Communication-optimal parallel 2.5D matrix multiplication and LU factorization algorithms, Edgar Solomonik and James Demmel, Tech. Rep. UCB/EECS-2011-10 (Feb. 2011).
Skip LU factorization (§4-6, and parts of §7-8)
To probe further into Communication-avoiding algorithms, visit the Bebop web site or Jim Demmel’s Course on Communication-Avoiding algorithms
Ten Ways to Fool the Masses When Giving Performance Results on GPUs, Scott Pakin, HPCWire. (Refers to David Bailey’s original 1991 paper: Twelve Ways to Fool the Masses When Giving Performance Results on Parallel Computers.)
Further reading on this subject: Misleading Performance Reporting in the Supercomputing Field, by David H. Bailey. Scientific Programming, 1(No. 2):141-151, Winter 1992.