Multi-GPU communication method and device, storage medium and electronic device
A sub-task and image processing technology, applied in the field of multi-GPU communication, can solve the problem of low overall GPU performance and achieve the effect of improving the overall performance of GPU
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0025] The method embodiment provided in Embodiment 1 of the present application may be executed in a multi-GPU system or a system with a similar architecture. Take running on a multi-GPU system as an example, figure 1 It is a schematic structural diagram of the GPU system implemented by the multi-GPU communication method of the embodiment of the present invention. like figure 1 Shown is a GPU system architecture with one master GPU and four slave GPUs. There is one-to-one communication between the master GPU and the slave GPU through a high-speed interconnection bus. The main GPU receives the drawing task sent by the PCIE bus of the upper layer, and transmits it to the other three slave GPUs through the high-speed interconnection bus. The data needed in the drawing process is shared through shared RAM and shared Cache.
[0026] In this embodiment, a multi-GPU communication method is provided, which is used to communicate between a master GPU and multiple slave GPUs, and a...
Embodiment 2
[0059] In this embodiment, a multi-GPU communication method is provided, which is used to communicate between a master GPU and multiple slave GPUs, and a one-to-one communication between the master GPU and each of the slave GPUs is performed through a high-speed interconnection bus. communication, such as Figure 4 As shown, the process includes the following steps:
[0060] Step S401, receiving multiple image processing subtasks obtained by dividing the image processing task by the main GPU;
[0061] Step S402, processing the plurality of image processing subtasks according to a preset method;
[0062] Step S403, sending the processing result of the image processing task to the main GPU.
[0063] The above-mentioned multi-GPU communication method, by receiving the drawing tasks sent by the upper system on the main GPU, completes the automatic drawing task cutting, distributes part of the drawing tasks to multiple slave GPUs, and manages the load of multiple slave GPUs at th...
Embodiment 3
[0078] In this embodiment, a device for multi-GPU communication is also provided, and the device is used to implement the above embodiments and preferred implementation manners, and what has been described will not be repeated. As used below, the term "module" may be a combination of software and / or hardware that realizes a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementations in hardware, or a combination of software and hardware are also possible and contemplated.
[0079] image 3 It is a structural block diagram of a device for multi-GPU communication according to an embodiment of the present invention, which is used to communicate between a master GPU and multiple slave GPUs, and the master GPU and each of the slave GPUs are connected through a high-speed interconnection bus One-to-one communication, such as image 3 As shown, the device includes:
[0080] The first receiving module 3...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More