Re: [hatari-devel] Re: 040/060 emulation issue with Q2

[ Thread Index | Date Index | More Archives ]

Ok, this is very strange. I found that disabling a bunch of checkboxes on the SYSTEM page caused the remaining faults to disappear and the project runs properly.

Narrowing them down, it appears that enabling either 'prefetch mode' or 'cycle accurate mode' (or both) will break my project. The other checkboxes don't seem to contribute to the problem.

The main difference between this project and BadMooD is use of the FPU, mainly within C code. I didn't see these issues with BadMooD.

Not sure if that helps make sense of whats going on.


On 3 March 2015 at 19:00, Douglas Little <doug694@xxxxxxxxxxxxxx> wrote:

$0002808a : f23c 4538 4020 0000                fcmp.s    #$40200000,fp2
$00028092 : f29d                               DC.W      $f29d
$000280ce : f201                               DC.W      $f201
*$000280d4 : ff70                               DC.W      $ff70  <--


just to be sure it's not a disassembler problem, could you run hatari with "--disasm uae" and see if you get different results (or more coherent ones) ?  (Hatari can use UAE/WinUAE disassember or the external one)

I'll try that - however the emulator itself is throwing an illegal instruction on that opcode. The disasm is wrong in a number of ways - true (e.g. half of the FPU opcodes are show as data, and some have src/dst swapped), but not sure that can be responsible for an exception?

Anyway I'll see if the other disassembler can make sense of that opcode as you suggest. I tried the online multi-arch disassembler but it can't cope with FPU ops :-z will need to do it by hand otherwise.


Mail converted by MHonArc 2.6.19+