Date   : Thu, 09 Mar 2006 12:45:23 +0100
From   : John Kortink <kortink@...>
Subject: Re: Warning: Sad case on list!

On Thu, 9 Mar 2006 11:05:08 +0000, Andy Armstrong wrote:

>On 9 Mar 2006, at 10:38, John Kortink wrote:
>> The depth of that FIFO won't matter too much, probably.
>> The ARM side will be waiting for 'slomo' to draw things
>> most of the time anyway. Saving a few cycles just on
>> getting the VDU bytes through won't make that much of
>> a difference.
>
>If you time how long it takes /just/ to draw things and how long it  
>takes /just/ to calculate things then - given a big enough FIFO the  
>smaller of those two times is the fastest you can get. If it's  
>currently slower than that then a sufficiently large FIFO will speed  
>it up :)

Obviously. But a FIFO can only speed up communications, not
actual processing. If one side is continuously waiting for
the other to complete an earlier operation, as opposed to
completing communications via the FIFO, there's no point
in having a larger FIFO. In practice this is very often
the case, certainly in non-interactive and pretty linear
processing like graphics. Speeding up processing is a
matter of balancing computing tasks first and foremost,
and only once these are balanced, perhaps of FIFO size
(and then only up to the point where it can contain the
largest single communication between the two sides).


John Kortink

-- 

Email    : kortink@...         
Homepage : http://www.inter.nl.net/users/J.Kortink

GoMMC, the ultimate BBC B/B+/Master storage system :
http://web.inter.nl.net/users/J.Kortink/home/hardware/gommc