cuda: update python interface to support GPU off-loading through SHT_ALLOW_GPU

README.md

@@ -68,7 +68,7 @@ DOCUMENTATION:
 
 CHANGE LOG:
 -----------
-* v2.10  (28 Feb 2018)
+* v3.0  (28 Feb 2018)
 	- support for nvidia Kepler & Pascal GPU with `--enable-cuda` (including automatic offload).
 	- Java JNI wrapper (beta, see `shtns_jni` folder -- thanks to Julien Pierret).
 

configure

 #! /bin/sh
 # Guess values for system-dependent variables and create Makefiles.
-# Generated by GNU Autoconf 2.69 for SHTns 2.9.
+# Generated by GNU Autoconf 2.69 for SHTns 3.0.
 #
 #
 # Copyright (C) 1992-1996, 1998-2012 Free Software Foundation, Inc.
@@ -577,8 +577,8 @@ MAKEFLAGS=
 # Identity of this package.
 PACKAGE_NAME='SHTns'
 PACKAGE_TARNAME='shtns'
-PACKAGE_VERSION='2.9'
-PACKAGE_STRING='SHTns 2.9'
+PACKAGE_VERSION='3.0'
+PACKAGE_STRING='SHTns 3.0'
 PACKAGE_BUGREPORT=''
 PACKAGE_URL='https://bitbucket.org/nschaeff/shtns'
 
@@ -1251,7 +1251,7 @@ if test "$ac_init_help" = "long"; then
   # Omit some internal or obsolete options to make the list less imposing.
   # This message is too long to be a string in the A/UX 3.1 sh.
   cat <<_ACEOF
-\`configure' configures SHTns 2.9 to adapt to many kinds of systems.
+\`configure' configures SHTns 3.0 to adapt to many kinds of systems.
 
 Usage: $0 [OPTION]... [VAR=VALUE]...
 
@@ -1312,7 +1312,7 @@ fi
 
 if test -n "$ac_init_help"; then
   case $ac_init_help in
-     short | recursive ) echo "Configuration of SHTns 2.9:";;
+     short | recursive ) echo "Configuration of SHTns 3.0:";;
    esac
   cat <<\_ACEOF
 
@@ -1420,7 +1420,7 @@ fi
 test -n "$ac_init_help" && exit $ac_status
 if $ac_init_version; then
   cat <<\_ACEOF
-SHTns configure 2.9
+SHTns configure 3.0
 generated by GNU Autoconf 2.69
 
 Copyright (C) 2012 Free Software Foundation, Inc.
@@ -1877,7 +1877,7 @@ cat >config.log <<_ACEOF
 This file contains any messages produced by compilers while
 running configure, to aid debugging if configure makes a mistake.
 
-It was created by SHTns $as_me 2.9, which was
+It was created by SHTns $as_me 3.0, which was
 generated by GNU Autoconf 2.69.  Invocation command line was
 
   $ $0 $@
@@ -5412,7 +5412,7 @@ cat >>$CONFIG_STATUS <<\_ACEOF || ac_write_fail=1
 # report actual input values of CONFIG_FILES etc. instead of their
 # values after options handling.
 ac_log="
-This file was extended by SHTns $as_me 2.9, which was
+This file was extended by SHTns $as_me 3.0, which was
 generated by GNU Autoconf 2.69.  Invocation command line was
 
   CONFIG_FILES    = $CONFIG_FILES
@@ -5475,7 +5475,7 @@ _ACEOF
 cat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1
 ac_cs_config="`$as_echo "$ac_configure_args" | sed 's/^ //; s/[\\""\`\$]/\\\\&/g'`"
 ac_cs_version="\\
-SHTns config.status 2.9
+SHTns config.status 3.0
 configured by $0, generated by GNU Autoconf 2.69,
   with options \\"\$ac_cs_config\\"
 

configure.ac

@@ -2,7 +2,7 @@
 # Process this file with autoconf to produce a configure script.
 
 AC_PREREQ([2.62])
-AC_INIT([SHTns],[2.9],[],[shtns],[https://bitbucket.org/nschaeff/shtns])
+AC_INIT([SHTns],[3.0],[],[shtns],[https://bitbucket.org/nschaeff/shtns])
 AC_LANG([C])
 AC_CONFIG_SRCDIR([sht_init.c])
 AC_CONFIG_HEADERS([sht_config.h])

shtns.py

@@ -118,6 +118,7 @@ SHT_PHI_CONTIGUOUS = _shtns.SHT_PHI_CONTIGUOUS
 SHT_SOUTH_POLE_FIRST = _shtns.SHT_SOUTH_POLE_FIRST
 SHT_SCALAR_ONLY = _shtns.SHT_SCALAR_ONLY
 SHT_LOAD_SAVE_CFG = _shtns.SHT_LOAD_SAVE_CFG
+SHT_ALLOW_GPU = _shtns.SHT_ALLOW_GPU
 class sht(_object):
     """Proxy of C shtns_info struct."""
 
@@ -147,6 +148,9 @@ class sht(_object):
     __swig_getmethods__["nspat"] = _shtns.sht_nspat_get
     if _newclass:
         nspat = _swig_property(_shtns.sht_nspat_get)
+    __swig_getmethods__["nlm_cplx"] = _shtns.sht_nlm_cplx_get
+    if _newclass:
+        nlm_cplx = _swig_property(_shtns.sht_nlm_cplx_get)
 
     def __init__(self, lmax, mmax=-1, mres=1, norm=sht_orthonormal, nthreads=0):
         """__init__(shtns_info self, int lmax, int mmax=-1, int mres=1, int norm=sht_orthonormal, int nthreads=0) -> sht"""
@@ -537,6 +541,10 @@ def nlm_calc(lmax, mmax, mres):
     """nlm_calc(long lmax, long mmax, long mres) -> long"""
     return _shtns.nlm_calc(lmax, mmax, mres)
 
+def nlm_cplx_calc(lmax, mmax, mres):
+    """nlm_cplx_calc(long lmax, long mmax, long mres) -> long"""
+    return _shtns.nlm_cplx_calc(lmax, mmax, mres)
+
 def set_verbosity(arg1):
     """set_verbosity(int arg1)"""
     return _shtns.set_verbosity(arg1)

shtns_numpy_wrap.c

@@ -3759,6 +3759,28 @@ fail:
 }
 
 
+SWIGINTERN PyObject *_wrap_sht_nlm_cplx_get(PyObject *SWIGUNUSEDPARM(self), PyObject *args) {
+  PyObject *resultobj = 0;
+  struct shtns_info *arg1 = (struct shtns_info *) 0 ;
+  void *argp1 = 0 ;
+  int res1 = 0 ;
+  PyObject * obj0 = 0 ;
+  unsigned int result;
+  
+  if (!PyArg_ParseTuple(args,(char *)"O:sht_nlm_cplx_get",&obj0)) SWIG_fail;
+  res1 = SWIG_ConvertPtr(obj0, &argp1,SWIGTYPE_p_shtns_info, 0 |  0 );
+  if (!SWIG_IsOK(res1)) {
+    SWIG_exception_fail(SWIG_ArgError(res1), "in method '" "sht_nlm_cplx_get" "', argument " "1"" of type '" "struct shtns_info *""'"); 
+  }
+  arg1 = (struct shtns_info *)(argp1);
+  result = (unsigned int)(unsigned int) ((arg1)->nlm_cplx);
+  resultobj = SWIG_From_unsigned_SS_int((unsigned int)(result));
+  return resultobj;
+fail:
+  return NULL;
+}
+
+
 SWIGINTERN PyObject *_wrap_new_sht(PyObject *SWIGUNUSEDPARM(self), PyObject *args, PyObject *kwargs) {
   PyObject *resultobj = 0;
   int arg1 ;
@@ -5605,6 +5627,46 @@ fail:
 }
 
 
+SWIGINTERN PyObject *_wrap_nlm_cplx_calc(PyObject *SWIGUNUSEDPARM(self), PyObject *args) {
+  PyObject *resultobj = 0;
+  long arg1 ;
+  long arg2 ;
+  long arg3 ;
+  long val1 ;
+  int ecode1 = 0 ;
+  long val2 ;
+  int ecode2 = 0 ;
+  long val3 ;
+  int ecode3 = 0 ;
+  PyObject * obj0 = 0 ;
+  PyObject * obj1 = 0 ;
+  PyObject * obj2 = 0 ;
+  long result;
+  
+  if (!PyArg_ParseTuple(args,(char *)"OOO:nlm_cplx_calc",&obj0,&obj1,&obj2)) SWIG_fail;
+  ecode1 = SWIG_AsVal_long(obj0, &val1);
+  if (!SWIG_IsOK(ecode1)) {
+    SWIG_exception_fail(SWIG_ArgError(ecode1), "in method '" "nlm_cplx_calc" "', argument " "1"" of type '" "long""'");
+  } 
+  arg1 = (long)(val1);
+  ecode2 = SWIG_AsVal_long(obj1, &val2);
+  if (!SWIG_IsOK(ecode2)) {
+    SWIG_exception_fail(SWIG_ArgError(ecode2), "in method '" "nlm_cplx_calc" "', argument " "2"" of type '" "long""'");
+  } 
+  arg2 = (long)(val2);
+  ecode3 = SWIG_AsVal_long(obj2, &val3);
+  if (!SWIG_IsOK(ecode3)) {
+    SWIG_exception_fail(SWIG_ArgError(ecode3), "in method '" "nlm_cplx_calc" "', argument " "3"" of type '" "long""'");
+  } 
+  arg3 = (long)(val3);
+  result = (long)nlm_cplx_calc(arg1,arg2,arg3);
+  resultobj = SWIG_From_long((long)(result));
+  return resultobj;
+fail:
+  return NULL;
+}
+
+
 SWIGINTERN PyObject *_wrap_set_verbosity(PyObject *SWIGUNUSEDPARM(self), PyObject *args) {
   PyObject *resultobj = 0;
   int arg1 ;
@@ -5647,6 +5709,7 @@ static PyMethodDef SwigMethods[] = {
 	 { (char *)"sht_nphi_get", _wrap_sht_nphi_get, METH_VARARGS, (char *)"sht_nphi_get(sht self) -> unsigned short const"},
 	 { (char *)"sht_nlat_get", _wrap_sht_nlat_get, METH_VARARGS, (char *)"sht_nlat_get(sht self) -> unsigned short const"},
 	 { (char *)"sht_nspat_get", _wrap_sht_nspat_get, METH_VARARGS, (char *)"sht_nspat_get(sht self) -> unsigned int const"},
+	 { (char *)"sht_nlm_cplx_get", _wrap_sht_nlm_cplx_get, METH_VARARGS, (char *)"sht_nlm_cplx_get(sht self) -> unsigned int const"},
 	 { (char *)"new_sht", (PyCFunction) _wrap_new_sht, METH_VARARGS | METH_KEYWORDS, (char *)"new_sht(int lmax, int mmax=-1, int mres=1, int norm=sht_orthonormal, int nthreads=0) -> sht"},
 	 { (char *)"delete_sht", _wrap_delete_sht, METH_VARARGS, (char *)"delete_sht(sht self)"},
 	 { (char *)"sht_set_grid", (PyCFunction) _wrap_sht_set_grid, METH_VARARGS | METH_KEYWORDS, (char *)"sht_set_grid(sht self, int nlat=0, int nphi=0, int flags=sht_quick_init, double polar_opt=1.0e-8, int nl_order=1)"},
@@ -5691,6 +5754,7 @@ static PyMethodDef SwigMethods[] = {
 	 { (char *)"sht_SHqst_to_spat_m", _wrap_sht_SHqst_to_spat_m, METH_VARARGS, (char *)"sht_SHqst_to_spat_m(sht self, PyObject * Qlm, PyObject * Slm, PyObject * Tlm, PyObject * Vr, PyObject * Vt, PyObject * Vp, PyObject * im)"},
 	 { (char *)"sht_swigregister", sht_swigregister, METH_VARARGS, NULL},
 	 { (char *)"nlm_calc", _wrap_nlm_calc, METH_VARARGS, (char *)"nlm_calc(long lmax, long mmax, long mres) -> long"},
+	 { (char *)"nlm_cplx_calc", _wrap_nlm_cplx_calc, METH_VARARGS, (char *)"nlm_cplx_calc(long lmax, long mmax, long mres) -> long"},
 	 { (char *)"set_verbosity", _wrap_set_verbosity, METH_VARARGS, (char *)"set_verbosity(int arg1)"},
 	 { (char *)"print_version", _wrap_print_version, METH_VARARGS, (char *)"print_version()"},
 	 { NULL, NULL, 0, NULL }
@@ -6435,6 +6499,7 @@ SWIG_init(void) {
   SWIG_Python_SetConstant(d, "SHT_SOUTH_POLE_FIRST",SWIG_From_int((int)((256*32))));
   SWIG_Python_SetConstant(d, "SHT_SCALAR_ONLY",SWIG_From_int((int)((256*16))));
   SWIG_Python_SetConstant(d, "SHT_LOAD_SAVE_CFG",SWIG_From_int((int)((256*64))));
+  SWIG_Python_SetConstant(d, "SHT_ALLOW_GPU",SWIG_From_int((int)((256*128))));
 #if PY_VERSION_HEX >= 0x03000000
   return m;
 #else