[llvm-commits] [llvm] r78567 - in /llvm/trunk: include/llvm/CodeGen/AsmPrinter.h include/llvm/Support/IOManip.h lib/CodeGen/AsmPrinter/AsmPrinter.cpp
Chris Lattner
clattner at apple.com
Mon Aug 17 22:47:57 PDT 2009
On Aug 11, 2009, at 12:59 PM, David A. Greene wrote:
> On Tuesday 11 August 2009 12:22, Chris Lattner wrote:
>
>> Hi David,
>>
>> Thanks for working on this, why not just put this under -asm-verbose?
>> The whole idea is that -asm-verbose is used when we don't care about
>> compile time, I don't see a need for yet-another flag. Also:
>
> Because the nightly tester runs with -asm-verbose and I got into
> enough
> trouble last time. :)
That's not a good reason, a better fix is to change the nightly
testers to not use -asm-verbose. :) I just did this. Please
eliminate exuberant asm.
>>> +++ llvm/trunk/include/llvm/Support/IOManip.h Mon Aug 10 11:38:07
>>> 2009
>>> @@ -0,0 +1,43 @@
>>> +//===----------------- IOManip.h - iostream manipulators ---------
>>> *- C++ -*===//
>>
>> Here we go again. *again*, please do not add this sort of thing. It
>> is not appreciated. As discussed several times before, I do not want
>> manipulators like this. Thanks.
>
> Ok, I didn't realize all manipulators are verboten.
They are when they cause more complexity than they prevent. Please
remove this.
>>> + if (loop->empty()) {
>>
>> I don't think that loops can be empty, they have to at least have a
>> header. I don't know why MachineLoop has an empty() predicate.
>
> I'm going to leave this here for now. When empty() goes away we can
> fix it.
Dan clarified that this is not dead it should stay.
>>> + << " Inside Loop BB" << getFunctionNumber() << "_"
>>> + << Header->getNumber() << " Depth " << CurLoop-
>>>
>>>> getLoopDepth();
>>
>> This output seems like it would be very verbose and redundant.
>
> It is, on purpose. It makes writing tools easier.
I thought the point was to make the .s files more human readable? IF
you use a simple algorithm, surely any tools can adapt.
Please eliminate IOManip.h and exuberant asm soon.
-Chris
More information about the llvm-commits
mailing list