https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111613
Bug ID: 111613
Summary: Bit field stores can be incorrectly optimized away
when -fstore-merging is in effect
Product: gcc
Version: 12.1.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c
Assignee: unassigned at gcc dot gnu.org
Reporter: gcc at kempniu dot pl
Target Milestone: ---
Created attachment 56002
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=56002&action=edit
Preprocessed reproducer
This is the simplest reproducer I could come up with:
$ cat bitfield.c
#include <stdio.h>
#include <stdlib.h>
struct bitfield {
unsigned int field1 : 1;
unsigned int field2 : 1;
unsigned int field3 : 1;
};
__attribute__((noinline)) static void
set_field1_and_field2(struct bitfield *b) {
b->field1 = 1;
b->field2 = 1;
}
__attribute__((noinline)) static struct bitfield *
new_bitfield(void) {
struct bitfield *b = (struct bitfield *)malloc(sizeof(*b));
b->field3 = 1;
set_field1_and_field2(b);
return b;
}
int main(void) {
struct bitfield *b = new_bitfield();
printf("%d\n", b->field3);
return 0;
}
$ gcc -O2 -o bitfield bitfield.c
$ ./bitfield
0
$ gcc -O2 -fno-store-merging -o bitfield bitfield.c
$ ./bitfield
1
Removing one of the stores from set_field1_and_field2() makes the issue
go away. Moving "b->field3 = 1;" below the call to
set_field1_and_field2() also makes the issue go away.
This was originally found for the following GCC version:
$ gcc -v
Using built-in specs.
COLLECT_GCC=/usr/bin/gcc
COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-pc-linux-gnu/13.2.1/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: /build/gcc/src/gcc/configure
--enable-languages=ada,c,c++,d,fortran,go,lto,objc,obj-c++ --enable-bootstrap
--prefix=/usr --libdir=/usr/lib --libexecdir=/usr/lib --mandir=/usr/share/man
--infodir=/usr/share/info --with-bugurl=https://bugs.archlinux.org/
--with-build-config=bootstrap-lto --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-libstdcxx-backtrace --enable-link-serialization=1
--enable-linker-build-id --enable-lto --enable-multilib --enable-plugin
--enable-shared --enable-threads=posix --disable-libssp --disable-libstdcxx-pch
--disable-werror
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 13.2.1 20230801 (GCC)
However, I have subsequently confirmed that it also happens on the
current master branch.
A git bisect points at the following commit as the culprit (first
included in GCC 12.1.0):
commit 22c242342e38ebffa6bbf7e86e7a1e4abdf0d686
Author: Martin Liska <mli...@suse.cz>
Date: Thu Nov 18 17:50:19 2021 +0100
IPA: fix reproducibility in IPA MOD REF
gcc/ChangeLog:
* ipa-modref.c (analyze_function): Do not execute the code
only if dump_file != NULL.
Reverting this change on top of current master makes the issue
disappear, so it looks legit to me.
Disassembly of new_bitfield() for "gcc -O2":
0x0000000000001190 <+0>: sub $0x8,%rsp
0x0000000000001194 <+4>: mov $0x4,%edi
0x0000000000001199 <+9>: call 0x1040 <malloc@plt>
0x000000000000119e <+14>: mov %rax,%rdi
0x00000000000011a1 <+17>: call 0x1180 <set_field1_and_field2>
0x00000000000011a6 <+22>: add $0x8,%rsp
0x00000000000011aa <+26>: ret
Disassembly of new_bitfield() for "gcc -O2 -fno-store-merging":
0x0000000000001190 <+0>: sub $0x8,%rsp
0x0000000000001194 <+4>: mov $0x4,%edi
0x0000000000001199 <+9>: call 0x1040 <malloc@plt>
0x000000000000119e <+14>: orb $0x4,(%rax)
0x00000000000011a1 <+17>: mov %rax,%rdi
0x00000000000011a4 <+20>: call 0x1180 <set_field1_and_field2>
0x00000000000011a9 <+25>: add $0x8,%rsp
0x00000000000011ad <+29>: ret
