* [PATCH] Only use NO_REGS in cost calculation when !hard_regno_mode_ok for GENERAL_REGS and mode.
@ 2023-05-17 6:57 liuhongt
2023-05-19 21:31 ` Jeff Law
2023-05-25 14:29 ` Vladimir Makarov
0 siblings, 2 replies; 4+ messages in thread
From: liuhongt @ 2023-05-17 6:57 UTC (permalink / raw)
To: gcc-patches; +Cc: vmakarov, linkw, segher
r14-172-g0368d169492017 replaces GENERAL_REGS with NO_REGS in cost
calculation when the preferred register class are not known yet.
It regressed powerpc PR109610 and PR109858, it looks too aggressive to use
NO_REGS when mode can be allocated with GENERAL_REGS.
The patch takes a step back, still use GENERAL_REGS when
hard_regno_mode_ok for mode and GENERAL_REGS, otherwise uses NO_REGS.
Kewen confirmed the patch fixed PR109858, I vefiried it also fixed PR109610.
Bootstrapped and regtested on x86_64-pc-linux-gnu{-m32,}.
No big performance impact for SPEC2017 on icelake server.
Ok for trunk?
gcc/ChangeLog:
* ira-costs.cc (scan_one_insn): Only use NO_REGS in cost
calculation when !hard_regno_mode_ok for GENERAL_REGS and
mode, otherwise still use GENERAL_REGS.
---
gcc/ira-costs.cc | 12 ++++++++----
1 file changed, 8 insertions(+), 4 deletions(-)
diff --git a/gcc/ira-costs.cc b/gcc/ira-costs.cc
index d2a801ab9b0..ae8304ff938 100644
--- a/gcc/ira-costs.cc
+++ b/gcc/ira-costs.cc
@@ -1572,12 +1572,16 @@ scan_one_insn (rtx_insn *insn)
&& (! ira_use_lra_p || ! pic_offset_table_rtx
|| ! contains_symbol_ref_p (XEXP (note, 0))))
{
- /* Costs for NO_REGS are used in cost calculation on the
- 1st pass when the preferred register classes are not
- known yet. In this case we take the best scenario. */
- enum reg_class cl = NO_REGS;
+ enum reg_class cl = GENERAL_REGS;
rtx reg = SET_DEST (set);
int num = COST_INDEX (REGNO (reg));
+ /* Costs for NO_REGS are used in cost calculation on the
+ 1st pass when the preferred register classes are not
+ known yet. In this case we take the best scenario when
+ mode can't be put into GENERAL_REGS. */
+ if (!targetm.hard_regno_mode_ok (ira_class_hard_regs[cl][0],
+ GET_MODE (reg)))
+ cl = NO_REGS;
COSTS (costs, num)->mem_cost
-= ira_memory_move_cost[GET_MODE (reg)][cl][1] * frequency;
--
2.39.1.388.g2fc9e9ca3c
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH] Only use NO_REGS in cost calculation when !hard_regno_mode_ok for GENERAL_REGS and mode.
2023-05-17 6:57 [PATCH] Only use NO_REGS in cost calculation when !hard_regno_mode_ok for GENERAL_REGS and mode liuhongt
@ 2023-05-19 21:31 ` Jeff Law
2023-05-25 14:29 ` Vladimir Makarov
1 sibling, 0 replies; 4+ messages in thread
From: Jeff Law @ 2023-05-19 21:31 UTC (permalink / raw)
To: liuhongt, gcc-patches; +Cc: vmakarov, linkw, segher
On 5/17/23 00:57, liuhongt via Gcc-patches wrote:
> r14-172-g0368d169492017 replaces GENERAL_REGS with NO_REGS in cost
> calculation when the preferred register class are not known yet.
> It regressed powerpc PR109610 and PR109858, it looks too aggressive to use
> NO_REGS when mode can be allocated with GENERAL_REGS.
> The patch takes a step back, still use GENERAL_REGS when
> hard_regno_mode_ok for mode and GENERAL_REGS, otherwise uses NO_REGS.
> Kewen confirmed the patch fixed PR109858, I vefiried it also fixed PR109610.
>
> Bootstrapped and regtested on x86_64-pc-linux-gnu{-m32,}.
> No big performance impact for SPEC2017 on icelake server.
> Ok for trunk?
>
> gcc/ChangeLog:
>
> * ira-costs.cc (scan_one_insn): Only use NO_REGS in cost
> calculation when !hard_regno_mode_ok for GENERAL_REGS and
> mode, otherwise still use GENERAL_REGS.
BTW, Vlad is on PTO right now. I'm sure he'll handle this after he
returns and starts digging out of all the stuff that's piled up.
jeff
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH] Only use NO_REGS in cost calculation when !hard_regno_mode_ok for GENERAL_REGS and mode.
2023-05-17 6:57 [PATCH] Only use NO_REGS in cost calculation when !hard_regno_mode_ok for GENERAL_REGS and mode liuhongt
2023-05-19 21:31 ` Jeff Law
@ 2023-05-25 14:29 ` Vladimir Makarov
2023-05-25 15:37 ` Segher Boessenkool
1 sibling, 1 reply; 4+ messages in thread
From: Vladimir Makarov @ 2023-05-25 14:29 UTC (permalink / raw)
To: liuhongt, gcc-patches; +Cc: linkw, segher
On 5/17/23 02:57, liuhongt wrote:
> r14-172-g0368d169492017 replaces GENERAL_REGS with NO_REGS in cost
> calculation when the preferred register class are not known yet.
> It regressed powerpc PR109610 and PR109858, it looks too aggressive to use
> NO_REGS when mode can be allocated with GENERAL_REGS.
> The patch takes a step back, still use GENERAL_REGS when
> hard_regno_mode_ok for mode and GENERAL_REGS, otherwise uses NO_REGS.
> Kewen confirmed the patch fixed PR109858, I vefiried it also fixed PR109610.
>
> Bootstrapped and regtested on x86_64-pc-linux-gnu{-m32,}.
> No big performance impact for SPEC2017 on icelake server.
> Ok for trunk?
>
> gcc/ChangeLog:
>
> * ira-costs.cc (scan_one_insn): Only use NO_REGS in cost
> calculation when !hard_regno_mode_ok for GENERAL_REGS and
> mode, otherwise still use GENERAL_REGS.
Thank you for the patch. It looks good for me. It is ok to commit it
into the trunk.
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH] Only use NO_REGS in cost calculation when !hard_regno_mode_ok for GENERAL_REGS and mode.
2023-05-25 14:29 ` Vladimir Makarov
@ 2023-05-25 15:37 ` Segher Boessenkool
0 siblings, 0 replies; 4+ messages in thread
From: Segher Boessenkool @ 2023-05-25 15:37 UTC (permalink / raw)
To: Vladimir Makarov; +Cc: liuhongt, gcc-patches, linkw
On Thu, May 25, 2023 at 10:29:47AM -0400, Vladimir Makarov wrote:
>
> On 5/17/23 02:57, liuhongt wrote:
> >r14-172-g0368d169492017 replaces GENERAL_REGS with NO_REGS in cost
> >calculation when the preferred register class are not known yet.
> >It regressed powerpc PR109610 and PR109858, it looks too aggressive to use
> >NO_REGS when mode can be allocated with GENERAL_REGS.
> >The patch takes a step back, still use GENERAL_REGS when
> >hard_regno_mode_ok for mode and GENERAL_REGS, otherwise uses NO_REGS.
> >Kewen confirmed the patch fixed PR109858, I vefiried it also fixed
> >PR109610.
> >
> >Bootstrapped and regtested on x86_64-pc-linux-gnu{-m32,}.
> >No big performance impact for SPEC2017 on icelake server.
> >Ok for trunk?
> >
> >gcc/ChangeLog:
> >
> > * ira-costs.cc (scan_one_insn): Only use NO_REGS in cost
> > calculation when !hard_regno_mode_ok for GENERAL_REGS and
> > mode, otherwise still use GENERAL_REGS.
>
> Thank you for the patch. It looks good for me. It is ok to commit it
> into the trunk.
Thanks everyone involved for fixing this nasty regression! Much
appreciated.
Segher
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2023-05-25 15:38 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-05-17 6:57 [PATCH] Only use NO_REGS in cost calculation when !hard_regno_mode_ok for GENERAL_REGS and mode liuhongt
2023-05-19 21:31 ` Jeff Law
2023-05-25 14:29 ` Vladimir Makarov
2023-05-25 15:37 ` Segher Boessenkool
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).