release_35 patches for unroll pragma

[Eric Christopher]

Seems like the feature missed the boat unless there's a particularly
compelling reason. We usually do time based releases and not feature
based.

Any reason that we need them in 3.5? Correctness?

-eric

On Mon, Aug 4, 2014 at 10:01 AM, Mark Heffernan <meheff at google.com> wrote:
> Ping?  Hal or Aaron?  Is it too late to get these in?  I was on vacation
> last week so wasn't able to follow up earlier on this.
>
> Thanks,
> Mark
>
>
> On Tue, Jul 29, 2014 at 12:35 PM, Mark Heffernan <meheff at google.com> wrote:
>>
>> Hi,
>>
>> Below are a handful of patches I submitted last week for the loop unroll
>> pragma which would be nice to get into 3.5.  The bug fix and metadata
>> renaming ones I'd particularly like to see get in.  The "#pragma nounroll"
>> is more of a nice-to-have.  All of the patches should only affect code which
>> uses one of the forms of the loop unroll pragmas.
>>
>> Rename metadata.  In unroll pragma syntax and loop hint metadata, change
>> "enable" forms to a new form using the string "full":
>> r213771 (clang)
>> r213772 (llvm)
>> r213775 (fix broken test caused by r213771)
>>
>> Add support for #pragma nounroll :
>> r213885
>>
>> Bug fixes:
>> r213789 : Do not add unroll disable metadata after unrolling pass for
>> loops with #pragma clang loop unroll(full).
>> r213900 : Fix bug where some loops with loop unroll_count(N) would
>> mistakenly not be unrolled.
>>
>> Let me know if these are reasonable.
>>
>> Thanks!
>> Mark
>
>
>
> _______________________________________________
> cfe-commits mailing list
> cfe-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>