MiBench：嵌入式处理器性能基准分析

需积分: 16 63 浏览量更新于2024-12-03 收藏 69KB PDF 举报

"Mibench是一个针对学生参考的嵌入式系统基准测试套件，它包含了一系列具有商业代表性的嵌入式程序，与现有的SPEC2000基准进行对比。研究中使用了适应ARM指令集的新版SimpleScalar工具来评估这些基准在类似当前和下一代嵌入式处理器上的性能。MiBench基准的特点包括不同的指令分布、内存行为和可用并行性，并且对所有研究人员免费开放。" Mibench是为了解决传统基准测试如Dhrystone、LINPACK、Whetstone和CPU2等在特定计算领域的局限性而设计的。这些传统的基准测试分别专注于系统整数性能、可向量化计算、数值计算（浮点运算）等方面，但它们可能无法充分反映嵌入式系统的性能特性。嵌入式系统通常有着独特的需求，例如低功耗、实时性以及特定应用领域的需求。Mibench的创建旨在更准确地模拟这些系统的行为，其特点包括： 1. **指令分布**：嵌入式程序的指令流可能与通用计算机上的程序不同，Mibench中的基准反映了这种差异，有助于评估处理器在处理嵌入式应用时的效率。 2. **内存行为**：由于资源限制，嵌入式系统通常有独特的内存访问模式。Mibench的基准测试考虑了这些因素，可以分析处理器的内存管理能力及其对系统性能的影响。 3. **可用并行性**：许多嵌入式应用存在潜在的并行性，如多任务处理或数据并行处理。Mibench通过分析这些基准，可以帮助研究者理解和优化嵌入式处理器的并行处理能力。 4. **跨平台兼容性**：Mibench支持ARM指令集，这使得它能够覆盖广泛的应用场景，包括手机、物联网设备以及其他基于ARM架构的嵌入式系统。 5. **开源和研究友好**：MiBench是免费提供给所有研究人员的，这促进了学术界和工业界的研究与合作，推动了嵌入式系统性能评估技术的发展。通过对Mibench基准的分析和测试，设计者可以更好地理解不同嵌入式处理器的性能潜力，从而优化硬件设计或调整软件算法，以适应特定的嵌入式应用需求。此外，Mibench也成为了衡量嵌入式系统性能和进行性能优化的重要工具，对于教育和研究具有重要意义。

2.2. Network

The Network category represents embedded

processors in network devices like switches and

routers. The work done by these embedded processors

involves shortest path calculations, tree and table

lookups and data input/output. The algorithms used to

demonstrate the networking category are finding a

shortest path in a graph and creating and searching a

Patricia trie data structure. Some of the benchmarks in

the Security and Telecommunications category are also

relevant to the Network category: CRC32, sha, and

blowfish. However, they are separated for organization.

dijkstra: The Dijkstra benchmark constructs a large

graph in an adjacency matrix representation and then

calculates the shortest path between every pair of nodes

using repeated applications of Dijkstra’s algorithm.

Dijkstra’s algorithm is a well known solution to the

shortest path problem and completes in O(n

) time.

patricia: A Patricia trie is a data structure used in

place of full trees with very sparse leaf nodes.

Branches with only a single leaf are collapsed upwards

in the trie to reduce traversal time at the expense of

code complexity. Often, Patricia tries are used to

represent routing tables in network applications. The

input data for this benchmark is a list of IP traffic from

a highly active web server for a 2 hour period. The IP

numbers are disguised.

2.3. Security

Data Security is going to have increased importance

as the Internet continues to gain popularity in e-

commerce activities. Therefore, Security is given its

own category in MiBench. The Security category

includes several common algorithms for data

encryption, decryption and hashing. One algorithm,

rijndael, is the new Advanced Encryption Standard

(AES) [12]. The other representative security

algorithms are Blowfish [13], PGP [15] and SHA [14].

blowfish encrypt/decrypt: Blowfish is a symmetric

block cipher with a variable length key. It was

developed in 1993 by Bruce Schneider. Since its key

length can range from 32 to 448 bits, it is ideal for

domestic and exportable encryption. The input data

sets are a large and small ASCII text file of an article

found online.

sha: SHA is the secure hash algorithm that produces

a 160-bit message digest for a given input. It is often

used in the secure exchange of cryptographic keys and

for generating digital signatures. It is also used in the

well-known MD4 and MD5 hashing functions. The

input data sets are the same as the ones used by

blowfish.

rijndael encrypt/decrypt: Rijndael was selected as

the National Institute of Standards and Technologies

Advanced Encryption Standard (AES). It is a block

cipher with the option of 128-, 192-, and 256-bit keys

and blocks. The input data sets are the same as the ones

used by blowfish.

pgp sign/verify: Pretty Good Privacy (PGP) is a

public key encryption algorithm developed by Phil

Zimmerman. It allows you to communicate securely

with people you’ve never met using digital signatures

and the RSA public key cryptosystem. The input data

for both the large and small tests is a small text file.

This is because PGP is usually only used to securely

exchange a key for a block cipher which can then

encrypt/decrypt data at a much faster rate.

2.4. Consumer Devices

The Consumer Devices benchmarks are intended to

represent the many consumer devices that have grown

in popularity during recent years like scanners, digital

cameras and Personal Digital Assistants (PDAs). The

Table 1: MiBench Benchmarks

Auto./Industrial Consumer Office Network Security Telecomm.

basicmath jpeg ghostscript dijkstra blowfish enc. CRC32

bitcount lame ispell patricia blowfish dec. FFT

qsort mad rsynth (CRC32) pgp sign IFFT

susan (edges) tiff2bw sphinx (sha) pgp verify ADPCM enc.

susan (corners) tiff2rgba stringsearch (blowfish) rijndael enc. ADPCM dec.

susan (smoothing) tiffdither rijndael dec. GSM enc.

tiffmedian sha GSM dec.

typeset

剩余11页未读，继续阅读

huang1167c

粉丝: 0
资源: 1

MiBench：嵌入式处理器性能基准分析

mibench_bitcnts

opos api 110 doc for reference

DDC & DUC for reference Design

Optimization algorithm based on matrix calculus for reference

C for AIX Compiler Reference

Reference Architecture for SOA

HDL Compiler for Verilog Reference Manual.pdf

C++ for Dummies Quick Reference Guide

Logdump Reference for Oracle GoldenGate

reference data for radio engineers

最新资源