- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I see that mvNCCompile has a "-s" switch for configuring the maximum number of shaves at compile-time. I have 2 questions related to this:
1) Is there any way to configure the exact number of shaves to be used and not the maximum ?
2) Can the shaves be turned on/off at run-time through through some API that is already available or otherwise ?
Thanks.
Link Copied
2 Replies
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Any thoughts guys ? Trying to get a feel for the impact of the number of shaves on perf - more precise/runtime control would help.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
@darrengreene At the moment, there isn't a way to configure the exact number of SHAVEs to be used via the API or the toolkit.
The reason for implementing the -s option the way we did was for power and performance reasons. The device runtime code may use fewer SHAVEs for some layers where measurements have typically shown no inference performance degradation (and consequently show a power benefit) from using fewer SHAVEs.

Reply
Topic Options
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page