Ram vs Rom speed

[emiespo]

Originally Posted by Thomas Richter

Check closer. The TTx registers always exist. They have a different name on the ECs, though the register mapping is exactly identical (obviously).

Worth adding that the 68030.library also (partly*) fixes Akiko on CD32s with a 68030. It marks its address non cacheable, so that when the CPU writes and reads from the same location multiple times, it won't just get back the cached data from it. Tested on a TF330 with an EC030.

* This is a partial fix as the OS will try to detect Akiko *before* the 68030 lib gets loaded, and marks it as broken anyway (so disabling the fast c2p routines).

I make a hacky fix that simply restores the hardware pointer, it would be nice to do a proper patch of the OS routine to turn the cache temporarily off when Akiko is getting assessed.

EDIT: sorry bumped in this thread whilst I was researching info on the audio.device I am trying to write a OS-friendly CDXL player on 1.3, and it seems I can't really get a non flickering display nor perfect audio with the docs I found around (RethinkDisplay() and CMD_STOP + CMD_WRITE (stereo) + CMD_START to sync channels).

[Thomas Richter]

Actually, Akiko is no longer used by the graphics.library, so that problem went away. The Os C2P functions were completely redone as of 3.1.4.

[emiespo]

Originally Posted by Thomas Richter

Actually, Akiko is no longer used by the graphics.library, so that problem went away. The Os C2P functions were completely redone as of 3.1.4.

Ah interesting didn't know that! Well Akiko does still do something up to a 68030@50MHZ (where it is marginally faster than the hybrid c2p CPU/Blitter, with DoomAttack).

Still much faster for the standard 680EC20 of a CD32 (basically "copyspeed"). Are the new routines blitter-assisted? or just plain CPU?

Are there 3.1.4 ROMs for the CD32 too? I thought 3.2 was the first system to actually have official working ROMs. I should probably get those for my CD32...

[Thomas Richter]

No, the blitter really does not pay off here. The trouble is that the 3.1 and before versions even followed a two-step approach where in the first step the c2p conversion was run, and in a second step, another blitter assisted copy was performed, so it was really slow.

The new functions are just plain CPU, but they do not use the typical "roxxler"-approach, but something faster. There is no 3.1.4 for the CD32, due to lack of hardware. For 3.2, I do not know.