[Bug tree-optimization/90018] [8/9 Regression] r265453 miscompiled 527.cam4_r in SPEC CPU 2017

marxin at gcc dot gnu.org gcc-bugzilla@gcc.gnu.org
Tue Apr 9 11:04:00 GMT 2019


https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90018

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |WAITING

--- Comment #6 from Martin Liška <marxin at gcc dot gnu.org> ---
I've just tested that on -march=skylake-avx512:
model name      : Intel(R) Xeon(R) Platinum 8164 CPU @ 2.00GHz

r265451 works for me, but I had to increase a stack limit. With default limit
8192 I hit segfault here:

Program received signal SIGSEGV, Segmentation fault.
0x00000000008c890f in dyn_comp::dyn_run (ptop=219.4067000000008, ndt=1800,
te0=0, dyn_state=..., dyn_in=..., dyn_out=..., rc=-1) at
dyn_comp.fppized.f90:1142
1142       call t_startf ('dyn_run_alloc')
(gdb) bt
#0  0x00000000008c890f in dyn_comp::dyn_run (ptop=219.4067000000008, ndt=1800,
te0=0, dyn_state=..., dyn_in=..., dyn_out=..., rc=-1) at
dyn_comp.fppized.f90:1142
#1  0x00000000009e18f9 in stepon::stepon_run1 (dtime_out=<optimized out>,
phys_state=..., phys_tend=..., dyn_in=..., dyn_out=...) at
stepon.fppized.f90:427
#2  0x0000000000a27840 in cam_comp::cam_run1 (cam_in=..., cam_out=...) at
cam_comp.fppized.f90:195
#3  0x0000000000a50a04 in atm_comp_mct::atm_init_mct (eclock=..., cdata_a=...,
x2a_a=..., a2x_a=..., nlfilename=<error reading variable: value requires
4200145 bytes, which is more than max-value-size>,
_nlfilename=_nlfilename@entry=0) at atm_comp_mct.fppized.f90:349
#4  0x0000000000ab0f02 in ccsm_comp_mod::ccsm_init () at
ccsm_comp_mod.fppized.f90:1577
#5  0x0000000000402bcd in ccsm_driver () at ccsm_driver.fppized.f90:57
#6  main (argc=<optimized out>, argv=<optimized out>) at
ccsm_driver.fppized.f90:25
#7  0x00007ffff6e35ea7 in __libc_start_main () from /lib64/libc.so.6

@H.J. Can you please check the stack limit?


More information about the Gcc-bugs mailing list