Call a stub for the write barrier on ARM. This is a step

Call a stub for the write barrier on ARM.  This is a step
towards incremental marking on ARM.
Committed: http://code.google.com/p/v8/source/detail?r=8195

[Lasse Reichstein, 2011-06-07 08:48:37]

LGTM

http://codereview.chromium.org/7113012/diff/1/src/arm/macro-assembler-arm.cc
File src/arm/macro-assembler-arm.cc (right):

http://codereview.chromium.org/7113012/diff/1/src/arm/macro-assembler-arm.cc#...
src/arm/macro-assembler-arm.cc:3142: void MacroAssembler::HasColour(Register
object,
I think we generally use the Bowdlerized US English spelling, however much it
might hurt.
Variables too.

http://codereview.chromium.org/7113012/diff/1/src/arm/macro-assembler-arm.cc#...
src/arm/macro-assembler-arm.cc:3150: MarkBits(object, bitmap_scratch,
mask_scratch);
Could MarkBits be named something that sounds less imperative? I get confused as
to whether it actually marks any bits, or just finds the address/mask of the
mark-bits.
Something like GetMarkBitAddressAndMask would perhaps be too long, but more
readable.

http://codereview.chromium.org/7113012/diff/1/src/arm/macro-assembler-arm.cc#...
src/arm/macro-assembler-arm.cc:3158: b(eq, &word_boundary);
Should this be optimized for speed or size?
If speed, you could probably detect the case where the color overlaps a word
boundary earlier (mask is negative) and have a faster case that tests both bits
in one test if it doesn't.

http://codereview.chromium.org/7113012/diff/1/src/arm/macro-assembler-arm.h
File src/arm/macro-assembler-arm.h (right):

http://codereview.chromium.org/7113012/diff/1/src/arm/macro-assembler-arm.h#n...
src/arm/macro-assembler-arm.h:1145: // the position of the first bit.  Uses ecx
as scratch and leaves addr_reg
ecx?

[Erik Corry, 2011-06-08 10:31:29]

http://codereview.chromium.org/7113012/diff/1/src/arm/macro-assembler-arm.cc
File src/arm/macro-assembler-arm.cc (right):

http://codereview.chromium.org/7113012/diff/1/src/arm/macro-assembler-arm.cc#...
src/arm/macro-assembler-arm.cc:3142: void MacroAssembler::HasColour(Register
object,
On 2011/06/07 08:48:37, Lasse Reichstein wrote:
> I think we generally use the Bowdlerized US English spelling, however much it
> might hurt.
> Variables too.

Auch!  That hert!

http://codereview.chromium.org/7113012/diff/1/src/arm/macro-assembler-arm.cc#...
src/arm/macro-assembler-arm.cc:3150: MarkBits(object, bitmap_scratch,
mask_scratch);
On 2011/06/07 08:48:37, Lasse Reichstein wrote:
> Could MarkBits be named something that sounds less imperative? I get confused
as
> to whether it actually marks any bits, or just finds the address/mask of the
> mark-bits.
> Something like GetMarkBitAddressAndMask would perhaps be too long, but more
> readable.

GetMarkBits

http://codereview.chromium.org/7113012/diff/1/src/arm/macro-assembler-arm.cc#...
src/arm/macro-assembler-arm.cc:3158: b(eq, &word_boundary);
On 2011/06/07 08:48:37, Lasse Reichstein wrote:
> Should this be optimized for speed or size?
> If speed, you could probably detect the case where the color overlaps a word
> boundary earlier (mask is negative) and have a faster case that tests both
bits
> in one test if it doesn't.

I think there's room for experimentation here, but I'm going to leave that for
later.

http://codereview.chromium.org/7113012/diff/1/src/arm/macro-assembler-arm.h
File src/arm/macro-assembler-arm.h (right):

http://codereview.chromium.org/7113012/diff/1/src/arm/macro-assembler-arm.h#n...
src/arm/macro-assembler-arm.h:1145: // the position of the first bit.  Uses ecx
as scratch and leaves addr_reg
On 2011/06/07 08:48:37, Lasse Reichstein wrote:
> ecx?

Yeah, we had ARM add it to the architecture specially for V8 :-)