[NFC][CodeGen] Add unary fneg tests to X86/fma-scalar-combine.ll

llvm-svn: 362741

1 files changed, 97 insertions, 3 deletions

diff --git a/llvm/test/CodeGen/X86/fma-scalar-combine.ll b/llvm/test/CodeGen/X86/fma-scalar-combine.ll
index f8e823525b8..6d287f3b1c8 100644
--- a/llvm/test/CodeGen/X86/fma-scalar-combine.ll
+++ b/llvm/test/CodeGen/X86/fma-scalar-combine.ll
@@ -428,6 +428,31 @@ entry:
   ret <2 x double> %12
 }
 
+define <2 x double> @combine_scalar_mask_fnmsub_f32_unary_fneg(<2 x double> %a, i8 zeroext %k, <2 x double> %b, <2 x double> %c) {
+; CHECK-LABEL: combine_scalar_mask_fnmsub_f32_unary_fneg:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vfnmsub213ss %xmm2, %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0xaf,0xc2]
+; CHECK-NEXT:    # xmm0 = -(xmm1 * xmm0) - xmm2
+; CHECK-NEXT:    retq # encoding: [0xc3]
+entry:
+  %0 = bitcast <2 x double> %a to <4 x float>
+  %1 = bitcast <2 x double> %b to <4 x float>
+  %2 = bitcast <2 x double> %c to <4 x float>
+  %3 = extractelement <4 x float> %0, i64 0
+  %4 = extractelement <4 x float> %1, i64 0
+  %5 = extractelement <4 x float> %2, i64 0
+  %sub = fneg fast float %5
+  %6 = fmul fast float %4, %3
+  %7 = fsub fast float %sub, %6
+  %8 = bitcast i8 %k to <8 x i1>
+  %9 = extractelement <8 x i1> %8, i64 0
+  %10 = select i1 %9, float %7, float %3
+  %11 = insertelement <4 x float> %0, float %10, i64 0
+  %12 = bitcast <4 x float> %11 to <2 x double>
+  ret <2 x double> %12
+}
+
 define <2 x double> @combine_scalar_mask_fnmsub_f64(<2 x double> %a, i8 zeroext %k, <2 x double> %b, <2 x double> %c) {
 ; CHECK-LABEL: combine_scalar_mask_fnmsub_f64:
 ; CHECK:       # %bb.0: # %entry
@@ -449,6 +474,27 @@ entry:
   ret <2 x double> %8
 }
 
+define <2 x double> @combine_scalar_mask_fnmsub_f64_unary_fneg(<2 x double> %a, i8 zeroext %k, <2 x double> %b, <2 x double> %c) {
+; CHECK-LABEL: combine_scalar_mask_fnmsub_f64_unary_fneg:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vfnmsub213sd %xmm2, %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0xaf,0xc2]
+; CHECK-NEXT:    # xmm0 = -(xmm1 * xmm0) - xmm2
+; CHECK-NEXT:    retq # encoding: [0xc3]
+entry:
+  %0 = extractelement <2 x double> %a, i64 0
+  %1 = extractelement <2 x double> %b, i64 0
+  %2 = extractelement <2 x double> %c, i64 0
+  %sub = fneg fast double %2
+  %3 = fmul fast double %1, %0
+  %4 = fsub fast double %sub, %3
+  %5 = bitcast i8 %k to <8 x i1>
+  %6 = extractelement <8 x i1> %5, i64 0
+  %7 = select i1 %6, double %4, double %0
+  %8 = insertelement <2 x double> %a, double %7, i64 0
+  ret <2 x double> %8
+}
+
 define <2 x double> @combine_scalar_maskz_fnmsub_32(i8 zeroext %k, <2 x double> %a, <2 x double> %b, <2 x double> %c) {
 ; CHECK-LABEL: combine_scalar_maskz_fnmsub_32:
 ; CHECK:       # %bb.0: # %entry
@@ -474,8 +520,8 @@ entry:
   ret <2 x double> %12
 }
 
-define <2 x double> @combine_scalar_maskz_fnmsub_64(i8 zeroext %k, <2 x double> %a, <2 x double> %b, <2 x double> %c) {
-; CHECK-LABEL: combine_scalar_maskz_fnmsub_64:
+define <2 x double> @combine_scalar_maskz_fnmsub_64_unary_fneg(i8 zeroext %k, <2 x double> %a, <2 x double> %b, <2 x double> %c) {
+; CHECK-LABEL: combine_scalar_maskz_fnmsub_64_unary_fneg:
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
 ; CHECK-NEXT:    vfnmsub213sd %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0xaf,0xc2]
@@ -485,7 +531,7 @@ entry:
   %0 = extractelement <2 x double> %a, i64 0
   %1 = extractelement <2 x double> %b, i64 0
   %2 = extractelement <2 x double> %c, i64 0
-  %sub = fsub fast double -0.000000e+00, %2
+  %sub = fneg fast double %2
   %3 = fmul fast double %1, %0
   %4 = fsub fast double %sub, %3
   %5 = bitcast i8 %k to <8 x i1>
@@ -521,6 +567,32 @@ entry:
   ret <2 x double> %12
 }
 
+define <2 x double> @combine_scalar_mask3_fnmsub_32_unary_fneg(<2 x double> %a, <2 x double> %b, <2 x double> %c, i8 zeroext %k) {
+; CHECK-LABEL: combine_scalar_mask3_fnmsub_32_unary_fneg:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vfnmsub231ss %xmm0, %xmm1, %xmm2 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0xbf,0xd0]
+; CHECK-NEXT:    # xmm2 = -(xmm1 * xmm0) - xmm2
+; CHECK-NEXT:    vmovaps %xmm2, %xmm0 # encoding: [0xc5,0xf8,0x28,0xc2]
+; CHECK-NEXT:    retq # encoding: [0xc3]
+entry:
+  %0 = bitcast <2 x double> %a to <4 x float>
+  %1 = bitcast <2 x double> %b to <4 x float>
+  %2 = bitcast <2 x double> %c to <4 x float>
+  %3 = extractelement <4 x float> %0, i64 0
+  %4 = extractelement <4 x float> %1, i64 0
+  %5 = extractelement <4 x float> %2, i64 0
+  %sub = fneg fast float %5
+  %6 = fmul fast float %4, %3
+  %7 = fsub fast float %sub, %6
+  %8 = bitcast i8 %k to <8 x i1>
+  %9 = extractelement <8 x i1> %8, i64 0
+  %10 = select i1 %9, float %7, float %5
+  %11 = insertelement <4 x float> %2, float %10, i64 0
+  %12 = bitcast <4 x float> %11 to <2 x double>
+  ret <2 x double> %12
+}
+
 define <2 x double> @combine_scalar_mask3_fnmsub_64(<2 x double> %a, <2 x double> %b, <2 x double> %c, i8 zeroext %k) {
 ; CHECK-LABEL: combine_scalar_mask3_fnmsub_64:
 ; CHECK:       # %bb.0: # %entry
@@ -542,3 +614,25 @@ entry:
   %8 = insertelement <2 x double> %c, double %7, i64 0
   ret <2 x double> %8
 }
+
+define <2 x double> @combine_scalar_mask3_fnmsub_64_unary_fneg(<2 x double> %a, <2 x double> %b, <2 x double> %c, i8 zeroext %k) {
+; CHECK-LABEL: combine_scalar_mask3_fnmsub_64_unary_fneg:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vfnmsub231sd %xmm0, %xmm1, %xmm2 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0xbf,0xd0]
+; CHECK-NEXT:    # xmm2 = -(xmm1 * xmm0) - xmm2
+; CHECK-NEXT:    vmovapd %xmm2, %xmm0 # encoding: [0xc5,0xf9,0x28,0xc2]
+; CHECK-NEXT:    retq # encoding: [0xc3]
+entry:
+  %0 = extractelement <2 x double> %a, i64 0
+  %1 = extractelement <2 x double> %b, i64 0
+  %2 = extractelement <2 x double> %c, i64 0
+  %sub = fneg fast double %2
+  %3 = fmul fast double %1, %0
+  %4 = fsub fast double %sub, %3
+  %5 = bitcast i8 %k to <8 x i1>
+  %6 = extractelement <8 x i1> %5, i64 0
+  %7 = select i1 %6, double %4, double %2
+  %8 = insertelement <2 x double> %c, double %7, i64 0
+  ret <2 x double> %8
+}