[Coco] megaread

Boisy G. Pitre boisy at tee-boy.com
Sat May 22 07:29:26 EDT 2010


On May 21, 2010, at 11:40 PM, Roger Taylor wrote:

> At 10:27 PM 5/21/2010, you wrote:
>> Can somebody nicely tell me if the megaread command is reading in any other pattern than 1 sector increments?
>> If it's not, then Houston we have a problem.  The problem would be that deblocking to increase speed won't be realized during such a test.
>> --

Sure.  The answer is in the megaread.asm source in nitros9/level1/cmds, and it shows that an I$Read call is reading 1024 bytes at a time.

> Ok, for starters, using code that will be optimized more, I've got a 78s megaread on the newest Drive Pak driver for OS-9/NitrOS-9 Level II

That is significantly better than the test we made at the fest.  Good going.

> I doubt 32 more seconds can be squeezed out......  but I'll try   ;)

You might want to try modifying megaread to use a different size than 1024 and see if you can get improvements in timing.

Like we discussed earlier, a benchmarking utility that tracks timing internally as well as allowing write benchmarking would be desirable.  The megaread source can be the start of such a new utility.

Good luck!


More information about the Coco mailing list