515313 - nanojit: finish killing reservations

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Nicholas Nethercote [inactive]]

Bug 514349 started killing reservations, mostly removing them from Assembler.cpp and the i386 backend.  Bug 515309, bug 515310, bug 515311, and bug 515312 are for each of the other backends.  Once they're all done, the final remnants can be removed from Assembler.{cpp,h} and LIR.{cpp,h}.

Comment 1

[Nicholas Nethercote [inactive]]

Nb: we see patterns like this a lot in the back-ends:

    Register sr = ( lhs->isUnusedOrHasUnknownReg()
                  ? findRegFor(lhs, FpRegs)
                  : lhs->getReg() );

with some variation on the ?-branch.  It might be worth factoring these out into a function.

Comment 2

[Nicholas Nethercote [inactive]]

(In reply to comment #1)
> Nb: we see patterns like this a lot in the back-ends:
> 
>     Register sr = ( lhs->isUnusedOrHasUnknownReg()
>                   ? findRegFor(lhs, FpRegs)
>                   : lhs->getReg() );
> 
> with some variation on the ?-branch.  It might be worth factoring these out
> into a function.

Bug 516348 covers the most common of these cases.

Comment 3

[Nicholas Nethercote [inactive]]

Attachment: patch

Here's a patch.  Nothing surprising, just finishes off removing the Reservation type.

Comment 4

[Nicholas Nethercote [inactive]]

BTW, I tested the patch on the TR try server, it passed all platforms.

Comment 5

[Nicholas Nethercote [inactive]]

hg.mozilla.org/projects/nanojit-central/rev/1b6cb808f304

Comment 6

[Leon Sha]

http://hg.mozilla.org/tracemonkey/rev/f34d86c36e47
http://hg.mozilla.org/tamarin-redux/rev/4f97e229b335

Comment 7

[Nicholas Nethercote [inactive]]

http://hg.mozilla.org/mozilla-central/rev/f34d86c36e47