https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99572
Bug ID: 99572
Summary: std::counting_semaphore coalescing wakes
Product: gcc
Version: 11.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: libstdc++
Assignee: unassigned at gcc dot gnu.org
Reporter: michaelkuc6 at gmail dot com
Target Milestone: ---
Created attachment 50379
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=50379&action=edit
Non-waking semaphores source code
When repeatedly calling `std::counting_semaphore::release()`, only a single
thread is woken, unless the releasing thread is made to sleep.
A minimal example of the error is at
https://github.com/Crystalix007/GCC-Semaphores, or as the attached file
(compiled with `g++ --std=c++20 -o Sem -lpthread source.cpp`).
Inspecting with GDB myself, the semaphore release function calls
`std::__atomic_semaphore<int>::_M_release()`, implemented
https://github.com/gcc-mirror/gcc/blob/5987d8a79cda1069c774e5c302d5597310270026/libstdc%2B%2B-v3/include/bits/semaphore_base.h#L270.
Notably the semaphore release only wakes a single thread, which would be OK if
a different thread were woken each time. Eventually the call chain ends up
performing a futex syscall
https://github.com/gcc-mirror/gcc/blob/5987d8a79cda1069c774e5c302d5597310270026/libstdc%2B%2B-v3/include/bits/atomic_wait.h#L112.
As far as I can tell, this doesn't correctly handle repeated wake-ups. Testing
against the POSIX semaphore implementation, GCC's implementation seems to wake
up only a single thread (i.e. coalescing all wakeup events into one).
By sleeping the thread between semaphore releases, the program can actually
finish. If the delay is too short, sometimes only some of the threads are
woken, but with a sufficient delay, all threads are woken correctly.
Unless I'm misunderstanding, is it possible that the woken thread is not marked
as woken, or the futex wake just doesn't care that the woken thread is already
awake?
GCC Version (compiled from commit a18ebd6c439):
Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-pc-linux-gnu/11.0.1/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: /home/michaelkuc6/.cache/yay/gcc-git/src/gcc/configure
--prefix=/usr --libdir=/usr/lib --libexecdir=/usr/lib --mandir=/usr/share/man
--infodir=/usr/share/info --with-bugurl=https://bugs.archlinux.org/
--enable-languages=c,c++,ada,fortran,go,lto,objc,obj-c++,d --with-isl
--with-linker-hash-style=gnu --with-system-zlib --enable-__cxa_atexit
--enable-cet=auto --enable-checking=release --enable-clocale=gnu
--enable-default-pie --enable-default-ssp --enable-gnu-indirect-function
--enable-gnu-unique-object --enable-install-libiberty --enable-linker-build-id
--enable-lto --enable-multilib --enable-plugin --enable-shared
--enable-threads=posix --disable-libssp --disable-libstdcxx-pch
--disable-libunwind-exceptions --disable-werror
gdc_include_dir=/usr/include/dlang/gdc
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 11.0.1 20210307 (experimental) (GCC)
System Type: Linux (x86_64) (specifically Arch Linux kernel 5.11.2)
