[CIG-SEISMO] patch for precision issue with sign() with gfortran 4.6
Brad Aagaard
baagaard at usgs.gov
Thu Feb 9 17:26:43 PST 2012
SPECFEM developers,
In order to build SPECFEM3D from the SVN repository with gfortran 4.6, I
needed to explicitly specify double precision for the 1.0 argument in
the call to the sign() function (see below). My config.log is attached
in case that helps identify whether this is an actual issue or user error.
Thanks,
Brad
Index: src/specfem3D/setup_sources_receivers.f90
===================================================================
--- src/specfem3D/setup_sources_receivers.f90 (revision 19607)
+++ src/specfem3D/setup_sources_receivers.f90 (working copy)
@@ -570,7 +570,7 @@
! determines factor +/-1 depending on sign of moment tensor
! (see e.g. Krebs et al., 2009. Fast full-wavefield seismic
inversion using encoded sources,
! Geophysics, 74 (6), WCC177-WCC188.)
- pm1_source_encoding(isource) = sign(1.0,Mxx(isource))
+ pm1_source_encoding(isource) = sign(1.0d0,Mxx(isource))
! source array interpolated on all element gll points (only
used for non point sources)
call
compute_arrays_source_acoustic(xi_source(isource),eta_source(isource),gamma_source(isource),&
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: config.log
Url: http://geodynamics.org/pipermail/cig-seismo/attachments/20120209/7f7c2be9/attachment-0001.txt
More information about the CIG-SEISMO
mailing list