Hi Richard,
On 22/06/2026 14:59, Christopher Bazley wrote:
I have confirmed that the same (bad) code generation occurs without my
changes in this patchset, and also at the commit upon which my patch
series is based (i.e., without any of my BB SLP patches).
I think it would be worthwhile to improve store_constructor for most
types even if it still has pathological behavior for one type, so I am
tempted to report it as a bug and move on. I can also note in the commit
log that the described non-fallback behaviour of store_constructor
relies on vec_init having been implemented for the mode and eltmode.
Please consider approving the latest version of this patch without the
bugfix:
https://forge.sourceware.org/gcc/gcc-TEST/pulls/177
https://inbox.sourceware.org/gcc-patches/bmm.hjj7t7yaru.gcc.gcc-
[email protected]/
I have since created PR 125941 for the issue of bad code generation
triggered by RUNTESTFLAGS="aarch64-sve-acle.exp=cops_bool.c".
--
Christopher Bazley
Staff Software Engineer, GNU Tools Team.
Arm Ltd, 110 Fulbourn Road, Cambridge, CB1 9NJ, UK.
http://www.arm.com/
