View Single Post
Old 26 May 2012, 02:40   #54
Registered User
Join Date: Dec 2009
Location: Madrid / Spain
Posts: 27
Originally Posted by Ezrec View Post
The JIT emulators I have used in the past usually take a ~40x performance hit - so on a 1Ghz 'host', you end up with a about 30Mhz-25Mhz of emulated speed.
That's really bad performance.

Often, you will get 'apparently higher' speeds (ie Rosetta on the Mac OSX platform), because the JIT is only recompiling the application - all drawing, disk, os, and input calls are handled by the host OS.
Apple software is overrated.

That would not be the case for the system you discuss - it would be emulating the entire system, not just the applications, so the 40x speed hit effects everything.
No problem, that 40x speed hit doesn't affect good JITs.

Find me a m68k JIT for PPC hosts with better performance numbers, and you would then be able to make a good case.
Trance. Feel free to send me any m68k bruteforce test and I'll execute it.

Example of Lame m68k running on trance powerup (old version) on 603e/200Mhz with no L2 cache and slow edo ram running twice faster than a 68040/40Mhz:

Example of Ocean Machine from The Black Lotus running on AGA+JIT on a 603e/240Mhz (no rtg, no ahi, just the ppc writting to custom chip registers and chipram using slow ram)
[ Show youtube player ]

A peg1 running at 600Mhz runs rings around 060s running m68060 code.

And despiting Petunia is slower than Trance it's still faster than 060.
Crumb is offline  
Page generated in 0.03969 seconds with 10 queries