On Mon, Sep 08, 2025 at 02:38:48PM +0200, Richard Biener wrote:
> When there's an asm goto in the latch of a loop we may not use
> IP_END IVs since instantiating those would (need to) split the
> latch edge which in turn invalidates IP_NORMAL position handling.
> This is a revision of the PR107997 fix.
>
> Bootstrap and regtest running on x86_64-unknown-linux-gnu.
>
> PR tree-optimization/107997
> PR tree-optimization/121844
> * tree-ssa-loop-ivopts.cc (allow_ip_end_pos_p): Do not allow
> IP_END for latches ending with a control stmt.
> (create_new_iv): Do not split the latch edge, instead assert
> that's not necessary.
LGTM.
> * gcc.dg/torture/pr121844.c: New testcase.
> --- /dev/null
> +++ b/gcc/testsuite/gcc.dg/torture/pr121844.c
> @@ -0,0 +1,19 @@
> +/* { dg-do compile } */
> +
> +#include <stdint.h>
Why? Does it rely on the size of any of those types in any way?
Can't you s/int32_t/int/ and s/int16_t/void/ and probably
unsigned long long * for type of e?
> +
> +typedef uint64_t a;
> +int32_t b[] = {};
> +int32_t **c;
> +int16_t d() {
> + a *e;
This one is uninitialized, is that needed for the repro?
Can't it be e.g. argument to the function instead?
> + int32_t f;
> + *c = &f;
> + for (;;)
> + asm goto("" : : : : g);
> + for (; f; f--) {
> + asm goto("" : : : : g);
> + g:
> + *e ^= b[f + 1];
> + }
> +}
Jakub
