git@vger.kernel.org mailing list mirror (one of many)
 help / color / mirror / code / Atom feed
From: Jeff King <peff@peff.net>
To: git@vger.kernel.org
Cc: "René Scharfe" <l.s.r@web.de>,
	"Rasmus Villemoes" <rv@rasmusvillemoes.dk>,
	"Eric Sunshine" <sunshine@sunshineco.com>
Subject: [PATCH 7/9] strbuf_getwholeline: use getdelim if it is available
Date: Thu, 16 Apr 2015 05:01:38 -0400	[thread overview]
Message-ID: <20150416090138.GG17938@peff.net> (raw)
In-Reply-To: <20150416084733.GA17811@peff.net>

We spend a lot of time in strbuf_getwholeline in a tight
loop reading characters from a stdio handle into a buffer.
The libc getdelim() function can do this for us with less
overhead. It's in POSIX.1-2008, and was a GNU extension
before that. Therefore we can't rely on it, but can fall
back to the existing getc loop when it is not available.

The HAVE_GETDELIM knob is turned on automatically for Linux,
where we have glibc. We don't need to set any new
feature-test macros, because we already define _GNU_SOURCE.
Other systems that implement getdelim may need to other
macros (probably _POSIX_C_SOURCE >= 200809L), but we can
address that along with setting the Makefile knob after
testing the feature on those systems.

Running "git rev-parse refs/heads/does-not-exist" on a repo
with an extremely large (1.6GB) packed-refs file went from
(best-of-5):

  real    0m8.601s
  user    0m8.084s
  sys     0m0.524s

to:

  real    0m6.768s
  user    0m6.340s
  sys     0m0.432s

for a wall-clock speedup of 21%.

Based on a patch from Rasmus Villemoes <rv@rasmusvillemoes.dk>.

Signed-off-by: Jeff King <peff@peff.net>
---
If somebody has a FreeBSD or OS X system to test on, I'd
love to see what is needed to compile with HAVE_GETDELIM
there. And to confirm that the performance is much better.
Sharing my 1.6GB packed-refs file would be hard, but you
should be able to generate something large and ridiculous.
I'll leave that as an exercise to the reader.

 Makefile         |  6 ++++++
 config.mak.uname |  1 +
 strbuf.c         | 42 ++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 49 insertions(+)

diff --git a/Makefile b/Makefile
index 5f3987f..36655d5 100644
--- a/Makefile
+++ b/Makefile
@@ -359,6 +359,8 @@ all::
 # compiler is detected to support it.
 #
 # Define HAVE_BSD_SYSCTL if your platform has a BSD-compatible sysctl function.
+#
+# Define HAVE_GETDELIM if your system has the getdelim() function.
 
 GIT-VERSION-FILE: FORCE
 	@$(SHELL_PATH) ./GIT-VERSION-GEN
@@ -1437,6 +1439,10 @@ ifdef HAVE_BSD_SYSCTL
 	BASIC_CFLAGS += -DHAVE_BSD_SYSCTL
 endif
 
+ifdef HAVE_GETDELIM
+	BASIC_CFLAGS += -DHAVE_GETDELIM
+endif
+
 ifeq ($(TCLTK_PATH),)
 NO_TCLTK = NoThanks
 endif
diff --git a/config.mak.uname b/config.mak.uname
index f4e77cb..d26665f 100644
--- a/config.mak.uname
+++ b/config.mak.uname
@@ -36,6 +36,7 @@ ifeq ($(uname_S),Linux)
 	HAVE_DEV_TTY = YesPlease
 	HAVE_CLOCK_GETTIME = YesPlease
 	HAVE_CLOCK_MONOTONIC = YesPlease
+	HAVE_GETDELIM = YesPlease
 endif
 ifeq ($(uname_S),GNU/kFreeBSD)
 	HAVE_ALLOCA_H = YesPlease
diff --git a/strbuf.c b/strbuf.c
index 921619e..0d4f4e5 100644
--- a/strbuf.c
+++ b/strbuf.c
@@ -435,6 +435,47 @@ int strbuf_getcwd(struct strbuf *sb)
 	return -1;
 }
 
+#ifdef HAVE_GETDELIM
+int strbuf_getwholeline(struct strbuf *sb, FILE *fp, int term)
+{
+	ssize_t r;
+
+	if (feof(fp))
+		return EOF;
+
+	strbuf_reset(sb);
+
+	/* Translate slopbuf to NULL, as we cannot call realloc on it */
+	if (!sb->alloc)
+		sb->buf = NULL;
+	r = getdelim(&sb->buf, &sb->alloc, term, fp);
+
+	if (r > 0) {
+		sb->len = r;
+		return 0;
+	}
+	assert(r == -1);
+
+	/*
+	 * Normally we would have called xrealloc, which will try to free
+	 * memory and recover. But we have no way to tell getdelim() to do so.
+	 * Worse, we cannot try to recover ENOMEM ourselves, because we have
+	 * no idea how many bytes were read by getdelim.
+	 *
+	 * Dying here is reasonable. It mirrors what xrealloc would do on
+	 * catastrophic memory failure. We skip the opportunity to free pack
+	 * memory and retry, but that's unlikely to help for a malloc small
+	 * enough to hold a single line of input, anyway.
+	 */
+	if (errno == ENOMEM)
+		die("Out of memory, getdelim failed");
+
+	/* Restore slopbuf that we moved out of the way before */
+	if (!sb->buf)
+		strbuf_init(sb, 0);
+	return EOF;
+}
+#else
 int strbuf_getwholeline(struct strbuf *sb, FILE *fp, int term)
 {
 	int ch;
@@ -458,6 +499,7 @@ int strbuf_getwholeline(struct strbuf *sb, FILE *fp, int term)
 	sb->buf[sb->len] = '\0';
 	return 0;
 }
+#endif
 
 int strbuf_getline(struct strbuf *sb, FILE *fp, int term)
 {
-- 
2.4.0.rc2.384.g7297a4a

  parent reply	other threads:[~2015-04-16  9:02 UTC|newest]

Thread overview: 44+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-04-05  1:06 [PATCH 0/6] address packed-refs speed regressions Jeff King
2015-04-05  1:07 ` [PATCH 1/6] strbuf_getwholeline: use getc macro Jeff King
2015-04-05  1:08 ` [PATCH 2/6] git-compat-util: add fallbacks for unlocked stdio Jeff King
2015-04-05  1:11 ` [PATCH 3/6] strbuf_getwholeline: use getc_unlocked Jeff King
2015-04-05  4:56   ` Jeff King
2015-04-05  5:27     ` Jeff King
2015-04-05  5:35       ` Jeff King
2015-04-05 20:49         ` Junio C Hamano
2015-04-05 14:36     ` Duy Nguyen
2015-04-05 18:24       ` Jeff King
2015-04-05 20:09     ` Junio C Hamano
2015-04-07 13:48     ` Rasmus Villemoes
2015-04-07 19:04       ` Jeff King
2015-04-07 22:43         ` Rasmus Villemoes
2015-04-08  0:17           ` Jeff King
2015-04-05  1:11 ` [PATCH 4/6] strbuf: add an optimized 1-character strbuf_grow Jeff King
2015-04-06  2:13   ` Eric Sunshine
2015-04-06  5:05     ` Jeff King
2015-04-05  1:11 ` [PATCH 5/6] t1430: add another refs-escape test Jeff King
2015-04-05  1:15 ` [PATCH 6/6] refname_is_safe: avoid expensive normalize_path_copy call Jeff King
2015-04-05 13:41 ` [PATCH 0/6] address packed-refs speed regressions René Scharfe
2015-04-05 18:52   ` Jeff King
2015-04-05 18:59     ` Jeff King
2015-04-05 23:04       ` René Scharfe
2015-04-05 22:39     ` René Scharfe
2015-04-06  4:49       ` Jeff King
2015-04-16  8:47 ` [PATCH v2 0/9] " Jeff King
2015-04-16  8:48   ` [PATCH 1/9] strbuf_getwholeline: use getc macro Jeff King
2015-04-16  8:48   ` [PATCH 2/9] git-compat-util: add fallbacks for unlocked stdio Jeff King
2015-04-16  8:49   ` [PATCH 3/9] strbuf_getwholeline: use getc_unlocked Jeff King
2015-04-16  8:51   ` [PATCH 4/9] config: use getc_unlocked when reading from file Jeff King
2015-04-16  8:53   ` [PATCH 5/9] strbuf_addch: avoid calling strbuf_grow Jeff King
2015-04-16  8:58   ` [PATCH 6/9] strbuf_getwholeline: " Jeff King
2015-04-16  9:01   ` Jeff King [this message]
2015-04-17 10:16     ` [PATCH 7/9] strbuf_getwholeline: use getdelim if it is available Eric Sunshine
2015-04-21 23:09       ` Jeff King
2015-05-08 23:56         ` Eric Sunshine
2015-05-09  1:09           ` Jeff King
2015-06-02 18:22             ` Eric Sunshine
2015-04-22 18:00       ` Johannes Schindelin
2015-04-22 18:06         ` Jeff King
2015-04-16  9:03   ` [PATCH 8/9] read_packed_refs: avoid double-checking sane refs Jeff King
2015-04-16  9:04   ` [PATCH 9/9] t1430: add another refs-escape test Jeff King
2015-04-16  9:25   ` [PATCH v2 0/9] address packed-refs speed regressions Jeff King

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

  List information: http://vger.kernel.org/majordomo-info.html

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20150416090138.GG17938@peff.net \
    --to=peff@peff.net \
    --cc=git@vger.kernel.org \
    --cc=l.s.r@web.de \
    --cc=rv@rasmusvillemoes.dk \
    --cc=sunshine@sunshineco.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.
Code repositories for project(s) associated with this public inbox

	https://80x24.org/mirrors/git.git

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).