Radeon RADV "ACO" Performance On Mesa 19.3 Looking Good

[phoronix]

Phoronix: Radeon RADV "ACO" Performance On Mesa 19.3 Looking Good

With the ACO shader compiler back-end merged for Mesa 19.3 as of mid-September and more improvements merged since then, it's now very easy to try out this alternative to AMD's longstanding AMDGPU LLVM compiler back-end. ACO, which is developed by Valve and their contractors outside of AMD, is looking quite good not only for quicker shader compilation (more punctual game load times) but also helping the performance of some Vulkan-powered games. Here are our latest ACO benchmarks.

http://www.phoronix.com/vr.php?view=28340

[Snaipersky]

Good stuff Michael, with the faster compiler being attributed to the performance wins of the proprietary PRO drivers, I wonder if ACO has closed that gap.

[xxmitsu]

And mesa master branch is not even containing all the available improvements. As it was mentioned in the merge forum thread:

from the ACO merge request:

> For full performance, the following MRs are required: !1240 !1239 !1236 !1664 !1893

which are not merged yet

Anyway, great to see that even that, it managed to obtain better performance.
Thank you for the testing, Michael!
And thank you Valve & Co. for your efforts, much appreciated!

[schmidtbag]

That's a pretty decent improvement. Not sure it was worth all the time and effort put into the new code, but, it's still a noteworthy performance improvement.

[LeJimster]

That's a pretty decent improvement. Not sure it was worth all the time and effort put into the new code, but, it's still a noteworthy performance improvement.

The main benefit of ACO is faster initial compilation times, and much reduced stutter in games. I would say the performance shown in Michaels benchmarks don't reflect the whole picture. Theres also optimizations to do with DXVK that aren't available with LLVM.

[schmidtbag]

The main benefit of ACO is faster initial compilation times, and much reduced stutter in games. I would say the performance shown in Michaels benchmarks don't reflect the whole picture. Theres also optimizations to do with DXVK that aren't available with LLVM.

Ah, good points. Yes - not everything is so easily shown in benchmarks.

[shmerl]

The Witcher 3 fails to start with mesa-git.

Are you using Navi? It works with Vega. For Navi, there is a separate ACO branch, but it's not working for TW3 yet either, beyond the main menu.

[shmerl]

It would be nice to compare frame rate stability between Mesa-default and Mesa-ACO, with an empty shader cache in both cases.

In case of some games, frame times are being logged into files when benchmarking:

The best way to log framerates and frametimes is using GALLIUM_HUD and Mesa Vulkan overlay. Works regardless of the application.

[tildearrow]

Typos:

But be aware currently the back-end supports just GFX8 and GFX9 hardware throguh Vega:

GFX8 isn't Vega?

[geearf]

Typos:



GFX8 isn't Vega?

From https://github.com/RadeonOpenCompute/ROCm/wiki it seems not.

GFX8 GPUs

• “Fiji” chips, such as on the AMD Radeon R9 Fury X and Radeon Instinct MI8
• “Polaris 10” chips, such as on the AMD Radeon RX 580 and Radeon Instinct MI6
• .....
• “Polaris 11” chips, such as on the AMD Radeon RX 570 and Radeon Pro WX 4100
• “Polaris 12” chips, such as on the AMD Radeon RX 550 and Radeon RX 540

GFX9 GPUs

• “Vega 10” chips, such as on the AMD Radeon RX Vega 64 and Radeon Instinct MI25
• “Vega 7nm” chips, such as on the Radeon Instinct MI50, Radeon Instinct MI60 or AMD Radeon VII