6 ms·

memcpy implementation is a rabbit hole. The main trap is trying to benchmark and profile it without the context of the whole application. I recommend to read

memcpy implementation is a rabbit hole. The main trap is trying to benchmark and profile it without the context of the whole application. I recommend to read a comment in ClickHouse source code (disclaimer: written by me) https://github.com/ClickHouse/ClickHouse/blob/master/base/glibc-compatibility/memcpy/memcpy.h#L6 https://github.com/ClickHouse/ClickHouse/blob/master/base/gl...

vanderZwan 4y ago
> The main trap is trying to benchmark and profile it without the context of the whole application. Emery Berger's "Performance (Really) Matters" talk also goes into this[0]. He paints a really, really depressing picture of how unreliable benchmarks truly are. Like, things like link order having ridiculous impacts and stuff like that. [0] https://www.youtube.com/watch?v=7g1Acy5eGbE&t=306s https://www.youtube.com/watch?v=7g1Acy5eGbE&t=306s