I could not reproduce the authors results on clang 8.0 or gcc 8.1 at O3. The uint8_t version performed roughly 6 times faster than the uint32_t version. The post was published in 2019 so perhaps something has changed…
I could not reproduce the authors results on clang 8.0 or gcc 8.1 at O3. The uint8_t version performed roughly 6 times faster than the uint32_t version. The post was published in 2019 so perhaps something has changed…