Closed
Bug 487482
Opened 15 years ago
Closed 6 years ago
Implement policy for flushing and reusing JIT code memory
Categories
(Tamarin Graveyard :: Baseline JIT (CodegenLIR), defect)
Tamarin Graveyard
Baseline JIT (CodegenLIR)
Tracking
(Not tracked)
RESOLVED
WONTFIX
Future
People
(Reporter: edwsmith, Unassigned)
References
Details
(Whiteboard: Tracking)
Attachments
(1 file)
13.06 KB,
patch
|
lhansen
:
review+
|
Details | Diff | Splinter Review |
Tracking bug for adding support for flushing cold code from jit memory
Reporter | ||
Updated•15 years ago
|
Assignee: nobody → edwsmith
Reporter | ||
Comment 1•15 years ago
|
||
* depends on the patches that are in dependent bugs. * conservative mark-sweep of stack to find active code. * known bug: the ExceptionHandlers object from jitted code sticks around when method is later interpreted.
Attachment #376272 -
Flags: review?
Reporter | ||
Updated•15 years ago
|
Attachment #376272 -
Flags: review? → review?(lhansen)
Reporter | ||
Comment 2•15 years ago
|
||
working repo is http://hg.mozilla.org/users/edwsmith_adobe.com/users/deopt
Comment 3•15 years ago
|
||
Found the repo at http://hg.mozilla.org/users/edwsmith_adobe.com/deopt
Comment 4•15 years ago
|
||
Comment on attachment 376272 [details] [diff] [review] use premark() callback to flush jitted code that isn't active on stack. Looks reasonable to me. (In sweep, the accounting only needs happen for verbose code, which may matter if CodeAlloc::size is anything but trivial. Another performance hack is that if sweeping will discard all unmarked methods /and/ there is an expectation of many active methods so that sweep has relatively low efficiency then you can dispense with sweeping altogether by having the marker unlink pinned methods from the method list (and add them to a secondary list); sweeping then just deletes everything remaining on the list and swaps in the secondary list for the first list. But that seems minor.)
Attachment #376272 -
Flags: review?(lhansen) → review+
Reporter | ||
Comment 6•15 years ago
|
||
Work in progress in these repos: https://hg.mozilla.org/users/edwsmith_adobe.com/deopt https://hg.mozilla.org/users/edwsmith_adobe.com/fragmentation https://hg.mozilla.org/users/edwsmith_adobe.com/second-jit all are ~3 months stale and need rebasing. * CodeAlloc fragmentation has been addressed in tamarin-redux * pointer classification may no longer be necessary; new MethodFrame gives us a way to exactly walk the AS3 stack and mark jit'd functions.
Updated•15 years ago
|
Target Milestone: --- → Future
Reporter | ||
Updated•14 years ago
|
Assignee: edwsmith → nobody
Comment 7•14 years ago
|
||
This is a mass change. Every comment has "assigned-to-new" in it. I didn't look through the bugs, so I'm sorry if I change a bug which shouldn't be changed. But I guess these bugs are just bugs that were once assigned and people forgot to change the Status back when unassigning.
Status: ASSIGNED → NEW
Reporter | ||
Updated•14 years ago
|
OS: Mac OS X → All
Hardware: x86 → All
Whiteboard: Tracking
Reporter | ||
Updated•14 years ago
|
Component: Virtual Machine → JIT Compiler (NanoJIT)
Comment 8•6 years ago
|
||
Tamarin is a dead project now. Mass WONTFIX.
Status: NEW → RESOLVED
Closed: 6 years ago
Resolution: --- → WONTFIX
Comment 9•6 years ago
|
||
Tamarin isn't maintained anymore. WONTFIX remaining bugs.
You need to log in
before you can comment on or make changes to this bug.
Description
•