elvin-n opened a new pull request, #11357: URL: https://github.com/apache/tvm/pull/11357
Previous PR11161 added topi compute and schedules which manages textures in dynamic mode through explicit call of cache_read in adreno primitive schedules. This PR adds top-down approach of memory annotation plus adding support in dependent parts. 1. Passes of analyzing memory scope for expressions in the relay graph and modifies relay graph by adding/modifying VirutalDevice with required memory scope 2. Add support of memory scope into graph memory planner 3. Support of static planned textures in json and opencl runtime 4. Modification of Relay->TIR settling taking into account relay expr virtual devices for variables -- 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]
