Hello my friends. I have two different DSPs: a C6711 and a C6713 DSKs boards,
and I need that one of them makes one (or some) specific task(s), and the other
makes the rest, at the same time. What is the best way to do it with CCS (in my
case the v3.1)??. I have thought to do it across McBSP port, but the algorithm
that I have poses some problems.
Thanks for your Help.
Andres T.
Two TMS320C6000 DSPs with different Tasks between them
Started by ●October 29, 2008
Reply by ●October 29, 20082008-10-29
Andres,
On Wed, Oct 29, 2008 at 4:53 PM, wrote:
> Hello my friends. I have two different DSPs: a C6711 and a C6713 DSKs
> boards, and I need that one of them makes one (or some) specific task(s),
> and the other makes the rest, at the same time. What is the best way to do
> it with CCS (in my case the v3.1)??. I have thought to do it across McBSP
> port, but the algorithm that I have poses some problems.
If you make the 6711DSK the master and the 6711DSK the slave [all data
transfers must be initiated by the master, but the slave could drive
an interrupt to the master], you can follow Jeff's answer from the
last time that this was posted [with a couple of comments added].
from Jeff's previous reply:
Another possibility -- one that I've seen some DSP labs / Professors
assign before -- is to use EMIF on DSK board A to control HPI on DSK
board B. The student typically runs a very short cable between the
daughterboard connectors.
I seem to remember this method can be implemented without additional
circuitry: the /CEn line on the DSK A
daughterboard is tied to /HCS, ARDY is tied to HRDY, ARW is tied to
HR/W, and addr lines 2:3 are used for HCNTL0 and HCNTL1. Plus 16-bit
data bus. One complication would be the HWIL bit (half-word signal),
it would need to connect to an EMIF address line also.
If you configure the daughtercard EMIF address 0xA0000000 for 16
bits, you could use BE0 to control HWIL [instead of an address line].
This would allow the 6711DSK to DMA to/from the 6713DSK. The
requirements to support this would be a 2-3 inch cable with
approximately 26-30 wires connected.
Note to OP -- this is just a general description, so don't try it
without first studing the EMIF and HPI interfaces
very carefully.
-Jeff
>
> Thanks for your Help.
>
> Andres T.
--
www.dsprelated.com/blogs-1/nf/Mike_Dunn.php
On Wed, Oct 29, 2008 at 4:53 PM, wrote:
> Hello my friends. I have two different DSPs: a C6711 and a C6713 DSKs
> boards, and I need that one of them makes one (or some) specific task(s),
> and the other makes the rest, at the same time. What is the best way to do
> it with CCS (in my case the v3.1)??. I have thought to do it across McBSP
> port, but the algorithm that I have poses some problems.
If you make the 6711DSK the master and the 6711DSK the slave [all data
transfers must be initiated by the master, but the slave could drive
an interrupt to the master], you can follow Jeff's answer from the
last time that this was posted [with a couple of comments added].
from Jeff's previous reply:
Another possibility -- one that I've seen some DSP labs / Professors
assign before -- is to use EMIF on DSK board A to control HPI on DSK
board B. The student typically runs a very short cable between the
daughterboard connectors.
I seem to remember this method can be implemented without additional
circuitry: the /CEn line on the DSK A
daughterboard is tied to /HCS, ARDY is tied to HRDY, ARW is tied to
HR/W, and addr lines 2:3 are used for HCNTL0 and HCNTL1. Plus 16-bit
data bus. One complication would be the HWIL bit (half-word signal),
it would need to connect to an EMIF address line also.
If you configure the daughtercard EMIF address 0xA0000000 for 16
bits, you could use BE0 to control HWIL [instead of an address line].
This would allow the 6711DSK to DMA to/from the 6713DSK. The
requirements to support this would be a 2-3 inch cable with
approximately 26-30 wires connected.
Note to OP -- this is just a general description, so don't try it
without first studing the EMIF and HPI interfaces
very carefully.
-Jeff
>
> Thanks for your Help.
>
> Andres T.
--
www.dsprelated.com/blogs-1/nf/Mike_Dunn.php
