From: Xi Ruoyao <xry111@xry111.site>
To: libc-alpha@sourceware.org
Cc: caiyinyu <caiyinyu@loongson.cn>, Wang Xuerui <i@xen0n.name>,
Adhemerval Zanella Netto <adhemerval.zanella@linaro.org>,
Andreas Schwab <schwab@suse.de>, Xi Ruoyao <xry111@xry111.site>
Subject: [PATCH] linux: Allow avoiding va_list for generic syscall and use it for LoongArch
Date: Fri, 24 Mar 2023 15:27:45 +0800 [thread overview]
Message-ID: <20230324072745.4138-1-xry111@xry111.site> (raw)
Currently GCC generates highly sub-optimal code on architectures where
the calling convention prefers registers for arugment passing. This is
GCC PR100955. While it's technically a missed-optimization in GCC, it
seems not trivial to fix (I've not seen any compiler which can optimize
this properly yet).
As the generic Linux syscall actually uses a fixed number of arguments,
we can avoid va_list if possible and make the compiler do right thing.
Add a macro __ASSUME_SYSCALL_NAMED_WORKS which should be defined if:
* The generic Linux syscall implementation will be used.
* The calling convention has at least 7 GARs.
* The kernel ABI uses registers for syscall numbers and arguments.
LoongArch is benefited from this (saving about 430 CPU cycles per
syscall, though I won't call it a significant improvement because
syscall is "slow" in nature). And in the future we may switch more
ports to use the generic syscall without a performance regression,
reducing the number of target-specific syscall.{c,S} files we need to
maintain.
---
.../sysv/linux/loongarch/kernel-features.h | 25 +++++++++++++
sysdeps/unix/sysv/linux/syscall.c | 35 ++++++++++++++-----
2 files changed, 52 insertions(+), 8 deletions(-)
create mode 100644 sysdeps/unix/sysv/linux/loongarch/kernel-features.h
diff --git a/sysdeps/unix/sysv/linux/loongarch/kernel-features.h b/sysdeps/unix/sysv/linux/loongarch/kernel-features.h
new file mode 100644
index 0000000000..4a1c115831
--- /dev/null
+++ b/sysdeps/unix/sysv/linux/loongarch/kernel-features.h
@@ -0,0 +1,25 @@
+/* Set flags signalling availability of kernel features based on given
+ kernel version number.
+ Copyright (C) 2023 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library. If not, see
+ <https://www.gnu.org/licenses/>. */
+
+#include_next <kernel-features.h>
+
+/* Define this if the calling convention for passing 7 named arguments is
+ same as passing one named argument and 6 variable arguments, and the
+ kernel ABI uses registers for syscall number and arguments. */
+#define __ASSUME_SYSCALL_NAMED_WORKS 1
diff --git a/sysdeps/unix/sysv/linux/syscall.c b/sysdeps/unix/sysv/linux/syscall.c
index a5a2843b73..ed5ad5afd5 100644
--- a/sysdeps/unix/sysv/linux/syscall.c
+++ b/sysdeps/unix/sysv/linux/syscall.c
@@ -16,9 +16,33 @@
License along with the GNU C Library. If not, see
<https://www.gnu.org/licenses/>. */
-#include <stdarg.h>
#include <sysdep.h>
+#ifndef __ASSUME_SYSCALL_NAMED_WORKS
+#include <stdarg.h>
+#endif
+
+static inline long int
+__syscall (long int number, long int a0, long int a1, long int a2, long int a3,
+ long int a4, long int a5)
+{
+ long int r = INTERNAL_SYSCALL_NCS_CALL (number, a0, a1, a2, a3, a4, a5);
+ if (__glibc_unlikely (INTERNAL_SYSCALL_ERROR_P (r)))
+ {
+ __set_errno (-r);
+ return -1;
+ }
+ return r;
+}
+
+#ifdef __ASSUME_SYSCALL_NAMED_WORKS
+long int
+syscall (long int number, long int a0, long int a1, long int a2, long int a3,
+ long int a4, long int a5)
+{
+ return __syscall (number, a0, a1, a2, a3, a4, a5);
+}
+#else
long int
syscall (long int number, ...)
{
@@ -33,11 +57,6 @@ syscall (long int number, ...)
long int a5 = va_arg (args, long int);
va_end (args);
- long int r = INTERNAL_SYSCALL_NCS_CALL (number, a0, a1, a2, a3, a4, a5);
- if (__glibc_unlikely (INTERNAL_SYSCALL_ERROR_P (r)))
- {
- __set_errno (-r);
- return -1;
- }
- return r;
+ return __syscall (number, a0, a1, a2, a3, a4, a5);
}
+#endif
--
2.39.2
next reply other threads:[~2023-03-24 7:29 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-03-24 7:27 Xi Ruoyao [this message]
2023-03-24 9:47 ` Florian Weimer
2023-03-24 10:36 ` Xi Ruoyao
2023-03-24 10:45 ` Florian Weimer
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=20230324072745.4138-1-xry111@xry111.site \
--to=xry111@xry111.site \
--cc=adhemerval.zanella@linaro.org \
--cc=caiyinyu@loongson.cn \
--cc=i@xen0n.name \
--cc=libc-alpha@sourceware.org \
--cc=schwab@suse.de \
/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).