Re: [PATCH] tree-optimizatio/118852 - wrong code with 502.gcc_r

2025-02-14 Thread Richard Biener
On Fri, 14 Feb 2025, Uros Bizjak wrote: > On Fri, Feb 14, 2025 at 3:10 PM Richard Biener wrote: > > > > 502.gcc_r when built with -fprofile-generate exposes a SLP discovery > > issue where an IV forced live due to early break is not properly > > discovered if its latch def is part of a different

Re: [PATCH] tree-optimizatio/118852 - wrong code with 502.gcc_r

2025-02-14 Thread Uros Bizjak
On Fri, Feb 14, 2025 at 3:10 PM Richard Biener wrote: > > 502.gcc_r when built with -fprofile-generate exposes a SLP discovery > issue where an IV forced live due to early break is not properly > discovered if its latch def is part of a different IVs SSA cycle. > To mitigate this we have to make s

[PATCH] tree-optimizatio/118852 - wrong code with 502.gcc_r

2025-02-14 Thread Richard Biener
502.gcc_r when built with -fprofile-generate exposes a SLP discovery issue where an IV forced live due to early break is not properly discovered if its latch def is part of a different IVs SSA cycle. To mitigate this we have to make sure to create an SLP instance for the original IV. Ideally we'd