Network cards use several technique for DMA interrupt mitigation. 

When sending data to the card. The command steam is usually located in on
card memory to reduce bus access. A the end of the stream you place a
special op code that causes the GPU to slowly loop on that opcode. Slow is
a relative term. When you are ready to run more commands you use the
shared memory interface to alter the opcode. The copy is usually one word
so that it can be altered atomically. This allows you to add more command
where or not the GPU has reached the special opcode. This eliminates the
need for an end of stream interrupt. In this model GPU input command
stream processing is never truly halted.

When transferring data off the card newer network cards have a programmable
interrupt mask. The idea is that at the end of the CPU processing each
packet it resets the timer and checks to see if another packet is waiting.
This allows you to receiver many packets at high speed without generating
an interrupt. When the packets stop coming you won't have reset the
interrupt mask and the next packet will generate an interrupt.

As for page flipping, the ATI cards can be programmed to automatically
flip on each vertical retrace. You can use this to eliminate the need for
retrace interrupts.

-- 
Jon Smirl
[EMAIL PROTECTED]



_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)

Reply via email to