Replace subtract-from-zero float in version with fneg in PowerPC special fma compiler builtins

[ajwock]

Summary

This patch adds a test for the PowerPC fma compiler builtins, some variations of which negate inputs and outputs. The code to generate IR for these builtins was untested before this patch.

Originally, the code used the outdated method of subtracting floating point values from -0.0 as floating point negation. This patch remedies that.

[cameron.mcinally]

LGTM, assuming Harbormaster whines are addressed.

[ajwock]

Rebased for harbormaster.

clang/lib/CodeGen/CGBuiltin.cpp

Value *CodeGenFunction::EmitPPCBuiltinExpr(unsigned BuiltinID,
   case PPC::BI__builtin_vsx_xvmsubadp:
   case PPC::BI__builtin_vsx_xvmsubasp:
   case PPC::BI__builtin_vsx_xvnmsubadp:
   case PPC::BI__builtin_vsx_xvnmsubasp: {
     llvm::Type *ResultType = ConvertType(E->getType());
     Value *X = EmitScalarExpr(E->getArg(0));
     Value *Y = EmitScalarExpr(E->getArg(1));
     Value *Z = EmitScalarExpr(E->getArg(2));
-    Value *Zero = llvm::ConstantFP::getZeroValueForNegation(ResultType);
     llvm::Function *F = CGM.getIntrinsic(Intrinsic::fma, ResultType);
     switch (BuiltinID) {
       case PPC::BI__builtin_vsx_xvmaddadp:
       case PPC::BI__builtin_vsx_xvmaddasp:
         return Builder.CreateCall(F, {X, Y, Z});
+
       case PPC::BI__builtin_vsx_xvnmaddadp:
       case PPC::BI__builtin_vsx_xvnmaddasp:
-        return Builder.CreateFSub(Zero,
-                                  Builder.CreateCall(F, {X, Y, Z}), "sub");
+        return Builder.CreateFNeg(Builder.CreateCall(F, {X, Y, Z}), "neg");
+
       case PPC::BI__builtin_vsx_xvmsubadp:
       case PPC::BI__builtin_vsx_xvmsubasp:
-        return Builder.CreateCall(F,
-                                  {X, Y, Builder.CreateFSub(Zero, Z, "sub")});
+        return Builder.CreateCall(F, {X, Y, Builder.CreateFNeg(Z, "neg")});
+
       case PPC::BI__builtin_vsx_xvnmsubadp:
       case PPC::BI__builtin_vsx_xvnmsubasp:
-        Value *FsubRes =
-          Builder.CreateCall(F, {X, Y, Builder.CreateFSub(Zero, Z, "sub")});
-        return Builder.CreateFSub(Zero, FsubRes, "sub");
+        return Builder.CreateFNeg(
+            Builder.CreateCall(F, {X, Y, Builder.CreateFNeg(Z, "neg")}), "neg");
     }
     llvm_unreachable("Unknown FMA operation");
     return nullptr; // Suppress no-return warning
   }
 
   case PPC::BI__builtin_vsx_insertword: {
     llvm::Function *F = CGM.getIntrinsic(Intrinsic::ppc_vsx_xxinsertw);
 

clang/test/CodeGen/builtins-ppc-fma.c

+// RUN: %clang_cc1 -triple powerpc64le-gnu-linux \
+// RUN: -target-feature +altivec -Wall -Wno-unused -Werror -emit-llvm %s -o - | FileCheck      \
+// RUN: %s
+
+typedef __attribute__((vector_size(4 * sizeof(float)))) float vec_float;
+typedef __attribute__((vector_size(2 * sizeof(double)))) double vec_double;
+
+volatile vec_double vd;
+volatile vec_float vf;
+
+void test_fma(void) {
+  vf = __builtin_vsx_xvmaddasp(vf, vf, vf);
+  // CHECK: @llvm.fma.v4f32(<4 x float> %{{.*}}, <4 x float> %{{.*}}, <4 x float> %{{.*}})
+
+  vd = __builtin_vsx_xvmaddadp(vd, vd, vd);
+  // CHECK: @llvm.fma.v2f64(<2 x double> %{{.*}}, <2 x double> %{{.*}}, <2 x double> %{{.*}})
+
+  vf = __builtin_vsx_xvnmaddasp(vf, vf, vf);
+  // CHECK: [[RESULT:%[^ ]+]] = call <4 x float> @llvm.fma.v4f32(<4 x float> %{{.*}}, <4 x float> %{{.*}}, <4 x float> %{{.*}})
+  // CHECK: fneg <4 x float> [[RESULT]]
+
+  vd = __builtin_vsx_xvnmaddadp(vd, vd, vd);
+  // CHECK: [[RESULT:%[^ ]+]] = call <2 x double> @llvm.fma.v2f64(<2 x double> %{{.*}}, <2 x double> %{{.*}}, <2 x double> %{{.*}})
+  // CHECK: fneg <2 x double> [[RESULT]]
+
+  vf = __builtin_vsx_xvmsubasp(vf, vf, vf);
+  // CHECK: [[RESULT:%[^ ]+]] fneg <4 x float> %{{.*}}
+  // CHECK: @llvm.fma.v4f32(<4 x float> %{{.*}}, <4 x float> %{{.*}}, <4 x float> [[RESULT]])
+
+  vd = __builtin_vsx_xvmsubadp(vd, vd, vd);
+  // CHECK: fneg <2 x double> [[RESULT]]
+  // CHECK: [[RESULT:%[^ ]+]] = call <2 x double> @llvm.fma.v2f64(<2 x double> %{{.*}}, <2 x double> %{{.*}}, <2 x double> %{{.*}})
+
+  vf = __builtin_vsx_xvnmsubasp(vf, vf, vf);
+  // CHECK: [[RESULT:%[^ ]+]] = fneg <4 x float> %{{.*}}
+  // CHECK: [[RESULT2:%[^ ]+]] = call <4 x float> @llvm.fma.v2f64(<4 x float> %{{.*}}, <4 x float> %{{.*}}, <4 x float> [[RESULT]])
+  // CHECK: fneg <4 x float> [[RESULT2]]
+
+  vd = __builtin_vsx_xvnmsubadp(vd, vd, vd);
+  // CHECK: [[RESULT:%[^ ]+]] = fneg <2 x double> %{{.*}}
+  // CHECK: [[RESULT2:%[^ ]+]] = call <2 x double> @llvm.fma.v2f64(<2 x double> %{{.*}}, <2 x double> %{{.*}}, <2 x double> [[RESULT]])
+  // CHECK: fneg <2 x double> [[RESULT2]]
+}