public inbox for gcc-cvs@sourceware.org help / color / mirror / Atom feed
From: Martin Liska <marxin@gcc.gnu.org> To: gcc-cvs@gcc.gnu.org Subject: [gcc(refs/users/marxin/heads/loop-unswitching-switch-v5)] Restore old code that can handle floating point types. Date: Mon, 8 Nov 2021 15:36:44 +0000 (GMT) [thread overview] Message-ID: <20211108153644.20B843858415@sourceware.org> (raw) https://gcc.gnu.org/g:31102db9992df9b12228e36cdc1b18cc7198a733 commit 31102db9992df9b12228e36cdc1b18cc7198a733 Author: Martin Liska <mliska@suse.cz> Date: Mon Nov 8 16:29:59 2021 +0100 Restore old code that can handle floating point types. Diff: --- gcc/tree-ssa-loop-unswitch.c | 52 ++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 52 insertions(+) diff --git a/gcc/tree-ssa-loop-unswitch.c b/gcc/tree-ssa-loop-unswitch.c index 4b4bd471acd..1d169802252 100644 --- a/gcc/tree-ssa-loop-unswitch.c +++ b/gcc/tree-ssa-loop-unswitch.c @@ -300,6 +300,39 @@ tree_may_unswitch_on (basic_block bb, class loop *loop, edge *cond_edge) return NULL_TREE; } +/* Simplifies COND using checks in front of the entry of the LOOP. Just very + simplish (sufficient to prevent us from duplicating loop in unswitching + unnecessarily). */ + +static tree +simplify_using_entry_checks (class loop *loop, tree cond) +{ + edge e = loop_preheader_edge (loop); + gimple *stmt; + + while (1) + { + stmt = last_stmt (e->src); + if (stmt + && gimple_code (stmt) == GIMPLE_COND + && gimple_cond_code (stmt) == TREE_CODE (cond) + && operand_equal_p (gimple_cond_lhs (stmt), + TREE_OPERAND (cond, 0), 0) + && operand_equal_p (gimple_cond_rhs (stmt), + TREE_OPERAND (cond, 1), 0)) + return (e->flags & EDGE_TRUE_VALUE + ? boolean_true_node + : boolean_false_node); + + if (!single_pred_p (e->src)) + return cond; + + e = single_pred_edge (e->src); + if (e->src == ENTRY_BLOCK_PTR_FOR_FN (cfun)) + return cond; + } +} + /* Unswitch single LOOP. NUM is number of unswitchings done; we do not allow it to grow too much, it is too easy to create example on that the code would grow exponentially. */ @@ -393,6 +426,25 @@ tree_unswitch_single_loop (class loop *loop, int num) update_stmt (condition); changed = true; } + else + { + /* FIXME: Remove legacy code for floating point type that + is currently not supported by ranger. */ + cond = simplify_using_entry_checks (loop, cond); + if (integer_nonzerop (cond)) + { + gimple_cond_set_condition_from_tree (condition, + boolean_true_node); + changed = true; + } + else if (integer_zerop (cond)) + { + /* Remove true path. */ + gimple_cond_set_condition_from_tree (condition, + boolean_false_node); + changed = true; + } + } } else if (swtch != NULL) {
next reply other threads:[~2021-11-08 15:36 UTC|newest] Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top 2021-11-08 15:36 Martin Liska [this message] -- strict thread matches above, loose matches on Subject: below -- 2021-11-08 15:30 Martin Liska
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=20211108153644.20B843858415@sourceware.org \ --to=marxin@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).