ports/openmpi
ports/openmpi
1
1
Форкнуть 0
Код Релизы Активность
938e7cd8d9
openmpi/ompi/mca/coll/tuned/coll_tuned.h

401 строка
20 KiB
C
Исходник Обычный вид История

snapshot as I move machines do not use This commit was SVN r7103.
2005-08-31 05:43:48 +04:00
/*
Update the copyright notices for IU and UTK. This commit was SVN r7999.
2005-11-05 22:57:48 +03:00
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
A few small changes that just expanded in the name of neatness... (1) As pointed out by Torsten after Jeff comment that there are 15 collectives yesterday.. nope.. I have 16 but miss counted them in my ifdefs (I had two #11s). Replaces with enum... (2) Added a readonly MCA param for how many backend algorithms are available per collective (used by benchmarker/STS) This allowed me to remove the tuned query internal functions and replace them with ompi_coll_tuned_forced_max_algorithms[COLL]. (3) I was reading the user forced MCA params for the collectives on each comm create (module init) but I then put the values into a global set of variables (like ompi_coll_tuned_reduce_forced_algorithm). To fix this and make the code neater: (a) The component looks up the MCA param indices on Open if dynamic_rules is set via the ompi_coll_tuned_COLLECTIVE_intra_check_forced_init () call. (b) Got rid of the ompi_coll_ompi_coll_tuned_COLLECTIVE_forced_algorithm/segmentsize/etc globals with a struct that is now cached on the module data hung off the communicator. i.e. done right. (c) On module init if dynamic rules enabled we call a general getvalues routine (in coll_tuned_forced.c) to get the CURRENT values using the MCA param indices and then put them on the modules data segment. A shorter version of getvalues exists for barrier which only needs the algorithm choice This commit was SVN r9663.
2006-04-20 03:42:06 +04:00
* Copyright (c) 2004-2006 The University of Tennessee and The University
Update the copyright notices for IU and UTK. This commit was SVN r7999.
2005-11-05 22:57:48 +03:00
* of Tennessee Research Foundation. All rights
* reserved.
snapshot as I move machines do not use This commit was SVN r7103.
2005-08-31 05:43:48 +04:00
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
* University of Stuttgart. All rights reserved.
* Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*/
#ifndef MCA_COLL_TUNED_EXPORT_H
#define MCA_COLL_TUNED_EXPORT_H
#include "ompi_config.h"
#include "mpi.h"
Next step in the project split, mainly source code re-arranging - move files out of toplevel include/ and etc/, moving it into the sub-projects - rather than including config headers with <project>/include, have them as <project> - require all headers to be included with a project prefix, with the exception of the config headers ({opal,orte,ompi}_config.h mpi.h, and mpif.h) This commit was SVN r8985.
2006-02-12 04:33:29 +03:00
#include "opal/mca/mca.h"
#include "ompi/mca/coll/coll.h"
#include "ompi/request/request.h"
#include "ompi/mca/pml/pml.h"
snapshot as I move machines do not use This commit was SVN r7103.
2005-08-31 05:43:48 +04:00
/* need to include our own topo prototypes so we can malloc data on the comm correctly */
#include "coll_tuned_topo.h"
snapshot while switching systems but, dynamic rules from a user defined config file is almost there now This commit was SVN r7943.
2005-11-01 03:19:05 +03:00
/* also need the dynamic rule structures */
#include "coll_tuned_dynamic_rules.h"
A few small changes that just expanded in the name of neatness... (1) As pointed out by Torsten after Jeff comment that there are 15 collectives yesterday.. nope.. I have 16 but miss counted them in my ifdefs (I had two #11s). Replaces with enum... (2) Added a readonly MCA param for how many backend algorithms are available per collective (used by benchmarker/STS) This allowed me to remove the tuned query internal functions and replace them with ompi_coll_tuned_forced_max_algorithms[COLL]. (3) I was reading the user forced MCA params for the collectives on each comm create (module init) but I then put the values into a global set of variables (like ompi_coll_tuned_reduce_forced_algorithm). To fix this and make the code neater: (a) The component looks up the MCA param indices on Open if dynamic_rules is set via the ompi_coll_tuned_COLLECTIVE_intra_check_forced_init () call. (b) Got rid of the ompi_coll_ompi_coll_tuned_COLLECTIVE_forced_algorithm/segmentsize/etc globals with a struct that is now cached on the module data hung off the communicator. i.e. done right. (c) On module init if dynamic rules enabled we call a general getvalues routine (in coll_tuned_forced.c) to get the CURRENT values using the MCA param indices and then put them on the modules data segment. A shorter version of getvalues exists for barrier which only needs the algorithm choice This commit was SVN r9663.
2006-04-20 03:42:06 +04:00
/* need the forced user choice structures */
#include "coll_tuned_forced.h"
snapshot while switching systems but, dynamic rules from a user defined config file is almost there now This commit was SVN r7943.
2005-11-01 03:19:05 +03:00
/* some fixed value index vars to simplify certain operations */
A few small changes that just expanded in the name of neatness... (1) As pointed out by Torsten after Jeff comment that there are 15 collectives yesterday.. nope.. I have 16 but miss counted them in my ifdefs (I had two #11s). Replaces with enum... (2) Added a readonly MCA param for how many backend algorithms are available per collective (used by benchmarker/STS) This allowed me to remove the tuned query internal functions and replace them with ompi_coll_tuned_forced_max_algorithms[COLL]. (3) I was reading the user forced MCA params for the collectives on each comm create (module init) but I then put the values into a global set of variables (like ompi_coll_tuned_reduce_forced_algorithm). To fix this and make the code neater: (a) The component looks up the MCA param indices on Open if dynamic_rules is set via the ompi_coll_tuned_COLLECTIVE_intra_check_forced_init () call. (b) Got rid of the ompi_coll_ompi_coll_tuned_COLLECTIVE_forced_algorithm/segmentsize/etc globals with a struct that is now cached on the module data hung off the communicator. i.e. done right. (c) On module init if dynamic rules enabled we call a general getvalues routine (in coll_tuned_forced.c) to get the CURRENT values using the MCA param indices and then put them on the modules data segment. A shorter version of getvalues exists for barrier which only needs the algorithm choice This commit was SVN r9663.
2006-04-20 03:42:06 +04:00
typedef enum COLLTYPE {ALLGATHER, ALLGATHERV, ALLREDUCE, ALLTOALL, ALLTOALLV, ALLTOALLW, BARRIER, BCAST,
EXSCAN, GATHER, GATHERV, REDUCE, REDUCESCATTER, SCAN, SCATTER, SCATTERV, COLLCOUNT} COLLTYPE_T;
snapshot while switching systems but, dynamic rules from a user defined config file is almost there now This commit was SVN r7943.
2005-11-01 03:19:05 +03:00
cleaned up ready for changes to move cached data off the MCW module to the component (where they belong) This commit was SVN r8407.
2005-12-08 06:14:57 +03:00
/* defined arg lists to simply auto inclusion of user overriding decision functions */
#define ALLGATHER_ARGS void *sbuf, int scount, struct ompi_datatype_t *sdtype, void *rbuf, int rcount, struct ompi_datatype_t *rdtype, struct ompi_communicator_t *comm
#define ALLGATHERV_ARGS void *sbuf, int scount, struct ompi_datatype_t *sdtype, void * rbuf, int *rcounts, int *disps, struct ompi_datatype_t *rdtype, struct ompi_communicator_t *comm
#define ALLREDUCE_ARGS void *sbuf, void *rbuf, int count, struct ompi_datatype_t *dtype, struct ompi_op_t *op, struct ompi_communicator_t *comm
#define ALLTOALL_ARGS void *sbuf, int scount, struct ompi_datatype_t *sdtype, void* rbuf, int rcount, struct ompi_datatype_t *rdtype, struct ompi_communicator_t *comm
#define ALLTOALLV_ARGS void *sbuf, int *scounts, int *sdisps, struct ompi_datatype_t *sdtype, void *rbuf, int *rcounts, int *rdisps, struct ompi_datatype_t *rdtype, struct ompi_communicator_t *comm
#define ALLTOALLW_ARGS void *sbuf, int *scounts, int *sdisps, struct ompi_datatype_t **sdtypes, void *rbuf, int *rcounts, int *rdisps, struct ompi_datatype_t **rdtypes, struct ompi_communicator_t *comm
#define BARRIER_ARGS struct ompi_communicator_t *comm
#define BCAST_ARGS void *buff, int count, struct ompi_datatype_t *datatype, int root, struct ompi_communicator_t *comm
#define EXSCAN_ARGS void *sbuf, void *rbuf, int count, struct ompi_datatype_t *dtype, struct ompi_op_t *op, struct ompi_communicator_t *comm
#define GATHER_ARGS void *sbuf, int scount, struct ompi_datatype_t *sdtype, void *rbuf, int rcount, struct ompi_datatype_t *rdtype, int root, struct ompi_communicator_t *comm
#define GATHERV_ARGS void *sbuf, int scount, struct ompi_datatype_t *sdtype, void *rbuf, int *rcounts, int *disps, struct ompi_datatype_t *rdtype, int root, struct ompi_communicator_t *comm
#define REDUCE_ARGS void *sbuf, void* rbuf, int count, struct ompi_datatype_t *dtype, struct ompi_op_t *op, int root, struct ompi_communicator_t *comm
#define REDUCESCATTER_ARGS void *sbuf, void *rbuf, int *rcounts, struct ompi_datatype_t *dtype, struct ompi_op_t *op, struct ompi_communicator_t *comm
#define SCAN_ARGS void *sbuf, void *rbuf, int count, struct ompi_datatype_t *dtype, struct ompi_op_t *op, struct ompi_communicator_t *comm
#define SCATTER_ARGS void *sbuf, int scount, struct ompi_datatype_t *sdtype, void *rbuf, int rcount, struct ompi_datatype_t *rdtype, int root, struct ompi_communicator_t *comm
#define SCATTERV_ARGS void *sbuf, int *scounts, int *disps, struct ompi_datatype_t *sdtype, void* rbuf, int rcount, struct ompi_datatype_t *rdtype, int root, struct ompi_communicator_t *comm
/* end defined arg lists to simply auto inclusion of user overriding decision functions */
snapshot while switching systems but, dynamic rules from a user defined config file is almost there now This commit was SVN r7943.
2005-11-01 03:19:05 +03:00
snapshot as I move machines do not use This commit was SVN r7103.
2005-08-31 05:43:48 +04:00
#if defined(c_plusplus) || defined(__cplusplus)
extern "C" {
#endif
A few small changes that just expanded in the name of neatness... (1) As pointed out by Torsten after Jeff comment that there are 15 collectives yesterday.. nope.. I have 16 but miss counted them in my ifdefs (I had two #11s). Replaces with enum... (2) Added a readonly MCA param for how many backend algorithms are available per collective (used by benchmarker/STS) This allowed me to remove the tuned query internal functions and replace them with ompi_coll_tuned_forced_max_algorithms[COLL]. (3) I was reading the user forced MCA params for the collectives on each comm create (module init) but I then put the values into a global set of variables (like ompi_coll_tuned_reduce_forced_algorithm). To fix this and make the code neater: (a) The component looks up the MCA param indices on Open if dynamic_rules is set via the ompi_coll_tuned_COLLECTIVE_intra_check_forced_init () call. (b) Got rid of the ompi_coll_ompi_coll_tuned_COLLECTIVE_forced_algorithm/segmentsize/etc globals with a struct that is now cached on the module data hung off the communicator. i.e. done right. (c) On module init if dynamic rules enabled we call a general getvalues routine (in coll_tuned_forced.c) to get the CURRENT values using the MCA param indices and then put them on the modules data segment. A shorter version of getvalues exists for barrier which only needs the algorithm choice This commit was SVN r9663.
2006-04-20 03:42:06 +04:00
/* these are the same across all modules and are loaded at component query time */
More protection around what we really want to get exported. This commit was SVN r11437.
2006-08-27 08:49:02 +04:00
extern int ompi_coll_tuned_stream;
extern int ompi_coll_tuned_priority;
extern int ompi_coll_tuned_preallocate_memory_comm_size_limit;
extern int ompi_coll_tuned_use_dynamic_rules;
extern char* ompi_coll_tuned_dynamic_rules_filename;
extern int ompi_coll_tuned_init_tree_fanout;
extern int ompi_coll_tuned_init_chain_fanout;
Last Cleanup BEFORE adding last two methods and final cross over points. - new mca param calls - move printfs to OPAL_OUTPUT This commit was SVN r7692.
2005-10-11 22:51:03 +04:00
Infastructure changes. started to add static (fixed if) statement based decision rules based on gigE numbers added mca params so that a user can force a certain algorithm/segment/topo on a per collective basis (this is not in the fixed call path but only in the dynamic (at com create) call path). (these params can be used by test suites such as OCC to choice which algorithm they are using). This commit was SVN r7854.
2005-10-25 07:55:58 +04:00
/* forced algorithm choices */
A few small changes that just expanded in the name of neatness... (1) As pointed out by Torsten after Jeff comment that there are 15 collectives yesterday.. nope.. I have 16 but miss counted them in my ifdefs (I had two #11s). Replaces with enum... (2) Added a readonly MCA param for how many backend algorithms are available per collective (used by benchmarker/STS) This allowed me to remove the tuned query internal functions and replace them with ompi_coll_tuned_forced_max_algorithms[COLL]. (3) I was reading the user forced MCA params for the collectives on each comm create (module init) but I then put the values into a global set of variables (like ompi_coll_tuned_reduce_forced_algorithm). To fix this and make the code neater: (a) The component looks up the MCA param indices on Open if dynamic_rules is set via the ompi_coll_tuned_COLLECTIVE_intra_check_forced_init () call. (b) Got rid of the ompi_coll_ompi_coll_tuned_COLLECTIVE_forced_algorithm/segmentsize/etc globals with a struct that is now cached on the module data hung off the communicator. i.e. done right. (c) On module init if dynamic rules enabled we call a general getvalues routine (in coll_tuned_forced.c) to get the CURRENT values using the MCA param indices and then put them on the modules data segment. A shorter version of getvalues exists for barrier which only needs the algorithm choice This commit was SVN r9663.
2006-04-20 03:42:06 +04:00
/* the indices to the MCA params so that modules can look them up at open / comm create time */
More protection around what we really want to get exported. This commit was SVN r11437.
2006-08-27 08:49:02 +04:00
extern coll_tuned_force_algorithm_mca_param_indices_t ompi_coll_tuned_forced_params[COLLCOUNT];
A few small changes that just expanded in the name of neatness... (1) As pointed out by Torsten after Jeff comment that there are 15 collectives yesterday.. nope.. I have 16 but miss counted them in my ifdefs (I had two #11s). Replaces with enum... (2) Added a readonly MCA param for how many backend algorithms are available per collective (used by benchmarker/STS) This allowed me to remove the tuned query internal functions and replace them with ompi_coll_tuned_forced_max_algorithms[COLL]. (3) I was reading the user forced MCA params for the collectives on each comm create (module init) but I then put the values into a global set of variables (like ompi_coll_tuned_reduce_forced_algorithm). To fix this and make the code neater: (a) The component looks up the MCA param indices on Open if dynamic_rules is set via the ompi_coll_tuned_COLLECTIVE_intra_check_forced_init () call. (b) Got rid of the ompi_coll_ompi_coll_tuned_COLLECTIVE_forced_algorithm/segmentsize/etc globals with a struct that is now cached on the module data hung off the communicator. i.e. done right. (c) On module init if dynamic rules enabled we call a general getvalues routine (in coll_tuned_forced.c) to get the CURRENT values using the MCA param indices and then put them on the modules data segment. A shorter version of getvalues exists for barrier which only needs the algorithm choice This commit was SVN r9663.
2006-04-20 03:42:06 +04:00
/* the actual max algorithm values (readonly), loaded at component open */
More protection around what we really want to get exported. This commit was SVN r11437.
2006-08-27 08:49:02 +04:00
extern int ompi_coll_tuned_forced_max_algorithms[COLLCOUNT];
snapshot as I move machines do not use This commit was SVN r7103.
2005-08-31 05:43:48 +04:00
/*
* coll API functions
*/
/* API functions */
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_init_query(bool enable_progress_threads,
snapshot as I move machines do not use This commit was SVN r7103.
2005-08-31 05:43:48 +04:00
bool enable_mpi_threads);
minor compile warnings fix This commit was SVN r8497.
2005-12-14 22:09:46 +03:00
snapshot as I move machines do not use This commit was SVN r7103.
2005-08-31 05:43:48 +04:00
const struct mca_coll_base_module_1_0_0_t *
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
ompi_coll_tuned_comm_query(struct ompi_communicator_t *comm, int *priority,
snapshot as I move machines do not use This commit was SVN r7103.
2005-08-31 05:43:48 +04:00
struct mca_coll_base_comm_t **data);
const struct mca_coll_base_module_1_0_0_t *
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
ompi_coll_tuned_module_init(struct ompi_communicator_t *comm);
minor compile warnings fix This commit was SVN r8497.
2005-12-14 22:09:46 +03:00
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_module_finalize(struct ompi_communicator_t *comm);
snapshot as I move machines do not use This commit was SVN r7103.
2005-08-31 05:43:48 +04:00
/* API functions of decision functions and any implementations */
Various clean ups Changed component so choice of decision functions controlled by mca params (for now fixed decision functions (if statements) default) started fixes for the various bcasts This commit was SVN r7117.
2005-09-01 03:38:09 +04:00
/*
* Note this gets long as we have to have a prototype for each
* MPI collective 4 times.. 2 for the comm type and 2 for each decision
* type.
* we might cut down the decision prototypes by conditional compiling
*/
cleaned up ready for changes to move cached data off the MCW module to the component (where they belong) This commit was SVN r8407.
2005-12-08 06:14:57 +03:00
/* All Gather */
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_allgather_intra_dec_fixed(ALLGATHER_ARGS);
int ompi_coll_tuned_allgather_intra_dec_dynamic(ALLGATHER_ARGS);
int ompi_coll_tuned_allgather_inter_dec_fixed(ALLGATHER_ARGS);
int ompi_coll_tuned_allgather_inter_dec_dynamic(ALLGATHER_ARGS);
cleaned up ready for changes to move cached data off the MCW module to the component (where they belong) This commit was SVN r8407.
2005-12-08 06:14:57 +03:00
/* All GatherV */
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_allgatherv_intra_dec_fixed(ALLGATHERV_ARGS);
int ompi_coll_tuned_allgatherv_intra_dec_dynamic(ALLGATHERV_ARGS);
int ompi_coll_tuned_allgatherv_inter_dec_fixed(ALLGATHERV_ARGS);
int ompi_coll_tuned_allgatherv_inter_dec_dynamic(ALLGATHERV_ARGS);
cleaned up ready for changes to move cached data off the MCW module to the component (where they belong) This commit was SVN r8407.
2005-12-08 06:14:57 +03:00
/* All Reduce */
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_allreduce_intra_dec_fixed(ALLREDUCE_ARGS);
int ompi_coll_tuned_allreduce_intra_dec_dynamic(ALLREDUCE_ARGS);
int ompi_coll_tuned_allreduce_intra_do_forced(ALLREDUCE_ARGS);
A few small changes that just expanded in the name of neatness... (1) As pointed out by Torsten after Jeff comment that there are 15 collectives yesterday.. nope.. I have 16 but miss counted them in my ifdefs (I had two #11s). Replaces with enum... (2) Added a readonly MCA param for how many backend algorithms are available per collective (used by benchmarker/STS) This allowed me to remove the tuned query internal functions and replace them with ompi_coll_tuned_forced_max_algorithms[COLL]. (3) I was reading the user forced MCA params for the collectives on each comm create (module init) but I then put the values into a global set of variables (like ompi_coll_tuned_reduce_forced_algorithm). To fix this and make the code neater: (a) The component looks up the MCA param indices on Open if dynamic_rules is set via the ompi_coll_tuned_COLLECTIVE_intra_check_forced_init () call. (b) Got rid of the ompi_coll_ompi_coll_tuned_COLLECTIVE_forced_algorithm/segmentsize/etc globals with a struct that is now cached on the module data hung off the communicator. i.e. done right. (c) On module init if dynamic rules enabled we call a general getvalues routine (in coll_tuned_forced.c) to get the CURRENT values using the MCA param indices and then put them on the modules data segment. A shorter version of getvalues exists for barrier which only needs the algorithm choice This commit was SVN r9663.
2006-04-20 03:42:06 +04:00
int ompi_coll_tuned_allreduce_intra_do_this(ALLREDUCE_ARGS, int algorithm, int faninout, int segsize);
int ompi_coll_tuned_allreduce_intra_check_forced_init (coll_tuned_force_algorithm_mca_param_indices_t *mca_param_indices);
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_allreduce_intra_nonoverlapping(ALLREDUCE_ARGS);
int ompi_coll_tuned_allreduce_intra_basic_linear(ALLREDUCE_ARGS);
int ompi_coll_tuned_allreduce_inter_dec_fixed(ALLREDUCE_ARGS);
int ompi_coll_tuned_allreduce_inter_dec_dynamic(ALLREDUCE_ARGS);
cleaned up ready for changes to move cached data off the MCW module to the component (where they belong) This commit was SVN r8407.
2005-12-08 06:14:57 +03:00
/* AlltoAll */
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_alltoall_intra_dec_fixed(ALLTOALL_ARGS);
int ompi_coll_tuned_alltoall_intra_dec_dynamic(ALLTOALL_ARGS);
int ompi_coll_tuned_alltoall_intra_do_forced(ALLTOALL_ARGS);
A few small changes that just expanded in the name of neatness... (1) As pointed out by Torsten after Jeff comment that there are 15 collectives yesterday.. nope.. I have 16 but miss counted them in my ifdefs (I had two #11s). Replaces with enum... (2) Added a readonly MCA param for how many backend algorithms are available per collective (used by benchmarker/STS) This allowed me to remove the tuned query internal functions and replace them with ompi_coll_tuned_forced_max_algorithms[COLL]. (3) I was reading the user forced MCA params for the collectives on each comm create (module init) but I then put the values into a global set of variables (like ompi_coll_tuned_reduce_forced_algorithm). To fix this and make the code neater: (a) The component looks up the MCA param indices on Open if dynamic_rules is set via the ompi_coll_tuned_COLLECTIVE_intra_check_forced_init () call. (b) Got rid of the ompi_coll_ompi_coll_tuned_COLLECTIVE_forced_algorithm/segmentsize/etc globals with a struct that is now cached on the module data hung off the communicator. i.e. done right. (c) On module init if dynamic rules enabled we call a general getvalues routine (in coll_tuned_forced.c) to get the CURRENT values using the MCA param indices and then put them on the modules data segment. A shorter version of getvalues exists for barrier which only needs the algorithm choice This commit was SVN r9663.
2006-04-20 03:42:06 +04:00
int ompi_coll_tuned_alltoall_intra_do_this(ALLTOALL_ARGS, int algorithm, int faninout, int segsize);
int ompi_coll_tuned_alltoall_intra_check_forced_init (coll_tuned_force_algorithm_mca_param_indices_t *mca_param_indices);
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_alltoall_intra_pairwise(ALLTOALL_ARGS);
int ompi_coll_tuned_alltoall_intra_bruck(ALLTOALL_ARGS);
int ompi_coll_tuned_alltoall_intra_basic_linear(ALLTOALL_ARGS);
int ompi_coll_tuned_alltoall_intra_two_procs(ALLTOALL_ARGS);
int ompi_coll_tuned_alltoall_inter_dec_fixed(ALLTOALL_ARGS);
int ompi_coll_tuned_alltoall_inter_dec_dynamic(ALLTOALL_ARGS);
cleaned up ready for changes to move cached data off the MCW module to the component (where they belong) This commit was SVN r8407.
2005-12-08 06:14:57 +03:00
/* AlltoAllV */
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_alltoallv_intra_dec_fixed(ALLTOALLV_ARGS);
int ompi_coll_tuned_alltoallv_intra_dec_dynamic(ALLTOALLV_ARGS);
int ompi_coll_tuned_alltoallv_inter_dec_fixed(ALLTOALLV_ARGS);
int ompi_coll_tuned_alltoallv_inter_dec_dynamic(ALLTOALLV_ARGS);
cleaned up ready for changes to move cached data off the MCW module to the component (where they belong) This commit was SVN r8407.
2005-12-08 06:14:57 +03:00
/* AlltoAllW */
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_alltoallw_intra_dec_fixed(ALLTOALLW_ARGS);
int ompi_coll_tuned_alltoallw_intra_dec_dynamic(ALLTOALLW_ARGS);
int ompi_coll_tuned_alltoallw_inter_dec_fixed(ALLTOALLW_ARGS);
int ompi_coll_tuned_alltoallw_inter_dec_dynamic(ALLTOALLW_ARGS);
cleaned up ready for changes to move cached data off the MCW module to the component (where they belong) This commit was SVN r8407.
2005-12-08 06:14:57 +03:00
/* Barrier */
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_barrier_intra_dec_fixed(BARRIER_ARGS);
int ompi_coll_tuned_barrier_intra_dec_dynamic(BARRIER_ARGS);
int ompi_coll_tuned_barrier_intra_do_forced(BARRIER_ARGS);
A few small changes that just expanded in the name of neatness... (1) As pointed out by Torsten after Jeff comment that there are 15 collectives yesterday.. nope.. I have 16 but miss counted them in my ifdefs (I had two #11s). Replaces with enum... (2) Added a readonly MCA param for how many backend algorithms are available per collective (used by benchmarker/STS) This allowed me to remove the tuned query internal functions and replace them with ompi_coll_tuned_forced_max_algorithms[COLL]. (3) I was reading the user forced MCA params for the collectives on each comm create (module init) but I then put the values into a global set of variables (like ompi_coll_tuned_reduce_forced_algorithm). To fix this and make the code neater: (a) The component looks up the MCA param indices on Open if dynamic_rules is set via the ompi_coll_tuned_COLLECTIVE_intra_check_forced_init () call. (b) Got rid of the ompi_coll_ompi_coll_tuned_COLLECTIVE_forced_algorithm/segmentsize/etc globals with a struct that is now cached on the module data hung off the communicator. i.e. done right. (c) On module init if dynamic rules enabled we call a general getvalues routine (in coll_tuned_forced.c) to get the CURRENT values using the MCA param indices and then put them on the modules data segment. A shorter version of getvalues exists for barrier which only needs the algorithm choice This commit was SVN r9663.
2006-04-20 03:42:06 +04:00
int ompi_coll_tuned_barrier_intra_do_this(BARRIER_ARGS, int algorithm, int faninout, int segsize);
int ompi_coll_tuned_barrier_intra_check_forced_init (coll_tuned_force_algorithm_mca_param_indices_t *mca_param_indices);
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_barrier_inter_dec_fixed(BARRIER_ARGS);
int ompi_coll_tuned_barrier_inter_dec_dynamic(BARRIER_ARGS);
int ompi_coll_tuned_barrier_intra_doublering(BARRIER_ARGS);
int ompi_coll_tuned_barrier_intra_recursivedoubling(BARRIER_ARGS);
int ompi_coll_tuned_barrier_intra_bruck(BARRIER_ARGS);
int ompi_coll_tuned_barrier_intra_two_procs(BARRIER_ARGS);
int ompi_coll_tuned_barrier_intra_linear(BARRIER_ARGS);
cleaned up ready for changes to move cached data off the MCW module to the component (where they belong) This commit was SVN r8407.
2005-12-08 06:14:57 +03:00
/* Bcast */
New architecture for broadcast. A generic broadcast working on a tree description. Most of the bcast algorithms can be completed using this generic function once we create the tree structure. Add all kind of trees. There are 2 versions of the generic bcast function. One using overlapping between receives (for intermediary nodes) and then blocking sends to all childs and another where all sends are non blocking. I still have to figure out which one give the smallest overhead. This commit was SVN r12530.
2006-11-10 08:53:50 +03:00
int ompi_coll_tuned_bcast_intra_generic( BCAST_ARGS, uint32_t count_by_segment, ompi_coll_tree_t* tree );
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_bcast_intra_dec_fixed(BCAST_ARGS);
int ompi_coll_tuned_bcast_intra_dec_dynamic(BCAST_ARGS);
int ompi_coll_tuned_bcast_intra_do_forced(BCAST_ARGS);
A few small changes that just expanded in the name of neatness... (1) As pointed out by Torsten after Jeff comment that there are 15 collectives yesterday.. nope.. I have 16 but miss counted them in my ifdefs (I had two #11s). Replaces with enum... (2) Added a readonly MCA param for how many backend algorithms are available per collective (used by benchmarker/STS) This allowed me to remove the tuned query internal functions and replace them with ompi_coll_tuned_forced_max_algorithms[COLL]. (3) I was reading the user forced MCA params for the collectives on each comm create (module init) but I then put the values into a global set of variables (like ompi_coll_tuned_reduce_forced_algorithm). To fix this and make the code neater: (a) The component looks up the MCA param indices on Open if dynamic_rules is set via the ompi_coll_tuned_COLLECTIVE_intra_check_forced_init () call. (b) Got rid of the ompi_coll_ompi_coll_tuned_COLLECTIVE_forced_algorithm/segmentsize/etc globals with a struct that is now cached on the module data hung off the communicator. i.e. done right. (c) On module init if dynamic rules enabled we call a general getvalues routine (in coll_tuned_forced.c) to get the CURRENT values using the MCA param indices and then put them on the modules data segment. A shorter version of getvalues exists for barrier which only needs the algorithm choice This commit was SVN r9663.
2006-04-20 03:42:06 +04:00
int ompi_coll_tuned_bcast_intra_do_this(BCAST_ARGS, int algorithm, int faninout, int segsize);
int ompi_coll_tuned_bcast_intra_check_forced_init (coll_tuned_force_algorithm_mca_param_indices_t *mca_param_indices);
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_bcast_intra_basic_linear(BCAST_ARGS);
int ompi_coll_tuned_bcast_intra_chain(BCAST_ARGS, uint32_t segsize, int32_t chains);
int ompi_coll_tuned_bcast_intra_pipeline(BCAST_ARGS, uint32_t segsize);
New architecture for broadcast. A generic broadcast working on a tree description. Most of the bcast algorithms can be completed using this generic function once we create the tree structure. Add all kind of trees. There are 2 versions of the generic bcast function. One using overlapping between receives (for intermediary nodes) and then blocking sends to all childs and another where all sends are non blocking. I still have to figure out which one give the smallest overhead. This commit was SVN r12530.
2006-11-10 08:53:50 +03:00
int ompi_coll_tuned_bcast_intra_binomial(BCAST_ARGS, uint32_t segsize);
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_bcast_intra_bintree(BCAST_ARGS, uint32_t segsize);
int ompi_coll_tuned_bcast_intra_split_bintree(BCAST_ARGS, uint32_t segsize);
int ompi_coll_tuned_bcast_inter_dec_fixed(BCAST_ARGS);
int ompi_coll_tuned_bcast_inter_dec_dynamic(BCAST_ARGS);
cleaned up ready for changes to move cached data off the MCW module to the component (where they belong) This commit was SVN r8407.
2005-12-08 06:14:57 +03:00
/* Exscan */
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_exscan_intra_dec_fixed(EXSCAN_ARGS);
int ompi_coll_tuned_exscan_intra_dec_dynamic(EXSCAN_ARGS);
int ompi_coll_tuned_exscan_inter_dec_fixed(EXSCAN_ARGS);
int ompi_coll_tuned_exscan_inter_dec_dynamic(EXSCAN_ARGS);
cleaned up ready for changes to move cached data off the MCW module to the component (where they belong) This commit was SVN r8407.
2005-12-08 06:14:57 +03:00
/* Gather */
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_gather_intra_dec_fixed(GATHER_ARGS);
int ompi_coll_tuned_gather_intra_dec_dynamic(GATHER_ARGS);
int ompi_coll_tuned_gather_inter_dec_fixed(GATHER_ARGS);
int ompi_coll_tuned_gather_inter_dec_dynamic(GATHER_ARGS);
cleaned up ready for changes to move cached data off the MCW module to the component (where they belong) This commit was SVN r8407.
2005-12-08 06:14:57 +03:00
/* GatherV */
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_gatherv_intra_dec_fixed(GATHERV_ARGS);
int ompi_coll_tuned_gatherv_intra_dec_dynamic(GATHER_ARGS);
int ompi_coll_tuned_gatherv_inter_dec_fixed(GATHER_ARGS);
int ompi_coll_tuned_gatherv_inter_dec_dynamic(GATHER_ARGS);
cleaned up ready for changes to move cached data off the MCW module to the component (where they belong) This commit was SVN r8407.
2005-12-08 06:14:57 +03:00
/* Reduce */
New architecture for broadcast. A generic broadcast working on a tree description. Most of the bcast algorithms can be completed using this generic function once we create the tree structure. Add all kind of trees. There are 2 versions of the generic bcast function. One using overlapping between receives (for intermediary nodes) and then blocking sends to all childs and another where all sends are non blocking. I still have to figure out which one give the smallest overhead. This commit was SVN r12530.
2006-11-10 08:53:50 +03:00
int ompi_coll_tuned_reduce_generic( REDUCE_ARGS, ompi_coll_tree_t* tree, int count_by_segment );
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_reduce_intra_dec_fixed(REDUCE_ARGS);
int ompi_coll_tuned_reduce_intra_dec_dynamic(REDUCE_ARGS);
int ompi_coll_tuned_reduce_intra_do_forced(REDUCE_ARGS);
A few small changes that just expanded in the name of neatness... (1) As pointed out by Torsten after Jeff comment that there are 15 collectives yesterday.. nope.. I have 16 but miss counted them in my ifdefs (I had two #11s). Replaces with enum... (2) Added a readonly MCA param for how many backend algorithms are available per collective (used by benchmarker/STS) This allowed me to remove the tuned query internal functions and replace them with ompi_coll_tuned_forced_max_algorithms[COLL]. (3) I was reading the user forced MCA params for the collectives on each comm create (module init) but I then put the values into a global set of variables (like ompi_coll_tuned_reduce_forced_algorithm). To fix this and make the code neater: (a) The component looks up the MCA param indices on Open if dynamic_rules is set via the ompi_coll_tuned_COLLECTIVE_intra_check_forced_init () call. (b) Got rid of the ompi_coll_ompi_coll_tuned_COLLECTIVE_forced_algorithm/segmentsize/etc globals with a struct that is now cached on the module data hung off the communicator. i.e. done right. (c) On module init if dynamic rules enabled we call a general getvalues routine (in coll_tuned_forced.c) to get the CURRENT values using the MCA param indices and then put them on the modules data segment. A shorter version of getvalues exists for barrier which only needs the algorithm choice This commit was SVN r9663.
2006-04-20 03:42:06 +04:00
int ompi_coll_tuned_reduce_intra_do_this(REDUCE_ARGS, int algorithm, int faninout, int segsize);
int ompi_coll_tuned_reduce_intra_check_forced_init (coll_tuned_force_algorithm_mca_param_indices_t *mca_param_indices);
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_reduce_intra_basic_linear(REDUCE_ARGS);
int ompi_coll_tuned_reduce_intra_chain(REDUCE_ARGS, uint32_t segsize, int fanout);
int ompi_coll_tuned_reduce_intra_pipeline(REDUCE_ARGS, uint32_t segsize);
Big collective commit. I lightly test it, but I think it should be quite stable. Anyway, the default decision functions (for broadcast, reduce and barrier) are based on a high performance network (not TCP). It should give good performance (really good) for any network having the following caracteristics: small latency (5 microseconds) and good bandwidth (more than 1Gb/s). + Cleanup of the reduce algorithms, plus 2 new algorithms (binary and binomial). Now most of the reduce algorithms use a generic tree based function for completing the reduce. + Added macros for computing the trees (they are used for bcast and reduce right now). + Allow the usage of all 5 topologies. + Jelena's implementation of a binary tree that can be used for non commutative operations. Right now only the tree building function is there, it will get activated soon. + Some others minor cleanups. This commit was SVN r12326.
2006-10-27 02:53:05 +04:00
int ompi_coll_tuned_reduce_intra_binary(REDUCE_ARGS, uint32_t segsize);
int ompi_coll_tuned_reduce_intra_binomial(REDUCE_ARGS, uint32_t segsize);
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_reduce_inter_dec_fixed(REDUCE_ARGS);
int ompi_coll_tuned_reduce_inter_dec_dynamic(REDUCE_ARGS);
cleaned up ready for changes to move cached data off the MCW module to the component (where they belong) This commit was SVN r8407.
2005-12-08 06:14:57 +03:00
/* Reduce-Scatter */
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_reduce_scatter_intra_dec_fixed(REDUCESCATTER_ARGS);
int ompi_coll_tuned_reduce_scatter_intra_dec_dynamic(REDUCESCATTER_ARGS);
int ompi_coll_tuned_reduce_scatter_inter_dec_fixed(REDUCESCATTER_ARGS);
int ompi_coll_tuned_reduce_scatter_inter_dec_dynamic(REDUCESCATTER_ARGS);
cleaned up ready for changes to move cached data off the MCW module to the component (where they belong) This commit was SVN r8407.
2005-12-08 06:14:57 +03:00
/* Scan */
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_scan_intra_dec_fixed(SCAN_ARGS);
int ompi_coll_tuned_scan_intra_dec_dynamic(SCAN_ARGS);
int ompi_coll_tuned_scan_inter_dec_fixed(SCAN_ARGS);
int ompi_coll_tuned_scan_inter_dec_dynamic(SCAN_ARGS);
cleaned up ready for changes to move cached data off the MCW module to the component (where they belong) This commit was SVN r8407.
2005-12-08 06:14:57 +03:00
/* Scatter */
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_scatter_intra_dec_fixed(SCATTER_ARGS);
int ompi_coll_tuned_scatter_intra_dec_dynamic(SCATTER_ARGS);
int ompi_coll_tuned_scatter_inter_dec_fixed(SCATTER_ARGS);
int ompi_coll_tuned_scatter_inter_dec_dynamic(SCATTER_ARGS);
cleaned up ready for changes to move cached data off the MCW module to the component (where they belong) This commit was SVN r8407.
2005-12-08 06:14:57 +03:00
/* ScatterV */
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
int ompi_coll_tuned_scatterv_intra_dec_fixed(SCATTERV_ARGS);
int ompi_coll_tuned_scatterv_intra_dec_dynamic(SCATTERV_ARGS);
int ompi_coll_tuned_scatterv_inter_dec_fixed(SCATTERV_ARGS);
int ompi_coll_tuned_scatterv_inter_dec_dynamic(SCATTERV_ARGS);
cleaned up ready for changes to move cached data off the MCW module to the component (where they belong) This commit was SVN r8407.
2005-12-08 06:14:57 +03:00
a few more functions (testing) This commit was SVN r7190.
2005-09-06 09:21:57 +04:00
snapshot as I move machines do not use This commit was SVN r7103.
2005-08-31 05:43:48 +04:00
/* Utility functions */
Update to have public symbols be consistent; use new prefix rule (apparently we've been doing this in opal and orte, but not in ompi yet). All public symbols begin with "ompi_coll_tuned_" (not mca_coll_tuned_) except the component struct. Now this component passes the illegal symbol report with no hits. This commit was SVN r8589.
2005-12-22 16:49:33 +03:00
static inline void ompi_coll_tuned_free_reqs(ompi_request_t **reqs, int count)
snapshot as I move machines do not use This commit was SVN r7103.
2005-08-31 05:43:48 +04:00
{
int i;
for (i = 0; i < count; ++i)
ompi_request_free(&reqs[i]);
}
Start moving stuff out of modules that should be in the component. This commit was SVN r8874.
2006-02-01 23:50:14 +03:00
struct mca_coll_tuned_component_t {
/** Base coll component */
mca_coll_base_component_1_0_0_t super;
/** MCA parameter: Priority of this component */
int tuned_priority;
/** global stuff that I need the component to store */
/* MCA parameters first */
/* cached decision table stuff (moved from MCW module) */
ompi_coll_alg_rule_t *all_base_rules;
};
/**
* Convenience typedef
*/
typedef struct mca_coll_tuned_component_t mca_coll_tuned_component_t;
/**
* Global component instance
*/
More protection around what we really want to get exported. This commit was SVN r11437.
2006-08-27 08:49:02 +04:00
OMPI_MODULE_DECLSPEC extern mca_coll_tuned_component_t mca_coll_tuned_component;
Start moving stuff out of modules that should be in the component. This commit was SVN r8874.
2006-02-01 23:50:14 +03:00
snapshot as I move machines do not use This commit was SVN r7103.
2005-08-31 05:43:48 +04:00
/*
* Data structure for hanging data off the communicator
Start moving stuff out of modules that should be in the component. This commit was SVN r8874.
2006-02-01 23:50:14 +03:00
* i.e. per module instance
snapshot as I move machines do not use This commit was SVN r7103.
2005-08-31 05:43:48 +04:00
*/
struct mca_coll_base_comm_t {
/* standard data for requests and PML usage */
checkpoint This commit was SVN r7168.
2005-09-03 05:41:13 +04:00
/* Precreate space for requests
* Note this does not effect basic,
* but if in wrong context can confuse a debugger
Multiple changes -added some alltoall calls (pairwise checked ok, bruck testing) -changes in use of data hung of communicator -making sendrecv call a true inline function -more use ompi_ddt routines This commit was SVN r7337.
2005-09-13 08:28:18 +04:00
* this is controlled by an MCA param
snapshot as I move machines do not use This commit was SVN r7103.
2005-08-31 05:43:48 +04:00
*/
checkpoint This commit was SVN r7168.
2005-09-03 05:41:13 +04:00
ompi_request_t **mcct_reqs;
int mcct_num_reqs;
snapshot as I move machines do not use This commit was SVN r7103.
2005-08-31 05:43:48 +04:00
/*
* tuned topo information caching per communicator
*
Various clean ups Changed component so choice of decision functions controlled by mca params (for now fixed decision functions (if statements) default) started fixes for the various bcasts This commit was SVN r7117.
2005-09-01 03:38:09 +04:00
* for each communicator we cache the topo information so we can
* reuse without regenerating if we change the root, [or fanout]
* then regenerate and recache this information
snapshot as I move machines do not use This commit was SVN r7103.
2005-08-31 05:43:48 +04:00
*
*/
checkpoint This commit was SVN r7168.
2005-09-03 05:41:13 +04:00
/* general tree with n fan out */
ompi_coll_tree_t *cached_ntree;
int cached_ntree_root;
int cached_ntree_fanout;
snapshot as I move machines do not use This commit was SVN r7103.
2005-08-31 05:43:48 +04:00
checkpoint This commit was SVN r7168.
2005-09-03 05:41:13 +04:00
/* binary tree */
ompi_coll_tree_t *cached_bintree;
int cached_bintree_root;
/* binomial tree */
ompi_coll_tree_t *cached_bmtree;
snapshot as I move machines do not use This commit was SVN r7103.
2005-08-31 05:43:48 +04:00
int cached_bmtree_root;
checkpoint This commit was SVN r7168.
2005-09-03 05:41:13 +04:00
/* chained tree (fanout followed by pipelines) */
One to rule them all. We only need one topology information: a tree. How we build it it's hat make the difference. This commit was SVN r12268.
2006-10-24 01:46:30 +04:00
ompi_coll_tree_t *cached_chain;
snapshot as I move machines do not use This commit was SVN r7103.
2005-08-31 05:43:48 +04:00
int cached_chain_root;
int cached_chain_fanout;
checkpoint This commit was SVN r7168.
2005-09-03 05:41:13 +04:00
/* pipeline */
One to rule them all. We only need one topology information: a tree. How we build it it's hat make the difference. This commit was SVN r12268.
2006-10-24 01:46:30 +04:00
ompi_coll_tree_t *cached_pipeline;
checkpoint This commit was SVN r7168.
2005-09-03 05:41:13 +04:00
int cached_pipeline_root;
Enabled, and running... todos. turn the debug messages into ompi ignorables and inot do some ops in ompi_bug mode This commit was SVN r8036.
2005-11-08 07:43:17 +03:00
/* extra data required by the decision functions */
ompi_coll_alg_rule_t *all_base_rules; /* stored only on MCW, all other coms ref it */
Start moving stuff out of modules that should be in the component. This commit was SVN r8874.
2006-02-01 23:50:14 +03:00
/* moving to the component */
Enabled, and running... todos. turn the debug messages into ompi ignorables and inot do some ops in ompi_bug mode This commit was SVN r8036.
2005-11-08 07:43:17 +03:00
ompi_coll_com_rule_t *com_rules[COLLCOUNT]; /* the communicator rules for each MPI collective for ONLY my comsize */
A few small changes that just expanded in the name of neatness... (1) As pointed out by Torsten after Jeff comment that there are 15 collectives yesterday.. nope.. I have 16 but miss counted them in my ifdefs (I had two #11s). Replaces with enum... (2) Added a readonly MCA param for how many backend algorithms are available per collective (used by benchmarker/STS) This allowed me to remove the tuned query internal functions and replace them with ompi_coll_tuned_forced_max_algorithms[COLL]. (3) I was reading the user forced MCA params for the collectives on each comm create (module init) but I then put the values into a global set of variables (like ompi_coll_tuned_reduce_forced_algorithm). To fix this and make the code neater: (a) The component looks up the MCA param indices on Open if dynamic_rules is set via the ompi_coll_tuned_COLLECTIVE_intra_check_forced_init () call. (b) Got rid of the ompi_coll_ompi_coll_tuned_COLLECTIVE_forced_algorithm/segmentsize/etc globals with a struct that is now cached on the module data hung off the communicator. i.e. done right. (c) On module init if dynamic rules enabled we call a general getvalues routine (in coll_tuned_forced.c) to get the CURRENT values using the MCA param indices and then put them on the modules data segment. A shorter version of getvalues exists for barrier which only needs the algorithm choice This commit was SVN r9663.
2006-04-20 03:42:06 +04:00
/* for forced algorithms we store the information on the module */
/* previously we only had one shared copy, ops, it really is per comm/module */
coll_tuned_force_algorithm_params_t user_forced[COLLCOUNT];
snapshot as I move machines do not use This commit was SVN r7103.
2005-08-31 05:43:48 +04:00
};
Start moving stuff out of modules that should be in the component. This commit was SVN r8874.
2006-02-01 23:50:14 +03:00
/**
* Convenience typedef
*/
typedef struct mca_coll_base_comm_t mca_coll_base_comm_t;
snapshot as I move machines do not use This commit was SVN r7103.
2005-08-31 05:43:48 +04:00
#if defined(c_plusplus) || defined(__cplusplus)
}
#endif
Big collective commit. I lightly test it, but I think it should be quite stable. Anyway, the default decision functions (for broadcast, reduce and barrier) are based on a high performance network (not TCP). It should give good performance (really good) for any network having the following caracteristics: small latency (5 microseconds) and good bandwidth (more than 1Gb/s). + Cleanup of the reduce algorithms, plus 2 new algorithms (binary and binomial). Now most of the reduce algorithms use a generic tree based function for completing the reduce. + Added macros for computing the trees (they are used for bcast and reduce right now). + Allow the usage of all 5 topologies. + Jelena's implementation of a binary tree that can be used for non commutative operations. Right now only the tree building function is there, it will get activated soon. + Some others minor cleanups. This commit was SVN r12326.
2006-10-27 02:53:05 +04:00
#define COLL_TUNED_UPDATE_BINTREE( OMPI_COMM, ROOT ) \
do { \
mca_coll_base_comm_t* coll_comm = (OMPI_COMM)->c_coll_selected_data; \
if( !( (coll_comm->cached_bintree) \
&& (coll_comm->cached_bintree_root == (ROOT)) ) ) { \
if( coll_comm->cached_bintree ) { /* destroy previous binomial if defined */ \
ompi_coll_tuned_topo_destroy_tree( &(coll_comm->cached_bintree) ); \
} \
coll_comm->cached_bintree = ompi_coll_tuned_topo_build_tree(2,(OMPI_COMM),(ROOT)); \
coll_comm->cached_bintree_root = (ROOT); \
} \
} while (0)
#define COLL_TUNED_UPDATE_BMTREE( OMPI_COMM, ROOT ) \
do { \
mca_coll_base_comm_t* coll_comm = (OMPI_COMM)->c_coll_selected_data; \
if( !( (coll_comm->cached_bmtree) \
&& (coll_comm->cached_bmtree_root == (ROOT)) ) ) { \
if( coll_comm->cached_bmtree ) { /* destroy previous binomial if defined */ \
ompi_coll_tuned_topo_destroy_tree( &(coll_comm->cached_bmtree) ); \
} \
coll_comm->cached_bmtree = ompi_coll_tuned_topo_build_bmtree( (OMPI_COMM), (ROOT) ); \
coll_comm->cached_bmtree_root = (ROOT); \
} \
} while (0)
#define COLL_TUNED_UPDATE_PIPELINE( OMPI_COMM, ROOT ) \
do { \
mca_coll_base_comm_t* coll_comm = (OMPI_COMM)->c_coll_selected_data; \
if( !( (coll_comm->cached_pipeline) \
&& (coll_comm->cached_pipeline_root == (ROOT)) ) ) { \
if (coll_comm->cached_pipeline) { /* destroy previous pipeline if defined */ \
ompi_coll_tuned_topo_destroy_tree( &(coll_comm->cached_pipeline) ); \
} \
coll_comm->cached_pipeline = ompi_coll_tuned_topo_build_chain( 1, (OMPI_COMM), (ROOT) ); \
coll_comm->cached_pipeline_root = (ROOT); \
} \
} while (0)
#define COLL_TUNED_UPDATE_CHAIN( OMPI_COMM, ROOT, FANOUT ) \
do { \
mca_coll_base_comm_t* coll_comm = (OMPI_COMM)->c_coll_selected_data; \
if( !( (coll_comm->cached_chain) \
&& (coll_comm->cached_chain_root == (ROOT)) \
&& (coll_comm->cached_chain_fanout == (FANOUT)) ) ) { \
if( coll_comm->cached_chain) { /* destroy previous chain if defined */ \
ompi_coll_tuned_topo_destroy_tree( &(coll_comm->cached_chain) ); \
} \
coll_comm->cached_chain = ompi_coll_tuned_topo_build_chain((FANOUT),(OMPI_COMM),(ROOT)); \
coll_comm->cached_chain_root = (ROOT); \
coll_comm->cached_chain_fanout = (FANOUT); \
} \
} while (0)
snapshot as I move machines do not use This commit was SVN r7103.
2005-08-31 05:43:48 +04:00
#endif /* MCA_COLL_TUNED_EXPORT_H */
Копировать постоянную ссылку