Developers Meeting Notes 3/16/12

6 replies [Last post]
rgore's picture
Offline
Joined: 01/27/2011

At the Developers' Meeting on 3/16/12 we discussed the following:

  • The group discussed the recent workshop at Boulder, CO. The feedback from the returned surveys has been positive thus far.
  • One participant in the workshop worked with Michael Spiegel attempting to build Open MX on a Windows 64-bit machine with access to npsol source code. Unfortunately a cryptic compilation error seemingly unrelated to OpenMx prevented building OpenMx successfully. The group plans to post to the R Mailing List and describe the error to see if any assistance is available.
  • The group discussed Mike Hunter's progress on the LISREL objective function in OpenMx. There is now a working test with LISREL. However, currently the test case does calculate the means, and it's not using the dimnames of the matrices. Mike is working with a new larger test case to implement these features in the upcoming weeks.
  • The group also discussed a proposal put forth by Ryne Estabrook and Tim Brick for a new OpenMx object class mxFitFunction (or more simply mxFit) and accompanying mxModel slot. It was decided the function should not take the location of the mean and covariance as parameters. Also deprecating the MxRAMObjective function to enable a more consistent interface was discussed.
  • Ross is continuing to make progress in implementing the solnp and subnp algorithms in the R package Rsolnp into R.
tbates's picture
Online
Joined: 07/31/2009
Bug for apple developer

Tangential to win compile issues, what specific bug can we submit to Apple about the problem compiling parallel on OSX? Perhaps a few clear replications will prompt change?

mspiegel's picture
Offline
Joined: 07/31/2009
I'd recommend that if you

I'd recommend that if you have a mac, you submit a bug report. Here's a description of the problem: https://plus.google.com/101546077160053841119/posts/9h35WKKqffL.

neale's picture
Offline
Joined: 07/31/2009
gcc 4.6 fixes?

Is it not the case that as one poster in that thread notes, gcc 4.6.1 eliminates the problem?

tbates's picture
Online
Joined: 07/31/2009
But that doesn't ship with Xcode

Vastly better for apple to get this right and maintain it if they can

mspiegel's picture
Offline
Joined: 07/31/2009
I've mentioned this before at

I've mentioned this before at the developer meetings: you're not going to get OpenMP support directly from Apple until a BSD licensed OpenMP runtime library will be available for llvm. Some statistics on the number of GPL packages in OS X as a function of time are available here: http://meta.ath0.com/2012/02/05/apples-great-gpl-purge. And they have no incentive to build their own OpenMP runtime, they want everyone to use Grand Central Dispatch.

mspiegel's picture
Offline
Joined: 07/31/2009
Yes, as we've discovered the

Yes, as we've discovered the OpenMx user must have gcc 4.6.1 installed on their machine to eliminate the problem.