public inbox for gcc-cvs@sourceware.org help / color / mirror / Atom feed
From: Richard Biener <rguenth@gcc.gnu.org> To: gcc-cvs@gcc.gnu.org Subject: [gcc(refs/users/rguenth/heads/vect-force-slp)] Fail vectorization when not SLP with --param vect-force-slp=1 Date: Thu, 9 Nov 2023 13:02:58 +0000 (GMT) [thread overview] Message-ID: <20231109130258.16A91385DC17@sourceware.org> (raw) https://gcc.gnu.org/g:98219d2638765a521c7ab61c7d9559ccba0cf4b4 commit 98219d2638765a521c7ab61c7d9559ccba0cf4b4 Author: Richard Biener <rguenther@suse.de> Date: Wed Aug 23 10:48:32 2023 +0200 Fail vectorization when not SLP with --param vect-force-slp=1 The following adds --param vect-force-slp allowing to indicate failure when not all stmts participating in loop vectorization are using SLP vectorization. This is intended for transitioning and debugging. Enabling this without further changes results in the following within vect.exp on x86_64 === g++ Summary === -# of expected passes 619 +# of expected passes 546 +# of unexpected failures 73 === gcc Summary === -# of expected passes 8835 -# of expected failures 256 +# of expected passes 7271 +# of unexpected failures 1564 +# of unexpected successes 12 +# of expected failures 244 === gfortran Summary === -# of expected passes 171 +# of expected passes 144 +# of unexpected failures 27 * params.opt (-param=vect-force-slp=): New, default to 0. * doc/invoke.texi (--param vect-force-slp): Document. * tree-vect-stmts.cc (vect_analyze_stmt): With --param vect-force-slp=1 fail vectorization when not using SLP. Diff: --- gcc/doc/invoke.texi | 4 ++++ gcc/params.opt | 4 ++++ gcc/tree-vect-stmts.cc | 6 ++++++ 3 files changed, 14 insertions(+) diff --git a/gcc/doc/invoke.texi b/gcc/doc/invoke.texi index 1bf71ce47a8..5ebe045c593 100644 --- a/gcc/doc/invoke.texi +++ b/gcc/doc/invoke.texi @@ -16357,6 +16357,10 @@ this parameter. The default value of this parameter is 50. @item vect-induction-float Enable loop vectorization of floating point inductions. +@item vect-force-slp +Fail vectorization when falling back to non-SLP. This is intended for +debugging only. + @item vrp-sparse-threshold Maximum number of basic blocks before VRP uses a sparse bitmap cache. diff --git a/gcc/params.opt b/gcc/params.opt index f1202abc00d..d0b7a6d770b 100644 --- a/gcc/params.opt +++ b/gcc/params.opt @@ -1191,6 +1191,10 @@ The maximum factor which the loop vectorizer applies to the cost of statements i Common Joined UInteger Var(param_vect_induction_float) Init(1) IntegerRange(0, 1) Param Optimization Enable loop vectorization of floating point inductions. +-param=vect-force-slp= +Common Joined UInteger Var(param_vect_force_slp) Init(0) IntegerRange(0, 1) Param Optimization +Fail vectorization when falling back to non-SLP. + -param=vrp-sparse-threshold= Common Joined UInteger Var(param_vrp_sparse_threshold) Init(3000) Optimization Param Maximum number of basic blocks before VRP uses a sparse bitmap cache. diff --git a/gcc/tree-vect-stmts.cc b/gcc/tree-vect-stmts.cc index ee89f47c468..0ca675ad47c 100644 --- a/gcc/tree-vect-stmts.cc +++ b/gcc/tree-vect-stmts.cc @@ -12882,6 +12882,12 @@ vect_analyze_stmt (vec_info *vinfo, return opt_result::success (); } + if (param_vect_force_slp && !node) + return opt_result::failure_at (stmt_info->stmt, + "not vectorized:" + " not part of SLP but SLP forced: %G", + stmt_info->stmt); + ok = true; if (!bb_vinfo && (STMT_VINFO_RELEVANT_P (stmt_info)
next reply other threads:[~2023-11-09 13:02 UTC|newest] Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top 2023-11-09 13:02 Richard Biener [this message] -- strict thread matches above, loose matches on Subject: below -- 2024-05-13 14:26 Richard Biener 2024-02-23 7:31 Richard Biener 2023-11-02 13:59 Richard Biener 2023-10-19 13:28 Richard Biener 2023-10-16 12:49 Richard Biener 2023-10-06 7:06 Richard Biener
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=20231109130258.16A91385DC17@sourceware.org \ --to=rguenth@gcc.gnu.org \ --cc=gcc-cvs@gcc.gnu.org \ /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: linkBe 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).