OSchip
/
llvm-project
mirror of https://github.com/THU-DSP-LAB/llvm-project.git
13
0
Fork 6
Code Issues Packages Projects Releases Wiki Activity

R600: Custom lower f32 = uint_to_fp i64 

llvm-svn: 193701
This commit is contained in:
Tom Stellard 2013-10-30 17:22:05 +00:00
parent 3c845b6f1b
commit c947d8ca64
3 changed files with 42 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
llvm/lib/Target/R600/AMDGPUISelLowering.cpp
View File

@ -134,6 +134,8 @@ AMDGPUTargetLowering::AMDGPUTargetLowering(TargetMachine &TM) :
setOperationAction(ISD::FNEG, MVT::v2f32, Expand);
setOperationAction(ISD::FNEG, MVT::v4f32, Expand);
setOperationAction(ISD::UINT_TO_FP, MVT::i64, Custom);
setOperationAction(ISD::MUL, MVT::i64, Expand);
setOperationAction(ISD::UDIV, MVT::i32, Expand);
@ -251,6 +253,7 @@ SDValue AMDGPUTargetLowering::LowerOperation(SDValue Op, SelectionDAG &DAG)
case ISD::INTRINSIC_WO_CHAIN: return LowerINTRINSIC_WO_CHAIN(Op, DAG);
case ISD::STORE: return LowerSTORE(Op, DAG);
case ISD::UDIVREM: return LowerUDIVREM(Op, DAG);
case ISD::UINT_TO_FP: return LowerUINT_TO_FP(Op, DAG);
}
return Op;
}
@ -672,6 +675,25 @@ SDValue AMDGPUTargetLowering::LowerUDIVREM(SDValue Op,
return DAG.getMergeValues(Ops, 2, DL);
}
SDValue AMDGPUTargetLowering::LowerUINT_TO_FP(SDValue Op,
SelectionDAG &DAG) const {
SDValue S0 = Op.getOperand(0);
SDLoc DL(Op);
if (Op.getValueType() != MVT::f32 || S0.getValueType() != MVT::i64)
return SDValue();
// f32 uint_to_fp i64
SDValue Lo = DAG.getNode(ISD::EXTRACT_ELEMENT, DL, MVT::i32, S0,
DAG.getConstant(0, MVT::i32));
SDValue FloatLo = DAG.getNode(ISD::UINT_TO_FP, DL, MVT::f32, Lo);
SDValue Hi = DAG.getNode(ISD::EXTRACT_ELEMENT, DL, MVT::i32, S0,
DAG.getConstant(1, MVT::i32));
SDValue FloatHi = DAG.getNode(ISD::UINT_TO_FP, DL, MVT::f32, Hi);
FloatHi = DAG.getNode(ISD::FMUL, DL, MVT::f32, FloatHi,
DAG.getConstantFP(4294967296.0f, MVT::f32)); // 2^32
return DAG.getNode(ISD::FADD, DL, MVT::f32, FloatLo, FloatHi);
}
//===----------------------------------------------------------------------===//
// Helper functions

1
llvm/lib/Target/R600/AMDGPUISelLowering.h
View File

@ -37,6 +37,7 @@ private:
/// \brief Split a vector store into multiple scalar stores.
/// \returns The resulting chain.
SDValue LowerUDIVREM(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerUINT_TO_FP(SDValue Op, SelectionDAG &DAG) const;
protected:

23
llvm/test/CodeGen/R600/uint_to_fp.ll
View File

@ -1,10 +1,10 @@
; RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s --check-prefix=R600-CHECK
; RUN: llc < %s -march=r600 -mcpu=SI -verify-machineinstrs | FileCheck %s --check-prefix=SI-CHECK
; R600-CHECK: @uint_to_fp_v2i32
; R600-CHECK-LABEL: @uint_to_fp_v2i32
; R600-CHECK-DAG: UINT_TO_FLT * T{{[0-9]+\.[XYZW]}}, KC0[2].W
; R600-CHECK-DAG: UINT_TO_FLT * T{{[0-9]+\.[XYZW]}}, KC0[3].X
; SI-CHECK: @uint_to_fp_v2i32
; SI-CHECK-LABEL: @uint_to_fp_v2i32
; SI-CHECK: V_CVT_F32_U32_e32
; SI-CHECK: V_CVT_F32_U32_e32
define void @uint_to_fp_v2i32(<2 x float> addrspace(1)* %out, <2 x i32> %in) {
@ -13,12 +13,12 @@ define void @uint_to_fp_v2i32(<2 x float> addrspace(1)* %out, <2 x i32> %in) {
ret void
}
; R600-CHECK: @uint_to_fp_v4i32
; R600-CHECK-LABEL: @uint_to_fp_v4i32
; R600-CHECK: UINT_TO_FLT * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
; R600-CHECK: UINT_TO_FLT * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
; R600-CHECK: UINT_TO_FLT * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
; R600-CHECK: UINT_TO_FLT * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
; SI-CHECK: @uint_to_fp_v4i32
; SI-CHECK-LABEL: @uint_to_fp_v4i32
; SI-CHECK: V_CVT_F32_U32_e32
; SI-CHECK: V_CVT_F32_U32_e32
; SI-CHECK: V_CVT_F32_U32_e32
@ -29,3 +29,18 @@ define void @uint_to_fp_v4i32(<4 x float> addrspace(1)* %out, <4 x i32> addrspac
store <4 x float> %result, <4 x float> addrspace(1)* %out
ret void
}
; R600-CHECK-LABEL: @uint_to_fp_i64_f32
; R600-CHECK: UINT_TO_FLT
; R600-CHECK: UINT_TO_FLT
; R600-CHECK: MULADD_IEEE
; SI-CHECK-LABEL: @uint_to_fp_i64_f32
; SI-CHECK: V_CVT_F32_U32_e32
; SI-CHECK: V_CVT_F32_U32_e32
; SI-CHECK: V_MAD_F32
define void @uint_to_fp_i64_f32(float addrspace(1)* %out, i64 %in) {
entry:
%0 = uitofp i64 %in to float
store float %0, float addrspace(1)* %out
ret void
}