On Fri, Feb 14, 2025 at 3:10 PM Richard Biener <rguent...@suse.de> wrote:
>
> 502.gcc_r when built with -fprofile-generate exposes a SLP discovery
> issue where an IV forced live due to early break is not properly
> discovered if its latch def is part of a different IVs SSA cycle.
> To mitigate this we have to make sure to create an SLP instance
> for the original IV. Ideally we'd handle all vect_induction_def
> the same but this is left for next stage1.
>
> Bootstrap and regtest running on x86_64-unknown-linux-gnu.
>
> PR tree-optimizatio/118852
> * tree-vect-slp.cc (vect_analyze_slp): For early-break
> forced-live IVs make sure we create an appropriate
> entry into the SLP graph.
>
> * gcc.dg/vect/pr118852.c: New testcase.
> ---
> gcc/testsuite/gcc.dg/vect/pr118852.c | 109 +++++++++++++++++++++++++++
> gcc/tree-vect-slp.cc | 24 +++++-
> 2 files changed, 130 insertions(+), 3 deletions(-)
> create mode 100644 gcc/testsuite/gcc.dg/vect/pr118852.c
>
> diff --git a/gcc/testsuite/gcc.dg/vect/pr118852.c
> b/gcc/testsuite/gcc.dg/vect/pr118852.c
> new file mode 100644
> index 00000000000..71bfb450f7c
> --- /dev/null
> +++ b/gcc/testsuite/gcc.dg/vect/pr118852.c
> @@ -0,0 +1,109 @@
> +/* { dg-add-options vect_early_break } */
> +/* { dg-require-profiling "-fprofile-generate" } */
> +/* { dg-additional-options "-fallow-store-data-races -fprofile-arcs" } */
> +/* { dg-additional-options "-msse4.2 -mprefer-vector-width=128" { target {
> x86_64-*-* i?86-*-* } } } */
> +
> +#include "tree-vect.h"
> +
> +typedef unsigned int hashval_t;
> +struct htab {
> + void ** entries;
> + unsigned long size;
> +};
> +typedef struct htab *htab_t;
> +unsigned long htab_size (htab_t h)
> +{
> + return h->size;
> +}
> +typedef struct
> +{
> + htab_t htab;
> + void * *slot;
> + void * *limit;
> +} htab_iterator;
> +
> +static inline void *
> +first_htab_element (htab_iterator *hti, htab_t table)
> +{
> + hti->htab = table;
> + hti->slot = table->entries;
> + hti->limit = hti->slot + htab_size (table);
> + do
> + {
> + void * x = *(hti->slot);
> + if (x != ((void *) 0) && x != ((void *) 1))
> + break;
> + } while (++(hti->slot) < hti->limit);
> +
> + if (hti->slot < hti->limit)
> + return *(hti->slot);
> + return
> + ((void *)0)
> + ;
"return (void *) 0;" ...
> +}
> +
> +static inline unsigned char
> +end_htab_p (const htab_iterator *hti)
> +{
> + if (hti->slot >= hti->limit)
> + return 1;
> + return 0;
> +}
> +
> +static inline void *
> +next_htab_element (htab_iterator *hti)
> +{
> + while (++(hti->slot) < hti->limit)
> + {
> + void * x = *(hti->slot);
> + if (x != ((void *) 0) && x != ((void *) 1))
> + return x;
> + }
> + return
> + ((void *)0)
> + ;
... and also above.
Uros.
