On Tue, 8 Jul 2025, Tamar Christina wrote:
> This patch adds support for niters ranges for partial
> vector loops.
>
> Due to the last iteration being partial the bounds should
> be at least 1 but niters // vf as the max.
>
> Bootstrapped Regtested on aarch64-none-linux-gnu,
> arm-none-linux-gnueabihf, x86_64-pc-linux-gnu
> -m32, -m64 and no issues.
>
> Tested testcase on riscv64-unknown-linux-gnu. Should use
> the same testcase as patch 1.
>
> Ok for master?
OK.
Richard.
> Thanks,
> Tamar
>
> gcc/ChangeLog:
>
> PR tree-optimization/120922
> * tree-vect-loop-manip.cc (vect_gen_vector_loop_niters): Support range
> for partial vectors.
>
> ---
> diff --git a/gcc/tree-vect-loop-manip.cc b/gcc/tree-vect-loop-manip.cc
> index
> 9c58a6a01d7c9c3d97a2f8595b1ea6cff2d653f6..2d01a4b0ed1c8431510ad5e6e2175dbe89371618
> 100644
> --- a/gcc/tree-vect-loop-manip.cc
> +++ b/gcc/tree-vect-loop-manip.cc
> @@ -2857,11 +2857,21 @@ vect_gen_vector_loop_niters (loop_vec_info
> loop_vinfo, tree niters,
> we set range information to make niters analyzer's life easier.
> Note the number of latch iteration value can be TYPE_MAX_VALUE so
> we have to represent the vector niter TYPE_MAX_VALUE + 1 / vf. */
> - if (stmts != NULL
> - && const_vf > 0
> - && !LOOP_VINFO_USING_PARTIAL_VECTORS_P (loop_vinfo))
> + if (stmts != NULL && const_vf > 0)
> {
> - if (niters_no_overflow)
> + if (niters_no_overflow
> + && LOOP_VINFO_USING_PARTIAL_VECTORS_P (loop_vinfo))
> + {
> + int_range<1> vr (type,
> + wi::one (TYPE_PRECISION (type)),
> + wi::div_ceil (wi::max_value
> + (TYPE_PRECISION (type),
> + TYPE_SIGN (type)),
> + const_vf,
> + TYPE_SIGN (type)));
> + set_range_info (niters_vector, vr);
> + }
> + else if (niters_no_overflow)
> {
> int_range<1> vr (type,
> wi::one (TYPE_PRECISION (type)),
>
>
>
--
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)
