🚀 go-pugleaf

In article <1js5r7INNevj@rs2.hrz.th-darmstadt.de> wallmann@backus.pu.informatik.th-darmstadt.de (Natuerlich!) writes:
>In article <1993Jan22.162113.13989@bsu-ucs> 01mbmccabe@leo.bsuvc.bsu.edu writes:
>>> You see, the CPU doesn't spend all it's time accessing memory.  Many
>>> instructions involve no memory access at all.  Others access memory,
>>> but also spend a lot of time processing the instruction.  And if an
>>> instruction or data is already in the on-chip cache, it will be just as
>>> fast regardless of the bus width.  
>What about if a DMA transfer is going on, is this done over the video 
>(32 bit part) ? 

Good question, dunno what a good test would be.
>
>>can someone make a list of typical sorts of programs that would be really
>>"memory-intensive" so as to really suffer from this 16-bit memory bus?
>>
>Anything that's video intensive f.e.

Note that in the monochrome video modes, the bustest program (posted here
last December?) shows no slowdown whatsoever for 16 bit accesses, and less
than .01% slowdown for 32 bit accesses. (I.e., the 68030 cache delivers
100% of the theoretical performance.) I posted my timing results to both
this group and comp.sys.amiga.advocacy; I am still waiting to see the
equivalent results for an A1200. If you need the source code again I will
send it to you; I rewrote it in MIT syntax (for GAS) and still have the
original Motorola syntax posting.
-- 
  -- Howard Chu @ Jet Propulsion Laboratory, Pasadena, CA

There's a rough border between genius and insanity, but I'm a dual citizen.