Superconducting system architecture for high-performance energy-efficient cryogenic computing
An energy efficient rapid single flux quantum (ERSFQ) logic register wheel includes a circular shift register having a plurality of destructive read out (DRO) cells. Each entry of the circular shift register includes a data block, a tag, and a valid bit. A compare and control logic is coupled to the circular shift register to compare a source specifier or a destination register specifier against a register tag stored in the wheel following each cycle of the register wheel. At least one or more read ports and at least one or more write ports are coupled to the circular shift register to write to or to read from a different entry each in the register wheel following each cycle of the register wheel. A RSFQ clearable FIFO with flushing and a crosspoint memory topology for integrating MRAM devices with ERSFQ circuits are also described.
1. An energy efficient rapid single flux quantum (ERSFQ) logic register wheel comprising:
a circular shift register having a plurality of destructive read out (DRO) cells, each entry of said circular shift register comprising a data block, a tag, and a valid bit which advance by one entry every cycle of the register wheel;
a compare and control logic coupled to said circular shift register to compare a source specifier or a destination register specifier against a register tag stored in the wheel following each cycle of the register wheel;
at least one or more read ports and at least one or more write ports, coupled to said circular shift register to write to or to read from a different entry each in the register wheel following each cycle of the register wheel; and
wherein said logic wheel comprises a Josephson junction (JJ) based energy ERSFQ logic disposed in a cryostat which maintains cryogenic temperatures.
2. The ERSFQ logic register wheel of claim 1 , wherein said ERSFQ logic wheel progresses through said circular register at a frequency of more than twice a core clock frequency of an associated processor to reduce average access latency.
3. The ERSFQ logic register wheel of claim 1 , wherein said ERSFQ logic register wheel further comprises at least one or more additional read port comprising a single JJ junction.
4. The ERSFQ logic register wheel of claim 1 , wherein said ERSFQ logic register wheel further comprises at least one or more additional write port comprising three JJ junctions.
5. The ERSFQ logic register wheel of claim 1 , wherein said ERSFQ logic register wheel comprises a random access memory (RAM) or a content addressed memory (CAM).
6. The ERSFQ logic register wheel of claim 1 , wherein said ERSFQ logic register wheel comprises a ERSFQ logic store wheel wherein each entry of said circular shift register further comprises an additional bit flag and said ERSFQ logic register wheel further comprises at least one or more search ports.
7. The ERSFQ logic store wheel of claim 6 , wherein said ERSFQ logic store wheel comprises an ERSFQ logic miss store holding register (MSHR) wheel.
8. The ERSFQ logic MSHR wheel of claim 7 , wherein said MSHR wheel comprises a CAM wheel with an address field for comparison and two fields for secondary misses.
9. The ERSFQ logic register wheel of claim 1 , wherein at least one or more of said ERSFQ logic register wheels provide a register file access component of a processor which is coupled to a decode component by asynchronous FIFOs and where said register file access component provides instructions to an execution component via asynchronous FIFOs such that each of said decode component, said register file access component, and said execution component can run at different clock rates.
10. The ERSFQ logic register wheel of claim 1 , further comprising a multiported store queue as a write buffer.
11. The ERSFQ logic register wheel of claim 1 , wherein at least one or more ERSFQ logic register wheels provide a load queue to support recovery from branch mispredictions.
12. The ERSFQ logic register wheel of claim 11 , wherein a first load wheel holds load instructions that have not yet been issued to a memory or searched by a store wheel and a second wheel holds loads after they are issued to a memory subsystem until they are committed.