Template for elpa_impl

Makefile.am

@@ -85,7 +85,8 @@ EXTRA_libelpa@SUFFIX@_private_la_DEPENDENCIES = \
   src/elpa1/elpa_reduce_add_vectors.F90 \
   src/elpa1/elpa_transpose_vectors.F90 \
   src/elpa_api_math_template.F90 \
-  src/elpa_impl_template.F90 \
+  src/elpa_impl_math_template.F90 \
+  src/elpa_impl_generalized_transform_template.F90 \
   src/elpa1/elpa1_compute_template.F90 \
   src/elpa2/elpa2_compute_real_template.F90 \
   src/elpa2/elpa2_compute_complex_template.F90 \
@@ -659,7 +660,8 @@ EXTRA_DIST = \
   nvcc_wrap \
   src/GPU/cuUtils_template.cu \
   src/elpa_api_math_template.F90 \
-  src/elpa_impl_template.F90 \
+  src/elpa_impl_math_template.F90 \
+  src/elpa_impl_generalized_transform.F90 \
   src/elpa1/elpa1_compute_template.F90 \
   src/elpa1/elpa1_merge_systems_real_template.F90 \
   src/elpa1/elpa1_solve_tridi_real_template.F90 \

configure.ac

@@ -31,7 +31,7 @@ AC_SUBST([ELPA_SO_VERSION], [11:0:0])
 
 # API Version
 AC_DEFINE([EARLIEST_API_VERSION], [20170403], [Earliest supported ELPA API version])
-AC_DEFINE([CURRENT_API_VERSION], [20171201], [Current ELPA API version])
+AC_DEFINE([CURRENT_API_VERSION], [20180301], [Current ELPA API version])
 # Autotune Version
 AC_DEFINE([EARLIEST_AUTOTUNE_VERSION], [20171201], [Earliest ELPA API version, which supports autotuning])
 

generated_headers.am

@@ -18,6 +18,7 @@ config-f90.h: config.h
 
 generated_headers += elpa/elpa_generated.h
 elpa/elpa_generated.h: $(top_srcdir)/src/elpa_impl.F90 \
+                       $(top_srcdir)/src/elpa_impl_math_template.F90 \
                        $(top_srcdir)/src/elpa_api.F90 | elpa
 	@rm -f $@
 	$(call extract_interface,!c>)

src/elpa_api.F90

@@ -513,14 +513,12 @@ module elpa_api
 #undef REALCASE
 #undef DOUBLE_PRECISION
 
-#ifdef WANT_SINGLE_PRECISION_REAL
 #define REALCASE 1
 #define SINGLE_PRECISION 1
 #include "general/precision_macros.h"
 #include "elpa_api_math_template.F90"
 #undef REALCASE
 #undef SINGLE_PRECISION
-#endif /* WANT_SINGLE_PRECISION_REAL */
 
 #define COMPLEXCASE 1
 #define DOUBLE_PRECISION 1
@@ -529,14 +527,12 @@ module elpa_api
 #undef DOUBLE_PRECISION
 #undef COMPLEXCASE
 
-#ifdef WANT_SINGLE_PRECISION_COMPLEX
 #define COMPLEXCASE 1
 #define SINGLE_PRECISION
 #include "general/precision_macros.h"
 #include "elpa_api_math_template.F90"
 #undef COMPLEXCASE
 #undef SINGLE_PRECISION
-#endif /* WANT_SINGLE_PRECISION_COMPLEX */
 
 ! end of math routines
 

src/elpa_impl_template.F90 → src/elpa_impl_generalized_transform_template.F90

-    subroutine elpa_transform_generalized_&
+   subroutine elpa_transform_generalized_&
             &ELPA_IMPL_SUFFIX&
             &(self, a, b, is_already_decomposed, error)
         implicit none
@@ -62,6 +62,7 @@
          &trmm("L", "U", BLAS_TRANS_OR_CONJ, "N", self%na, self%na, &
                ONE, b, self%na, a, self%na)
 #endif
+
      call self%timer_stop("scalapack multiply inv(U)^T * A")
 #endif /* DO_USE_ELPA_HERMITIAN_MULTIPLY */