Functional Description
5-52
ADSP-BF537 Blackfin Processor Hardware Reference
If two MDMA streams are used (S0-D0 and S1-D1), the user may choose
to allocate bandwidth either by fixed stream priority or by a round robin
scheme. This is selected by programming the
MDMA_ROUND_ROBIN_PERIOD
field in the
DMA_TC_PER
register (see
“Static Channel Prioritization” on
).
If this field is set to 0, then MDMA is scheduled by fixed priority.
MDMA stream 0 takes precedence over MDMA stream 1 whenever
stream 0 is ready to perform transfers. Since an MDMA stream is typically
capable of transferring data on every available cycle, this could cause
MDMA stream 1 traffic to be delayed for an indefinite time until any and
all MDMA stream 0 operations are complete. This scheme could be
appropriate in systems where low duration but latency sensitive data buff-
ers need to be moved immediately, interrupting long duration, low
priority background transfers.
If the
MDMA_ROUND_ROBIN_PERIOD
field is set to some nonzero value in the
range
1 <= P <= 31
, then a round robin scheduling method is used. The
two MDMA streams are granted bus access in alternation in bursts of up
to
P
data transfers. This could be used in systems where two transfer pro-
cesses need to coexist, each with a guaranteed fraction of the available
bandwidth. For example, one stream might be programmed for inter-
nal-to-external moves while the other is programmed for
external-to-internal moves, and each would be allocated approximately
equal data bandwidth.
In round robin operation, the MDMA stream selection at any time is
either “free” or “locked.” Initially, the selection is free. On any free cycle
available to MDMA (when no peripheral DMA accesses take precedence),
if either or both MDMA streams request access, the higher precedence
stream will be granted (stream 0 in case of conflict), and that stream’s
selection is then “locked.” The
MDMA_ROUND_ROBIN_COUNT
counter field in
the
DMA_TC_CNT
register is loaded with the period
P
from
MDMA_ROUND_
ROBIN_PERIOD
, and MDMA transfers begin. The counter is decremented
on every data transfer (as each data word is written to memory). After the
Summary of Contents for Blackfin ADSP-BF537
Page 42: ...Contents xlii ADSP BF537 Blackfin Processor Hardware Reference ...
Page 90: ...Development Tools 1 32 ADSP BF537 Blackfin Processor Hardware Reference ...
Page 138: ...Programming Examples 4 26 ADSP BF537 Blackfin Processor Hardware Reference ...
Page 340: ...SDC Programming Examples 6 84 ADSP BF537 Blackfin Processor Hardware Reference ...
Page 606: ...Programming Examples 9 94 ADSP BF537 Blackfin Processor Hardware Reference ...
Page 660: ...Programming Examples 10 54 ADSP BF537 Blackfin Processor Hardware Reference ...
Page 720: ...Electrical Specifications 11 60 ADSP BF537 Blackfin Processor Hardware Reference ...
Page 840: ...Programming Examples 13 42 ADSP BF537 Blackfin Processor Hardware Reference ...
Page 876: ...Programming Examples 14 36 ADSP BF537 Blackfin Processor Hardware Reference ...
Page 938: ...Programming Examples 15 62 ADSP BF537 Blackfin Processor Hardware Reference ...
Page 958: ...Programming Examples 17 12 ADSP BF537 Blackfin Processor Hardware Reference ...
Page 986: ...Programming Examples 18 28 ADSP BF537 Blackfin Processor Hardware Reference ...
Page 1162: ...G 26 ADSP BF537 Blackfin Processor Hardware Reference ...
Page 1218: ...Index I 56 ADSP BF537 Blackfin Processor Hardware Reference ...