This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Testing patch to make FDO profile data tolerate more source changes
Just to clarify is gcc.dg/tree-prof/wcoverage-mismatch.c an example of
what you are talking about? If so, I really like this is approach.
Thanks for the feedback!
Neil
On Mon, Mar 30, 2009 at 4:41 PM, Andrew Pinski <pinskia@gmail.com> wrote:
> On Mon, Mar 30, 2009 at 2:50 PM, Janis Johnson <janis187@us.ibm.com> wrote:
>> On Mon, 2009-03-30 at 13:53 -0700, Neil Vachharajani wrote:
>>> Hello,
>>>
>>> I'm working on a patch which makes FDO profile data more robust to
>>> source changes. ?Before I sent the patch to the list for review, I was
>>> wondering how to write a test for this situation. ?In particular, what
>>> I would like to have is two source files which are similar but not the
>>> same. ?The first should be compiled and run to generate profile data,
>>> and the generated .gcda file should be used to compile the second
>>> file. ?I wasn't quite sure how to do this within the DejaGNU framework
>>> and was wondering if anyone had some pointers.
>>
>> Existing FDO tests use lib/profopt.exp via g++.dg/bprob/bprob.exp
>> and gcc.misc-tests/bprob.exp. ?You can copy those .exp files to
>> variants that use different versions of the source file for the
>> two compiles. ?Let me know if you have further questions.
>
>
> Or even better just use a define which allows you to select the before
> and after code. ?I think there might be already testcases in the
> testsuite that do that.
>
> Thanks,
> Andrew Pinski
>
--
Neil Vachharajani
Google
650-214-1804