RPCS3/llvm
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm.git synced 2024-12-26 22:26:16 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

R600/SI: FMA is faster than fmul and fadd for f64

Browse Source 
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188136 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Niels Ole Salscheider 2013-08-10 10:38:54 +00:00
parent 4ab92e4d4b
commit 6e4dbcd115
3 changed files with 50 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
lib/Target/R600/SIISelLowering.cpp
View File

@ -338,6 +338,24 @@ MVT SITargetLowering::getScalarShiftAmountTy(EVT VT) const {
return MVT::i32;
}
bool SITargetLowering::isFMAFasterThanFMulAndFAdd(EVT VT) const {
VT = VT.getScalarType();
if (!VT.isSimple())
return false;
switch (VT.getSimpleVT().SimpleTy) {
case MVT::f32:
return false; /* There is V_MAD_F32 for f32 */
case MVT::f64:
return true;
default:
break;
}
return false;
}
//===----------------------------------------------------------------------===//
// Custom DAG Lowering Operations
//===----------------------------------------------------------------------===//

1
lib/Target/R600/SIISelLowering.h
View File

@ -55,6 +55,7 @@ public:
MachineBasicBlock * BB) const;
virtual EVT getSetCCResultType(LLVMContext &Context, EVT VT) const;
virtual MVT getScalarShiftAmountTy(EVT VT) const;
virtual bool isFMAFasterThanFMulAndFAdd(EVT VT) const;
virtual SDValue LowerOperation(SDValue Op, SelectionDAG &DAG) const;
virtual SDValue PerformDAGCombine(SDNode *N, DAGCombinerInfo &DCI) const;
virtual SDNode *PostISelFolding(MachineSDNode *N, SelectionDAG &DAG) const;

31
test/CodeGen/R600/fmuladd.ll Normal file
View File

@ -0,0 +1,31 @@
; RUN: llc < %s -march=r600 -mcpu=SI | FileCheck %s
; CHECK: @fmuladd_f32
; CHECK: V_MAD_F32 {{VGPR[0-9]+, VGPR[0-9]+, VGPR[0-9]+, VGPR[0-9]+}}
define void @fmuladd_f32(float addrspace(1)* %out, float addrspace(1)* %in1,
float addrspace(1)* %in2, float addrspace(1)* %in3) {
%r0 = load float addrspace(1)* %in1
%r1 = load float addrspace(1)* %in2
%r2 = load float addrspace(1)* %in3
%r3 = tail call float @llvm.fmuladd.f32(float %r0, float %r1, float %r2)
store float %r3, float addrspace(1)* %out
ret void
}
declare float @llvm.fmuladd.f32(float, float, float)
; CHECK: @fmuladd_f64
; CHECK: V_FMA_F64 {{VGPR[0-9]+_VGPR[0-9]+, VGPR[0-9]+_VGPR[0-9]+, VGPR[0-9]+_VGPR[0-9]+, VGPR[0-9]+_VGPR[0-9]+}}
define void @fmuladd_f64(double addrspace(1)* %out, double addrspace(1)* %in1,
double addrspace(1)* %in2, double addrspace(1)* %in3) {
%r0 = load double addrspace(1)* %in1
%r1 = load double addrspace(1)* %in2
%r2 = load double addrspace(1)* %in3
%r3 = tail call double @llvm.fmuladd.f64(double %r0, double %r1, double %r2)
store double %r3, double addrspace(1)* %out
ret void
}
declare double @llvm.fmuladd.f64(double, double, double)