Reviewers: Kevin Millikin, Mads Ager,
Description:
Implementing inline caches for GenericBinaryOpStub.
There are 3 fast cases: HeapNumber operands, String operands and Object
operands.
This CL implements it for ia32 only.
Please review this at http://codereview.chromium.org/551093
SVN Base: htt
http://codereview.chromium.org/551093
--
v8-dev mailing list
v8-dev@googlegroups.com
http://groups.google.com/group/v8-dev
I have addressed all the comments. Now I will split this CL into two as Mads
suggested: one dealing with registers/smis improvements, another dealing
with
ICs.
http://codereview.chromium.org/551093/diff/1/3
File src/ia32/codegen-ia32.cc (right):
http://codereview.chromium.org/551093/diff/1/3
ARM codegen currently does not have binary op ICs in trunk. My change got
reverted when it broke debug tests. The funny thing is: BinaryOpIC::Clear
was the reason. I am re-submitting ARM change today with this issue fixed.
Vlad
On Thu, Apr 8, 2010 at 5:36 PM, wrote:
> LGTM, thanks!
>
> Isn't
That and a change in BinaryOpIC::GetTypeInfo that also have been
breaking debug tests.
On Thu, Apr 8, 2010 at 7:12 PM, wrote:
>
> LGTM, I suppose the only change to the previous submit is the removal of the
> clearing on GC?
>
> http://codereview.chromium.org/1629008/show
--
v8-dev mailing list
question.
Vlad
On Thu, Apr 8, 2010 at 7:25 PM, Mads Sig Ager wrote:
> Can you quickly explain why that assert is triggered on ARM and not on
> the other platforms?
>
> Thanks, -- Mads
>
> On Thu, Apr 8, 2010 at 5:17 PM, Vladislav Kaznacheev
> wrote:
>> That
Summarizing the chat we had with Mads:
1. Doesn't this mean that we will always enter the runtime system for
> calling
> functions in arrays: a[0]()? This used to be a KeyedLoadIC followed by a
> call
> which is pretty fast. We should make sure that this is still fast.
>
This has been my oversigh
OK. BTW, similar change on x64 broke something, so I know it is not that
simple.
Vlad
On Wed, Jun 2, 2010 at 6:36 PM, wrote:
> Vlad,
>
> Thanks a lot for spotting this!
>
> Unfortunately, you can't just overwrite the index register. The contract of
> the
> generator is to preserve it. It may
Fixed before submitting.
Thanks,
Vlad
On Mon, Jul 5, 2010 at 3:25 PM, wrote:
> There is a lint error on objects.h.
>
>
> http://codereview.chromium.org/2812041/show
>
--
v8-dev mailing list
v8-dev@googlegroups.com
http://groups.google.com/group/v8-dev
--
v8-dev mailing list
v8-dev@googlegroups.com
http://groups.google.com/group/v8-dev
Thanks for the comments!
We had a long discussion with Vitaly on how to proceed, and here is what we
came up with (let me know if you find anything of the following stupid):
- Updating the slack estimation on every map transition is too fragile and
does not provide any performance benefit. I will
11 matches
Mail list logo