[clang] ba87430 - [PowerPC] Replace subtract-from-zero float in version with fneg in PowerPC special fma compiler builtins

[Kevin P. Neal via cfe-commits]

Author: Andrew Wock
Date: 2020-04-03T14:59:33-04:00
New Revision: ba87430cadb2d5d0ee8e4b75101d7abcf6b321bf

URL: https://github.com/llvm/llvm-project/commit/ba87430cadb2d5d0ee8e4b75101d7abcf6b321bf
DIFF: https://github.com/llvm/llvm-project/commit/ba87430cadb2d5d0ee8e4b75101d7abcf6b321bf.diff

LOG: [PowerPC] Replace subtract-from-zero float in version with fneg in PowerPC special fma compiler builtins

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Originally, the code used the outdated method of subtracting floating point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.

Patch by: Drew Wock <drew.wock at sas.com>
Differential Revision: https://reviews.llvm.org/D76949

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

Modified: 
    clang/lib/CodeGen/CGBuiltin.cpp

Removed: 
    


################################################################################
diff  --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index 880fe0e271f5..3545ccda5d4d 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -13214,25 +13214,24 @@ Value *CodeGenFunction::EmitPPCBuiltinExpr(unsigned BuiltinID,
     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

diff  --git a/clang/test/CodeGen/builtins-ppc-fma.c b/clang/test/CodeGen/builtins-ppc-fma.c
new file mode 100644
index 000000000000..3a628f7613b5
--- /dev/null
+++ b/clang/test/CodeGen/builtins-ppc-fma.c
@@ -0,0 +1,43 @@
+// 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]]
+}