OpenCL: large graph support through RAM->VRAM streaming #211

Activity

1. EgorOrachyov

   added

   enhancementNew feature or requestNew feature or request

   help wantedExtra attention is neededExtra attention is needed

   topic:gpuOpenCL (Compute) relatedOpenCL (Compute) related

   topic:apiLibrary API relatedLibrary API related

   topic:perfPerformance optimziationsPerformance optimziations

   on Mar 11, 2023
2. EgorOrachyov

   added this to the Stage 4. New horizon milestone on Mar 11, 2023
3. EgorOrachyov

   assigned

   gsvgit

   and

   EgorOrachyov

   on Mar 11, 2023
4. gsvgit commented on Mar 16, 2023

   

   gsvgit

   on Mar 16, 2023

   Last edited by gsvgit

   Member

   More actions

   What do you think about SVM ? Can it be helpful in this task?
5. EgorOrachyov commented on Mar 16, 2023

   

   EgorOrachyov

   on Mar 16, 2023

   Last edited by EgorOrachyov

   ContributorAuthor

   More actions

   @gsvgit By def, shared virtual memory is a virtual address space for multiple processors, devices which can access the same address in seamles fashion.

   The question is, how it is implemented for a gpu.

   When and how gpu loads pages?
   What happens if there not enough GPU space?
   What is the maximum allocation size for a svm allocation?

   Streaming solves the problem of large vram usage and overcomes limit of single allocation, which also is a limit of maximum range, which can be addressed inside a kernel.

   Svm seems like a gimmick. But how it is implemented across vendors and which overhead has?

   All in all, seems svm wont solve problem of large data.