Trying The LLVM SI Machine Instruction Scheduler With Mesa 17.2 + LLVM 5.0
A Phoronix reader was recently making comments about the LLVM SI Machine Instruction Scheduler "sisched", so I decided to run some fresh benchmarks of this opt-in feature for RadeonSI Gallium3D.
With a Radeon R9 Fury on the i7-7700K box I decided to run some "sisched" tests using the Padoka PPA built against Mesa 17.2-dev this week with LLVM 5.0 SVN AMDGPU.
It's been over one year since last looking at sisched benchmarks so curiosity was there over the impact with modern Mesa and hardware. This scheduler is designed to boost performance for some workloads.
Though in my tests this week, I didn't see any impact with the tested hardware/software configuration...
Those curious can poke at this result file but it didn't end up yielding any meaningful difference in the tests carried out.
With a Radeon R9 Fury on the i7-7700K box I decided to run some "sisched" tests using the Padoka PPA built against Mesa 17.2-dev this week with LLVM 5.0 SVN AMDGPU.
It's been over one year since last looking at sisched benchmarks so curiosity was there over the impact with modern Mesa and hardware. This scheduler is designed to boost performance for some workloads.
Though in my tests this week, I didn't see any impact with the tested hardware/software configuration...
Those curious can poke at this result file but it didn't end up yielding any meaningful difference in the tests carried out.
11 Comments