https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94600
Bug ID: 94600
Summary: Ignored volatile specifier on loop unrolling and
bitfield misoptimization
Product: gcc
Version: 10.0
Status: UNCONFIRMED
Keywords: wrong-code
Severity: normal
Priority: P3
Component: middle-end
Assignee: unassigned at gcc dot gnu.org
Reporter: hp at gcc dot gnu.org
Target Milestone: ---
Host: x86_64-unknown-linux-gnu
Target: arm-eabi, cris-elf, aarch64-linux
Created attachment 48275
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=48275&action=edit
test-case in need of decoration for the gcc test-suite. Compile e.g. at -O2 or
-Os.
Observe the unwieldy code in function foo compiled at -O2 or -Os for any of the
listed targets (and likely many more, except curiously not x86_64 and i386).
Compare this to the compiled bar; nice as expected.
Note in particular the reads in the generated code from the LHS specified as
volatile and which is only stored into in the source code. Note also the
smaller-than-struct-sized stores. The latter may be target-dependent but the
reads should not happen anywhere.
