Direct access to low-latency memory
View Patent ↗A content aware application processing system is provided for allowing directed access to data stored in a non-cache memory thereby bypassing cache coherent memory. The processor includes a system interface to cache coherent memory and a low latency memory interface to a non-cache coherent memory. The system interface directs memory access for ordinary load/store instructions executed by the processor to the cache coherent memory. The low latency memory interface directs memory access for non-ordinary load/store instructions executed by the processor to the non-cache memory, thereby bypassing the cache coherent memory. The non-ordinary load/store instruction can be a coprocessor instruction. The memory can be a low-latency type memory. The processor can include a plurality of processor cores.
1. A processor, comprising:
at least one processor core;
a system interface to cache coherent memory which directs memory access for ordinary load/store instructions executed by the at least one processor core to the cache coherent memory; and
a low latency memory interface to a non-cache memory which directs memory access for non-ordinary load/store instructions executed by the at least one processor core to the non-cache memory, the low latency memory interface directly coupled to the at least one processor core over a dedicated bus, thereby bypassing the cache coherent memory,
wherein the non-ordinary load/store instructions include an instruction type field, the instruction type field is related to moving data between the non-cache memory and a main register file, the instruction type field having a first instruction type for moving data between the non-cache memory and at least one holding register and a second instruction type for moving the data between the at least one holding register and the main register file;
wherein the non-cache memory is accessed by the low latency memory interface via a two instruction sequence, the two instruction sequence comprising the first and second instruction types.
2. The processor of claim 1 , wherein the non-ordinary load/store instruction is a coprocessor instruction.
3. The processor of claim 1 , wherein the at least one holding register is located within an execution unit in the at least one processor.
4. The processor of claim 3 , wherein the at least one holding register is separate from the main register file located within the execution unit in the at least processor core.
5. The processor of claim 1 , wherein the data is stored in a deterministic finite automata (DFA) graph in the non-cache memory for performing content aware application processing.
6. The processor of claim 1 , wherein the non-cache memory is a low-latency memory.
7. The processor of claim 6 , wherein the low latency memory is selected from a group consisting of dynamic random access memory (DRAM), reduced latency dynamic random access memory (RLDRAM), static random access memory (SRAM), and fast cycle random access memory (FCRAM).
8. The processor of claim 7 , wherein the RLDRAM has a latency of less than or equal to 20 nanoseconds.
9. A method comprising:
directing memory access for ordinary load/store instructions executed by a processor core to cache coherent memory via a system interface to the cache coherent memory; and
directing memory access for non-ordinary load/store instructions executed by the processor core to non-cache memory via a low latency memory interface to the non-cache cache coherent memory, the low latency memory interface directly coupled to the processor core over a dedicated bus, thereby bypassing the cache coherent memory;
wherein the non-ordinary load/store instructions include an instruction type field, the instruction type field is related to moving data between the non-cache memory and a main register file, the instruction type field having a first instruction type for moving data between the non-cache memory and at least one holding register and a second instruction type for moving the data between the at least one holding register and the main register file;
wherein the non-cache memory is accessed by the low latency memory interface via a two instruction sequence, the two instruction sequence comprising the first and second instruction types.
10. The method of claim 9 , wherein the non-ordinary load/store instruction is a coprocessor instruction.
11. The method of claim 9 , wherein the at least one holding register is located within an execution unit within the processor core.
12. The method of claim 11 , wherein the at least one holding register located within the processor core is separate from the main register file located within the execution unit in the processor core.
13. The method of claim 9 , wherein the data is stored in a deterministic finite automata (DFA) graph in the non-cache memory for performing content aware application processing.
14. The method of claim 9 , wherein the non-cache memory is a low-latency memory.
15. The method of claim 14 , wherein the low-latency memory is selected from a group consisting of dynamic random access memory (DRAM), reduced latency dynamic random access memory (RLDRAM), static random access memory (SRAM), and fast cycle random access memory (FCRAM).
16. The method of claim 15 , wherein the processor core accesses the RLDRAM with less than or equal to 20 nanosecond latency.
17. A system for increasing processor speed for content aware application processing, comprising:
a processor, the processor including at least one processor core and an interface to a non-cache memory; and
non-ordinary load/store instructions included in a general purpose instruction set for moving data between the at least one processor core and the non-cache memory, the non-cache memory interface directly coupled to the at least one processor core over a dedicated bus, thereby bypassing cache coherent memory;
wherein the non-ordinary load/store instructions include an instruction type field, the instruction type field is related to moving data between the non-cache memory and a main register file, the instruction type field having a first instruction type for moving data between the non-cache memory and at least one holding register and a second instruction type for moving the data between the at least one holding register and the main register file;
wherein the non-cache memory is accessed by a low latency memory interface via a two instruction sequence, the two instruction sequence comprising the first and second instruction types.