[hatari-users] Note on hard disk images and GEMDOS drives

Uwe Seimet Uwe.Seimet at seimet.de
Sat Nov 14 07:43:12 CET 2009


Hi,

> Well, an error is always possible, but regarding the mfp timer precision 
> I'm quite sure we only have a 4 or 8 cycles difference with real hardware 
> (which is required to run very precise demo such as Overlander's 3D 
> fullscreen or Paulo Simoes Overscan Demos).

Yes, I also assumed that other applications rely much more on correct
timings than a hard disk driver.

> Perhaps the problem lies in some badly emulated bits or combination of 
> mask/pending/... mfp status ?

This may be possible.

> Just to be sure, try turning off the "patch timer D" option.

Changing this (and other) emulation options did not make any difference.

Let me go a bit more into detail on what I observe when debugging with
this debug setup in hdc.c:

#define HDC_VERBOSE           /* display operations */
#define HDC_REALLY_VERBOSE    /* display command packets */

I run my tests with the HDDRUTIL "Device Check". The device check sends
an INQUIRY command to all ACSI devices from 0 to 7. Essentially I get
this debug output:

Byte 0: 0x72
HDC opcode 0x12 : INQUIRY
Controller: 3
Drive: 4

Byte 0: 0x32
HDC opcode 0x12 : INQUIRY
Controller: 1
Drive: 2

(I have extended the standard debug output by a line displaying the
complete first command byte).

The next time I run the check I get a similar output, but not exactly
the same. Even though the low nibble of the first command byte always is
0x12 the data in the high nibble are different each time, resulting in
senseless data for the controller and the drive number.
Also note that even though I should see 8 INQUIRY commands (one for each
of the 8 possible ACSI devices) the emulation only reports 2 commands.
Sometimes for a complete devices check only one INQUIRY command is
reported, sometimes none at all.

The command bytes somehow seem to get garbled and/or thrown away, but I
wonder how this is possible within an emulation.

Best regards

Uwe



More information about the hatari-users mailing list