From: Andrew Pinski <quic_apinski@quicinc.com>
To: <gcc-patches@gcc.gnu.org>
Cc: Andrew Pinski <quic_apinski@quicinc.com>
Subject: [PATCH] tree-cfg: Move the returns_twice check to be last statement only [PR114301]
Date: Tue, 14 May 2024 08:51:32 -0700 [thread overview]
Message-ID: <20240514155132.2890624-1-quic_apinski@quicinc.com> (raw)
When I was checking to making sure that all of the bugs dealing
with the case where gimple_can_duplicate_bb_p would return false was fixed,
I noticed that the code which was checking if a call statement was
returns_twice was checking all call statements rather than just the
last statement. Since calling gimple_call_flags has a small non-zero
overhead due to a few string comparison, removing the uses of it
can have a small performance improvement. In the case of returns_twice
functions calls, will always end the basic-block due to the check in
stmt_can_terminate_bb_p (and others). So checking only the last statement
is a small optimization and will be safe.
Bootstrapped and tested pon x86_64-linux-gnu with no regressions.
PR tree-optimization/114301
gcc/ChangeLog:
* tree-cfg.cc (gimple_can_duplicate_bb_p): Check returns_twice
only on the last call statement rather than all.
Signed-off-by: Andrew Pinski <quic_apinski@quicinc.com>
---
gcc/tree-cfg.cc | 14 +++++++++-----
1 file changed, 9 insertions(+), 5 deletions(-)
diff --git a/gcc/tree-cfg.cc b/gcc/tree-cfg.cc
index b2d47b72084..7fb7b92966b 100644
--- a/gcc/tree-cfg.cc
+++ b/gcc/tree-cfg.cc
@@ -6495,6 +6495,13 @@ gimple_can_duplicate_bb_p (const_basic_block bb)
&& gimple_call_internal_p (last)
&& gimple_call_internal_unique_p (last))
return false;
+
+ /* Prohibit duplication of returns_twice calls, otherwise associated
+ abnormal edges also need to be duplicated properly.
+ return_twice functions will always be the last statement. */
+ if (is_gimple_call (last)
+ && (gimple_call_flags (last) & ECF_RETURNS_TWICE))
+ return false;
}
for (gimple_stmt_iterator gsi = gsi_start_bb (CONST_CAST_BB (bb));
@@ -6502,15 +6509,12 @@ gimple_can_duplicate_bb_p (const_basic_block bb)
{
gimple *g = gsi_stmt (gsi);
- /* Prohibit duplication of returns_twice calls, otherwise associated
- abnormal edges also need to be duplicated properly.
- An IFN_GOMP_SIMT_ENTER_ALLOC/IFN_GOMP_SIMT_EXIT call must be
+ /* An IFN_GOMP_SIMT_ENTER_ALLOC/IFN_GOMP_SIMT_EXIT call must be
duplicated as part of its group, or not at all.
The IFN_GOMP_SIMT_VOTE_ANY and IFN_GOMP_SIMT_XCHG_* are part of such a
group, so the same holds there. */
if (is_gimple_call (g)
- && (gimple_call_flags (g) & ECF_RETURNS_TWICE
- || gimple_call_internal_p (g, IFN_GOMP_SIMT_ENTER_ALLOC)
+ && (gimple_call_internal_p (g, IFN_GOMP_SIMT_ENTER_ALLOC)
|| gimple_call_internal_p (g, IFN_GOMP_SIMT_EXIT)
|| gimple_call_internal_p (g, IFN_GOMP_SIMT_VOTE_ANY)
|| gimple_call_internal_p (g, IFN_GOMP_SIMT_XCHG_BFLY)
--
2.34.1
next reply other threads:[~2024-05-14 15:51 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-05-14 15:51 Andrew Pinski [this message]
2024-05-15 8:02 ` 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=20240514155132.2890624-1-quic_apinski@quicinc.com \
--to=quic_apinski@quicinc.com \
--cc=gcc-patches@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: 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).