A low-energy task scheduling strategy for cpu-gpu heterogeneity
A technology for task scheduling and low energy consumption, applied in energy-saving computing, multi-programming devices, instruments, etc., can solve the problem that the distinction between heterogeneous cores is not detailed enough, the optimization of task scheduling length is not considered, and soft real-time requirements are not considered. Problems such as lack of information in the algorithm initialization phase
A technology for task scheduling and low energy consumption, applied in energy-saving computing, multi-programming devices, instruments, etc., can solve the problem that the distinction between heterogeneous cores is not detailed enough, the optimization of task scheduling length is not considered, and soft real-time requirements are not considered. Problems such as lack of information in the algorithm initialization phase
CN109960576BActive Publication Date: 2021-04-16BEIJING UNIV OF TECH
Image
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0075] In order to make the purpose, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings.
[0076] The present invention is based on the task scheduling strategy under the CPU-GPU master-slave architecture environment, and the schematic diagram of the heterogeneous architecture is as follows figure 1 As shown, here we take the illustration as an example. There is one CPU and three GPUs in the processor, and data is exchanged between different processing cores through the PCI main line. The processed tasks are as follows: figure 2 As shown in the DAG of , it can be divided into 14 subtasks with dependencies and large differences in communication volume. Before starting task scheduling, the entire system needs to set its own processor performance-related parameters (step 1), and obtain task-related information when accepting tasks (step 2)...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
A low-energy task scheduling strategy for CPU-GPU heterogeneity. Aiming at the characteristics of heterogeneous multi-core systems and the problems of traditional ant colony algorithm that only optimizes a single objective and the convergence speed is too slow, a new method that pays attention to real-time constraints is proposed. and ant colony task scheduling algorithm for system energy consumption. The method firstly provides guidance information in the process of pheromone initialization according to the energy consumption of tasks on heterogeneous cores to speed up the algorithm convergence speed, and then filters the cores according to the real-time constraints of tasks, and then according to the computing power of tasks on heterogeneous cores. According to the energy consumption of different tasks, the energy consumption of inter-core communication and the pheromone content of different tasks to select the appropriate execution core, and finally through the multiple iterations of the ant colony algorithm, the scheduling scheme with lower energy consumption is continuously searched, and the pheromone content is adjusted according to the obtained results. Speed up algorithm convergence. After several iterations, the final task scheduling scheme is obtained, which can optimize the energy consumption of the system under the condition of satisfying the real-time constraints of the task.
Description
technical field [0001] The invention belongs to the field of heterogeneous system task allocation and resource scheduling, and specifically relates to a resource allocation and task scheduling strategy oriented to CPU-GPU heterogeneous system structure while considering real-time constraints and system energy consumption. Background technique [0002] Nowadays, the processor architecture is gradually developing towards multi-core and heterogeneous. Multi-core processors have become the current mainstream processors. With multiple processing cores, the application program can deliver parallel threads to multiple cores for processing separately, so the running speed of the program is greatly improved. According to whether there are differences in their core structures, multi-core processors are divided into homogeneous multi-core processors and heterogeneous multi-core processors. A homogeneous multi-core processor mostly means that all processor cores have the same architec...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
16 Apr 2021
Publication
CN109960576B
- IPC
- G06F9/48; G06N3/00
- CPC
- G06F9/4881; G06F9/4893; G06F2209/483; G06F2209/484; G06N3/006; Y02D10/00
- Inventors
- 方娟; 周宽