NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/sys/amiga/programmer/7373

Internet Message Format | 1996-08-05 | 2.8 KB

Path: informatik.tu-muenchen.de!fischerj
From: [e-mail] (Juergen "Rally" Fischer)
Newsgroups: comp.sys.amiga.programmer
Subject: Re: CHIP RAM speed test resul
Date: 15 Apr 1996 14:05:16 GMT
Organization: Technische Universitaet Muenchen, Germany
Distribution: world
Message-ID: <4ktl2s$[e-mail]>
References: <4j6jv0$[e-mail]> <[e-mail]> <[e-mail]> <4k1kk3$[e-mail]> <4k1oqk$[e-mail]>
NNTP-Posting-Host: hphalle5.informatik.tu-muenchen.de
Originator: [e-mail]


In article <4k1oqk$[e-mail]>, [e-mail] (K. C. Lee) writes:
|> Organization: The National Capital FreeNet
|> Lines: 41
|> Sender: [e-mail] (K. C. Lee)
|> Message-ID: <4k1oqk$[e-mail]>
|> References: <4j6jv0$[e-mail]> <[e-mail]> <[e-mail]> <4k1kk3$[e-mail]>
|> Reply-To: [e-mail] (K. C. Lee)
|> NNTP-Posting-Host: freenet2.carleton.ca
|>
|>
|> Juergen "Rally" Fischer ([e-mail]) writes:
|> >
|> > ok, so why my 020 needs _12_ cycles , i.e. _846_ ns (!!!!) to load a
|> > byte/.w/.l from chipmem ?
|>
|> Remember that the CPU is not the only one that have access to your chip
|> ram... ( I don't know too much about AGA 64-bit fetch and all the funny

This is not the reason. same goes for master dma off.
This also won't explain why reads are slower than writes although the
020 can read in 6 cyles, i.e. beyond 8 cycles.

|> screen modes.)  Your CPU have 1/2 of the bandwidth available and also have
|> to wait if it trys to access the memory at the wrong time.
|>
|> > that's unlogic, because any acess should be delayed by a fix amount
|> > of time. but: load 6 -> 12 cycles (difference: 6), store 4 -> 8 cycles
|> > (difference: 4).
|>
|> May be it is the memory access pattern ?  I wouldn't even try to figure
|> things in a CPU with caches.

It is not the cache, my cpu got no datacache.
Anyway, wrong results dues to caching would mean _faster_ results, not
slower ;->

|>
|> > BTW imho it should cost almost nothing to add a A3000-alike chipmembuffer,
|> > did you do it in walker ? I really hope so. 4 longwords would even
|>
|> What chip buffer is that ?  The 020 already have 32-bit access to the chip
|> ram.

The A3000 has 32bit acess, too. but it is buffered so cpu can go on
with fastmemacess while the write is done.

|> >    [e-mail] (Juergen "Rally" Fischer)   =:)
|>
|> K. C. Lee
------------------------------------------------------------------------
   [e-mail] (Juergen "Rally" Fischer)   =:)