undef/git-commit-vandalism
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
779ea9303a
git-commit-vandalism/xdiff/xprepare.c

479 lines
12 KiB
C
Raw Normal View History

Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
/*
* LibXDiff by Davide Libenzi ( File Differential Library )
* Copyright (C) 2003 Davide Libenzi
*
* This 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.
*
* This 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
Replace Free Software Foundation address in license notices The mailing address for the FSF has changed over the years. Rather than updating the address across all files, refer readers to gnu.org, as the GNU GPL documentation now suggests for license notices. The mailing address is retained in the full license files (COPYING and LGPL-2.1). The old address is still present in t/diff-lib/COPYING. This is intentional, as the file is used in tests and the contents are not expected to change. Signed-off-by: Todd Zullinger <tmz@pobox.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2017-11-07 06:39:33 +01:00
* License along with this library; if not, see
* <http://www.gnu.org/licenses/>.
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
*
* Davide Libenzi <davidel@xmailserver.org>
*
*/
#include "xinclude.h"
#define XDL_KPDIS_RUN 4
Clean-up trivially redundant diff. Also corrects the line numbers in unified output when using zero lines context.
2006-04-04 03:47:55 +02:00
#define XDL_MAX_EQLIMIT 1024
xdiff: give up scanning similar lines early In a corner case of large files whose lines do not match uniquely, the loop to eliminate a line that matches multiple locations adjacent to a run of lines that do not uniquely match wasted too much cycles. Fix this by giving up early after scanning 100 lines in both direction. Signed-off-by: Junio C Hamano <gitster@pobox.com>
2008-11-08 06:24:33 +01:00
#define XDL_SIMSCAN_WINDOW 100
xdiff/xprepare: use a smaller sample size for histogram diff For histogram diff, we can afford a smaller sample size and thus a poorer estimate of the number of lines, as the hash table (rhash) won't be filled up/grown. This is safe as the final count of lines (xdf.nrecs) will be updated correctly anyway by xdl_prepare_ctx(). This gives us a small boost in performance. Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-07-12 08:10:27 +02:00
#define XDL_GUESS_NLINES1 256
#define XDL_GUESS_NLINES2 20
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
typedef struct s_xdlclass {
struct s_xdlclass *next;
unsigned long ha;
char const *line;
long size;
long idx;
xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records() In xdl_cleanup_records(), we see O(n*m) performance, where n is the number of records from xdf->dstart to xdf->dend, and m is the size of a bucket in xdf->rhash (<= by mlim). Here, we improve this to O(n) by pre-computing nm (in rcrec->len(1|2)) in xdl_classify_record(). Reported-by: Marat Radchenko <marat@slonopotamus.org> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-17 03:53:57 +02:00
long len1, len2;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
} xdlclass_t;
typedef struct s_xdlclassifier {
unsigned int hbits;
long hsize;
xdlclass_t **rchash;
chastore_t ncha;
xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records() In xdl_cleanup_records(), we see O(n*m) performance, where n is the number of records from xdf->dstart to xdf->dend, and m is the size of a bucket in xdf->rhash (<= by mlim). Here, we improve this to O(n) by pre-computing nm (in rcrec->len(1|2)) in xdl_classify_record(). Reported-by: Marat Radchenko <marat@slonopotamus.org> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-17 03:53:57 +02:00
xdlclass_t **rcrecs;
long alloc;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
long count;
Teach diff about -b and -w flags This adds -b (--ignore-space-change) and -w (--ignore-all-space) flags to diff. The main part of the patch is teaching libxdiff about it. [jc: renamed xdl_line_match() to xdl_recmatch() since the former is used for different purposes in xpatchi.c which is in the parts of the upstream source we do not use.] Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-06-14 17:40:23 +02:00
long flags;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
} xdlclassifier_t;
Teach diff about -b and -w flags This adds -b (--ignore-space-change) and -w (--ignore-all-space) flags to diff. The main part of the patch is teaching libxdiff about it. [jc: renamed xdl_line_match() to xdl_recmatch() since the former is used for different purposes in xpatchi.c which is in the parts of the upstream source we do not use.] Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-06-14 17:40:23 +02:00
static int xdl_init_classifier(xdlclassifier_t *cf, long size, long flags);
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
static void xdl_free_classifier(xdlclassifier_t *cf);
xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records() In xdl_cleanup_records(), we see O(n*m) performance, where n is the number of records from xdf->dstart to xdf->dend, and m is the size of a bucket in xdf->rhash (<= by mlim). Here, we improve this to O(n) by pre-computing nm (in rcrec->len(1|2)) in xdl_classify_record(). Reported-by: Marat Radchenko <marat@slonopotamus.org> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-17 03:53:57 +02:00
static int xdl_classify_record(unsigned int pass, xdlclassifier_t *cf, xrecord_t **rhash,
unsigned int hbits, xrecord_t *rec);
static int xdl_prepare_ctx(unsigned int pass, mmfile_t *mf, long narec, xpparam_t const *xpp,
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
xdlclassifier_t *cf, xdfile_t *xdf);
static void xdl_free_ctx(xdfile_t *xdf);
static int xdl_clean_mmatch(char const *dis, long i, long s, long e);
xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records() In xdl_cleanup_records(), we see O(n*m) performance, where n is the number of records from xdf->dstart to xdf->dend, and m is the size of a bucket in xdf->rhash (<= by mlim). Here, we improve this to O(n) by pre-computing nm (in rcrec->len(1|2)) in xdl_classify_record(). Reported-by: Marat Radchenko <marat@slonopotamus.org> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-17 03:53:57 +02:00
static int xdl_cleanup_records(xdlclassifier_t *cf, xdfile_t *xdf1, xdfile_t *xdf2);
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
static int xdl_trim_ends(xdfile_t *xdf1, xdfile_t *xdf2);
xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records() In xdl_cleanup_records(), we see O(n*m) performance, where n is the number of records from xdf->dstart to xdf->dend, and m is the size of a bucket in xdf->rhash (<= by mlim). Here, we improve this to O(n) by pre-computing nm (in rcrec->len(1|2)) in xdl_classify_record(). Reported-by: Marat Radchenko <marat@slonopotamus.org> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-17 03:53:57 +02:00
static int xdl_optimize_ctxs(xdlclassifier_t *cf, xdfile_t *xdf1, xdfile_t *xdf2);
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
Teach diff about -b and -w flags This adds -b (--ignore-space-change) and -w (--ignore-all-space) flags to diff. The main part of the patch is teaching libxdiff about it. [jc: renamed xdl_line_match() to xdl_recmatch() since the former is used for different purposes in xpatchi.c which is in the parts of the upstream source we do not use.] Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-06-14 17:40:23 +02:00
static int xdl_init_classifier(xdlclassifier_t *cf, long size, long flags) {
cf->flags = flags;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
cf->hbits = xdl_hashbits((unsigned int) size);
cf->hsize = 1 << cf->hbits;
if (xdl_cha_init(&cf->ncha, sizeof(xdlclass_t), size / 4 + 1) < 0) {
return -1;
}
if (!(cf->rchash = (xdlclass_t **) xdl_malloc(cf->hsize * sizeof(xdlclass_t *)))) {
xdl_cha_free(&cf->ncha);
return -1;
}
xdiff/xprepare: use memset() Use memset() instead of a for loop to initialize. This could give a performance advantage. Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-07-07 06:23:55 +02:00
memset(cf->rchash, 0, cf->hsize * sizeof(xdlclass_t *));
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records() In xdl_cleanup_records(), we see O(n*m) performance, where n is the number of records from xdf->dstart to xdf->dend, and m is the size of a bucket in xdf->rhash (<= by mlim). Here, we improve this to O(n) by pre-computing nm (in rcrec->len(1|2)) in xdl_classify_record(). Reported-by: Marat Radchenko <marat@slonopotamus.org> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-17 03:53:57 +02:00
cf->alloc = size;
if (!(cf->rcrecs = (xdlclass_t **) xdl_malloc(cf->alloc * sizeof(xdlclass_t *)))) {
xdl_free(cf->rchash);
xdl_cha_free(&cf->ncha);
return -1;
}
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
cf->count = 0;
return 0;
}
static void xdl_free_classifier(xdlclassifier_t *cf) {
xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records() In xdl_cleanup_records(), we see O(n*m) performance, where n is the number of records from xdf->dstart to xdf->dend, and m is the size of a bucket in xdf->rhash (<= by mlim). Here, we improve this to O(n) by pre-computing nm (in rcrec->len(1|2)) in xdl_classify_record(). Reported-by: Marat Radchenko <marat@slonopotamus.org> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-17 03:53:57 +02:00
xdl_free(cf->rcrecs);
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
xdl_free(cf->rchash);
xdl_cha_free(&cf->ncha);
}
xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records() In xdl_cleanup_records(), we see O(n*m) performance, where n is the number of records from xdf->dstart to xdf->dend, and m is the size of a bucket in xdf->rhash (<= by mlim). Here, we improve this to O(n) by pre-computing nm (in rcrec->len(1|2)) in xdl_classify_record(). Reported-by: Marat Radchenko <marat@slonopotamus.org> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-17 03:53:57 +02:00
static int xdl_classify_record(unsigned int pass, xdlclassifier_t *cf, xrecord_t **rhash,
unsigned int hbits, xrecord_t *rec) {
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
long hi;
char const *line;
xdlclass_t *rcrec;
xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records() In xdl_cleanup_records(), we see O(n*m) performance, where n is the number of records from xdf->dstart to xdf->dend, and m is the size of a bucket in xdf->rhash (<= by mlim). Here, we improve this to O(n) by pre-computing nm (in rcrec->len(1|2)) in xdl_classify_record(). Reported-by: Marat Radchenko <marat@slonopotamus.org> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-17 03:53:57 +02:00
xdlclass_t **rcrecs;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
line = rec->ptr;
hi = (long) XDL_HASHLONG(rec->ha, cf->hbits);
for (rcrec = cf->rchash[hi]; rcrec; rcrec = rcrec->next)
Teach diff about -b and -w flags This adds -b (--ignore-space-change) and -w (--ignore-all-space) flags to diff. The main part of the patch is teaching libxdiff about it. [jc: renamed xdl_line_match() to xdl_recmatch() since the former is used for different purposes in xpatchi.c which is in the parts of the upstream source we do not use.] Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-06-14 17:40:23 +02:00
if (rcrec->ha == rec->ha &&
xdl_recmatch(rcrec->line, rcrec->size,
rec->ptr, rec->size, cf->flags))
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
break;
if (!rcrec) {
if (!(rcrec = xdl_cha_alloc(&cf->ncha))) {
return -1;
}
rcrec->idx = cf->count++;
xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records() In xdl_cleanup_records(), we see O(n*m) performance, where n is the number of records from xdf->dstart to xdf->dend, and m is the size of a bucket in xdf->rhash (<= by mlim). Here, we improve this to O(n) by pre-computing nm (in rcrec->len(1|2)) in xdl_classify_record(). Reported-by: Marat Radchenko <marat@slonopotamus.org> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-17 03:53:57 +02:00
if (cf->count > cf->alloc) {
cf->alloc *= 2;
if (!(rcrecs = (xdlclass_t **) xdl_realloc(cf->rcrecs, cf->alloc * sizeof(xdlclass_t *)))) {
return -1;
}
cf->rcrecs = rcrecs;
}
cf->rcrecs[rcrec->idx] = rcrec;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
rcrec->line = line;
rcrec->size = rec->size;
rcrec->ha = rec->ha;
xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records() In xdl_cleanup_records(), we see O(n*m) performance, where n is the number of records from xdf->dstart to xdf->dend, and m is the size of a bucket in xdf->rhash (<= by mlim). Here, we improve this to O(n) by pre-computing nm (in rcrec->len(1|2)) in xdl_classify_record(). Reported-by: Marat Radchenko <marat@slonopotamus.org> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-17 03:53:57 +02:00
rcrec->len1 = rcrec->len2 = 0;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
rcrec->next = cf->rchash[hi];
cf->rchash[hi] = rcrec;
}
xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records() In xdl_cleanup_records(), we see O(n*m) performance, where n is the number of records from xdf->dstart to xdf->dend, and m is the size of a bucket in xdf->rhash (<= by mlim). Here, we improve this to O(n) by pre-computing nm (in rcrec->len(1|2)) in xdl_classify_record(). Reported-by: Marat Radchenko <marat@slonopotamus.org> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-17 03:53:57 +02:00
(pass == 1) ? rcrec->len1++ : rcrec->len2++;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
rec->ha = (unsigned long) rcrec->idx;
hi = (long) XDL_HASHLONG(rec->ha, hbits);
rec->next = rhash[hi];
rhash[hi] = rec;
return 0;
}
xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records() In xdl_cleanup_records(), we see O(n*m) performance, where n is the number of records from xdf->dstart to xdf->dend, and m is the size of a bucket in xdf->rhash (<= by mlim). Here, we improve this to O(n) by pre-computing nm (in rcrec->len(1|2)) in xdl_classify_record(). Reported-by: Marat Radchenko <marat@slonopotamus.org> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-17 03:53:57 +02:00
static int xdl_prepare_ctx(unsigned int pass, mmfile_t *mf, long narec, xpparam_t const *xpp,
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
xdlclassifier_t *cf, xdfile_t *xdf) {
unsigned int hbits;
xdiff/xprepare: use memset() Use memset() instead of a for loop to initialize. This could give a performance advantage. Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-07-07 06:23:55 +02:00
long nrec, hsize, bsize;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
unsigned long hav;
char const *blk, *cur, *top, *prev;
xrecord_t *crec;
xrecord_t **recs, **rrecs;
xrecord_t **rhash;
unsigned long *ha;
char *rchg;
long *rindex;
xdiff/xprepare: refactor abort cleanups Group free()'s that are called when a malloc() fails in xdl_prepare_ctx(), making for more readable code. Also add a free() on ha, in case future git hackers add allocs after the ha malloc. Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-07-07 06:23:56 +02:00
ha = NULL;
rindex = NULL;
rchg = NULL;
rhash = NULL;
recs = NULL;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
xdiff/xprepare: refactor abort cleanups Group free()'s that are called when a malloc() fails in xdl_prepare_ctx(), making for more readable code. Also add a free() on ha, in case future git hackers add allocs after the ha malloc. Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-07-07 06:23:56 +02:00
if (xdl_cha_init(&xdf->rcha, sizeof(xrecord_t), narec / 4 + 1) < 0)
goto abort;
if (!(recs = (xrecord_t **) xdl_malloc(narec * sizeof(xrecord_t *))))
goto abort;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
diff histogram: intern strings Histogram is the only diff algorithm not to call xdl_classify_record(). xdl_classify_record() ensures that the hash values of two strings that are not equal differ which means that it is not necessary to use xdl_recmatch() when comparing lines, all that is necessary is to compare the hash values. This gives a 7% reduction in the runtime of "git log --patch" when using the histogram diff algorithm. Test HEAD^ HEAD ----------------------------------------------------------------------------- 4000.1: log -3000 (baseline) 0.18(0.14+0.04) 0.19(0.17+0.02) +5.6% 4000.2: log --raw -3000 (tree-only) 0.99(0.77+0.21) 0.98(0.78+0.20) -1.0% 4000.3: log -p -3000 (Myers) 4.84(4.31+0.51) 4.81(4.15+0.64) -0.6% 4000.4: log -p -3000 --histogram 6.34(5.86+0.46) 5.87(5.19+0.66) -7.4% 4000.5: log -p -3000 --patience 5.39(4.60+0.76) 5.35(4.60+0.73) -0.7% Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-11-17 12:20:23 +01:00
hbits = xdl_hashbits((unsigned int) narec);
hsize = 1 << hbits;
if (!(rhash = (xrecord_t **) xdl_malloc(hsize * sizeof(xrecord_t *))))
goto abort;
memset(rhash, 0, hsize * sizeof(xrecord_t *));
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
nrec = 0;
tree-wide: apply equals-null.cocci Signed-off-by: Junio C Hamano <gitster@pobox.com>
2022-05-02 18:50:37 +02:00
if ((cur = blk = xdl_mmfile_first(mf, &bsize))) {
xdiff: do away with xdl_mmfile_next() Given our simple mmfile structure, xdl_mmfile_next() calls are redundant. Do away with calls to them. Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-01 06:20:07 +02:00
for (top = blk + bsize; cur < top; ) {
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
prev = cur;
Teach diff about -b and -w flags This adds -b (--ignore-space-change) and -w (--ignore-all-space) flags to diff. The main part of the patch is teaching libxdiff about it. [jc: renamed xdl_line_match() to xdl_recmatch() since the former is used for different purposes in xpatchi.c which is in the parts of the upstream source we do not use.] Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-06-14 17:40:23 +02:00
hav = xdl_hash_record(&cur, top, xpp->flags);
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
if (nrec >= narec) {
narec *= 2;
xdiff/xprepare: refactor abort cleanups Group free()'s that are called when a malloc() fails in xdl_prepare_ctx(), making for more readable code. Also add a free() on ha, in case future git hackers add allocs after the ha malloc. Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-07-07 06:23:56 +02:00
if (!(rrecs = (xrecord_t **) xdl_realloc(recs, narec * sizeof(xrecord_t *))))
goto abort;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
recs = rrecs;
}
xdiff/xprepare: refactor abort cleanups Group free()'s that are called when a malloc() fails in xdl_prepare_ctx(), making for more readable code. Also add a free() on ha, in case future git hackers add allocs after the ha malloc. Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-07-07 06:23:56 +02:00
if (!(crec = xdl_cha_alloc(&xdf->rcha)))
goto abort;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
crec->ptr = prev;
crec->size = (long) (cur - prev);
crec->ha = hav;
recs[nrec++] = crec;
diff histogram: intern strings Histogram is the only diff algorithm not to call xdl_classify_record(). xdl_classify_record() ensures that the hash values of two strings that are not equal differ which means that it is not necessary to use xdl_recmatch() when comparing lines, all that is necessary is to compare the hash values. This gives a 7% reduction in the runtime of "git log --patch" when using the histogram diff algorithm. Test HEAD^ HEAD ----------------------------------------------------------------------------- 4000.1: log -3000 (baseline) 0.18(0.14+0.04) 0.19(0.17+0.02) +5.6% 4000.2: log --raw -3000 (tree-only) 0.99(0.77+0.21) 0.98(0.78+0.20) -1.0% 4000.3: log -p -3000 (Myers) 4.84(4.31+0.51) 4.81(4.15+0.64) -0.6% 4000.4: log -p -3000 --histogram 6.34(5.86+0.46) 5.87(5.19+0.66) -7.4% 4000.5: log -p -3000 --patience 5.39(4.60+0.76) 5.35(4.60+0.73) -0.7% Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-11-17 12:20:23 +01:00
if (xdl_classify_record(pass, cf, rhash, hbits, crec) < 0)
xdiff/xprepare: refactor abort cleanups Group free()'s that are called when a malloc() fails in xdl_prepare_ctx(), making for more readable code. Also add a free() on ha, in case future git hackers add allocs after the ha malloc. Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-07-07 06:23:56 +02:00
goto abort;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
}
}
xdiff/xprepare: refactor abort cleanups Group free()'s that are called when a malloc() fails in xdl_prepare_ctx(), making for more readable code. Also add a free() on ha, in case future git hackers add allocs after the ha malloc. Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-07-07 06:23:56 +02:00
if (!(rchg = (char *) xdl_malloc((nrec + 2) * sizeof(char))))
goto abort;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
memset(rchg, 0, (nrec + 2) * sizeof(char));
xdiff: avoid unnecessary memory allocations rindex and ha are only used by xdl_cleanup_records() which is not called by the histogram or patience algorithms. The perf test results show a small reduction in run time but that is probably within the noise. Test HEAD^ HEAD ----------------------------------------------------------------------------- 4000.1: log -3000 (baseline) 0.19(0.17+0.02) 0.19(0.12+0.07) +0.0% 4000.2: log --raw -3000 (tree-only) 0.98(0.78+0.20) 0.98(0.81+0.16) +0.0% 4000.3: log -p -3000 (Myers) 4.81(4.15+0.64) 4.81(4.23+0.56) +0.0% 4000.4: log -p -3000 --histogram 5.87(5.19+0.66) 5.83(5.11+0.70) -0.7% 4000.5: log -p -3000 --patience 5.35(4.60+0.73) 5.31(4.61+0.69) -0.7% Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-11-17 12:20:24 +01:00
if ((XDF_DIFF_ALG(xpp->flags) != XDF_PATIENCE_DIFF) &&
(XDF_DIFF_ALG(xpp->flags) != XDF_HISTOGRAM_DIFF)) {
if (!(rindex = xdl_malloc((nrec + 1) * sizeof(*rindex))))
goto abort;
if (!(ha = xdl_malloc((nrec + 1) * sizeof(*ha))))
goto abort;
}
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
xdf->nrec = nrec;
xdf->recs = recs;
xdf->hbits = hbits;
xdf->rhash = rhash;
xdf->rchg = rchg + 1;
xdf->rindex = rindex;
xdf->nreff = 0;
xdf->ha = ha;
xdf->dstart = 0;
xdf->dend = nrec - 1;
return 0;
xdiff/xprepare: refactor abort cleanups Group free()'s that are called when a malloc() fails in xdl_prepare_ctx(), making for more readable code. Also add a free() on ha, in case future git hackers add allocs after the ha malloc. Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-07-07 06:23:56 +02:00
abort:
xdl_free(ha);
xdl_free(rindex);
xdl_free(rchg);
xdl_free(rhash);
xdl_free(recs);
xdl_cha_free(&xdf->rcha);
return -1;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
}
static void xdl_free_ctx(xdfile_t *xdf) {
xdl_free(xdf->rhash);
xdl_free(xdf->rindex);
xdl_free(xdf->rchg - 1);
xdl_free(xdf->ha);
xdl_free(xdf->recs);
xdl_cha_free(&xdf->rcha);
}
int xdl_prepare_env(mmfile_t *mf1, mmfile_t *mf2, xpparam_t const *xpp,
xdfenv_t *xe) {
xdiff/xprepare: use a smaller sample size for histogram diff For histogram diff, we can afford a smaller sample size and thus a poorer estimate of the number of lines, as the hash table (rhash) won't be filled up/grown. This is safe as the final count of lines (xdf.nrecs) will be updated correctly anyway by xdl_prepare_ctx(). This gives us a small boost in performance. Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-07-12 08:10:27 +02:00
long enl1, enl2, sample;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
xdlclassifier_t cf;
xdiff/xprepare: initialise xdlclassifier_t cf in xdl_prepare_env() Ensure that the xdl_free_classifier() call on xdlclassifier_t cf is safe even if xdl_init_classifier() isn't called. This may occur in the case where diff is run with --histogram and a call to, say, xdl_prepare_ctx() fails. Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-31 06:48:46 +02:00
memset(&cf, 0, sizeof(cf));
xdiff/xprepare: use a smaller sample size for histogram diff For histogram diff, we can afford a smaller sample size and thus a poorer estimate of the number of lines, as the hash table (rhash) won't be filled up/grown. This is safe as the final count of lines (xdf.nrecs) will be updated correctly anyway by xdl_prepare_ctx(). This gives us a small boost in performance. Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-07-12 08:10:27 +02:00
/*
* For histogram diff, we can afford a smaller sample size and
* thus a poorer estimate of the number of lines, as the hash
* table (rhash) won't be filled up/grown. The number of lines
* (nrecs) will be updated correctly anyway by
* xdl_prepare_ctx().
*/
xdiff: PATIENCE/HISTOGRAM are not independent option bits Because the default Myers, patience and histogram algorithms cannot be in effect at the same time, XDL_PATIENCE_DIFF and XDL_HISTOGRAM_DIFF are not independent bits. Instead of wasting one bit per algorithm, define a few macros to access the few bits they occupy and update the code that access them. Signed-off-by: Junio C Hamano <gitster@pobox.com>
2012-02-20 00:36:55 +01:00
sample = (XDF_DIFF_ALG(xpp->flags) == XDF_HISTOGRAM_DIFF
? XDL_GUESS_NLINES2 : XDL_GUESS_NLINES1);
xdiff/xprepare: use a smaller sample size for histogram diff For histogram diff, we can afford a smaller sample size and thus a poorer estimate of the number of lines, as the hash table (rhash) won't be filled up/grown. This is safe as the final count of lines (xdf.nrecs) will be updated correctly anyway by xdl_prepare_ctx(). This gives us a small boost in performance. Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-07-12 08:10:27 +02:00
enl1 = xdl_guess_lines(mf1, sample) + 1;
enl2 = xdl_guess_lines(mf2, sample) + 1;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
diff histogram: intern strings Histogram is the only diff algorithm not to call xdl_classify_record(). xdl_classify_record() ensures that the hash values of two strings that are not equal differ which means that it is not necessary to use xdl_recmatch() when comparing lines, all that is necessary is to compare the hash values. This gives a 7% reduction in the runtime of "git log --patch" when using the histogram diff algorithm. Test HEAD^ HEAD ----------------------------------------------------------------------------- 4000.1: log -3000 (baseline) 0.18(0.14+0.04) 0.19(0.17+0.02) +5.6% 4000.2: log --raw -3000 (tree-only) 0.99(0.77+0.21) 0.98(0.78+0.20) -1.0% 4000.3: log -p -3000 (Myers) 4.84(4.31+0.51) 4.81(4.15+0.64) -0.6% 4000.4: log -p -3000 --histogram 6.34(5.86+0.46) 5.87(5.19+0.66) -7.4% 4000.5: log -p -3000 --patience 5.39(4.60+0.76) 5.35(4.60+0.73) -0.7% Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-11-17 12:20:23 +01:00
if (xdl_init_classifier(&cf, enl1 + enl2 + 1, xpp->flags) < 0)
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
return -1;
xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records() In xdl_cleanup_records(), we see O(n*m) performance, where n is the number of records from xdf->dstart to xdf->dend, and m is the size of a bucket in xdf->rhash (<= by mlim). Here, we improve this to O(n) by pre-computing nm (in rcrec->len(1|2)) in xdl_classify_record(). Reported-by: Marat Radchenko <marat@slonopotamus.org> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-17 03:53:57 +02:00
if (xdl_prepare_ctx(1, mf1, enl1, xpp, &cf, &xe->xdf1) < 0) {
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
xdl_free_classifier(&cf);
return -1;
}
xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records() In xdl_cleanup_records(), we see O(n*m) performance, where n is the number of records from xdf->dstart to xdf->dend, and m is the size of a bucket in xdf->rhash (<= by mlim). Here, we improve this to O(n) by pre-computing nm (in rcrec->len(1|2)) in xdl_classify_record(). Reported-by: Marat Radchenko <marat@slonopotamus.org> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-17 03:53:57 +02:00
if (xdl_prepare_ctx(2, mf2, enl2, xpp, &cf, &xe->xdf2) < 0) {
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
xdl_free_ctx(&xe->xdf1);
xdl_free_classifier(&cf);
return -1;
}
xdiff: PATIENCE/HISTOGRAM are not independent option bits Because the default Myers, patience and histogram algorithms cannot be in effect at the same time, XDL_PATIENCE_DIFF and XDL_HISTOGRAM_DIFF are not independent bits. Instead of wasting one bit per algorithm, define a few macros to access the few bits they occupy and update the code that access them. Signed-off-by: Junio C Hamano <gitster@pobox.com>
2012-02-20 00:36:55 +01:00
if ((XDF_DIFF_ALG(xpp->flags) != XDF_PATIENCE_DIFF) &&
(XDF_DIFF_ALG(xpp->flags) != XDF_HISTOGRAM_DIFF) &&
xdl_optimize_ctxs(&cf, &xe->xdf1, &xe->xdf2) < 0) {
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
xdl_free_ctx(&xe->xdf2);
xdl_free_ctx(&xe->xdf1);
xdiff/xprepare: fix a memory leak The xdl_prepare_env() function may initialise an xdlclassifier_t data structure via xdl_init_classifier(), which allocates memory to several fields, for example 'rchash', 'rcrecs' and 'ncha'. If this function later exits due to the failure of xdl_optimize_ctxs(), then this xdlclassifier_t structure, and the memory allocated to it, is not cleaned up. In order to fix the memory leak, insert a call to xdl_free_classifier() before returning. Signed-off-by: Ramsay Jones <ramsay@ramsayjones.plus.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-03-05 00:10:50 +01:00
xdl_free_classifier(&cf);
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
return -1;
}
diff histogram: intern strings Histogram is the only diff algorithm not to call xdl_classify_record(). xdl_classify_record() ensures that the hash values of two strings that are not equal differ which means that it is not necessary to use xdl_recmatch() when comparing lines, all that is necessary is to compare the hash values. This gives a 7% reduction in the runtime of "git log --patch" when using the histogram diff algorithm. Test HEAD^ HEAD ----------------------------------------------------------------------------- 4000.1: log -3000 (baseline) 0.18(0.14+0.04) 0.19(0.17+0.02) +5.6% 4000.2: log --raw -3000 (tree-only) 0.99(0.77+0.21) 0.98(0.78+0.20) -1.0% 4000.3: log -p -3000 (Myers) 4.84(4.31+0.51) 4.81(4.15+0.64) -0.6% 4000.4: log -p -3000 --histogram 6.34(5.86+0.46) 5.87(5.19+0.66) -7.4% 4000.5: log -p -3000 --patience 5.39(4.60+0.76) 5.35(4.60+0.73) -0.7% Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-11-17 12:20:23 +01:00
xdl_free_classifier(&cf);
xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records() In xdl_cleanup_records(), we see O(n*m) performance, where n is the number of records from xdf->dstart to xdf->dend, and m is the size of a bucket in xdf->rhash (<= by mlim). Here, we improve this to O(n) by pre-computing nm (in rcrec->len(1|2)) in xdl_classify_record(). Reported-by: Marat Radchenko <marat@slonopotamus.org> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-17 03:53:57 +02:00
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
return 0;
}
void xdl_free_env(xdfenv_t *xe) {
xdl_free_ctx(&xe->xdf2);
xdl_free_ctx(&xe->xdf1);
}
static int xdl_clean_mmatch(char const *dis, long i, long s, long e) {
Clean-up trivially redundant diff. Also corrects the line numbers in unified output when using zero lines context.
2006-04-04 03:47:55 +02:00
long r, rdis0, rpdis0, rdis1, rpdis1;
xdiff: give up scanning similar lines early In a corner case of large files whose lines do not match uniquely, the loop to eliminate a line that matches multiple locations adjacent to a run of lines that do not uniquely match wasted too much cycles. Fix this by giving up early after scanning 100 lines in both direction. Signed-off-by: Junio C Hamano <gitster@pobox.com>
2008-11-08 06:24:33 +01:00
/*
* Limits the window the is examined during the similar-lines
* scan. The loops below stops when dis[i - r] == 1 (line that
* has no match), but there are corner cases where the loop
* proceed all the way to the extremities by causing huge
* performance penalties in case of big files.
*/
if (i - s > XDL_SIMSCAN_WINDOW)
s = i - XDL_SIMSCAN_WINDOW;
if (e - i > XDL_SIMSCAN_WINDOW)
e = i + XDL_SIMSCAN_WINDOW;
Clean-up trivially redundant diff. Also corrects the line numbers in unified output when using zero lines context.
2006-04-04 03:47:55 +02:00
/*
* Scans the lines before 'i' to find a run of lines that either
* have no match (dis[j] == 0) or have multiple matches (dis[j] > 1).
* Note that we always call this function with dis[i] > 1, so the
* current line (i) is already a multimatch line.
*/
for (r = 1, rdis0 = 0, rpdis0 = 1; (i - r) >= s; r++) {
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
if (!dis[i - r])
Clean-up trivially redundant diff. Also corrects the line numbers in unified output when using zero lines context.
2006-04-04 03:47:55 +02:00
rdis0++;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
else if (dis[i - r] == 2)
Clean-up trivially redundant diff. Also corrects the line numbers in unified output when using zero lines context.
2006-04-04 03:47:55 +02:00
rpdis0++;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
else
break;
}
Clean-up trivially redundant diff. Also corrects the line numbers in unified output when using zero lines context.
2006-04-04 03:47:55 +02:00
/*
* If the run before the line 'i' found only multimatch lines, we
* return 0 and hence we don't make the current line (i) discarded.
* We want to discard multimatch lines only when they appear in the
* middle of runs with nomatch lines (dis[j] == 0).
*/
if (rdis0 == 0)
return 0;
for (r = 1, rdis1 = 0, rpdis1 = 1; (i + r) <= e; r++) {
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
if (!dis[i + r])
Clean-up trivially redundant diff. Also corrects the line numbers in unified output when using zero lines context.
2006-04-04 03:47:55 +02:00
rdis1++;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
else if (dis[i + r] == 2)
Clean-up trivially redundant diff. Also corrects the line numbers in unified output when using zero lines context.
2006-04-04 03:47:55 +02:00
rpdis1++;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
else
break;
}
Clean-up trivially redundant diff. Also corrects the line numbers in unified output when using zero lines context.
2006-04-04 03:47:55 +02:00
/*
* If the run after the line 'i' found only multimatch lines, we
* return 0 and hence we don't make the current line (i) discarded.
*/
if (rdis1 == 0)
return 0;
rdis1 += rdis0;
rpdis1 += rpdis0;
return rpdis1 * XDL_KPDIS_RUN < (rpdis1 + rdis1);
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
}
/*
* Try to reduce the problem complexity, discard records that have no
* matches on the other file. Also, lines that have multiple matches
* might be potentially discarded if they happear in a run of discardable.
*/
xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records() In xdl_cleanup_records(), we see O(n*m) performance, where n is the number of records from xdf->dstart to xdf->dend, and m is the size of a bucket in xdf->rhash (<= by mlim). Here, we improve this to O(n) by pre-computing nm (in rcrec->len(1|2)) in xdl_classify_record(). Reported-by: Marat Radchenko <marat@slonopotamus.org> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-17 03:53:57 +02:00
static int xdl_cleanup_records(xdlclassifier_t *cf, xdfile_t *xdf1, xdfile_t *xdf2) {
Revert removal of multi-match discard heuristic in 27af01 27af01d (xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records(), 2011-08-17) was supposed to be a performance boost only. However, it unexpectedly changed the behaviour of diff. Revert a part of 27af01d that removes logic that mark lines as "multi-match" (ie. dis[i] == 2). This was preventing the multi-match discard heuristic (performed in xdl_cleanup_records() and xdl_clean_mmatch()) from executing. Reported-by: Alexander Pepper <pepper@inf.fu-berlin.de> Signed-off-by: René Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-09-25 15:39:08 +02:00
long i, nm, nreff, mlim;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
xrecord_t **recs;
xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records() In xdl_cleanup_records(), we see O(n*m) performance, where n is the number of records from xdf->dstart to xdf->dend, and m is the size of a bucket in xdf->rhash (<= by mlim). Here, we improve this to O(n) by pre-computing nm (in rcrec->len(1|2)) in xdl_classify_record(). Reported-by: Marat Radchenko <marat@slonopotamus.org> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-17 03:53:57 +02:00
xdlclass_t *rcrec;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
char *dis, *dis1, *dis2;
Clean-up trivially redundant diff. Also corrects the line numbers in unified output when using zero lines context.
2006-04-04 03:47:55 +02:00
if (!(dis = (char *) xdl_malloc(xdf1->nrec + xdf2->nrec + 2))) {
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
return -1;
}
Clean-up trivially redundant diff. Also corrects the line numbers in unified output when using zero lines context.
2006-04-04 03:47:55 +02:00
memset(dis, 0, xdf1->nrec + xdf2->nrec + 2);
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
dis1 = dis;
dis2 = dis1 + xdf1->nrec + 1;
Revert removal of multi-match discard heuristic in 27af01 27af01d (xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records(), 2011-08-17) was supposed to be a performance boost only. However, it unexpectedly changed the behaviour of diff. Revert a part of 27af01d that removes logic that mark lines as "multi-match" (ie. dis[i] == 2). This was preventing the multi-match discard heuristic (performed in xdl_cleanup_records() and xdl_clean_mmatch()) from executing. Reported-by: Alexander Pepper <pepper@inf.fu-berlin.de> Signed-off-by: René Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-09-25 15:39:08 +02:00
if ((mlim = xdl_bogosqrt(xdf1->nrec)) > XDL_MAX_EQLIMIT)
mlim = XDL_MAX_EQLIMIT;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
for (i = xdf1->dstart, recs = &xdf1->recs[xdf1->dstart]; i <= xdf1->dend; i++, recs++) {
xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records() In xdl_cleanup_records(), we see O(n*m) performance, where n is the number of records from xdf->dstart to xdf->dend, and m is the size of a bucket in xdf->rhash (<= by mlim). Here, we improve this to O(n) by pre-computing nm (in rcrec->len(1|2)) in xdl_classify_record(). Reported-by: Marat Radchenko <marat@slonopotamus.org> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-17 03:53:57 +02:00
rcrec = cf->rcrecs[(*recs)->ha];
nm = rcrec ? rcrec->len2 : 0;
Revert removal of multi-match discard heuristic in 27af01 27af01d (xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records(), 2011-08-17) was supposed to be a performance boost only. However, it unexpectedly changed the behaviour of diff. Revert a part of 27af01d that removes logic that mark lines as "multi-match" (ie. dis[i] == 2). This was preventing the multi-match discard heuristic (performed in xdl_cleanup_records() and xdl_clean_mmatch()) from executing. Reported-by: Alexander Pepper <pepper@inf.fu-berlin.de> Signed-off-by: René Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-09-25 15:39:08 +02:00
dis1[i] = (nm == 0) ? 0: (nm >= mlim) ? 2: 1;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
}
Revert removal of multi-match discard heuristic in 27af01 27af01d (xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records(), 2011-08-17) was supposed to be a performance boost only. However, it unexpectedly changed the behaviour of diff. Revert a part of 27af01d that removes logic that mark lines as "multi-match" (ie. dis[i] == 2). This was preventing the multi-match discard heuristic (performed in xdl_cleanup_records() and xdl_clean_mmatch()) from executing. Reported-by: Alexander Pepper <pepper@inf.fu-berlin.de> Signed-off-by: René Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-09-25 15:39:08 +02:00
if ((mlim = xdl_bogosqrt(xdf2->nrec)) > XDL_MAX_EQLIMIT)
mlim = XDL_MAX_EQLIMIT;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
for (i = xdf2->dstart, recs = &xdf2->recs[xdf2->dstart]; i <= xdf2->dend; i++, recs++) {
xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records() In xdl_cleanup_records(), we see O(n*m) performance, where n is the number of records from xdf->dstart to xdf->dend, and m is the size of a bucket in xdf->rhash (<= by mlim). Here, we improve this to O(n) by pre-computing nm (in rcrec->len(1|2)) in xdl_classify_record(). Reported-by: Marat Radchenko <marat@slonopotamus.org> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-17 03:53:57 +02:00
rcrec = cf->rcrecs[(*recs)->ha];
nm = rcrec ? rcrec->len1 : 0;
Revert removal of multi-match discard heuristic in 27af01 27af01d (xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records(), 2011-08-17) was supposed to be a performance boost only. However, it unexpectedly changed the behaviour of diff. Revert a part of 27af01d that removes logic that mark lines as "multi-match" (ie. dis[i] == 2). This was preventing the multi-match discard heuristic (performed in xdl_cleanup_records() and xdl_clean_mmatch()) from executing. Reported-by: Alexander Pepper <pepper@inf.fu-berlin.de> Signed-off-by: René Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-09-25 15:39:08 +02:00
dis2[i] = (nm == 0) ? 0: (nm >= mlim) ? 2: 1;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
}
for (nreff = 0, i = xdf1->dstart, recs = &xdf1->recs[xdf1->dstart];
i <= xdf1->dend; i++, recs++) {
if (dis1[i] == 1 ||
(dis1[i] == 2 && !xdl_clean_mmatch(dis1, i, xdf1->dstart, xdf1->dend))) {
xdf1->rindex[nreff] = i;
xdf1->ha[nreff] = (*recs)->ha;
nreff++;
} else
xdf1->rchg[i] = 1;
}
xdf1->nreff = nreff;
for (nreff = 0, i = xdf2->dstart, recs = &xdf2->recs[xdf2->dstart];
i <= xdf2->dend; i++, recs++) {
if (dis2[i] == 1 ||
(dis2[i] == 2 && !xdl_clean_mmatch(dis2, i, xdf2->dstart, xdf2->dend))) {
xdf2->rindex[nreff] = i;
xdf2->ha[nreff] = (*recs)->ha;
nreff++;
} else
xdf2->rchg[i] = 1;
}
xdf2->nreff = nreff;
xdl_free(dis);
return 0;
}
/*
* Early trim initial and terminal matching records.
*/
static int xdl_trim_ends(xdfile_t *xdf1, xdfile_t *xdf2) {
long i, lim;
xrecord_t **recs1, **recs2;
recs1 = xdf1->recs;
recs2 = xdf2->recs;
for (i = 0, lim = XDL_MIN(xdf1->nrec, xdf2->nrec); i < lim;
i++, recs1++, recs2++)
if ((*recs1)->ha != (*recs2)->ha)
break;
xdf1->dstart = xdf2->dstart = i;
recs1 = xdf1->recs + xdf1->nrec - 1;
recs2 = xdf2->recs + xdf2->nrec - 1;
for (lim -= i, i = 0; i < lim; i++, recs1--, recs2--)
if ((*recs1)->ha != (*recs2)->ha)
break;
xdf1->dend = xdf1->nrec - i - 1;
xdf2->dend = xdf2->nrec - i - 1;
return 0;
}
xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records() In xdl_cleanup_records(), we see O(n*m) performance, where n is the number of records from xdf->dstart to xdf->dend, and m is the size of a bucket in xdf->rhash (<= by mlim). Here, we improve this to O(n) by pre-computing nm (in rcrec->len(1|2)) in xdl_classify_record(). Reported-by: Marat Radchenko <marat@slonopotamus.org> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-17 03:53:57 +02:00
static int xdl_optimize_ctxs(xdlclassifier_t *cf, xdfile_t *xdf1, xdfile_t *xdf2) {
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
if (xdl_trim_ends(xdf1, xdf2) < 0 ||
xdiff/xprepare: improve O(n*m) performance in xdl_cleanup_records() In xdl_cleanup_records(), we see O(n*m) performance, where n is the number of records from xdf->dstart to xdf->dend, and m is the size of a bucket in xdf->rhash (<= by mlim). Here, we improve this to O(n) by pre-computing nm (in rcrec->len(1|2)) in xdl_classify_record(). Reported-by: Marat Radchenko <marat@slonopotamus.org> Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-17 03:53:57 +02:00
xdl_cleanup_records(cf, xdf1, xdf2) < 0) {
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
return -1;
}
return 0;
}
Reference in New Issue Copy Permalink