[gmx-users] Re: about crashes and seg fault on opls meoh under OSX

[Erik Lindahl]

ioannis.nicolis at free.fr wrote:

>Hello again,
>
>Could the bugs of Apple gcc3 you talk about be related to the crashes I 
>experience with the methanol box that I have send you?
>I used the box for another simulation (3 small molecules solvated) and I 
>had again weird results:
>The first time I had a "pressure scaling more than 1%" stop after about 
>217000 steps of 1fs each, while the pressure and density where normal.
>The second time I did an energy minimisation of the system and 
>launched again the dynamics. It stopped after 228590 steps with 
>"Segmentation fault" and no other message.
>
>  
>
>>There are a couple of bugs in the Apple version of gcc3. I've worked
>>around these and significantly improved the altivec innerloops in
>>the prerelease version available at
>>    
>>
>
>If you think is relevant I can download and try the beta version you 
>mention.
>
You might want to give it a try (I plan to look at the files some time 
next week); the
gcc3 bugfixes isn't the only new feature, I've rewritten large parts of 
the Altivec code
too, so if there was any bug there (didn't found any as far as I know) 
it might be
fixed.

Cheers,

Erik