[Bug target/81935] New: [8 Regression] r251260 miscompiles libstdc++ (on ppc64le)
trippels at gcc dot gnu.org
gcc-bugzilla@gcc.gnu.org
Wed Aug 23 08:28:00 GMT 2017
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81935
Bug ID: 81935
Summary: [8 Regression] r251260 miscompiles libstdc++ (on
ppc64le)
Product: gcc
Version: 8.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: target
Assignee: unassigned at gcc dot gnu.org
Reporter: trippels at gcc dot gnu.org
CC: aldyh at gcc dot gnu.org
Target Milestone: ---
Target: ppc64le
Since r251260 several programs spin in libstdc++.so.6.0.25.
For example cmake:
- 99.99% 1.59% cmake libstdc++.so.6.0.25 [.] __cxa_guard_acquire
- 98.40% __cxa_guard_acquire
+ 10.87% futex_wait_setup
+ 6.86% gup_pte_range
+ 6.86% get_futex_key
6.63% fast_exception_return
+ 6.10% drop_futex_key_refs.isra.11
+ 6.09% put_page
+ 4.97% syscall
+ 4.90% _raw_spin_lock
+ 4.86% unlock_page
+ 3.08% __get_user_pages_fast
+ 3.02% system_call_relon_pSeries
+ 2.77% system_call_common
+ 2.57% system_call
+ 2.46% __smp_mb__after_atomic
+ 2.44% syscall_exit_work
+ 2.28% __audit_syscall_exit
+ 2.16% do_futex
+ 2.01% __audit_syscall_entry
+ 1.94% do_syscall_trace_leave
+ 1.42% kfree
+ 1.37% dput
1.17% ret_from_except_lite
+ 1.10% do_syscall_trace_enter
+ 1.04% mntput
+ 1.01% futex_wait
0.94% current_kernel_time
+ 0.88% __cxa_guard_acquire
+ 0.87% sys_futex
0.76% __syscall_error
+ 0.74% hash_futex
+ 0.73% memset
0.61% path_put
0.59% get_user_pages_fast
+ 1.59% _dl_start_user
+ 99.99% 0.00% cmake libstdc++.so.6.0.25 [.]
_GLOBAL__sub_I_compatibility_thread_c__0x.cc
+ 99.99% 0.00% cmake ld-2.17.so [.] _dl_init_internal
+ 99.99% 0.00% cmake ld-2.17.so [.] _dl_start_user
+ 87.27% 0.00% cmake libstdc++.so.6.0.25 [.]
std::future_error::future_error
+ 23.29% 10.87% cmake [kernel.kallsyms] [k] futex_wait_setup
+ 21.54% 2.57% cmake [kernel.kallsyms] [k] system_call
+ 18.08% 6.86% cmake [kernel.kallsyms] [k] get_futex_key
+ 16.45% 0.87% cmake [kernel.kallsyms] [k] sys_futex
+ 15.82% 2.16% cmake [kernel.kallsyms] [k] do_futex
ranlib also spins in the same place.
Reverting the commit "fixes" the problem.
More information about the Gcc-bugs
mailing list