On Tue, 2 Sep 2025, Tamar Christina wrote:
> If the user has requested loop unrolling through pragma GCC unroll then at the
> moment we only set LOOP_VINFO_USER_UNROLL if the vectorizer has not overrode
> the
> unroll factor (through backend costing) or if the VF made the requested unroll
> factor be 1.
>
> But of these events are costing related, and so it stands to reason that we
> should set LOOP_VINFO_USER_UNROLL to we return the RTL unroller to use the
> backend costing for any further unrolling.
I can't parse this fully.
Hmm, but shouldn't we at least special-case when !new_res (aka
the requested unrolling wasn't used) but VF == 1, aka vectorization
didn't do any unrolling?
Or is the reasoning that the vector cost model is as good as the
RTL one (which doesn't exist...) and thus if that decided not to
unroll do not let the RTL cost model override it? But that means
to set loop->unroll to 1, not 0 as we do for LOOP_VINFO_USER_UNROLL.
The RTL "cost mode" will simply unroll by 8 if possible.
What was it that made you propose this change?
Richard.
> Bootstrapped Regtested on aarch64-none-linux-gnu,
> arm-none-linux-gnueabihf, x86_64-pc-linux-gnu
> -m32, -m64 and no issues.
>
> Ok for master? Tests using this are in the next patch.
>
> Thanks,
> Tamar
>
> gcc/ChangeLog:
>
> * tree-vect-loop.cc (vect_analyze_loop_1): If the unroll pragma was set
> mark it as handled.
>
> ---
> diff --git a/gcc/tree-vect-loop.cc b/gcc/tree-vect-loop.cc
> index
> 3b56f865ec7c3fdb8e9def37ec446a25086b4e28..f393b45232b3390959a2be756db3aa8b35002bf0
> 100644
> --- a/gcc/tree-vect-loop.cc
> +++ b/gcc/tree-vect-loop.cc
> @@ -2941,11 +2941,13 @@ vect_analyze_loop_1 (class loop *loop,
> vec_info_shared *shared,
> {
> delete loop_vinfo;
> loop_vinfo = unroll_vinfo;
> - LOOP_VINFO_USER_UNROLL (loop_vinfo) = user_unroll > 1;
> }
> else
> delete unroll_vinfo;
> }
> +
> + /* Record that we have honored a user unroll factor. */
> + LOOP_VINFO_USER_UNROLL (loop_vinfo) = user_unroll > 1;
> }
>
> /* Remember the autodetected vector mode. */
>
>
>
--
Richard Biener <rguent...@suse.de>
SUSE Software Solutions Germany GmbH,
Frankenstrasse 146, 90461 Nuernberg, Germany;
GF: Ivo Totev, Andrew McDonald, Werner Knoblich; (HRB 36809, AG Nuernberg)
