Data storage device and method for providing an adaptive data path
A data storage device and method for providing an adaptive data path are disclosed. In one embodiment, a data storage device is in communication with a host comprising a first processor (e.g., a graphics processing unit (GPU)), a second processor (e.g., a central processing unit (CPU)), and a queue. The data storage device chooses a data path to use to communicate with the queue based on whether the queue is associated with the first processor or with the second processor. Other embodiments are possible, and each of the embodiments can be used alone or together in combination.
1. A data storage device comprising:
an interface configured to communicate with a host comprising a central processing unit (CPU), a graphics processing unit (GPU), and a plurality of submission queues;
a memory; and
one or more processors configured to communicate with the interface and the memory and further configured to:
classify each submission queue of the plurality of submission queues as either a CPU queue or a GPU queue;
determine a data path for a transmission to one submission queue of the plurality of submission queues based on whether the one submission queue is classified as a CPU queue or as a GPU queue; and
determine a maximum packet size and/or a maximum read request size based on whether the one submission queue is classified as a CPU queue or as a GPU queue.
2. The data storage device of claim 1 , wherein each submission queue is classified by its address range.
3. The data storage device of claim 1 , wherein the host further comprises at least one additional GPU, each GPU with its own data path to the data storage device.
4. The data storage device of claim 1 , wherein there are more switches in a data path from the data storage device to the GPU than there are switches in a data path from the data storage device to the CPU.
5. The data storage device of claim 1 , wherein:
a first data path is used in response to the one submission queue being classified as a CPU queue;
a second data path is used in response to the one submission queue being classified as a GPU queue; and
the first and the second data paths have different latency and/or throughput attributes.
6. The data storage device of claim 1 , wherein the transmission relates to command fetching, physical region pages (PRP) fetching, scatter gather lists (SGL) fetching, data transfer, completion posting, and/or interrupt posting.
7. The data storage device of claim 1 , wherein the one or more processors is further configured to determine a maximum number of outstanding transactions to issue based on whether the one submission queue is classified as a CPU queue or as a GPU queue.
8. The data storage device of claim 1 , wherein the one or more processors is further configured to associate a request from a direct memory access (DMA) engine with the one submission queue.
9. The data storage device of claim 1 , wherein the memory comprises a three-dimensional memory.
10. The data storage device of claim 1 , wherein each submission queue is classified by its creation time.
11. A method comprising:
performing the following in a data storage device in communication with a host comprising a first processor, a second processor, and a queue:
determine whether the queue is associated with the first processor or with the second processor;
in response to determining that the queue is associated with the first processor, select a first data path to the queue;
in response to determining that the queue is associated with the second processor, select a second data path to the queue, wherein the first and the second data paths have different latency and/or throughput attributes; and
determine a maximum packet size, a maximum read request size, and/or a maximum number of outstanding transactions to issue based on whether the queue is associated with the first processor or with the second processor.
12. The method of claim 11 , wherein the first processor comprises a graphics processing unit (GPU) and the second processor comprises a central processing unit (CPU).
13. The method of claim 11 , wherein the determination whether the queue is associated with the first processor or with the second processor is made based on an address range of the queue.
14. The method of claim 11 , wherein the determination whether the queue is associated with the first processor or with the second processor is made based on a creation time of the queue.
15. The method of claim 11 , further comprising sending a transmission on the selected data path, wherein the transmission relates to command fetching, physical region pages (PRP) fetching, and/or scatter gather lists (SGL) fetching.
16. The method of claim 11 , further comprising associating a request from a direct memory access (DMA) engine with the queue.
17. The method of claim 11 , wherein there are more switches in the first data path than there are switches in the second data path.
18. The method of claim 11 , further comprising sending a transmission on the selected data path, wherein the transmission relates to data transfer.
19. The method of claim 11 , further comprising sending a transmission on the selected data path, wherein the transmission relates to completion posting and/or interrupt posting.
20. A data storage device comprising:
an interface configured to communicate with a host comprising a central processing unit (CPU), a graphics processing unit (GPU), and a plurality of submission queues;
a memory; and
means for:
classifying each submission queue of the plurality of submission queues as either a CPU queue or a GPU queue;
determining a data path for a transmission to one submission queue of the plurality of submission queues based on whether the one submission queue is classified as a CPU queue or as a GPU queue; and
determining a maximum number of outstanding transactions to issue based on whether the one submission queue is classified as a CPU queue or as a GPU queue.