Entry tags:
hard things in computer science
Now at work I learn the hard way that the saying "There are only two hard problems in Computer Science: cache invalidation and naming things" is so true.
And I mean cache invalidation, or rather cache replacement. I am running a STREAM-scale like u-benchmark on a Skylake server, performance is at 14.3-17.1GB/sec, and this ~20% variation of performance is driven by very obscure factors..
And I mean cache invalidation, or rather cache replacement. I am running a STREAM-scale like u-benchmark on a Skylake server, performance is at 14.3-17.1GB/sec, and this ~20% variation of performance is driven by very obscure factors..