I feel like I just read the introduction to a really interesting article.
It really seemrd like there was more to be said there.
Agreed, although that is his personal style. Most of his blog posts are interesting although short. I see them more as very well formatted musings.
It's not necessarily a bad thing, in-fact the world might benefit from more of this style. It's "here's some data, decide what it means for yourself"
I think the expectation of more comes from the experience of predominantly encountering articles with a different form.
[dead]
It’s a very interesting benchmark (https://github.com/lemire/TestingMLP) — probably worth adding to the Phoronix or some wider suite.
Every couple of years I refresh my own parallel reduction benchmarks (https://github.com/ashvardanian/ParallelReductionsBenchmark), which are also memory-bound. Mine mostly focus on the boring simple throughput-maximizing cases on CPUs and GPUs.
Lately, as GPUs are pulled into more general data-processing tasks, I keep running into non-coalesced, pointer-chasing patterns — but I still don’t have a good mental model for estimating the cost of different access strategies. A crossover between these two topics — running MLP-style loads on GPUs — might be exactly the benchmark missing, in case someone is looking for a good weekend project!
I wish the chart extended past 28, otherwise how do we know that it tops out there?
You don't; the author explains that testing beyond that produces noise that makes it hard to analyze.
It's pretty trivial to keep randomising the array and plot some min/max bands, or just the average.
[dead]