Please be more descriptive in the readme in how to install, include, test, benchmark this library.
I figured the benchmarks are *-bench, would be nice if make benchmark ran them all, the same way make test runs tests.
- Also what are the new type names? half, float16_t, ...
- Mention if/that it is storage only type, and for CPU computation it needs to be always converted to float32?
- what are some "good" speeds for an intel CPU? How to obtain better? (PSIMD, ...)
Please be more descriptive in the readme in how to install, include, test, benchmark this library.
I figured the benchmarks are
*-bench, would be nice ifmake benchmarkran them all, the same waymake testruns tests.