Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Combination and optimization methods of access and storage based on analysis of data stream

A technology of data flow analysis and optimization method, which is applied in the field of general-purpose and embedded compiler memory access optimization, can solve the problems of lack of accurate description, etc., and achieve the effect of easy implementation, good optimization performance, and reduced bandwidth requirements

Inactive Publication Date: 2008-05-28
G CLOUD TECH
View PDF3 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The disadvantage of this method is that it is only for the incremental or decremental access to the array elements in the program loop, and the width of the data access is also limited to the 64-bit instruction width; it lacks the precise description of the bit information of the access operation of the entire program ; At the same time, for embedded processors, the performance bottleneck of the program does not lie in the cyclic array access, but in the operation and access to specific data, which often spreads throughout the entire program

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 more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Combination and optimization methods of access and storage based on analysis of data stream
  • Combination and optimization methods of access and storage based on analysis of data stream
  • Combination and optimization methods of access and storage based on analysis of data stream

Examples

Experimental program
Comparison scheme
Effect test

example 1

[0073] Example 1: It is assumed that each basic block contains at most one read operation or write operation, that is, there is at most one memory access operation. Let Out(x), Out(y) be the output data sets of the successor block x and the successor block y of the basic block i respectively, and In(i) be the input data set of the basic block i. There are two memory access operations in the output data set Out(x) of the successor block x, namely r(l, 20, 8) and r(m, 32, 4), and in the output data set Out(x) of the successor block y There are two memory access operations in y), namely r(l, 16, 8) and r(q, 64, 32), where r(l, 20, 8) and r(l, 16, 8) represent For the two read operations implemented in the basic block l, since the basic block l can only have one memory access operation, the two read operations must be the same read operation and the same element. Therefore, an AND operation is performed on the data areas of the two operations. In the read operation r(1, 20, 8), ...

example 2

[0078]Example 2: Assuming that basic block i contains write operation w(i, 16, 8), the input data set of basic block i is {r(l, 8, 16), r(m, 20, 24)}, since the The write operation resets the data of 16 to 24 bits, so all relevant data bits in the input data set should be set as invalid. The data bits in r(l, 8, 16) are from 8 to 24 bits, and r(m , 20, 24) the data bits are from 20 to 44 bits, and the write operation resets the data of 16 to 24 bits. Therefore, the effective bits of the data in r (1, 8, 16) are from 8 to 16 bits , the effective bits of the data in r(m, 20, 24) are from 24 to 44 bits, and the changed input data set is finally {r(l, 8, 8), r(m, 24, 20)}. The specific implementation method of setting the data bits related to the write operation as invalid is to first invert the valid data bits of the write operation, and then perform an "AND" operation with the data bits of all operations in the data set, where valid data is represented by 1, and invalid The dat...

example 3

[0094] Example 3: For the read operation r(i, x, s) at the basic block i, suppose any read operation r(j, y, t) is located at the basic block j, if the input set in(i ) exists r(j, y', t') and y=y', t=t', then r(j, y, t) is active at basic block i, and it can be compared with r(i, x , s) merge. Construct a set avai_set(i), and all read operations that can be combined with r(i, x, s) are elements in the set. The same process can be done for the write operation at the basic block i.

[0095] Step 60, judging whether the set used to save the active memory access operation is empty, if it is empty, then jump to step 140, otherwise, execute the next step;

[0096] Step 70 , for all elements in the set obtained in step 50 , respectively calculate the merge density between the memory access operation corresponding to each element and the current memory access operation. The method for calculating the combined density is to add the data width of the first memory access operation to...

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

No PUM Login to View More

Abstract

This invention provides one visit merge and optimization method based on data flow analysis, which merges two or more visit orders into one multi-bit visit order of adjacent address through total visit memory information for data flow analysis. The merged visit data is in register or local memory, the original visit memory order is taken place of register or local memory direct visit.

Description

technical field [0001] The invention relates to the technical field of memory access optimization for general and embedded compilation, and relates to a method for optimizing memory access and merging based on data flow analysis. Background technique [0002] In computer technology, the speed of the computer processor is much faster than the memory access speed, and the speed of the computer processor is much higher than the memory access speed. Therefore, the processor spends a lot of time during the operation of the computer. Waiting to fetch data, which makes memory fetching increasingly a bottleneck of computer performance. The optimization technology related to memory access is an important type of compilation optimization technology. This technology reduces the delay of memory access and improves the utilization rate of memory access bandwidth through various methods. [0003] Memory access optimization technology mainly uses multi-level caching on hardware to reduce ...

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
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06F9/45
Inventor 刘弢吴承勇
Owner G CLOUD TECH
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com