A new Odyssey is in development

@tao

Quote:
The WebKit browser engine (and Chromium and Firefox too) have Javascript JIT compilers written for x86/x64 and ARM CPUs. So if we switched to x64 or ARM then we'd be able to use the ready-made JIT.

If we want the Javascript JIT compiler for PowerPC,** then we'd need someone to write and maintain that code.

Hans

** And we DO want the JIT compiler, because it boosts the performance of Javascript code. Modern websites tend to be Javascript heavy...

this is a great news!

@tlosm

But then we need a new Petunia JIT for ARM , or else 680x0 software will run slow on AmigaOS4.x under ARM.

That’s kind of my point, we moving to a unsupported CPU, and break everything… if we need a new everything.. this not the time to keep what is broken.

ExecSG team should by now have identify the biggest problems with the OS design, and its time to think about how to avoid these problems. L1 cache issues, funky undocumented behavior of forbids / permit and disabled / enable


Unless we are talking about some kind of stop gap solution, based on QEMU, with some backdoor libraries into the host OS to accelerate parts using ARM optimized code on critical parts. Note that’s tricky sense this breaks the OS4 scheduler, it’s like running inside forbid permit, where the emulator has to wait for host to execute a function.

@LiveForIt
Quote:Unless you use a very slow ARM CPU, like the Orange Pi in the A600GS/A1200NG, you don't need a new m68k ARM JIT. With any current ARM CPU, especially if you can use multiple cores, just a fast PPC JIT should be emulating the PPC m68k Petuinia JIT generated PPC code inside the ARM PPC JIT faster even than on the best PPC OS4 systems, X1000 and X5000.
Maybe not quite as fast as Trance on a 970 Mac running MorphOS, though.

@Hans

Thanks. What confused me is that porting the JavaScript JIT (and apparently also the Webassembly engine...I see the Power64 Linux fans are still struggling with that) must be an order of magnitude more difficult than porting the rest of the browser. I'm assuming Timberwolf never had a JS JIT either. I also hadn't considered that the interpreter and the JIT are two discrete engines and that you can have one without the other. All good to know.

I read a blog entry a year ago or more, which suggested that a well-tuned interpreter could rival a JIT. Wish I could find it now to get your opinion. As it stands, I'll side with you.

@nubechecorre

I still have faith in IBrowse, too.

@tao

Quote:

Majority of that is servers, not desktop systems, unless you count old G5 Mac’s, the loss of Apple naturally means less people work on it and maintain it. There was a browser called TenForeFox that supported JS JIT for PowerPC many years ago. But that’s the Gecko web engine.

When you have unique OS, you need to support it, EUAE did not have BSDScoket support up until recently for AmigaOS4, and AHI support is not working, unless you count AHI on Paula. The lack of developers in this community and the split in community between MorphOS and AmigaOS (doubling up the work, does not help.), thankfully some software is open source. The community is also too small to support the developer’s economy. This means part time work and slow progress. The few developers around here are basically Plummers running from job to job, fixing pipes. Sometimes you need to take on some work you’re not experienced with. Again, slow progress.