replica
/
hanchenye-llvm-project
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[X86][AVX512] Add support for VPERMPD/VPERMQ masked shuffle comments 

llvm-svn: 274469
This commit is contained in:
Simon Pilgrim 2016-07-03 18:40:24 +00:00
parent a0d73835b2
commit 68ea80649b
3 changed files with 24 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
llvm/lib/Target/X86/InstPrinter/X86InstComments.cpp
View File

@ -132,6 +132,14 @@ using namespace llvm;
CASE_AVX512_INS_COMMON(Inst, Z256, src##i) \
CASE_AVX_INS_COMMON(Inst, Y, src##i)
#define CASE_MASK_VPERM(Inst, src) \
CASE_MASK_INS_COMMON(Inst, Z, src##i) \
CASE_MASK_INS_COMMON(Inst, Z256, src##i)
#define CASE_MASKZ_VPERM(Inst, src) \
CASE_MASKZ_INS_COMMON(Inst, Z, src##i) \
CASE_MASKZ_INS_COMMON(Inst, Z256, src##i)
#define CASE_VSHUF(Inst, src) \
CASE_AVX512_INS_COMMON(SHUFF##Inst, Z, r##src##i) \
CASE_AVX512_INS_COMMON(SHUFI##Inst, Z, r##src##i) \
@ -261,6 +269,10 @@ static std::string getMaskName(const MCInst *MI, const char *DestName,
CASE_MASKZ_VPERMILPI(PSHUFHW, r)
CASE_MASKZ_VPERMILPI(PSHUFLW, m)
CASE_MASKZ_VPERMILPI(PSHUFLW, r)
CASE_MASKZ_VPERM(PERMPD, m)
CASE_MASKZ_VPERM(PERMPD, r)
CASE_MASKZ_VPERM(PERMQ, m)
CASE_MASKZ_VPERM(PERMQ, r)
CASE_MASKZ_VSHUF(64X2, m)
CASE_MASKZ_VSHUF(64X2, r)
CASE_MASKZ_VSHUF(32X4, m)
@ -322,6 +334,10 @@ static std::string getMaskName(const MCInst *MI, const char *DestName,
CASE_MASK_VPERMILPI(PSHUFHW, r)
CASE_MASK_VPERMILPI(PSHUFLW, m)
CASE_MASK_VPERMILPI(PSHUFLW, r)
CASE_MASK_VPERM(PERMPD, m)
CASE_MASK_VPERM(PERMPD, r)
CASE_MASK_VPERM(PERMQ, m)
CASE_MASK_VPERM(PERMQ, r)
CASE_MASK_VSHUF(64X2, m)
CASE_MASK_VSHUF(64X2, r)
CASE_MASK_VSHUF(32X4, m)

8
llvm/test/CodeGen/X86/avx512-intrinsics.ll
View File

@ -6352,8 +6352,8 @@ define <8 x double>@test_int_x86_avx512_mask_perm_df_512(<8 x double> %x0, i32 %
; CHECK-LABEL: test_int_x86_avx512_mask_perm_df_512:
; CHECK: ## BB#0:
; CHECK-NEXT: kmovw %esi, %k1
; CHECK-NEXT: vpermpd {{.*#+}} zmm1 = zmm0[3,0,0,0,7,4,4,4]
; CHECK-NEXT: vpermpd {{.*#+}} zmm2 = zmm0[3,0,0,0,7,4,4,4]
; CHECK-NEXT: vpermpd {{.*#+}} zmm1 {%k1} = zmm0[3,0,0,0,7,4,4,4]
; CHECK-NEXT: vpermpd {{.*#+}} zmm2 {%k1} {z} = zmm0[3,0,0,0,7,4,4,4]
; CHECK-NEXT: vpermpd {{.*#+}} zmm0 = zmm0[3,0,0,0,7,4,4,4]
; CHECK-NEXT: vaddpd %zmm2, %zmm1, %zmm1
; CHECK-NEXT: vaddpd %zmm0, %zmm1, %zmm0
@ -6372,8 +6372,8 @@ define <8 x i64>@test_int_x86_avx512_mask_perm_di_512(<8 x i64> %x0, i32 %x1, <8
; CHECK-LABEL: test_int_x86_avx512_mask_perm_di_512:
; CHECK: ## BB#0:
; CHECK-NEXT: kmovw %esi, %k1
; CHECK-NEXT: vpermq {{.*#+}} zmm1 = zmm0[3,0,0,0,7,4,4,4]
; CHECK-NEXT: vpermq {{.*#+}} zmm2 = zmm0[3,0,0,0,7,4,4,4]
; CHECK-NEXT: vpermq {{.*#+}} zmm1 {%k1} = zmm0[3,0,0,0,7,4,4,4]
; CHECK-NEXT: vpermq {{.*#+}} zmm2 {%k1} {z} = zmm0[3,0,0,0,7,4,4,4]
; CHECK-NEXT: vpermq {{.*#+}} zmm0 = zmm0[3,0,0,0,7,4,4,4]
; CHECK-NEXT: vpaddq %zmm2, %zmm1, %zmm1
; CHECK-NEXT: vpaddq %zmm0, %zmm1, %zmm0

8
llvm/test/CodeGen/X86/avx512vl-intrinsics.ll
View File

@ -8076,9 +8076,9 @@ define <4 x double>@test_int_x86_avx512_mask_perm_df_256(<4 x double> %x0, i32 %
; CHECK: ## BB#0:
; CHECK-NEXT: kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT: vpermpd $3, %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x01,0xc8,0x03]
; CHECK-NEXT: ## ymm1 = ymm0[3,0,0,0]
; CHECK-NEXT: ## ymm1 {%k1} = ymm0[3,0,0,0]
; CHECK-NEXT: vpermpd $3, %ymm0, %ymm2 {%k1} {z} ## encoding: [0x62,0xf3,0xfd,0xa9,0x01,0xd0,0x03]
; CHECK-NEXT: ## ymm2 = ymm0[3,0,0,0]
; CHECK-NEXT: ## ymm2 {%k1} {z} = ymm0[3,0,0,0]
; CHECK-NEXT: vpermpd $3, %ymm0, %ymm0 ## encoding: [0x62,0xf3,0xfd,0x28,0x01,0xc0,0x03]
; CHECK-NEXT: ## ymm0 = ymm0[3,0,0,0]
; CHECK-NEXT: vaddpd %ymm2, %ymm1, %ymm1 ## encoding: [0x62,0xf1,0xf5,0x28,0x58,0xca]
@ -8099,9 +8099,9 @@ define <4 x i64>@test_int_x86_avx512_mask_perm_di_256(<4 x i64> %x0, i32 %x1, <4
; CHECK: ## BB#0:
; CHECK-NEXT: kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT: vpermq $3, %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x00,0xc8,0x03]
; CHECK-NEXT: ## ymm1 = ymm0[3,0,0,0]
; CHECK-NEXT: ## ymm1 {%k1} = ymm0[3,0,0,0]
; CHECK-NEXT: vpermq $3, %ymm0, %ymm2 {%k1} {z} ## encoding: [0x62,0xf3,0xfd,0xa9,0x00,0xd0,0x03]
; CHECK-NEXT: ## ymm2 = ymm0[3,0,0,0]
; CHECK-NEXT: ## ymm2 {%k1} {z} = ymm0[3,0,0,0]
; CHECK-NEXT: vpermq $3, %ymm0, %ymm0 ## encoding: [0x62,0xf3,0xfd,0x28,0x00,0xc0,0x03]
; CHECK-NEXT: ## ymm0 = ymm0[3,0,0,0]
; CHECK-NEXT: vpaddq %ymm2, %ymm1, %ymm1 ## encoding: [0x62,0xf1,0xf5,0x28,0xd4,0xca]