[thunderbird32]

So it looks like the R4000 core code in mame is taking about 60% of the total CPU time and if there is room for improvements, that would be the first place to start. I've tried with -drc and -nodrc and it didn't make any difference, so in infer that there is no dynamic recompile code active, which could potentially create lots of speed improvements.

We'll have to see what the future brings.

I'm not expecting any speed improvement, because according to the perverted philosphy of the MAME team, dynamic recompilation, HLE and graphical acceleration are detrimental to "accuracy". On the contrary, I expect the emulation to become even slower, in the name of "increased accuracy".

I'm sorry to be this blunt, but as long as a complex platform like this is only emulated by MAME, it might as well be not emulated at all: even on the fastest machines, it will always be ridiculously slow and unusable. Sure, the MAME team might lull you into a false sense of hope by telling you lies such as "code optimization is not necessary because hardware will eventually become fast enough on its own", but history shows that the sluggishness of MAME has increased much more than hardware speed.

At this point, if this was an Italian emulation newsgroup 10 years ago, MAME zealots would jump at my throat stating that "the purpose of MAME is not to run the software it emulates" and "being able to use the emulated software is a nice side effect". I hope nobody in this forum drinks that Kool-Aid.

MAME is open source these days, if someone wants to put in the work to move the emulation to a dedicated emulator, the hardest work is already done (I would think). The MAME devs are the only ones who have been willing to put in the work to get SGI emulation working to this advanced state, as yet.