On Mon, Nov 19, 2018 at 5:31 PM Timothy Arceri <tarc...@itsqueeze.com> wrote:
> On 20/11/18 2:46 am, Jason Ekstrand wrote: > > I also saw a case last week where the induction variable is <= > > imax(thing, 4) and it'd be nice to unroll those too. > > That is an interesting one. It means we could do a simple unroll for the > first 4 iterations i.e no if statements required, which is a bit of a > different use case from the new functions introduced in this series but > shouldn't bee too hard to do. > Actually, I think it was imin(thing, 4) so we would do a conditional unroll of 4 iterations. Still, it's potentially similar. Mostly that was to point out that more creative loop unrolling is a good thing. --Jason > > > > --Jason > > > > On Mon, Nov 19, 2018 at 12:18 AM Timothy Arceri <tarc...@itsqueeze.com > > <mailto:tarc...@itsqueeze.com>> wrote: > > > > This series add support for partial loop unrolling for loops with an > > unknown trip count. The new partial_unroll function allows the caller > > to specifiy how may times the loop should be unrolled and then the > > loop is inserted in the innermost continue branch of the unrolled > loop. > > > > For now we only do partial unrolling for loops where we can guess the > > iteration count based on array access that uses an induction variable > > as its index (see patch 7). > > > > Patches 1-6 are tidy-ups/refactors. > > > > Patch 7 adds a method for guessing the trip count for the loop. > > > > Patch 8 adds partial unrolling support and also support for removing > > redundant load/stores from the remaining loop when considering if > > we would be accessing an array out bounds. > > > > > > _______________________________________________ > > mesa-dev mailing list > > mesa-dev@lists.freedesktop.org <mailto: > mesa-dev@lists.freedesktop.org> > > https://lists.freedesktop.org/mailman/listinfo/mesa-dev > > >
_______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev
