Thank you for checking! Definitely 32 seconds on my Tandy 200. Is this only a problem for Virtual T's 200 emulation? I think this would have been noticed before if it affected the Model 100, or perhaps it's a regression?
--b9 On February 9, 2026 8:12:15 AM PST, Kenneth Pettit <[email protected]> wrote: >I tried that for loop in VirtualT 2.4MHz speed. And yes, it only took 17 >seconds. I did not validate this on an real T200 regarding the 32s speed. > >Makes me wonder if the "2.4MHz" calculation is wrong (after all these years). >The emulation doesn't try to execute code at precisely the same timing as the >real hardware (i.e. it doesn't check the clock after every instruction) but >instead check it every 1000 or maybe 10000 instructions (or maybe cycles, I >dont' remember which) and then adds micro-sleeps to keep the *average* speed >at the desired setting. > >So there could be an error in either: > >1. The calculation logic for those little micro-sleeps >2. The accumulation of how long each OPCODE takes in terms of cycles (this is >what is used to determine how many cycles the emulation has progressed). > >Would need some investigation. > >Ken > >On 2/8/26 10:44 AM, John R. Hogerhuis wrote: >> >> Virtual t has speed settings. And Ken worked to make it cycle accurate. >> >> You might have it on (host) CPU friendly which is not the most cycle >> accurate, I think. >> >> My own emulator CoudT is definitely not cycle accurate being written in >> JavaScript and running in the browser. Though it does have some code to put >> it in the ballpark. >> >> -- John.
