Joined: 18-Jan-2003 Posts: 2320
From: Minnesota, USA
@ppcamiga1
By the time Commodore went under in 1994, Unix, OS/2 and other multitasking kernels were becoming commonplace. But the chipset remained unique until the Minimig core came out for the FPGAs much later.
One thing I do agree with you about AGA: the blitter and copper should have been updated to use the page-mode bus accesses for a 4x speed increase in 64 bit aligned operations. That would have increased the resolution of the copper-chunky 12-bit graphics color mode to 320×200 and made the blitter competitively fast as well.
Last edited by Samurai_Crow on 01-Nov-2021 at 01:51 PM.
Joined: 6-May-2007 Posts: 11228
From: Greensborough, Australia
@ppcamiga1
I'd actually like to see some tests for blitter vs. CPU speed. And for that I mean real operations. It's easy to say the CPU is faster when doing a straight copy of bitplanes. But the blitter is more complicated than a simple memory block copier. No doubt the 020 with the exra bitfield operations can compete with the blitter. But how fast is the CPU when doing real blitter simulation?
That is, copying an interleaved bitmap. Copying an offset bitmap into another offset bitmap. drawing lines.
Interleaved requires doing lots of smaller burst copies for each line separately against an easy full bitmap.
Offset copy requires to rotate and mask data as it transfers the bitmap in source, desination or both as well as perform a logic operation to merge data.
Line drawing requires lots of little shift and mask operations including texture mapping a straight line pattern into a bitmap line at any angle.
Not saying it can't be done. But, easier said than done! When hardware does all these operations for free the CPU should be way faster if it is able to efficiently simulate the exact same operations at a faster rate with less expense on the CPU.