On 27/02/2024 09:59, Chaitanya S Prakash wrote:
+ /* Generate an unrepresentable length/address */ + do { + len = (1 << (pageshift++)) + ((count++ % pagesize) * pagesize);
Is there a particular rationale for the added multiplier (count++ % pagesize)? Does that yield an unrepresentable length faster?
Kevin
It was done to randomize the generation of the unrepresentable length/address.
Discussed offline - probably better to keep it simple (just an offset of one page) since there's no particular reason to use a different offset.
v8 is looking good otherwise so I can make that change when merging (once Amit's series is ready).
Kevin
+ } while (len == cheri_representable_length(len));