SYSTEMS AND METHOD FOR MAPPING FIFOS TO PROCESSOR ADDRESS SPACE

• US 20180137075A1
• Filed: 11/06/2017
• Published: 05/17/2018
• Est. Priority Date: 11/17/2016
• Status: Active Grant

• Current Assignee: US Department of The Air Force (U.S. Department Of Defense)
• Original Assignee: Goverment of The United States As Represented By The Secretary of The Air Force
• Inventors: LINDERMAN, MARK H, WU, QING, FITZGERALD, DENNIS
• Granted Patent: US 10,521,390 B2
• Time in Patent Office: Days
• CPC Class Codes: F16C 33/741 by means of a fluid
• CPC Class Codes: F16J 15/32 with elastic sealings, e.g. O-rings
• CPC Class Codes: G06F 13/4022 using switching circuits, e.g. switching matrix, connection or expansion network G06F13/4009 takes precedence
• CPC Class Codes: G06F 13/4031 with arbitration
• CPC Class Codes: G06F 13/404 with address mapping
• CPC Class Codes: G06F 13/4081 Live connection to bus, e.g. hot-plugging current or voltage limitation during live insertion H02H9/004
• CPC Class Codes: Y02D 10/00 Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

An apparatus for a microprocessor computer system and method for configuring the same where said microprocessor computer system comprises a processor core and at least one hardware buffer FIFO with memory-mapped head and tail that handles data movement among the processor cores, networks, raw data input and outputs, and memory. The method for configuring said microprocessor computer system comprises utilizing a FIFO auxiliary processor to process said data traversing said hardware FIFO; utilizing said hardware FIFOs to efficiently pipe data through functional blocks; and utilizing a FIFO controller to perform DMA operations that include non-unit-stride access patterns and transfers among processor cores, networks, raw data input and outputs, memory, and other memory-mapped hardware FIFOs.

Claims

1. A microprocessor computer system comprising:

a processor core capable of at least one of

reading data from a separate storage means by presenting an address and retrieving data; and

writing data to a separate storage means by presenting an address and data to be stored by said storage means;

a storage means comprising at least one bank of at least one hardware first in first out (FIFO), said FIFO further comprising a head, a tail, and a buffer comprising any number of additional buffer locations to hold data enqueued in the FIFO;

wherein at least one of said FIFO head and FIFO tail is addressable by said processor core and mapped to at least one fixed memory address; and

busses and control signals to operably couple said processor core and said hardware FIFO, wherein said busses transfer addresses and data between said processor core and said hardware FIFO, and said control signals indicate failed attempts to write to said hardware FIFO when it is full and failed attempts to read from said hardware FIFO when it is empty.

2. The microprocessor computer system according to claim 1 further comprising a controllable FIFO,

wherein said controllable FIFI further comprises a FIFO controller, and wherein said FIFO controller further comprises at least one of a state machine and a programmable microcontroller that is configured to control operations on said hardware FIFO.

3. The microprocessor computer system according to claim 2 further comprising a FIFO bank.

4. The microprocessor computer system according to claim 1 further comprising memory hierarchy comprising at least one cache memory and at least one higher-level memory within a memory hierarchy above said cache memory wherein

said cache memory operably couples said processor core to said higher-level memory according to customary computer architecture practices,

wherein said control signals further indicate to said processor core that said processor core'"'"'s requested read operation has resulted in a cache miss such that the requested data is not immediately available and

wherein said control signals further indicate to said processor core that said processor core'"'"'s requested write operation cannot immediately be executed.

5. The microprocessor computer system according to claim 4 further comprising at least one arbiter that mediates contention for bus and signal resources shared by at least two of said processor core, said hardware FIFO, said cache memory, and said higher-level memory.

6. The microprocessor computer system according to claim 1 wherein said processor core includes lower-level caches through which said processor core accesses said at least one of said hardware FIFO.

7. The microprocessor computer system according to claim 1 wherein said read operation on said FIFO head comprises at least one of a blocking dequeueing read, a blocking non-dequeuing read, a non-blocking dequeueing read, and a non-blocking non-dequeuing read.

8. The microprocessor computer system according to claim 7 wherein said processor core indicates the type of said read operation to be performed by mapping a different address to said FIFO head for each type of said read operation.

9. The microprocessor computer system according to claim 1 wherein said write operation on said FIFO tail comprises at least one of a blocking enqueueing write, a non-blocking enqueueing write, and a non-blocking non-enqueueing write.

10. The microprocessor computer system according to claim 9 wherein said processor core indicates the type of said write operation to be performed by mapping a different address to said FIFO tail for each type of said write operation.

11. The microprocessor computer system according to claim 1, wherein attempts by said processor core to perform an enqueuing write on a full FIFO are treated like a cache miss that stalls an invoking processor core thread until capacity is available;

and

attempts by said processor core to perform a dequeuing read on an empty FIFO are treated like a cache miss that stalls an invoking processor core thread until data is available.

12. The microprocessor computer system according to claim 8 wherein said buffer write operation further comprises a block-oriented data transfer operation using said memory write operation.

13. The microprocessor computer system according to claim 1 wherein said buffer read operation further comprises a block-oriented data retrieval operation using the memory read operation.

14. The microprocessor computer system according to claim 1 wherein said hardware FIFO services data exchanges between at least two of:

said processor cores, at least one of incoming raw input/output (I/O), outgoing raw I/O, incoming network, and outgoing network.

15. The microprocessor computer system according to claim 2 wherein said FIFO controller further comprises an auxiliary processor configured by said processor core to control at least one of:

manipulation of data enqueued in said hardware FIFO;

transfers of data into said hardware FIFO from at least one of cache memory, other memory, external networks, raw data inputs, and other hardware FIFOs; and

transfers of data from said hardware FIFO to at least one of cache memory, other memory, external networks, raw data inputs, and other hardware FIFOs.

16. The microprocessor computer system according to claim 15 wherein said auxiliary processor is of a processor type selected from the group consisting of a field-programmable gate array (FPGA), a microprocessor, and a microcontroller.

17. The microprocessor computer system according to claim 15 wherein addresses for at least one of transfers to cache, memory and other FIFOS, and transfers from cache, memory, and other FIFOs use addresses computed by said FIFO controller.

18. The microprocessor computer system according to claim 1 wherein said hardware FIFO is of a computer memory type selected from the group consisting of:

a static random-access memory (SRAM), a flip-flop, a register file, and a latch.

19. The microprocessor computer system according to claim 15 further comprising at least one other hardware FIFOs

wherein said FIFO controller of said controllable FIFO is further configured to at least one of:

write data from said controllable FIFO'"'"'s head to at least one of said other FIFO'"'"'s tail, and

read data from at least one of said other FIFO'"'"'s head to said controllable FIFO'"'"'s tail.

20. The microprocessor system according to claim 1 wherein at least one of said memory-mapped FIFO head and tail is pinned at a logical buffer address managed by the cache memory.

21. The microprocessor computer system according to claim 1 further comprising:

at least one additional buffer that is located apart from said hardware FIFO and that uses an alternate memory to expand the capacity of said hardware FIFO, and busses and signals to transfer data between said hardware FIFO and said additional buffer.

22. The microprocessor computer system according to claim 7 wherein said FIFO controller adjusts the size of said additional buffer.

23. The microprocessor computer system according to claim 4 further comprising:

at least one additional processor core; and

a data communications network;

wherein said at least one additional processor core exchanges data over said data communications network with said processor core by means of said hardware FIFOs.

24. The microprocessor computer system according to claim 9 wherein

said at least one additional processor core comprises a fixed functionality computational block.

25. The microprocessor computer system according to claim 24, wherein said fixed functionality computational block consumes data from at least one memory-mapped hardware FIFO and produces data that is enqueued on at least one memory-mapped hardware FIFO.

35. The microprocessor computer system according to claim 1 wherein said FIFO buffer further comprises a functional block that processes data as it moves from said FIFO tail towards said FIFO head.

26. A method of configuring a microprocessor computer system, wherein said microprocessor computer system comprises

a processor core capable of at least one of reading data from a storage means by presenting a memory address and retrieving data and writing data to a storage means by presenting a memory address and data to be stored by said storage means;

a storage means comprising at least one bank of at least one hardware first in first out hardware FIFO, said hardware FIFO comprising a head, a tail, and a buffer comprising any number of additional buffer locations to hold data enqueued in said hardware FIFO, wherein at least one of said hardware FIFO head and hardware FIFO tail is addressable by said processor core and mapped to at least one fixed said memory address;

busses and control signals to operably couple said processor core and said hardware FIFO wherein said busses transfer addresses and data between said processor core and said hardware FIFO; and

control signals for indicating failed attempts to write to said hardware FIFO when it is full and failed attempts to read from said hardware FIFO when it is empty;

a FIFO controller comprising at least one of a state machine and a programmable microcontroller that is configured to control operations on said hardware FIFO and a set of configuration and status registers; and

a new data packet configuration comprising a header of any length and data of any length;

wherein said method comprises the steps of;

filling with new data packet at said tail and advancing towards said head of said hardware FIFO as prior data packets are dequeued;

detecting said new data packet header at said head of said hardware FIFO;

determining if said FIFO controller has already been configured to process said new data packet;

when said FIFO controller has not been configured to process said new data packet header,

sending from said FIFO controller to said processor core an interrupt; and

wherein said processor core;

dequeues and processes said new data packet'"'"'s header;

performs any operations that precede packet data processing; and

configures said FIFO controller for processing a remainder of said new data packet;

processing by said configured FIFO controller of said new data packet according to said configuration until, and in accordance with a FIFO controller configuration, a termination condition is satisfied;

when said FIFO controller is configured to interrupt said processor core at a completion of said processing;

interrupting said processor core; and

whereupon said processor core inspects said FIFO configuration and status registers and completes processing of said new data packet.

27. The method according to claim 26 wherein said processing of said new packet data comprises a direct memory access (DMA) operation that writes data to at least one of cache memory, other memory, and at least one other FIFO.

28. A method of configuring a microprocessor computer system that comprises

a processor core capable of at least one of

reading data from a storage means by presenting an address;

and

retrieving data and writing data to a storage means by presenting an address and data to be stored by said storage means;

said storage means comprising at least one bank of at least one hardware first in first out (FIFO) queue, denoted hardware FIFO, comprising a head, a tail, and a buffer comprising any number of additional buffer locations to hold data enqueued in the FIFO wherein at least one of said FIFO head and FIFO tail is addressable by said processor core and mapped to at least one fixed memory address;

busses and control signals to operably couple said processor core and said hardware FIFO wherein said busses transfer addresses and data between said processor core and said hardware FIFO, and said control signals indicate failed attempts to write to said hardware FIFO when it is full and failed attempts to read from said hardware FIFO when it is empty;

a FIFO controller comprising at least one of a state machine and a programmable microcontroller that is configured to control operations on said hardware FIFO and a set of configuration and status registers;

wherein said method comprising the steps of;

configuring by said processor core, said FIFO controller to retrieve and process data from at least one of cache memory, other memory, as at least one of memory-mapped FIFO;

writing by said processor core, any required header data to said FIFO prior initiating FIFO controller processing;

initiating by said processor core, FIFO controller processing using said configuration, the following steps until FIFO controller satisfies termination condition specified in said configuration;

computing an address of a next data item according to said configuration;

retrieving said next data item from said address and

processing said next data item according to said configuration; and

when said FIFO controller is configured to interrupt said processor core at the completion of said processing,

interrupting said processor core, whereupon said processor core inspects said FIFO configuration and status registers and completes processing.

29. The method according to claim 28 wherein said steps of computing address, retrieving next data item, and processing next data item collectively perform a direct memory access (DMA) operation that reads data from at least one of cache memory, other memory, and at least one of another FIFO.

30. The method according to claim 29 wherein said DMA operation collects data as a pre-fetch for subsequent processing.

31. A method of configuring a microprocessor computer system, wherein said microprocessor computer system comprises

a processor core capable of at least one

of reading data from a storage means by presenting an address and retrieving data;

and

writing data to a storage means by presenting an address and data to be stored by said storage means;

a storage means comprising at least one bank of at least one hardware first in first out (FIFO) queue comprising a head, a tail, and a buffer comprising any number of additional buffer locations to hold data enqueued in said hardware FIFO wherein at least one of said hardware FIFO head and said hardware FIFO tail is addressable by said processor core and mapped to at least one fixed memory address;

busses and control signals to operably couple said processor core and said hardware FIFO wherein said busses transfer addresses and data between said processor core and said hardware FIFO, and said control signals indicate failed attempts to write to said hardware FIFO when it is full and failed attempts to read from said hardware FIFO when it is empty; and

when a bit width of said hardware FIFO tail exceeds a bit width of said processor core'"'"'s ability to write to said hardware FIFO tail in a single write operation, said method comprises the steps of

performing a sequence of non-enqueuing write operations until an occurrence of a last word to be written; and

performing an enqueuing write for the last write operation that commits said word to said hardware FIFO.

32. A method of configuring a microprocessor computer system, wherein said microprocessor computer system comprises:

a processor core capable of at least one of

reading data from a storage means by presenting an address; and

retrieving data and writing data to a storage means by presenting an address and data to be stored by said storage means,

said storage means comprising at least one bank of at least one hardware first in first out (FIFO) queue, denoted hardware FIFO, comprising a head, a tail, and a buffer comprising any number of additional buffer locations to hold data enqueued in the FIFO wherein at least one of said FIFO head and FIFO tail is addressable by the processor core and mapped to at least one fixed memory address,

busses and control signals to operably couple said processor core and said hardware FIFO wherein said busses transfer addresses and data between said processor core and said hardware FIFO, and said control signals indicate failed attempts to write to said hardware FIFO when it is full and failed attempts to read from said hardware FIFO when it is empty,

wherein when bit width of said FIFO head exceeds bit width of said processor core'"'"'s ability to read from said FIFO head in a single read operation, said method comprises

performing a sequence of non-dequeuing read operations until an occurrence of a last word to be read; and

performing a dequeuing read for said last read operation that removes said word from said hardware FIFO.

33. A method for configuring a microprocessor computer system, wherein said microprocessor computer system comprises a processor core executing vector operations and at least one memory-mapped FIFO;

said method comprising the steps of;

filling said at least one memory-mapped FIFO with vector data as needed; and

performing said vector operation repeatedly to process each element of said vector data by;

dequeuing a value from said at least one memory-mapped FIFO; and

processing said value according to an operation to be performed.

34. A method of configuring a microprocessor computer system, wherein said microprocessor computer system comprises a processor core executing vector operations and at least one memory-mapped FIFO;

said method comprising the steps of;

performing said vector operation repeatedly to generate each value of an vector result comprising the steps;

executing an iteration of said vector operation resulting in an element of an output vector,

enqueuing said value onto said at least one memory-mapped FIFO.