Assembler

[omarsis81]

Probably a dumb question, but here it goes: could we get significant performance benefit IF PCem would be written in assembler? I know it´s almost impossible because of its difficulty, but I ask anyway

[Battler]

The CPU dynamic recompiler already is written mostly in assembler.

[SarahWalker]

You might get a very small gain? Like a couple of percent. Not at all worth the months, if not years, it would take to rewrite. And it would ruin any hopes of portability - I still want to get PCem running on something ARM based at some point.

[omarsis81]

That´s what I thought, not worth on today's processors.
Thank you for your thoughts

[omarsis81]

I still want to get PCem running on something ARM based at some point.

Just out of curiosity, do you have a rough idea of what could be the maximum CPU PCem can achieve on a Raspberry Pi 3 at current stage? 386, 486/66 maybe?

[SarahWalker]

Nothing at the moment as I'm pretty sure it won't compile.

With whatever fixes are required... something 386SXy maybe? It's not going to be even remotely fast.

[omarsis81]

Nothing at the moment as I'm pretty sure it won't compile.

With whatever fixes are required... something 386SXy maybe? It's not going to be even remotely fast.

Ok, I just wanted to have an idea. Thanks!
I noticed the Pi is widely used for emulation

[leilei]

I did try once. The code doesn't have enough generic paths to easily switch on (or at least, the x86 parts aren't def'd out enough) so it's not immediately compilable.

It's mostly the references to recompiler parts that occur outside the recompiler code at fault.

I too am also expecting something around a 386 interpreter. No Voodoo ever.

a pi3 booting a OS that's just a tiny linux that only runs pcem while housed in an old computer case would be something. though it'd need proper vga out and be able to make the same ambient/idling noises for it to really work.