Dell Latitude 5330, i7-1265U, 1.8GHz, 16GB, Windows 10 Enterprise
…(TTOL-device).qsch: 8.1sec fast math enabled ; 3.3sec fast math disabled
…(vanease).qsch: 2.5sec fast math enabled ; 5.8sec fast math disabled
Dell Alienware m15 R2, i7-9750H, 2.6GHz, 16GB, Windows 11 Home
…(TTOL-device).qsch: 9.7sec fast math enabled ; 4sec fast math disabled
…(vanease).qsch: 3.7sec fast math enabled ; 7.4sec fast math disabled
If I put all our test results for comparison, we can have this observation, which one is faster depends on which simulator is used.
- Qspice64.exe (fast math enable) : …(vanease).qsch faster than …(TTOL-device).qsch
- Qspice80.exe (fast math disable) : …(TTOL-device).qsch faster than …(vanease).qsch
You observation is correct, …(vanease).qsch with Qspice64.exe yield fastest simulation time in all computers, but this one with trap oscillation.
…(TTOL-device).qsch with Qspice80.exe is 2nd fastest in all computers, with trap oscillation eliminated.
As a conclusion
- we both run these simulation with same “relative” simulation time, our difference is “absolute” simulation time.
- “fast math” is not guarantee “faster”, as here with a counter example that running …(TTOL-device).qsch can yield a faster result with Qspice80.exe (fast math disabled)