Re: [PATCH] Take register pressure into account for vec_construct/scalar_to_vec when the components are not loaded from memory.

2023-12-04 Thread Hongtao Liu
On Mon, Dec 4, 2023 at 3:51 PM Uros Bizjak wrote: > > On Mon, Dec 4, 2023 at 8:11 AM Hongtao Liu wrote: > > > > On Fri, Dec 1, 2023 at 10:26 PM Richard Biener > > wrote: > > > > > > On Fri, Dec 1, 2023 at 3:39 AM liuhongt wrote: > > > > > > > > > Hmm, I would suggest you put reg_needed into the

Re: [PATCH] Take register pressure into account for vec_construct/scalar_to_vec when the components are not loaded from memory.

2023-12-03 Thread Uros Bizjak
On Mon, Dec 4, 2023 at 8:11 AM Hongtao Liu wrote: > > On Fri, Dec 1, 2023 at 10:26 PM Richard Biener > wrote: > > > > On Fri, Dec 1, 2023 at 3:39 AM liuhongt wrote: > > > > > > > Hmm, I would suggest you put reg_needed into the class and accumulate > > > > over all vec_construct, with your patch

Re: [PATCH] Take register pressure into account for vec_construct/scalar_to_vec when the components are not loaded from memory.

2023-12-03 Thread Hongtao Liu
On Fri, Dec 1, 2023 at 10:26 PM Richard Biener wrote: > > On Fri, Dec 1, 2023 at 3:39 AM liuhongt wrote: > > > > > Hmm, I would suggest you put reg_needed into the class and accumulate > > > over all vec_construct, with your patch you pessimize a single v32qi > > > over two separate v16qi for exa

Re: [PATCH] Take register pressure into account for vec_construct/scalar_to_vec when the components are not loaded from memory.

2023-12-01 Thread Richard Biener
On Fri, Dec 1, 2023 at 3:39 AM liuhongt wrote: > > > Hmm, I would suggest you put reg_needed into the class and accumulate > > over all vec_construct, with your patch you pessimize a single v32qi > > over two separate v16qi for example. Also currently the whole block is > > gated with INTEGRAL_TY

[PATCH] Take register pressure into account for vec_construct/scalar_to_vec when the components are not loaded from memory.

2023-11-30 Thread liuhongt
> Hmm, I would suggest you put reg_needed into the class and accumulate > over all vec_construct, with your patch you pessimize a single v32qi > over two separate v16qi for example. Also currently the whole block is > gated with INTEGRAL_TYPE_P but register pressure would be also > a concern for f