Author here. First of all, thanks for the compliment! It’s tough to get myself to write these days, so any motivation is appreciated.
And yes, once all the usual tricks have been exhausted, the nest step is looking at the cache/cache line sizes of the exact CPU you’re targeting and dividing the workload into units that fit inside the (lowest level possible) cache, so it’s always hot. And if you’re into this stuff, then you’re probably aware of cache-oblivious algorithms[0] as well :)
Personally, I almost never had the need to go too far into platform-specific code (except SIMD, of course), doing all the stuff in the post is 99% of the way there.
And yeah, C# is criminally underrated, I might write a post comparing high-perf code in C++ and C# in the future.
>> C# has an awesome situation in here with its support for value types (ref structs), slices (spans), stack allocation, SIMD intrinsics (including AVX512!). You can even go bare-metal and GC-free with bflat.
There's been a really solid effort by the maintainers to improve performance in C# , especially with regard to keeping stuff off the heap. I think it's a fantastic language for doing backends in. It's unfortunate that one of the big language users, Unity, has not yet updated to the modern runtime.
One other trick I use reasonably often is using something more complicated than AoS or SoA layouts. Reasons vary (the false sharing padding in your article is one example), but cache lines are another good one. You might, e.g., want an AoSoA structure to keep the SoA portion of things on a single cache line if you know you'll always need both data elements (the entire struct), want to pack as much data in a cache line as possible, and also want that data to be aligned.
Author here, kinda sorta. I should've been a bit more specific than that.
You can have a profile showing a function taking up 99% of the time, but when you dive into it, there's no clear bottleneck. But just because there's no bottleneck, that doesn't mean it's optimized; vice versa-a well-optimized program can have a bottleneck that's already been cycle-squeezed to hell and back.
What I wanted to say was that a spiky profile provides a clear path to optimizing a piece of code, whereas a flat profile usually means there are more fundamental issues (inefficient memory management, pointer chasing all over the place, convoluted object system, etc.).
It sounds like a flat profile essentially is a local optimum, compared to cases where there's a path "upwards" along a hill to some place more optimal that doesn't require completely changing your strategy.
That's actually a good observation, yeah. It's often the case that you dig deeper and deeper and find some incomprehensible spaghetti and just say "fuck it, I'll just do what I can here, should be enough".
And yes, once all the usual tricks have been exhausted, the nest step is looking at the cache/cache line sizes of the exact CPU you’re targeting and dividing the workload into units that fit inside the (lowest level possible) cache, so it’s always hot. And if you’re into this stuff, then you’re probably aware of cache-oblivious algorithms[0] as well :)
Personally, I almost never had the need to go too far into platform-specific code (except SIMD, of course), doing all the stuff in the post is 99% of the way there.
And yeah, C# is criminally underrated, I might write a post comparing high-perf code in C++ and C# in the future.
[0]: https://en.wikipedia.org/wiki/Cache-oblivious_algorithm