Logical elements with switchable connections in a reconfigurable fabric
View Patent ↗Techniques are disclosed for designing a reconfigurable fabric. The reconfigurable fabric is designed using logical elements, configurable connections between and among the logical elements, and rotating circular buffers. The circular buffers contain configuration instructions. The configuration instructions control connections between and among logical elements. The logical elements change operation based on the instructions that rotate through the circular buffers. Clusters of logical elements are interconnected by a switching fabric. Each cluster contains processing elements, storage elements, and switching elements. A circular buffer within a cluster contains multiple switching instructions to control the flow of data throughout the switching fabric. The circular buffer provides a pipelined execution of switching instructions for the implementation of multiple functions. Each cluster contains multiple processing elements, and each cluster further comprises an additional circular buffer for each processing element. Logical operations are controlled by the circular buffers.
1. A computer-implemented method of designing a reconfigurable fabric comprising:
implementing a plurality of logical elements;
connecting configurable connections between and among the plurality of logical elements; and
implementing a plurality of rotating circular buffers residing on a semiconductor chip where the plurality of rotating circular buffers contain configuration instructions which control different configurable connections between and among the plurality of logical elements and dynamically change operation of the logical elements based on the rotating of a circular buffer within the plurality of rotating circular buffers, wherein the plurality of rotating circular buffers have differing lengths and resynchronize with a zeroth pipeline stage for each of the plurality of rotating circular buffers.
2. The method of claim 1 wherein the rotating of the circular buffer within the plurality of rotating circular buffers enables two or more logic functions to be implemented in run time.
3. The method of claim 1 wherein the configuration instructions are preprocessed by a compiler and statically scheduled to avoid collisions among the plurality of logical elements.
4. The method of claim 3 wherein the preprocessing inserts further instructions to prevent the collisions.
5. The method of claim 4 wherein the further instructions include one or more of storage instructions, sleep instructions, or no-op instructions.
6. The method of claim 3 further comprising replacing multiple instructions with a single fan-in instruction in preprocessing.
7. The method of claim 1 wherein the configurable connections are time multiplexed.
8. The method of claim 1 wherein the circular buffer within the plurality of rotating circular buffers is programmed and instructions are preprocessed to generate input to the circular buffer within the plurality of rotating circular buffers for dynamic programming.
9. The method of claim 1 wherein the circular buffer within the plurality of rotating circular buffers controls passing data to a quad of processing elements through switching elements, where each of the quad of processing elements is controlled by four other circular buffers, where data is passed back through the switching elements from the quad of processing elements where the switching elements are again controlled by the circular buffer within the plurality of rotating circular buffers.
10. The method of claim 1 wherein a column within the circular buffer within the plurality of rotating circular buffers can be skipped in a cycle.
11. The method of claim 10 wherein not skipping indicates a valid instruction.
12. The method of claim 1 wherein the plurality of logical elements includes a processing element that can be placed in a sleep state where the sleep state is exited based on data being valid.
13. The method of claim 12 wherein the sleep state can be entered by an instruction within the processing element.
14. The method of claim 12 wherein the processing element wakes up from the sleep state when valid data is applied to inputs of the processing element.
15. The method of claim 14 wherein the sleep state can only be exited by stimulus external to the processing element and not based on programming of the processing element.
16. The method of claim 15 wherein the sleep state is exited based on an instruction applied to a switching fabric.
17. The method of claim 1 wherein the configurable connections provide three-dimensional routing.
18. The method of claim 1 further comprising storage elements coupled to the configurable connections wherein the storage elements include one or more of registers, caches, or on-chip memories.
19. The method of claim 18 wherein the storage elements store data while the configurable connections are busy with other data.
20. The method of claim 1 wherein the configurable connections enable bypassing of neighboring logical elements.
21. The method of claim 1 wherein communication through the configurable connections is based on data being valid.
22. The method of claim 1 where collisions are avoided by intermediate data being stored in registers for pipeline cycles before being sent to an output port for a cluster.
23. A computer program embodied in a non-transitory computer readable medium for implementation of a reconfigurable fabric, the computer program comprising code which configures one or more processors to perform operations of:
designing a reconfigurable fabric by:
implementing a plurality of logical elements;
connecting configurable connections between and among the plurality of logical elements; and
implementing a plurality of rotating circular buffers residing on a semiconductor chip where the plurality of rotating circular buffers contain configuration instructions which control different configurable connections between and among the plurality of logical elements and dynamically change operation of the logical elements based on the rotating of a circular buffer within the plurality of rotating circular buffers, wherein the plurality of rotating circular buffers have differing lengths and resynchronize with a zeroth pipeline stage for each of the plurality of rotating circular buffers.
24. A computer system for implementation of a reconfigurable fabric comprising:
a memory which stores instructions; and
one or more processors coupled to the memory, wherein the one or more processors are configured to:
design a reconfigurable fabric by:
implementing a plurality of logical elements;
connecting configurable connections between the plurality of logical elements; and
implementing a plurality of rotating circular buffers residing on a semiconductor chip where the plurality of rotating circular buffers contain configuration instructions which control different configurable connections between the plurality of logical elements and dynamically change operation of the logical elements based on the rotating of a circular buffer within the plurality of rotating circular buffers, wherein the plurality of rotating circular buffers have differing lengths and resynchronize with a zeroth pipeline stage for each of the plurality of rotating circular buffers.