[NFC][CodeGen] Add unary FNeg tests to X86/avx512vl-intrinsics-fast-isel.ll

Patch 3 of 3 for X86/avx512vl-intrinsics-fast-isel.ll

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@363200 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/test/CodeGen/X86/avx512vl-intrinsics-fast-isel.ll b/test/CodeGen/X86/avx512vl-intrinsics-fast-isel.ll
index 24479988b589c3f335316f45b59b94cf0529f5fd..b51402c6981149fcc3dba92089ea2d3549713c97 100644 (file)
--- a/test/CodeGen/X86/avx512vl-intrinsics-fast-isel.ll
+++ b/test/CodeGen/X86/avx512vl-intrinsics-fast-isel.ll
@@ -7142,6 +7142,31 @@ entry:
   ret <4 x float> %2
 }
 
+define <4 x float> @test_mm_mask3_fnmsub_ps_unary_fneg(<4 x float> %__A, <4 x float> %__B, <4 x float> %__C, i8 zeroext %__U) {
+; X86-LABEL: test_mm_mask3_fnmsub_ps_unary_fneg:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
+; X86-NEXT:    kmovw %eax, %k1
+; X86-NEXT:    vfnmsub231ps {{.*#+}} xmm2 = -(xmm0 * xmm1) - xmm2
+; X86-NEXT:    vmovaps %xmm2, %xmm0
+; X86-NEXT:    retl
+;
+; X64-LABEL: test_mm_mask3_fnmsub_ps_unary_fneg:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    kmovw %edi, %k1
+; X64-NEXT:    vfnmsub231ps {{.*#+}} xmm2 = -(xmm0 * xmm1) - xmm2
+; X64-NEXT:    vmovaps %xmm2, %xmm0
+; X64-NEXT:    retq
+entry:
+  %neg.i = fneg <4 x float> %__B
+  %neg1.i = fneg <4 x float> %__C
+  %0 = tail call <4 x float> @llvm.fma.v4f32(<4 x float> %__A, <4 x float> %neg.i, <4 x float> %neg1.i) #9
+  %1 = bitcast i8 %__U to <8 x i1>
+  %extract.i = shufflevector <8 x i1> %1, <8 x i1> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %2 = select <4 x i1> %extract.i, <4 x float> %0, <4 x float> %__C
+  ret <4 x float> %2
+}
+
 define <8 x float> @test_mm256_mask_fnmsub_ps(<8 x float> %__A, i8 zeroext %__U, <8 x float> %__B, <8 x float> %__C) {
 ; X86-LABEL: test_mm256_mask_fnmsub_ps:
 ; X86:       # %bb.0: # %entry
@@ -7164,6 +7189,28 @@ entry:
   ret <8 x float> %2
 }
 
+define <8 x float> @test_mm256_mask_fnmsub_ps_unary_fneg(<8 x float> %__A, i8 zeroext %__U, <8 x float> %__B, <8 x float> %__C) {
+; X86-LABEL: test_mm256_mask_fnmsub_ps_unary_fneg:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
+; X86-NEXT:    kmovw %eax, %k1
+; X86-NEXT:    vfnmsub132ps {{.*#+}} ymm0 = -(ymm0 * ymm1) - ymm2
+; X86-NEXT:    retl
+;
+; X64-LABEL: test_mm256_mask_fnmsub_ps_unary_fneg:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    kmovw %edi, %k1
+; X64-NEXT:    vfnmsub132ps {{.*#+}} ymm0 = -(ymm0 * ymm1) - ymm2
+; X64-NEXT:    retq
+entry:
+  %neg.i = fneg <8 x float> %__B
+  %neg1.i = fneg <8 x float> %__C
+  %0 = tail call <8 x float> @llvm.fma.v8f32(<8 x float> %__A, <8 x float> %neg.i, <8 x float> %neg1.i) #9
+  %1 = bitcast i8 %__U to <8 x i1>
+  %2 = select <8 x i1> %1, <8 x float> %0, <8 x float> %__A
+  ret <8 x float> %2
+}
+
 define <8 x float> @test_mm256_mask3_fnmsub_ps(<8 x float> %__A, <8 x float> %__B, <8 x float> %__C, i8 zeroext %__U) {
 ; X86-LABEL: test_mm256_mask3_fnmsub_ps:
 ; X86:       # %bb.0: # %entry
@@ -7188,6 +7235,30 @@ entry:
   ret <8 x float> %2
 }
 
+define <8 x float> @test_mm256_mask3_fnmsub_ps_unary_fneg(<8 x float> %__A, <8 x float> %__B, <8 x float> %__C, i8 zeroext %__U) {
+; X86-LABEL: test_mm256_mask3_fnmsub_ps_unary_fneg:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
+; X86-NEXT:    kmovw %eax, %k1
+; X86-NEXT:    vfnmsub231ps {{.*#+}} ymm2 = -(ymm0 * ymm1) - ymm2
+; X86-NEXT:    vmovaps %ymm2, %ymm0
+; X86-NEXT:    retl
+;
+; X64-LABEL: test_mm256_mask3_fnmsub_ps_unary_fneg:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    kmovw %edi, %k1
+; X64-NEXT:    vfnmsub231ps {{.*#+}} ymm2 = -(ymm0 * ymm1) - ymm2
+; X64-NEXT:    vmovaps %ymm2, %ymm0
+; X64-NEXT:    retq
+entry:
+  %neg.i = fneg <8 x float> %__B
+  %neg1.i = fneg <8 x float> %__C
+  %0 = tail call <8 x float> @llvm.fma.v8f32(<8 x float> %__A, <8 x float> %neg.i, <8 x float> %neg1.i) #9
+  %1 = bitcast i8 %__U to <8 x i1>
+  %2 = select <8 x i1> %1, <8 x float> %0, <8 x float> %__C
+  ret <8 x float> %2
+}
+
 define <2 x double> @test_mm_mask_expandloadu_pd(<2 x double> %__W, i8 zeroext %__U, i8* readonly %__P) {
 ; X86-LABEL: test_mm_mask_expandloadu_pd:
 ; X86:       # %bb.0: # %entry