[m-rev.] for review: improve the worst case of sparse bitset list_to_set

[Zoltan Somogyi]

For review by anyone.

For very small bitsets, e.g. those that contain only values
in the range 0-63, the old implementation is probably a bit faster.
Is this important enough to keep the old implementation alongside the new?
And if so, should the old one be list_to_set, and the new one
something like long_list_to_set, or should the new one be
list_to_set, with the old one being short_list_to_set?

Peter, is your setup for benchmarking digraph ops, which involve
sparse bitsets. capable of getting useful data on the performance characteristics
of list_to_set?

Zoltan.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Log.l2s
Type: application/octet-stream
Size: 245 bytes
Desc: not available
URL: <http://lists.mercurylang.org/archives/reviews/attachments/20230123/4398a459/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: DIFF.l2s
Type: application/octet-stream
Size: 10248 bytes
Desc: not available
URL: <http://lists.mercurylang.org/archives/reviews/attachments/20230123/4398a459/attachment-0001.obj>