diff options
| author | Craig Topper <craig.topper@gmail.com> | 2017-03-29 06:55:28 +0000 |
|---|---|---|
| committer | Craig Topper <craig.topper@gmail.com> | 2017-03-29 06:55:28 +0000 |
| commit | d284606327a58fd4684a63b126079938ff4513cc (patch) | |
| tree | b95dbfdbe21b923ee8805237ec26b8baec7fed2f /llvm/test/CodeGen/X86/fma-fneg-combine.ll | |
| parent | 60c24878748987d193c91a863a3dc822ee664afc (diff) | |
| download | bcm5719-llvm-d284606327a58fd4684a63b126079938ff4513cc.tar.gz bcm5719-llvm-d284606327a58fd4684a63b126079938ff4513cc.zip | |
[AVX-512] Remove explicit KMOVWrk/KMOVWKr instructions from patterns where we can just use COPY_TO_REGCLASS instead.
This will result in a KMOVW or KMOVD being emitted during register allocation. And in at least some cases this might allow the register coalescer to remove the copy all together.
llvm-svn: 298984
Diffstat (limited to 'llvm/test/CodeGen/X86/fma-fneg-combine.ll')
| -rw-r--r-- | llvm/test/CodeGen/X86/fma-fneg-combine.ll | 24 |
1 files changed, 16 insertions, 8 deletions
diff --git a/llvm/test/CodeGen/X86/fma-fneg-combine.ll b/llvm/test/CodeGen/X86/fma-fneg-combine.ll index 6cdb3c3d6c6..bb332f7282a 100644 --- a/llvm/test/CodeGen/X86/fma-fneg-combine.ll +++ b/llvm/test/CodeGen/X86/fma-fneg-combine.ll @@ -142,7 +142,7 @@ define <4 x float> @test11(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 ze ; SKX: # BB#0: # %entry ; SKX-NEXT: vxorps {{.*}}(%rip){1to4}, %xmm2, %xmm0 ; SKX-NEXT: andl $1, %edi -; SKX-NEXT: kmovw %edi, %k1 +; SKX-NEXT: kmovd %edi, %k1 ; SKX-NEXT: vfmadd231ss %xmm1, %xmm1, %xmm0 {%k1} ; SKX-NEXT: retq ; @@ -183,13 +183,21 @@ entry: } define <2 x double> @test13(<2 x double> %a, <2 x double> %b, <2 x double> %c, i8 %mask) { -; CHECK-LABEL: test13: -; CHECK: # BB#0: # %entry -; CHECK-NEXT: vxorpd {{.*}}(%rip), %xmm0, %xmm0 -; CHECK-NEXT: andl $1, %edi -; CHECK-NEXT: kmovw %edi, %k1 -; CHECK-NEXT: vfmadd213sd %xmm2, %xmm1, %xmm0 {%k1} -; CHECK-NEXT: retq +; SKX-LABEL: test13: +; SKX: # BB#0: # %entry +; SKX-NEXT: vxorpd {{.*}}(%rip), %xmm0, %xmm0 +; SKX-NEXT: andl $1, %edi +; SKX-NEXT: kmovd %edi, %k1 +; SKX-NEXT: vfmadd213sd %xmm2, %xmm1, %xmm0 {%k1} +; SKX-NEXT: retq +; +; KNL-LABEL: test13: +; KNL: # BB#0: # %entry +; KNL-NEXT: vxorpd {{.*}}(%rip), %xmm0, %xmm0 +; KNL-NEXT: andl $1, %edi +; KNL-NEXT: kmovw %edi, %k1 +; KNL-NEXT: vfmadd213sd %xmm2, %xmm1, %xmm0 {%k1} +; KNL-NEXT: retq entry: %sub.i = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %a %0 = tail call <2 x double> @llvm.x86.avx512.mask.vfmadd.sd(<2 x double> %sub.i, <2 x double> %b, <2 x double> %c, i8 %mask, i32 4) |
