Add -flimit-float-precision to enable some faster,

but less accurate (non-IEEE) code sequences for
certain math library functions.  Add the first of
several such expansions.  Don't worry, if you don't
turn it on it won't affect you.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55823 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Dale Johannesen 2008-09-05 01:48:15 +00:00
parent 4f22bb0676
commit 601d3c0145
2 changed files with 57 additions and 3 deletions

View File

@ -49,6 +49,17 @@
#include <algorithm>
using namespace llvm;
/// LimitFloatPrecision - Generate low-precision inline sequences for
/// some float libcalls (6, 8 or 12 bits).
static unsigned LimitFloatPrecision;
static cl::opt<unsigned, true>
LimitFPPrecision("limit-float-precision",
cl::desc("Generate low-precision inline sequences "
"for some float libcalls"),
cl::location(LimitFloatPrecision),
cl::init(0));
/// ComputeLinearIndex - Given an LLVM IR aggregate type and a sequence
/// insertvalue or extractvalue indices that identify a member, return
/// the linearized index of the start of the member.
@ -2742,6 +2753,49 @@ SelectionDAGLowering::implVisitBinaryAtomic(CallInst& I, ISD::NodeType Op) {
return 0;
}
/// visitExp2 - lower an exp2 intrinsic. Handles the special sequences
/// for limited-precision mode.
void
SelectionDAGLowering::visitExp2(CallInst &I) {
SDValue result;
if (getValue(I.getOperand(1)).getValueType() == MVT::f32 &&
LimitFloatPrecision>0 && LimitFloatPrecision<=12) {
SDValue operand = getValue(I.getOperand(1));
SDValue t0 = DAG.getNode(ISD::FP_TO_SINT, MVT::i32, operand);
SDValue t1 = DAG.getNode(ISD::SINT_TO_FP, MVT::f32, t0);
SDValue t2 = DAG.getNode(ISD::FSUB, MVT::f32, operand, t1);
SDValue t3 = DAG.getNode(ISD::SHL, MVT::i32, t0,
DAG.getConstant(23, MVT::i32));
SDValue t4 = DAG.getNode(ISD::FP_EXTEND, MVT::f64, t2);
SDValue t5 = DAG.getNode(ISD::FMUL, MVT::f64, t4,
DAG.getConstantFP(APFloat(
APInt(64, 0x3fb446bc609aa9cdULL)), MVT::f64));
SDValue t6 = DAG.getNode(ISD::FADD, MVT::f64, t5,
DAG.getConstantFP(APFloat(
APInt(64, 0x3fccb71e629f3a20ULL)), MVT::f64));
SDValue t7 = DAG.getNode(ISD::FMUL, MVT::f64, t6, t4);
SDValue t8 = DAG.getNode(ISD::FADD, MVT::f64, t7,
DAG.getConstantFP(APFloat(
APInt(64, 0x3fe64960db7bd5feULL)), MVT::f64));
SDValue t9 = DAG.getNode(ISD::FMUL, MVT::f64, t8, t4);
SDValue t10 = DAG.getNode(ISD::FADD, MVT::f64, t9,
DAG.getConstantFP(APFloat(
APInt(64, 0x3fefff1f934bd549ULL)), MVT::f64));
SDValue t11 = DAG.getNode(ISD::FP_ROUND, MVT::f32, t10,
DAG.getConstant(0, MVT::i32));
SDValue t12 = DAG.getNode(ISD::BIT_CONVERT, MVT::i32, t11);
SDValue t13 = DAG.getNode(ISD::ADD, MVT::i32, t12, t3);
result = DAG.getNode(ISD::BIT_CONVERT, MVT::f32, t13);
} else {
// No special expansion.
result = DAG.getNode(ISD::FEXP2,
getValue(I.getOperand(1)).getValueType(),
getValue(I.getOperand(1)));
}
setValue(&I, result);
}
/// visitIntrinsicCall - Lower the call to the specified intrinsic function. If
/// we want to emit this as a call to a named external function, return the name
/// otherwise lower it and return null.
@ -3037,9 +3091,7 @@ SelectionDAGLowering::visitIntrinsicCall(CallInst &I, unsigned Intrinsic) {
getValue(I.getOperand(1))));
return 0;
case Intrinsic::exp2:
setValue(&I, DAG.getNode(ISD::FEXP2,
getValue(I.getOperand(1)).getValueType(),
getValue(I.getOperand(1))));
visitExp2(I);
return 0;
case Intrinsic::pow:
setValue(&I, DAG.getNode(ISD::FPOW,

View File

@ -505,6 +505,8 @@ private:
const char *visitIntrinsicCall(CallInst &I, unsigned Intrinsic);
void visitTargetIntrinsic(CallInst &I, unsigned Intrinsic);
void visitExp2(CallInst &I);
void visitVAStart(CallInst &I);
void visitVAArg(VAArgInst &I);
void visitVAEnd(CallInst &I);