ANN: Severe bug in VICE-1.17 [message #141680] |
Mon, 07 November 2005 03:39 |
Andreas Matthies
Messages: 16 Registered: October 2003
Karma: 0
|
Junior Member |
|
|
Hi.
We have to confirm that theres a severe bug in VICE-1.17. The emulator
will crash after ~2^32 machine cycles which is about 72 minutes running
x64 at 100% or even earlier if you use warp mode.
Every platform is affected so we will release a fixes version 1.18
soon.
I apologize for any trouble this bug is responsible for.
- Andreas
|
|
|
Re: ANN: Severe bug in VICE-1.17 [message #141683 is a reply to message #141680] |
Tue, 08 November 2005 05:13 |
A Grosz
Messages: 37 Registered: April 2005
Karma: 0
|
Member |
|
|
Hi,
That's a bummer. I presume it's because of the clock counter overflow?
Any reason why you are not using 64 bit integers for the clocking? They
virtually can not overflow (well, one must be patient :-) ) and most
compilers do support it.
Regards,
Attila
|
|
|
Re: ANN: Severe bug in VICE-1.17 [message #141685 is a reply to message #141683] |
Wed, 09 November 2005 04:10 |
Spiro Trikaliotis
Messages: 380 Registered: December 2011
Karma: 0
|
Senior Member |
|
|
Hello Attila,
A Grosz <grosza@hotmail.com> schrieb:
> That's a bummer. I presume it's because of the clock counter overflow?
> Any reason why you are not using 64 bit integers for the clocking?
> They virtually can not overflow (well, one must be patient :-) ) and
> most compilers do support it.
I have not looked to deep into that problem. Anyway, as I understand it,
some code is executed when it should not have been executed (the device
is disabled). Now, while executing this code path, if an overflow
occurs, some division is done, which divides through an uninitialized
variable. This variable is uninitialized as the device is not used
(deactivated).
Now, this variable happens to be zero -> BUMM
Thus, the problem is not the timer itself, but executing code that
should not be executed (and using uninitialized variables). Using a 64
bit integer for the clocking would not necessarily help here.
At least, this is how I understand it.
Regards,
Spiro.
--
Spiro R. Trikaliotis
http://www.trikaliotis.net/
http://cbm4win.sf.net/
|
|
|
Re: ANN: Severe bug in VICE-1.17 [message #141687 is a reply to message #141685] |
Wed, 09 November 2005 08:22 |
A Grosz
Messages: 37 Registered: April 2005
Karma: 0
|
Member |
|
|
Hi Spiro,
Thanks for the explanation. My question was more of a general one
though, like is there a specific reason for using 32 bit integers in
VICE, instead of 64 bit ones. The latter is also natively supported by
most compilers. They have a little more overhead (at least on 32 bit
systems ;-) ) but using 64 bits one could totally avoid this whole
clock overflow handling misery (or am I missing something here?) And
even when running 20x faster than the original, it'll take a few
thousand years before a 64 bit clock counter overflows. Most likely the
hw would be defect by then anyway :-)
Attila
|
|
|
Re: ANN: Severe bug in VICE-1.17 [message #141688 is a reply to message #141687] |
Wed, 09 November 2005 11:18 |
Patryk 'Silver Dream
Messages: 737 Registered: July 2003
Karma: 0
|
Senior Member |
|
|
A Grosz wrote:
> Hi Spiro,
>
> Thanks for the explanation. My question was more of a general one
> though, like is there a specific reason for using 32 bit integers in
> VICE, instead of 64 bit ones. The latter is also natively supported by
> most compilers. They have a little more overhead (at least on 32 bit
> systems ;-) ) but using 64 bits one could totally avoid this whole
> clock overflow handling misery (or am I missing something here?)
It would have caused the bug not to be discovered that quickly :-)
|
|
|
Re: ANN: Severe bug in VICE-1.17 [message #141689 is a reply to message #141687] |
Thu, 10 November 2005 17:32 |
BlackJack
Messages: 152 Registered: December 2004
Karma: 0
|
Senior Member |
|
|
On Wed, 09 Nov 2005 05:22:29 -0800, A Grosz wrote:
> Thanks for the explanation. My question was more of a general one
> though, like is there a specific reason for using 32 bit integers in
> VICE, instead of 64 bit ones.
This slows down the emulator on 32 bit machines without making the
emulation more perfect or better. So what's the specific reason to use
64 bit integers?
I think the majority of PCs out there are still 32 bit ones and I don't
see why we should lose speed without getting anything in return.
Ciao,
Marc 'BlackJack' Rintsch
|
|
|
Re: ANN: Severe bug in VICE-1.17 [message #141690 is a reply to message #141689] |
Fri, 11 November 2005 15:20 |
A Grosz
Messages: 37 Registered: April 2005
Karma: 0
|
Member |
|
|
BlackJack wrote:
> This slows down the emulator on 32 bit machines without making the
> emulation more perfect or better. So what's the specific reason to use
> 64 bit integers?
As I wrote earlier, you won't have to check for integer overflow
and walk thru all the events in the queue to update them. I was not
saying it would be quicker, although it save quite a lot of codework
(and potentiel bugs)
> I think the majority of PCs out there are still 32 bit ones and I don't
> see why we should lose speed without getting anything in return.
See above.
Attila
|
|
|
Re: ANN: Severe bug in VICE-1.17 [message #141691 is a reply to message #141680] |
Fri, 11 November 2005 22:02 |
Martin Brunner
Messages: 147 Registered: January 2012
Karma: 0
|
Senior Member |
|
|
Andreas Matthies schrieb:
> We have to confirm that theres a severe bug in VICE-1.17. The emulator
> will crash after ~2^32 machine cycles which is about 72 minutes running
> x64 at 100% or even earlier if you use warp mode.
Ah... That's why I wasn't able to record the game for more than about 69
minutes. Ok, seems I have to wait for the next release.
|
|
|
|
Re: ANN: Severe bug in VICE-1.17 [message #141693 is a reply to message #141692] |
Sat, 12 November 2005 08:32 |
A Grosz
Messages: 37 Registered: April 2005
Karma: 0
|
Member |
|
|
Hi,
Spiro Trikaliotis wrote:
> Why would you want to go "through all events in the queue"?
Yeah, I just see it's not really necessary, you are right. It really
depends on the implementation. I have no idea how it's done in VICE,
though, I thought you might have to update the event clock timings on
your event queue list. But one can avoid it and only check for the clock
overflow the rest goes "automatically".
Well I have always used 64 bit clock counters, hence never really
bothered about 32 bit integer overflow ;-)
Regards,
Attila
|
|
|
Re: ANN: Severe bug in VICE-1.17 [message #141695 is a reply to message #141680] |
Mon, 14 November 2005 14:35 |
|
Originally posted by: LocalH
Andreas Matthies wrote:
> Hi.
>
> We have to confirm that theres a severe bug in VICE-1.17. The emulator
> will crash after ~2^32 machine cycles which is about 72 minutes running
> x64 at 100% or even earlier if you use warp mode.
>
> Every platform is affected so we will release a fixes version 1.18
> soon.
>
> I apologize for any trouble this bug is responsible for.
>
> - Andreas
>
> From - Thu
I want to bring to your attention another bug (not so severe) with
recording video output of NTSC mode. In PAL mode, output is correctly at
50fps (notwithstanding any possible minor audio skew which can be easily
corrected for). However, in NTSC mode, output is still 50fps, which
means that one out of every 6 frames are being dropped. As I'd like to
put some C64 demos on DVD, I'd like the capability to dump both NTSC and
PAL video, and right now only PAL is possible with any sort of quality.
|
|
|