Jef Poskanzer :batman: on Nostr: nprofile1q…6gyrf I wrote two or three versions of bitblit in the '80s / '90s and my ...

nprofile1qy2hwumn8ghj7un9d3shjtnddaehgu3wwp6kyqpqf3nlrmpjja65pzmvlegyl27l0y2m007nncxttdwqndrejajvdzask6gyrf (nprofile…gyrf) I wrote two or three versions of bitblit in the '80s / '90s and my experience was different. I found that if I wrote the inner loop in simple & clear C code, the regular old non-specialized compiler was able to optimize it to run as fast as memory bandwidth would allow. This was easier back then because on-chip L1/L2 caches weren't a thing yet.

One difference is that I was working in the Unix world and gcc was already quite good. Maybe Mac & Windows compilers were not as good back then.