This makes sure that task switches can happen during deserialization, so the LS doesn't appear unresponsive. It also speeds up deserialization by a factor of 10 to 20 for e.g. CUDA.jl. Lastly, progress reporting now includes a rough percentage and more info as to what's actually going on (instead of the generic "indexing packages...").
This makes sure that task switches can happen during deserialization, so the LS doesn't appear unresponsive. It also speeds up deserialization by a factor of 10 to 20 for e.g. CUDA.jl. Lastly, progress reporting now includes a rough percentage and more info as to what's actually going on (instead of the generic "indexing packages...").