A Deep Dive Into The Performance-Focused AMDGPU "Bulk Moves" Functionality

Recently on Phoronix you have likely heard a lot about the LRU "bulk moves" functionality for the AMDGPU driver after it was talked up by a Valve Linux developer for the performance help to Linux games and then the change landing in Linux 5.4 as a "fix".

Huang Rui was the developer involved at AMD leading the charge on this bulk moving mechanism and he presented at last week's X.Org Developer's Conference on the topic. He mentioned how it came about when they were looking at the performance of the F1 2017 game's benchmark and ultimately seeing a need to redesign their kernel driver's buffer migration code.

This buffer migration code for the "Least Recently Used" code did end up helping out not only Vulkan workloads in better frame-rates but also latency improvements and boosting OpenCL performance too.

Those wanting to learn more about this bulk moves functionality can do so via this PDF slide deck. A few days back I delivered our own preliminary Linux 5.4 tests with AMDGPU bulk moves.