README.md

fast-code-examples

This repository houses implementations of optimization examples from the paper:

“How To Write Fast Numerical Code: - A Small Introduction” By Srinivas Chellappa, Franz Franchetti, and Markus Püschel

All examples can be executed from the Makefile

Make Targets

$ make help                    # list outs all available options
$ make isRightServer           # tests that on node sandbox
$ make show_cache_info         # outputs cache info from getconf
$ make show_simpleMMM          # displays the code SimpleMMM.cpp
$ make show_simpleDynamicMMM   # displays the code SimpleDynamicMMM.cpp
$ make show_blockingMMM        # displays the code BlockingMMM.cpp
$ make show_blockingDynamicMMM # displays the code BlockingDynamicMMM.cpp

make run_simpleMMM M=300 N=300 P=300 OUTPUT=0 CACHE=1

Compiles and executes a simple matrix-matrix multiplication example. The example will create an MxN matrix and an NxP matrix, and multiply them to create an MxP matrix.

The OUTPUT option specifies whether the three matrics should be included in the output.

The CACHE option specifies the type of PAPI information to display. CACHE=1 displays cache information from PAPI, while CACHE=0 displays instruction count and FP operations.

make run_blockingMMM M=300 N=300 P=300 NB=100 OUTPUT=0 CACHE=1

Compiles and executes a blocking matrix-matrix multiplication example. The example will create an MxN matrix and an NxP matrix, and multiply them to create an MxP matrix. The program will do block multiplication with block sizes NB. The block size must divide evenly into the matrices.

The OUTPUT option specifies whether the three matrics should be included in the output.

The CACHE option specifies the type of PAPI information to display. CACHE=1 displays cache information from PAPI, while CACHE=0 displays instruction count and FP operations.