Not in the form of the above (I actually did use the above test when I built my Pentium III Tualatin retro rig)

How do you test in general - a small C program directly run from something like FreeDOS with inline assembly (testing all the commands) or more test suites / stress tests kind of thing? For the latter, you could use something like y-cruncher which has a specific SSE3 x86 mode, with an older version without C++11 dependency: https://www.numberworld.org/y-cruncher/version_history.html#2_17_2013

With Intel Generic, older versions actually detect SSE3:

I don't know anything about the coverage (all SSE3 commands?), tho (and unfrotunately know nothing for SSSE3 instead of writing assembly)

Furthermore, all SSE instructions currently fall back to the interpreter, so if someone manages to write recompiled versions of the SSE instructions, that would be a MASSIVE boon to performance.

Grateful I found this statement, as I've spent the past few days trying to troubleshoot performance of the Virtual Acoustic voices for the Yamaha S-YXG100, whose polyphonic variant does seem to use some SSE instructions. Even hacked a custom setting file that only allowed up to two voices with only a marginal improvement, and mind you, my host machine is an M4 Max Mac Studio.

Sadly, while I do know some coding, my experience in C++ is largely limited to sound processing specifically, though Torinde's comment about SIMDe seems like a step in the right direction from my knowing some other projects use it. Hoping for the best with improving performance, preferably soon!

PCBox mainly uses C, not C++. Also, I'd like the code to be as portable as humanly possible, which means no SIMDe, since that can only work on certain architectures.

Thanks for the clarification between C and C++, though for better or for worse, it doesn't change my statement about my admittedly-limited experience with either language. As for your position on SIMDe, while it's admirable that you want PCBox to be as portable as possible, I'm personally struggling to make sense of a better solution insofar as SSE, as from what I've gleaned of that project, it's mostly just pointing opcodes from one architecture to a compatible opcode for another, which will be important for non-Intel users like myself. Granted, your point may be valid in that performance may not improve much on an Intel-based host.

no SIMDe, since that can only work on certain architectures.

@fuel-pcbox, SIMDe seems to support every relevant host. Do you have in mind any other architectures to be supported by PCBox?

It's a generic intel cpu, with more instructions. exactly what it says on the tin :p