RE: [patch] cilkplus array notation for C (clean, independent patchset, take 1)

["Iyer, Balaji V" <balaji dot v dot iyer at intel dot com>]

> -----Original Message-----
> From: Aldy Hernandez [mailto:aldyh@redhat.com]
> Sent: Thursday, March 21, 2013 9:09 AM
> To: Iyer, Balaji V
> Cc: Jakub Jelinek; Jeff Law; Joseph S. Myers; gcc-patches
> Subject: Re: [patch] cilkplus array notation for C (clean, independent patchset,
> take 1)
> 
> On 03/21/13 08:06, Iyer, Balaji V wrote:
> >>
> >> Balaji, please check the corresponding .sum files before and after
> >> your patch to make sure that the same number of tests are being
> >> tested.  We have a nifty script in contrib/compare_tests for this task.
> >
> > That's how I verify it. (I grep for the ^FAIL in trunk and the applied branch and
> make sure the output files are the same by going through it). Did I miss
> anything?
> 
> If you're using compare_tests, you should be fine.  But just grepping for FAIL
> won't do because there are tests that could have passed before, but are no
> longer being tested, so they don't show up as a fail.  I believe compare_tests
> complains with "tests that used to pass but have disappeared" (or something
> similar).

I first look at the expected passes, expected fails, etc. If those numbers match up, then I do what I said above . Otherwise I look at things that have failed that shouldn't and/or passed that shouldn't have (this, should almost never happen because all the Cilk plus related code are all enclosed between inside an if (flag_enable_cilkplus) statement).


> 
> >
> >>
> >> And as Jakub has said, check (with and) without parallelization.
> >>
> >
> > Yes, I am doing that also for the patch I am submitting.
> 
> Thank you.