On Mar 7, 2007, at 11:22 PM, Evan Cheng wrote:
>
> On Mar 7, 2007, at 9:15 PM, Chris Lattner wrote:
>
>>> + // Read-only data.
>>> + bool isIntFPLiteral = Type->isInteger() || Type-
>>> >isFloatingPoint();
>>> + if (C->ContainsRelocations() && Subtarget-
>>> >isTargetDarwin() &&
>>> + TM.getRelocationModel() != Reloc::Static)
>>
>> Please rearrange this to:
>> if (isdarwin && not static && containsrelocations)
>>
>> putting the expensive check last.
>>
>>> + SwitchToDataSection("\t.const_data\n");
>>> + else if (isIntFPLiteral && Size == 4 &&
>>> + TAI->getFourByteConstantSection())
>>> + SwitchToDataSection(TAI->getFourByteConstantSection
>>> (), I);
>>> + else if (isIntFPLiteral && Size == 8 &&
>>> + TAI->getEightByteConstantSection())
>>> + SwitchToDataSection(TAI->getEightByteConstantSection
>>> (), I);
>>> + else if (isIntFPLiteral && Size == 16 &&
>>> + TAI->getSixteenByteConstantSection())
>>> + SwitchToDataSection(TAI-
>>> >getSixteenByteConstantSection(), I);
>>
>> Do we really need to check isIntFPLiteral here? Why can't
>> something like {short,short} go in literal4 ?
>
> We don't. But ContainsRelocations has to be checked first. We can't
> use FourByteConstantSection, etc. if the constant has relocations.
Ah, ok, so instead of "isIntFPLiteral", you really mean "!contains
relocations"
-Chris
_______________________________________________
llvm-commits mailing list
llvm-commits@cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
