From: Richard Biener <rguenther@suse.de>
To: liuhongt <hongtao.liu@intel.com>
Cc: gcc-patches@gcc.gnu.org, tamar.christina@arm.com
Subject: Re: [PATCH] [vect] Don't peel nonlinear iv(mult or shift) for epilog when vf is not constant.
Date: Thu, 2 Feb 2023 08:45:43 +0000 (UTC) [thread overview]
Message-ID: <nycvar.YFH.7.77.849.2302020845350.6551@jbgna.fhfr.qr> (raw)
In-Reply-To: <20230202013052.2070569-1-hongtao.liu@intel.com>
On Thu, 2 Feb 2023, liuhongt wrote:
> Normally when vf is not constant, it will be prevented by
> vectorizable_nonlinear_inductions, but for this case, it failed going
> into
>
> if (STMT_VINFO_RELEVANT_P (stmt_info))
> {
> need_to_vectorize = true;
> if (STMT_VINFO_DEF_TYPE (stmt_info) == vect_induction_def
> && ! PURE_SLP_STMT (stmt_info))
> ok = vectorizable_induction (loop_vinfo,
> stmt_info, NULL, NULL,
> &cost_vec);
>
> since the iv is never used outside of the loop, and will be dce later, so
> vectorizer doesn't bother checking if it's vectorizable. it's
> true but hit gcc_assert in vect_can_peel_nonlinear_iv_p when vf is not
> constant. One solution is ignoring the nonlinear iv peeling if it's
> !STMT_VINFO_RELEVANT_P (stmt_info) just like the upper code, the other
> solution is returning false earlier in the
> vect_can_peel_nonlinear_iv_p when vf is not constant, the patch chooses
> the second incase there's other cases using vect_can_advance_ivs_p which
> calls vect_can_peel_nonlinear_iv_p.
>
> Also remove vect_can_peel_nonlinear_iv_p from vectorizable_nonlinear_inductions.
>
> Bootstrapped and regtested on x86_64-pc-linux-gnu{-m32,} and aarch64-linux-gnu{-m32,}
> Ok for trunk?
OK.
Thanks,
Richard.
> gcc/ChangeLog:
>
> PR tree-optimization/108601
> * tree-vectorizer.h (vect_can_peel_nonlinear_iv_p): Remove declare.
> * tree-vect-loop.cc
> (vectorizable_nonlinear_induction): Remove
> vect_can_peel_nonlinear_iv_p.
> (vect_can_peel_nonlinear_iv_p): Don't peel
> nonlinear iv(mult or shift) for epilog when vf is not
> constant and moved the defination to ..
> * tree-vect-loop-manip.cc (vect_can_peel_nonlinear_iv_p):
> .. Here.
>
> gcc/testsuite/ChangeLog:
>
> * gcc.target/aarch64/pr108601.c: New test.
> ---
> gcc/testsuite/gcc.target/aarch64/pr108601.c | 11 +++++
> gcc/tree-vect-loop-manip.cc | 44 ++++++++++++++++++++
> gcc/tree-vect-loop.cc | 46 ---------------------
> gcc/tree-vectorizer.h | 3 --
> 4 files changed, 55 insertions(+), 49 deletions(-)
> create mode 100644 gcc/testsuite/gcc.target/aarch64/pr108601.c
>
> diff --git a/gcc/testsuite/gcc.target/aarch64/pr108601.c b/gcc/testsuite/gcc.target/aarch64/pr108601.c
> new file mode 100644
> index 00000000000..deb8b3061d8
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/aarch64/pr108601.c
> @@ -0,0 +1,11 @@
> +/* { dg-do compile } */
> +/* { dg-options "-O3 -fprofile-generate -mcpu=neoverse-v1" } */
> +
> +int
> +foo() {
> + int flag = 1;
> + for (; flag <= 1 << 21; flag <<= 1)
> + ;
> + return 0;
> +}
> +
> diff --git a/gcc/tree-vect-loop-manip.cc b/gcc/tree-vect-loop-manip.cc
> index b5c5f859144..c04fcf40c44 100644
> --- a/gcc/tree-vect-loop-manip.cc
> +++ b/gcc/tree-vect-loop-manip.cc
> @@ -1390,6 +1390,50 @@ iv_phi_p (stmt_vec_info stmt_info)
> return true;
> }
>
> +/* Return true if vectorizer can peel for nonlinear iv. */
> +static bool
> +vect_can_peel_nonlinear_iv_p (loop_vec_info loop_vinfo,
> + enum vect_induction_op_type induction_type)
> +{
> + tree niters_skip;
> + /* Init_expr will be update by vect_update_ivs_after_vectorizer,
> + if niters or vf is unkown:
> + For shift, when shift mount >= precision, there would be UD.
> + For mult, don't known how to generate
> + init_expr * pow (step, niters) for variable niters.
> + For neg, it should be ok, since niters of vectorized main loop
> + will always be multiple of 2. */
> + if ((!LOOP_VINFO_NITERS_KNOWN_P (loop_vinfo)
> + || !LOOP_VINFO_VECT_FACTOR (loop_vinfo).is_constant ())
> + && induction_type != vect_step_op_neg)
> + {
> + if (dump_enabled_p ())
> + dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
> + "Peeling for epilogue is not supported"
> + " for nonlinear induction except neg"
> + " when iteration count is unknown.\n");
> + return false;
> + }
> +
> + /* Also doens't support peel for neg when niter is variable.
> + ??? generate something like niter_expr & 1 ? init_expr : -init_expr? */
> + niters_skip = LOOP_VINFO_MASK_SKIP_NITERS (loop_vinfo);
> + if ((niters_skip != NULL_TREE
> + && TREE_CODE (niters_skip) != INTEGER_CST)
> + || (!vect_use_loop_mask_for_alignment_p (loop_vinfo)
> + && LOOP_VINFO_PEELING_FOR_ALIGNMENT (loop_vinfo) < 0))
> + {
> + if (dump_enabled_p ())
> + dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
> + "Peeling for alignement is not supported"
> + " for nonlinear induction when niters_skip"
> + " is not constant.\n");
> + return false;
> + }
> +
> + return true;
> +}
> +
> /* Function vect_can_advance_ivs_p
>
> In case the number of iterations that LOOP iterates is unknown at compile
> diff --git a/gcc/tree-vect-loop.cc b/gcc/tree-vect-loop.cc
> index f0801c23671..01b60a8de33 100644
> --- a/gcc/tree-vect-loop.cc
> +++ b/gcc/tree-vect-loop.cc
> @@ -8808,49 +8808,6 @@ vect_update_nonlinear_iv (gimple_seq* stmts, tree vectype,
>
> }
>
> -/* Return true if vectorizer can peel for nonlinear iv. */
> -bool
> -vect_can_peel_nonlinear_iv_p (loop_vec_info loop_vinfo,
> - enum vect_induction_op_type induction_type)
> -{
> - tree niters_skip;
> - /* Init_expr will be update by vect_update_ivs_after_vectorizer,
> - if niters is unkown:
> - For shift, when shift mount >= precision, there would be UD.
> - For mult, don't known how to generate
> - init_expr * pow (step, niters) for variable niters.
> - For neg, it should be ok, since niters of vectorized main loop
> - will always be multiple of 2. */
> - if (!LOOP_VINFO_NITERS_KNOWN_P (loop_vinfo)
> - && induction_type != vect_step_op_neg)
> - {
> - if (dump_enabled_p ())
> - dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
> - "Peeling for epilogue is not supported"
> - " for nonlinear induction except neg"
> - " when iteration count is unknown.\n");
> - return false;
> - }
> -
> - /* Also doens't support peel for neg when niter is variable.
> - ??? generate something like niter_expr & 1 ? init_expr : -init_expr? */
> - niters_skip = LOOP_VINFO_MASK_SKIP_NITERS (loop_vinfo);
> - if ((niters_skip != NULL_TREE
> - && TREE_CODE (niters_skip) != INTEGER_CST)
> - || (!vect_use_loop_mask_for_alignment_p (loop_vinfo)
> - && LOOP_VINFO_PEELING_FOR_ALIGNMENT (loop_vinfo) < 0))
> - {
> - if (dump_enabled_p ())
> - dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
> - "Peeling for alignement is not supported"
> - " for nonlinear induction when niters_skip"
> - " is not constant.\n");
> - return false;
> - }
> -
> - return true;
> -}
> -
> /* Function vectorizable_induction
>
> Check if STMT_INFO performs an nonlinear induction computation that can be
> @@ -8921,9 +8878,6 @@ vectorizable_nonlinear_induction (loop_vec_info loop_vinfo,
> return false;
> }
>
> - if (!vect_can_peel_nonlinear_iv_p (loop_vinfo, induction_type))
> - return false;
> -
> if (!INTEGRAL_TYPE_P (TREE_TYPE (vectype)))
> {
> if (dump_enabled_p ())
> diff --git a/gcc/tree-vectorizer.h b/gcc/tree-vectorizer.h
> index a2aa71bedc0..4ba653712e9 100644
> --- a/gcc/tree-vectorizer.h
> +++ b/gcc/tree-vectorizer.h
> @@ -2347,9 +2347,6 @@ extern tree cse_and_gimplify_to_preheader (loop_vec_info, tree);
> /* Nonlinear induction. */
> extern tree vect_peel_nonlinear_iv_init (gimple_seq*, tree, tree,
> tree, enum vect_induction_op_type);
> -extern bool
> -vect_can_peel_nonlinear_iv_p (loop_vec_info loop_vinfo,
> - enum vect_induction_op_type induction_type);
>
> /* In tree-vect-slp.cc. */
> extern void vect_slp_init (void);
>
--
Richard Biener <rguenther@suse.de>
SUSE Software Solutions Germany GmbH, Frankenstrasse 146, 90461 Nuernberg,
Germany; GF: Ivo Totev, Andrew Myers, Andrew McDonald, Boudien Moerman;
HRB 36809 (AG Nuernberg)
prev parent reply other threads:[~2023-02-02 8:45 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-02-02 1:30 liuhongt
2023-02-02 8:45 ` Richard Biener [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=nycvar.YFH.7.77.849.2302020845350.6551@jbgna.fhfr.qr \
--to=rguenther@suse.de \
--cc=gcc-patches@gcc.gnu.org \
--cc=hongtao.liu@intel.com \
--cc=tamar.christina@arm.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).