How-to install SimpleScalar on Ubuntu
EDIT June 2017: Using the Dockerfile, you can use this script whatever your OS. Install Docker, then follow comments at the top of Dockerfile.
SimpleScalar is a VLIW processor simulator. It allows to access measures like execution time, caches misses, and branch mis-prediction rates. It is entirely scalable: cache size and type, number of UAL entities, branch predictor types are customizable. It can be used as a tool for understanding Design-Space-Exploration. Unfortunately, the installation is a hassle due to an outdated code source.
The following script intends to automate the installation process with Ubuntu. It has been tested with Ubuntu 16.04 LTS and gcc 4.6. It may not resolve all the errors you will have to face: don't hesitate to contribute if you think the script has to be improved. If you can choose not to use this simulator, you might want to take a glance at other simulators like SoCLib or gem5. A more exhaustive list is available at wikipedia (en)
The script
install_simple_scalar.sh will create ~/simple_scalar, and install SimpleScalar in it.
Set the script as executable (chmod +x install_simple_scalar.sh), run it, this it it! You should be able to compile your first program using a command-line such as: "sslittle-na-sstrix-gcc -x c++ main.c"
Miscellaneous
- gcc 2.6 does not accept inline comments with C. A simple workaround is to compile your C code as if it was a C++ code: use for instance
sslittle-na-sstrix-gcc -x c++ main.c - You might want to change some compilation flags in the SimpleScalar Makefile: Replace "
-O0" by "-O3 -march=native -fstrict-aliasing" and removing the "-DDEBUG" flag gives a substantial speed gain (near x2.0) - Dimensioning your architecture is always a trade off: Cache speed decreases with size. For a quantitative idea of the speed you can expect from a cache, you might be interested by the tool CACTI, luckily far more simple to install than SimpleScalar.
Connected links
Are you stuck with a bug not described here? The following link could be useful: