1009967 - IonMonkey: Implement BitXor Recover Instruction

Status: RESOLVED FIXED

(Core :: JavaScript Engine: JIT, defect)

Description

[Nicolas B. Pierron [:nbp]]

Implement RBitXor in js/src/jit/Recover.cpp.
See Bug 1003801 comment 0 for explanation.

Comment 1

[Inanc Seylan]

I would like to work on this bug; if it's ok.

Comment 2

[Inanc Seylan]

Attachment: rbitxor.patch

An implementation of the recover code for BitXor and a test case.

Comment 3

[Nicolas B. Pierron [:nbp]]

Comment on attachment 8422425 [details] [diff] [review]
rbitxor.patch

Review of attachment 8422425 [details] [diff] [review]:
-----------------------------------------------------------------

This looks good, but …

::: js/src/jit-test/tests/ion/dce-with-rinstructions.js
@@ +78,5 @@
>  
>  for (i = 0; i < 100; i++) {
>      rbitnot_number(i);
>      rbitnot_object(i);
> +    rbitxor_number(i);

Add an object test case,

::: js/src/jit/MIR.h
@@ +3517,5 @@
>      }
>      void computeRange(TempAllocator &alloc);
> +
> +    bool writeRecoverData(CompactBufferWriter &writer) const;
> +    bool canRecoverOnBailout() const { return true; }

And have a look at MBinaryBitwiseInstruction ;)

Comment 4

[Inanc Seylan]

(In reply to Nicolas B. Pierron [:nbp] from comment #3)
> Comment on attachment 8422425 [details] [diff] [review]
> rbitxor.patch
> 
> Add an object test case,
Done.
> 
> ::: js/src/jit/MIR.h
> 
> And have a look at MBinaryBitwiseInstruction ;)
Good point ;) Since I did not see any getAliasSet() method in MBitXor, my initial implementation returned true from canRecoverOnBailout().

Comment 5

[Inanc Seylan]

Attachment: rbitxor.patch

Improvements to the previous version:
1. Test case for rbitxor_object.
2. A new canRecoverOnBailout() implementation for MBitXor so that it does not try to recover for objects.

Comment 6

[Nicolas B. Pierron [:nbp]]

Comment on attachment 8422468 [details] [diff] [review]
rbitxor.patch

Review of attachment 8422468 [details] [diff] [review]:
-----------------------------------------------------------------

Nice work :)

Comment 7

[Nicolas B. Pierron [:nbp]]

If you want your patch to be landed, do not set checkin+, but checkin?.  Otherwise we might think that the patch is already landed.  Also once the patch has a r+ from all reviewers without any fix-up needed, then you can add the checkin-needed keyword.

Comment 8

[Inanc Seylan]

(In reply to Nicolas B. Pierron [:nbp] from comment #7)
> If you want your patch to be landed, do not set checkin+, but checkin?. 
> Otherwise we might think that the patch is already landed.  Also once the
> patch has a r+ from all reviewers without any fix-up needed, then you can
> add the checkin-needed keyword.

Ok, thanks!

Comment 9

[Nicolas B. Pierron [:nbp]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/22298ae9c687

Comment 10

[Carsten Book [:Tomcat]]

also thanks for contributing to mozilla!

Comment 11

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/22298ae9c687