[AMBER] AMD GPU executable reads different gb parameters

From: Wong, Sergio E. via AMBER <amber.ambermd.org>
Date: Sun, 23 Mar 2025 08:06:30 +0000

Hi;

    I’m compiling amber 24 on an AMD MI300A GPU system. The compilation was successful ( it finished ). However, I noticed something odd while running the tests.

   The following line is part of the test diff for irest1_ntt0_igb8_ntc2 SPFP. The same result is true for the DPFP executable.

   possible FAILURE: check irest1_ntt0_igb8_ntc2.out.dif

51c51
< gbalphaP = 1.00000, gbbetaP = 0.80000, gbgammaP = 4.85000
> gbalphaP = 0.41836, gbbetaP = 0.29005, gbgammaP = 0.10642


As far as I can tell, the gbalphaP, etc are part of the input/parmtop file . Is the difference due to a problem reading the prmtop file or is it a problem writing the mdout file?

Note, when I compiled the CPU version of amber24, there is no error involving gbalphaP , gbbetaP and gbgammaP. Can anyone shed any light?

Thank you in advance;

-Sergio

_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sun Mar 23 2025 - 01:30:04 PDT
Custom Search