[cfe-commits] r146430 - in /cfe/trunk: lib/Sema/SemaChecking.cpp test/SemaCXX/Inputs/array-bounds-system-header.h test/SemaCXX/array-bounds-system-header.cpp

[Matt Beaumont-Gay]

Done, r146562. Let me know what your performance tests say -- on my
desktop, it brought the syntax-only time for aes.c down to within
noise of r146429.

On Tue, Dec 13, 2011 at 23:10, Chandler Carruth <chandlerc at google.com> wrote:
> On Tue, Dec 13, 2011 at 10:53 PM, Chris Lattner <clattner at apple.com> wrote:
>>
>>
>> On Dec 13, 2011, at 3:08 PM, Matt Beaumont-Gay wrote:
>>
>> > Thanks for the heads up. I'm working on optimizing that codepath some
>> > more.
>>
>> In the meantime, would it be best to temporarily revert the patch so that
>> other performance regressions aren't lost in this noise?
>
>
> The test case is an extreme one. In the preprocessor it expands to an
> amazing number of array dereferences, many of them in trivially dead code.
> We end up producing warnings for a large number of them and then filter them
> out. Anyways, it seems very unlikely that much other code will tickle this
> performance problem. Regressions elsewhere should still show up.
>
> That said, I think Matt almost had the performance fixes done, so I'm not
> too worried about it?