Getting Mozjpeg faster on x86 and ARM
(The TL;DR version of this for those who just want the code, grab the fork here) Why optimize? Software is more of an art than a science. In software, there are a variety of programming languages and within each there are nearly an infinite number of ways to solve the same problem with different code. Solving the same problem on the same hardware with different solutions uses different amounts of time (and energy). The difference can be dramatic and when millions (billions?) of people are using the same software every day, the impact can be measured in productivity, dollars or pollution avoided. Computers are not like a wood fireplace where the wood burns the same whether you make use of the heat or not. Computers do work on demand and sleep (use less energy) when not in use. The faster the work can get done, the more time can be spent sleeping or the quicker the next task can begin.
The internet is run on mostly open source code. Some of this code has been painstakingly optimized, but unfortunately much has not. Code optimization is a specialized set of skills that most software developers don’t possess. It’s not practical or even feasible to fully optimize every project / library / tool in use today; there’s just too much code and not enough skilled people to dedicate to such a task. It is beneficial when some low hanging fruit is discovered that benefits a lot of people from a small investment of time. One such open source tool that we use in our image optimization service is MozJPEG. Optimal performance is critical for our customers and the environment, so we took some time to examine if there was room for optimization.
