MiBench

Source Code from: http://vhosts.eecs.umich.edu/mibench/source.html

Not Ported

• consumer/tiff* Needs system wide installation
• office/ghostscript Not compatible with stdlib.h
• office/ispell input file has wrong format
• office/sphinx Not compatible with stdlib.h
• security/rijndael Not compatible with stdlib.h/stdio.h
• security/pgp No compatible system provided by Makefile
• consumer/mad uses configure
• office/rsynth uses configure

Build

• set PLATFORM, CC and CFLAGS
• run make
• after successful compilation all binaries can be found in bin

Input Generation

automotive

• bitcount number of runs can be adjusted by export BASICMATH_RUNS=<X> where as <X> is a positive integer (larger <X> means longer runtime)
• bitcount input number be adjusted by export BITCOUNT_INPUT=<X> where as <X> is a positive integer (larger <X> means longer runtime)
• qsort_small run python input_generation/generate_qsort_small_input.py [NUMBER_OF_WORDS] which generates the file input_data/qsort_small_input.txt
• qsort_large run python input_generation/generate_qsort_large_input.py [NUMBER_OF_VECTORS] which generates the file input_data/qsort_large_input.txt
• susan run python input_generation/generate_susan_input.py [WIDTH] [HEIGHT] which generates the file input_data/susan_input.pgm

consumer

• jpeg run python input_generation/generate_jpeg_input.py [WIDTH] [HEIGHT] which generates the file input_data/jpeg_input.ppm
• lame run python input_generation/generate_lame_input.py [SECONDS] which generates the file input_data/lame_input.wav
• typeset run python input_generation/generate_typeset_input.py [WORDS] which generates the file input_data/typeset_input.lout

network

• dijkstra run python input_generation/generate_dijkstra_input.py [NUM_NODES] which generates the file input_data/dijkstra_input.dat
• patricia run python input_generateion/generate_patricia_input.py [DURATION] [EVENTS] which generates the file input_data/patricia_input.udp

office

• stringsearch number of runs can be adjusted by export STRINGSEARCH_RUNS=<X> where as <X> is a positive integer (larger <X> means longer runtime)

security

• blowfish run python input_generation/generate_blowfish_input.py [CHARS] which generates the file input_data/blowfish_input.asc
• sha run python input_generation/generate_sha_input.py [CHARS] which generates the file input_data/sha_input.asc

telecomm

• crc run python input_generation/generate_crc_input.py [BYTES] which generates the file input_data/crc_input.pcm
• fft input waves ans lengths in runme_variable.sh can be adjusted by export FFT_WAVES=<X> and export FFT_LENGTH=<Y> where as <X> and <Y> are postive integers (larger <X> and <Y> mean longer runtime)
• adpcm run python input_generation/generate_adpcm_input.py [BYTES] which generates the file input_data/adpcm_input.pcm
• gsm run python input_generation/generate_gsm_input.py [BYTES] which generates the file input_data/gsm_input.pcm

Run

Each benchmark has its own script which is located in bin.

cd bin
./run_${PLATFORM}

Results

Those results are obtained with the input generation inputs as specified in the Makefile. Both Intel Systems run Scientific Linux 7.6. The AMD System runs Ubuntu 18.04. All benchmarks are executed three times using the /usr/bin/time command which measures the system, user, and real time. Each bar of the plot shows the real time separated into system time, user time, and the difference of the sum of system and user time compared to the real time. The error bars show the standard deviation of the real time for all three executions.