A couple of questions about the emulated 3dfx performance

[xenaretos]

Recently, while playing around with a machine that has an emulated voodoo1, I found out that, at least in the recompiler mode it runs at unlimited speed. However, when switched to the interpreter mode, it runs at about as fast as a voodoo1 should, and peak fillrate is also a bit more stable.

So, the first question is, is the interpreter actually trying to emulate the real hardware speed, or is it just a coincidence on my machine (at least the control thread is often, but not always maxed out, but the render threads almost never are)?

The second question is about the 3dfx section of the machine monitor. The fillrate monitors, for both pixel and texel, have a primary value and a secondary one in brackets. I haven't found any explanation of what exactly those mean and what the difference is (the second number seems more stable and related to the theoretical throughput though) and would like to know that.

[SarahWalker]

So, the first question is, is the interpreter actually trying to emulate the real hardware speed, or is it just a coincidence on my machine (at least the control thread is often, but not always maxed out, but the render threads almost never are)?

Total coincidence I'm afraid. There's no attempt at speed limiting on the 3DFX emulation.

The second question is about the 3dfx section of the machine monitor. The fillrate monitors, for both pixel and texel, have a primary value and a secondary one in brackets. I haven't found any explanation of what exactly those mean and what the difference is (the second number seems more stable and related to the theoretical throughput though) and would like to know that.

First number is what was actually rendered in the last second. Second number is the nominal "peak performance", ie the fillrate if the render thread was running at 100% utilisation.

[xenaretos]

OK, thanks, good to know.