734415 - IonMonkey: Specialize phis with int32 and double operands as double

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Jan de Mooij [:jandem]]

Attachment: Patch

With this patch Kraken gaussian-blur drops from 290 ms to 185 ms.

Comment 1

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

Comment on attachment 604442 [details] [diff] [review]
Patch

Review of attachment 604442 [details] [diff] [review]:
-----------------------------------------------------------------

Nice.

Comment 2

[Jan de Mooij [:jandem]]

https://hg.mozilla.org/projects/ionmonkey/rev/2eddf60c86e2

Comment 3

[Hannes Verschore [:h4writer]]

If I'm reading this diff right.
Among other things you changed the following:

for (...) {
  if (type == MIRType_None) {
    ...
  } else if (type != in->type()) {
    ...
  }
}

into

for (...) {
  if (type == MIRType_None) {
    ...
    continue;
  }
  if (type != in->type()) {
    ...
  }
}

Also with changes on the "..." lines.
But I'm curious why altering "else if" into a "continue". 
Is there a specific reason for doing this?

Comment 4

[Jan de Mooij [:jandem]]

(In reply to Hannes Verschore from comment #3)
>
> But I'm curious why altering "else if" into a "continue". 
> Is there a specific reason for doing this?

No important reason, it just seemed a bit more readable and it matched the other "if .. continue;" before it.

Comment 5

[Hannes Verschore [:h4writer]]

(In reply to Jan de Mooij (:jandem) from comment #4)
> No important reason, it just seemed a bit more readable and it matched the
> other "if .. continue;" before it.

Ah ... I was thinking there were some style rules unknown to me involved.
Nice win on kraken btw.