514532 - NJ: LIR changes to help x64 code generation

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[David Anderson [:dvander] - inactive, e-mail if emergency]

Attachment: patch

Adds ins_peq0, insImmWord, merges some spew changes from tamarin-redux.

Comment 1

[Andreas Gal :gal]

shouldn't that be insImmp?

Comment 2

[David Anderson [:dvander] - inactive, e-mail if emergency]

That takes a (void*), I want to avoid having (void*) casts everywhere.

Comment 3

[David Anderson [:dvander] - inactive, e-mail if emergency]

Attachment: added ins_i2p and ins_u2p

This version lets me get rid of some ugly helpers in jstracer.cpp which I had to duplicate in jsregexp.cpp.

Comment 4

[Rick Reitmaier]

Comment on attachment 398490 [details] [diff] [review]
added ins_i2p and ins_u2p

Not sure we want to head back in this direction with implicit types; deferring to Ed for his comments.

Comment 5

[David Anderson [:dvander] - inactive, e-mail if emergency]

(In reply to comment #4)
> (From update of attachment 398490 [details] [diff] [review])
> Not sure we want to head back in this direction with implicit types; deferring
> to Ed for his comments.

What are the implicit types? The helpers just redirect to pointer-width versions as helpers -- compile-time decision, not run-time sniffing of LIR.

Comment 6

[Edwin Smith]

Comment on attachment 398490 [details] [diff] [review]
added ins_i2p and ins_u2p

lir itself should not have any 'pointer' type instructions (32bit or 64bit must be explicit) but its okay to have helpers for pointers.  we already had some, this adds a few more.

Comment 7

[Blake Kaplan (:mrbkap) (inactive)]

http://hg.mozilla.org/mozilla-central/rev/bc9354a9fe70

Comment 8

[Robert Sayre]

http://hg.mozilla.org/releases/mozilla-1.9.2/rev/bd96a5387880