RPCS3
Need for Speed: Most Wanted - LLVM runs slower than ASMJIT
Quick summary
Using LLVM recompiler runs significantly slower than ASMJIT.
Details
Heres the performance comparison between LLVM and ASMJIT.
LLVM recompiler: 9.2/10.8/11.8 FPS
Screenshot
ASMJIT recompiler: 14.8/16.8/18.4 FPS
Screenshot
I disabled the RSX tiled memory to show better performance results. The strange thing about LLVM is that it's supposed to be faster than/tied with ASMJIT, which is why I thought it would be necessary to post this bug.
Attach a log file
Attach capture files for visual issues
No response
System configuration
AMD Ryzen 5900X 12-Core Processor | 24 Threads | 15.89 GiB RAM | RTX 3080 driver 545.84.0.0 | Windows 10 Pro 22H2
Other details
No response
I've done some research by testing other SPU options. I've noticed that SPU Interpreter dynamic slightly improves the performance over ASMJIT, although this does causes a huge audio slowdown when enabling these options.
ASMJIT: 15.3/16.8/18.1 FPS
Screenshot
Interpreter dynamic: 16.4/17.4/18.5 FPS
Screenshot
Interpreter static: 10.5/12.2/13.2 FPS
Screenshot
I've ranked each SPU options from slowest to fastest:
- Interpreter dynamic (fastest)
- ASMJIT recompiler
- Interpreter static
- LLVM recompiler (slowest)
The ASMJIT performance seems to have regressed when enabling RSX tiled memory, while the LLVM performance works totally fine.
