search

KomputeProject / kompute

General purpose GPU compute framework built on Vulkan to support 1000s of cross vendor graphics cards (AMD, Qualcomm, NVIDIA & friends). Blazing fast, mobile-enabled, asynchronous and optimized for advanced GPU data processing usecases. Backed by the Linux Foundation.
http://kompute.cc/
Apache License 2.0
1.94k stars 146 forks source link

Allow for extending memory allocations for optimization #35

Open axsaucedo opened 4 years ago

axsaucedo commented 4 years ago

When running in specialised hardware there are memory optimizations required on the type of memory that is used and allocated, so would be required to explore efficient ways to expose the memory types that are used (and whether the current abstractions are enough)

axsaucedo commented 4 years ago

Really good resource on memory management, would be interesting to explore if transfer queues could be used for asynchronous memory defragmentation https://www.youtube.com/watch?v=rXSdDE7NWmA