[jsc-dev] [webkit-gtk] Enable WebAssembly on Linux based ports (GTK and JSCOnly)

Konstantin Tokarev annulen at yandex.ru
Wed Mar 1 09:16:01 PST 2017



01.03.2017, 20:10, "Andrea Giammarchi" <andrea.giammarchi at gmail.com>:
> Chiming in, it'd be awesome to be able to test WASM on 2.16 behind a flag, I'm targeting mostly SBC including Raspberry Pi 0 where WebKitGTK+ already works like a charm (enabling WebGL and with HW accelerated layers) but I can't compare performance against WASM and building from scratch targeting Pi Zero ain't the easiest thing to do.

AFAIU Zero is based on 32-bit ARM, and WASM requires B3 that is 64-bit only.

>
> Thanks for eventual consideration.
>
> On Wed, Mar 1, 2017 at 4:57 PM, Yusuke SUZUKI <utatane.tea at gmail.com> wrote:
>> On Thu, Mar 2, 2017 at 1:36 AM, Saam Barati <sbarati at apple.com> wrote:
>>> On Mar 1, 2017, at 7:46 AM, Michael Catanzaro <mcatanzaro at igalia.com> wrote:
>>>
>>>> On Wed, 2017-03-01 at 15:44 +0100, Carlos Garcia Campos wrote:
>>>> We have recently branched for 2.16 so it's a perfect moment to enable
>>>> it in trunk with very little risk.
>>>>
>>>> Is Apple planning to enable it within the next six months? If so, now
>>>> is indeed the ideal time to enable it for WebKitGTK+.
>>>
>>> What do you mean by enable it? It's been on by default on the Mac port for a while now and has been in the various STP releases.
>>>
>>> I think it's a good idea to enable it on other ports too. Note that Keith is currently working on a patch to make loads faster in Wasm. It may require some OS specific handling to run well on Linux: https://bugs.webkit.org/show_bug.cgi?id=162693
>>
>> Awesome! I think enabling WASM on Linux after landing Keith's faster memory access patch (by using SEGV signal handling instead of bound checking) is good.
>> Once that patch is implemented and landed, we can start implementing platform-specific part of that (extracting instruction pointer from mcontext_t and signal handling part should be implemented for Linux).
>> I think that should be easy (since retrieving instruction pointer from mcontext_t is already once implemented for sampling profiler. Thus, we can share / re-implement the same thing for that).
>>
>>>> If not, then I'd prefer to wait until it's going to be enabled in the
>>>> macOS port.
>>>>
>>>> Michael
>>>> _______________________________________________
>>>> jsc-dev mailing list
>>>> jsc-dev at lists.webkit.org
>>>> https://lists.webkit.org/mailman/listinfo/jsc-dev
>>
>> _______________________________________________
>> webkit-gtk mailing list
>> webkit-gtk at lists.webkit.org
>> https://lists.webkit.org/mailman/listinfo/webkit-gtk
> ,
>
> _______________________________________________
> webkit-gtk mailing list
> webkit-gtk at lists.webkit.org
> https://lists.webkit.org/mailman/listinfo/webkit-gtk


-- 
Regards,
Konstantin


More information about the jsc-dev mailing list