Changed the handling of variadic intrinsics a bit.

Removed the -fp80 option and made real be 80bit floats on X86, this is what the D spec really says it should be and fixes a bunch of issues.
Changed the handling of parameter attributes to a bit more generalized approach.
Added sext/zext attributes for byte/short/ubyte/ushort parameters, fixes #60 .
Parameter attribs now properly set for intrinsic calls if necessary.
Made the tango.math.Math patch less intrusive.
Fixed/added some mini tests.
This commit is contained in:
Tomas Lindquist Olsen
2008-08-01 17:59:58 +02:00
parent 858dd3e4fa
commit 9b45fc5533
32 changed files with 417 additions and 531 deletions

View File

@@ -35,6 +35,25 @@ bool DtoIsReturnedInArg(Type* type)
return (t == Tstruct || t == Tarray || t == Tdelegate || t == Tsarray || typ->iscomplex());
}
unsigned DtoShouldExtend(Type* type)
{
type = type->toBasetype();
if (type->isintegral())
{
switch(type->ty)
{
case Tint8:
case Tint16:
return llvm::ParamAttr::SExt;
case Tuns8:
case Tuns16:
return llvm::ParamAttr::ZExt;
}
}
return llvm::ParamAttr::None;
}
Type* DtoDType(Type* t)
{
if (t->ty == Ttypedef) {
@@ -79,7 +98,10 @@ const LLType* DtoType(Type* t)
return LLType::DoubleTy;
case Tfloat80:
case Timaginary80:
return (global.params.useFP80) ? LLType::X86_FP80Ty : LLType::DoubleTy;
if (global.params.cpu == ARCHx86)
return LLType::X86_FP80Ty;
else
return LLType::DoubleTy;
// complex
case Tcomplex32:
@@ -485,11 +507,9 @@ llvm::ConstantInt* DtoConstUbyte(unsigned char i)
llvm::ConstantFP* DtoConstFP(Type* t, long double value)
{
TY ty = DtoDType(t)->ty;
if (ty == Tfloat32 || ty == Timaginary32)
return llvm::ConstantFP::get(llvm::APFloat(float(value)));
else if (ty == Tfloat64 || ty == Timaginary64 || ty == Tfloat80 || ty == Timaginary80)
return llvm::ConstantFP::get(llvm::APFloat(double(value)));
const LLType* llty = DtoType(t);
assert(llty->isFloatingPoint());
return LLConstantFP::get(llty, value);
}
//////////////////////////////////////////////////////////////////////////////////////////