[mesa-users] mesa release 3708

Bill Paxton paxton at kitp.ucsb.edu
Mon Oct 31 10:28:37 EDT 2011


The new mesa release has some very lovely things in it!   
Special thanks go to Rich Townsend and Joergen Christensen-Dalsgaard.
Here are the highlights of the changes since last time.


add mesa sdk
	Rich Townsend got tired of hearing me complain about the problems with mesa installations,
	so he's done something about it!
	it is optional of course
		if you want to do your installations the hard way, you can still fight with compilers etc yourself.
	but now there is an alternative to fighting -- switch by these simple steps:
		unpack the mesa sdk, 
		add a few lines to your startup script,
	no more worrying about finding the right compiler and no more hassle getting the right version of pgplot installed.
	this is how I'm going to work from now on, so the sdk install will be what I've used and tested.
	the sdk includes the most recent gfortran, and that's now my compiler of choice.
		the performance gap between ifort and gfortran seems to be vanishing quickly,
		and the convenience of having gfortran ready-to-go in the sdk is wonderful!!!!!
	the sdk also includes a pre-built pgplot for mesa, plus prebuilt hdf5 and se (for Falk and friends)
	there is a version for Mac and a version for linux (no, we still don't do Windows)

Here's how I made the switch to the mesasdk on my Mac
	 (customize this as you wish -- for example, the mesasdk directory can go anywhere you like)

unpack mesasdk-x86_64-darwin-20111029.tar.gz
put the resulting mesasdk directory in my top level directory (/Users/bpaxton)

add lines to my .tcshrc file
setenv MESASDK_ROOT /Users/bpaxton/mesasdk
source ${MESASDK_ROOT}/bin/mesasdk_init.csh

(btw: make sure you are setting OMP_NUM_THREADS too!)

copy mesa/utils/makefile_header.mesasdk to mesa/utils/makefile_header

start up a new shell and type 'which gfortran' to make sure it finds the new mesasdk

do ./clean and ./install for mesa to check it out

add adipls (and remove pulse package since it is now obsolete)
	many thanks to Joergen Christensen-Dalsgaard for making the 2010 version of ADIPLS available in mesa
	it can be used "stand-alone" with his usual control files
	or it can be called directly from star between steps (see the new astero package)

add astero package
	tools for using star + adipls so can evolve stars and test the oscillation modes on the fly
	astero/test has an example -- match data for HD49385 (based on results in S. Deheuvels & E. Michel, 2011)
	with option for using BOBYQA to search parameter space for good match to spectral and seismic observations
	(BOBYQA is "bounded optimization by quadratic approximation" -- it is available in num of course)

add other_mlt.f in star/lib
	so you can intercept calls from star to mlt and do-it-yourself -- be famous!  invent a replacement for MLT!

add solar_calibration in test_suite 
	with options for using BOBYQA optimization to search for parameters giving best solar model.
	btw: I seem to have fixed so many bugs that now our solar model isn't as good as it was last year!  ;-(
	it is still okay, but let me know if you find a better match by tweaking parameters!

add radiation turbulence and turbulent diffusion (see star_defaults.dek and test_suite/1.5M_with_diffusion)

fix bug in net -- using incorrect Q for weaklib reactions (ouch)

add qsort to num lib

fix memory leak in mesh adjust

add distance in HR diagram timestep controls and pgstar file output controls (useful for making movies)

add the other kind of Nieuwenhuijzen wind

add newton_decsol_switch so automatically change from dense to sparse blocks when go to large nets

add logg-Teff target box for pgstar (for asteroseismology)

add mode propagation plot for pgstar (also for asteroseismology)

add by_logT as xaxis option for pgstar

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20111031/f7299d6f/attachment.html>

More information about the Mesa-users mailing list