Intel vs AMD

[MrTyton]

Hello,

We have reason to believe that this fix causes the game to run faster on AMD processors when compared to Intel ones. We are still compiling data to validate this, but could it be due to the Make here? It contains the -masm=intel flag, which might cause unexpected calls/instructions for an AMD processor.

Would it be possible to create a version of the .dll that was compiled with AMD instructions so that we can investigate further?

Thanks.

[Nucleoprotein]

I dont think release build are using MinGW, they compiled using Visual Studio but which one is question to @rebtd7 Maybe this is related to frame limiter unlock here: https://github.com/rebtd7/FF13Fix/blob/master/d3d9ex/AutoFix.cpp#L300 as maybe some floating point math is giving different results? Like this issue: https://cookieplmonster.github.io/2020/07/19/silentpatch-mass-effect/

[MrTyton]

Ok, I wasn't sure if they were or not - but it's the only obvious start point, hence why I brought it up.

FWIW all of the tests are being done with a target FPS of 60. I don't see anything in that func that directly uses floating point math though, unless you're concerned with how the value ends up getting used by the game later? It's a possibility. I'm unsure how to proceed.

Thanks for the help.

[rebtd7]

As @Nucleoprotein mentioned, there are no specific Intel optimizations enabled in this project afaik. (even if it did, the compiled code of the DLL itself basically does nothing that could interfere with the game performance)

It is possible that there are some floating point math in the original game that is causing issues; they might even be unrelated to any change done by this mod, but as the game performs better with it, I suppose it may make it more noticeable.

Are you able to isolate the other variables like OS version, OS timer accuracy, GPU models/drivers, etc? I think any of those could possibly make a small difference on something like this