Icemist commented on code in PR #11465:
URL: https://github.com/apache/tvm/pull/11465#discussion_r904436666
##########
src/runtime/graph_executor/debug/graph_executor_debug.cc:
##########
@@ -55,48 +56,54 @@ class GraphExecutorDebug : public GraphExecutor {
* By default, one `repeat` contains `number` runs. If this parameter
is set,
* the parameters `number` will be dynamically adjusted to meet the
* minimum duration requirement of one `repeat`.
- * \return Comma seperated string containing the elapsed time per op for the
last
- * iteration only, because returning a long string over rpc can be
expensive.
+ * \param cooldown_interval_ms The cooldown interval in milliseconds between
the number of repeats
+ * defined by `repeats_to_cooldown`.
+ * \param repeats_to_cooldown The number of repeats before the
+ * cooldown is activated.
+ * \return Comma separated string containing the elapsed time per op for
+ * the last iteration only, because returning a long string over rpc
can be expensive.
Review Comment:
I wonder what caused it. I didn't have any problems even when there were
tens of thousands of numbers here.
I would guess that the right thing to do here would be to pack the data as
it is done in `time_evaluator`. And then extract the data with `struct.unpack`.
I would like to know the reason for the comment:
// Have problems returning FloatImm so serialize to string results as hack.
@AndrewZhaoLuo can you help with that?
In my previous version of the patch, I did dimanic array packing, and this
was not a problem for `time_evaluator`.
--
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]
