Re: [PATCH] hppa64: Improve hppa_rtx_costs for DImode shifts by constants.

Hi Roger, On 2020-09-07 3:59 p.m., Roger Sayle wrote: > 2020-09-07 Roger Sayle > > gcc/ChangeLog > * config/pa/pa.c (hppa_rtx_costs) [ASHIFT, ASHIFTRT, LSHIFTRT]: > Provide accurate costs for DImode shifts of integer constants. I committed this change.  However, it doesn't fix PR mi

Re: [PATCH] hppa64: Improve hppa_rtx_costs for DImode shifts by constants.

On 2020-09-07 3:59 p.m., Roger Sayle wrote: > Please let me know what you think. > > > 2020-09-07 Roger Sayle > > gcc/ChangeLog > * config/pa/pa.c (hppa_rtx_costs) [ASHIFT, ASHIFTRT, LSHIFTRT]: > Provide accurate costs for DImode shifts of integer constants. Will do. I'm still testi

[PATCH] hppa64: Improve hppa_rtx_costs for DImode shifts by constants.

Hi Dave, Here's another patch for you to try on hppa64. I've no idea whether it helps PR middle-end/87256 on hppa64 or makes things worse, but this updates the DImode shift by constant costs on TARGET_64BIT, in the same way that the previous rtx_costs patch improved the SImode costs. Generating op