This is a summary of discussions relative to the merge request created by Karl
Meakin (karmea01) <[email protected]> titled
aarch64: Relax type-checking assert [PR126064]
since its creation.
Description: Comparing `arg_type` and `TREE_TYPE (b)` by pointer address causes
an
assertion failure when one type is a `typedef` for the other (eg
`uint32x2_t` and `__Uint32x2_t`. Fix by using a more relaxed comparison.
gcc/ChangeLog:
* config/aarch64/aarch64-neon-builtins-base.cc: Compare
`arg_type` and `TREE_TYPE (b)` with `types_compatible_p`, rather than
their pointer address.
gcc/testsuite/ChangeLog:
* gcc.target/aarch64/pr126064.c: New test.
ChangeLog:
* v1: Initial revision
* v2: Use `types_compatible_p` rather than comparing types mods and signedness
* v3: Fix commit author email
The full and up to date discussion can be found at
https://forge.sourceware.org/gcc/gcc-TEST/pulls/183
The merge request has been closed without being merged directly on the forge
repository.
On 2026-07-09 14:12:27+00:00, Alex Coplan (acoplan) <[email protected]>
requested changes to the code:
For the subject, the PR number should be in square parens. So how about
something like: "aarch64: Relax type-checking assert [PR126064]"? Otherwise,
with the test updated, this LGTM.
> +++ gcc/testsuite/gcc.target/aarch64/pr126064.c
> @@ -0,0 +4,4 @@
> +#include <arm_neon.h>
> +
> +void
> +repro (uint8_t *dst, uint8x8_t a, uint8x8_t b)
Can you instead use the more minimal reproducer that Tamar gave in the PR
(https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126064#c1)?
On 2026-07-09 14:16:51+00:00, Alex Coplan (acoplan) wrote:
Also, it's not usual to use backticks in the ChangeLog, so please remove those.
On 2026-07-11 04:57:52+00:00, Drea Pinski (pinskia) wrote:
@acoplan wrote in
https://forge.sourceware.org/gcc/gcc-TEST/pulls/183#issuecomment-6613:
> Also, it's not usual to use backticks in the ChangeLog, so please remove
> those.
I use them all the time and have for a few years now. Though it is usually
with respect to naming the match pattern that is being changed rather than
function names.
But others are starting to use them including for function names. Examples
include r17-1545-g896be69c09cda1, r17-1540-g191fa9195ff05a,
r17-660-gbf69bded7b47b5 and (myself) r17-536-gf108e23d2bff50.
On 2026-07-13 10:25:54+00:00, Alex Coplan (acoplan) wrote:
@pinskia wrote in
https://forge.sourceware.org/gcc/gcc-TEST/pulls/183#issuecomment-6645:
> @acoplan wrote in #183 (comment):
>
> > Also, it's not usual to use backticks in the ChangeLog, so please remove
> > those.
>
> I use them all the time and have for a few years now. Though it is usually
> with respect to naming the match pattern that is being changed rather than
> function names. But others are starting to use them including for function
> names. Examples include r17-1545-g896be69c09cda1, r17-1540-g191fa9195ff05a,
> r17-660-gbf69bded7b47b5 and (myself) r17-536-gf108e23d2bff50.
Fair enough, feel free to ignore that then.
On 2026-07-14 16:34:35+00:00, Karl Meakin (karmea01) wrote:
Merged
