Arm desktop: x86 emulation | Lobsters

Cannot compare Geekbench results cause it crashes randomly after configuration changes

Hm, with FEX 2507 Geekbench always crashes after the ray tracer on the Snapdragon X1E.


I got some additional hints from FEX-emu folks to make emulation a bit faster. One is reducing precision of x86 FPU, second is disabling TSO.

Oh wow, thanks for the tip – I was not expecting such a huge difference so I haven’t tried it before, but now I’m sure glad I did!

Turns out, the CPU bottleneck I’ve been experiencing in Severed Steel (one of the best FPSes ever, btw) was entirely due to TSO emulation. "TSOEnabled": "0" brought it from “tanking to the 20s-30s when not looking at a wall” to a solid 60 frames per second (on a low enough resolution to exclude GPU perf from the equation). Now it runs better on the Snapdragon than it ever did on my previous Ryzen 5850U laptop!

(upd: huh, one planned optimization “will have a disproportionately large improvement on Qualcomm’s Oryon which has stalls when a cacheline is accessed as LRCPC and then non-LRCPC”!)

Read more here: Source link