Device and method for determining performance of floating-point calculation

Abstract

The invention discloses a method for determining the floating-point computing performance of a computing unit, which includes obtaining the number of various floating-point micro-operations in the computing unit within a period of time, and based on the number of various floating-point micro-operations and The length of the time period is used to determine the floating-point computing performance of the computing unit, where various floating-point micro-operations include the following types of floating-point micro-operations: floating-point micro-operations of the X87 instruction type, the number of floating-point micro-operations of the MMX instruction type, and SSE Instruction-type floating-point micro-operations. The invention also discloses a corresponding device for determining the floating-point calculation performance of the calculation unit.

Description

technical field [0001] The invention relates to the field of performance analysis of a computer system, in particular to a device and a method for determining the performance of a computer system's CPU. Background technique [0002] The central processing unit (Central Processing Unit, hereinafter referred to as CPU) is one of the most important devices in a computer. Its main function is to interpret computer instructions and process data in computer software. Due to the important position of the CPU in the entire electronic computer system, its computing power and execution speed have become the indicators that people care about. A very important index to measure the CPU computing performance of a computer system is the computing peak value or floating-point computing peak value, which refers to the maximum number of floating-point operations that a computer can complete within a unit of time. [0003] A measure of the ability of a computer to perform floating-point calcu...

AI Technical Summary

Problems solved by technology

[0004] At present, there is no device and method that can accurately and timely count the corresponding floating-point computing performance Gflops data when the computing unit executes the application program. Existing industry users usually calculate it according to the computing scale and data volume of their own applications. The overall computing power of the application, and then deduce the Gflops value of each computing node or single CPU core in the system on the computer according to the execution time

Images