[AMDGPU] Always use rcp + mul with fast math

Regardless of relaxation options such as -cl-fast-relaxed-math
we are producing rather long code for fdiv via amdgcn_fdiv_fast
intrinsic. This intrinsic is used to replace fdiv with 2.5ulp
metadata and does not handle denormals, thus believed to be fast.

An fdiv instruction can also have fast math flag either by itself
or together with fpmath metadata. Clang used with a relaxation flag
always produces both metadata and fast flag:

%div = fdiv fast float %v, %0, !fpmath !12
!12 = !{float 2.500000e+00}

Current implementation ignores fast flag and favors metadata. An
instruction with just fast flag would be lowered to a fastest rcp +
mul, but that never happen on practice because of described mutual
clang and BE behavior.

This change allows an "fdiv fast" to be always lowered as rcp + mul.

Differential Revision: https://reviews.llvm.org/D34844

llvm-svn: 307308

2 files changed, 8 insertions, 8 deletions

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUCodeGenPrepare.cpp b/llvm/lib/Target/AMDGPU/AMDGPUCodeGenPrepare.cpp
index b312dbc8d14..31ee9206ae2 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUCodeGenPrepare.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUCodeGenPrepare.cpp
@@ -380,7 +380,9 @@ bool AMDGPUCodeGenPrepare::visitFDiv(BinaryOperator &FDiv) {
   FastMathFlags FMF = FPOp->getFastMathFlags();
   bool UnsafeDiv = HasUnsafeFPMath || FMF.unsafeAlgebra() ||
                                       FMF.allowReciprocal();
-  if (ST->hasFP32Denormals() && !UnsafeDiv)
+
+  // With UnsafeDiv node will be optimized to just rcp and mul.
+  if (ST->hasFP32Denormals() || UnsafeDiv)
     return false;
 
   IRBuilder<> Builder(FDiv.getParent(), std::next(FDiv.getIterator()), FPMath);
diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index b1d2f0a8b79..04c79b59308 100644
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -3736,7 +3736,9 @@ SDValue SITargetLowering::lowerFastUnsafeFDIV(SDValue Op,
   SDValue LHS = Op.getOperand(0);
   SDValue RHS = Op.getOperand(1);
   EVT VT = Op.getValueType();
-  bool Unsafe = DAG.getTarget().Options.UnsafeFPMath;
+  const SDNodeFlags Flags = Op->getFlags();
+  bool Unsafe = DAG.getTarget().Options.UnsafeFPMath ||
+                Flags.hasUnsafeAlgebra() || Flags.hasAllowReciprocal();
 
   if (!Unsafe && VT == MVT::f32 && Subtarget->hasFP32Denormals())
     return SDValue();
@@ -3771,15 +3773,11 @@ SDValue SITargetLowering::lowerFastUnsafeFDIV(SDValue Op,
     }
   }
 
-  const SDNodeFlags Flags = Op->getFlags();
-
-  if (Unsafe || Flags.hasAllowReciprocal()) {
+  if (Unsafe) {
     // Turn into multiply by the reciprocal.
     // x / y -> x * (1.0 / y)
-    SDNodeFlags NewFlags;
-    NewFlags.setUnsafeAlgebra(true);
     SDValue Recip = DAG.getNode(AMDGPUISD::RCP, SL, VT, RHS);
-    return DAG.getNode(ISD::FMUL, SL, VT, LHS, Recip, NewFlags);
+    return DAG.getNode(ISD::FMUL, SL, VT, LHS, Recip, Flags);
   }
 
   return SDValue();