Closed Bug 524578 Opened 15 years ago Closed 15 years ago

NJ merge: cumulative changes to finish nanojit merge

Categories

(Tamarin Graveyard :: Baseline JIT (CodegenLIR), defect, P2)

defect

Tracking

(Not tracked)

VERIFIED FIXED
flash10.1

People

(Reporter: graydon, Assigned: edwsmith)

References

Details

Attachments

(1 file, 1 obsolete file)

These are problematic for a couple reasons. 

First and most importantly, they're being used in a negative sense at a CPP level (#ifndef) as antonyms for ARM_VFP, which is a runtime test on TM. That much simply has to go, it's nonsense and leads us to having to undefine NJ_SOFTFLOAT in order to get the VFP-support code into our builds.

Once those go, the only remaining instance in all of nanojit is in insCall, which is already ARM_VFP guarded, and in TR those appear to be set as compile-time mutual opposites, as they're runtime mutual opposites in TM. To reduce confusion here I'm proposing just removing the guards there as well.
This patch contains corrected guards (tested) and includes, cumulatively, all remaining differences I think TR needs to absorb. Covers the evict issue and the loop-carried exprs issue.

Sorry to be doing all at once, figured your time is scarce.
Attachment #408482 - Attachment is obsolete: true
Assignee: rreitmai → edwsmith
Summary: NJ merge: remove NJ_SOFTFLOAT guards from nanojit → NJ merge: cumulative changes to finish nanojit merge
Attachment #409149 - Flags: review+
Attachment #409149 - Flags: review?(rreitmai)
tamarin-redux and tracemonkey match exactly now.  recording URL's for posterity.

http://hg.mozilla.org/tamarin-redux/rev/6e89a44f87de
http://hg.mozilla.org/tracemonkey/rev/a658f130ad09
Priority: -- → P2
Target Milestone: --- → flash10.1
Attachment #409149 - Flags: review?(rreitmai) → review+
Status: NEW → RESOLVED
Closed: 15 years ago
Resolution: --- → FIXED
Status: RESOLVED → VERIFIED
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: