Hoho, I've probably got it. And thanks to... quake ;-) Today I was trying to get -m68060 quake compilation done but I received some strange errors, nothing worked... After short discussion with Vincent I realized he is using (for his quake tests -- it's really cool source code for testing purposes ;-))) unmodified gcc, i.e. multilib is generated for m68000, m68020-60 + mshort variants + m68020-60=m68060 so when he tells gcc to use -m68060, he was using m68020-60 libs.Any idea what to enable to make it working for 060 compilation?
I was, on the contrary, used 060 multilib enabled + 060 mintlib. So it should work except... except the fact I overlooked there's 3rd player involved -- math lib. Sparemint/Easymint uses fdlibm, Vincent PML. Both of them are compiled for mshort, m68000 and m68020-60. Now you probably know what was the problem -- missing m68060 math lib! So I patched Vincent's makefiles for PML to produce 060 math lib and voila, quake worked!
How does this rely on gcc? If you remember, there was a loooooooooooooot of error messages about... about math functions! Simply, gcc wanted to use some nasty math stuff, it was told to compile for 060 so it was searching for... 060 math lib which wasn't found so standard 68000 math lib was used (with missing / other named functions probably). I'm going to try it this night, we'll see.
Funny observation #1 -- if you use '-g' switch for -m68060 stuff, m68000 libc is used (actually, libc_g but since mintlib provides this only for 68000, this one is used)
Funny observation #2 -- quake is able to run in aranym-jit @ 1.66 GHz dualcore in 800x608 in native videl resolution nearly as fast as 320x200 on original atari (in fact, 320x200 on aranym runs at 110 FPS ;-) When I will be bored, I might try to use VDI services to set this screen and use FVDI for rendering.. this could be quite killing ;-)
-- MiKRO / Mystic Bytes http://mikro.atari.org