Shumway - PDF.js for Flash

[Omega X]

The real alternative was silverlight. But flash is so ingrained that it will take a lot more effort to knock it off the tracks.

[cyko_01]

Is there a (active) blog or something where we can follow shumway development?

[Omega X]

https://github.com/mozilla/shumway

[iwod]

Another candidate for asm.js , along with pdf.js

[cyko_01]

Another candidate for asm.js , along with pdf.js

I was thinking the same thing

[_Alexander]

Anything *.js will never be as optimized as native C++ code with hand optimized assembly for performance critical areas.
Considering that some of you have problems as is with rendering swf, a .js solution will only make things worse.

Mozilla should invest time in fixing its plugin support - flash plugins jump on a 4.5Ghz quad core. That is not good coding, that is absolute embarrassment.

[bjherbison]

Anything *.js will never be as optimized as native C++ code with hand optimized assembly for performance critical areas.
Considering that some of you have problems as is with rendering swf, a .js solution will only make things worse.

On the other hand, code written by Firefox in JavaScript will be much better integrated with Firefox than anything using the plugin API. That could result in performance benefits.

Mozilla should invest time in fixing its plugin support - flash plugins jump on a 4.5Ghz quad core. That is not good coding, that is absolute embarrassment.

I'm not sure what changes you are suggesting. If you have ideas for specific Firefox code changes that would improve plugin performance (without introducing bugs) the Firefox developers would be glad to hear your suggestions, but last I looked the API interface was fairly straightforward so I expect there's not much room for improvement.

A different plugin API would not be well received by plugin developers as they would need to change their code and have different code for Chrome and Firefox.

[Sintendo]

A different plugin API would not be well received by plugin developers as they would need to change their code and have different code for Chrome and Firefox.

Didn't Chrome already have a different API for plugins?

[Ver Greeneyes]

Anything *.js will never be as optimized as native C++ code with hand optimized assembly for performance critical areas.

I think technically the JITed code could be just as fast, especially with dynamic optimizations that are statically invalid, but the problem is compilation overhead. JIT compilers will probably never be able to do some of the computationally expensive optimizations that normal compilers can attempt because compilation time directly impacts performance. Doing compilation on a background thread might help with that, of course, but I'm not sure how aggressive JIT developers are willing to be with that (especially with the possibility of parallel execution cropping up more and more). As long as the code is enough like C++ though, asm.js might help get a lot closer to the ideal.

There is of course the question of whether Flash is as slow as it is / often seems to be because its execution is simply that complex, or if its codebase is simply a mess of ancient code, hacks and workarounds, and could be made faster by rewriting it from scratch. Without having seen the code I'm not going to make any statements about that, but the instability of Flash and its tendency to hang does drive me up the wall sometimes. For one thing, I still can't run flash with its Protected Mode turned on, despite how long it has been since it was first introduced. I tried it again a few weeks ago, and the result was that the amount of hangs and crashes I saw in Firefox more than tripled.

[Zlip792]

A different plugin API would not be well received by plugin developers as they would need to change their code and have different code for Chrome and Firefox.

Didn't Chrome already have a different API for plugins?

Chrome use PPAPI interface for Plugins.

[cyko_01]

Anything *.js will never be as optimized as native C++ code with hand optimized assembly for performance critical areas.

They are fully aware of that. They are setting the target at 2x native c++ which is a reasonable goal and a huge improvement.

Mozilla should invest time in fixing its plugin support - flash plugins jump on a 4.5Ghz quad core. That is not good coding, that is absolute embarrassment.

Last I checked it was adobe, not mozilla that makes the flash plugin...although they are trying to kill it with shumway - yes, it has gotten so bad that they are convinced they can do it better with javascript/html5

[MINIDOU]

the only flaw i see with firefox implementation of flash, is the lags that appears when one or several tab uses flash (it appears more when using at least 2 tabs) every time there is an action in the browser (like open a tab with some js)
it doesn't happen in chrome, and it's pretty easy to notice it when you watch some twitch streams while browsing

not sure if OMTC will/can change that

[Drumbrake]

Does Shumway actually work ? I've installed it and now every youtube page has this shumway widget,I click on it and each time a normal flash player plays the video...

Is there some other setting I need to tweak to make this work ?

[faulty_fox]

Shumway is in a very early stage of development. And what they are trying to achieve is very difficult.
Currently hardly anything works with it, complex stuff like the youtube player won't work any time soon.

Edit : https://github.com/mozilla/shumway/issues/187

[Drumbrake]

I see,that will take (quite) a while...I hope this works out,eventually.
Thanks.