diff options
| author | Sanjay Patel <spatel@rotateright.com> | 2019-02-22 15:47:45 +0000 |
|---|---|---|
| committer | Sanjay Patel <spatel@rotateright.com> | 2019-02-22 15:47:45 +0000 |
| commit | a9e289174a1c21698ea0c1f1a43f1f504e5e75ef (patch) | |
| tree | 103fb47d274aba00af6c22098d1e8fdaff08899b /llvm/test/CodeGen/X86 | |
| parent | 1baf7896cc2b48a15ba0590a5ece6143ef1ab552 (diff) | |
| download | bcm5719-llvm-a9e289174a1c21698ea0c1f1a43f1f504e5e75ef.tar.gz bcm5719-llvm-a9e289174a1c21698ea0c1f1a43f1f504e5e75ef.zip | |
[x86] allow narrowing of vector UINT_TO_FP
As discussed in:
D56864
D58197
Always use the narrow (128-bit) instruction when possible.
We already had the signed int version of this transform.
llvm-svn: 354675
Diffstat (limited to 'llvm/test/CodeGen/X86')
| -rw-r--r-- | llvm/test/CodeGen/X86/vec_int_to_fp-widen.ll | 8 | ||||
| -rw-r--r-- | llvm/test/CodeGen/X86/vec_int_to_fp.ll | 24 |
2 files changed, 8 insertions, 24 deletions
diff --git a/llvm/test/CodeGen/X86/vec_int_to_fp-widen.ll b/llvm/test/CodeGen/X86/vec_int_to_fp-widen.ll index 7441ce58027..41524f36d63 100644 --- a/llvm/test/CodeGen/X86/vec_int_to_fp-widen.ll +++ b/llvm/test/CodeGen/X86/vec_int_to_fp-widen.ll @@ -684,9 +684,7 @@ define <2 x double> @uitofp_4i32_to_2f64(<4 x i32> %a) { ; ; AVX512VL-LABEL: uitofp_4i32_to_2f64: ; AVX512VL: # %bb.0: -; AVX512VL-NEXT: vcvtudq2pd %xmm0, %ymm0 -; AVX512VL-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 -; AVX512VL-NEXT: vzeroupper +; AVX512VL-NEXT: vcvtudq2pd %xmm0, %xmm0 ; AVX512VL-NEXT: retq ; ; AVX512DQ-LABEL: uitofp_4i32_to_2f64: @@ -699,9 +697,7 @@ define <2 x double> @uitofp_4i32_to_2f64(<4 x i32> %a) { ; ; AVX512VLDQ-LABEL: uitofp_4i32_to_2f64: ; AVX512VLDQ: # %bb.0: -; AVX512VLDQ-NEXT: vcvtudq2pd %xmm0, %ymm0 -; AVX512VLDQ-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 -; AVX512VLDQ-NEXT: vzeroupper +; AVX512VLDQ-NEXT: vcvtudq2pd %xmm0, %xmm0 ; AVX512VLDQ-NEXT: retq %cvt = uitofp <4 x i32> %a to <4 x double> %shuf = shufflevector <4 x double> %cvt, <4 x double> undef, <2 x i32> <i32 0, i32 1> diff --git a/llvm/test/CodeGen/X86/vec_int_to_fp.ll b/llvm/test/CodeGen/X86/vec_int_to_fp.ll index ab3fd395d33..02b55f3ffc6 100644 --- a/llvm/test/CodeGen/X86/vec_int_to_fp.ll +++ b/llvm/test/CodeGen/X86/vec_int_to_fp.ll @@ -684,9 +684,7 @@ define <2 x double> @uitofp_4i32_to_2f64(<4 x i32> %a) { ; ; AVX512VL-LABEL: uitofp_4i32_to_2f64: ; AVX512VL: # %bb.0: -; AVX512VL-NEXT: vcvtudq2pd %xmm0, %ymm0 -; AVX512VL-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 -; AVX512VL-NEXT: vzeroupper +; AVX512VL-NEXT: vcvtudq2pd %xmm0, %xmm0 ; AVX512VL-NEXT: retq ; ; AVX512DQ-LABEL: uitofp_4i32_to_2f64: @@ -699,9 +697,7 @@ define <2 x double> @uitofp_4i32_to_2f64(<4 x i32> %a) { ; ; AVX512VLDQ-LABEL: uitofp_4i32_to_2f64: ; AVX512VLDQ: # %bb.0: -; AVX512VLDQ-NEXT: vcvtudq2pd %xmm0, %ymm0 -; AVX512VLDQ-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 -; AVX512VLDQ-NEXT: vzeroupper +; AVX512VLDQ-NEXT: vcvtudq2pd %xmm0, %xmm0 ; AVX512VLDQ-NEXT: retq %cvt = uitofp <4 x i32> %a to <4 x double> %shuf = shufflevector <4 x double> %cvt, <4 x double> undef, <2 x i32> <i32 0, i32 1> @@ -5699,9 +5695,7 @@ define double @extract0_uitofp_v4i32_f64(<4 x i32> %x) nounwind { ; ; AVX512VL-LABEL: extract0_uitofp_v4i32_f64: ; AVX512VL: # %bb.0: -; AVX512VL-NEXT: vcvtudq2pd %xmm0, %ymm0 -; AVX512VL-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 -; AVX512VL-NEXT: vzeroupper +; AVX512VL-NEXT: vcvtudq2pd %xmm0, %xmm0 ; AVX512VL-NEXT: retq ; ; AVX512DQ-LABEL: extract0_uitofp_v4i32_f64: @@ -5714,9 +5708,7 @@ define double @extract0_uitofp_v4i32_f64(<4 x i32> %x) nounwind { ; ; AVX512VLDQ-LABEL: extract0_uitofp_v4i32_f64: ; AVX512VLDQ: # %bb.0: -; AVX512VLDQ-NEXT: vcvtudq2pd %xmm0, %ymm0 -; AVX512VLDQ-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 -; AVX512VLDQ-NEXT: vzeroupper +; AVX512VLDQ-NEXT: vcvtudq2pd %xmm0, %xmm0 ; AVX512VLDQ-NEXT: retq %e = extractelement <4 x i32> %x, i32 0 %r = uitofp i32 %e to double @@ -5855,9 +5847,7 @@ define double @extract3_uitofp_v4i32_f64(<4 x i32> %x) nounwind { ; AVX512VL-LABEL: extract3_uitofp_v4i32_f64: ; AVX512VL: # %bb.0: ; AVX512VL-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,1,2,3] -; AVX512VL-NEXT: vcvtudq2pd %xmm0, %ymm0 -; AVX512VL-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 -; AVX512VL-NEXT: vzeroupper +; AVX512VL-NEXT: vcvtudq2pd %xmm0, %xmm0 ; AVX512VL-NEXT: retq ; ; AVX512DQ-LABEL: extract3_uitofp_v4i32_f64: @@ -5871,9 +5861,7 @@ define double @extract3_uitofp_v4i32_f64(<4 x i32> %x) nounwind { ; AVX512VLDQ-LABEL: extract3_uitofp_v4i32_f64: ; AVX512VLDQ: # %bb.0: ; AVX512VLDQ-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,1,2,3] -; AVX512VLDQ-NEXT: vcvtudq2pd %xmm0, %ymm0 -; AVX512VLDQ-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 -; AVX512VLDQ-NEXT: vzeroupper +; AVX512VLDQ-NEXT: vcvtudq2pd %xmm0, %xmm0 ; AVX512VLDQ-NEXT: retq %e = extractelement <4 x i32> %x, i32 3 %r = uitofp i32 %e to double |
