coll: Add Rabenseifner's algorithm for Reduce and Allreduce

A component with implementation of R. Rabenseifner's algorithm for Reduce and Allreduce.
This algorithm is a combination of a reduce-scatter implemented with recursive vector halving
and recursive distance doubling, followed either by a gather or an allgather.

Current limitations:
  -- count >= 2^{\floor{\log_2 p}}
  -- commutative operations only
  -- intra-communicators onl

Signed-off-by: Mikhail Kurnosov <[email protected]>

coll/spacc: Modify implementation to use `ompi_coll_base_sendrecv()`

Replace irecv() + isend() + ompi_request_wait() to ompi_coll_base_sendrecv().

Signed-off-by: Mikhail Kurnosov <[email protected]>

Author: mkurnosov

ompi/mca/coll/spacc/Makefile.am

@@ -0,0 +1,35 @@
+#
+# $COPYRIGHT$
+#
+# Additional copyrights may follow
+#
+# $HEADER$
+#
+
+sources = \
+        coll_spacc.h \
+        coll_spacc_component.c \
+        coll_spacc_module.c \
+        coll_spacc_allreduce.c \
+        coll_spacc_reduce.c
+
+# Make the output library in this directory, and name it either
+# mca_<type>_<name>.la (for DSO builds) or libmca_<type>_<name>.la
+# (for static builds).
+
+if MCA_BUILD_ompi_coll_spacc_DSO
+component_noinst =
+component_install = mca_coll_spacc.la
+else
+component_noinst = libmca_coll_spacc.la
+component_install =
+endif
+
+mcacomponentdir = $(ompilibdir)
+mcacomponent_LTLIBRARIES = $(component_install)
+mca_coll_spacc_la_SOURCES = $(sources)
+mca_coll_spacc_la_LDFLAGS = -module -avoid-version
+
+noinst_LTLIBRARIES = $(component_noinst)
+libmca_coll_spacc_la_SOURCES =$(sources)
+libmca_coll_spacc_la_LDFLAGS = -module -avoid-version

ompi/mca/coll/spacc/coll_spacc.h

@@ -0,0 +1,83 @@
+/*
+ * $COPYRIGHT$
+ *
+ * Additional copyrights may follow
+ *
+ * $HEADER$
+ */
+
+#ifndef MCA_COLL_SPACC_EXPORT_H
+#define MCA_COLL_SPACC_EXPORT_H
+
+#include "ompi_config.h"
+
+#include "mpi.h"
+#include "ompi/mca/coll/coll.h"
+
+BEGIN_C_DECLS
+
+/* Globally exported variables */
+extern int ompi_coll_spacc_stream;
+extern int ompi_coll_spacc_priority;
+
+/* API functions */
+
+int mca_coll_spacc_init_query(bool enable_progress_threads,
+                              bool enable_mpi_threads);
+mca_coll_base_module_t
+*mca_coll_spacc_comm_query(struct ompi_communicator_t *comm, int *priority);
+
+int mca_coll_spacc_module_enable(mca_coll_base_module_t *module,
+                                 struct ompi_communicator_t *comm);
+
+int mca_coll_spacc_allreduce_intra_redscat_allgather(
+    const void *sbuf, void *rbuf, int count, struct ompi_datatype_t *dtype,
+    struct ompi_op_t *op, struct ompi_communicator_t *comm,
+    mca_coll_base_module_t *module);
+
+int mca_coll_spacc_reduce_intra_redscat_gather(
+    const void *sbuf, void *rbuf, int count, struct ompi_datatype_t *dtype,
+    struct ompi_op_t *op, int root, struct ompi_communicator_t *comm,
+    mca_coll_base_module_t *module);
+
+/*
+ * coll API functions
+ */
+
+/* API functions */
+
+int ompi_coll_spacc_init_query(bool enable_progress_threads,
+                                bool enable_mpi_threads);
+
+mca_coll_base_module_t *
+ompi_coll_spacc_comm_query(struct ompi_communicator_t *comm, int *priority);
+
+struct mca_coll_spacc_component_t {
+    /* Base coll component */
+    mca_coll_base_component_2_0_0_t super;
+
+    /* MCA parameter: priority of this component */
+    int spacc_priority;
+
+    /* global stuff that I need the component to store */
+
+    /* MCA parameters first */
+};
+
+/*
+ * Convenience typedef
+ */
+typedef struct mca_coll_spacc_component_t mca_coll_spacc_component_t;
+
+/*
+ * Global component instance
+ */
+OMPI_MODULE_DECLSPEC extern mca_coll_spacc_component_t mca_coll_spacc_component;
+
+struct mca_coll_spacc_module_t {
+    mca_coll_base_module_t super;
+};
+typedef struct mca_coll_spacc_module_t mca_coll_spacc_module_t;
+OBJ_CLASS_DECLARATION(mca_coll_spacc_module_t);
+
+#endif  /* MCA_COLL_SPACC_EXPORT_H */