Part 2 notes

#Part 2.1

Command:

xmem -l -j1 -f part2q1-w4k.csv > part2q1-w4k-out.txt -v -w4

with -w# starting from #=4 (4KiB) in increasing multiples of 2, up through 524288 (524288KiB/512MiB)

(a)

I plotted working set size vs. mean (average) latency, as on page 8 of the paper.

#Part 2.2

Command for (a):

xmem -c64 -w4 -j1 -s -S1 -v -f part2q2_c64_stride1.csv > part2q2_c64_stride1_out.txt

(replace stride1 with stride1, 2, 4, 8, 16 and negative versions)

Repeat with -c32 for (b).