Receive buffer architecture method and apparatus

• US 10,397,144 B2
• Filed: 12/22/2016
• Issued: 08/27/2019
• Est. Priority Date: 12/22/2016
• Status: Active Grant

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Dcruz, Lloyd, Bathia, Abhay H.
• Primary Examiner(s): Coby, Frantz
• Application Number:
  US15/389,172
• Publication Number:
  US 20180183733A1
• Time in Patent Office: 978 Days
• Field of Search: 709220, 709223, 709234, 709236, 709237
• CPC Class Codes: G06F 3/0679 Non-volatile semiconductor memory device, e.g. flash memory, one time programmable memory [OTP]
• CPC Class Codes: H04L 49/9005 using dynamic buffer space allocation
• CPC Class Codes: H04W 24/10 Scheduling measurement reports ; Arrangements for measurement reports
• CPC Class Codes: H04W 84/12 WLAN [Wireless Local Area Networks]

Abstract

Apparatus and method to facilitate delivery of data packets to compute nodes are disclosed herein. A receive buffer may store a first data packet associated with a first context, among a plurality of contexts, and a second data packet associated with a second context, among the plurality of contexts, wherein the first and second contexts identify respective first and second processes to be performed by a compute node in association with the respective first and second data packets, and wherein the receive buffer includes logic that is to determine that the compute node is unavailable to perform the first process with respect to the first data packet, determine that the compute node is available to perform the second process with respect to the second data packet, and read the second data packet for delivery to the compute node while the first data packet is retained.

Claims

1. An apparatus comprising:

a communication chip comprising a receive buffer that stores a first data packet associated with a first context, among a plurality of contexts, and a second data packet associated with a second context, among the plurality of contexts, wherein the first and second contexts identify respective first and second processes to be performed by a compute node in association with the respective first and second data packets, and the first and second data packets are stored in the receive buffer to be delivered to the compute node for operation of the first and second processes,

wherein the receive buffer includes logic that is to determine that the compute node is unavailable to perform the first process with respect to the first data packet, determine that the compute node is available to perform the second process with respect to the second data packet, and read the second data packet for delivery to the compute node while the first data packet is retained.

2. The apparatus of claim 1, wherein the receive buffer is to receive and store the first data packet prior to receipt and storage of the second data packet.

3. The apparatus of claim 1, wherein the receive buffer comprises a first-in-first-out (FIFO) buffer, a ring buffer, or a ring queue buffer.

4. The apparatus of claim 1, wherein the logic included in the receive buffer comprises hardware, firmware, an application specific integrated circuit (ASIC), programmable array logic (PAL), field programmable gate array (FPGA), circuitry, on-chip circuitry, or circuitry.

5. The apparatus of claim 1, wherein the receive buffer includes a receive buffer data memory organized as a plurality of buffer credits, a buffer credit of the plurality of buffer credits having a pre-determined memory size and a unique buffer address, and each of the first and second data packet is stored in select one or more buffer credits of the plurality of buffer credits.

6. The apparatus of claim 5, wherein the receive buffer includes context specific receive queue (CSRQ) registers, the CSRQ registers including a first buffer address in which a data packet associated with a particular context, among the plurality of contexts, is first stored in the receive buffer data memory, a last buffer address in which a data packet associated with the particular context is last stored in the receive buffer data memory, a count of a number of data packets associated with the particular context stored in the receive buffer data memory, an end of packet (EOP) flag indicative of the data packet associated with the particular context that is first stored in the receive buffer data memory being a single buffer credit size, and a host buffer location flag indicative of the compute node being available to perform a process associated with the particular context with respect to the data packet associated with the particular context that is first stored in the receive buffer data memory, and wherein the CSRQ registers include information about each context of the plurality of contexts.

7. The apparatus of claim 5, wherein the receive buffer includes a packet buffer link list (PBLL) memory, the PBLL memory including a next buffer link list that identifies successive buffer addresses of the receive buffer data memory in which data packets associated with a particular context of the plurality of contexts are stored and a next buffer end of packet (EOP) list that identifies whether a respective buffer address is a second to last buffer address in which a data packet of the data packets associated with the particular context is stored.

8. The apparatus of claim 7, wherein the PBLL memory includes a read port and a write port, the read and write ports accessed by the logic to deliver the second data packet to the compute node.

9. The apparatus of claim 1, wherein the receive buffer comprises a free buffer list first-in-first-out (FBLF) memory and wherein the FBLF memory includes a read port and a write port, the read and write ports accessed by the logic to deliver the second data packet to the compute node.

10. A computerized method comprising:

determining, by a receive buffer within a communication chip, whether a host buffer location is available to accept delivery of a first data packet associated with a first context of a plurality of contexts, wherein the first data packet is buffered in the receive buffer and the first context identifies a first process to be performed by a host compute node including the host buffer location;

when the host buffer location is determined to be unavailable for the first data packet, determining, by the receiver buffer, whether a host buffer location is available to accept delivery of a second data packet associated with a second context of the plurality of contexts, wherein the second data packet is buffered in the receive buffer and the second context identifies a second process to be performed by the host compute node including the host buffer location; and

when the host buffer location is determined to be available for the second data packet, retaining the first data packet in the receive buffer and reading out at least a portion of the second data packet from the receive buffer.

11. The method of claim 10, further comprising arbitrating between head data packets associated with respective contexts of the plurality of contexts to obtain a grant for a next transmission of a data packet to the host compute node to be the at least a portion of the second data packet, wherein the head data packets associated with respective contexts of the plurality of contexts comprise data packets first buffered in the receive buffer for respective contexts and the second data packet comprises a head data packet for the second context.

12. The method of claim 11, further comprising updating one or more fields included in the receive buffer in accordance with the reading out of the at least a portion of the second data packet, wherein the one or more fields includes a first field that identifies a receive buffer location address in which a first portion of a head data packet associated with the second context is stored, a second field that identifies a count of data packets stored in the receive buffer associated with the second context, a third field that identifies a current availability or unavailability of the host buffer location in connection with the second context, a fourth field that identifies whether the head data packet associated with the second context comprises a data packet that is entirely stored in a single receive buffer location address that is of a pre-defined size, a fifth field that identifies a first available receive buffer location address to store a future data packet, and a sixth field that identifies a second to last receive buffer location address for respective data packets stored in the receive buffer.

13. The method of claim 10, further comprising sequentially reading out remaining portions of the second data packet from the receive buffer until an end of a second data packet indicator included in the receive buffer matches a receive buffer location address from which the second data packet is read out.

14. The method of claim 10, wherein receiving and buffering the first data packet comprises receiving the first data packet from a second compute node via a network through which the host compute node and the second compute node are communicatively coupled.

15. A system comprising:

a compute node in a network; and

a compute node interface in communication with the compute node, the compute node interface including a receive buffer in a communication chip that stores a first data packet associated with a first context of a plurality of contexts and a second data packet associated with a second context of the plurality of contexts, wherein the first and second contexts identify respective first and second processes to be performed by the compute node in association with the respective first and second data packets, and the first and second data packets are stored in the receive buffer to be delivered to the compute node for operation of the first and second processes, and wherein, when the compute node is unavailable to perform the first process with respect to the first data packet, the receive buffer is to determine that the compute node is available to perform the second process with respect to the second data packet and transmit the second data packet to the compute node prior to transmission of the first data packet.

16. The system of claim 15, wherein the compute node interface includes a direct memory access (DMA) engine in communication with the receive buffer, and wherein the second data packet is transmitted from the receive buffer to the DMA engine for transmission to the compute node.

17. The system of claim 15, further comprising:

a second compute node in the network; and

a second compute node interface in communication with the second compute node, wherein the first data packet is transmitted from the second compute node to the second compute node interface to the compute node interface, via the network through which the compute node interface and the second compute node interface are communicatively coupled.

18. The system of claim 15, wherein the receive buffer comprises hardware, firmware, an application specific integrated circuit (ASIC), programmable array logic (PAL), field programmable gate array (FPGA), circuitry, on-chip circuitry, or circuitry.

19. The system of claim 15, wherein the receive buffer includes a receive buffer data memory organized as a plurality of buffer credits, a buffer credit of the plurality of buffer credits having a pre-determined memory size and a unique buffer address, and each of the first and second data packet is stored in select one or more buffer credits of the plurality of buffer credits.

20. The system of claim 19, wherein the receive buffer includes a free buffer list first-in-first-out (FIFO) (FBLF) memory, the FBLF memory including a free buffer FIFO list that identifies buffer addresses available to store incoming data packets, a head free buffer pointer indicative of a first available buffer address among the buffer addresses available that is available to store the incoming data packets, and a tail free buffer pointer indicative of a last available buffer address among the buffer addresses available that is available to store the incoming data packets.

21. The system of claim 15, wherein the compute node interface includes hardware-based logic, context specific receive queue (CSRQ) registers, a packet buffer link list (PBLL) memory, and a free buffer list first-in-first-out (FIFO) (FBLF) memory, and wherein the hardware-based logic is to control creation and maintenance of characteristics and storage information associated with the first and second data packets in the receive buffer.

22. An apparatus comprising:

means for determining a first host buffer location in a host compute node to which a first data packet associated with a first context of a plurality of contexts is to be delivered, wherein the first data packet is buffered in a receive buffer in a communications chip and the first context identifies a first process to be performed by the host compute node;

when the first host buffer location is determined to be unavailable for the first data packet, means for determining whether a second host buffer location in the host compute node is available to accept delivery of a second data packet associated with a second context of the plurality of contexts, wherein the second data packet is buffered in the receive buffer and the second context identifies a second process to be performed by the host compute node; and

when the second host buffer location is determined to be available for the second data packet, means for retaining the first data packet in the receive buffer and means for delivering at least a portion of the second data packet from the receive buffer.

23. The apparatus of claim 22, further comprising means for updating one or more fields included in the receive buffer in accordance with delivery of the at least a portion of the second data packet, wherein the one or more fields includes a first field that identifies a receive buffer location address in which a first portion of a head data packet associated with the second context is stored, a second field that identifies a count of data packets stored in the receive buffer associated with the second context, a third field that identifies a current availability or unavailability of the host buffer location in connection with the second context, a fourth field that identifies whether the head data packet associated with the second context comprises a data packet that is entirely stored in a single receive buffer location address that is of a pre-defined size, a fifth field that identifies a first available receive buffer location address to store a future data packet, and a sixth field that identifies a second to last receive buffer location address for respective data packets stored in the receive buffer.

24. The apparatus of claim 22, further comprising means for determining, at a later point in time after delivery of the second data packet to the host compute node, whether the first host buffer location is available to receive the first data packet, and upon an affirmative determination, means for delivering the first data packet to the host compute node.

25. The apparatus of claim 22, wherein the means for determining the first host buffer location to which the first data packet is to be delivered comprises means for performing a lookup of a host buffer location address to which the host compute node can accept the first data packet to perform the first process.