fdo-mirrors/cairo
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/cairo/cairo.git synced 2026-05-21 09:48:10 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions
2531 commits 12 branches 118 tags 122 MiB
Commit graph

2531 commits

This branch
This branch
All branches
Author SHA1 Message Date
Behdad Esfahbod
3d398b1a21 [test/Makefile] Unset DIST_SUBDIRS 2006-09-06 13:49:53 -04:00
Behdad Esfahbod
c4d863afad [perf/Makefile] Unset DIST_SUBDIRS 2006-09-06 13:49:27 -04:00
Behdad Esfahbod
6b377b62ac Update 2006-09-06 13:49:02 -04:00
Behdad Esfahbod
94c361cd2a [Makefile] Add perf to DIST_SUBDIRS 2006-09-06 13:47:37 -04:00
Behdad Esfahbod
8c9ea020ca [Makefile] Fix typo 2006-09-06 12:45:07 -04:00
Carl Worth
a7844338da perf: Add tessellate test case (in -16, -64, and -256 varieties) 2006-09-06 05:17:01 -07:00
Carl Worth
221c02098f perf: Allow CAIRO_TEST_TARGET to limit targets tested. 
Also, don't print the image size for a test if there is only one.
 2006-09-06 05:09:19 -07:00
Carl Worth
b866069210 perf: Handle 32-bit wraparound of performance counter 2006-09-06 05:07:08 -07:00
Carl Worth
b64645c567 perf: Fix cairo_perf_ticks_per_second to avoid wraparound 2006-09-06 03:32:07 -07:00
Carl Worth
d9082d59db Prefer CPU performance counters (if available) over gettimeofday. 
I've seen this improve the std. deviation often by a factor of 2
and occasionally up to a factor of 10. It is sometimes not much
better, but never seems to be appreciably worse compared to using
gettimeofday.

Thanks to David A. Schleef <ds@schleef.org> and his liboil for
the implementation.
 2006-09-06 01:37:13 -07:00
Carl Worth
1085d99bc3 perf/README: Update due to cairo_perf_timer API changes 2006-09-06 00:56:56 -07:00
Carl Worth
3c407aa80a perf-paint: Spend more iterations on smaller sizes to balance testing. 
The values here are chosen to try to equalize the standard deviation
of the various tests.
 2006-09-06 00:53:52 -07:00
Carl Worth
3519887f18 perf: Move sorting and discarding outside of compute_stats. Adjust discard to slowest 15% only. 2006-09-06 00:52:06 -07:00
Carl Worth
d31037e421 perf: Report times not rates. Interface in integers not doubles. 2006-09-06 00:15:49 -07:00
Carl Worth
9d0d38e0a0 perf: Rename functions in line with cairo's naming guidelines 2006-09-05 22:58:33 -07:00
Carl Worth
9f51fcc888 perf: Fold cairo-perf-timer.h into cairo-perf.h 2006-09-05 22:56:00 -07:00
Carl Worth
499a3a7c47 perf: Remove unused alarm functions 2006-09-05 22:52:00 -07:00
Benjamin Otte
df8cc10073 perf: Make cairo_perf_timer structure private. Make timer functions void. 2006-09-05 22:48:38 -07:00
Benjamin Otte
1bb6f9fb10 perf: Add yield and fix double comparison 
- add a yield () function that's called before every test. It reduced the std
dev slightly for me
- fix double comparisons to not just compare the integer part
 2006-09-05 22:36:56 -07:00
Carl Worth
bcb7863f00 Move target tolerance to cairo_test_target structure (should let single-pixel SVG errors pass) 
Previously we were setting the target tolerance based on the surface
type. But that doesn't work as multiple backends will provide a surface
of type meta. So instead we put the tolerance as a value in the
cairo_test_target data structure.

With this change, some single-pixel errors of 1 in the SVG backend
should now be ignored.
 2006-09-05 22:25:21 -07:00
Carl Worth
4620b929ae RELEASING: Add note on checking for local symbol PLT entries. 2006-09-05 17:22:13 -07:00
Carl Worth
60c4add26a slim_hidden_proto: Move smeicolon from definition to use for consistency and legibility 2006-09-05 17:22:07 -07:00
Carl Worth
d7f08794ef Add many missing slim_hidden calls to bypass PLT entries for local use of public functions 2006-09-05 17:22:00 -07:00
Carl Worth
e06246b9b1 pixman: Add pixman_private decorations to hide pixman symbols from public interface 2006-09-05 17:21:48 -07:00
Adrian Johnson
d716020db2 Generate Type 1 fonts from glyph outlines 
This patch generates Type 1 fonts for the PS/PDF backends when TrueType
or Type 1 subsetting fails. This has the advantage over the current
Type 3 fallback of reduced font size and better quality rendering
in some PDF viewers. xpdf shows a large improvement in text display
quality with this patch.
 2006-09-05 15:57:47 -04:00
Adrian Johnson
ad78eb7692 Improve error checking in TrueType subsetting 
TrueType subsetting will SEGV when using OpenType fonts with CFF outlines.
 2006-09-05 14:53:33 -04:00
Adrian Johnson
49fa8d353b Correct the value of /LastChar in the PDF Type 1 font dictionary. 2006-09-05 14:19:48 -04:00
Adrian Johnson
d81de15804 Fix Type 1 embedding in PDF 
PDF Files with Type 1 fonts fail to open in any version of
ghostscript prior to 8.54. The problem is the hex encoding of the
encrypted portion of the font. The PDF reference says this should
only be in binary.
 2006-09-05 14:19:26 -04:00
Behdad Esfahbod
e5a9c23308 [configure] Invalidate cached warning flags if list of flags changes 
such that one doesn't need to remove config.cache when when we add
new warning flags to be checked.
 2006-09-02 19:52:38 -04:00
Carl Worth
7c97a787eb bug 8104: Eliminate unused variables. Replicate assert statement to identify branch of interest. 2006-09-01 16:10:25 -07:00
Carl Worth
0a1aef157e perf/README: Fix typo 2006-08-31 18:25:04 -07:00
Carl Worth
91d5541705 perf: Add README file explaining how to add a new test 2006-08-31 18:11:25 -07:00
Carl Worth
2fa0228d63 Two big improvements to bring the std. deviation down to where we want it. 
1. Remove all the alarm/signal code, which just isn't doing what we want for some reason.
   Instead, for now we'll simply run for a fixed number of iterations, (perhaps we
   can tune that per test later).

2. Before computing mean and stdandard deviation of runs, sort them all and discard the
   top and bottom 20% of the values.

Now the standard deviation for the paint test is generally 2% or less.
 2006-08-31 17:43:40 -07:00
Carl Worth
6ae6d91c0c perf: Switch from alarm to setitimer for more fine-grained control of timers 2006-08-31 14:08:43 -07:00
Benjamin Otte
2ebb9af434 boilerplate: Fix compilation for glitz. 2006-08-31 13:30:10 -07:00
Carl Worth
19a5b8b9b5 perf: Run for multiple iterations and print std. deviation 2006-08-31 13:22:17 -07:00
Carl Worth
689e9c446e perf: Add indication of content to output 2006-08-31 12:34:21 -07:00
Carl Worth
7917e1201f perf: Fix typo in backend blackballing so that image tests now run 2006-08-31 11:53:16 -07:00
Carl Worth
0c741675e1 Make perf interface return a rate. Start print target and test names. 2006-08-31 11:51:28 -07:00
Carl Worth
fd13e874a7 perf: Drop separate setup function from each test case. 
We are already doing loop measurement internally, so each function
can already do any setup it needs without it affecting the measurement.
 2006-08-31 11:42:51 -07:00
Carl Worth
578b74473d perf: More consistency improvements for names. 
Rename bench_timer_t to cairo_perf_timer_t.
Rename PERF_LOOP macros to CAIRO_PERF_LOOP
Rename DECL_PERF_FUNC to CAIRO_PERF_DECL
 2006-08-31 11:38:23 -07:00
Carl Worth
cf75da4842 perf: Rename timer-alarm files to cairo-perf-timer 2006-08-31 11:38:23 -07:00
Behdad Esfahbod
7ad6e94101 [test] Use FcFreeTypeQuery 2006-08-31 14:22:10 -04:00
Carl Worth
7f6e901a23 perf: More simplification of PERF_LOOP macros. 
Here we drop both the start_timing and stop_timing functions,
foliding their contents into main, PERF_LOOP_INIT and
PERF_LOOP_FINI.
 2006-08-31 11:19:48 -07:00
Carl Worth
b9f629d542 perf: Don't require a separate counter from the timer for perf loops. 2006-08-31 11:02:20 -07:00
Carl Worth
13bcba68ae perf: Collapse timing.[ch] down into cairo-perf.[ch] 2006-08-31 10:41:00 -07:00
Carl Worth
a60ed68dae perf: Remove unused util.c 2006-08-31 10:33:24 -07:00
Carl Worth
739c476767 Initial hookup of Vlad's timer/alarm code 2006-08-31 10:29:52 -07:00
Vladimir Vukicevic
3d279da621 Add useful pieces from Vladimir's cairo-bench 2006-08-31 10:28:06 -07:00
Behdad Esfahbod
851dd63719 [test] Use $(srcdir) to find valgrind supressions 2006-08-31 13:18:47 -04:00