_x86_avx512_mask_inserti64x2_512(<8 x i64> %x0, <2 x i64> %x1, <8 x i64> %x3, i8 %x4) {
|
|
; X86-LABEL: test_int_x86_avx512_mask_inserti64x2_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vinserti64x2 $1, %xmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x38,0xd1,0x01]
|
|
; X86-NEXT: vmovdqa64 %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc2]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_mask_inserti64x2_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vinserti64x2 $1, %xmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x38,0xd1,0x01]
|
|
; X64-NEXT: vmovdqa64 %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc2]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
%res = call <8 x i64> @llvm.x86.avx512.mask.inserti64x2.512(<8 x i64> %x0, <2 x i64> %x1, i32 1, <8 x i64> %x3, i8 %x4)
|
|
ret <8 x i64> %res
|
|
}
|
|
|
|
define <8 x i64>@test_int_x86_avx512_maskz_inserti64x2_512(<8 x i64> %x0, <2 x i64> %x1, i8 %x4) {
|
|
; X86-LABEL: test_int_x86_avx512_maskz_inserti64x2_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vinserti64x2 $1, %xmm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0xc9,0x38,0xc1,0x01]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_maskz_inserti64x2_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vinserti64x2 $1, %xmm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0xc9,0x38,0xc1,0x01]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
%res = call <8 x i64> @llvm.x86.avx512.mask.inserti64x2.512(<8 x i64> %x0, <2 x i64> %x1, i32 1, <8 x i64> zeroinitializer, i8 %x4)
|
|
ret <8 x i64> %res
|
|
}
|
|
|
|
|
|
declare <16 x i32> @llvm.x86.avx512.cvtmask2d.512(i16)
|
|
|
|
define <16 x i32>@test_int_x86_avx512_cvtmask2d_512(i16 %x0) {
|
|
; X86-LABEL: test_int_x86_avx512_cvtmask2d_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k0 # encoding: [0xc5,0xf8,0x90,0x44,0x24,0x04]
|
|
; X86-NEXT: vpmovm2d %k0, %zmm0 # encoding: [0x62,0xf2,0x7e,0x48,0x38,0xc0]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_cvtmask2d_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: kmovw %edi, %k0 # encoding: [0xc5,0xf8,0x92,0xc7]
|
|
; X64-NEXT: vpmovm2d %k0, %zmm0 # encoding: [0x62,0xf2,0x7e,0x48,0x38,0xc0]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
%res = call <16 x i32> @llvm.x86.avx512.cvtmask2d.512(i16 %x0)
|
|
ret <16 x i32> %res
|
|
}
|
|
|
|
declare <8 x i64> @llvm.x86.avx512.cvtmask2q.512(i8)
|
|
|
|
define <8 x i64>@test_int_x86_avx512_cvtmask2q_512(i8 %x0) {
|
|
; X86-LABEL: test_int_x86_avx512_cvtmask2q_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: kmovb {{[0-9]+}}(%esp), %k0 # encoding: [0xc5,0xf9,0x90,0x44,0x24,0x04]
|
|
; X86-NEXT: vpmovm2q %k0, %zmm0 # encoding: [0x62,0xf2,0xfe,0x48,0x38,0xc0]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_cvtmask2q_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: kmovw %edi, %k0 # encoding: [0xc5,0xf8,0x92,0xc7]
|
|
; X64-NEXT: vpmovm2q %k0, %zmm0 # encoding: [0x62,0xf2,0xfe,0x48,0x38,0xc0]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
%res = call <8 x i64> @llvm.x86.avx512.cvtmask2q.512(i8 %x0)
|
|
ret <8 x i64> %res
|
|
}
|
|
|
|
declare <16 x float> @llvm.x86.avx512.mask.broadcastf32x8.512(<8 x float>, <16 x float>, i16)
|
|
|
|
define <16 x float>@test_int_x86_avx512_broadcastf32x8_512(<8 x float> %x0, <16 x float> %x2) {
|
|
; CHECK-LABEL: test_int_x86_avx512_broadcastf32x8_512:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
|
|
; CHECK-NEXT: vinsertf64x4 $1, %ymm0, %zmm0, %zmm0 # encoding: [0x62,0xf3,0xfd,0x48,0x1a,0xc0,0x01]
|
|
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
|
|
|
|
%res = call <16 x float> @llvm.x86.avx512.mask.broadcastf32x8.512(<8 x float> %x0, <16 x float> %x2, i16 -1)
|
|
ret <16 x float> %res
|
|
}
|
|
|
|
define <16 x float>@test_int_x86_avx512_mask_broadcastf32x8_512(<8 x float> %x0, <16 x float> %x2, i16 %mask) {
|
|
; X86-LABEL: test_int_x86_avx512_mask_broadcastf32x8_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
|
|
; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vinsertf32x8 $1, %ymm0, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x1a,0xc8,0x01]
|
|
; X86-NEXT: vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_mask_broadcastf32x8_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vinsertf32x8 $1, %ymm0, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x1a,0xc8,0x01]
|
|
; X64-NEXT: vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
|
|
%res = call <16 x float> @llvm.x86.avx512.mask.broadcastf32x8.512(<8 x float> %x0, <16 x float> %x2, i16 %mask)
|
|
ret <16 x float> %res
|
|
}
|
|
|
|
define <16 x float>@test_int_x86_avx512_maskz_broadcastf32x8_512(<8 x float> %x0, i16 %mask) {
|
|
; X86-LABEL: test_int_x86_avx512_maskz_broadcastf32x8_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
|
|
; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vinsertf32x8 $1, %ymm0, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xc9,0x1a,0xc0,0x01]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_maskz_broadcastf32x8_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vinsertf32x8 $1, %ymm0, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xc9,0x1a,0xc0,0x01]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
|
|
%res = call <16 x float> @llvm.x86.avx512.mask.broadcastf32x8.512(<8 x float> %x0, <16 x float> zeroinitializer, i16 %mask)
|
|
ret <16 x float> %res
|
|
}
|
|
|
|
define <16 x float>@test_int_x86_avx512_mask_broadcastf32x8_512_load(<8 x float>* %x0ptr, <16 x float> %x2, i16 %mask) {
|
|
; X86-LABEL: test_int_x86_avx512_mask_broadcastf32x8_512_load:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
|
|
; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
|
|
; X86-NEXT: vbroadcastf32x8 (%eax), %zmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x1b,0x00]
|
|
; X86-NEXT: # zmm0 {%k1} = mem[0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_mask_broadcastf32x8_512_load:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
|
|
; X64-NEXT: vbroadcastf32x8 (%rdi), %zmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x1b,0x07]
|
|
; X64-NEXT: # zmm0 {%k1} = mem[0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
|
|
%x0 = load <8 x float>, <8 x float>* %x0ptr
|
|
%res = call <16 x float> @llvm.x86.avx512.mask.broadcastf32x8.512(<8 x float> %x0, <16 x float> %x2, i16 %mask)
|
|
ret <16 x float> %res
|
|
}
|
|
|
|
declare <8 x double> @llvm.x86.avx512.mask.broadcastf64x2.512(<2 x double>, <8 x double>, i8)
|
|
|
|
define <8 x double>@test_int_x86_avx512_broadcastf64x2_512(<2 x double> %x0, <8 x double> %x2) {
|
|
; CHECK-LABEL: test_int_x86_avx512_broadcastf64x2_512:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: # kill: def $xmm0 killed $xmm0 def $ymm0
|
|
; CHECK-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0x7d,0x18,0xc0,0x01]
|
|
; CHECK-NEXT: vinsertf64x4 $1, %ymm0, %zmm0, %zmm0 # encoding: [0x62,0xf3,0xfd,0x48,0x1a,0xc0,0x01]
|
|
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
|
|
|
|
%res = call <8 x double> @llvm.x86.avx512.mask.broadcastf64x2.512(<2 x double> %x0, <8 x double> %x2, i8 -1)
|
|
ret <8 x double> %res
|
|
}
|
|
|
|
define <8 x double>@test_int_x86_avx512_mask_broadcastf64x2_512(<2 x double> %x0, <8 x double> %x2, i8 %mask) {
|
|
; X86-LABEL: test_int_x86_avx512_mask_broadcastf64x2_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: # kill: def $xmm0 killed $xmm0 def $ymm0
|
|
; X86-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0x7d,0x18,0xc0,0x01]
|
|
; X86-NEXT: kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vinsertf64x4 $1, %ymm0, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x1a,0xc8,0x01]
|
|
; X86-NEXT: vmovapd %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x28,0xc1]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_mask_broadcastf64x2_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: # kill: def $xmm0 killed $xmm0 def $ymm0
|
|
; X64-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0x7d,0x18,0xc0,0x01]
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vinsertf64x4 $1, %ymm0, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x1a,0xc8,0x01]
|
|
; X64-NEXT: vmovapd %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x28,0xc1]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
|
|
%res = call <8 x double> @llvm.x86.avx512.mask.broadcastf64x2.512(<2 x double> %x0, <8 x double> %x2, i8 %mask)
|
|
ret <8 x double> %res
|
|
}
|
|
|
|
define <8 x double>@test_int_x86_avx512_maskz_broadcastf64x2_512(<2 x double> %x0, i8 %mask) {
|
|
; X86-LABEL: test_int_x86_avx512_maskz_broadcastf64x2_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: # kill: def $xmm0 killed $xmm0 def $ymm0
|
|
; X86-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0x7d,0x18,0xc0,0x01]
|
|
; X86-NEXT: kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vinsertf64x4 $1, %ymm0, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0xc9,0x1a,0xc0,0x01]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_maskz_broadcastf64x2_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: # kill: def $xmm0 killed $xmm0 def $ymm0
|
|
; X64-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0x7d,0x18,0xc0,0x01]
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vinsertf64x4 $1, %ymm0, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0xc9,0x1a,0xc0,0x01]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
|
|
%res = call <8 x double> @llvm.x86.avx512.mask.broadcastf64x2.512(<2 x double> %x0, <8 x double> zeroinitializer, i8 %mask)
|
|
ret <8 x double> %res
|
|
}
|
|
|
|
define <8 x double>@test_int_x86_avx512_mask_broadcastf64x2_512_load(<2 x double>* %x0ptr, <8 x double> %x2, i8 %mask) {
|
|
; X86-LABEL: test_int_x86_avx512_mask_broadcastf64x2_512_load:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
|
|
; X86-NEXT: kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x08]
|
|
; X86-NEXT: vbroadcastf64x2 (%eax), %zmm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x49,0x1a,0x00]
|
|
; X86-NEXT: # zmm0 {%k1} = mem[0,1,0,1,0,1,0,1]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_mask_broadcastf64x2_512_load:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
|
|
; X64-NEXT: vbroadcastf64x2 (%rdi), %zmm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x49,0x1a,0x07]
|
|
; X64-NEXT: # zmm0 {%k1} = mem[0,1,0,1,0,1,0,1]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
|
|
%x0 = load <2 x double>, <2 x double>* %x0ptr
|
|
%res = call <8 x double> @llvm.x86.avx512.mask.broadcastf64x2.512(<2 x double> %x0, <8 x double> %x2, i8 %mask)
|
|
ret <8 x double> %res
|
|
}
|
|
|
|
declare <16 x i32> @llvm.x86.avx512.mask.broadcasti32x8.512(<8 x i32>, <16 x i32>, i16)
|
|
|
|
define <16 x i32>@test_int_x86_avx512_broadcasti32x8_512(<8 x i32> %x0, <16 x i32> %x2) {
|
|
; CHECK-LABEL: test_int_x86_avx512_broadcasti32x8_512:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
|
|
; CHECK-NEXT: vinsertf64x4 $1, %ymm0, %zmm0, %zmm0 # encoding: [0x62,0xf3,0xfd,0x48,0x1a,0xc0,0x01]
|
|
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
|
|
|
|
%res = call <16 x i32> @llvm.x86.avx512.mask.broadcasti32x8.512(<8 x i32> %x0, <16 x i32> %x2, i16 -1)
|
|
ret <16 x i32> %res
|
|
}
|
|
|
|
define <16 x i32>@test_int_x86_avx512_mask_broadcasti32x8_512(<8 x i32> %x0, <16 x i32> %x2, i16 %mask) {
|
|
; X86-LABEL: test_int_x86_avx512_mask_broadcasti32x8_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
|
|
; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vinserti32x8 $1, %ymm0, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x3a,0xc8,0x01]
|
|
; X86-NEXT: vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_mask_broadcasti32x8_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vinserti32x8 $1, %ymm0, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x3a,0xc8,0x01]
|
|
; X64-NEXT: vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
|
|
%res = call <16 x i32> @llvm.x86.avx512.mask.broadcasti32x8.512(<8 x i32> %x0, <16 x i32> %x2, i16 %mask)
|
|
ret <16 x i32> %res
|
|
}
|
|
|
|
define <16 x i32>@test_int_x86_avx512_maskz_broadcasti32x8_512(<8 x i32> %x0, i16 %mask) {
|
|
; X86-LABEL: test_int_x86_avx512_maskz_broadcasti32x8_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
|
|
; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vinserti32x8 $1, %ymm0, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xc9,0x3a,0xc0,0x01]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_maskz_broadcasti32x8_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vinserti32x8 $1, %ymm0, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xc9,0x3a,0xc0,0x01]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
|
|
%res = call <16 x i32> @llvm.x86.avx512.mask.broadcasti32x8.512(<8 x i32> %x0, <16 x i32> zeroinitializer, i16 %mask)
|
|
ret <16 x i32> %res
|
|
}
|
|
|
|
define <16 x i32>@test_int_x86_avx512_mask_broadcasti32x8_512_load(<8 x i32>* %x0ptr, <16 x i32> %x2, i16 %mask) {
|
|
; X86-LABEL: test_int_x86_avx512_mask_broadcasti32x8_512_load:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
|
|
; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
|
|
; X86-NEXT: vbroadcasti32x8 (%eax), %zmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x5b,0x00]
|
|
; X86-NEXT: # zmm0 {%k1} = mem[0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_mask_broadcasti32x8_512_load:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
|
|
; X64-NEXT: vbroadcasti32x8 (%rdi), %zmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x5b,0x07]
|
|
; X64-NEXT: # zmm0 {%k1} = mem[0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
|
|
%x0 = load <8 x i32>, <8 x i32>* %x0ptr
|
|
%res = call <16 x i32> @llvm.x86.avx512.mask.broadcasti32x8.512(<8 x i32> %x0, <16 x i32> %x2, i16 %mask)
|
|
ret <16 x i32> %res
|
|
}
|
|
|
|
declare <8 x i64> @llvm.x86.avx512.mask.broadcasti64x2.512(<2 x i64>, <8 x i64>, i8)
|
|
|
|
define <8 x i64>@test_int_x86_avx512_broadcasti64x2_512(<2 x i64> %x0, <8 x i64> %x2) {
|
|
; CHECK-LABEL: test_int_x86_avx512_broadcasti64x2_512:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: # kill: def $xmm0 killed $xmm0 def $ymm0
|
|
; CHECK-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0x7d,0x18,0xc0,0x01]
|
|
; CHECK-NEXT: vinsertf64x4 $1, %ymm0, %zmm0, %zmm0 # encoding: [0x62,0xf3,0xfd,0x48,0x1a,0xc0,0x01]
|
|
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
|
|
|
|
%res = call <8 x i64> @llvm.x86.avx512.mask.broadcasti64x2.512(<2 x i64> %x0, <8 x i64> %x2, i8 -1)
|
|
ret <8 x i64> %res
|
|
}
|
|
|
|
define <8 x i64>@test_int_x86_avx512_mask_broadcasti64x2_512(<2 x i64> %x0, <8 x i64> %x2, i8 %mask) {
|
|
; X86-LABEL: test_int_x86_avx512_mask_broadcasti64x2_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: # kill: def $xmm0 killed $xmm0 def $ymm0
|
|
; X86-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0x7d,0x38,0xc0,0x01]
|
|
; X86-NEXT: kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x3a,0xc8,0x01]
|
|
; X86-NEXT: vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_mask_broadcasti64x2_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: # kill: def $xmm0 killed $xmm0 def $ymm0
|
|
; X64-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0x7d,0x38,0xc0,0x01]
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x3a,0xc8,0x01]
|
|
; X64-NEXT: vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
|
|
%res = call <8 x i64> @llvm.x86.avx512.mask.broadcasti64x2.512(<2 x i64> %x0, <8 x i64> %x2, i8 %mask)
|
|
ret <8 x i64> %res
|
|
}
|
|
|
|
define <8 x i64>@test_int_x86_avx512_maskz_broadcasti64x2_512(<2 x i64> %x0, i8 %mask) {
|
|
; X86-LABEL: test_int_x86_avx512_maskz_broadcasti64x2_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: # kill: def $xmm0 killed $xmm0 def $ymm0
|
|
; X86-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0x7d,0x38,0xc0,0x01]
|
|
; X86-NEXT: kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0xc9,0x3a,0xc0,0x01]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_maskz_broadcasti64x2_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: # kill: def $xmm0 killed $xmm0 def $ymm0
|
|
; X64-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0x7d,0x38,0xc0,0x01]
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0xc9,0x3a,0xc0,0x01]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
|
|
%res = call <8 x i64> @llvm.x86.avx512.mask.broadcasti64x2.512(<2 x i64> %x0, <8 x i64> zeroinitializer, i8 %mask)
|
|
ret <8 x i64> %res
|
|
}
|
|
|
|
define <8 x i64>@test_int_x86_avx512_mask_broadcasti64x2_512_load(<2 x i64>* %x0ptr, <8 x i64> %x2, i8 %mask) {
|
|
; X86-LABEL: test_int_x86_avx512_mask_broadcasti64x2_512_load:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
|
|
; X86-NEXT: kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x08]
|
|
; X86-NEXT: vbroadcasti64x2 (%eax), %zmm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x49,0x5a,0x00]
|
|
; X86-NEXT: # zmm0 {%k1} = mem[0,1,0,1,0,1,0,1]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_mask_broadcasti64x2_512_load:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
|
|
; X64-NEXT: vbroadcasti64x2 (%rdi), %zmm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x49,0x5a,0x07]
|
|
; X64-NEXT: # zmm0 {%k1} = mem[0,1,0,1,0,1,0,1]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
|
|
%x0 = load <2 x i64>, <2 x i64>* %x0ptr
|
|
%res = call <8 x i64> @llvm.x86.avx512.mask.broadcasti64x2.512(<2 x i64> %x0, <8 x i64> %x2, i8 %mask)
|
|
ret <8 x i64> %res
|
|
}
|
|
|
|
declare i16 @llvm.x86.avx512.cvtd2mask.512(<16 x i32>)
|
|
|
|
define i16@test_int_x86_avx512_cvtd2mask_512(<16 x i32> %x0) {
|
|
; CHECK-LABEL: test_int_x86_avx512_cvtd2mask_512:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vpmovd2m %zmm0, %k0 # encoding: [0x62,0xf2,0x7e,0x48,0x39,0xc0]
|
|
; CHECK-NEXT: kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
|
|
; CHECK-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
|
|
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
|
|
%res = call i16 @llvm.x86.avx512.cvtd2mask.512(<16 x i32> %x0)
|
|
ret i16 %res
|
|
}
|
|
|
|
declare i8 @llvm.x86.avx512.cvtq2mask.512(<8 x i64>)
|
|
|
|
define i8@test_int_x86_avx512_cvtq2mask_512(<8 x i64> %x0) {
|
|
; CHECK-LABEL: test_int_x86_avx512_cvtq2mask_512:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vpmovq2m %zmm0, %k0 # encoding: [0x62,0xf2,0xfe,0x48,0x39,0xc0]
|
|
; CHECK-NEXT: kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
|
|
; CHECK-NEXT: # kill: def $al killed $al killed $eax
|
|
; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
|
|
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
|
|
%res = call i8 @llvm.x86.avx512.cvtq2mask.512(<8 x i64> %x0)
|
|
ret i8 %res
|
|
}
|
|
|
|
declare i8 @llvm.x86.avx512.mask.fpclass.pd.512(<8 x double>, i32, i8)
|
|
|
|
define i8 @test_int_x86_avx512_mask_fpclass_pd_512(<8 x double> %x0) {
|
|
; CHECK-LABEL: test_int_x86_avx512_mask_fpclass_pd_512:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vfpclasspd $2, %zmm0, %k1 # encoding: [0x62,0xf3,0xfd,0x48,0x66,0xc8,0x02]
|
|
; CHECK-NEXT: vfpclasspd $4, %zmm0, %k0 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x66,0xc0,0x04]
|
|
; CHECK-NEXT: kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
|
|
; CHECK-NEXT: # kill: def $al killed $al killed $eax
|
|
; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
|
|
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
|
|
%res = call i8 @llvm.x86.avx512.mask.fpclass.pd.512(<8 x double> %x0, i32 4, i8 -1)
|
|
%res1 = call i8 @llvm.x86.avx512.mask.fpclass.pd.512(<8 x double> %x0, i32 2, i8 %res)
|
|
ret i8 %res1
|
|
}
|
|
declare i16 @llvm.x86.avx512.mask.fpclass.ps.512(<16 x float>, i32, i16)
|
|
|
|
define i16@test_int_x86_avx512_mask_fpclass_ps_512(<16 x float> %x0) {
|
|
; CHECK-LABEL: test_int_x86_avx512_mask_fpclass_ps_512:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vfpclassps $2, %zmm0, %k1 # encoding: [0x62,0xf3,0x7d,0x48,0x66,0xc8,0x02]
|
|
; CHECK-NEXT: vfpclassps $4, %zmm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x66,0xc0,0x04]
|
|
; CHECK-NEXT: kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
|
|
; CHECK-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
|
|
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
|
|
%res = call i16 @llvm.x86.avx512.mask.fpclass.ps.512(<16 x float> %x0, i32 4, i16 -1)
|
|
%res1 = call i16 @llvm.x86.avx512.mask.fpclass.ps.512(<16 x float> %x0, i32 2, i16 %res)
|
|
ret i16 %res1
|
|
}
|
|
|
|
declare <8 x double> @llvm.x86.avx512.mask.cvtqq2pd.512(<8 x i64>, <8 x double>, i8, i32)
|
|
|
|
define <8 x double>@test_int_x86_avx512_mask_cvt_qq2pd_512(<8 x i64> %x0, <8 x double> %x1, i8 %x2) {
|
|
; X86-LABEL: test_int_x86_avx512_mask_cvt_qq2pd_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vcvtqq2pd %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0xfe,0x49,0xe6,0xc8]
|
|
; X86-NEXT: vcvtqq2pd {rn-sae}, %zmm0, %zmm0 # encoding: [0x62,0xf1,0xfe,0x18,0xe6,0xc0]
|
|
; X86-NEXT: vaddpd %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0xf5,0x48,0x58,0xc0]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_mask_cvt_qq2pd_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vcvtqq2pd %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0xfe,0x49,0xe6,0xc8]
|
|
; X64-NEXT: vcvtqq2pd {rn-sae}, %zmm0, %zmm0 # encoding: [0x62,0xf1,0xfe,0x18,0xe6,0xc0]
|
|
; X64-NEXT: vaddpd %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0xf5,0x48,0x58,0xc0]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
%res = call <8 x double> @llvm.x86.avx512.mask.cvtqq2pd.512(<8 x i64> %x0, <8 x double> %x1, i8 %x2, i32 4)
|
|
%res1 = call <8 x double> @llvm.x86.avx512.mask.cvtqq2pd.512(<8 x i64> %x0, <8 x double> %x1, i8 -1, i32 8)
|
|
%res2 = fadd <8 x double> %res, %res1
|
|
ret <8 x double> %res2
|
|
}
|
|
|
|
declare <8 x float> @llvm.x86.avx512.mask.cvtqq2ps.512(<8 x i64>, <8 x float>, i8, i32)
|
|
|
|
define <8 x float> @test_int_x86_avx512_mask_cvt_qq2ps_512(<8 x i64> %x0, <8 x float> %x1, i8 %x2) {
|
|
; X86-LABEL: test_int_x86_avx512_mask_cvt_qq2ps_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vcvtqq2ps %zmm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xfc,0x49,0x5b,0xc8]
|
|
; X86-NEXT: vcvtqq2ps {rn-sae}, %zmm0, %ymm0 # encoding: [0x62,0xf1,0xfc,0x18,0x5b,0xc0]
|
|
; X86-NEXT: vaddps %ymm0, %ymm1, %ymm0 # encoding: [0xc5,0xf4,0x58,0xc0]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_mask_cvt_qq2ps_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vcvtqq2ps %zmm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xfc,0x49,0x5b,0xc8]
|
|
; X64-NEXT: vcvtqq2ps {rn-sae}, %zmm0, %ymm0 # encoding: [0x62,0xf1,0xfc,0x18,0x5b,0xc0]
|
|
; X64-NEXT: vaddps %ymm0, %ymm1, %ymm0 # encoding: [0xc5,0xf4,0x58,0xc0]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
%res = call <8 x float> @llvm.x86.avx512.mask.cvtqq2ps.512(<8 x i64> %x0, <8 x float> %x1, i8 %x2, i32 4)
|
|
%res1 = call <8 x float> @llvm.x86.avx512.mask.cvtqq2ps.512(<8 x i64> %x0, <8 x float> %x1, i8 -1, i32 8)
|
|
%res2 = fadd <8 x float> %res, %res1
|
|
ret <8 x float> %res2
|
|
}
|
|
|
|
declare <8 x double> @llvm.x86.avx512.mask.cvtuqq2pd.512(<8 x i64>, <8 x double>, i8, i32)
|
|
|
|
define <8 x double> @test_int_x86_avx512_mask_cvt_uqq2pd_512(<8 x i64> %x0, <8 x double> %x1, i8 %x2) {
|
|
; X86-LABEL: test_int_x86_avx512_mask_cvt_uqq2pd_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vcvtuqq2pd %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0xfe,0x49,0x7a,0xc8]
|
|
; X86-NEXT: vcvtuqq2pd {rn-sae}, %zmm0, %zmm0 # encoding: [0x62,0xf1,0xfe,0x18,0x7a,0xc0]
|
|
; X86-NEXT: vaddpd %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0xf5,0x48,0x58,0xc0]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_mask_cvt_uqq2pd_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vcvtuqq2pd %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0xfe,0x49,0x7a,0xc8]
|
|
; X64-NEXT: vcvtuqq2pd {rn-sae}, %zmm0, %zmm0 # encoding: [0x62,0xf1,0xfe,0x18,0x7a,0xc0]
|
|
; X64-NEXT: vaddpd %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0xf5,0x48,0x58,0xc0]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
%res = call <8 x double> @llvm.x86.avx512.mask.cvtuqq2pd.512(<8 x i64> %x0, <8 x double> %x1, i8 %x2, i32 4)
|
|
%res1 = call <8 x double> @llvm.x86.avx512.mask.cvtuqq2pd.512(<8 x i64> %x0, <8 x double> %x1, i8 -1, i32 8)
|
|
%res2 = fadd <8 x double> %res, %res1
|
|
ret <8 x double> %res2
|
|
}
|
|
|
|
declare <8 x float> @llvm.x86.avx512.mask.cvtuqq2ps.512(<8 x i64>, <8 x float>, i8, i32)
|
|
|
|
define <8 x float> @test_int_x86_avx512_mask_cvt_uqq2ps_512(<8 x i64> %x0, <8 x float> %x1, i8 %x2) {
|
|
; X86-LABEL: test_int_x86_avx512_mask_cvt_uqq2ps_512:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
|
|
; X86-NEXT: vcvtuqq2ps %zmm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xff,0x49,0x7a,0xc8]
|
|
; X86-NEXT: vcvtuqq2ps {rn-sae}, %zmm0, %ymm0 # encoding: [0x62,0xf1,0xff,0x18,0x7a,0xc0]
|
|
; X86-NEXT: vaddps %ymm0, %ymm1, %ymm0 # encoding: [0xc5,0xf4,0x58,0xc0]
|
|
; X86-NEXT: retl # encoding: [0xc3]
|
|
;
|
|
; X64-LABEL: test_int_x86_avx512_mask_cvt_uqq2ps_512:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
|
|
; X64-NEXT: vcvtuqq2ps %zmm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xff,0x49,0x7a,0xc8]
|
|
; X64-NEXT: vcvtuqq2ps {rn-sae}, %zmm0, %ymm0 # encoding: [0x62,0xf1,0xff,0x18,0x7a,0xc0]
|
|
; X64-NEXT: vaddps %ymm0, %ymm1, %ymm0 # encoding: [0xc5,0xf4,0x58,0xc0]
|
|
; X64-NEXT: retq # encoding: [0xc3]
|
|
%res = call <8 x float> @llvm.x86.avx512.mask.cvtuqq2ps.512(<8 x i64> %x0, <8 x float> %x1, i8 %x2, i32 4)
|
|
%res1 = call <8 x float> @llvm.x86.avx512.mask.cvtuqq2ps.512(<8 x i64> %x0, <8 x float> %x1, i8 -1, i32 8)
|
|
%res2 = fadd <8 x float> %res, %res1
|
|
ret <8 x float> %res2
|
|
}
|