This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Drop frequencies from basic blocks
> Can be also seen in GCC PGO:
>
> checking for ssize_t... ../../libdecnumber/decNumber.c: In function ‘decDecap’:
> ../../libdecnumber/decNumber.c:7640:25: error: corrupted value profile: interval profile counter (0 out of 0) inconsistent with basic-block count (4356)
> if (cut!=DECDPUN) *msu%=powers[cut]; /* clear left digits */
Hmm, I have restarted profiledbootstrap and it also reproduces to me.
It is odd that those counters are 0 in basic block that is executed. Having small
reproducer would be cool, but I will try to debug it from libdecnumber.
That other ICE is another issue with Theresa's code for dropping profiles that
are mismatched (it is only place we mismatch profile in GCC). I will take a look
what went wrong here.
Thanks,
Honza
> ^~
> no
> checking unwind.h usability... ../../libdecnumber/decNumber.c: In function ‘decNumberRotate’:
> ../../libdecnumber/decNumber.c:2526:9: error: corrupted value profile: interval profile counter (0 out of 0) inconsistent with basic-block count (8)
> uInt save=res->lsu[0]%powers[shift]; /* save low digit(s) */
> ^~~~
> ../../libdecnumber/decNumber.c:2529:11: error: corrupted value profile: interval profile counter (0 out of 0) inconsistent with basic-block count (4)
> uInt rem=save%powers[shift-msudigits];/* split save */
> ^~~
> ../../libdecnumber/decNumber.c:2546:11: error: corrupted value profile: interval profile counter (0 out of 0) inconsistent with basic-block count (2)
> uInt save=res->lsu[0]%powers[shift]; /* save low digit(s) */
> ^~~~