- Allow the use of branch delay slots.

- Allow the use of branch delay slots.
Committed: https://code.google.com/p/dart/source/detail?r=19669

[Ivan Posva, 2013-03-07 16:57:41]

https://codereview.chromium.org/12623002/diff/1/runtime/vm/assembler.h
File runtime/vm/assembler.h (right):

https://codereview.chromium.org/12623002/diff/1/runtime/vm/assembler.h#newcode93
runtime/vm/assembler.h:93: template<typename T> void Backup() {
If you have a better suggestions for this name, such as UnEmit, I would like to
hear about them.

[srdjan, 2013-03-07 17:07:30]

lgtm

[zra, 2013-03-07 17:11:34]

lgtm

https://codereview.chromium.org/12623002/diff/1/runtime/vm/assembler.h
File runtime/vm/assembler.h (right):

https://codereview.chromium.org/12623002/diff/1/runtime/vm/assembler.h#newcode93
runtime/vm/assembler.h:93: template<typename T> void Backup() {
On 2013/03/07 16:57:41, Ivan Posva wrote:
> If you have a better suggestions for this name, such as UnEmit, I would like
to
> hear about them.
Remit?

[regis, 2013-03-07 17:35:09]

LGTM

"Remit" is not bad, if a bit fancy.
"RemoveLast" or "UndoEmit" may be clearer.

We will have to be careful not to bind a label and then use the delay slot. Bind
should make the delay slot unavailable.

Cute syntactic sugar:
#define ____ assembler->delay_slot()->

__ jr(RA);
____ ori(V0, ZR, Immediate(42));

[Ivan Posva, 2013-03-08 07:25:25]

Committed patchset #3 manually as r19669 (presubmit successful).

[Ivan Posva, 2013-03-08 09:44:20]

On 2013/03/07 17:35:09, regis wrote:
> LGTM
> 
> "Remit" is not bad, if a bit fancy.
> "RemoveLast" or "UndoEmit" may be clearer.

I used "Remit". I like it and Zach is a native speaker.

> We will have to be careful not to bind a label and then use the delay slot.
Bind
> should make the delay slot unavailable.
> 
> Cute syntactic sugar:
> #define ____ assembler->delay_slot()->
> 
> __ jr(RA);
> ____ ori(V0, ZR, Immediate(42));

This could be a bit too subtle.

-Ivan