sched-deps: respect deps->readonly in macro-fusion (PR 84566)

Message ID alpine.LNX.2.20.13.1804101323460.24851@monopod.intra.ispras.ru
State New
Headers show
Series
  • sched-deps: respect deps->readonly in macro-fusion (PR 84566)
Related show

Commit Message

Alexander Monakov April 10, 2018, 10:40 a.m.
Hi,

this fixes a simple "regression" under the qsort_chk umbrella: sched-deps
analysis has deps->readonly flag, but macro-fusion code does not respect it
and mutates instructions. This breaks an assumption in sel_rank_for_schedule
and manifests as qsort checking error.

Since sched_macro_fuse_insns is only called to set SCHED_GROUP_P on suitable
insns, guard the call with !deps->readonly.

Bootstrapped/regtested on x86_64 with sel-sched active and
--with-cpu=sandybridge to exercise macro-fusion code and verified on aarch64
cross-compiler that the failing testcase given in the PR is fixed.

OK to apply?

Thanks.
Alexander

	PR rtl-optimization/84566
	* sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
	sched_macro_fuse_insns.

Comments

Andrey Belevantsev April 11, 2018, 10:19 a.m. | #1
On 10.04.2018 13:40, Alexander Monakov wrote:
> Hi,
> 
> this fixes a simple "regression" under the qsort_chk umbrella: sched-deps
> analysis has deps->readonly flag, but macro-fusion code does not respect it
> and mutates instructions. This breaks an assumption in sel_rank_for_schedule
> and manifests as qsort checking error.
> 
> Since sched_macro_fuse_insns is only called to set SCHED_GROUP_P on suitable
> insns, guard the call with !deps->readonly.
> 
> Bootstrapped/regtested on x86_64 with sel-sched active and
> --with-cpu=sandybridge to exercise macro-fusion code and verified on aarch64
> cross-compiler that the failing testcase given in the PR is fixed.
> 
> OK to apply?

Fine with me but you need a scheduler maintainer approval.

Andrey

> 
> Thanks.
> Alexander
> 
> 	PR rtl-optimization/84566
> 	* sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
> 	sched_macro_fuse_insns.
> 
> diff --git a/gcc/sched-deps.c b/gcc/sched-deps.c
> index 9a5cbebea40..120b5f0ddc1 100644
> --- a/gcc/sched-deps.c
> +++ b/gcc/sched-deps.c
> @@ -2897,7 +2897,8 @@ sched_analyze_insn (struct deps_desc *deps, rtx x, rtx_insn *insn)
>  			 && code == SET);
>  
>    /* Group compare and branch insns for macro-fusion.  */
> -  if (targetm.sched.macro_fusion_p
> +  if (!deps->readonly
> +      && targetm.sched.macro_fusion_p
>        && targetm.sched.macro_fusion_p ())
>      sched_macro_fuse_insns (insn);
>  
>
Vladimir Makarov April 11, 2018, 2:19 p.m. | #2
On 04/11/2018 06:19 AM, Andrey Belevantsev wrote:
> On 10.04.2018 13:40, Alexander Monakov wrote:
>> Hi,
>>
>> this fixes a simple "regression" under the qsort_chk umbrella: sched-deps
>> analysis has deps->readonly flag, but macro-fusion code does not respect it
>> and mutates instructions. This breaks an assumption in sel_rank_for_schedule
>> and manifests as qsort checking error.
>>
>> Since sched_macro_fuse_insns is only called to set SCHED_GROUP_P on suitable
>> insns, guard the call with !deps->readonly.
>>
>> Bootstrapped/regtested on x86_64 with sel-sched active and
>> --with-cpu=sandybridge to exercise macro-fusion code and verified on aarch64
>> cross-compiler that the failing testcase given in the PR is fixed.
>>
>> OK to apply?
> Fine with me but you need a scheduler maintainer approval.
>
>
Ok with me.  Thanks for the patch.

Patch

diff --git a/gcc/sched-deps.c b/gcc/sched-deps.c
index 9a5cbebea40..120b5f0ddc1 100644
--- a/gcc/sched-deps.c
+++ b/gcc/sched-deps.c
@@ -2897,7 +2897,8 @@  sched_analyze_insn (struct deps_desc *deps, rtx x, rtx_insn *insn)
 			 && code == SET);
 
   /* Group compare and branch insns for macro-fusion.  */
-  if (targetm.sched.macro_fusion_p
+  if (!deps->readonly
+      && targetm.sched.macro_fusion_p
       && targetm.sched.macro_fusion_p ())
     sched_macro_fuse_insns (insn);