Open achirkin opened 3 years ago
@achirkin could you perhaps use the atomics provided in https://github.com/rapidsai/raft/blob/branch-21.08/cpp/include/raft/device_atomics.cuh? We would like to deprecate/remove conflicting versions of these very soon
Sure, thanks for pointing that out!
This issue has been labeled inactive-90d
due to no recent activity in the past 90 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed.
This issue has been labeled inactive-30d
due to no recent activity in the past 30 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed. This issue will be labeled inactive-90d
if there is no activity in the next 60 days.
Describe the bug Just looking at the code of
myAtomicMax
, it looks like this operation may return the result after some other thread writes into the memory location. An easy fix would be to return the actual old or new value frommyAtomicReduce
. Not sure if this can hurt anybody for now, but it's probably worth fixing.Expected behavior
myAtomicMax
/myAtomicMin
should consistently return the old value to comply with the intrinsic versions.