git/list[1] front-page[2] threads[3] people[4] search[5] about
 

[PATCH v4 16/19] grep: add support for the PCRE v1 JIT API

From
Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Date
Apr 25, 2017, 21:05 UTC
Message-ID
<20170425210548.24612-17-avarab@gmail.com>
In-Reply-To
<20170425210548.24612-1-avarab@gmail.com>

Change the grep PCRE v1 code to use JIT when available. When PCRE support was initially added in commit 63e7e9d8b6 ("git-grep: Learn PCRE", 2011-05-09) PCRE had no JIT support, it was integrated into 8.20 released on 2011-10-21.

When JIT support is enabled the PCRE performance usually improves by more than 50%. The pattern compilation times are relatively slower, but those relative numbers are tiny, and are easily made back in all but the most trivial cases of grep. Detailed benchhmarks are available at: http://sljit.sourceforge.net/pcre.html

With this change the difference in a t/perf/p7820-grep-engines.sh run is, shown with git --word-diff:

    7820.1: extended with how.to                               [-0.28(1.23+0.44)-]{+0.28(1.18+0.39)+}
    7820.2: extended with ^how to                              [-0.26(1.15+0.38)-]{+0.27(1.13+0.40)+}
    7820.3: extended with \w+our\w*                            [-6.06(38.44+0.35)-]{+6.11(38.66+0.32)+}
    7820.4: extended with -?-?-?-?-?-?-?-?-?-?-?-----------$   [-0.37(1.57+0.38)-]{+0.37(1.56+0.42)+}
    7820.5: pcre1 with how.to                                  [-0.26(1.15+0.37)-]{+0.19(0.39+0.55)+}
    7820.6: pcre1 with ^how to                                 [-0.46(2.66+0.31)-]{+0.22(0.67+0.44)+}
    7820.7: pcre1 with \w+our\w*                               [-16.42(99.42+0.48)-]{+0.51(3.05+0.24)+}
    7820.8: pcre1 with -?-?-?-?-?-?-?-?-?-?-?-----------$      [-81.52(275.37+0.41)-]{+5.16(19.31+0.33)+}

The conditional support for JIT is implemented as suggested in the pcrejit(3) man page. E.g. defining PCRE_STUDY_JIT_COMPILE to 0 if it's not present.

There's no graceful fallback if pcre_jit_stack_alloc() fails under PCRE_CONFIG_JIT, instead the program will abort. I don't think this is worth handling, it'll only fail in cases where malloc() doesn't work, in which case we're screwed anyway.

Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
---
 grep.c | 27 ++++++++++++++++++++++++++-
 grep.h |  5 +++++
 2 files changed, 31 insertions(+), 1 deletion(-)
diff --git a/grep.c b/grep.c
index c61e69deaa..8e3ba164b5 100644
--- a/grep.c
+++ b/grep.c
@@ -329,6 +329,9 @@ static void compile_pcre1_regexp(struct grep_pat *p, const struct grep_opt *opt)
 	const char *error;
 	int erroffset;
 	int options = PCRE_MULTILINE;
+#ifdef PCRE_CONFIG_JIT
+	int canjit;
+#endif
 
 	if (opt->ignore_case) {
 		if (has_non_ascii(p->pattern))
@@ -343,9 +346,19 @@ static void compile_pcre1_regexp(struct grep_pat *p, const struct grep_opt *opt)
 	if (!p->pcre1_regexp)
 		compile_regexp_failed(p, error);
 
-	p->pcre1_extra_info = pcre_study(p->pcre1_regexp, 0, &error);
+	p->pcre1_extra_info = pcre_study(p->pcre1_regexp, PCRE_STUDY_JIT_COMPILE, &error);
 	if (!p->pcre1_extra_info && error)
 		die("%s", error);
+
+#ifdef PCRE_CONFIG_JIT
+	pcre_config(PCRE_CONFIG_JIT, &canjit);
+	if (canjit == 1) {
+		p->pcre1_jit_stack = pcre_jit_stack_alloc(1, 1024 * 1024);
+		if (!p->pcre1_jit_stack)
+			die("BUG: Couldn't allocate PCRE JIT stack");
+		pcre_assign_jit_stack(p->pcre1_extra_info, NULL, p->pcre1_jit_stack);
+	}
+#endif
 }
 
 static int pcre1match(struct grep_pat *p, const char *line, const char *eol,
@@ -356,8 +369,15 @@ static int pcre1match(struct grep_pat *p, const char *line, const char *eol,
 	if (eflags & REG_NOTBOL)
 		flags |= PCRE_NOTBOL;
 
+#ifdef PCRE_CONFIG_JIT
+	ret = pcre_jit_exec(p->pcre1_regexp, p->pcre1_extra_info, line, eol - line,
+			    0, flags, ovector, ARRAY_SIZE(ovector),
+			    p->pcre1_jit_stack);
+#else
 	ret = pcre_exec(p->pcre1_regexp, p->pcre1_extra_info, line, eol - line,
 			0, flags, ovector, ARRAY_SIZE(ovector));
+#endif
+
 	if (ret < 0 && ret != PCRE_ERROR_NOMATCH)
 		die("pcre_exec failed with error code %d", ret);
 	if (ret > 0) {
@@ -372,7 +392,12 @@ static int pcre1match(struct grep_pat *p, const char *line, const char *eol,
 static void free_pcre1_regexp(struct grep_pat *p)
 {
 	pcre_free(p->pcre1_regexp);
+#ifdef PCRE_CONFIG_JIT
+	pcre_free_study(p->pcre1_extra_info);
+	pcre_jit_stack_free(p->pcre1_jit_stack);
+#else
 	pcre_free(p->pcre1_extra_info);
+#endif
 	pcre_free((void *)p->pcre1_tables);
 }
 #else /* !USE_LIBPCRE1 */
diff --git a/grep.h b/grep.h
index fa2ab9485f..29e20bf837 100644
--- a/grep.h
+++ b/grep.h
@@ -3,9 +3,13 @@
 #include "color.h"
 #ifdef USE_LIBPCRE1
 #include <pcre.h>
+#ifndef PCRE_STUDY_JIT_COMPILE
+#define PCRE_STUDY_JIT_COMPILE 0
+#endif
 #else
 typedef int pcre;
 typedef int pcre_extra;
+typedef int pcre_jit_stack;
 #endif
 #include "kwset.h"
 #include "thread-utils.h"
@@ -48,6 +52,7 @@ struct grep_pat {
 	regex_t regexp;
 	pcre *pcre1_regexp;
 	pcre_extra *pcre1_extra_info;
+	pcre_jit_stack *pcre1_jit_stack;
 	const unsigned char *pcre1_tables;
 	kwset_t kws;
 	unsigned fixed:1;
-- 
2.11.0
Previous: Ævar Arnfjörð BjarmasonNext: Ævar Arnfjörð Bjarmason
Message 21 of 24 in “PCRE v1 improvements & PCRE v2 support”
  1. 00/19 PCRE v1 improvements & PCRE v2 supportÆvar Arnfjörð Bjarmason, Apr 25, 2017
  2. 01/19 grep: amend submodule recursion test in preparation for rx engine testingÆvar Arnfjörð Bjarmason, Apr 25, 2017
  3. 02/19 grep: add tests for grep pattern types being passed to submodulesÆvar Arnfjörð Bjarmason, Apr 25, 2017
  4. 03/19 grep: submodule-related case statements should die if new fields are addedÆvar Arnfjörð Bjarmason, Apr 25, 2017
  5. 04/19 grep: remove redundant regflags assignment under PCREÆvar Arnfjörð Bjarmason, Apr 25, 2017
  6. 05/19 grep: remove redundant `regflags &= ~REG_EXTENDED` assignmentsÆvar Arnfjörð Bjarmason, Apr 25, 2017
  7. Junio C HamanoApr 26, 2017
  8. Junio C HamanoApr 26, 2017
  9. 05/19 grep: remove redundant `regflags &= ~REG_EXTENDED` assignmentsÆvar Arnfjörð Bjarmason, Apr 26, 2017
  10. Junio C HamanoApr 27, 2017
  11. 06/19 Makefile & configure: reword outdated comment about PCREÆvar Arnfjörð Bjarmason, Apr 25, 2017
  12. 07/19 grep: add a test for backreferences in PCRE patternsÆvar Arnfjörð Bjarmason, Apr 25, 2017
  13. 09/19 log: add -P as a synonym for --perl-regexpÆvar Arnfjörð Bjarmason, Apr 25, 2017
  14. 08/19 log: add exhaustive tests for pattern style options & configÆvar Arnfjörð Bjarmason, Apr 25, 2017
  15. 10/19 grep & rev-list doc: stop promising libpcre for --perl-regexpÆvar Arnfjörð Bjarmason, Apr 25, 2017
  16. 12/19 test-lib: rename the LIBPCRE prerequisite to PCREÆvar Arnfjörð Bjarmason, Apr 25, 2017
  17. 13/19 grep: change the internal PCRE macro names to be PCRE1Ævar Arnfjörð Bjarmason, Apr 25, 2017
  18. 14/19 grep: change the internal PCRE code & header names to be PCRE1Ævar Arnfjörð Bjarmason, Apr 25, 2017
  19. 15/19 perf: add a performance comparison test of grep -E and -PÆvar Arnfjörð Bjarmason, Apr 25, 2017
  20. 11/19 grep: make grep.patternType=[pcre|pcre1] a synonym for "perl"Ævar Arnfjörð Bjarmason, Apr 25, 2017
  21. 16/19 grep: add support for the PCRE v1 JIT APIÆvar Arnfjörð Bjarmason, Apr 25, 2017
  22. 17/19 grep: add support for PCRE v2Ævar Arnfjörð Bjarmason, Apr 25, 2017
  23. 18/19 grep: remove support for concurrent use of both PCRE v1 & v2Ævar Arnfjörð Bjarmason, Apr 25, 2017
  24. 19/19 Makefile & configure: make PCRE v2 the default PCRE implementationÆvar Arnfjörð Bjarmason, Apr 25, 2017

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.