[0/4] Addressing modulo-scheduling bugs

zhroma zhroma@ispras.ru
Tue Apr 16 11:59:00 GMT 2019


Last few days I’ve added my comments/patches into almost all 
modulo-scheduler PRs appeared in the last two years.  Here I want to 
discuss five PRs.  First of all, I have four patches which fix 
regressions, so technically I can ask about applying them right now on 
stage4.  But we don’t have a maintainer in modulo-sched pass, so not 
sure about their fast approval.

Last one PR90040 is meta-bug I’ve created.  It consolidates 5-6 other 
bugzilla PRs, which really are about the same issue.  Unfortunately, I 
can’t solve that myself.  Jakub, Honza, would you kindly help me and 
take a look at that PR to provide any suggestions you might have.

All my patches were successfully bootstrapped and regtested on x86_64.  
Also I’ve checked that in cross-compiler mode to s390, spu, aarch64, 
arm, ia64, ppc and ppc64 regtest shows no new regressions.  The same 
testing was done with -fmodulo-sched enabled.  Also the same testing was 
done with -fmodulo-sched -fmodulo-sched-allow-regmoves enabled.  
Moreover, all that was also done on 8 branch. No new issues introduced.  
So, since I’ve made really heavy testing,
IMO all 4 fixes can go into trunk and even 8 branch.

When development goes into stage1 and after somehow fixing PR90040 issue 
I will introduce updated patchset described here:
https://gcc.gnu.org/ml/gcc-patches/2017-02/msg01647.html
(the set is supported locally on all branches since 4.9 with making a 
lot of regtesting).

Regarding the modulo scheduling maintainership, if my patches look good, 
I'm happy to volunteer to maintain the code as I looked at the code a 
lot during recent years.  Please let me know the steps I should do to 
make that happen.

Roman



More information about the Gcc-patches mailing list