Re: [hatari-devel] WinUAE CPU core cycle-exact option impact for 68000

["Christian Zietz" <czietz@xxxxxxx>]

Remark: I didn't say that I "get the same result in Hatari, independent of the cycle-exact setting." I simply didn't test with cycle exact off. But with cycle exact on I get the same result as on real HW (39 loops), which is therefore correct. But Thorsten doesn't get the correct result.

 

I'm using Hatari 2.2.1, official download for Windows from the website.

 

Regards

Christian

 

 

Gesendet: Freitag, 17. Januar 2020 um 11:53 Uhr
Von: "Thorsten Otto" <admin@xxxxxxxxxxx>
An: hatari-devel@xxxxxxxxxxxxxxxxxxx
Betreff: Re: [hatari-devel] WinUAE CPU core cycle-exact option impact for 68000

While trying to do some benchmarks for EmuTOS, i currently have a strange problem with cycle-exact mode. The attached program gives different results on my machine: 39 loops when cycle-exact is off, 30 when cycle-exact is on. It is a simple test that just does a vr_recfl() on the whole screen, running for 5 seconds. According to Christian, 39 is the correct value on real HW, and on his machine he gets the same result in Hatari, independent of the cycle-exact setting.

 

The tests were done with a current snapshot version of 256k EmuTOS. The behaviour is the same with current Hatari, and official 2.2.1 release. Any idea what is causing this? I also tried the windows version, but that just crashes with a segmentation fault in VirtualBox...