Adaptive fault-tolerant data parallel computing implementation method for multi-core processor
A multi-core processor and parallel computing technology, applied in the field of multi-core processors, can solve problems such as parallel programs cannot be executed correctly, multi-core processors fail, and fail.
Active Publication Date: 2020-02-18
NAT UNIV OF DEFENSE TECH
View PDF5 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0004] However, some cores of multi-core processors may fail. For example, satellites operating in space are affected by space particle radiation. Multi-core processors will also have yield problems during production. Some multi-core processors will have partial core failure problems. Now In existing technologies, when multi-core processors are used to perform data parallel computing, the data parallel program divides data tasks according to the original number of processing cores, so a problem is usually faced: when some cores of the multi-core processor fail, the parallel program cannot Correct execution, in order to ensure correct execution, certain modifications must be made
Method used
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
View moreImage
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
[0034] The present invention will be further described below in conjunction with the accompanying drawings and specific preferred embodiments, but the protection scope of the present invention is not limited thereby.
[0035] The multi-core processor used in this embodiment is as figure 1 As shown, including isomorphic cores numbered 0, 1, 2, ..., q-1 in sequence, each core shares the off-chip memory, and the calculation data and calculation results are stored in the shared off-chip memory, and each core passes the data Parallel computing works together to complete computing tasks. In a specific application embodiment, multi-core processors perform block matrix multiplication (C=C+A*B) data parallel calculations such as figure 2 As shown, using the parallel computing method of data parallelism, the C and B matrices are divided into blocks according to the columns with the basic block size BLOCK as the unit. The single calculation task under the full load of each core is:
...
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
The invention discloses an adaptive fault-tolerant data parallel computing implementation method for a multi-core processor. The method comprises the following steps of: 1, setting a global register R; 2, when the target multi-core processor needs to execute data parallel computing, reading a global register R after each core of the target multi-core processor is started; 3, during each calculation of each core of the target multi-core processor, respectively calculating the address of the data block needing to be calculated by each core and the residual calculation data volume of the currentcalculation task, and calculating the actual load of each core during each calculation; 4, enabling each core of the target multi-core processor to execute calculation according to the data block address and the actual load during each calculation; and step 5, completing calculation of all calculation data volumes and then exiting. Data parallel computing can still be adaptively and correctly executed by the method of the invention when part of cores fail, and the method has the advantages of being simple in implementation method, good in adaptive fault tolerance performance, flexible in application and the like.
Description
technical field [0001] The invention relates to the technical field of multi-core processors, in particular to a multi-core processor-oriented method for realizing adaptive fault-tolerant data parallel computing. Background technique [0002] With the increasingly prominent problems of power consumption and heat dissipation, energy consumption has gradually become an increasingly important factor affecting high-performance computing systems, making the architecture of processors develop towards multi-core. At present, multi-core processors have occupied high-performance computing For the vast majority of the market, in terms of embedded application platforms, the application of multi-core processors is becoming more and more extensive. [0003] There are two main types of parallel methods for multi-core processors: data parallelism and task parallelism, in which data parallel applications such as multi-channel network traffic processing, multiple audio / video channel data pro...
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
Login to View More IPC IPC(8): G06F9/50
CPCG06F9/5044G06F9/505Y02D10/00
Inventor 刘仲郭阳扈啸田希邓林陈海燕李勇吴虎成孙书为李程
Owner NAT UNIV OF DEFENSE TECH