Memory arbitration technique for turbo decoding
View Patent ↗A technique for resolving access contention in a parallel turbo decoder is described. The technique includes associating a plurality of buffer memories with the subdecoders so that accesses to banks of a shared interleaver memory can be rescheduled. Accesses can be rescheduled to prevent simultaneous accesses to a single bank of the shared interleaver memory based on an interleaver pattern.
1. A memory arbitrator for a parallel turbo decoder for resolving subdecoder access contention to a shared interleaver memory having multiple banks comprising:
a plurality of N subdecoders configured for parallel sub-block decoding of a turbo code;
a plurality of N buffer memories, each buffer memory associated with a different one of the plurality of subdecoders and having a depth M so as to buffer a plurality of M access attempts to the shared interleaver memory by the associated subdecoder;
a multiplexer coupled between the plurality of N buffer memories and the shared interleaver memory and configured to enable access to the shared interleaver memory by the plurality of buffer memories; and
a scheduler coupled to the plurality of subdecoders and configured to reschedule the accesses to the shared interleaver memory based on an interleaver pattern of the turbo decoder to avoid simultaneous access to at least one of the multiple banks of the shared interleaver memory.
2. The memory arbitrator of claim 1 , wherein the scheduler is further configured to reschedule the accesses to the shared interleaver memory so as to minimize aggregate delay associated with access to the shared interleaver memory.
3. The memory arbitrator of claim 1 , wherein the scheduler further comprises a read only memory configured to hold a predetermined reschedule of the accesses to the shared interleaver memory.
4. The memory arbitrator of claim 1 , wherein M<N.
5. The memory arbitrator of claim 1 , wherein the buffer memories are each configured to buffer read access to the shared interleaver memory by the associated subdecoder.
6. The memory arbitrator of claim 1 , where the buffer memories are each configured to buffer write access to the shared interleaver memory by the associated subdecoder.
7. The memory arbitrator of claim 1 , wherein each of the buffer memories is a first-in first-out memory.
8. The memory arbitrator of claim 1 , wherein each of the buffer memories is a last-in first-out memory.
9. A memory arbitrator for a parallel turbo decoder for resolving subdecoder access contention to a shared interleaver memory having multiple banks comprising:
a plurality of N subdecoders configured for parallel sub-block decoding of a turbo code;
a plurality of N buffer memories, each buffer memory associated with a different one of the plurality of subdecoders and having a depth M so as to buffer a plurality of M access attempts to the shared interleaver memory by the associated subdecoder, wherein each buffer memory is a last-in first-out memory; and
a multiplexer coupled between the plurality of N buffer memories and the shared interleaver memory and configured to enable access to the shared interleaver memory by the plurality of buffer memories.