Nebulasray visuals improvement in MAME at last! (WIP)

[BPzeBanshee]

Key commits:
http://git.redump.net/mame/commit/?id=5 ... 4c643d32ce
http://git.redump.net/mame/commit/?id=b ... 98576b451d

I don't think this one's a very popular title here anyway but it's been a long-standing video bug in MAME for some time now. For comparison, how it currently looks in public MAME builds, and the actual board videos.

Haven't really got time or energy to compile my own builds for now but will get around to it soon. Haze may also update his UME builds to support it if he thinks it's worthwhile so soon after the last build (and the rest of the team figure out all that same-label-different-hash nonsense that's currently the rage in his blog's comments section).

Also, not sure if this happened in 0.154 or after but there's been some recent alterations to the CV1000 driver and the SH4 CPU that runs it (no that's not a typo, SH3 is a subsidiary of the SH4 in MAME structure), those who got borderline performance should probably give it a spin now to see how it behaves.

[IseeThings]

Nebulasray is currently very slow, and the sprites glitch quite badly with the raster interrupt, so it sill needs work, the driver wasn't optimized for a scanline based drawing case, and that effect requires every scanline to be drawn individually because they change the scroll values mid-display on every line.

CV1K is significantly slower now, some of the changes made were basically removing optimizations I did in the past, in order to convert things to proper C++. Until you see 'SH3/4 recompiler' things are only going to get worse.

[BPzeBanshee]

Ah, so it's not quite done but 'getting there' then. I'll edit the thread description for accuracy then.

Do you really think a SH4 recompiler is going to happen? It seems like it'll have to if things like NAOMI is ever going to be practical to use but I find it difficult to see anything happening soon with that.