Issue 5486054: two loops for setting stage descs

Keyboard Shortcuts

	File
u :	up to issue
m :	publish + mail comments
M :	edit review message
j / k :	jump to file after / before current file
J / K :	jump to next file with a comment after / before current file
	Side-by-side diff
i :	toggle intra-line diffs
e :	expand all comments
c :	collapse all comments
s :	toggle showing all comments
n / p :	next / previous diff chunk or comment
N / P :	next / previous comment
<Up> / <Down> :	next / previous line
<Enter> :	respond to / edit current comment
d :	mark current comment as done

	Issue
u :	up to list of issues
m :	publish + mail comments
j / k :	jump to patch after / before current patch
o / <Enter> :	open current patch in side-by-side view
i :	open current patch in unified diff view

	Issue List
j / k :	jump to issue after / before current issue
o / <Enter> :	open current issue
# :	close issue

	Comment/message editing
<Ctrl> + s or <Ctrl> + Enter :	save comment
<Esc> :	cancel edit

Issue 5486054: two loops for setting stage descs (Closed)

Can't Edit
Can't Publish+Mail
Start Review

Created:
12 years, 11 months ago by bsalomon

Modified:
12 years, 10 months ago

Reviewers:
TomH

Base URL:
http://skia.googlecode.com/svn/trunk/

Visibility:
Public.

Patch Set 1 #

Patch Set 2 : cleanup #

Created: 12 years, 11 months ago

Download [raw] [tar.bz2]

	Unified diffs	Side-by-side diffs	Delta from patch set	Stats (+134 lines, -108 lines)			Patch
M	src/gpu/GrGLProgram.h	View		1 chunk	+3 lines, -0 lines	0 comments	Download
M	src/gpu/GrGpuGL.h	View	1	5 chunks	+14 lines, -13 lines	0 comments	Download
M	src/gpu/GrGpuGLShaders.cpp	View		3 chunks	+117 lines, -95 lines	0 comments	Download

Messages

Total messages: 3

Expand All Messages | Collapse All Messages

TomH

Yes, like the refactoring. LGTM. What's the perf gain coming from - was there a ...

12 years, 11 months ago (2011-12-14 15:18:25 UTC) #2

bsalomon

12 years, 11 months ago (2011-12-14 15:33:19 UTC) #3

On 2011/12/14 15:18:25, TomH wrote:
> Yes, like the refactoring. LGTM.
> What's the perf gain coming from - was there a rewrite in the code that got
> pulled out into the new function? If so, it wasn't easy to spot; the logic for
> what gets copied & what gets skipped looks equivalent...

Now I'm not sure if there really was a perf gain at all given the flakiness of
the bench runs. But it looked like there were some extra store/loads around the
stage enabled bool and also in determining whether to check skipColor or
skipCoverage based on the stage index and desc.firstCoverageStage(). In any
event I'm confident the performance isn't worse and I think it is an improvement
to the code.

Expand All Messages | Collapse All Messages