Device and method for accelerating graphics processor units, and computer readable storage medium
A method for accelerating graphics processing units (GPUs) receives a request for usage of GPU resource sent by a user, calculates a quantity of GPUs which are necessary, and arrange the GPUs in several ways to maximize data transmission from and between the GPUs, and between the GPUs and one or more central processing units (CPUs) connected by switches between the GPUs and the CPUs. A device for accelerating GPUs is also provided.
1. A method for accelerating graphics processor units (GPUs), a plurality of GPUs exchanging data with central processing units (CPUs) through switches, a number of the GPUs being greater than or equal to a number of switches being greater than or equal to a number of the CPUs, the method comprising:
receiving a request for usage of GPU resource sent by a user;
calculating a quantity of the GPUs necessary for the usage;
arranging the GPUs according to the usage to maximize data transmission of the GPUs; and
processing the request for GPU resource by the arranged GPUs;
wherein when the usage of the GPUs is less than or equal to a first threshold, a plurality of GPUs connected to different switches is selected, and the switches perform data interaction with one of the CPUs;
when the usage of the GPUs is greater than the first threshold and less than or equal to a second threshold, the GPUs are distributed into groups, each group of GPUs communicates with one of the switches to form a joint body; and
when the usage of the GPUs is greater than or equal to the second threshold, the GPUs are arranged to form a ring index according to a preset index rule.
2. The method for accelerating GPUs of claim 1 , wherein
when the usage of the GPUs is greater than the first threshold and less than or equal to a second threshold, the joint bodies perform data interaction with at least two CPUs.
3. The method for accelerating GPUs of claim 2 , wherein each group of GPUs comprises one or more GPUs.
4. The method for accelerating GPUs of claim 2 , wherein p 1 when the usage of the GPUs is greater than or equal to the second threshold, the GPUs in the ring index perform data interaction with the CPU through at least one switch.
5. The method for accelerating GPUs of claim 4 , wherein the preset index rule is to use an NVlink connection to change index numbers of the GPUs.
6. The method for accelerating GPUs of claim 4 , wherein the first threshold is two, and the second threshold is eight.
7. A device for accelerating graphics processor units (GPUs), a plurality of GPUs exchanging data with central processing units (CPUs) through switches, a number of the GPUs being greater than or equal to a number of switches being greater than or equal to a number of the CPUs, the device comprising:
a communication unit configured for to establish a communication connection between the GPUs and the switches, and the switches and the CPUs;
a processor; and
a storage device storing one or more programs, when executed by the processor, the one or more programs causing the processor to:
receive a request for usage of GPU resource sent by a user;
calculate a quantity of the GPUs necessary for the usage;
arrange the GPUs according to the usage to maximize data transmission of the GPUs; and
process the request for GPU resource by the arranged GPUs;
wherein when the usage of the GPUs is less than or equal to a first threshold, a plurality of GPUs connected to different switches are selected, and the switches perform data interaction with one of the CPUs;
when the usage of the GPUs is greater than the first threshold and less than or equal to a second threshold, the GPUs are distributed into groups, each group of GPUs communicates with one of the switches to form a joint body; and
when the usage of the GPUs is greater than or equal to the second threshold, the GPUs are arranged to form a ring index according to a preset index rule.
8. The device for accelerating GPUs of claim 7 , wherein
when the usage of the GPUs is greater than the first threshold and less than or equal to a second threshold, the joint bodies perform data interaction with at least two CPUs.
9. The device for accelerating GPUs of claim 8 , wherein each group of GPUs comprises one or more GPUs.
10. The device for accelerating GPUs of claim 8
wherein when the usage of the GPUs is greater than or equal to the second threshold, the GPUs in the ring index perform data interaction with the CPU through at least one switch.
11. The device for accelerating GPUs of claim 10 , wherein the preset index rule is to use an NVlink connection to change index numbers of the GPUs.
12. The device for accelerating GPUs of claim 8 , wherein the first threshold is two, and the second threshold is eight.
13. A non-transitory computer readable storage medium having stored thereon instructions that, when executed by at least one processor of a computing device, causes the processor to perform a method for accelerating graphic processor units (GPUs), a plurality of GPUs exchanging data with central processing units (CPUs) through switches, a number of the GPUs being greater than or equal to a number of switches being greater than or equal to a number of the CPUs, wherein the method comprises:
receiving a request for usage of GPU resource sent by a user;
calculating a quantity of the GPUs necessary for the usage;
arranging the GPUs according to the usage to maximize data transmission of the GPUs; and
processing the request for GPU resource by the arranged GPUs;
wherein when the usage of the GPUs is less than or equal to a first threshold, a plurality of GPUs connected to different switches are selected, and the switches perform data interaction with one of the CPUs;
when the usage of the GPUs is greater than the first threshold and less than or equal to a second threshold, the GPUs are distributed into groups, each group of GPUs communicates with one of the switches to form a joint body; and
when the usage of the GPUs is greater than or equal to the second threshold, the GPUs are arranged to form a ring index according to a preset index rule.
14. The non-transitory computer readable storage medium of claim 13 , wherein
when the usage quantity of the GPUs is greater than the first threshold and less than or equal to a second threshold, the joint bodies perform data interaction with at least two CPUs.
15. The non-transitory computer readable storage medium of claim 13 , wherein
when the usage of the GPUs is greater than or equal to the second threshold, the GPUs in the ring index perform data interaction with the CPU through at least one switch.