[uae] Re: MacOSX Intel supports JIT?
- From: Richard Drummond <evilrich@xxxxxxxxxxxxxx>
- To: uae@xxxxxxxxxxxxx
- Date: Fri, 23 Mar 2007 00:21:39 -0400
Hi Anders
On Wednesday 21 March 2007 06:26, Anders Hasselqvist wrote:
> I just tried the CVS version of E-UAE on my MacBook Pro.
> When I compile without JIT enabled everything seems to work
> ok (although a bit slow). However when I enable JIT it crashes
> during the startup procedure. If I disable the harddrive it
> starts to the Kickstart screen (Floppy screen).
>
> The last output I see in my terminal is:
> JIT: Building compiler function table.
> JIT: Supposedly 42346 compileable opcodes.
> JIT: Allocated 1024 KB translation cache.
> SDL sound driver found and configured for 16 bits at 44100 Hz, buffer is
> 100 ms (17640 bytes).
> reset at 0
> PAL mode, 50Hz (h=227 v=312)
> chipmem cleared
> framewait overflow
> SERIAL: period=372, baud=9600, hsyncs=16 PC=f8018a
>
> Program received signal EXC_BAD_ACCESS, Could not access memory.
> Reason: KERN_INVALID_ADDRESS at address: 0x2e0b5ffc
> 0x17362860 in ?? ()
> (gdb) bt
> #0 0x17362860 in ?? ()
Can you check what your comp_trust*= options are set to in your config file
(if they are set at all). I'm just looking at the code right now, and it
seems they default to 'direct', which probably ain't a great idea, although
I've never noticed it to be a problem before. However, see my other mail on
this topic. I suspect on fast machines E-UAE doesn't back these out
to 'indirect' quickly enough.
> I tried checking the backtrace to see where the crash is but it seems like
> the stack is trashed.
You probably want to pass 'CFLAGS="-ggdb"' to configure to be able to get
worthwhile backtraces. Of course, if the crash occurs in code generated by
the JIT - which seems likely in this case - then traditional debugging tools
won't be of much use.
> Is JIT supported on MacOSX Intel?
It's not supported in the sense that I don't have an MacIntel box to test it
on, and so I don't provide binaries. But others do and claim it works. There
appear to be problems, but nobody has provided me with solid bug reports on
what these possible problems might be.
Cheers,
Rich
--
Richard Drummond
Web: http://www.rcdrummond.net/
Mail: mailto:evilrich@xxxxxxxxxxxxxx
Other related posts:
