Optimized arbiter using multi-level arbitration
An apparatus comprising a first sub-arbiter circuit and a second sub-arbiter circuit. The first sub-arbiter circuit may be configured to determine a winning channel from a plurality of channel requests based on a first criteria. The second sub-arbiter circuit may be configured to determine a winning channel received from the plurality of channel requests based on a second criteria. The second sub-arbiter may also be configured to optimize the order of the winning channels from the first sub-arbiter by overriding the first sub-arbiter if the second criteria creates a more efficient data transfer.
1. An apparatus comprising:
a first sub-arbiter circuit configured to determine a first winning channel from a plurality of channel requests based on a first criteria; and
a second sub-arbiter circuit configured to determine a second winning channel received from said plurality of channel requests based on a second criteria, wherein (a) said second sub-arbiter is configured to further prioritize the order of said first winning channel when determining said second winning channel by overriding said first winning channel if said second criteria creates a more efficient data transfer and (b) said second criteria of said second sub-arbiter circuit is different than said first criteria of said first sub-arbiter circuit to maintain consecutive read or write operations in a particular bank of a plurality of banks of a memory.
2. The apparatus according to claim 1 , further comprising:
a channel router circuit configured to receive a signal defining said first channel received.
3. The apparatus according to claim 1 , further comprising:
a memory circuit configured to receive said signal.
4. The apparatus according to claim 1 , wherein said second sub-arbiter uses bank history to determine said second winning channel.
5. The apparatus according to claim 1 , wherein said second sub-arbiter is configured to use a history of reads and writes to determine a subsequent winning channel.
6. The apparatus according to claim 1 , wherein said first and said second sub-arbiter are configured to statistically trade off (i) client bandwidth, (ii) client latency, (iii) fairness and (iv) memory protocol efficiency for optimal performance.
7. The apparatus according to claim 2 , wherein said plurality of channel requests are received from said channel router circuit.
8. The apparatus according to claim 1 , wherein said first winning channel from said first sub-arbiter is presented to a protocol engine circuit.
9. The apparatus according to claim 1 , wherein said second winning channel is a previous winner of said first sub-arbiter.
10. The apparatus according to claim 1 , wherein said first criteria is based on a time slice priority set in a programmable register.
11. The apparatus according to claim 1 , wherein said second criteria is based on assigning fairness thresholds to each of said plurality of channel requests.
12. The apparatus according to claim 1 , wherein said first sub-arbiter and said second sub-arbiter determine simultaneously determine respective winning channels at a regular interval of one or more clock cycles.
13. An apparatus comprising:
an arbiter circuit configured to sort a first set of input requests received from a plurality of channels based on a first plurality of criteria, wherein (i) each combination of the sorting using the plurality of criteria generates a second set of requests based on the combinations of criteria, (ii) each said second set of requests is independently arbitrated by a first priority process to produce an intermediate winning channel for each of the combinations of criteria, (iii) each of the intermediate winning channels is used as a third set of requests which are arbitrated by a priority of the combinations of the said sorting of the said plurality of criteria to generate a final winning channel and (iv) said intermediate winning channels maintain consecutive read or write operations in a particular bank of a plurality of banks of a memory.
14. The apparatus according to claim 13 , wherein said criteria includes one or more of fairness criteria, bank optimality criteria, and read/write optimality criteria.
15. The apparatus according to claim 14 , wherein said apparatus has said priority of combinations orders bank optimality criteria as a first priority, fairness criteria as a second criteria, and read/write optimality as a third priority.
16. The apparatus according to claim 13 , wherein said final winning channel always presents a result in response to any input requester.
17. The apparatus according to claim 13 , wherein said apparatus presents an indication of latency based on which of the criteria the winning channel has.
18. The apparatus according to claim 1 , wherein said second criteria maintains the same read/write direction as a previous winner within said particular bank of said plurality of banks of said memory.
19. The apparatus according to claim 1 , wherein said second criteria implements directional organization of said plurality of banks of said memory.