Why don't you just use JPC? JPC was able to run native unmodified x86 code in Java 2 years ago. It also has the much stronger double security of the JVM and being an emulator.
Ian, i don't think that an emulator running in a VM (Even with JIT support) could come close to native execution (With the few alignment and instructions limits here) for any processor intensive application.
JPC claim 20% of native speed, Native Client claim 95% of native speed... When speaking about anything doing with Raytracing, FFTs, image processing or video processing it's a _really_ big gap.
And for any other use, if 20% of native speed is ok, why bother with anything else than javascript ? Especially with SpiderMonkey and V8 becoming common place in the future.
James, I assume you didn't read the paper. If you had, you would know they use a totally different security model, and don't assume NaCl authors are trustworthy (there isn't even a digital signature to fake).
Follow
5 kommentarer :
Why don't you just use JPC? JPC was able to run native unmodified x86 code in Java 2 years ago. It also has the much stronger double security of the JVM and being an emulator.
http://www-jpc.physics.ox.ac.uk
Ian, i don't think that an emulator running in a VM (Even with JIT support) could come close to native execution (With the few alignment and instructions limits here) for any processor intensive application.
JPC claim 20% of native speed, Native Client claim 95% of native speed... When speaking about anything doing with Raytracing, FFTs, image processing or video processing it's a _really_ big gap.
And for any other use, if 20% of native speed is ok, why bother with anything else than javascript ? Especially with SpiderMonkey and V8 becoming common place in the future.
Anyway good work google.
I assume you are too young to remember Active/X and all the problems it created.
I'd rather see GWT support built into the browser providing smaller downloads and faster execution.
James, I assume you didn't read the paper. If you had, you would know they use a totally different security model, and don't assume NaCl authors are trustworthy (there isn't even a digital signature to fake).
Good post
Send en kommentar