18:25 <Company> bluetail: in C, different integer types are first promoted to the largest type, and when one of those is unsigned, then the other will be turned unsigned

18:26 <Company> so int64_t + uint32_t will be int64_t I believe (because you can turn a uint32 into int64 without data loss

18:27 <Company> but int32_t + uint32_t will end up as uint32_t (which does lose data and that can be really annoying

18:30 <Company> and that gets really annoying if you (usually accidentally) do int32_t x = -1; uint32_t y = 0; int_64_t z = x + y;

18:32 <Company> because the addition happens first and gets done in uint32_t which then gets converted again for the assignment

19:07 <mareko> int32_t + uint32_t --> uint32_t doesn't lose data, you can cast the result back to int32_t to get signed bits because the addition produces the same bits regardless of the signedness of the type

19:08 <mareko> the uint32_t result type only affects later operations like conversion to float where the type matters

19:09 <mareko> conversions between signed and unsigned int are noops

20:33 <bl4ckb0ne> why does vkCmdBlitImage may perform a format conversion but not vkCmdCopyImage?

20:53 <karolherbst> because one is a blit and the other is a plain copy

20:56 <bl4ckb0ne> might have to check my definition of blit then