From: "Martin Liška" <mliska@suse.cz>
To: Jan Hubicka <hubicka@ucw.cz>
Cc: Nick Desaulniers <ndesaulniers@google.com>, gcc-patches@gcc.gnu.org
Subject: Re: [PATCH] inline: do not inline when no_profile_instrument_function is different
Date: Wed, 18 Aug 2021 10:38:26 +0200 [thread overview]
Message-ID: <eead9e43-623e-9452-f69a-b22e160406ce@suse.cz> (raw)
In-Reply-To: <7fed627d-7213-918c-3ca6-0db60623a32b@suse.cz>
[-- Attachment #1: Type: text/plain, Size: 534 bytes --]
On 6/23/21 3:15 PM, Martin Liška wrote:
> On 6/23/21 2:38 PM, Jan Hubicka wrote:
>> Is there reason to prevent the inlining once instrumentation is done?
>
> No ;)
>
>> I think you can just block it for early inliner.
>
> Sure. Do you have a handy predicate function that tells if einliner is done?
>
> Thanks,
> Martin
Hello.
There's updated version of that patch that blocks inlining only during einline IPA pass.
Patch can bootstrap on x86_64-linux-gnu and survives regression tests.
Ready to be installed?
Thanks,
Martin
[-- Attachment #2: 0001-inline-do-not-einline-when-no_profile_instrument_fun.patch --]
[-- Type: text/x-patch, Size: 2395 bytes --]
From e2adaff9ed92bcc380e1368418da5ad2801fef4e Mon Sep 17 00:00:00 2001
From: Martin Liska <mliska@suse.cz>
Date: Tue, 22 Jun 2021 10:09:01 +0200
Subject: [PATCH] inline: do not einline when no_profile_instrument_function is
different
PR gcov-profile/80223
gcc/ChangeLog:
* ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
options, do not inline when no_profile_instrument_function
attributes are different in early inliner. It's fine to inline
it after PGO instrumentation.
gcc/testsuite/ChangeLog:
* gcc.dg/no_profile_instrument_function-attr-2.c: New test.
---
gcc/ipa-inline.c | 17 +++++++++++++++++
.../no_profile_instrument_function-attr-2.c | 15 +++++++++++++++
2 files changed, 32 insertions(+)
create mode 100644 gcc/testsuite/gcc.dg/no_profile_instrument_function-attr-2.c
diff --git a/gcc/ipa-inline.c b/gcc/ipa-inline.c
index 413446bcc46..012b326b5e9 100644
--- a/gcc/ipa-inline.c
+++ b/gcc/ipa-inline.c
@@ -396,6 +396,23 @@ can_inline_edge_p (struct cgraph_edge *e, bool report,
e->inline_failed = CIF_SANITIZE_ATTRIBUTE_MISMATCH;
inlinable = false;
}
+ else if (profile_arc_flag
+ && (lookup_attribute ("no_profile_instrument_function",
+ DECL_ATTRIBUTES (caller->decl)) == NULL_TREE)
+ != (lookup_attribute ("no_profile_instrument_function",
+ DECL_ATTRIBUTES (callee->decl)) == NULL_TREE))
+ {
+ cgraph_node *origin = caller;
+ while (origin->clone_of)
+ origin = origin->clone_of;
+
+ if (!DECL_STRUCT_FUNCTION (origin->decl)->always_inline_functions_inlined)
+ {
+ e->inline_failed = CIF_UNSPECIFIED;
+ inlinable = false;
+ }
+ }
+
if (!inlinable && report)
report_inline_failed_reason (e);
return inlinable;
diff --git a/gcc/testsuite/gcc.dg/no_profile_instrument_function-attr-2.c b/gcc/testsuite/gcc.dg/no_profile_instrument_function-attr-2.c
new file mode 100644
index 00000000000..472eca88efd
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/no_profile_instrument_function-attr-2.c
@@ -0,0 +1,15 @@
+/* { dg-require-effective-target global_constructor } */
+/* { dg-options "-O2 -fprofile-generate -fprofile-update=single -fdump-tree-optimized" } */
+
+__attribute__ ((no_profile_instrument_function))
+int foo()
+{
+ return 0;
+}
+
+int bar()
+{
+ return foo();
+}
+
+/* { dg-final { scan-tree-dump-not" = foo \\(\\)" "optimized"} } */
--
2.32.0
next prev parent reply other threads:[~2021-08-18 8:38 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-06-23 11:53 Martin Liška
2021-06-23 12:38 ` Jan Hubicka
2021-06-23 13:15 ` Martin Liška
2021-06-25 19:00 ` Nick Desaulniers
2021-08-18 8:38 ` Martin Liška [this message]
2021-09-06 17:06 ` Jeff Law
2021-09-07 20:41 ` [Committed] Fix fatal typo in gcc.dg/no_profile_instrument_function-attr-2.c Hans-Peter Nilsson
2021-09-08 7:29 ` Martin Liška
2021-09-08 11:25 ` Hans-Peter Nilsson
2021-09-08 11:31 ` Rainer Orth
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=eead9e43-623e-9452-f69a-b22e160406ce@suse.cz \
--to=mliska@suse.cz \
--cc=gcc-patches@gcc.gnu.org \
--cc=hubicka@ucw.cz \
--cc=ndesaulniers@google.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).