diff options
Diffstat (limited to 'llvm/test/CodeGen/X86/avx512-cvt.ll')
| -rw-r--r-- | llvm/test/CodeGen/X86/avx512-cvt.ll | 200 |
1 files changed, 44 insertions, 156 deletions
diff --git a/llvm/test/CodeGen/X86/avx512-cvt.ll b/llvm/test/CodeGen/X86/avx512-cvt.ll index 4c089ac379c..e99cdaf1ce9 100644 --- a/llvm/test/CodeGen/X86/avx512-cvt.ll +++ b/llvm/test/CodeGen/X86/avx512-cvt.ll @@ -484,32 +484,12 @@ define <4 x float> @ulto4f32(<4 x i64> %a) { define <8 x double> @ulto8f64(<8 x i64> %a) { ; NODQ-LABEL: ulto8f64: ; NODQ: # %bb.0: -; NODQ-NEXT: vextracti32x4 $3, %zmm0, %xmm1 -; NODQ-NEXT: vpextrq $1, %xmm1, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm2, %xmm2 -; NODQ-NEXT: vmovq %xmm1, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm3, %xmm1 -; NODQ-NEXT: vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0] -; NODQ-NEXT: vextracti32x4 $2, %zmm0, %xmm2 -; NODQ-NEXT: vpextrq $1, %xmm2, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm3, %xmm3 -; NODQ-NEXT: vmovq %xmm2, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm4, %xmm2 -; NODQ-NEXT: vmovlhps {{.*#+}} xmm2 = xmm2[0],xmm3[0] -; NODQ-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm1 -; NODQ-NEXT: vextracti128 $1, %ymm0, %xmm2 -; NODQ-NEXT: vpextrq $1, %xmm2, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm4, %xmm3 -; NODQ-NEXT: vmovq %xmm2, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm4, %xmm2 -; NODQ-NEXT: vmovlhps {{.*#+}} xmm2 = xmm2[0],xmm3[0] -; NODQ-NEXT: vpextrq $1, %xmm0, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm4, %xmm3 -; NODQ-NEXT: vmovq %xmm0, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm4, %xmm0 -; NODQ-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm3[0] -; NODQ-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; NODQ-NEXT: vinsertf64x4 $1, %ymm1, %zmm0, %zmm0 +; NODQ-NEXT: vpandq {{.*}}(%rip){1to8}, %zmm0, %zmm1 +; NODQ-NEXT: vporq {{.*}}(%rip){1to8}, %zmm1, %zmm1 +; NODQ-NEXT: vpsrlq $32, %zmm0, %zmm0 +; NODQ-NEXT: vporq {{.*}}(%rip){1to8}, %zmm0, %zmm0 +; NODQ-NEXT: vsubpd {{.*}}(%rip){1to8}, %zmm0, %zmm0 +; NODQ-NEXT: vaddpd %zmm0, %zmm1, %zmm0 ; NODQ-NEXT: retq ; ; VLDQ-LABEL: ulto8f64: @@ -524,32 +504,12 @@ define <8 x double> @ulto8f64(<8 x i64> %a) { ; ; KNL_WIDEN-LABEL: ulto8f64: ; KNL_WIDEN: # %bb.0: -; KNL_WIDEN-NEXT: vextracti32x4 $3, %zmm0, %xmm1 -; KNL_WIDEN-NEXT: vpextrq $1, %xmm1, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm2, %xmm2 -; KNL_WIDEN-NEXT: vmovq %xmm1, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm3, %xmm1 -; KNL_WIDEN-NEXT: vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0] -; KNL_WIDEN-NEXT: vextracti32x4 $2, %zmm0, %xmm2 -; KNL_WIDEN-NEXT: vpextrq $1, %xmm2, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm3, %xmm3 -; KNL_WIDEN-NEXT: vmovq %xmm2, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm4, %xmm2 -; KNL_WIDEN-NEXT: vmovlhps {{.*#+}} xmm2 = xmm2[0],xmm3[0] -; KNL_WIDEN-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm1 -; KNL_WIDEN-NEXT: vextracti128 $1, %ymm0, %xmm2 -; KNL_WIDEN-NEXT: vpextrq $1, %xmm2, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm4, %xmm3 -; KNL_WIDEN-NEXT: vmovq %xmm2, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm4, %xmm2 -; KNL_WIDEN-NEXT: vmovlhps {{.*#+}} xmm2 = xmm2[0],xmm3[0] -; KNL_WIDEN-NEXT: vpextrq $1, %xmm0, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm4, %xmm3 -; KNL_WIDEN-NEXT: vmovq %xmm0, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm4, %xmm0 -; KNL_WIDEN-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm3[0] -; KNL_WIDEN-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; KNL_WIDEN-NEXT: vinsertf64x4 $1, %ymm1, %zmm0, %zmm0 +; KNL_WIDEN-NEXT: vpandq {{.*}}(%rip){1to8}, %zmm0, %zmm1 +; KNL_WIDEN-NEXT: vporq {{.*}}(%rip){1to8}, %zmm1, %zmm1 +; KNL_WIDEN-NEXT: vpsrlq $32, %zmm0, %zmm0 +; KNL_WIDEN-NEXT: vporq {{.*}}(%rip){1to8}, %zmm0, %zmm0 +; KNL_WIDEN-NEXT: vsubpd {{.*}}(%rip){1to8}, %zmm0, %zmm0 +; KNL_WIDEN-NEXT: vaddpd %zmm0, %zmm1, %zmm0 ; KNL_WIDEN-NEXT: retq %b = uitofp <8 x i64> %a to <8 x double> ret <8 x double> %b @@ -558,58 +518,22 @@ define <8 x double> @ulto8f64(<8 x i64> %a) { define <16 x double> @ulto16f64(<16 x i64> %a) { ; NODQ-LABEL: ulto16f64: ; NODQ: # %bb.0: -; NODQ-NEXT: vextracti32x4 $3, %zmm0, %xmm2 -; NODQ-NEXT: vpextrq $1, %xmm2, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm3, %xmm3 -; NODQ-NEXT: vmovq %xmm2, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm4, %xmm2 -; NODQ-NEXT: vmovlhps {{.*#+}} xmm2 = xmm2[0],xmm3[0] -; NODQ-NEXT: vextracti32x4 $2, %zmm0, %xmm3 -; NODQ-NEXT: vpextrq $1, %xmm3, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm4, %xmm4 -; NODQ-NEXT: vmovq %xmm3, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm3 -; NODQ-NEXT: vmovlhps {{.*#+}} xmm3 = xmm3[0],xmm4[0] -; NODQ-NEXT: vinsertf128 $1, %xmm2, %ymm3, %ymm2 -; NODQ-NEXT: vextracti128 $1, %ymm0, %xmm3 -; NODQ-NEXT: vpextrq $1, %xmm3, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm4 -; NODQ-NEXT: vmovq %xmm3, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm3 -; NODQ-NEXT: vmovlhps {{.*#+}} xmm3 = xmm3[0],xmm4[0] -; NODQ-NEXT: vpextrq $1, %xmm0, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm4 -; NODQ-NEXT: vmovq %xmm0, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm0 -; NODQ-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm4[0] -; NODQ-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0 -; NODQ-NEXT: vinsertf64x4 $1, %ymm2, %zmm0, %zmm0 -; NODQ-NEXT: vextracti32x4 $3, %zmm1, %xmm2 -; NODQ-NEXT: vpextrq $1, %xmm2, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm3 -; NODQ-NEXT: vmovq %xmm2, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm2 -; NODQ-NEXT: vmovlhps {{.*#+}} xmm2 = xmm2[0],xmm3[0] -; NODQ-NEXT: vextracti32x4 $2, %zmm1, %xmm3 -; NODQ-NEXT: vpextrq $1, %xmm3, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm4 -; NODQ-NEXT: vmovq %xmm3, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm3 -; NODQ-NEXT: vmovlhps {{.*#+}} xmm3 = xmm3[0],xmm4[0] -; NODQ-NEXT: vinsertf128 $1, %xmm2, %ymm3, %ymm2 -; NODQ-NEXT: vextracti128 $1, %ymm1, %xmm3 -; NODQ-NEXT: vpextrq $1, %xmm3, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm4 -; NODQ-NEXT: vmovq %xmm3, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm3 -; NODQ-NEXT: vmovlhps {{.*#+}} xmm3 = xmm3[0],xmm4[0] -; NODQ-NEXT: vpextrq $1, %xmm1, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm4 -; NODQ-NEXT: vmovq %xmm1, %rax -; NODQ-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm1 -; NODQ-NEXT: vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm4[0] -; NODQ-NEXT: vinsertf128 $1, %xmm3, %ymm1, %ymm1 -; NODQ-NEXT: vinsertf64x4 $1, %ymm2, %zmm1, %zmm1 +; NODQ-NEXT: vpbroadcastq {{.*#+}} zmm2 = [4294967295,4294967295,4294967295,4294967295,4294967295,4294967295,4294967295,4294967295] +; NODQ-NEXT: vpandq %zmm2, %zmm0, %zmm3 +; NODQ-NEXT: vpbroadcastq {{.*#+}} zmm4 = [4841369599423283200,4841369599423283200,4841369599423283200,4841369599423283200,4841369599423283200,4841369599423283200,4841369599423283200,4841369599423283200] +; NODQ-NEXT: vporq %zmm4, %zmm3, %zmm3 +; NODQ-NEXT: vpsrlq $32, %zmm0, %zmm0 +; NODQ-NEXT: vpbroadcastq {{.*#+}} zmm5 = [4985484787499139072,4985484787499139072,4985484787499139072,4985484787499139072,4985484787499139072,4985484787499139072,4985484787499139072,4985484787499139072] +; NODQ-NEXT: vporq %zmm5, %zmm0, %zmm0 +; NODQ-NEXT: vbroadcastsd {{.*#+}} zmm6 = [1.9342813118337666E+25,1.9342813118337666E+25,1.9342813118337666E+25,1.9342813118337666E+25,1.9342813118337666E+25,1.9342813118337666E+25,1.9342813118337666E+25,1.9342813118337666E+25] +; NODQ-NEXT: vsubpd %zmm6, %zmm0, %zmm0 +; NODQ-NEXT: vaddpd %zmm0, %zmm3, %zmm0 +; NODQ-NEXT: vpandq %zmm2, %zmm1, %zmm2 +; NODQ-NEXT: vporq %zmm4, %zmm2, %zmm2 +; NODQ-NEXT: vpsrlq $32, %zmm1, %zmm1 +; NODQ-NEXT: vporq %zmm5, %zmm1, %zmm1 +; NODQ-NEXT: vsubpd %zmm6, %zmm1, %zmm1 +; NODQ-NEXT: vaddpd %zmm1, %zmm2, %zmm1 ; NODQ-NEXT: retq ; ; VLDQ-LABEL: ulto16f64: @@ -626,58 +550,22 @@ define <16 x double> @ulto16f64(<16 x i64> %a) { ; ; KNL_WIDEN-LABEL: ulto16f64: ; KNL_WIDEN: # %bb.0: -; KNL_WIDEN-NEXT: vextracti32x4 $3, %zmm0, %xmm2 -; KNL_WIDEN-NEXT: vpextrq $1, %xmm2, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm3, %xmm3 -; KNL_WIDEN-NEXT: vmovq %xmm2, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm4, %xmm2 -; KNL_WIDEN-NEXT: vmovlhps {{.*#+}} xmm2 = xmm2[0],xmm3[0] -; KNL_WIDEN-NEXT: vextracti32x4 $2, %zmm0, %xmm3 -; KNL_WIDEN-NEXT: vpextrq $1, %xmm3, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm4, %xmm4 -; KNL_WIDEN-NEXT: vmovq %xmm3, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm3 -; KNL_WIDEN-NEXT: vmovlhps {{.*#+}} xmm3 = xmm3[0],xmm4[0] -; KNL_WIDEN-NEXT: vinsertf128 $1, %xmm2, %ymm3, %ymm2 -; KNL_WIDEN-NEXT: vextracti128 $1, %ymm0, %xmm3 -; KNL_WIDEN-NEXT: vpextrq $1, %xmm3, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm4 -; KNL_WIDEN-NEXT: vmovq %xmm3, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm3 -; KNL_WIDEN-NEXT: vmovlhps {{.*#+}} xmm3 = xmm3[0],xmm4[0] -; KNL_WIDEN-NEXT: vpextrq $1, %xmm0, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm4 -; KNL_WIDEN-NEXT: vmovq %xmm0, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm0 -; KNL_WIDEN-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm4[0] -; KNL_WIDEN-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0 -; KNL_WIDEN-NEXT: vinsertf64x4 $1, %ymm2, %zmm0, %zmm0 -; KNL_WIDEN-NEXT: vextracti32x4 $3, %zmm1, %xmm2 -; KNL_WIDEN-NEXT: vpextrq $1, %xmm2, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm3 -; KNL_WIDEN-NEXT: vmovq %xmm2, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm2 -; KNL_WIDEN-NEXT: vmovlhps {{.*#+}} xmm2 = xmm2[0],xmm3[0] -; KNL_WIDEN-NEXT: vextracti32x4 $2, %zmm1, %xmm3 -; KNL_WIDEN-NEXT: vpextrq $1, %xmm3, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm4 -; KNL_WIDEN-NEXT: vmovq %xmm3, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm3 -; KNL_WIDEN-NEXT: vmovlhps {{.*#+}} xmm3 = xmm3[0],xmm4[0] -; KNL_WIDEN-NEXT: vinsertf128 $1, %xmm2, %ymm3, %ymm2 -; KNL_WIDEN-NEXT: vextracti128 $1, %ymm1, %xmm3 -; KNL_WIDEN-NEXT: vpextrq $1, %xmm3, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm4 -; KNL_WIDEN-NEXT: vmovq %xmm3, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm3 -; KNL_WIDEN-NEXT: vmovlhps {{.*#+}} xmm3 = xmm3[0],xmm4[0] -; KNL_WIDEN-NEXT: vpextrq $1, %xmm1, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm4 -; KNL_WIDEN-NEXT: vmovq %xmm1, %rax -; KNL_WIDEN-NEXT: vcvtusi2sdq %rax, %xmm5, %xmm1 -; KNL_WIDEN-NEXT: vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm4[0] -; KNL_WIDEN-NEXT: vinsertf128 $1, %xmm3, %ymm1, %ymm1 -; KNL_WIDEN-NEXT: vinsertf64x4 $1, %ymm2, %zmm1, %zmm1 +; KNL_WIDEN-NEXT: vpbroadcastq {{.*#+}} zmm2 = [4294967295,4294967295,4294967295,4294967295,4294967295,4294967295,4294967295,4294967295] +; KNL_WIDEN-NEXT: vpandq %zmm2, %zmm0, %zmm3 +; KNL_WIDEN-NEXT: vpbroadcastq {{.*#+}} zmm4 = [4841369599423283200,4841369599423283200,4841369599423283200,4841369599423283200,4841369599423283200,4841369599423283200,4841369599423283200,4841369599423283200] +; KNL_WIDEN-NEXT: vporq %zmm4, %zmm3, %zmm3 +; KNL_WIDEN-NEXT: vpsrlq $32, %zmm0, %zmm0 +; KNL_WIDEN-NEXT: vpbroadcastq {{.*#+}} zmm5 = [4985484787499139072,4985484787499139072,4985484787499139072,4985484787499139072,4985484787499139072,4985484787499139072,4985484787499139072,4985484787499139072] +; KNL_WIDEN-NEXT: vporq %zmm5, %zmm0, %zmm0 +; KNL_WIDEN-NEXT: vbroadcastsd {{.*#+}} zmm6 = [1.9342813118337666E+25,1.9342813118337666E+25,1.9342813118337666E+25,1.9342813118337666E+25,1.9342813118337666E+25,1.9342813118337666E+25,1.9342813118337666E+25,1.9342813118337666E+25] +; KNL_WIDEN-NEXT: vsubpd %zmm6, %zmm0, %zmm0 +; KNL_WIDEN-NEXT: vaddpd %zmm0, %zmm3, %zmm0 +; KNL_WIDEN-NEXT: vpandq %zmm2, %zmm1, %zmm2 +; KNL_WIDEN-NEXT: vporq %zmm4, %zmm2, %zmm2 +; KNL_WIDEN-NEXT: vpsrlq $32, %zmm1, %zmm1 +; KNL_WIDEN-NEXT: vporq %zmm5, %zmm1, %zmm1 +; KNL_WIDEN-NEXT: vsubpd %zmm6, %zmm1, %zmm1 +; KNL_WIDEN-NEXT: vaddpd %zmm1, %zmm2, %zmm1 ; KNL_WIDEN-NEXT: retq %b = uitofp <16 x i64> %a to <16 x double> ret <16 x double> %b |
