tqchen commented on code in PR #15305:
URL: https://github.com/apache/tvm/pull/15305#discussion_r1261831523
##########
src/runtime/profiling.cc:
##########
@@ -878,6 +878,10 @@ PackedFunc WrapTimeEvaluator(PackedFunc pf, Device dev,
int number, int repeat,
// skip first time call, to activate lazy compilation components.
pf.CallPacked(args, &temp);
+ // allocate two large arrays to flush L2 cache
+ NDArray arr1 = NDArray::Empty({256 * 1000000 / 4}, {kDLInt, 32, 1}, dev);
Review Comment:
How about we add a parameter, L2 flush_size, which enables us to choose to
do so or not, and when it is 0 we skip
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
