Closed Bug 515822 Opened 15 years ago Closed 15 years ago

TM merge: Assembler and LIR changes from tracemonkey

Categories

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

defect
Not set
normal

Tracking

(Not tracked)

VERIFIED FIXED

People

(Reporter: rreitmai, Assigned: rreitmai)

References

Details

Attachments

(5 files)

A catch-all bug for landing changes to tamarin redux taken from tracemonkey.
Attached patch hashcode improvments (deleted) — Splinter Review
Mirror of bug 513539
Attachment #399893 - Flags: superreview?(edwsmith)
Attached patch remove AVMPLUS_PORTING_API (deleted) — Splinter Review
Mirror of bug 513535
Attachment #399898 - Flags: superreview?(edwsmith)
Attached patch dead code - NJ_STACK_OFFSET (deleted) — Splinter Review
Attachment #399903 - Flags: superreview?(edwsmith)
Attached patch LIR_float added (deleted) — Splinter Review
Mirror of change from bug 513838
Attachment #399909 - Flags: superreview?(edwsmith)
Mirror of changes in bugs 513884 and 513843
Attachment #399913 - Flags: superreview?(edwsmith)
Blocks: 503556
Depends on: 513539, 513535, 513838, 513843, 513884
Attachment #399903 - Flags: superreview?(edwsmith) → superreview+
Attachment #399893 - Flags: superreview?(edwsmith) → superreview+
Attachment #399898 - Flags: superreview?(edwsmith) → superreview+
(In reply to comment #4) > Created an attachment (id=399909) [details] > LIR_float added > > Mirror of change from bug 513838 does it cause problems if we are still using LIR_quad for double constants from Tamarin?
Comment on attachment 399913 [details] [diff] [review] gen() signature change / introduce LIR_flive code looks safe, and +1 for allocating loop variables to registers, but we need to make sure the expressions we use with LIR_live in Tamarin, match up here. we often use LIR_live(LIR_alloc expr), which should *not* get a register, so we should at least tweak this to include LIR_alloc in the list of instruction types that dont get registers assigned.
Attachment #399913 - Flags: superreview?(edwsmith) → superreview-
comment #6: no the code still works fine, but the suggestion from TM folks is that we should migrate over to LIR_float as there are fp optimizations that kick-in , when doing so. comment #7: nice catch; I had problems with that particular patch and ended up commenting out the handleLoopCarriedExprs. Would like to push the rest of the patch, commenting out that change (i.e leaving it as fingRegMem(). There will be some postmortem work where we need further changes, but I'd like to keep the continuity moving forward with the remaining patches.
(In reply to comment #8) > comment #7: nice catch; I had problems with that particular patch and ended up > commenting out the handleLoopCarriedExprs. Would like to push the rest of > the patch, commenting out that change (i.e leaving it as fingRegMem(). findMemFor, i think. > There will be some postmortem work where we need further changes, but I'd like > to keep the continuity moving forward with the remaining patches. cool
Blocks: 517308
Attachment #399909 - Flags: superreview?(edwsmith) → superreview+
Resolved fixed engineering / work item that has been pushed. Setting status to verified.
Status: RESOLVED → VERIFIED
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: