This CL changes our convergence logic from
- did the last run take more than x milliseconds?
to
- did the last run take more x milliseconds and are the last two runs within y% of each other?
There's also now an upper limit where we bail out with an error if we haven't yet met the convergence criteria. Keeping the lower bound is important for benches where the constant overhead is much larger than the work done in the loop; without it we'll see T(1 loop) == T(2 loops) and converge way too early.
This CL also exposed that DeferredCanvasBench had a bug: it was running N^2 loops when we told it to run N. (My fault.)
I threw in a couple other linty changes that I'd be happy to split off.
BUG=
R=robertphillips@google.com
Review URL: https://codereview.chromium.org/23536046
git-svn-id: http://skia.googlecode.com/svn/trunk@11267 2bbb7eff-a529-9590-31e7-b0007b416f81
The objective of this change is to remove all calls to
SkCanvas::setDevice. The factory API is hidden behind
a build flag in order to ease the roll into chromium.
A side-effect of the factory pattern is that it will
no longer be possible to allocate a SkDeferredCanvas on
the stack. This changes nothing for chrome, but it
impacts skia test programs.
Review URL: https://codereview.chromium.org/16040002
git-svn-id: http://skia.googlecode.com/svn/trunk@9298 2bbb7eff-a529-9590-31e7-b0007b416f81
This adds the same set of options in unit tests, i.e:
--match [~][^]match[$] [~][^]match[$] ...
~ causes a matching test to always be skipped
^ requires the start of the test to match
$ requires the end of the test to match
^ and $ requires an exact match
If a test does not match any list entry,
it is skipped unless some list entry starts with ~
Review URL: https://codereview.chromium.org/14746017
git-svn-id: http://skia.googlecode.com/svn/trunk@9096 2bbb7eff-a529-9590-31e7-b0007b416f81
Currently benches that set fIsRendering = false are run as part of the first config run. This is problematic for several reasons.
1) Which config they are run as depends on the --config options passed to bench. If I run bench --config GPU --config 8888 then they will be run in the GPU config and not the 8888 config.
2) Their presence makes bench take longer to run when testing rendering changes or comparing one rendering config to another (e.g. GPU v 8888). This is especially true on Android.
3) When comparing runs of a single config across multiple bench runs (typically with code changes) it isn't obvious from the output which benchs to ignore because they say are listed under as being part of the config.
This CL adds NONRENDERING as a config (open to name suggestions). Non-rendering tests are run in and only in the NONRENDERING config.
Will this screw up our bench bots/scripts/graphs? If so I think we should change them since running non-rendering tests separately makes a lot more sense than the current madness.
Review URL: https://codereview.chromium.org/12475012
git-svn-id: http://skia.googlecode.com/svn/trunk@8174 2bbb7eff-a529-9590-31e7-b0007b416f81
There were only a few warnings left, so I fixed them and enabled the unused variable warning by removing the "-Wno-unused" flag. Only the -Wno-unused-parameter remains for now (could be removed later).
Review URL: https://codereview.chromium.org/12480002
git-svn-id: http://skia.googlecode.com/svn/trunk@8030 2bbb7eff-a529-9590-31e7-b0007b416f81
Share common code between bench_ and render_ to set up the PictureRenderer.
Fix an include error in SkPictureRenderer.h.
Simplified parameter passing in render_pictures_main.
Switch to using an SkAutoTUnref for the PictureRenderer.
I also changed the input format somewhat, so the buildbots need to be updated as well: https://codereview.appspot.com/7441044/
Fixed a bug in PictureBenchmark where calling setTimeIndividualTiles(false) sets the member variable to true.
Removed setDeviceType from PictureBenchmark, since only the PictureRenderer needs to know which device type to use.
Some changes to the input format:
'--logPerIter' no longer takes a 1 or 0. Instead, '--logPerIter'
turns it on and '--nologPerIter' turns it off (with off as the
default). (Note that this is for bench_pictures; bench still uses the old format)
Change '--device' to '--config' and 'bitmap' to '8888' to be
the same as gm.
Requires '--r' before inputs (to match gm), though there can be multiple inputs following it.
Changed --enable-deferred-image-decoding (which no one uses but me yet anyway) to --deferImageDecoding, since the former is incompatible with the flag parser.
Changes to behavior:
Show a short error message on failure (rather than the explanation of all
flags).
BUG=https://code.google.com/p/skia/issues/detail?id=1094
Review URL: https://codereview.appspot.com/7230053
git-svn-id: http://skia.googlecode.com/svn/trunk@7961 2bbb7eff-a529-9590-31e7-b0007b416f81
Chrome uses this flag for recording to skpicture in order to improve performance.
Therefore, skai benchmarks should run with this flag enabled, and we need
gm and render_pictures test coverage to validate it. In gm, the vanilla
SkPicture test step will still run without the flag to ensure that case still gets test
coverage, while the SkPicture test steps that use rtree and tileGrid will now run with
the flag enabled.
Review URL: https://codereview.appspot.com/7111043
git-svn-id: http://skia.googlecode.com/svn/trunk@7234 2bbb7eff-a529-9590-31e7-b0007b416f81
- Add nacl_make script to build Skia targets for NaCl using gyp
- Add nacl_interface for command-line apps
- Add nacl_sample as front-end for SampleApp
- Add freetype to DEPS
- Various gyp tweaks for NaCl
TODO:
- Implement GL interface
- Implement font host
- Fix plumbing so that SampleApp works properly
Review URL: https://codereview.appspot.com/6671044
git-svn-id: http://skia.googlecode.com/svn/trunk@6245 2bbb7eff-a529-9590-31e7-b0007b416f81
