Closed Bug 97211 Opened 23 years ago Closed 23 years ago

CFM Mac OS X build should use system allocators, not MSL

Categories

(SeaMonkey :: Build Config, defect, P2)

Product:
SeaMonkey
Component:
Build Config
Platform:
PowerPC
macOS
Type:
defect

Tracking

(Not tracked)

Status:
RESOLVED WONTFIX
Milestone:
mozilla0.9.9

People

(Reporter: sfraser_bugs, Assigned: sdagley)

References

Details

(Keywords: perf)

Attachments

(2 files, 1 obsolete file)

Page load results with MSL malloc()
(deleted), text/html
Details
Page load results with system malloc()
(deleted), text/html
Details
We appear to be about 7% faster if our malloc() goes straight through to the system malloc(), rather than using MSL's alloc pool stuff. I'll attach page-load results. This is an easy change. We just have to define '_MSL_OS_DIRECT_MALLOC' in the appropriate MemAllocator.mcp prefix files.
Attached file Page load results with MSL malloc() (deleted) —
A 7% improvement we definitely can use - nominating for nsenterprise and nsbranch
Keywords: nsbranch, nsenterprise
Forgot to change milestone & priority when I added the nsbranch and nsenterprise keywords
Priority: -- → P2
Target Milestone: --- → mozilla0.9.5
Whiteboard: [OSX+] let's try this on the trunk then move to branch
7% is good. Simon doesnt have time to work on this one this week. Anyone else wann take this ? Pink ? Dagley ?
nsbranch+ing.
Keywords: nsbranch → nsbranch+
--> sdagley
Assignee: sfraser → sdagley
Attached patch Patch to use OS version of malloc (obsolete) (deleted) — — Splinter Review
Patch attached. I thought this version of the fix we only wanted on the branch and we'd have a fancier one for the trunk.
Attachment #50327 - Attachment is obsolete: true
Ignore that first patch, I forgot to make it Carbon specific.
Pls only check this one in on the branch.
Whiteboard: [OSX+] let's try this on the trunk then move to branch → [OSX+] let's try this on the trunk then move to branch [ETA 09.21]
Fix checked in on 094 branch but leaving bug open until we address it on the trunk. Are we running jgm's tests on the 094 branch so I can verify the improvment or do I need to compare the 09/21 and 09/24 builds myself? Also changing OSX+ in status to OSX since it is no longer blocking the branch
Status: NEW → ASSIGNED
Whiteboard: [OSX+] let's try this on the trunk then move to branch [ETA 09.21] → [OSX] let's try this on the trunk then move to branch
Whiteboard: [OSX] let's try this on the trunk then move to branch → [OSX] let's try this on the trunk then move to branch,PDT+
Should have used better phrasing - bug will stay open until a better/permanent fix is landed on the trunk. The Q&D fix to make all Carbon builds use the OS malloc is already on the branch.
Whiteboard: [OSX] let's try this on the trunk then move to branch,PDT+ → [OSX] quickfix on branch - trunk will get different fix, PDT+
Blocks: 102998
Target Milestone: mozilla0.9.5 → mozilla0.9.6
marking nsenterprise-; will be reevaluated for nsenterprise in future release.
Keywords: nsenterprise-
has this been verified on the 094 branch.
Whiteboard: [OSX] quickfix on branch - trunk will get different fix, PDT+ → [OSX] quickfix on branch - trunk will get different fix, [PDT+]
http://bugzilla.mozilla.org/show_bug.cgi?id=71718 should have some perf timings which we ran for OS X. Do those have the build dates you needed the data for to verify this on the branch?
QA Contact: granrose → sujay
Blocks: 107066
Keywords: nsbranch+
Resetting target milestone, 096 is past. Please retarget when you get back to this bug (no current activity apparent.)
Target Milestone: mozilla0.9.6 → ---
Target Milestone: --- → mozilla0.9.8
Keywords: nsenterprise, nsenterprise-
Whiteboard: [OSX] quickfix on branch - trunk will get different fix, [PDT+]
deferring to 0.9.9 (ccarlen is doing some research on the Pro 7 mem allocators which we want before deciding on changing them)
Target Milestone: mozilla0.9.8 → mozilla0.9.9
this was of importance at one time, and it looks like it can be fixed before M1.0 ... nominating nsbeta1.
Keywords: nsbeta1
I'm not the qa_contact for this component...Hong, please re-assign.
QA Contact: sujay → hong
No longer blocks: 107066
Blocks: 107066
No longer blocks: 107066
nsbeta1+ per ADT triage team, adding perf keyword
Keywords: nsbeta1 → nsbeta1+, perf
My testing with Pro 7 shows we're actually slightly (approx 3%) slower if we switch to the OS native malloc() under OS X so closing this bug as WONTFIX
Status: ASSIGNED → RESOLVED
Closed: 23 years ago
Resolution: --- → WONTFIX
Product: Browser → Seamonkey
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: