GYRE 2.1.1 Crash with Increasing alpha_

[ehsan]

Hello Rich,
I use GYRE to get the adiabatic frequencies of a secondary clump RGB (3.3Msun, Yc=0.50, f_ov=0.016), and noticed the disorder in mode order. To fix this, I tried to increase alpha_osc and alpha_exp to larger and larger values, where the following crash happened for both of them being set to 50, whereas there was no crash if I set them to 25.

Code: Select all

       0      33      33       0    0.4948668237750915E+02    0.0000000000000000E+00    0.4171699880215338E-13       5
       0      34      34       0    0.5099200365771901E+02    0.0000000000000000E+00    0.1163559792594714E-12       5
       0      35      35       0    0.5249749655743828E+02    0.0000000000000000E+00    0.3385534609589886E-12       5
       0      36      36       0    0.5400344087736774E+02    0.0000000000000000E+00    0.3092865962835469E-12       5
       0      37      37       0    0.5551010868792122E+02    0.0000000000000000E+00    0.2537126239856431E-12       5

Completed mode search; time elapsed:      2.414 s

Discriminant Scan
=================


Completed scan; time elapsed:    221.584 s

Scan Mode Search
================

       l    n_pg     n_p     n_g                 Re(omega)                 Im(omega)                       chi  n_iter

Program received signal SIGBUS: Access to an undefined portion of a memory object.

Backtrace for this error:
#0  0x10e1df092
#1  0x10e1df85e
#2  0x7fff9355ccf9
#3  0x10dd77d77
#4  0x10e3525fc
#5  0x7fff935088be
Bus error: 10

I have attached the model and the inlist.

Is it like increasing the alpha_ parameters not essentially a remedy to fix disorder in mode ordering?

Cheers.
Ehsan.

UPDATE:
I repeated the same experience by setting alpha_osc = 5 and alpha_exp = 2'; it succeeded charmingly, and the modes were perfectly in order. I am going to check period spacing and see how the periods/frequencies behave.

[rhtownsend]

Hello Rich,
I use GYRE to get the adiabatic frequencies of a secondary clump RGB (3.3Msun, Yc=0.50, f_ov=0.016), and noticed the disorder in mode order. To fix this, I tried to increase alpha_osc and alpha_exp to larger and larger values, where the following crash happened for both of them being set to 50, whereas there was no crash if I set them to 25.

Code: Select all

       0      33      33       0    0.4948668237750915E+02    0.0000000000000000E+00    0.4171699880215338E-13       5
       0      34      34       0    0.5099200365771901E+02    0.0000000000000000E+00    0.1163559792594714E-12       5
       0      35      35       0    0.5249749655743828E+02    0.0000000000000000E+00    0.3385534609589886E-12       5
       0      36      36       0    0.5400344087736774E+02    0.0000000000000000E+00    0.3092865962835469E-12       5
       0      37      37       0    0.5551010868792122E+02    0.0000000000000000E+00    0.2537126239856431E-12       5

Completed mode search; time elapsed:      2.414 s

Discriminant Scan
=================


Completed scan; time elapsed:    221.584 s

Scan Mode Search
================

       l    n_pg     n_p     n_g                 Re(omega)                 Im(omega)                       chi  n_iter

Program received signal SIGBUS: Access to an undefined portion of a memory object.

Backtrace for this error:
#0  0x10e1df092
#1  0x10e1df85e
#2  0x7fff9355ccf9
#3  0x10dd77d77
#4  0x10e3525fc
#5  0x7fff935088be
Bus error: 10

I have attached the model and the inlist.

Is it like increasing the alpha_ parameters not essentially a remedy to fix disorder in mode ordering?

Cheers.
Ehsan.

UPDATE:
I repeated the same experience by setting alpha_osc = 5 and alpha_exp = 2'; it succeeded charmingly, and the modes were perfectly in order. I am going to check period spacing and see how the periods/frequencies behave.

Hi Ehsan --

Going beyond alpha ~ 5-10 won't do anything further to fix issues in mode ordering; and it will make the execution much slower, as the grid sizes become very large. Eventually, the grids will be so large that GYRE runs out of memory; I think this is what you've encountered above with alpha = 50.

I'm going to update the documentation explaining this.

cheers,

Rich