On Jul 25, 2014, at 8:13 AM, Aaron Ballman <[email protected]> wrote:
> On Thu, Jul 24, 2014 at 6:48 PM, Tyler Nowicki <[email protected]> wrote:
>> Hi Aaron,
>>
>> Thanks for the review!
>>
>> There were a number of conflicts with the work on pragma unroll and
>> nounroll. To accommodate these directives I changed the Enabled argument of
>> the loop hint attribute to a state argument. The state argument can be
>> default, enable, or disable. Attributes like ‘#pragma unroll’ that don’t
>> have any arguments use the default state.
>
> I would prefer this to be a separate change if at all possible (as a
> predecessor to this patch) as it seems to be fairly substantive --
> there's about a 17kb difference between this patch and the last one.
> ;-)
The change I was describing was between the previous version of my patch and
the current. Right now in trunk there is no ‘Enabled’ argument. Currently an
integer is used to keep track of both the state and value. Ideally we should
have a union of both the state and constant expression. That will require a
little tablegen magic.
I agree the patch is a bit bigger. If you think it is required I can try to
split it up a bit.
Tyler
_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
