Re: [PATCH v6] forwprop: Try to blend two isomorphic VEC_PERM sequences

2024-11-21 Thread Christoph Müllner
On Thu, Nov 21, 2024 at 1:34 PM Sam James wrote: > > The default on trunk is --enable-checking=yes,extra (when gcc/DEV-PHASE > contains "experimental"), otherwise it's --enable-checking=release. > > I personally do most testing with --enable-checking=yes,rtl,extra but > you can do less than that i

Re: [PATCH v6] forwprop: Try to blend two isomorphic VEC_PERM sequences

2024-11-21 Thread Sam James
The default on trunk is --enable-checking=yes,extra (when gcc/DEV-PHASE contains "experimental"), otherwise it's --enable-checking=release. I personally do most testing with --enable-checking=yes,rtl,extra but you can do less than that if you want to quickly get results. The minimum for testing p

Re: [PATCH v6] forwprop: Try to blend two isomorphic VEC_PERM sequences

2024-11-21 Thread Christoph Müllner
On Thu, Nov 21, 2024 at 1:17 PM Richard Biener wrote: > > On Thu, 21 Nov 2024, Christoph Müllner wrote: > > > This extends forwprop by yet another VEC_PERM optimization: > > It attempts to blend two isomorphic vector sequences by using the > > redundancy in the lane utilization in these sequences.

Re: [PATCH v6] forwprop: Try to blend two isomorphic VEC_PERM sequences

2024-11-21 Thread Richard Biener
On Thu, 21 Nov 2024, Christoph Müllner wrote: > This extends forwprop by yet another VEC_PERM optimization: > It attempts to blend two isomorphic vector sequences by using the > redundancy in the lane utilization in these sequences. > This redundancy in lane utilization comes from the way how spec

[PATCH v6] forwprop: Try to blend two isomorphic VEC_PERM sequences

2024-11-20 Thread Christoph Müllner
This extends forwprop by yet another VEC_PERM optimization: It attempts to blend two isomorphic vector sequences by using the redundancy in the lane utilization in these sequences. This redundancy in lane utilization comes from the way how specific scalar statements end up vectorized: two VEC_PERMs