[Wrf-users] Re: runtime error: WRFV3 on Intel Mac (ifort w/ gcc)

[drb at atmos.umd.edu]

Hi, Todd:

I ran into the same problem on Intel Mac w/ifort (although I did compile
64-bit). I had to recompile WRF with the following edit to my LDFLAGS:

LDFLAGS         =       -Wl,-stack_addr,0xF10000000
-Wl,-stack_size,0x64000000

I also had to take "unset limits" and "MP_STACK_SIZE" variables out of my
.tcshrc. See if this works.

Debra Baker
University of Maryland


>
> I am having a few problems getting WFRV3 to run on my Mac OS X (Intel)
> system. I am using OS X 10.5.3 with ifort 10.1.014 and gcc 4.0.1. I am
> trying to move to 64 bit, but that has had its own problems so I am
> still creating 32 bit executables.
>
> When compiling for 32 bit executables, the only modification I need to
> make to the WRFV3 distribution is to add $(WRF_SRC_ROOT_DIR)/frame/
> module_domain_type.o to LDFLAGS_LOCAL in the arch/
> configure_new.defaults (as described at WRFBLOG). I am compiling in
> debug mode and using Totalview.
>
> With this setup I get executables. Whether it is in ideal.exe or
> wrf.exe, the executable dies (with a Segmentation Fault) on a call to
> "model_to_grid_config_rec". In this case (from ideal.exe built with ./
> compile em_grav2d_x in serial mode) the calling sequence is:
>
> CALL med_initialdata_output( head_grid , config_flags )
> 	CALL init_domain ( grid )
> 		CALL init_domain_rk( grid &
> 			CALL model_to_grid_config_rec ( grid%id , model_config_rec ,
> config_flags )
>
> If I look at the stack immediately before the call to
> model_to_grid_config_rec I do see a whole host of arrays with bad
> addresses, such as: rqvcuten:                         <Bad address:
> 0xbbf56540> (REAL*4(:,:,:)) (this may or may not be relevant).
>
> My hope is that someone has seen this error before and can provide
> some advice on how to fix it.
>
> Cheers,
> Todd