Analysis
It isn't always possible to find recent quantitative analysis for various software approaches to
base design and implementation decisions on. In light of this, I have decided to keep track of
measurements I have made on this web page in a manner that is somewhat presentable to others.
-
AMD64
During my time at a previous job, I was involved in an effort to quantify and explain certain
performance mismatches between a high performance computing application across different
clusters. These are some of the measurements associated with this effort.
-
BPF
In order to justify some design changes to the FreeBSD BPF buffering model for a packet
capture framework I was working on, I decided to develop a stochastic model of the
FreeBSD BPF implementation. These are the measurements taken to verify accuracy of
that model.
-
Spinlocks
As a follow-up to the AMD64 performance measurements, I decided to begin implementing
fair spinlocks. During this time, ticket spinlocks were committed to Linux. I decided
to (re-)implement various spinlock implementations in order to have a solid ground
to justify the design and implementation decisions I made for my algorithm.
© 2007 - 2010 Samy Al Bahra.