144337 - Can we short-circuit QuickDraw's implicit LockPortBits()

Status: VERIFIED FIXED

(SeaMonkey :: UI Design, defect)

Description

[Robert John Churchill]

According to TechNote 2051, on Mac OS X we should be able to short-circuit
QuickDraw's implicit LockPortBits() on every QD call by stratetic usage of
::LockPortBits() and ::UnlockPortBits().

I'll attached a test patch momentarily.

Pink, care to take a gander?  In a quick jury-rigged tes, I see a 8% speedup
when forcing the entire content area to re-draw on a semi-complex HTML page
(using MacSurfer, and using the HOME/END keys to show the top/bottom).

Comment 1

[Robert John Churchill]

Attachment: Patch to bracket our drawing rtns w/LockPortBits & UnlockPortBits

Comment 2

[Mike Pinkerton (not reading bugmail)]

can you test this on the cowtools page-load tests and see what kind of perf
improvment it gives us there?

Comment 3

[Robert John Churchill]

cowtools seems to average things out much more, indicating a couple of
percentage point improvements.

Anyway, let's try this out on the trunk.  Reviews?

Comment 4

[Steve Dagley]

Comment on attachment 83457 [details] [diff] [review]
Patch to bracket our drawing rtns w/LockPortBits & UnlockPortBits

r=sdagley

Comment 5

[Ben Goodger (use ben at mozilla dot org for email)]

Comment on attachment 83457 [details] [diff] [review]
Patch to bracket our drawing rtns w/LockPortBits & UnlockPortBits

sr=[e-mail]

Comment 6

[Robert John Churchill]

Checked into trunk, nominating from Moz1.0 branch

Comment 7

[Jaime Rodriguez, Jr.]

Nominating for adt1.0.1 as part of Cathleen's request during ADT meeting.

PAW - Can you find someone to verify this on the trunk, and check around for any
possible regressions, so we can take it in as early as possible?

Comment 8

[Paul Wyskoczka]

qa --> petersen

Comment 9

[Paul Wyskoczka]

Chris, could you verify this fix on the trunk.

Comment 10

[Chris Petersen]

I have today's OS X trunk build. What's the best way to verify this ?

Comment 11

[Mike Pinkerton (not reading bugmail)]

page-loader tests?

Comment 12

[Jaime Rodriguez, Jr.]

this has been checked into the trunk, so i am resolving this as fixed. [Note: it
has been nominated for the 1.0 branch with adt1.0.1 and mozilla1.0.1 keywords].

Comment 13

[Chris Petersen]

Spend time loading top 150 sites in the OS X trunk build (2002-06-04-08). No
problems found. Marking verified.

Comment 14

[Jaime Rodriguez, Jr.]

adt1.0.1+ (on ADT's behalf) approval for checkin to the 1.0 branch, pending
Driver's approval. pls check this in asap, then add the keyword "fixed1.0.1".

Comment 15

[Judson Valeski]

please checkin to the 1.0.1 branch. once there remove the "mozilla1.0.1+"
keyword and add the "fixed1.0.1" keyword.

Comment 16

[Robert John Churchill]

Fixed on Moz-1.0 branch.

Comment 17

[Simon Fraser [no longer active]]

So did this have any measurable performance impact?

Comment 18

[Jaime Rodriguez, Jr.]

chris - can you pls verify this on the branch, then mark it as "verified1.0.1"?
thanks!

Comment 19

[Chris Petersen]

Ran same page load tests in branch build (2002-07-16-05) with no regressions
found.  Adding verified1.0.1 keyword.