Changeset 2228 for LMDZ5/trunk/libf/dyn3d_common/coefpoly_m.F90

Timestamp:

Mar 12, 2015, 2:15:18 PM (9 years ago)

Author:

lguez

Message:

Correcting a problem from revision 2218. The type double precision
with option "-fdefault-real-8" of gfortran is promoted to 16-byte
precision and there is no specific procedure in arth with this
precision. Could not add a specific procedure in arth with double
precision because, with ifort, the option "-real-size 64" does not
promote the double precision, so that would make two identical
specific procedures in arth.

In module nrtype, replaced double precision by a parameterized real
kind so that the effective precision does not depend on a compiler
option.

In coefpoly, fxhyp, fyhyp and invert_zoom_x, use the parameterized
real kind defined in nrtype, instead of double precision.

Also, in module nrtype, removed unused derived types sprs2_sp and
sprs2_dp.

File:

• LMDZ5/trunk/libf/dyn3d_common/coefpoly_m.F90 (modified) (1 diff)

LMDZ5/trunk/libf/dyn3d_common/coefpoly_m.F90

@@ -25 +25 @@
     ! a0, a1, a2, a3.
 
-    DOUBLE PRECISION, intent(in):: xf1, xf2, xprim1, xprim2, xtild1, xtild2
-    DOUBLE PRECISION, intent(out):: a0, a1, a2, a3
+    use nrtype, only: k8
+
+    REAL(K8), intent(in):: xf1, xf2, xprim1, xprim2, xtild1, xtild2
+    REAL(K8), intent(out):: a0, a1, a2, a3
 
     ! Local:
-    DOUBLE PRECISION xtil1car, xtil2car, derr, x1x2car
+    REAL(K8) xtil1car, xtil2car, derr, x1x2car
 
     !------------------------------------------------------------