Time  Nick      Message
18:02 sfan5     added testing code to #15531
18:02 ShadowBot https://github.com/minetest/minetest/issues/15531 -- [no squash] Group sparse mesh buffers over entire scene for rendering by sfan5
18:49 sfan5     std::vector<S3DVertex> does some inefficient things when inserting :(
18:52 Krock     yes. it's a constant realloc if you're appending without planning beforehand
18:54 sfan5     not that
18:55 sfan5     since S3DVertex is not a trivial type (has a default constructor) it uses a for-loop copy instead of invoking memcpy
18:55 sfan5     and apparently the compiler can't turn that into a memcpy either
18:57 sfan5     (memcpy is typically SSE/AVX optimized)
19:15 sfan5     however that's not a difference in practice on my machine at least
19:15 sfan5     waste of time to investigate this further
19:17 Krock     sfan5: did you know? 99% of all opimizations were stopped before the next major breakthrough? https://i.kym-cdn.com/photos/images/newsfeed/002/244/642/a5f.jpg
19:17 Krock     (this is satire, if that wasn't already obvious)
20:21 sfan5     merging #15523, #15521, #15472 in 10m
20:21 ShadowBot https://github.com/minetest/minetest/issues/15523 -- [no sq] Don't recalculate meshnode normals unnecessarily by sfan5
20:21 ShadowBot https://github.com/minetest/minetest/issues/15521 -- [no squash] VoxelArea fixes v2 by sfan5
20:21 ShadowBot https://github.com/minetest/minetest/issues/15472 -- Use openssl's sha1 and sha256, optionally by Desour
20:42 Krock     hmm. the OpenSSL functions could also be loaded during runtime without linking. But I suppose that would be rather messy to get to work cross-platform.