Unlock instant, AI-driven research and patent intelligence for your innovation.

Fixed-point processing method and device thereof

A fixed-point, processor technology, applied in the field of data processing, can solve problems such as powerlessness, and achieve the effect of improving operating efficiency

Inactive Publication Date: 2017-04-19
FUJITSU LTD
View PDF5 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, in abnormal situations such as large value ranges, it seems powerless

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
  • Fixed-point processing method and device thereof
  • Fixed-point processing method and device thereof
  • Fixed-point processing method and device thereof

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0036] figure 1 It is the flow chart of the fixed-point processing method in Embodiment 1 of the present invention, such as figure 1 As shown, the method includes

[0037] Step 101, determining the numerical range of the floating-point variable in the floating-point code that does not include complex operations;

[0038] Step 102, analyzing the target platform processor that performs fixed-point processing on the floating-point variable, so as to obtain the special operation of the target platform processor, and determine the fixed-point strategy according to the analysis result;

[0039] Step 103, according to the numerical range of the floating-point variable and the fixed-point strategy, or according to the numerical range of the floating-point variable, the fixed-point strategy and the predetermined overall fixed-point accuracy, determine the fixed-point mode of each floating-point variable;

[0040] Step 104, perform fixed-point processing on the floating-point variable...

Embodiment 2

[0066] In the above-mentioned embodiment 1, the fixed-point processing method solves the problem of fixed-point precision and improves the operation efficiency. In addition, in some cases, when the test code is tested on a large scale on the target platform processor, some key calculations take a long time and become the bottleneck of the entire fixed-point code. For example, in the fixed-point code of the ARM9 processor, some key operations use 64-bit long integer data processing or even pseudo-floating point processing due to the large range of floating-point variable values. Because these key operations occupy a large part of the original floating-point code Part of the calculation time is longer after fixed-point conversion, which has become the speed bottleneck of the entire fixed-point code. The fixed-point processing method in Embodiment 2 can solve the above problem.

[0067] figure 2 It is the flow chart of the fixed-point processing method in Embodiment 2 of the pr...

Embodiment 3

[0084] Embodiment 3 of the present invention also provides a fixed-point processing method. Based on Embodiment 2, in this embodiment, after the modified fixed-point code is obtained in step 213, accuracy and performance tests are performed on the modified fixed-point code.

[0085]In this embodiment, the difference from Embodiment 2 is that after modifying the bottleneck operation according to the distribution statistics of the numerical range of the floating-point variables related to the bottleneck operation and the fixed-point strategy, and obtaining the modified fixed-point code , first use the test data to perform precision testing and target platform performance testing on the modified part of the fixed-point code, that is, first use the test data to perform precision testing and target platform performance testing on the fixed-point code with bottlenecks, where the accuracy requirements of the precision test It can be consistent with the overall accuracy test requireme...

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

The invention provides a fixed-point processing method and device. The method comprises the steps that the numerical range of floating-point variables in a floating-point code not including complex operation is determined; a target platform processor performing fixed-point processing on the floating-point variables is analyzed so that operation, at which is good, of the target platform processor can be obtained, and a fixed-point strategy is determined according to an analyzing result; the fixed-point modes of all the floating-point variables are determined according to the numerical range of the floating-point variables and the fixed-point strategy or according to the numerical range of the floating-point variables, the fixed-point strategy and predetermined total fixed-point accuracy; according to the fixed-point modes, fixed-point processing is performed on the floating-point variables, so that a fixed-point code is generated; total accuracy testing is performed on the fixed-point code by using test data; when the fixed-point code passes the total accuracy testing, target platform performance testing is performed on the fixed-point code; when the fixed-point code passes the performance testing, fixed-point processing is ended. Based on the fixed-point processing method, the operation efficiency of the fixed-point code can be effectively improved.

Description

technical field [0001] The invention relates to the field of data processing, in particular to a fixed-point processing method and device thereof. Background technique [0002] For reasons of low power consumption, low cost, and ease of design, many processors do not support floating-point operations. Because the operand must be an integer, it also limits the operations that the processor can undertake. If the source operand is a floating-point number, fixed-point processing is required. Its solution also varies according to different numerical characteristics. [0003] The existing fixed-point processing methods often do not consider the characteristics of the target platform but only consider the overall fixed-point accuracy. For different platforms, the fixed-point methods are almost the same, and it is impossible to optimize the fixed-point processing for a specific platform. And in the usual fixed-point method, when the value range is narrow and the operation is rela...

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): G06F9/30G06F11/36
Inventor 张国成
Owner FUJITSU LTD