Closed Bug 2211 Opened 26 years ago Closed 26 years ago

[PP] Text rendering is slow

Categories

(Core Graveyard :: GFX, defect, P1)

PowerPC
All
defect

Tracking

(Not tracked)

VERIFIED FIXED

People

(Reporter: pierre, Assigned: sfraser_bugs)

Details

Text rendering is particularly slow when we have plenty of different chunks of text. The font handling is not optimized at all. I described the problem in a comment in gfx.
Status: NEW → ASSIGNED
Assignee: pierre → sfraser
Status: ASSIGNED → NEW
Reassigned to <sfraser>
Pierre, do you have a sample URL that display slow rendering?
The default page (test0.html) is a very good example. Click the sizebox and shake the mouse like hell: the redraw should go much faster. Try running it on a non-G3 machine for more visible results. An advice too: recompile everything with Profile information (back on the G3 ;-), you will certainly see interesting things.
sujay/elig - here's another one to do time comparison per platform on. Let find out just how slow, slow is on the Mac.
I'm not done yet on the optimizations for this, which will help a little. 2254 is the big slowness bug.
Eli, you're a mac person...you can take this one...
Status: NEW → RESOLVED
Closed: 26 years ago
OS: All
Resolution: --- → FIXED
The GetFNum optimization that this bug was aimed at has been fixed, so I'm resolving this. For QA to test, you'll need to compare a viewer build from before ATSUI was turned on with today's build, on a page with many <font> tags. It should be a little faster.
Eli - check this out with 01/15 compared to next build when available. Let us know how it looks.
elig, could you please Verify this today? If better, let's mark as Verified.
Oops. This one fell through the cracks here; will verify tonight.
I really don't see a performance difference in page layout or resizing between the 1.12.99, 1.15.99 or 1.19.99 builds on an 8500/233, using the demo0 page. Generally, laying out the entire page takes about 2.5 seconds, and a minimal ---> maximal (on a 1024x768 display) resize takes about 1.3 seconds. I'll see if I can find a more intensive test case, or a slower Macintosh.
Per conversation with sfraser, I'd like to just mark this as verified: * Simon doesn't think the results will be perceptible unless the testing was done with a particularly long and complex case (in which slowness from bug #2254 dominates the results, and renders it impossible to check this bug) * I spend enough time using Simon's software to know that he knows how to do Mac optimization...and if he says it's been optimized, it's optimized. ;) * The only cases I can find aren't any faster * The only intensive cases I can find don't even render on the 1.15.99 or earlier builds (i.e. http://www.w3.org/Style/CSS/Test/current/...) * Page layout is consistently slower than on Win32, irregardless of whether multiple fonts are present. So, if there are no protests by tomorrow, I'll mark this Verified. Otherwise, I can check again with tomorrow's build (with 2254 fixed)
Status: RESOLVED → VERIFIED
Per conversation w/sfraser, marking this one as verified (for reasons noted in prior comment.)
Inserting Milestone info.
Product: Core → Core Graveyard
You need to log in before you can comment on or make changes to this bug.