Anonymous
Not applicable
Jul 13, 2009
02:04 PM
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Jul 13, 2009
02:04 PM
Hi,
In AN4051 (Doc No. 001-15284) about the GPIF Flow State feature for
UDMA, there is a strange (to me) quote on page 4 :
"It should be noted that in the case where MSTB is a CTL pin, the
GPIF will produce at most one extra active edge on MSTB in response to
a not ready indication from the slave."
So am I interpreting this correctly : if the GPIF is in a flow state
and reading data from an external slave, and the slave says "no data
available", then the GPIF *might* (or might not) go ahead and read the
data anyway ?
If this is the case, how does one deal with this situation ? In my
case, I'm reading from a FIFO inside an FPGA, so I can configure it to
have an "almost empty" flag in addition to an "empty flag". It would
seem that the logic would be to pass the FX2 a signal like :
RDY0 = (almost_empty) OR (empty),
and have the flowstate use RDY0 to throttle the data. In principle
this should transfer all but the final word from the FIFO. ("almost
empty" means there is one word remaining.) That, I suppose, could be
read from the FIFO by a normal (i.e., not flow-) state of the GPIF.
Is this the right approach ?
Unfortunately, I gave it a shot and it's not quite working yet.
Before pouring too much effort into it, I'd like to know if there's an
easier way...
BTW, I am basically just using the GPIF Primer example that
writes/reads to an external FIFO. I got rid of the OUT EP and the
write, and instead pipe external video data into the FIFO in the
FPGA. For the flow state logic I use
func = (FifoFlag) OR (empty),
if (func == 0) assert RE#, clock data
if (func == 1) de-assert RE#
where "FifoFlag" is the FX2's FULL flag for the IN EP, and "empty" is
the fpga FIFO empty flag.
Thanks
In AN4051 (Doc No. 001-15284) about the GPIF Flow State feature for
UDMA, there is a strange (to me) quote on page 4 :
"It should be noted that in the case where MSTB is a CTL pin, the
GPIF will produce at most one extra active edge on MSTB in response to
a not ready indication from the slave."
So am I interpreting this correctly : if the GPIF is in a flow state
and reading data from an external slave, and the slave says "no data
available", then the GPIF *might* (or might not) go ahead and read the
data anyway ?
If this is the case, how does one deal with this situation ? In my
case, I'm reading from a FIFO inside an FPGA, so I can configure it to
have an "almost empty" flag in addition to an "empty flag". It would
seem that the logic would be to pass the FX2 a signal like :
RDY0 = (almost_empty) OR (empty),
and have the flowstate use RDY0 to throttle the data. In principle
this should transfer all but the final word from the FIFO. ("almost
empty" means there is one word remaining.) That, I suppose, could be
read from the FIFO by a normal (i.e., not flow-) state of the GPIF.
Is this the right approach ?
Unfortunately, I gave it a shot and it's not quite working yet.
Before pouring too much effort into it, I'd like to know if there's an
easier way...
BTW, I am basically just using the GPIF Primer example that
writes/reads to an external FIFO. I got rid of the OUT EP and the
write, and instead pipe external video data into the FIFO in the
FPGA. For the flow state logic I use
func = (FifoFlag) OR (empty),
if (func == 0) assert RE#, clock data
if (func == 1) de-assert RE#
where "FifoFlag" is the FX2's FULL flag for the IN EP, and "empty" is
the fpga FIFO empty flag.
Thanks
0 Replies