Experimental Physics and
Industrial Control System

Till Straumann <[email protected]> · Tue, 15 Apr 2003 08:51:15 -0700

Marty Kraimer wrote:

Mark Rivers wrote:
Marty wrote:

If programed transfers are used instead of DMA there are two big
advantages.

1) A24 DMA is no longer required. Thus it will no longer be necessary
to modify the board support packages in order to support the NI1014.

I'm not quite sure I understand what you mean by this. Are you
suggesting eliminating support for A24 DMA entirely in the BSP, or
just for some quirk of the NI1014? We use the Kinetic Systems 2917
VME/CAMAC adapter, which uses A24 DMA. We definitely need to be able
to continue to use these, there are a dozen or more in use on
beamlines where they control a critical CAMAC device for multi-element
fluorescence detectors.
Mark Rivers

Thanks for the info.
My interest is the NI1014. I don't see much reason for it to use DMA.

Sorry for putting in my 2 cents:
I don't know anything about the NI1014 or the CPU you are using
nor the amount of data you want to transfer.
But: using DMA for data transfer from/to VME into/out of memory
is almost always a good idea on modern CPUs. PIO on such a slow
bus can stall a fast CPU for a lot of cycles.
-- Till
Marty Kraimer

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System