[ADMB Users] Help test admb-11.2

[Johnoel Ancheta]

Hi all,

I need to get some feedback.  Can someone test out the

prerelease?

http://www.admb-project.org/downloads/admb-11.2pre

If there is no issues, the admb-11.2 will release next week.

Johnoel

[John Sibert]

I ate my own dog food from the head of the repository. Everything seems
to work.

Still can't get my head around how to utilize pad() though.

Thanks,
John

John Sibert
Emeritus Researcher, SOEST
University of Hawaii at Manoa
Honolulu HI (GMT-10)
808-294-3842 (mobile)

Visit the ADMB project http://admb-project.org/

> _______________________________________________
> Developers mailing list
> Devel...@admb-project.org
> http://lists.admb-project.org/mailman/listinfo/developers

_______________________________________________
Users mailing list
Us...@admb-project.org
http://lists.admb-project.org/mailman/listinfo/users

[Johnoel Ancheta]

Hi Cole,

The main core objective is getting more users developing and running models with ADMB.

Building models (especially different types) will help improve the software by tracking down 

issues and defects.  The more users that are building with ADMB will take care of the need

to test for arch/os/compiler combinations.   

Below is a check list for user testing.

[ ] Can you successfully install using the installation procedures?

[ ] Can you successfully build models?

[ ] Can you run the built models?

[ ] Are the output file results (ie, .par, .std...) similar to previous versions?

[ ] Are the runtime speed about the same as previous versions?

[ ] Are the manuals and Readmes up to date?

Thanks for assisting.  Let know know if you need more access to the repository.

Johnoel

On Fri, Dec 12, 2014 at 3:15 PM, Cole Monnahan <mon...@uw.edu> wrote:

Hi Johnoel, I'm not a real developer but could help out. Do you mostly want people with different machines/compilers to make sure ADMB builds, or is it more that our own models still compile and run without change? I guess I'm not sure what's the most useful on my end. Thanks, Cole

[Allan Hicks - NOAA Federal]

Thanks Johnoel (and developers),

I've been working with Visual C++ and I feel that the installation instructions could use some improvement. I am willing to work on that, but I have a couple of questions that should probably be answered before that:

1) Should ADMB continue to support the Visual C++ compiler?    The reason I ask is that Microsoft requires you to create a Visual Studio account to download the free version, the SDK does not contain the compiler anymore, and there the mingw compiler seems to be working pretty well with an easy install.

2) Is there a 64-bit version of ADMB for the Visual C++ compiler that the project will release?   There does not seem to be one on the pre-release site.  I would probably switch to mingw if a Visual C++ 64-bit version is not available, even though Windows does not actually support the 64-bit variables with the declarations that ADMB uses.

If the ADMB-project feels that supporting the Visual C++ compiler is still worthwhile, I will test the install and supply more comments. Otherwise, I'll work on testing the mingw compiler.

Thanks,

Allan

On Fri, Dec 12, 2014 at 4:18 PM, Johnoel Ancheta <joh...@hawaii.edu> wrote:

[Jim Ianelli - NOAA Federal]

It worked great for me yesterday on Windows. Nice work!

On Tuesday, December 16, 2014, Johnoel Ancheta <joh...@hawaii.edu> wrote:

Hi all,

Please email feedback for the admb-11.2pre release.

the final admb-11.2 release should be available this week.

Thanks,

Johnoel

--

James Ianelli
NMFS/NOAA Building 4
7600 Sand Pt Way NE
Seattle WA 98115

 206 526 6510

[CHRIS GRANDIN]

Thanks for your comments Allan,

I wrote these install instructions (http://www.admb-project.org/documentation/install/admb-installation-visual-c) for ADMB 11.1 with VS 2010, SDK 7.1, and .NET 4. Since then there are new versions of all of these and it seems that the information is out of date. I wish I had the time necessary to keep updating these instructions but I just don't. I am actually unsure how many people used them and found them useful. At this point I would have to vote to remove support for VS and only support 64-bit mingw. If we did that, we could focus on making one compiler work well with ADMB and write a single set of detailed install instructions.

Thanks,
Chris

[Tom Wainwright]

Tested on my CentOS system:
  Linux 2.6.32-504.1.3.el6.centos.plus.x86_64
  g++ (GCC) 4.4.7 20120313 (Red Hat 4.4.7-11)

1) admb-11.2pre-centos6-intel2013-64bit.zip installed and cleanly built a few examples and recent projects.

2) I had a boring conference call this afternoon, so also did a source install. Built cleanly (except for a few type-conversion warnings), built and ran "simple" cleanly.

Seems fine. Didn't test any new features.

  Tom Wainwright

On Fri, Dec 12, 2014 at 4:18 PM, Johnoel Ancheta <joh...@hawaii.edu> wrote:

_______________________________________________
Users mailing list
Us...@admb-project.org
http://lists.admb-project.org/mailman/listinfo/users

--

Tom Wainwright

Research Fisheries Biologist

Newport Research Station

Northwest Fisheries Science Center

National Oceanic and Atmospheric Administration

www.nwfsc.noaa.gov

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The contents of this message are mine personally and do not
necessarily reflect any position of the Government or the
National Oceanic and Atmospheric Administration.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

[Tom Wainwright]

Johnoel,

You also asked about runtime speeds, so I did a quick comparison on a random-effects model I'm currently working on. I have three versions of ADMB installed as of today: 11.1 installed from source, 11.2pre installed from the intel 64bit binary, and 11.2pre installed from source.  I built and ran the model 3 times with each compiler, with these results:

1) ADMB 11.1 built from source (g++ 4.4.7):

BUILD:
$ /usr/bin/time admb -r euly
0.67user 0.03system 0:00.73elapsed
0.67user 0.05system 0:00.75elapsed
0.64user 0.03system 0:00.71elapsed

RUN:
$ /usr/bin/time ./euly
7.23user 1.17system 0:08.41elapsed
7.25user 1.17system 0:08.43elapsed
7.29user 1.14system 0:08.44elapsed

2) ADMB 11.2pre prebuilt binary admb-11.2pre-centos6-intel2013-64bit.zip

BUILD:
$ /usr/bin/time admb11.2 -r euly
0.66user 0.05system 0:00.74elapsed 96%CPU
0.66user 0.05system 0:00.74elapsed
0.64user 0.04system 0:00.71elapsed

RUN:
$ /usr/bin/time ./euly
7.60user 1.19system 0:08.79elapsed
7.60user 1.16system 0:08.77elapsed
7.28user 1.17system 0:08.45elapsed

3) ADMB 11.2pre built from source (g++ 4.4.7):

BUILD:
$ /usr/bin/time ~/tmp/admb/admb -r euly
0.64user 0.03system 0:00.70elapsed
0.65user 0.04system 0:00.72elapsed
0.64user 0.05system 0:00.72elapsed

RUN:
$ /usr/bin/time ./euly
6.84user 1.22system 0:08.07elapsed
6.89user 1.20system 0:08.10elapsed
7.22user 1.20system 0:08.43elapsed

All three versions are comparable on build times, but 11.2 built from source seems a bit faster in execution than 11.1 (Hooray!). 11.2pre from binary is slowest.

System:
Linux 2.6.32-504.1.3.el6.centos.plus.x86_64

Intel Xeon E3-1240 v3 (3.40 GHz)

g++ (GCC) 4.4.7 20120313 (Red Hat 4.4.7-11)

  Tom Wainwright

[Johnoel Ancheta]

Thanks Allan, for looking over the docs.  Any feedback and suggestion is greatly appreciated.

Yes, the ADMB project will continue to support as many possible arch/operating system/compilers 

combinations for commercial and open source compilers.  It is a good idea to use multiple compilers

because each has a different tools for improving source code development.  Compilers are another

set of eyes for checking code.

The VC++ distributions on the downloads page are 64Bit distributions.  It has been added to the 

description.

Cheers,

Johnoel 

[Johnoel Ancheta]

Hi all,

Please email feedback for the admb-11.2pre release.

the final admb-11.2 release should be available this week.

Thanks,

Johnoel

[Johnoel Ancheta]

Hi Tom,

Thank you for testing and confirming the new version is faster for your model.

It is greatly appreciated.

Cheers,

Johnoel