https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123268
Jeffrey A. Law <law at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Blocks|120763 |
CC| |rguenth at gcc dot gnu.org
--- Comment #1 from Jeffrey A. Law <law at gcc dot gnu.org> ---
This looks kind of like a FRE issue to me. In the .slsr dump we have:
vect_b_22.17_19 = .COND_LEN_MUL ({ -1, -1, -1, -1 }, vect__17.11_15,
vect__19.14_16, { 0.0, 0.0, 0.0, 0.0 }, 4, 0);
vectp_f.8_20 = vectp_f.9_29 + 16;
In the .fre5 dump that's transformed into:
_40 = VIEW_CONVERT_EXPR<vector(4) unsigned int>(vect__17.11_15);
_30 = .COND_LEN_AND ({ -1, -1, -1, -1 }, _40, { 4294967295, 0, 0, 0 }, { 0.0,
0.0, 0.0, 0.0 }, 4, 0);
vect_b_22.17_19 = VIEW_CONVERT_EXPR<vector(4) float>(_30);
vectp_f.8_20 = vectp_f.9_29 + 16;
Note how we have FP values for the else part of the COND_LEN_AND. That seems
wrong.
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120763
[Bug 120763] [meta-bug] Tracker for bugs to visit during weekly RISC-V meeting
