Path: utzoo!utgpu!jarvis.csri.toronto.edu!clyde.concordia.ca!uunet!lll-winken!decwrl!ucbvax!UMIX.CC.UMICH.EDU!krowitz%richter
From: krowitz%richter@UMIX.CC.UMICH.EDU (David Krowitz)
Newsgroups: comp.sys.apollo
Subject: Re:  DMA on DN4000 AT-bus
Message-ID: <9001032011.AA02924@richter.mit.edu>
Date: 3 Jan 90 20:11:24 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 42

It should be possible to do if you can wire (lock into physical
memory) and map a large enough I/O buffer to hold the entire
stream of data. I have wired buffers of up to 128Kb under SR10.
I don't know what the limit would be under SR9.6, you'll have
to try it. You'll have to use a single buffer for the entire
transfer, since at 20K words/sec you only have 0.5 msec 
between transfers. That it not enough time to wire and map
a second buffer. The sequence of calls you want to use would
be something like:


   	pbu_$aquire (device_name, name_len, debug_flg, unit_num, status)

	pbu_$allocate_map (unit, length_of_buffer, force_flag, I/O_vir_addr, status)

	pbu_$map (unit, big_buffer, length_of_buffer, I/O_vir_addr, status)

	pbu_$wire (unit, big_buffer, length_of_buffer, status)

	set dma-enable, interrupt-enable on interface board

	pbu_$dma_start (unit, dma_channel_from_ddf_blk, pbu_$read, buffer,length_of_buffer, option, status);

	pbu_$wait (unit, timeout, quit_flg, status);

	pbu_$dma_stop (unit, dma_channel, status)

	clear dma-enable, interrupt-enable on interface board

	pbu_$unwire (unit, buffer, length_of_buffer, flag, status)

	pbu_$unmap (unit, buffer, length, I/O_vir_addr, status);

	pbu_$free_map ( unit, status)


 -- David Krowitz

krowitz@richter.mit.edu   (18.83.0.109)
krowitz%richter.mit.edu@eddie.mit.edu
krowitz%richter.mit.edu@mitvma.bitnet
(in order of decreasing preference)