A multi-dimensional loop automatic vectorization block factor block method and device
A block and factor technology, applied in the field of compilation optimization, can solve the problems of low income of vectorization, data misalignment, etc., and achieve the effect of simple structure, good program speed-up ratio, and improved computing efficiency
Active Publication Date: 2021-07-13
XI AN JIAOTONG UNIV
View PDF5 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0003] The purpose of the present invention is to provide a multi-dimensional loop automatic vectorization block factor block method and device to overcome the problem of low vectorization revenue due to data misalignment in the existing method in the process of multi-dimensional loop automatic vectorization
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
[0026] The present invention is described in further detail below in conjunction with accompanying drawing:
[0027] In order to make the purpose, technical solution and advantages of the present invention more clear, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the implementation examples described here are only used to explain the basic idea of the present invention, and are not used to limit the protection scope of the present invention.
[0028] The present invention adopts matrix multiplication matmul program as program instance, and input array is respectively A, B, and output array is C, and problem scale is N, and row block factor is I, and the block factor of vectorizable loop layer is J. figure 1 It is the data layout of the array C in memory in the matmul program. The data type involved in the program example of the present invention is a double precisio...
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 a multi-dimensional loop automatic vectorization block factor block method and device, which recognizes a vectorizable loop layer for a loop code that has undergone conventional block, selects a block factor for a vectorizable loop layer, and selects a block factor for a vectorizable loop layer. The block factor of the loop layer is selected to maximize the number of all vectorizable data in the loop body, and the block factor of the vectorizable loop layer is adjusted to optimize the parallelism and locality of the program. Calculated by the method of mathematical programming and adjust the block factors of other loop layers, for loops with ill-conditioned scale, a better program speed-up ratio can be obtained than existing methods, and at the same time, it has good scalability. The block factor enhances the alignment of data and increases the number of vectorizable data, so that the program can achieve better performance on the vector processor.
Description
technical field [0001] The invention belongs to the field of compilation optimization, and in particular relates to a multi-dimensional loop automatic vectorization block factor block method and device. Background technique [0002] In computing-intensive applications, especially scientific and engineering computing applications, loop nested calculations consume most of the running time, which is called a program hotspot that needs to be solved urgently. The loop block of the processor calculation program is a widely used loop optimization technology. The code conversion of the nested loop part of the program calculation is performed through affine transformation. On the one hand, it increases the data locality of the program and reduces the cache failure rate; On the one hand, develop the coarse-grained parallelism of the loop code, and make full use of the computing performance of the multi-core processor. The loop iteration after the block resets the memory access order ...
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 Patent Type & Authority Patents(China)
IPC IPC(8): G06F8/41
CPCG06F8/443G06F8/4441
Inventor 刘松伍卫国柴晓菲屈彬马洁
Owner XI AN JIAOTONG UNIV