Bug 174645 - Reduce memory usage during OSR
Summary: Reduce memory usage during OSR
Status: ASSIGNED
Alias: None
Product: WebKit
Classification: Unclassified
Component: JavaScriptCore (show other bugs)
Version: WebKit Nightly Build
Hardware: Unspecified Unspecified
: P2 Normal
Assignee: Mark Lam
URL:
Keywords: InRadar
Depends on: 175144 175145 175453 174694 174696 174697 174809 174837 175446 175549 175617 175688
Blocks:
  Show dependency tree / graph
 
Reported: 2017-07-18 16:30 PDT by Mark Lam
Modified: 2017-08-17 14:52 PDT (History)
8 users (show)

See Also:

Attachments
Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Mark Lam 2017-07-18 16:30:58 PDT 
Details coming.

<rdar://problem/33183238>
Comment 1 Mark Lam 2017-07-18 16:35:58 PDT 
Currently, when we have to do an OSR exit (DFG->baseline or FTL->baseline), we generate >=1kb of machine code that does the exit.  We don’t need to generate any code for every exit; we could just have one shared JIT probe that calls a C++ function that does the exit.  This would save a lot of executable memory.

This also means that supporting the JIT probe mechanism will be required in order to use the DFG and FTL going forward.  Currently, the only CPU targets that support the JIT probe are ARM, ARMv7, ARM64, x86, and x86_64.