The cs61-lectures/net8 directory contains a program called
matrixmultiply (source in matrixmultiply.cc). This program creates
two random square matrices, multiplies them together, prints some
information about their product, then exits.
Can you use the knowledge you’ve gained in the class to radically
speed up matrixmultiply, while producing (almost) exactly the same
results as the original for all arguments?
Note 1: You can copy matrixmultiply.cc into other filenames; for
instance, given a file like matrixmultiply-magic.cc, the makefile
will generate binaries like matrixmultiply-magic.
Note 2: Docker overheads can overwhelm some performance improvements for this exercise. Try running outside Docker if you can.