=================================== MVAPICH-Plus Environment Variables =================================== :Author: MVAPICH Team :Date: 2024/07/26 CVAR Environment Variables -------------------------- These are set as environment variables setting the CVAR or it's alias to potential values. For example, setting the barrier intra-node algorithm: .. code:: sh MVP_BARRIER_INTRA_ALGORITHM="osu_pairwise" MPIR\_CVAR\_BARRIER\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BARRIER\_INTRA\_ALGORITHM - MPICH\_BARRIER\_INTRA\_ALGORITHM - **Description:** - Variable to select barrier algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - dynamic - Internal algorithm selection using dynamic tuning - nb - - Force nonblocking algorithm - smp - - Force smp algorithm - k\_dissemination - Force high radix dissemination algorithm - recexch - Force recursive exchange algorithm - osu\_pairwise - Force OSU mcast based design - **Default:** MPIR\_CVAR\_BARRIER\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_BARRIER\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BARRIER\_INTER\_ALGORITHM - MPICH\_BARRIER\_INTER\_ALGORITHM - **Description:** - Variable to select barrier algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - bcast - Force bcast algorithm - nb - Force nonblocking algorithm - **Default:** MPIR\_CVAR\_BARRIER\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_BARRIER\_DISSEM\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BARRIER\_DISSEM\_KVAL - MPICH\_BARRIER\_DISSEM\_KVAL - **Description:** - k value for dissemination exchange based barrier algorithm - **Default:** 2 MPIR\_CVAR\_BARRIER\_RECEXCH\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BARRIER\_RECEXCH\_KVAL - MPICH\_BARRIER\_RECEXCH\_KVAL - **Description:** - k value for recursive exchange based allreduce based - barrier - **Default:** 2 MPIR\_CVAR\_BARRIER\_RECEXCH\_SINGLE\_PHASE\_RECV ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BARRIER\_RECEXCH\_SINGLE\_PHASE\_RECV - MPICH\_BARRIER\_RECEXCH\_SINGLE\_PHASE\_RECV - **Description:** - This CVAR controls whether the recv is posted for one - phase or two phases in recexch algos. By default, we post the recvs - for 2 phases. - **Default:** 0 MPIR\_CVAR\_IBARRIER\_RECEXCH\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IBARRIER\_RECEXCH\_KVAL - MPICH\_IBARRIER\_RECEXCH\_KVAL - **Description:** - k value for recursive exchange based ibarrier - **Default:** 2 MPIR\_CVAR\_IBARRIER\_DISSEM\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IBARRIER\_DISSEM\_KVAL - MPICH\_IBARRIER\_DISSEM\_KVAL - **Description:** - k value for dissemination exchange based ibarrier - **Default:** 2 MPIR\_CVAR\_IBARRIER\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IBARRIER\_INTRA\_ALGORITHM - MPICH\_IBARRIER\_INTRA\_ALGORITHM - **Description:** - Variable to select ibarrier algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_recursive\_doubling - Force recursive doubling algorithm - tsp\_recexch - Force generic transport based recursive exchange - algorithm - tsp\_k\_dissemination - Force generic transport based high-radix - dissemination algorithm - **Default:** MPIR\_CVAR\_IBARRIER\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_IBARRIER\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IBARRIER\_INTER\_ALGORITHM - MPICH\_IBARRIER\_INTER\_ALGORITHM - **Description:** - Variable to select ibarrier algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_bcast - Force bcast algorithm - **Default:** MPIR\_CVAR\_IBARRIER\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_BCAST\_MIN\_PROCS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_MIN\_PROCS - MPICH\_BCAST\_MIN\_PROCS - **Description:** - Let's define short messages as messages with size < - MPIR\_CVAR\_BCAST\_SHORT\_MSG\_SIZE, and medium messages as messages - with size >= MPIR\_CVAR\_BCAST\_SHORT\_MSG\_SIZE but < - MPIR\_CVAR\_BCAST\_LONG\_MSG\_SIZE, and long messages as messages with - size >= MPIR\_CVAR\_BCAST\_LONG\_MSG\_SIZE. The broadcast algorithms - selection procedure is as follows. For short messages or when the - number of processes is < MPIR\_CVAR\_BCAST\_MIN\_PROCS, we do broadcast - using the binomial tree algorithm. Otherwise, for medium messages - and with a power-of-two number of processes, we do broadcast based - on a scatter followed by a recursive doubling allgather algorithm. - Otherwise, for long messages or with non power-of-two number of - processes, we do broadcast based on a scatter followed by a ring - allgather algorithm. (See also: MPIR\_CVAR\_BCAST\_SHORT\_MSG\_SIZE, - MPIR\_CVAR\_BCAST\_LONG\_MSG\_SIZE) - **Default:** 8 MPIR\_CVAR\_BCAST\_SHORT\_MSG\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_SHORT\_MSG\_SIZE - MPICH\_BCAST\_SHORT\_MSG\_SIZE - **Description:** - Let's define short messages as messages with size < - MPIR\_CVAR\_BCAST\_SHORT\_MSG\_SIZE, and medium messages as messages - with size >= MPIR\_CVAR\_BCAST\_SHORT\_MSG\_SIZE but < - MPIR\_CVAR\_BCAST\_LONG\_MSG\_SIZE, and long messages as messages with - size >= MPIR\_CVAR\_BCAST\_LONG\_MSG\_SIZE. The broadcast algorithms - selection procedure is as follows. For short messages or when the - number of processes is < MPIR\_CVAR\_BCAST\_MIN\_PROCS, we do broadcast - using the binomial tree algorithm. Otherwise, for medium messages - and with a power-of-two number of processes, we do broadcast based - on a scatter followed by a recursive doubling allgather algorithm. - Otherwise, for long messages or with non power-of-two number of - processes, we do broadcast based on a scatter followed by a ring - allgather algorithm. (See also: MPIR\_CVAR\_BCAST\_MIN\_PROCS, - MPIR\_CVAR\_BCAST\_LONG\_MSG\_SIZE) - **Default:** 12288 MPIR\_CVAR\_BCAST\_LONG\_MSG\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_LONG\_MSG\_SIZE - MPICH\_BCAST\_LONG\_MSG\_SIZE - **Description:** - Let's define short messages as messages with size < - MPIR\_CVAR\_BCAST\_SHORT\_MSG\_SIZE, and medium messages as messages - with size >= MPIR\_CVAR\_BCAST\_SHORT\_MSG\_SIZE but < - MPIR\_CVAR\_BCAST\_LONG\_MSG\_SIZE, and long messages as messages with - size >= MPIR\_CVAR\_BCAST\_LONG\_MSG\_SIZE. The broadcast algorithms - selection procedure is as follows. For short messages or when the - number of processes is < MPIR\_CVAR\_BCAST\_MIN\_PROCS, we do broadcast - using the binomial tree algorithm. Otherwise, for medium messages - and with a power-of-two number of processes, we do broadcast based - on a scatter followed by a recursive doubling allgather algorithm. - Otherwise, for long messages or with non power-of-two number of - processes, we do broadcast based on a scatter followed by a ring - allgather algorithm. (See also: MPIR\_CVAR\_BCAST\_MIN\_PROCS, - MPIR\_CVAR\_BCAST\_SHORT\_MSG\_SIZE) - **Default:** 524288 MPIR\_CVAR\_MAX\_SMP\_BCAST\_MSG\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_MAX\_SMP\_BCAST\_MSG\_SIZE - MPICH\_MAX\_SMP\_BCAST\_MSG\_SIZE - **Description:** - Maximum message size for which SMP-aware broadcast is - used. A value of '0' uses SMP-aware broadcast for all message - sizes. - **Default:** 0 MPIR\_CVAR\_BCAST\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_INTRA\_ALGORITHM - MPICH\_BCAST\_INTRA\_ALGORITHM - **Description:** - Variable to select bcast algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - dynamic - Internal algorithm selection using dynamic tuning - binomial - - - - - Force Binomial Tree - nb - - - - - Force nonblocking - algorithm - smp - - - - - Force smp algorithm - scatter\_recursive\_doubling\_allgather - Force Scatter - Recursive-Doubling Allgather - scatter\_ring\_allgather - - - Force Scatter Ring - pipelined\_tree - - - - Force tree-based - pipelined algorithm - tree - - - - - Force tree-based - algorithm - osu\_knomial - - - - Force OSU knomial - algorithm - osu\_pairwise - - - - Force OSU pairwise - algorithm - osu\_pipelined - - - - Force OSU pipelined - algorithm - **Default:** MPIR\_CVAR\_BCAST\_INTRA\_ALGORITHM\_dynamic MPIR\_CVAR\_BCAST\_TREE\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_TREE\_KVAL - MPICH\_BCAST\_TREE\_KVAL - **Description:** - k value for tree (kary, knomial, etc.) based bcast - **Default:** 2 MPIR\_CVAR\_BCAST\_TREE\_TYPE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_TREE\_TYPE - MPICH\_BCAST\_TREE\_TYPE - **Description:** - Tree type for tree based bcast kary - kary tree type - knomial\_1 - knomial\_1 tree type knomial\_2 - knomial\_2 tree type - topology\_aware - topology\_aware tree type topology\_aware\_k - - topology\_aware tree type with branching factor k topology\_wave - - topology\_wave tree type - **Default:** "kary" MPIR\_CVAR\_BCAST\_TOPO\_REORDER\_ENABLE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_TOPO\_REORDER\_ENABLE - MPICH\_BCAST\_TOPO\_REORDER\_ENABLE - **Description:** - This cvar controls if the leaders are reordered based on - the number of ranks in each group. - **Default:** 1 MPIR\_CVAR\_BCAST\_TOPO\_OVERHEAD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_TOPO\_OVERHEAD - MPICH\_BCAST\_TOPO\_OVERHEAD - **Description:** - This cvar controls the size of the overhead. - **Default:** 200 MPIR\_CVAR\_BCAST\_TOPO\_DIFF\_GROUPS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_TOPO\_DIFF\_GROUPS - MPICH\_BCAST\_TOPO\_DIFF\_GROUPS - **Description:** - This cvar controls the latency between different groups. - **Default:** 2800 MPIR\_CVAR\_BCAST\_TOPO\_DIFF\_SWITCHES ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_TOPO\_DIFF\_SWITCHES - MPICH\_BCAST\_TOPO\_DIFF\_SWITCHES - **Description:** - This cvar controls the latency between different switches - in the same groups. - **Default:** 1900 MPIR\_CVAR\_BCAST\_TOPO\_SAME\_SWITCHES ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_TOPO\_SAME\_SWITCHES - MPICH\_BCAST\_TOPO\_SAME\_SWITCHES - **Description:** - This cvar controls the latency in the same switch. - **Default:** 1600 MPIR\_CVAR\_BCAST\_IS\_NON\_BLOCKING ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_IS\_NON\_BLOCKING - MPICH\_BCAST\_IS\_NON\_BLOCKING - **Description:** - If set to true, MPI\_Bcast will use non-blocking send. - **Default:** 1 MPIR\_CVAR\_BCAST\_TREE\_PIPELINE\_CHUNK\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_TREE\_PIPELINE\_CHUNK\_SIZE - MPICH\_BCAST\_TREE\_PIPELINE\_CHUNK\_SIZE - **Description:** - Indicates the chunk size for pipelined bcast. - **Default:** 8192 MPIR\_CVAR\_BCAST\_RECV\_PRE\_POST ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_RECV\_PRE\_POST - MPICH\_BCAST\_RECV\_PRE\_POST - **Description:** - If set to true, MPI\_Bcast will pre-post all the receives. - **Default:** 0 MPIR\_CVAR\_BCAST\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_INTER\_ALGORITHM - MPICH\_BCAST\_INTER\_ALGORITHM - **Description:** - Variable to select bcast algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - nb - - - Force nonblocking algorithm - remote\_send\_local\_bcast - Force remote-send-local-bcast algorithm - **Default:** MPIR\_CVAR\_BCAST\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_IBCAST\_TREE\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IBCAST\_TREE\_KVAL - MPICH\_IBCAST\_TREE\_KVAL - **Description:** - k value for tree (kary, knomial, etc.) based ibcast - **Default:** 2 MPIR\_CVAR\_IBCAST\_TREE\_TYPE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IBCAST\_TREE\_TYPE - MPICH\_IBCAST\_TREE\_TYPE - **Description:** - Tree type for tree based ibcast kary - kary tree type - knomial\_1 - knomial\_1 tree type knomial\_2 - knomial\_2 tree type - **Default:** "kary" MPIR\_CVAR\_IBCAST\_TREE\_PIPELINE\_CHUNK\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IBCAST\_TREE\_PIPELINE\_CHUNK\_SIZE - MPICH\_IBCAST\_TREE\_PIPELINE\_CHUNK\_SIZE - **Description:** - Maximum chunk size (in bytes) for pipelining in tree based - ibcast. Default value is 0, that is, no pipelining by default - **Default:** 0 MPIR\_CVAR\_IBCAST\_RING\_CHUNK\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IBCAST\_RING\_CHUNK\_SIZE - MPICH\_IBCAST\_RING\_CHUNK\_SIZE - **Description:** - Maximum chunk size (in bytes) for pipelining in ibcast - ring algorithm. Default value is 0, that is, no pipelining by - default - **Default:** 0 MPIR\_CVAR\_IBCAST\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IBCAST\_INTRA\_ALGORITHM - MPICH\_IBCAST\_INTRA\_ALGORITHM - **Description:** - Variable to select ibcast algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_binomial - - - - Force Binomial - algorithm - sched\_smp - - - - - Force smp algorithm - sched\_scatter\_recursive\_doubling\_allgather - Force Scatter - Recursive Doubling Allgather algorithm - sched\_scatter\_ring\_allgather - - Force Scatter Ring - Allgather algorithm - tsp\_tree - - - - Force Generic Transport - Tree algorithm - tsp\_scatterv\_recexch\_allgatherv - - Force Generic Transport - Scatterv followed by Recursive Exchange Allgatherv algorithm - tsp\_scatterv\_ring\_allgatherv - - Force Generic Transport - Scatterv followed by Ring Allgatherv algorithm - tsp\_ring - - - - Force Generic Transport - Ring algorithm - **Default:** MPIR\_CVAR\_IBCAST\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_IBCAST\_SCATTERV\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IBCAST\_SCATTERV\_KVAL - MPICH\_IBCAST\_SCATTERV\_KVAL - **Description:** - k value for tree based scatter in - scatter\_recexch\_allgather algorithm - **Default:** 2 MPIR\_CVAR\_IBCAST\_ALLGATHERV\_RECEXCH\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IBCAST\_ALLGATHERV\_RECEXCH\_KVAL - MPICH\_IBCAST\_ALLGATHERV\_RECEXCH\_KVAL - **Description:** - k value for recursive exchange based allgather in - scatter\_recexch\_allgather algorithm - **Default:** 2 MPIR\_CVAR\_IBCAST\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IBCAST\_INTER\_ALGORITHM - MPICH\_IBCAST\_INTER\_ALGORITHM - **Description:** - Variable to select ibcast algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_flat - Force flat algorithm - **Default:** MPIR\_CVAR\_IBCAST\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_GATHER\_INTER\_SHORT\_MSG\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_GATHER\_INTER\_SHORT\_MSG\_SIZE - MPICH\_GATHER\_INTER\_SHORT\_MSG\_SIZE - **Description:** - use the short message algorithm for intercommunicator - MPI\_Gather if the send buffer size is < this value (in bytes) (See - also: MPIR\_CVAR\_GATHER\_VSMALL\_MSG\_SIZE) - **Default:** 2048 MPIR\_CVAR\_GATHER\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_GATHER\_INTRA\_ALGORITHM - MPICH\_GATHER\_INTRA\_ALGORITHM - **Description:** - Variable to select gather algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - dynamic - Internal algorithm selection using dynamic tuning - binomial - Force binomial algorithm - nb - Force nonblocking algorithm - osu\_direct - Force OSU direct algorithm - osu\_direct\_block - Force OSU direct block algorithm - **Default:** MPIR\_CVAR\_GATHER\_INTRA\_ALGORITHM\_dynamic MPIR\_CVAR\_GATHER\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_GATHER\_INTER\_ALGORITHM - MPICH\_GATHER\_INTER\_ALGORITHM - **Description:** - Variable to select gather algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - linear - - - Force linear algorithm - local\_gather\_remote\_send - Force local-gather-remote-send algorithm - nb - - - Force nonblocking algorithm - **Default:** MPIR\_CVAR\_GATHER\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_IGATHER\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IGATHER\_INTRA\_ALGORITHM - MPICH\_IGATHER\_INTRA\_ALGORITHM - **Description:** - Variable to select igather algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_binomial - Force binomial algorithm - tsp\_tree - Force genetric transport based tree algorithm - **Default:** MPIR\_CVAR\_IGATHER\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_IGATHER\_TREE\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IGATHER\_TREE\_KVAL - MPICH\_IGATHER\_TREE\_KVAL - **Description:** - k value for tree based igather - **Default:** 2 MPIR\_CVAR\_IGATHER\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IGATHER\_INTER\_ALGORITHM - MPICH\_IGATHER\_INTER\_ALGORITHM - **Description:** - Variable to select igather algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_long - Force long inter algorithm - sched\_short - Force short inter algorithm - **Default:** MPIR\_CVAR\_IGATHER\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_GATHERV\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_GATHERV\_INTRA\_ALGORITHM - MPICH\_GATHERV\_INTRA\_ALGORITHM - **Description:** - Variable to select gatherv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - dynamic - Internal algorithm selection using dynamic tuning - linear - Force linear algorithm - nb - Force nonblocking algorithm - **Default:** MPIR\_CVAR\_GATHERV\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_GATHERV\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_GATHERV\_INTER\_ALGORITHM - MPICH\_GATHERV\_INTER\_ALGORITHM - **Description:** - Variable to select gatherv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - linear - Force linear algorithm - nb - Force nonblocking algorithm - **Default:** MPIR\_CVAR\_GATHERV\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_IGATHERV\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IGATHERV\_INTRA\_ALGORITHM - MPICH\_IGATHERV\_INTRA\_ALGORITHM - **Description:** - Variable to select igatherv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_linear - - Force linear algorithm - tsp\_linear - Force generic transport based linear algorithm - **Default:** MPIR\_CVAR\_IGATHERV\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_IGATHERV\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IGATHERV\_INTER\_ALGORITHM - MPICH\_IGATHERV\_INTER\_ALGORITHM - **Description:** - Variable to select igatherv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_linear - Force linear algorithm - tsp\_linear - Force generic transport based linear algorithm - **Default:** MPIR\_CVAR\_IGATHERV\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_SCATTER\_INTER\_SHORT\_MSG\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_SCATTER\_INTER\_SHORT\_MSG\_SIZE - MPICH\_SCATTER\_INTER\_SHORT\_MSG\_SIZE - **Description:** - use the short message algorithm for intercommunicator - MPI\_Scatter if the send buffer size is < this value (in bytes) - **Default:** 2048 MPIR\_CVAR\_SCATTER\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_SCATTER\_INTRA\_ALGORITHM - MPICH\_SCATTER\_INTRA\_ALGORITHM - **Description:** - Variable to select scatter algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - dynamic - Internal algorithm selection using dynamic tuning - binomial - Force binomial algorithm - nb - Force nonblocking algorithm - osu\_direct - Force OSU direct alogirthm - **Default:** MPIR\_CVAR\_SCATTER\_INTRA\_ALGORITHM\_dynamic MPIR\_CVAR\_SCATTER\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_SCATTER\_INTER\_ALGORITHM - MPICH\_SCATTER\_INTER\_ALGORITHM - **Description:** - Variable to select scatter algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - linear - - - Force linear algorithm - nb - - - - Force nonblocking algorithm - remote\_send\_local\_scatter - Force remote-send-local-scatter - algorithm - **Default:** MPIR\_CVAR\_SCATTER\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_ISCATTER\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ISCATTER\_INTRA\_ALGORITHM - MPICH\_ISCATTER\_INTRA\_ALGORITHM - **Description:** - Variable to select iscatter algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_binomial - Force binomial algorithm - tsp\_tree - Force genetric transport based tree algorithm - **Default:** MPIR\_CVAR\_ISCATTER\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_ISCATTER\_TREE\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ISCATTER\_TREE\_KVAL - MPICH\_ISCATTER\_TREE\_KVAL - **Description:** - k value for tree based iscatter - **Default:** 2 MPIR\_CVAR\_ISCATTER\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ISCATTER\_INTER\_ALGORITHM - MPICH\_ISCATTER\_INTER\_ALGORITHM - **Description:** - Variable to select iscatter algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_linear - - - Force linear algorithm - sched\_remote\_send\_local\_scatter - Force remote-send-local-scatter - algorithm - **Default:** MPIR\_CVAR\_ISCATTER\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_SCATTERV\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_SCATTERV\_INTRA\_ALGORITHM - MPICH\_SCATTERV\_INTRA\_ALGORITHM - **Description:** - Variable to select scatterv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - dynamic - Internal algorithm selection using dynamic tuning - linear - Force linear algorithm - nb - Force nonblocking algorithm - **Default:** MPIR\_CVAR\_SCATTERV\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_SCATTERV\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_SCATTERV\_INTER\_ALGORITHM - MPICH\_SCATTERV\_INTER\_ALGORITHM - **Description:** - Variable to select scatterv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - linear - Force linear algorithm - nb - Force nonblocking algorithm - **Default:** MPIR\_CVAR\_SCATTERV\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_ISCATTERV\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ISCATTERV\_INTRA\_ALGORITHM - MPICH\_ISCATTERV\_INTRA\_ALGORITHM - **Description:** - Variable to select iscatterv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_linear - Force linear algorithm - tsp\_linear - Force generic transport based linear algorithm - **Default:** MPIR\_CVAR\_ISCATTERV\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_ISCATTERV\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ISCATTERV\_INTER\_ALGORITHM - MPICH\_ISCATTERV\_INTER\_ALGORITHM - **Description:** - Variable to select iscatterv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_linear - Force linear algorithm - tsp\_linear - Force generic transport based linear algorithm - **Default:** MPIR\_CVAR\_ISCATTERV\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_ALLGATHER\_SHORT\_MSG\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLGATHER\_SHORT\_MSG\_SIZE - MPICH\_ALLGATHER\_SHORT\_MSG\_SIZE - **Description:** - For MPI\_Allgather and MPI\_Allgatherv, the short message - algorithm will be used if the send buffer size is < this value (in - bytes). (See also: MPIR\_CVAR\_ALLGATHER\_LONG\_MSG\_SIZE) - **Default:** 81920 MPIR\_CVAR\_ALLGATHER\_LONG\_MSG\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLGATHER\_LONG\_MSG\_SIZE - MPICH\_ALLGATHER\_LONG\_MSG\_SIZE - **Description:** - For MPI\_Allgather and MPI\_Allgatherv, the long message - algorithm will be used if the send buffer size is >= this value (in - bytes) (See also: MPIR\_CVAR\_ALLGATHER\_SHORT\_MSG\_SIZE) - **Default:** 524288 MPIR\_CVAR\_ALLGATHER\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLGATHER\_INTRA\_ALGORITHM - MPICH\_ALLGATHER\_INTRA\_ALGORITHM - **Description:** - Variable to select allgather algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - dynamic - Internal algorithm selection using dynamic tuning - brucks - - Force brucks algorithm - k\_brucks - - Force brucks algorithm - nb - - - Force nonblocking algorithm - recursive\_doubling - Force recursive doubling algorithm - ring - - - Force ring algorithm - recexch\_doubling - Force recexch distance doubling algorithm - recexch\_halving - Force recexch distance halving algorithm - osu\_direct - - Force MVAPICH direct algorithm - osu\_direct\_spread - Force MVAPICH direct spread algorithm - osu\_direct\_throttling - Force MVAPICH direct throttling algorithm - osu\_gather\_bcast - Force MVAPICH gather-bcast algorithm - osu\_gpu\_compression - Force MVAPICH gpu compression algorithm - **Default:** MPIR\_CVAR\_ALLGATHER\_INTRA\_ALGORITHM\_dynamic MPIR\_CVAR\_ALLGATHER\_BRUCKS\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLGATHER\_BRUCKS\_KVAL - MPICH\_ALLGATHER\_BRUCKS\_KVAL - **Description:** - radix (k) value for generic transport brucks based - allgather - **Default:** 2 MPIR\_CVAR\_ALLGATHER\_RECEXCH\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLGATHER\_RECEXCH\_KVAL - MPICH\_ALLGATHER\_RECEXCH\_KVAL - **Description:** - k value for recursive exchange based allgather - **Default:** 2 MPIR\_CVAR\_ALLGATHER\_RECEXCH\_SINGLE\_PHASE\_RECV ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLGATHER\_RECEXCH\_SINGLE\_PHASE\_RECV - MPICH\_ALLGATHER\_RECEXCH\_SINGLE\_PHASE\_RECV - **Description:** - This CVAR controls whether the recv is posted for one - phase or two phases in recexch algos. By default, we post the recvs - for 2 phases. - **Default:** 0 MPIR\_CVAR\_ALLGATHER\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLGATHER\_INTER\_ALGORITHM - MPICH\_ALLGATHER\_INTER\_ALGORITHM - **Description:** - Variable to select allgather algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - local\_gather\_remote\_bcast - Force local-gather-remote-bcast - algorithm - nb - - - - Force nonblocking algorithm - **Default:** MPIR\_CVAR\_ALLGATHER\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_IALLGATHER\_RECEXCH\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLGATHER\_RECEXCH\_KVAL - MPICH\_IALLGATHER\_RECEXCH\_KVAL - **Description:** - k value for recursive exchange based iallgather - **Default:** 2 MPIR\_CVAR\_IALLGATHER\_BRUCKS\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLGATHER\_BRUCKS\_KVAL - MPICH\_IALLGATHER\_BRUCKS\_KVAL - **Description:** - k value for radix in brucks based iallgather - **Default:** 2 MPIR\_CVAR\_IALLGATHER\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLGATHER\_INTRA\_ALGORITHM - MPICH\_IALLGATHER\_INTRA\_ALGORITHM - **Description:** - Variable to select iallgather algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_ring - - Force ring algorithm - sched\_brucks - - Force brucks algorithm - sched\_recursive\_doubling - Force recursive doubling algorithm - tsp\_ring - Force generic transport ring algorithm - tsp\_brucks - Force generic transport based brucks algorithm - tsp\_recexch\_doubling - Force generic transport recursive exchange - with neighbours doubling in distance in each phase - tsp\_recexch\_halving - Force generic transport recursive exchange - with neighbours halving in distance in each phase - **Default:** MPIR\_CVAR\_IALLGATHER\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_IALLGATHER\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLGATHER\_INTER\_ALGORITHM - MPICH\_IALLGATHER\_INTER\_ALGORITHM - **Description:** - Variable to select iallgather algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_local\_gather\_remote\_bcast - Force local-gather-remote-bcast - algorithm - **Default:** MPIR\_CVAR\_IALLGATHER\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_ALLGATHERV\_PIPELINE\_MSG\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLGATHERV\_PIPELINE\_MSG\_SIZE - MPICH\_ALLGATHERV\_PIPELINE\_MSG\_SIZE - **Description:** - The smallest message size that will be used for the - pipelined, large-message, ring algorithm in the MPI\_Allgatherv - implementation. - **Default:** 32768 MPIR\_CVAR\_ALLGATHERV\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLGATHERV\_INTRA\_ALGORITHM - MPICH\_ALLGATHERV\_INTRA\_ALGORITHM - **Description:** - Variable to select allgatherv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - dynamic - Internal algorithm selection using dynamic tuning - brucks - - Force brucks algorithm - nb - - - Force nonblocking algorithm - recursive\_doubling - Force recursive doubling algorithm - ring - - Force ring algorithm - osu\_direct\_throttling - Force MVAPICH direct throttling algorithm - **Default:** MPIR\_CVAR\_ALLGATHERV\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_ALLGATHERV\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLGATHERV\_INTER\_ALGORITHM - MPICH\_ALLGATHERV\_INTER\_ALGORITHM - **Description:** - Variable to select allgatherv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - nb - - - - Force nonblocking algorithm - remote\_gather\_local\_bcast - Force remote-gather-local-bcast - algorithm - **Default:** MPIR\_CVAR\_ALLGATHERV\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_IALLGATHERV\_RECEXCH\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLGATHERV\_RECEXCH\_KVAL - MPICH\_IALLGATHERV\_RECEXCH\_KVAL - **Description:** - k value for recursive exchange based iallgatherv - **Default:** 2 MPIR\_CVAR\_IALLGATHERV\_BRUCKS\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLGATHERV\_BRUCKS\_KVAL - MPICH\_IALLGATHERV\_BRUCKS\_KVAL - **Description:** - k value for radix in brucks based iallgatherv - **Default:** 2 MPIR\_CVAR\_IALLGATHERV\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLGATHERV\_INTRA\_ALGORITHM - MPICH\_IALLGATHERV\_INTRA\_ALGORITHM - **Description:** - Variable to select iallgatherv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_brucks - - Force brucks algorithm - sched\_recursive\_doubling - Force recursive doubling algorithm - sched\_ring - - Force ring algorithm - tsp\_recexch\_doubling - Force generic transport recursive exchange - with neighbours doubling in distance in each phase - tsp\_recexch\_halving - Force generic transport recursive exchange - with neighbours halving in distance in each phase - tsp\_ring - - Force generic transport ring algorithm - tsp\_brucks - - Force generic transport based brucks - algorithm - **Default:** MPIR\_CVAR\_IALLGATHERV\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_IALLGATHERV\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLGATHERV\_INTER\_ALGORITHM - MPICH\_IALLGATHERV\_INTER\_ALGORITHM - **Description:** - Variable to select iallgatherv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_remote\_gather\_local\_bcast - Force remote-gather-local-bcast - algorithm - **Default:** MPIR\_CVAR\_IALLGATHERV\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_ALLTOALL\_SHORT\_MSG\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLTOALL\_SHORT\_MSG\_SIZE - MPICH\_ALLTOALL\_SHORT\_MSG\_SIZE - **Description:** - the short message algorithm will be used if the - per-destination message size (sendcount*size(sendtype)) is <= this - value (See also: MPIR\_CVAR\_ALLTOALL\_MEDIUM\_MSG\_SIZE) - **Default:** 256 MPIR\_CVAR\_ALLTOALL\_MEDIUM\_MSG\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLTOALL\_MEDIUM\_MSG\_SIZE - MPICH\_ALLTOALL\_MEDIUM\_MSG\_SIZE - **Description:** - the medium message algorithm will be used if the - per-destination message size (sendcount*size(sendtype)) is <= this - value and larger than MPIR\_CVAR\_ALLTOALL\_SHORT\_MSG\_SIZE (See also: - MPIR\_CVAR\_ALLTOALL\_SHORT\_MSG\_SIZE) - **Default:** 32768 MPIR\_CVAR\_ALLTOALL\_THROTTLE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLTOALL\_THROTTLE - MPICH\_ALLTOALL\_THROTTLE - **Description:** - Maximum number of concurrent irecvs/isends posted at a - time in some alltoall algorithms. Setting it to 0 causes all - irecvs/isends to be posted at once - **Default:** 32 MPIR\_CVAR\_ALLREDUCE\_THROTTLE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_THROTTLE - MPICH\_ALLREDUCE\_THROTTLE - **Description:** - Maximum number of concurrent irecvs/isends posted at a - time in direct allreduce algorithms. Setting it to 0 causes all - irecvs/isends to be posted at once - **Default:** 32 MPIR\_CVAR\_ALLGATHER\_THROTTLE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLGATHER\_THROTTLE - MPICH\_ALLGATHER\_THROTTLE - **Description:** - Maximum number of concurrent irecvs/isends posted at a - time in direct allgather algorithms. Setting it to 0 causes all - irecvs/isends to be posted at once - **Default:** 8 MPIR\_CVAR\_REDUCE\_SCATTER\_THROTTLE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_SCATTER\_THROTTLE - MPICH\_REDUCE\_SCATTER\_THROTTLE - **Description:** - Maximum number of concurrent irecvs/isends posted at a - time in direct reduce scatter algorithms. Setting it to 0 causes - all irecvs/isends to be posted at once - **Default:** 8 MPIR\_CVAR\_ALLTOALL\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLTOALL\_INTRA\_ALGORITHM - MPICH\_ALLTOALL\_INTRA\_ALGORITHM - **Description:** - Variable to select alltoall algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - dynamic - Internal algorithm selection using dynamic tuning - brucks - - - Force brucks algorithm - k\_brucks - - - Force Force radix k brucks algorithm - nb - - - - Force nonblocking algorithm - pairwise - - - Force pairwise algorithm - pairwise\_sendrecv\_replace - Force pairwise sendrecv replace - algorithm - scattered - - - Force scattered algorithm - osu\_gpu\_compression - Force compression algorithm - **Default:** MPIR\_CVAR\_ALLTOALL\_INTRA\_ALGORITHM\_dynamic MPIR\_CVAR\_ALLTOALL\_BRUCKS\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLTOALL\_BRUCKS\_KVAL - MPICH\_ALLTOALL\_BRUCKS\_KVAL - **Description:** - radix (k) value for generic transport brucks based - alltoall - **Default:** 2 MPIR\_CVAR\_ALLTOALL\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLTOALL\_INTER\_ALGORITHM - MPICH\_ALLTOALL\_INTER\_ALGORITHM - **Description:** - Variable to select alltoall algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - nb - - - Force nonblocking algorithm - pairwise\_exchange - Force pairwise exchange algorithm - **Default:** MPIR\_CVAR\_ALLTOALL\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_IALLTOALL\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLTOALL\_INTRA\_ALGORITHM - MPICH\_IALLTOALL\_INTRA\_ALGORITHM - **Description:** - Variable to select ialltoall algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_brucks - - Force brucks algorithm - sched\_inplace - - Force inplace algorithm - sched\_pairwise - - Force pairwise algorithm - sched\_permuted\_sendrecv - Force permuted sendrecv algorithm - tsp\_ring - - Force generic transport based ring algorithm - tsp\_brucks - - Force generic transport based brucks - algorithm - tsp\_scattered - Force generic transport based scattered - algorithm - **Default:** MPIR\_CVAR\_IALLTOALL\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_IALLTOALL\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLTOALL\_INTER\_ALGORITHM - MPICH\_IALLTOALL\_INTER\_ALGORITHM - **Description:** - Variable to select ialltoall algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_pairwise\_exchange - Force pairwise exchange algorithm - **Default:** MPIR\_CVAR\_IALLTOALL\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_ALLTOALLV\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLTOALLV\_INTRA\_ALGORITHM - MPICH\_ALLTOALLV\_INTRA\_ALGORITHM - **Description:** - Variable to select alltoallv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - dynamic - Internal algorithm selection using dynamic tuning - nb - - - - Force nonblocking algorithm - pairwise\_sendrecv\_replace - Force pairwise\_sendrecv\_replace - algorithm - scattered - - - Force scattered algorithm - **Default:** MPIR\_CVAR\_ALLTOALLV\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_ALLTOALLV\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLTOALLV\_INTER\_ALGORITHM - MPICH\_ALLTOALLV\_INTER\_ALGORITHM - **Description:** - Variable to select alltoallv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - pairwise\_exchange - Force pairwise exchange algorithm - nb - - - Force nonblocking algorithm - **Default:** MPIR\_CVAR\_ALLTOALLV\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_IALLTOALLV\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLTOALLV\_INTRA\_ALGORITHM - MPICH\_IALLTOALLV\_INTRA\_ALGORITHM - **Description:** - Variable to select ialltoallv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_blocked - - Force blocked algorithm - sched\_inplace - - Force inplace algorithm - tsp\_scattered - Force generic transport based scattered - algorithm - tsp\_blocked - - Force generic transport blocked algorithm - tsp\_inplace - - Force generic transport inplace algorithm - **Default:** MPIR\_CVAR\_IALLTOALLV\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_IALLTOALLV\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLTOALLV\_INTER\_ALGORITHM - MPICH\_IALLTOALLV\_INTER\_ALGORITHM - **Description:** - Variable to select ialltoallv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_pairwise\_exchange - Force pairwise exchange algorithm - **Default:** MPIR\_CVAR\_IALLTOALLV\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_IALLTOALLV\_SCATTERED\_OUTSTANDING\_TASKS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLTOALLV\_SCATTERED\_OUTSTANDING\_TASKS - MPICH\_IALLTOALLV\_SCATTERED\_OUTSTANDING\_TASKS - **Description:** - Maximum number of outstanding sends and recvs posted at a - time - **Default:** 64 MPIR\_CVAR\_IALLTOALLV\_SCATTERED\_BATCH\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLTOALLV\_SCATTERED\_BATCH\_SIZE - MPICH\_IALLTOALLV\_SCATTERED\_BATCH\_SIZE - **Description:** - Number of send/receive tasks that scattered algorithm - waits for completion before posting another batch of send/receives - of that size - **Default:** 4 MPIR\_CVAR\_ALLTOALLW\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLTOALLW\_INTRA\_ALGORITHM - MPICH\_ALLTOALLW\_INTRA\_ALGORITHM - **Description:** - Variable to select alltoallw algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - dynamic - Internal algorithm selection using dynamic tuning - nb - - - - Force nonblocking algorithm - pairwise\_sendrecv\_replace - Force pairwise sendrecv replace - algorithm - scattered - - - Force scattered algorithm - **Default:** MPIR\_CVAR\_ALLTOALLW\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_ALLTOALLW\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLTOALLW\_INTER\_ALGORITHM - MPICH\_ALLTOALLW\_INTER\_ALGORITHM - **Description:** - Variable to select alltoallw algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - nb - - - Force nonblocking algorithm - pairwise\_exchange - Force pairwise exchange algorithm - **Default:** MPIR\_CVAR\_ALLTOALLW\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_IALLTOALLW\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLTOALLW\_INTRA\_ALGORITHM - MPICH\_IALLTOALLW\_INTRA\_ALGORITHM - **Description:** - Variable to select ialltoallw algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_blocked - - Force blocked algorithm - sched\_inplace - - Force inplace algorithm - tsp\_blocked - Force generic transport based blocked algorithm - tsp\_inplace - Force generic transport based inplace algorithm - **Default:** MPIR\_CVAR\_IALLTOALLW\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_IALLTOALLW\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLTOALLW\_INTER\_ALGORITHM - MPICH\_IALLTOALLW\_INTER\_ALGORITHM - **Description:** - Variable to select ialltoallw algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_pairwise\_exchange - Force pairwise exchange algorithm - **Default:** MPIR\_CVAR\_IALLTOALLW\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_REDUCE\_SHORT\_MSG\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_SHORT\_MSG\_SIZE - MPICH\_REDUCE\_SHORT\_MSG\_SIZE - **Description:** - the short message algorithm will be used if the send - buffer size is <= this value (in bytes) - **Default:** 2048 MPIR\_CVAR\_MAX\_SMP\_REDUCE\_MSG\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_MAX\_SMP\_REDUCE\_MSG\_SIZE - MPICH\_MAX\_SMP\_REDUCE\_MSG\_SIZE - **Description:** - Maximum message size for which SMP-aware reduce is used. - A value of '0' uses SMP-aware reduce for all message sizes. - **Default:** 0 MPIR\_CVAR\_REDUCE\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_INTRA\_ALGORITHM - MPICH\_REDUCE\_INTRA\_ALGORITHM - **Description:** - Variable to select reduce algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - dynamic - Internal algorithm selection using dynamic tuning - binomial - - Force binomial algorithm - nb - - - Force nonblocking algorithm - smp - - - Force smp algorithm - reduce\_scatter\_gather - Force reduce scatter gather algorithm - osu\_knomial - - Force MVAPICH knomial algorithm - osu\_allreduce - - Force MVAPICH allreduce algorithm - **Default:** MPIR\_CVAR\_REDUCE\_INTRA\_ALGORITHM\_dynamic MPIR\_CVAR\_REDUCE\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_INTER\_ALGORITHM - MPICH\_REDUCE\_INTER\_ALGORITHM - **Description:** - Variable to select reduce algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - local\_reduce\_remote\_send - Force local-reduce-remote-send algorithm - nb - - - Force nonblocking algorithm - **Default:** MPIR\_CVAR\_REDUCE\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_IREDUCE\_TREE\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IREDUCE\_TREE\_KVAL - MPICH\_IREDUCE\_TREE\_KVAL - **Description:** - k value for tree (kary, knomial, etc.) based ireduce - **Default:** 2 MPIR\_CVAR\_IREDUCE\_TREE\_TYPE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IREDUCE\_TREE\_TYPE - MPICH\_IREDUCE\_TREE\_TYPE - **Description:** - Tree type for tree based ireduce kary - kary tree - knomial\_1 - knomial\_1 tree knomial\_2 - knomial\_2 tree - topology\_aware - topology\_aware tree type topology\_aware\_k - - topology\_aware tree type with branching factor k topology\_wave - - topology\_wave tree type - **Default:** "kary" MPIR\_CVAR\_IREDUCE\_TOPO\_REORDER\_ENABLE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IREDUCE\_TOPO\_REORDER\_ENABLE - MPICH\_IREDUCE\_TOPO\_REORDER\_ENABLE - **Description:** - This cvar controls if the leaders are reordered based on - the number of ranks in each group. - **Default:** 1 MPIR\_CVAR\_IREDUCE\_TOPO\_OVERHEAD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IREDUCE\_TOPO\_OVERHEAD - MPICH\_IREDUCE\_TOPO\_OVERHEAD - **Description:** - This cvar controls the size of the overhead. - **Default:** 200 MPIR\_CVAR\_IREDUCE\_TOPO\_DIFF\_GROUPS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IREDUCE\_TOPO\_DIFF\_GROUPS - MPICH\_IREDUCE\_TOPO\_DIFF\_GROUPS - **Description:** - This cvar controls the latency between different groups. - **Default:** 2800 MPIR\_CVAR\_IREDUCE\_TOPO\_DIFF\_SWITCHES ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IREDUCE\_TOPO\_DIFF\_SWITCHES - MPICH\_IREDUCE\_TOPO\_DIFF\_SWITCHES - **Description:** - This cvar controls the latency between different switches - in the same groups. - **Default:** 1900 MPIR\_CVAR\_IREDUCE\_TOPO\_SAME\_SWITCHES ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IREDUCE\_TOPO\_SAME\_SWITCHES - MPICH\_IREDUCE\_TOPO\_SAME\_SWITCHES - **Description:** - This cvar controls the latency in the same switch. - **Default:** 1600 MPIR\_CVAR\_IREDUCE\_TREE\_PIPELINE\_CHUNK\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IREDUCE\_TREE\_PIPELINE\_CHUNK\_SIZE - MPICH\_IREDUCE\_TREE\_PIPELINE\_CHUNK\_SIZE - **Description:** - Maximum chunk size (in bytes) for pipelining in tree based - ireduce. Default value is 0, that is, no pipelining by default - **Default:** -1 MPIR\_CVAR\_IREDUCE\_RING\_CHUNK\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IREDUCE\_RING\_CHUNK\_SIZE - MPICH\_IREDUCE\_RING\_CHUNK\_SIZE - **Description:** - Maximum chunk size (in bytes) for pipelining in ireduce - ring algorithm. Default value is 0, that is, no pipelining by - default - **Default:** 0 MPIR\_CVAR\_IREDUCE\_TREE\_BUFFER\_PER\_CHILD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IREDUCE\_TREE\_BUFFER\_PER\_CHILD - MPICH\_IREDUCE\_TREE\_BUFFER\_PER\_CHILD - **Description:** - If set to true, a rank in tree algorithms will allocate a - dedicated buffer for every child it receives data from. This would - mean more memory consumption but it would allow preposting of the - receives and hence reduce the number of unexpected messages. If set - to false, there is only one buffer that is used to receive the data - from all the children. The receives are therefore serialized, that - is, only one receive can be posted at a time. - **Default:** 0 MPIR\_CVAR\_IREDUCE\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IREDUCE\_INTRA\_ALGORITHM - MPICH\_IREDUCE\_INTRA\_ALGORITHM - **Description:** - Variable to select ireduce algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_smp - - - Force smp algorithm - sched\_binomial - - Force binomial algorithm - sched\_reduce\_scatter\_gather - Force reduce scatter gather algorithm - tsp\_tree - - - Force Generic Transport Tree - tsp\_ring - - - Force Generic Transport Ring - **Default:** MPIR\_CVAR\_IREDUCE\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_IREDUCE\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IREDUCE\_INTER\_ALGORITHM - MPICH\_IREDUCE\_INTER\_ALGORITHM - **Description:** - Variable to select ireduce algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_local\_reduce\_remote\_send - Force local-reduce-remote-send - algorithm - **Default:** MPIR\_CVAR\_IREDUCE\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_ALLREDUCE\_SHORT\_MSG\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_SHORT\_MSG\_SIZE - MPICH\_ALLREDUCE\_SHORT\_MSG\_SIZE - **Description:** - the short message algorithm will be used if the send - buffer size is <= this value (in bytes) - **Default:** 2048 MPIR\_CVAR\_MAX\_SMP\_ALLREDUCE\_MSG\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_MAX\_SMP\_ALLREDUCE\_MSG\_SIZE - MPICH\_MAX\_SMP\_ALLREDUCE\_MSG\_SIZE - **Description:** - Maximum message size for which SMP-aware allreduce is - used. A value of '0' uses SMP-aware allreduce for all message - sizes. - **Default:** 0 MPIR\_CVAR\_ALLREDUCE\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_INTRA\_ALGORITHM - MPICH\_ALLREDUCE\_INTRA\_ALGORITHM - **Description:** - Variable to select allreduce algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - dynamic - Internal algorithm selection using dynamic tuning - nb - - - Force nonblocking algorithm - smp - - - Force smp algorithm - recursive\_doubling - Force recursive doubling algorithm - recursive\_multiplying - Force recursive multiplying algorithm - reduce\_scatter\_allgather - Force reduce scatter allgather algorithm - tree - - - Force pipelined tree algorithm - recexch - - - Force generic transport recursive - exchange algorithm - ring - - - Force ring algorithm - k\_reduce\_scatter\_allgather - Force reduce scatter allgather - algorithm - osu\_rd\_compression - Force GPU compression recursive doubling - osu\_rsa\_ring\_compression - Force GPU compression ring - osu\_gpu\_rd - - Force recursive doubling algorithm using - GPU buffer - osu\_gpu\_rsa - - Force reduce scatter allgather algorithm - using GPU buffer - osu\_gpu\_krsa - - Force k reduce scatter allgather - algorithm - using GPU buffer - osu\_gpu\_ring - - Force ring RSA algorithm using GPU - buffer - osu\_gpu\_direct - - Force direct RSA algorithm using GPU - buffer - **Default:** MPIR\_CVAR\_ALLREDUCE\_INTRA\_ALGORITHM\_dynamic MPIR\_CVAR\_ALLREDUCE\_RECURSIVE\_MULTIPLYING\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_RECURSIVE\_MULTIPLYING\_KVAL - MPICH\_ALLREDUCE\_RECURSIVE\_MULTIPLYING\_KVAL - **Description:** - radix value (power of k) for generic recursive multiplying - based allreduce - **Default:** 2 MPIR\_CVAR\_ALLREDUCE\_TREE\_TYPE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_TREE\_TYPE - MPICH\_ALLREDUCE\_TREE\_TYPE - **Description:** - Tree type for tree based allreduce knomial\_1 is default as - it supports both commutative and non-commutative reduce operations - kary - kary tree type knomial\_1 - knomial\_1 tree type (tree - grows starting from the left of the root) knomial\_2 - knomial\_2 - tree type (tree grows starting from the right of the root) - topology\_aware - topology\_aware tree type topology\_aware\_k - - topology\_aware tree type with branching factor k topology\_wave - - topology\_wave tree type - **Default:** "knomial\_1" MPIR\_CVAR\_ALLREDUCE\_TREE\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_TREE\_KVAL - MPICH\_ALLREDUCE\_TREE\_KVAL - **Description:** - Indicates the branching factor for kary or knomial trees. - **Default:** 2 MPIR\_CVAR\_ALLREDUCE\_TOPO\_REORDER\_ENABLE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_TOPO\_REORDER\_ENABLE - MPICH\_ALLREDUCE\_TOPO\_REORDER\_ENABLE - **Description:** - This cvar controls if the leaders are reordered based on - the number of ranks in each group. - **Default:** 1 MPIR\_CVAR\_ALLREDUCE\_TOPO\_OVERHEAD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_TOPO\_OVERHEAD - MPICH\_ALLREDUCE\_TOPO\_OVERHEAD - **Description:** - This cvar controls the size of the overhead. - **Default:** 200 MPIR\_CVAR\_ALLREDUCE\_TOPO\_DIFF\_GROUPS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_TOPO\_DIFF\_GROUPS - MPICH\_ALLREDUCE\_TOPO\_DIFF\_GROUPS - **Description:** - This cvar controls the latency between different groups. - **Default:** 2800 MPIR\_CVAR\_ALLREDUCE\_TOPO\_DIFF\_SWITCHES ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_TOPO\_DIFF\_SWITCHES - MPICH\_ALLREDUCE\_TOPO\_DIFF\_SWITCHES - **Description:** - This cvar controls the latency between different switches - in the same groups. - **Default:** 1900 MPIR\_CVAR\_ALLREDUCE\_TOPO\_SAME\_SWITCHES ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_TOPO\_SAME\_SWITCHES - MPICH\_ALLREDUCE\_TOPO\_SAME\_SWITCHES - **Description:** - This cvar controls the latency in the same switch. - **Default:** 1600 MPIR\_CVAR\_ALLREDUCE\_TREE\_PIPELINE\_CHUNK\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_TREE\_PIPELINE\_CHUNK\_SIZE - MPICH\_ALLREDUCE\_TREE\_PIPELINE\_CHUNK\_SIZE - **Description:** - Maximum chunk size (in bytes) for pipelining in tree based - allreduce. Default value is 0, that is, no pipelining by default - **Default:** 0 MPIR\_CVAR\_ALLREDUCE\_TREE\_BUFFER\_PER\_CHILD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_TREE\_BUFFER\_PER\_CHILD - MPICH\_ALLREDUCE\_TREE\_BUFFER\_PER\_CHILD - **Description:** - If set to true, a rank in tree\_kary and tree\_knomial - algorithms will allocate a dedicated buffer for every child it - receives data from. This would mean more memory consumption but it - would allow preposting of the receives and hence reduce the number - of unexpected messages. If set to false, there is only one buffer - that is used to receive the data from all the children. The - receives are therefore serialized, that is, only one receive can be - posted at a time. - **Default:** 0 MPIR\_CVAR\_ALLREDUCE\_RECEXCH\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_RECEXCH\_KVAL - MPICH\_ALLREDUCE\_RECEXCH\_KVAL - **Description:** - k value for recursive exchange based allreduce - **Default:** 2 MPIR\_CVAR\_ALLREDUCE\_RECEXCH\_SINGLE\_PHASE\_RECV ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_RECEXCH\_SINGLE\_PHASE\_RECV - MPICH\_ALLREDUCE\_RECEXCH\_SINGLE\_PHASE\_RECV - **Description:** - This CVAR controls whether the recv is posted for one - phase or two phases in recexch algos. By default, we post the recvs - for 2 phases. - **Default:** 0 MPIR\_CVAR\_ALLREDUCE\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_INTER\_ALGORITHM - MPICH\_ALLREDUCE\_INTER\_ALGORITHM - **Description:** - Variable to select allreduce algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - nb - - - Force nonblocking algorithm - reduce\_exchange\_bcast - Force reduce-exchange-bcast algorithm - **Default:** MPIR\_CVAR\_ALLREDUCE\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_IALLREDUCE\_TREE\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLREDUCE\_TREE\_KVAL - MPICH\_IALLREDUCE\_TREE\_KVAL - **Description:** - k value for tree based iallreduce (for tree\_kary and - tree\_knomial) - **Default:** 2 MPIR\_CVAR\_IALLREDUCE\_TREE\_TYPE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLREDUCE\_TREE\_TYPE - MPICH\_IALLREDUCE\_TREE\_TYPE - **Description:** - Tree type for tree based ibcast kary - kary tree type - knomial\_1 - knomial\_1 tree type knomial\_2 - knomial\_2 tree type - **Default:** "kary" MPIR\_CVAR\_IALLREDUCE\_TREE\_PIPELINE\_CHUNK\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLREDUCE\_TREE\_PIPELINE\_CHUNK\_SIZE - MPICH\_IALLREDUCE\_TREE\_PIPELINE\_CHUNK\_SIZE - **Description:** - Maximum chunk size (in bytes) for pipelining in tree based - iallreduce. Default value is 0, that is, no pipelining by default - **Default:** 0 MPIR\_CVAR\_IALLREDUCE\_TREE\_BUFFER\_PER\_CHILD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLREDUCE\_TREE\_BUFFER\_PER\_CHILD - MPICH\_IALLREDUCE\_TREE\_BUFFER\_PER\_CHILD - **Description:** - If set to true, a rank in tree\_kary and tree\_knomial - algorithms will allocate a dedicated buffer for every child it - receives data from. This would mean more memory consumption but it - would allow preposting of the receives and hence reduce the number - of unexpected messages. If set to false, there is only one buffer - that is used to receive the data from all the children. The - receives are therefore serialized, that is, only one receive can be - posted at a time. - **Default:** 0 MPIR\_CVAR\_IALLREDUCE\_RECEXCH\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLREDUCE\_RECEXCH\_KVAL - MPICH\_IALLREDUCE\_RECEXCH\_KVAL - **Description:** - k value for recursive exchange based iallreduce - **Default:** 2 MPIR\_CVAR\_IALLREDUCE\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLREDUCE\_INTRA\_ALGORITHM - MPICH\_IALLREDUCE\_INTRA\_ALGORITHM - **Description:** - Variable to select iallreduce algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_naive - - - Force naive algorithm - sched\_smp - - - - Force smp algorithm - sched\_recursive\_doubling - - Force recursive doubling - algorithm - sched\_reduce\_scatter\_allgather - Force reduce scatter allgather - algorithm - tsp\_recexch\_single\_buffer - Force generic transport recursive - exchange with single buffer for receives - tsp\_recexch\_multiple\_buffer - Force generic transport recursive - exchange with multiple buffers for receives - tsp\_tree - - - Force generic transport tree - algorithm - tsp\_ring - - - Force generic transport ring - algorithm - tsp\_recexch\_reduce\_scatter\_recexch\_allgatherv - Force generic - transport recursive exchange with reduce scatter and allgatherv - **Default:** MPIR\_CVAR\_IALLREDUCE\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_IALLREDUCE\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLREDUCE\_INTER\_ALGORITHM - MPICH\_IALLREDUCE\_INTER\_ALGORITHM - **Description:** - Variable to select iallreduce algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_remote\_reduce\_local\_bcast - Force remote-reduce-local-bcast - algorithm - **Default:** MPIR\_CVAR\_IALLREDUCE\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_REDUCE\_SCATTER\_COMMUTATIVE\_LONG\_MSG\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_SCATTER\_COMMUTATIVE\_LONG\_MSG\_SIZE - MPICH\_REDUCE\_SCATTER\_COMMUTATIVE\_LONG\_MSG\_SIZE - **Description:** - the long message algorithm will be used if the operation - is commutative and the send buffer size is >= this value (in bytes) - **Default:** 524288 MPIR\_CVAR\_REDUCE\_SCATTER\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_SCATTER\_INTRA\_ALGORITHM - MPICH\_REDUCE\_SCATTER\_INTRA\_ALGORITHM - **Description:** - Variable to select reduce\_scatter algorithm - auto - - - Internal algorithm selection (can be - overridden with MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - dynamic - Internal algorithm selection using dynamic tuning - nb - - - Force nonblocking algorithm - noncommutative - Force noncommutative algorithm - pairwise - - Force pairwise algorithm - recursive\_doubling - Force recursive doubling algorithm - recursive\_halving - Force recursive halving algorithm - osu\_basic - - Force MVAPICH basic algorithm - osu\_ring - - Force MVAPICH ring algorithm - osu\_ring\_compression - Force compression algorithm - osu\_gpu\_direct - Force MVAPICH gpu direct algorithm - osu\_gpu\_ring - - Force MVAPICH gpu ring algorithm - **Default:** MPIR\_CVAR\_REDUCE\_SCATTER\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_REDUCE\_SCATTER\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_SCATTER\_INTER\_ALGORITHM - MPICH\_REDUCE\_SCATTER\_INTER\_ALGORITHM - **Description:** - Variable to select reduce\_scatter algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - nb - - - - Force nonblocking algorithm - remote\_reduce\_local\_scatter - Force remote-reduce-local-scatter - algorithm - **Default:** MPIR\_CVAR\_REDUCE\_SCATTER\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_IREDUCE\_SCATTER\_RECEXCH\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IREDUCE\_SCATTER\_RECEXCH\_KVAL - MPICH\_IREDUCE\_SCATTER\_RECEXCH\_KVAL - **Description:** - k value for recursive exchange based ireduce\_scatter - **Default:** 2 MPIR\_CVAR\_IREDUCE\_SCATTER\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IREDUCE\_SCATTER\_INTRA\_ALGORITHM - MPICH\_IREDUCE\_SCATTER\_INTRA\_ALGORITHM - **Description:** - Variable to select ireduce\_scatter algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_noncommutative - Force noncommutative algorithm - sched\_recursive\_doubling - Force recursive doubling algorithm - sched\_pairwise - - Force pairwise algorithm - sched\_recursive\_halving - Force recursive halving algorithm - tsp\_recexch - - Force generic transport recursive exchange - algorithm - **Default:** MPIR\_CVAR\_IREDUCE\_SCATTER\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_IREDUCE\_SCATTER\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IREDUCE\_SCATTER\_INTER\_ALGORITHM - MPICH\_IREDUCE\_SCATTER\_INTER\_ALGORITHM - **Description:** - Variable to select ireduce\_scatter algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_remote\_reduce\_local\_scatterv - Force - remote-reduce-local-scatterv algorithm - **Default:** MPIR\_CVAR\_IREDUCE\_SCATTER\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_REDUCE\_SCATTER\_BLOCK\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_SCATTER\_BLOCK\_INTRA\_ALGORITHM - MPICH\_REDUCE\_SCATTER\_BLOCK\_INTRA\_ALGORITHM - **Description:** - Variable to select reduce\_scatter\_block algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - dynamic - Internal algorithm selection using dynamic tuning - noncommutative - Force noncommutative algorithm - recursive\_doubling - Force recursive doubling algorithm - pairwise - - Force pairwise algorithm - recursive\_halving - Force recursive halving algorithm - nb - - - Force nonblocking algorithm - **Default:** MPIR\_CVAR\_REDUCE\_SCATTER\_BLOCK\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_REDUCE\_SCATTER\_BLOCK\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_SCATTER\_BLOCK\_INTER\_ALGORITHM - MPICH\_REDUCE\_SCATTER\_BLOCK\_INTER\_ALGORITHM - **Description:** - Variable to select reduce\_scatter\_block algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - nb - - - - Force nonblocking algorithm - remote\_reduce\_local\_scatter - Force remote-reduce-local-scatter - algorithm - **Default:** MPIR\_CVAR\_REDUCE\_SCATTER\_BLOCK\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_IREDUCE\_SCATTER\_BLOCK\_RECEXCH\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IREDUCE\_SCATTER\_BLOCK\_RECEXCH\_KVAL - MPICH\_IREDUCE\_SCATTER\_BLOCK\_RECEXCH\_KVAL - **Description:** - k value for recursive exchange based ireduce\_scatter\_block - **Default:** 2 MPIR\_CVAR\_IREDUCE\_SCATTER\_BLOCK\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IREDUCE\_SCATTER\_BLOCK\_INTRA\_ALGORITHM - MPICH\_IREDUCE\_SCATTER\_BLOCK\_INTRA\_ALGORITHM - **Description:** - Variable to select ireduce\_scatter\_block algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_noncommutative - Force noncommutative algorithm - sched\_recursive\_doubling - Force recursive doubling algorithm - sched\_pairwise - - Force pairwise algorithm - sched\_recursive\_halving - Force recursive halving algorithm - tsp\_recexch - - Force generic transport recursive exchange - algorithm - **Default:** MPIR\_CVAR\_IREDUCE\_SCATTER\_BLOCK\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_IREDUCE\_SCATTER\_BLOCK\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IREDUCE\_SCATTER\_BLOCK\_INTER\_ALGORITHM - MPICH\_IREDUCE\_SCATTER\_BLOCK\_INTER\_ALGORITHM - **Description:** - Variable to select ireduce\_scatter\_block algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_remote\_reduce\_local\_scatterv - Force - remote-reduce-local-scatterv algorithm - **Default:** MPIR\_CVAR\_IREDUCE\_SCATTER\_BLOCK\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_SCAN\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_SCAN\_INTRA\_ALGORITHM - MPICH\_SCAN\_INTRA\_ALGORITHM - **Description:** - Variable to select allgather algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - dynamic - Internal algorithm selection using dynamic tuning - nb - - - Force nonblocking algorithm - smp - - - Force smp algorithm - recursive\_doubling - Force recursive doubling algorithm - **Default:** MPIR\_CVAR\_SCAN\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_ISCAN\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ISCAN\_INTRA\_ALGORITHM - MPICH\_ISCAN\_INTRA\_ALGORITHM - **Description:** - Variable to select allgather algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_smp - - - Force smp algorithm - sched\_recursive\_doubling - Force recursive doubling algorithm - tsp\_recursive\_doubling - Force generic transport recursive doubling - algorithm - **Default:** MPIR\_CVAR\_ISCAN\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_EXSCAN\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_EXSCAN\_INTRA\_ALGORITHM - MPICH\_EXSCAN\_INTRA\_ALGORITHM - **Description:** - Variable to select allgather algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - dynamic - Internal algorithm selection using dynamic tuning - nb - - - Force nonblocking algorithm - recursive\_doubling - Force recursive doubling algorithm - **Default:** MPIR\_CVAR\_EXSCAN\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_IEXSCAN\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IEXSCAN\_INTRA\_ALGORITHM - MPICH\_IEXSCAN\_INTRA\_ALGORITHM - **Description:** - Variable to select iexscan algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_recursive\_doubling - Force recursive doubling algorithm - **Default:** MPIR\_CVAR\_IEXSCAN\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_NEIGHBOR\_ALLGATHER\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEIGHBOR\_ALLGATHER\_INTRA\_ALGORITHM - MPICH\_NEIGHBOR\_ALLGATHER\_INTRA\_ALGORITHM - **Description:** - Variable to select ineighbor\_allgather algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - nb - Force nonblocking algorithm - **Default:** MPIR\_CVAR\_NEIGHBOR\_ALLGATHER\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_NEIGHBOR\_ALLGATHER\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEIGHBOR\_ALLGATHER\_INTER\_ALGORITHM - MPICH\_NEIGHBOR\_ALLGATHER\_INTER\_ALGORITHM - **Description:** - Variable to select ineighbor\_allgather algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - nb - Force nonblocking algorithm - **Default:** MPIR\_CVAR\_NEIGHBOR\_ALLGATHER\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_INEIGHBOR\_ALLGATHER\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_INEIGHBOR\_ALLGATHER\_INTRA\_ALGORITHM - MPICH\_INEIGHBOR\_ALLGATHER\_INTRA\_ALGORITHM - **Description:** - Variable to select ineighbor\_allgather algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_linear - Force linear algorithm - tsp\_linear - Force generic transport based linear algorithm - **Default:** MPIR\_CVAR\_INEIGHBOR\_ALLGATHER\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_INEIGHBOR\_ALLGATHER\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_INEIGHBOR\_ALLGATHER\_INTER\_ALGORITHM - MPICH\_INEIGHBOR\_ALLGATHER\_INTER\_ALGORITHM - **Description:** - Variable to select ineighbor\_allgather algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_linear - Force linear algorithm - tsp\_linear - Force generic transport based linear algorithm - **Default:** MPIR\_CVAR\_INEIGHBOR\_ALLGATHER\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_NEIGHBOR\_ALLGATHERV\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEIGHBOR\_ALLGATHERV\_INTRA\_ALGORITHM - MPICH\_NEIGHBOR\_ALLGATHERV\_INTRA\_ALGORITHM - **Description:** - Variable to select neighbor\_allgatherv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - nb - Force nb algorithm - **Default:** MPIR\_CVAR\_NEIGHBOR\_ALLGATHERV\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_NEIGHBOR\_ALLGATHERV\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEIGHBOR\_ALLGATHERV\_INTER\_ALGORITHM - MPICH\_NEIGHBOR\_ALLGATHERV\_INTER\_ALGORITHM - **Description:** - Variable to select neighbor\_allgatherv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - nb - Force nb algorithm - **Default:** MPIR\_CVAR\_NEIGHBOR\_ALLGATHERV\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_INEIGHBOR\_ALLGATHERV\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_INEIGHBOR\_ALLGATHERV\_INTRA\_ALGORITHM - MPICH\_INEIGHBOR\_ALLGATHERV\_INTRA\_ALGORITHM - **Description:** - Variable to select ineighbor\_allgatherv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_linear - - Force linear algorithm - tsp\_linear - - Force generic transport based linear algorithm - **Default:** MPIR\_CVAR\_INEIGHBOR\_ALLGATHERV\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_INEIGHBOR\_ALLGATHERV\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_INEIGHBOR\_ALLGATHERV\_INTER\_ALGORITHM - MPICH\_INEIGHBOR\_ALLGATHERV\_INTER\_ALGORITHM - **Description:** - Variable to select ineighbor\_allgatherv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_linear - - Force linear algorithm - tsp\_linear - - Force generic transport based linear algorithm - **Default:** MPIR\_CVAR\_INEIGHBOR\_ALLGATHERV\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_NEIGHBOR\_ALLTOALL\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEIGHBOR\_ALLTOALL\_INTRA\_ALGORITHM - MPICH\_NEIGHBOR\_ALLTOALL\_INTRA\_ALGORITHM - **Description:** - Variable to select neighbor\_alltoall algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - nb - Force nb algorithm - **Default:** MPIR\_CVAR\_NEIGHBOR\_ALLTOALL\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_NEIGHBOR\_ALLTOALL\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEIGHBOR\_ALLTOALL\_INTER\_ALGORITHM - MPICH\_NEIGHBOR\_ALLTOALL\_INTER\_ALGORITHM - **Description:** - Variable to select neighbor\_alltoall algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - nb - Force nb algorithm - **Default:** MPIR\_CVAR\_NEIGHBOR\_ALLTOALL\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_INEIGHBOR\_ALLTOALL\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_INEIGHBOR\_ALLTOALL\_INTRA\_ALGORITHM - MPICH\_INEIGHBOR\_ALLTOALL\_INTRA\_ALGORITHM - **Description:** - Variable to select ineighbor\_alltoall algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_linear - - Force linear algorithm - tsp\_linear - - Force generic transport based linear algorithm - **Default:** MPIR\_CVAR\_INEIGHBOR\_ALLTOALL\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_INEIGHBOR\_ALLTOALL\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_INEIGHBOR\_ALLTOALL\_INTER\_ALGORITHM - MPICH\_INEIGHBOR\_ALLTOALL\_INTER\_ALGORITHM - **Description:** - Variable to select ineighbor\_alltoall algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_linear - - Force linear algorithm - tsp\_linear - - Force generic transport based linear algorithm - **Default:** MPIR\_CVAR\_INEIGHBOR\_ALLTOALL\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_NEIGHBOR\_ALLTOALLV\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEIGHBOR\_ALLTOALLV\_INTRA\_ALGORITHM - MPICH\_NEIGHBOR\_ALLTOALLV\_INTRA\_ALGORITHM - **Description:** - Variable to select neighbor\_alltoallv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - nb - Force nb algorithm - **Default:** MPIR\_CVAR\_NEIGHBOR\_ALLTOALLV\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_NEIGHBOR\_ALLTOALLV\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEIGHBOR\_ALLTOALLV\_INTER\_ALGORITHM - MPICH\_NEIGHBOR\_ALLTOALLV\_INTER\_ALGORITHM - **Description:** - Variable to select neighbor\_alltoallv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - nb - Force nb algorithm - **Default:** MPIR\_CVAR\_NEIGHBOR\_ALLTOALLV\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_INEIGHBOR\_ALLTOALLV\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_INEIGHBOR\_ALLTOALLV\_INTRA\_ALGORITHM - MPICH\_INEIGHBOR\_ALLTOALLV\_INTRA\_ALGORITHM - **Description:** - Variable to select ineighbor\_alltoallv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_linear - - Force linear algorithm - tsp\_linear - Force generic transport based linear algorithm - **Default:** MPIR\_CVAR\_INEIGHBOR\_ALLTOALLV\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_INEIGHBOR\_ALLTOALLV\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_INEIGHBOR\_ALLTOALLV\_INTER\_ALGORITHM - MPICH\_INEIGHBOR\_ALLTOALLV\_INTER\_ALGORITHM - **Description:** - Variable to select ineighbor\_alltoallv algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_linear - - Force linear algorithm - tsp\_linear - Force generic transport based linear algorithm - **Default:** MPIR\_CVAR\_INEIGHBOR\_ALLTOALLV\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_NEIGHBOR\_ALLTOALLW\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEIGHBOR\_ALLTOALLW\_INTRA\_ALGORITHM - MPICH\_NEIGHBOR\_ALLTOALLW\_INTRA\_ALGORITHM - **Description:** - Variable to select neighbor\_alltoallw algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - nb - Force nb algorithm - **Default:** MPIR\_CVAR\_NEIGHBOR\_ALLTOALLW\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_NEIGHBOR\_ALLTOALLW\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEIGHBOR\_ALLTOALLW\_INTER\_ALGORITHM - MPICH\_NEIGHBOR\_ALLTOALLW\_INTER\_ALGORITHM - **Description:** - Variable to select neighbor\_alltoallw algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - nb - Force nb algorithm - **Default:** MPIR\_CVAR\_NEIGHBOR\_ALLTOALLW\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_INEIGHBOR\_ALLTOALLW\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_INEIGHBOR\_ALLTOALLW\_INTRA\_ALGORITHM - MPICH\_INEIGHBOR\_ALLTOALLW\_INTRA\_ALGORITHM - **Description:** - Variable to select ineighbor\_alltoallw algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_linear - - Force linear algorithm - tsp\_linear - - Force generic transport based linear algorithm - **Default:** MPIR\_CVAR\_INEIGHBOR\_ALLTOALLW\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_INEIGHBOR\_ALLTOALLW\_INTER\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_INEIGHBOR\_ALLTOALLW\_INTER\_ALGORITHM - MPICH\_INEIGHBOR\_ALLTOALLW\_INTER\_ALGORITHM - **Description:** - Variable to select ineighbor\_alltoallw algorithm - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - sched\_auto - Internal algorithm selection for sched-based - algorithms - sched\_linear - - Force linear algorithm - tsp\_linear - - Force generic transport based linear algorithm - **Default:** MPIR\_CVAR\_INEIGHBOR\_ALLTOALLW\_INTER\_ALGORITHM\_auto MPIR\_CVAR\_BARRIER\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BARRIER\_DEVICE\_COLLECTIVE - MPICH\_BARRIER\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Barrier will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_IBARRIER\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IBARRIER\_DEVICE\_COLLECTIVE - MPICH\_IBARRIER\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Ibarrier will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_BARRIER\_INIT\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BARRIER\_INIT\_DEVICE\_COLLECTIVE - MPICH\_BARRIER\_INIT\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Barrier will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_BCAST\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_DEVICE\_COLLECTIVE - MPICH\_BCAST\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Bcast will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_IBCAST\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IBCAST\_DEVICE\_COLLECTIVE - MPICH\_IBCAST\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Ibcast will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_BCAST\_INIT\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_INIT\_DEVICE\_COLLECTIVE - MPICH\_BCAST\_INIT\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Bcast\_init will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_GATHER\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_GATHER\_DEVICE\_COLLECTIVE - MPICH\_GATHER\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Gather will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_IGATHER\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IGATHER\_DEVICE\_COLLECTIVE - MPICH\_IGATHER\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Igather will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_GATHER\_INIT\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_GATHER\_INIT\_DEVICE\_COLLECTIVE - MPICH\_GATHER\_INIT\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Gather\_init will allow - the device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_GATHERV\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_GATHERV\_DEVICE\_COLLECTIVE - MPICH\_GATHERV\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Gatherv will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_IGATHERV\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IGATHERV\_DEVICE\_COLLECTIVE - MPICH\_IGATHERV\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Igatherv will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_GATHERV\_INIT\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_GATHERV\_INIT\_DEVICE\_COLLECTIVE - MPICH\_GATHERV\_INIT\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Gatherv\_init will allow - the device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_SCATTER\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_SCATTER\_DEVICE\_COLLECTIVE - MPICH\_SCATTER\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Scatter will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_ISCATTER\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ISCATTER\_DEVICE\_COLLECTIVE - MPICH\_ISCATTER\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Iscatter will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_SCATTER\_INIT\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_SCATTER\_INIT\_DEVICE\_COLLECTIVE - MPICH\_SCATTER\_INIT\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Scatter\_init will allow - the device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_SCATTERV\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_SCATTERV\_DEVICE\_COLLECTIVE - MPICH\_SCATTERV\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Scatterv will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_ISCATTERV\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ISCATTERV\_DEVICE\_COLLECTIVE - MPICH\_ISCATTERV\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Iscatterv will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_SCATTERV\_INIT\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_SCATTERV\_INIT\_DEVICE\_COLLECTIVE - MPICH\_SCATTERV\_INIT\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Scatterv\_init will allow - the device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_ALLGATHER\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLGATHER\_DEVICE\_COLLECTIVE - MPICH\_ALLGATHER\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Allgather will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_IALLGATHER\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLGATHER\_DEVICE\_COLLECTIVE - MPICH\_IALLGATHER\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Iallgather will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_ALLGATHER\_INIT\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLGATHER\_INIT\_DEVICE\_COLLECTIVE - MPICH\_ALLGATHER\_INIT\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Allgather\_init will allow - the device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_ALLGATHERV\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLGATHERV\_DEVICE\_COLLECTIVE - MPICH\_ALLGATHERV\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Allgatherv will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_IALLGATHERV\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLGATHERV\_DEVICE\_COLLECTIVE - MPICH\_IALLGATHERV\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Iallgatherv will allow - the device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_ALLGATHERV\_INIT\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLGATHERV\_INIT\_DEVICE\_COLLECTIVE - MPICH\_ALLGATHERV\_INIT\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Allgatherv\_init will - allow the device to override the MPIR-level collective algorithms. - The device might still call the MPIR-level algorithms manually. If - set to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_ALLTOALL\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLTOALL\_DEVICE\_COLLECTIVE - MPICH\_ALLTOALL\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Alltoall will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_IALLTOALL\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLTOALL\_DEVICE\_COLLECTIVE - MPICH\_IALLTOALL\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Ialltoall will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_ALLTOALL\_INIT\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLTOALL\_INIT\_DEVICE\_COLLECTIVE - MPICH\_ALLTOALL\_INIT\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Alltoall\_init will allow - the device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_ALLTOALLV\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLTOALLV\_DEVICE\_COLLECTIVE - MPICH\_ALLTOALLV\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Alltoallv will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_IALLTOALLV\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLTOALLV\_DEVICE\_COLLECTIVE - MPICH\_IALLTOALLV\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Ialltoallv will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_ALLTOALLV\_INIT\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLTOALLV\_INIT\_DEVICE\_COLLECTIVE - MPICH\_ALLTOALLV\_INIT\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Alltoallv\_init will allow - the device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_ALLTOALLW\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLTOALLW\_DEVICE\_COLLECTIVE - MPICH\_ALLTOALLW\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Alltoallw will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_IALLTOALLW\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLTOALLW\_DEVICE\_COLLECTIVE - MPICH\_IALLTOALLW\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Ialltoallw will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_ALLTOALLW\_INIT\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLTOALLW\_INIT\_DEVICE\_COLLECTIVE - MPICH\_ALLTOALLW\_INIT\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Alltoallw\_init will allow - the device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_REDUCE\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_DEVICE\_COLLECTIVE - MPICH\_REDUCE\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Reduce will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_IREDUCE\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IREDUCE\_DEVICE\_COLLECTIVE - MPICH\_IREDUCE\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Ireduce will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_REDUCE\_INIT\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_INIT\_DEVICE\_COLLECTIVE - MPICH\_REDUCE\_INIT\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Reduce\_init will allow - the device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_ALLREDUCE\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_DEVICE\_COLLECTIVE - MPICH\_ALLREDUCE\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Allreduce will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_IALLREDUCE\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLREDUCE\_DEVICE\_COLLECTIVE - MPICH\_IALLREDUCE\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Iallreduce will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_ALLREDUCE\_INIT\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_INIT\_DEVICE\_COLLECTIVE - MPICH\_ALLREDUCE\_INIT\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Allreduce\_init will allow - the device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_REDUCE\_SCATTER\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_SCATTER\_DEVICE\_COLLECTIVE - MPICH\_REDUCE\_SCATTER\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Reduce\_scatter will allow - the device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_IREDUCE\_SCATTER\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IREDUCE\_SCATTER\_DEVICE\_COLLECTIVE - MPICH\_IREDUCE\_SCATTER\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Ireduce\_scatter will - allow the device to override the MPIR-level collective algorithms. - The device might still call the MPIR-level algorithms manually. If - set to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_REDUCE\_SCATTER\_INIT\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_SCATTER\_INIT\_DEVICE\_COLLECTIVE - MPICH\_REDUCE\_SCATTER\_INIT\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Reduce\_scatter\_init will - allow the device to override the MPIR-level collective algorithms. - The device might still call the MPIR-level algorithms manually. If - set to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_REDUCE\_SCATTER\_BLOCK\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_SCATTER\_BLOCK\_DEVICE\_COLLECTIVE - MPICH\_REDUCE\_SCATTER\_BLOCK\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Reduce\_scatter\_block will - allow the device to override the MPIR-level collective algorithms. - The device might still call the MPIR-level algorithms manually. If - set to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_IREDUCE\_SCATTER\_BLOCK\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IREDUCE\_SCATTER\_BLOCK\_DEVICE\_COLLECTIVE - MPICH\_IREDUCE\_SCATTER\_BLOCK\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Ireduce\_scatter\_block - will allow the device to override the MPIR-level collective - algorithms. The device might still call the MPIR-level algorithms - manually. If set to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_REDUCE\_SCATTER\_BLOCK\_INIT\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_SCATTER\_BLOCK\_INIT\_DEVICE\_COLLECTIVE - MPICH\_REDUCE\_SCATTER\_BLOCK\_INIT\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Reduce\_scatter\_block\_init - will allow the device to override the MPIR-level collective - algorithms. The device might still call the MPIR-level algorithms - manually. If set to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_SCAN\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_SCAN\_DEVICE\_COLLECTIVE - MPICH\_SCAN\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Scan will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_ISCAN\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ISCAN\_DEVICE\_COLLECTIVE - MPICH\_ISCAN\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Iscan will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_SCAN\_INIT\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_SCAN\_INIT\_DEVICE\_COLLECTIVE - MPICH\_SCAN\_INIT\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Scan\_init will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_EXSCAN\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_EXSCAN\_DEVICE\_COLLECTIVE - MPICH\_EXSCAN\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Exscan will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_IEXSCAN\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IEXSCAN\_DEVICE\_COLLECTIVE - MPICH\_IEXSCAN\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Iexscan will allow the - device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_EXSCAN\_INIT\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_EXSCAN\_INIT\_DEVICE\_COLLECTIVE - MPICH\_EXSCAN\_INIT\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Exscan\_init will allow - the device to override the MPIR-level collective algorithms. The - device might still call the MPIR-level algorithms manually. If set - to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_NEIGHBOR\_ALLGATHER\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEIGHBOR\_ALLGATHER\_DEVICE\_COLLECTIVE - MPICH\_NEIGHBOR\_ALLGATHER\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Neighbor\_allgather will - allow the device to override the MPIR-level collective algorithms. - The device might still call the MPIR-level algorithms manually. If - set to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_INEIGHBOR\_ALLGATHER\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_INEIGHBOR\_ALLGATHER\_DEVICE\_COLLECTIVE - MPICH\_INEIGHBOR\_ALLGATHER\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Ineighbor\_allgather will - allow the device to override the MPIR-level collective algorithms. - The device might still call the MPIR-level algorithms manually. If - set to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_NEIGHBOR\_ALLGATHER\_INIT\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEIGHBOR\_ALLGATHER\_INIT\_DEVICE\_COLLECTIVE - MPICH\_NEIGHBOR\_ALLGATHER\_INIT\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Neighbor\_allgather\_init - will allow the device to override the MPIR-level collective - algorithms. The device might still call the MPIR-level algorithms - manually. If set to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_NEIGHBOR\_ALLGATHERV\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEIGHBOR\_ALLGATHERV\_DEVICE\_COLLECTIVE - MPICH\_NEIGHBOR\_ALLGATHERV\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Neighbor\_allgatherv will - allow the device to override the MPIR-level collective algorithms. - The device might still call the MPIR-level algorithms manually. If - set to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_INEIGHBOR\_ALLGATHERV\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_INEIGHBOR\_ALLGATHERV\_DEVICE\_COLLECTIVE - MPICH\_INEIGHBOR\_ALLGATHERV\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Ineighbor\_allgatherv will - allow the device to override the MPIR-level collective algorithms. - The device might still call the MPIR-level algorithms manually. If - set to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_NEIGHBOR\_ALLGATHERV\_INIT\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEIGHBOR\_ALLGATHERV\_INIT\_DEVICE\_COLLECTIVE - MPICH\_NEIGHBOR\_ALLGATHERV\_INIT\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Neighbor\_allgatherv\_init - will allow the device to override the MPIR-level collective - algorithms. The device might still call the MPIR-level algorithms - manually. If set to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_NEIGHBOR\_ALLTOALL\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEIGHBOR\_ALLTOALL\_DEVICE\_COLLECTIVE - MPICH\_NEIGHBOR\_ALLTOALL\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Neighbor\_alltoall will - allow the device to override the MPIR-level collective algorithms. - The device might still call the MPIR-level algorithms manually. If - set to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_INEIGHBOR\_ALLTOALL\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_INEIGHBOR\_ALLTOALL\_DEVICE\_COLLECTIVE - MPICH\_INEIGHBOR\_ALLTOALL\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Ineighbor\_alltoall will - allow the device to override the MPIR-level collective algorithms. - The device might still call the MPIR-level algorithms manually. If - set to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_NEIGHBOR\_ALLTOALL\_INIT\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEIGHBOR\_ALLTOALL\_INIT\_DEVICE\_COLLECTIVE - MPICH\_NEIGHBOR\_ALLTOALL\_INIT\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Neighbor\_alltoall\_init - will allow the device to override the MPIR-level collective - algorithms. The device might still call the MPIR-level algorithms - manually. If set to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_NEIGHBOR\_ALLTOALLV\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEIGHBOR\_ALLTOALLV\_DEVICE\_COLLECTIVE - MPICH\_NEIGHBOR\_ALLTOALLV\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Neighbor\_alltoallv will - allow the device to override the MPIR-level collective algorithms. - The device might still call the MPIR-level algorithms manually. If - set to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_INEIGHBOR\_ALLTOALLV\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_INEIGHBOR\_ALLTOALLV\_DEVICE\_COLLECTIVE - MPICH\_INEIGHBOR\_ALLTOALLV\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Ineighbor\_alltoallv will - allow the device to override the MPIR-level collective algorithms. - The device might still call the MPIR-level algorithms manually. If - set to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_NEIGHBOR\_ALLTOALLV\_INIT\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEIGHBOR\_ALLTOALLV\_INIT\_DEVICE\_COLLECTIVE - MPICH\_NEIGHBOR\_ALLTOALLV\_INIT\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Neighbor\_alltoallv\_init - will allow the device to override the MPIR-level collective - algorithms. The device might still call the MPIR-level algorithms - manually. If set to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_NEIGHBOR\_ALLTOALLW\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEIGHBOR\_ALLTOALLW\_DEVICE\_COLLECTIVE - MPICH\_NEIGHBOR\_ALLTOALLW\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Neighbor\_alltoallw will - allow the device to override the MPIR-level collective algorithms. - The device might still call the MPIR-level algorithms manually. If - set to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_INEIGHBOR\_ALLTOALLW\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_INEIGHBOR\_ALLTOALLW\_DEVICE\_COLLECTIVE - MPICH\_INEIGHBOR\_ALLTOALLW\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Ineighbor\_alltoallw will - allow the device to override the MPIR-level collective algorithms. - The device might still call the MPIR-level algorithms manually. If - set to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_NEIGHBOR\_ALLTOALLW\_INIT\_DEVICE\_COLLECTIVE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEIGHBOR\_ALLTOALLW\_INIT\_DEVICE\_COLLECTIVE - MPICH\_NEIGHBOR\_ALLTOALLW\_INIT\_DEVICE\_COLLECTIVE - **Description:** - This CVAR is only used when MPIR\_CVAR\_DEVICE\_COLLECTIVES - is set to "percoll". If set to true, MPI\_Neighbor\_alltoallw\_init - will allow the device to override the MPIR-level collective - algorithms. The device might still call the MPIR-level algorithms - manually. If set to false, the device-override will be disabled. - **Default:** 1 MPIR\_CVAR\_COLL\_HYBRID\_MEMORY ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COLL\_HYBRID\_MEMORY - MPICH\_COLL\_HYBRID\_MEMORY - **Description:** - This cvar indicates if the memory used in the collective - operations are the same type. It set to true, it means in a - collective operation, some buffers could be on the CPU and some - buffers could be on the GPU. If set to false, it means all the data - in a collective operation are on the same type of memory. - **Default:** 0 MPIR\_CVAR\_GATHER\_VSMALL\_MSG\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_GATHER\_VSMALL\_MSG\_SIZE - MPICH\_GATHER\_VSMALL\_MSG\_SIZE - **Description:** - use a temporary buffer for intracommunicator MPI\_Gather if - the send buffer size is < this value (in bytes) (See also: - MPIR\_CVAR\_GATHER\_INTER\_SHORT\_MSG\_SIZE) - **Default:** 1024 MPIR\_CVAR\_IALLTOALL\_BRUCKS\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLTOALL\_BRUCKS\_KVAL - MPICH\_IALLTOALL\_BRUCKS\_KVAL - **Description:** - radix (k) value for generic transport brucks based - ialltoall - **Default:** 2 MPIR\_CVAR\_IALLTOALL\_BRUCKS\_BUFFER\_PER\_NBR ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLTOALL\_BRUCKS\_BUFFER\_PER\_NBR - MPICH\_IALLTOALL\_BRUCKS\_BUFFER\_PER\_NBR - **Description:** - If set to true, the tsp based brucks algorithm will - allocate dedicated send and receive buffers for every neighbor in - the brucks algorithm. Otherwise, it would reuse a single buffer for - sending and receiving data to/from neighbors - **Default:** 0 MPIR\_CVAR\_IALLTOALL\_SCATTERED\_OUTSTANDING\_TASKS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLTOALL\_SCATTERED\_OUTSTANDING\_TASKS - MPICH\_IALLTOALL\_SCATTERED\_OUTSTANDING\_TASKS - **Description:** - Maximum number of outstanding sends and recvs posted at a - time - **Default:** 64 MPIR\_CVAR\_IALLTOALL\_SCATTERED\_BATCH\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IALLTOALL\_SCATTERED\_BATCH\_SIZE - MPICH\_IALLTOALL\_SCATTERED\_BATCH\_SIZE - **Description:** - Number of send/receive tasks that scattered algorithm - waits for completion before posting another batch of send/receives - of that size - **Default:** 4 MPIR\_CVAR\_ENABLE\_DYNAMIC\_TUNING ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ENABLE\_DYNAMIC\_TUNING - MPICH\_ENABLE\_DYNAMIC\_TUNING - **Description:** - Enable or disable dynamic tuning - **Default:** 1 MPIR\_CVAR\_DYNAMIC\_TUNE\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_DYNAMIC\_TUNE\_THRESHOLD - MPICH\_DYNAMIC\_TUNE\_THRESHOLD - **Description:** - Number of iterations before dynamic tuning will be - re-attempted for a given collective and message range - **Default:** 1000 MPIR\_CVAR\_DYNAMIC\_TUNE\_TEST\_COUNT ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_DYNAMIC\_TUNE\_TEST\_COUNT - MPICH\_DYNAMIC\_TUNE\_TEST\_COUNT - **Description:** - Number of times each candidate algorithm will be tested - when performing dynamic retuning. - **Default:** 5 MPIR\_CVAR\_DYNAMIC\_TUNE\_LOCAL\_EVAL\_METRIC ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_DYNAMIC\_TUNE\_LOCAL\_EVAL\_METRIC - MPICH\_DYNAMIC\_TUNE\_LOCAL\_EVAL\_METRIC - **Description:** - Value reported by local processes ot the dynamic tuning - framework - avg - Report the average latency over - MPIR\_CVAR\_DYNAMIC\_TUNE\_TEST\_COUNT tests (default) - max - Report the maximum latency over - MPIR\_CVAR\_DYNAMIC\_TUNE\_TEST\_COUNT tests - min - Report the minimum latency over - MPIR\_CVAR\_DYNAMIC\_TUNE\_TEST\_COUNT tests - **Default:** MPIR\_CVAR\_DYNAMIC\_TUNE\_LOCAL\_EVAL\_METRIC\_avg MPIR\_CVAR\_DYNAMIC\_TUNE\_SELECTION\_METRIC ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_DYNAMIC\_TUNE\_SELECTION\_METRIC - MPICH\_DYNAMIC\_TUNE\_SELECTION\_METRIC - **Description:** - Metric used to select the best algorithm for dynamic - tuning - max - Use the algorithm that minimizes the maximum per process - latency (default) - min - Use the algorithm that minimizes the minimum per process - latency - avg - Use the algorithm that minimizes the average per process - latency - **Default:** MPIR\_CVAR\_DYNAMIC\_TUNE\_SELECTION\_METRIC\_max MPIR\_CVAR\_DYNAMIC\_TUNE\_VERBOSITY ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_DYNAMIC\_TUNE\_VERBOSITY - MPICH\_DYNAMIC\_TUNE\_VERBOSITY - **Description:** - Debuging verbosity for dynamic tuning - silent - No debugging output (default) - debug - Logs collective selection in debug output - basic - Identifies selected collectives and the relevant message - size - verbose - Full debug output - **Default:** MPIR\_CVAR\_DYNAMIC\_TUNE\_VERBOSITY\_silent MPIR\_CVAR\_BCAST\_INTER\_KNOMIAL\_FACTOR ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_INTER\_KNOMIAL\_FACTOR - MPICH\_BCAST\_INTER\_KNOMIAL\_FACTOR - **Description:** - This defines the degree of the knomial operation during - the inter-node knomial broadcast phase. - **Default:** 4 MPIR\_CVAR\_BCAST\_INTRA\_KNOMIAL\_FACTOR ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_INTRA\_KNOMIAL\_FACTOR - MPICH\_BCAST\_INTRA\_KNOMIAL\_FACTOR - **Description:** - This defines the degree of the knomial operation during - the intra-node knomial broadcast phase. - **Default:** 4 MPIR\_CVAR\_BCAST\_SEGMENT\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_SEGMENT\_SIZE - MPICH\_BCAST\_SEGMENT\_SIZE - **Description:** - Size of the segments used for PIPELINED bcast - **Default:** 8192 MPIR\_CVAR\_REDUCE\_INTER\_KNOMIAL\_FACTOR ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_INTER\_KNOMIAL\_FACTOR - MPICH\_REDUCE\_INTER\_KNOMIAL\_FACTOR - **Description:** - This defines the degree of the knomial operation during - the inter-node knomial reduce phase. - **Default:** 4 MPIR\_CVAR\_REDUCE\_INTRA\_KNOMIAL\_FACTOR ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_INTRA\_KNOMIAL\_FACTOR - MPICH\_REDUCE\_INTRA\_KNOMIAL\_FACTOR - **Description:** - This defines the degree of the knomial operation during - the intra-node knomial reduce phase. - **Default:** 4 MPIR\_CVAR\_DEVICE\_COLLECTIVES ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_DEVICE\_COLLECTIVES - MPICH\_DEVICE\_COLLECTIVES - **Description:** - Variable to select whether the device can override the - MPIR-level collective algorithms. - all - Always prefer the device collectives - none - Never pick the device collectives - percoll - Use the per-collective CVARs to decide - **Default:** MPIR\_CVAR\_DEVICE\_COLLECTIVES\_percoll MPIR\_CVAR\_COLLECTIVE\_FALLBACK ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COLLECTIVE\_FALLBACK - MPICH\_COLLECTIVE\_FALLBACK - **Description:** - Variable to control what the MPI library should do if the - user-specified collective algorithm does not work for the - arguments passed in by the user. - error - throw an error - print - print an error message and fallback to the internally - selected algorithm - silent - silently fallback to the internally selected algorithm - **Default:** MPIR\_CVAR\_COLLECTIVE\_FALLBACK\_silent MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COLL\_SELECTION\_TUNING\_JSON\_FILE - MPICH\_COLL\_SELECTION\_TUNING\_JSON\_FILE - **Description:** - Defines the location of tuning file. - **Default:** "" MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE\_DYNAMIC ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COLL\_SELECTION\_TUNING\_JSON\_FILE\_DYNAMIC - MPICH\_COLL\_SELECTION\_TUNING\_JSON\_FILE\_DYNAMIC - **Description:** - Defines the location of dynamic tuning file. - **Default:** "" MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE\_GPU ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COLL\_SELECTION\_TUNING\_JSON\_FILE\_GPU - MPICH\_COLL\_SELECTION\_TUNING\_JSON\_FILE\_GPU - **Description:** - Defines the location of gpu tuning file. - **Default:** "" MPIR\_CVAR\_COLL\_SELECTION\_TUNING\_JSON\_FILE\_GPU\_DYNAMIC ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COLL\_SELECTION\_TUNING\_JSON\_FILE\_GPU\_DYNAMIC - MPICH\_COLL\_SELECTION\_TUNING\_JSON\_FILE\_GPU\_DYNAMIC - **Description:** - Defines the location of dynamic gpu tuning file. - **Default:** "" MPIR\_CVAR\_HIERARCHY\_DUMP ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_HIERARCHY\_DUMP - MPICH\_HIERARCHY\_DUMP - **Description:** - If set to true, each rank will dump the hierarchy data - structure to a file named "hierarchy[rank]" in the current folder. - If set to false, the hierarchy data structure will not be dumped. - **Default:** 0 MPIR\_CVAR\_COORDINATES\_FILE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COORDINATES\_FILE - MPICH\_COORDINATES\_FILE - **Description:** - Defines the location of the input coordinates file. - **Default:** "" MPIR\_CVAR\_COLL\_TREE\_DUMP ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COLL\_TREE\_DUMP - MPICH\_COLL\_TREE\_DUMP - **Description:** - If set to true, each rank will dump the tree to a file - named "colltree[rank].json" in the current folder. If set to false, - the tree will not be dumped. - **Default:** 0 MPIR\_CVAR\_COORDINATES\_DUMP ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COORDINATES\_DUMP - MPICH\_COORDINATES\_DUMP - **Description:** - If set to true, rank 0 will dump the network coordinates - to a file named "coords" in the current folder. If set to false, - the network coordinates will not be dumped. - **Default:** 0 MPIR\_CVAR\_PROGRESS\_MAX\_COLLS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_PROGRESS\_MAX\_COLLS - MPICH\_PROGRESS\_MAX\_COLLS - **Description:** - Maximum number of collective operations at a time that the - progress engine should make progress on - **Default:** 0 MPIR\_CVAR\_COMM\_SPLIT\_USE\_QSORT ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COMM\_SPLIT\_USE\_QSORT - MPICH\_COMM\_SPLIT\_USE\_QSORT - **Description:** - Use qsort(3) in the implementation of MPI\_Comm\_split - instead of bubble sort. - **Default:** 1 MPIR\_CVAR\_CTXID\_EAGER\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CTXID\_EAGER\_SIZE - MPICH\_CTXID\_EAGER\_SIZE - **Description:** - The MPIR\_CVAR\_CTXID\_EAGER\_SIZE environment variable allows - you to specify how many words in the context ID mask will be set - aside for the eager allocation protocol. If the application is - running out of context IDs, reducing this value may help. - **Default:** 2 MPIR\_CVAR\_DATALOOP\_FAST\_SEEK ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_DATALOOP\_FAST\_SEEK - MPICH\_DATALOOP\_FAST\_SEEK - **Description:** - use a datatype-specialized algorithm to shortcut seeking - to the correct location in a noncontiguous buffer - **Default:** 1 MPIR\_CVAR\_YAKSA\_COMPLEX\_SUPPORT ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_YAKSA\_COMPLEX\_SUPPORT - MPICH\_YAKSA\_COMPLEX\_SUPPORT - **Description:** - This CVAR indicates that complex type reduction is not - supported in yaksa. - **Default:** 0 MPIR\_CVAR\_GPU\_DOUBLE\_SUPPORT ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_GPU\_DOUBLE\_SUPPORT - MPICH\_GPU\_DOUBLE\_SUPPORT - **Description:** - This CVAR indicates that double type is not supported on - the GPU. - **Default:** 0 MPIR\_CVAR\_GPU\_LONG\_DOUBLE\_SUPPORT ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_GPU\_LONG\_DOUBLE\_SUPPORT - MPICH\_GPU\_LONG\_DOUBLE\_SUPPORT - **Description:** - This CVAR indicates that double type is not supported on - the GPU. - **Default:** 0 MPIR\_CVAR\_ENABLE\_YAKSA\_REDUCTION ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ENABLE\_YAKSA\_REDUCTION - MPICH\_ENABLE\_YAKSA\_REDUCTION - **Description:** - This cvar enables yaksa based reduction for local reduce. - **Default:** 1 MPIR\_CVAR\_YAKSA\_REDUCTION\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_YAKSA\_REDUCTION\_THRESHOLD - MPICH\_YAKSA\_REDUCTION\_THRESHOLD - **Description:** - This cvar disables yaksa based reduction for messages - above the threshold. The default is no limit. - **Default:** -1 MPIR\_CVAR\_ENABLE\_GDRCOPY ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ENABLE\_GDRCOPY - MPICH\_ENABLE\_GDRCOPY - **Description:** - This cvar enables gdrcopy based staging for yaksa - operations - **Default:** 1 MPIR\_CVAR\_GDRCOPY\_MAX\_SIZE\_H2D ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_GDRCOPY\_MAX\_SIZE\_H2D - MPICH\_GDRCOPY\_MAX\_SIZE\_H2D - **Description:** - This cvar controls the maximum number of bytes for which - gdrcopy will be used for staging from host to device - **Default:** 32768 MPIR\_CVAR\_GDRCOPY\_MAX\_SIZE\_D2H ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_GDRCOPY\_MAX\_SIZE\_D2H - MPICH\_GDRCOPY\_MAX\_SIZE\_D2H - **Description:** - This cvar controls the maximum number of bytes for which - gdrcopy will be used for staging from a device to host - **Default:** 2048 MPIR\_CVAR\_GPU\_UNIFIED\_DIRECT\_COPY\_MAX ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_GPU\_UNIFIED\_DIRECT\_COPY\_MAX - MPICH\_GPU\_UNIFIED\_DIRECT\_COPY\_MAX - **Description:** - This cvar controls the maximum number of bytes for which - systems with unified GPU memory will use the direct CPU memcpy - support. For all messages at or above this value, the kernel driven - Yaksa approach will be used instead. - **Default:** (4 * 1024 * 1024) MPIR\_CVAR\_PROCTABLE\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_PROCTABLE\_SIZE - MPICH\_PROCTABLE\_SIZE - **Description:** - Size of the "MPIR" debugger interface proctable (process - table). - **Default:** 64 MPIR\_CVAR\_PROCTABLE\_PRINT ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_PROCTABLE\_PRINT - MPICH\_PROCTABLE\_PRINT - **Description:** - If true, dump the proctable entries at - MPII\_Wait\_for\_debugger-time. - **Default:** 0 MPIR\_CVAR\_PRINT\_ERROR\_STACK ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_PRINT\_ERROR\_STACK - MPICH\_PRINT\_ERROR\_STACK - **Description:** - If true, print an error stack trace at error handling - time. - **Default:** 1 MPIR\_CVAR\_CHOP\_ERROR\_STACK ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CHOP\_ERROR\_STACK - MPICH\_CHOP\_ERROR\_STACK - **Description:** - If >0, truncate error stack output lines this many - characters wide. If 0, do not truncate, and if <0 use a sensible - default. - **Default:** 0 MPIR\_CVAR\_ASYNC\_PROGRESS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ASYNC\_PROGRESS - MPICH\_ASYNC\_PROGRESS - **Description:** - If set to true, MPICH will initiate an additional thread - to make asynchronous progress on all communication operations - including point-to-point, collective, one-sided operations and I/O. - Setting this variable will automatically increase the - thread-safety level to MPI\_THREAD\_MULTIPLE. While this improves - the progress semantics, it might cause a small amount of - performance overhead for regular MPI operations. The user is - encouraged to leave one or more hardware threads vacant in order to - prevent contention between the application threads and the progress - thread(s). The impact of oversubscription is highly system - dependent but may be substantial in some cases, hence this - recommendation. - **Default:** 0 MPIR\_CVAR\_PROGRESS\_THREAD\_AFFINITY ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_PROGRESS\_THREAD\_AFFINITY - MPICH\_PROGRESS\_THREAD\_AFFINITY - **Description:** - Specifies affinity for all progress threads of local - processes. Can be set to auto or comma-separated list of logical - processors. When set to auto - MPICH will automatically select - logical CPU cores to decide affinity of the progress threads. When - set to comma-separated list of logical processors - In case of N - progress threads per process, the first N logical processors from - list will be assigned to threads of first local process, the next N - logical processors from list - to second local process and so on. - For example, thread affinity is "0,1,2,3", 2 progress threads per - process and 2 processes per node. Progress threads of first local - process will be pinned on logical processors "0,1", progress - threads of second local process - on "2,3". Cannot work together - with MPIR\_CVAR\_NUM\_CLIQUES or MPIR\_CVAR\_ODD\_EVEN\_CLIQUES. - **Default:** "" MPIR\_CVAR\_SUPPRESS\_ABORT\_MESSAGE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_SUPPRESS\_ABORT\_MESSAGE - MPICH\_SUPPRESS\_ABORT\_MESSAGE - **Description:** - Disable printing of abort error message. - **Default:** 0 MPIR\_CVAR\_COREDUMP\_ON\_ABORT ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COREDUMP\_ON\_ABORT - MPICH\_COREDUMP\_ON\_ABORT - **Description:** - Call libc abort() to generate a corefile - **Default:** 0 MPIR\_CVAR\_ERROR\_CHECKING ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ERROR\_CHECKING - MPICH\_ERROR\_CHECKING - **Description:** - If true, perform checks for errors, typically to verify - valid inputs to MPI routines. Only effective when MPICH is - configured with --enable-error-checking=runtime . - **Default:** 1 MPIR\_CVAR\_MEMDUMP ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_MEMDUMP - MPICH\_MEMDUMP - **Description:** - If true, list any memory that was allocated by MPICH and - that remains allocated when MPI\_Finalize completes. - **Default:** 1 MPIR\_CVAR\_DEBUG\_SUMMARY ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_DEBUG\_SUMMARY - MPICH\_DEBUG\_SUMMARY - MPIR\_CVAR\_MEM\_CATEGORY\_INFORMATION - MPIR\_CVAR\_CH4\_OFI\_CAPABILITY\_SETS\_DEBUG - MPIR\_CVAR\_CH4\_UCX\_CAPABILITY\_DEBUG - MVP\_MEM\_CATEGORY\_INFORMATION - MVP\_CH4\_OFI\_CAPABILITY\_SETS\_DEBUG - MVP\_CH4\_UCX\_CAPABILITY\_DEBUG - MPICH\_MEM\_CATEGORY\_INFORMATION - MPICH\_CH4\_OFI\_CAPABILITY\_SETS\_DEBUG - MPICH\_CH4\_UCX\_CAPABILITY\_DEBUG - **Description:** - 1: Print internal summary of various debug information, - such as memory allocation by category. Each layer may print their - own summary information. For example, ch4-ofi may print its - provider capability settings. 2: Also print the preferred NIC for - each rank - **Default:** 0 MPIR\_CVAR\_DEFAULT\_THREAD\_LEVEL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_DEFAULT\_THREAD\_LEVEL - MPICH\_DEFAULT\_THREAD\_LEVEL - **Description:** - Sets the default thread level to use when using MPI\_INIT. - This variable is case-insensitive. - **Default:** "MPI\_THREAD\_SINGLE" MPIR\_CVAR\_DEBUG\_HOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_DEBUG\_HOLD - MPICH\_DEBUG\_HOLD - **Description:** - If true, causes processes to wait in MPI\_Init and - MPI\_Initthread for a debugger to be attached. Once the debugger - has attached, the variable 'hold' should be set to 0 in order to - allow the process to continue (e.g., in gdb, "set hold=0"). - **Default:** 0 MPIR\_CVAR\_GPU\_USE\_IMMEDIATE\_COMMAND\_LIST ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_GPU\_USE\_IMMEDIATE\_COMMAND\_LIST - MPICH\_GPU\_USE\_IMMEDIATE\_COMMAND\_LIST - **Description:** - If true, mpl/ze will use immediate command list for - copying - **Default:** 0 MPIR\_CVAR\_GPU\_ROUND\_ROBIN\_COMMAND\_QUEUES ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_GPU\_ROUND\_ROBIN\_COMMAND\_QUEUES - MPICH\_GPU\_ROUND\_ROBIN\_COMMAND\_QUEUES - **Description:** - If true, mpl/ze will use command queues in a round-robin - fashion. If false, only command queues of index 0 will be used. - **Default:** 0 MPIR\_CVAR\_NO\_COLLECTIVE\_FINALIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NO\_COLLECTIVE\_FINALIZE - MPICH\_NO\_COLLECTIVE\_FINALIZE - **Description:** - If true, prevent MPI\_Finalize to invoke collective - behavior such as barrier or communicating to other processes. - Consequently, it may result in leaking memory or losing messages - due to pre-mature exiting. The default is false, which may invoke - collective behaviors at finalize. - **Default:** 0 MPIR\_CVAR\_FINALIZE\_WAIT ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_FINALIZE\_WAIT - MPICH\_FINALIZE\_WAIT - **Description:** - If true, poll progress at MPI\_Finalize until reference - count on MPI\_COMM\_WORLD and MPI\_COMM\_SELF reaches zero. This may be - necessary to prevent remote processes hanging if it has pending - communication protocols, e.g. a rendezvous send. - **Default:** 0 MPIR\_CVAR\_INIT\_SKIP\_PMI\_BARRIER ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_INIT\_SKIP\_PMI\_BARRIER - MPICH\_INIT\_SKIP\_PMI\_BARRIER - **Description:** - Skip MPIR\_pmi\_barrier() in MPI\_Init - **Default:** 1 MPIR\_CVAR\_GPU\_FAST\_COPY\_MAX\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_GPU\_FAST\_COPY\_MAX\_SIZE - MPICH\_GPU\_FAST\_COPY\_MAX\_SIZE - **Description:** - If a send message size is less than or equal to - MPIR\_CVAR\_GPU\_FAST\_COPY\_MAX\_SIZE (in bytes), then enable GPU-based - fast memcpy. The environment variable is valid only when then GPU - IPC shmmod is enabled. - **Default:** 4096 MPIR\_CVAR\_GPU\_FAST\_COPY\_MAX\_SIZE\_H2D ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_GPU\_FAST\_COPY\_MAX\_SIZE\_H2D - MPICH\_GPU\_FAST\_COPY\_MAX\_SIZE\_H2D - **Description:** - If a receive message size is less than or equal to - MPIR\_CVAR\_GPU\_FAST\_COPY\_MAX\_SIZE\_H2D (in bytes), then enable - GPU-based fast memcpy. - **Default:** 1048576 MPIR\_CVAR\_GPU\_FAST\_COPY\_MAX\_SIZE\_D2H ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_GPU\_FAST\_COPY\_MAX\_SIZE\_D2H - MPICH\_GPU\_FAST\_COPY\_MAX\_SIZE\_D2H - **Description:** - If a send message size is less than or equal to - MPIR\_CVAR\_GPU\_FAST\_COPY\_MAX\_SIZE\_D2H (in bytes), then enable - GPU-based fast memcpy. - **Default:** 32768 MPIR\_CVAR\_REQUEST\_ERR\_FATAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REQUEST\_ERR\_FATAL - MPICH\_REQUEST\_ERR\_FATAL - **Description:** - By default, MPI\_Waitall, MPI\_Testall, MPI\_Waitsome, and - MPI\_Testsome return MPI\_ERR\_IN\_STATUS when one of the request - fails. If MPIR\_CVAR\_REQUEST\_ERR\_FATAL is set to true, these - routines will return the error code of the request immediately. The - default MPI\_ERRS\_ARE\_FATAL error handler will dump a error stack in - this case, which maybe more convenient for debugging. This cvar - will also make nonblocking shched return error right away as it - issues operations. - **Default:** 0 MPIR\_CVAR\_REQUEST\_POLL\_FREQ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REQUEST\_POLL\_FREQ - MPICH\_REQUEST\_POLL\_FREQ - **Description:** - How frequent to poll during MPI\_{Waitany,Waitsome} in - terms of number of processed requests before polling. - **Default:** 8 MPIR\_CVAR\_REQUEST\_BATCH\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REQUEST\_BATCH\_SIZE - MPICH\_REQUEST\_BATCH\_SIZE - **Description:** - The number of requests to make completion as a batch in - MPI\_Waitall and MPI\_Testall implementation. A large number is - likely to cause more cache misses. - **Default:** 64 MPIR\_CVAR\_DEBUG\_PROGRESS\_TIMEOUT ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_DEBUG\_PROGRESS\_TIMEOUT - MPICH\_DEBUG\_PROGRESS\_TIMEOUT - **Description:** - Sets the timeout in seconds to dump outstanding requests - when progress wait is not making progress for some time. - **Default:** 0 MPIR\_CVAR\_DIMS\_VERBOSE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_DIMS\_VERBOSE - MPICH\_DIMS\_VERBOSE - **Description:** - If true, enable verbose output about the actions of the - implementation of MPI\_Dims\_create. - **Default:** 0 MPIR\_CVAR\_QMPI\_TOOL\_LIST ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_QMPI\_TOOL\_LIST - MPICH\_QMPI\_TOOL\_LIST - **Description:** - Set the number and order of QMPI tools to be loaded by the - MPI library when it is initialized. - **Default:** NULL MPIR\_CVAR\_NAMESERV\_FILE\_PUBDIR ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NAMESERV\_FILE\_PUBDIR - MPICH\_NAMESERV\_FILE\_PUBDIR - MPIR\_CVAR\_NAMEPUB\_DIR - MVP\_NAMEPUB\_DIR - MPICH\_NAMEPUB\_DIR - **Description:** - Sets the directory to use for MPI service publishing in - the file nameserv implementation. Allows the user to override - where the publish and lookup information is placed for - connect/accept based applications. - **Default:** NULL MPIR\_CVAR\_ENABLE\_COMPRESSION ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ENABLE\_COMPRESSION - MPICH\_ENABLE\_COMPRESSION - **Description:** - Possible values: 0, 1 If set to 1, turn on compression If - set to 2, turn off ompression - **Default:** 0 MPIR\_CVAR\_COMPRESSION\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COMPRESSION\_ALGORITHM - MPICH\_COMPRESSION\_ALGORITHM - **Description:** - Used to force a particular GPU compression algorithm - mpc - Use MPC lossless compression algorithm - zfp - Use ZFP lossy compression algorithm - **Default:** MPIR\_CVAR\_COMPRESSION\_ALGORITHM\_mpc MPIR\_CVAR\_COMPRESSION\_GPU\_BLOCKS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COMPRESSION\_GPU\_BLOCKS - MPICH\_COMPRESSION\_GPU\_BLOCKS - **Description:** - GPU thread blocks - **Default:** 216 MPIR\_CVAR\_COMPRESSION\_DIMENSION ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COMPRESSION\_DIMENSION - MPICH\_COMPRESSION\_DIMENSION - **Description:** - Dimensionality in compression Possible values: integer - between 1 and 32 - **Default:** 1 MPIR\_CVAR\_COMPRESSION\_NUM\_STREAM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COMPRESSION\_NUM\_STREAM - MPICH\_COMPRESSION\_NUM\_STREAM - **Description:** - Number of streams in compression - **Default:** 512 MPIR\_CVAR\_COMPRESSION\_DATA\_TYPE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COMPRESSION\_DATA\_TYPE - MPICH\_COMPRESSION\_DATA\_TYPE - **Description:** - Used to force a particular GPU compression algorithm - FLOAT - use compression with floating point type - DOUBLE - use compression with double precision type - **Default:** MPIR\_CVAR\_COMPRESSION\_DATA\_TYPE\_FLOAT MPIR\_CVAR\_ENABLE\_PT2PT\_GPU\_COMPRESSION ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ENABLE\_PT2PT\_GPU\_COMPRESSION - MPICH\_ENABLE\_PT2PT\_GPU\_COMPRESSION - **Description:** - Possible values: 0, 1 If set to 0, turn off point to point - compression If set to 1, use point to point compression - **Default:** 0 MPIR\_CVAR\_PT2PT\_GPU\_COMPRESSION\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_PT2PT\_GPU\_COMPRESSION\_THRESHOLD - MPICH\_PT2PT\_GPU\_COMPRESSION\_THRESHOLD - **Description:** - Use compression for point-to-point messages only when the - message size exceeds this setting, if point to point compression is - enabled. - **Default:** 1048576 MPIR\_CVAR\_COMPRESSION\_BUFFER\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COMPRESSION\_BUFFER\_SIZE - MPICH\_COMPRESSION\_BUFFER\_SIZE - **Description:** - Specifies the size of buffers for - compressing/decompressing active messages in each block of the - pool. The size here should be the maximum size needed for - compression algorithms, but in the future will be the chunk size - used for a pipelined compression. - **Default:** 33554432 MPIR\_CVAR\_COMPRESSION\_NUM\_BUFFERS\_PER\_CHUNK ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COMPRESSION\_NUM\_BUFFERS\_PER\_CHUNK - MPICH\_COMPRESSION\_NUM\_BUFFERS\_PER\_CHUNK - **Description:** - Specifies the number of buffers for - compressing/decompressing active messages in each block of the - pool. Setting to one for now because the buffers are so large. In - the future, we will set this higher once the default buffer size is - lowered. - **Default:** 1 MPIR\_CVAR\_COMPRESSION\_MAX\_NUM\_COMPRESSION\_BUFFERS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COMPRESSION\_MAX\_NUM\_COMPRESSION\_BUFFERS - MPICH\_COMPRESSION\_MAX\_NUM\_COMPRESSION\_BUFFERS - **Description:** - Specifies the max number of buffers for - compressing/decompressing in the pool. Capped at 64 by default, - meaning a maximum of 2GB can be allocated by this pool. - **Default:** 64 MPIR\_CVAR\_COMPRESSION\_REDUCE\_SCATTER\_CHUNK\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COMPRESSION\_REDUCE\_SCATTER\_CHUNK\_SIZE - MPICH\_COMPRESSION\_REDUCE\_SCATTER\_CHUNK\_SIZE - **Description:** - Chunk data size for COLLECTIVEective compression in - ring-based reduce-scatter - **Default:** (2*1024*1024) MPIR\_CVAR\_COMPRESSION\_BCAST\_CHUNKS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COMPRESSION\_BCAST\_CHUNKS - MPICH\_COMPRESSION\_BCAST\_CHUNKS - **Description:** - Default number of chunks for chunked-chain bcast with - compression - **Default:** 2 MPIR\_CVAR\_COMPRESSION\_BCAST\_CHUNK\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COMPRESSION\_BCAST\_CHUNK\_SIZE - MPICH\_COMPRESSION\_BCAST\_CHUNK\_SIZE - **Description:** - Default chunk size for chunked-chain bcast with - compression - **Default:** (512*1024) MPIR\_CVAR\_ZFP\_NUM\_STREAM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ZFP\_NUM\_STREAM - MPICH\_ZFP\_NUM\_STREAM - **Description:** - number of zfp streams to preallocate pool Possible values: - integer greater than 1 - **Default:** 64 MPIR\_CVAR\_ZFP\_RATE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ZFP\_RATE - MPICH\_ZFP\_RATE - **Description:** - Possible values: [1,32] for float, [1-64] for double - **Default:** 16 MPIR\_CVAR\_ZFP\_NX ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ZFP\_NX - MPICH\_ZFP\_NX - **Description:** - For 2D dimension Possible values: Positive integer, - multiple of 4 - **Default:** 4 MPIR\_CVAR\_ZFP\_NY ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ZFP\_NY - MPICH\_ZFP\_NY - **Description:** - For 3D dimension Possible values: Positive integer, - multiple of 4 - **Default:** 4 MPIR\_CVAR\_ZFP\_ALLOC\_WARNING ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ZFP\_ALLOC\_WARNING - MPICH\_ZFP\_ALLOC\_WARNING - **Description:** - Possible values: 0, 1 If set to 0, turn off warning for - allocation of zfp stream pool If set to 1, turn on warning for - allocation of zfp stream pool - **Default:** 1 MPIR\_CVAR\_ZFP\_PATH ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ZFP\_PATH - MPICH\_ZFP\_PATH - **Description:** - Path to an alternate ZFP library to be opened with dlsym - at runtime. By default ZFP is linked to an included version. - **Default:** NULL MPIR\_CVAR\_ENABLE\_ALLREDUCE\_COMPRESSION ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ENABLE\_ALLREDUCE\_COMPRESSION - MPICH\_ENABLE\_ALLREDUCE\_COMPRESSION - **Description:** - Possible values: 0, 1 If set to 0, turn off collective - compression for allreduce If set to 1, use collective compression - for allreduce - **Default:** 0 MPIR\_CVAR\_ABORT\_ON\_LEAKED\_HANDLES ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ABORT\_ON\_LEAKED\_HANDLES - MPICH\_ABORT\_ON\_LEAKED\_HANDLES - **Description:** - If true, MPI will call MPI\_Abort at MPI\_Finalize if any - MPI object handles have been leaked. For example, if MPI\_Comm\_dup - is called without calling a corresponding MPI\_Comm\_free. For - uninteresting reasons, enabling this option may prevent all known - object leaks from being reported. MPICH must have been configure - with "--enable-g=handlealloc" or better in order for this - functionality to work. - **Default:** 0 MPIR\_CVAR\_NETLOC\_NODE\_FILE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NETLOC\_NODE\_FILE - MPICH\_NETLOC\_NODE\_FILE - **Description:** - Subnet json file - **Default:** "auto" MPIR\_CVAR\_NOLOCAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NOLOCAL - MPICH\_NOLOCAL - MPIR\_CVAR\_NO\_LOCAL - MVP\_NO\_LOCAL - MPICH\_NO\_LOCAL - **Description:** - If true, force all processes to operate as though all - processes are located on another node. For example, this disables - shared memory communication hierarchical collectives. - **Default:** 0 MPIR\_CVAR\_ODD\_EVEN\_CLIQUES ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ODD\_EVEN\_CLIQUES - MPICH\_ODD\_EVEN\_CLIQUES - MPIR\_CVAR\_EVEN\_ODD\_CLIQUES - MVP\_EVEN\_ODD\_CLIQUES - MPICH\_EVEN\_ODD\_CLIQUES - **Description:** - If true, odd procs on a node are seen as local to each - other, and even procs on a node are seen as local to each other. - Used for debugging on a single machine. Deprecated in favor of - MPIR\_CVAR\_NUM\_CLIQUES. - **Default:** 0 MPIR\_CVAR\_NUM\_CLIQUES ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NUM\_CLIQUES - MPICH\_NUM\_CLIQUES - **Description:** - Specify the number of cliques that should be used to - partition procs on a local node. Procs with the same clique number - are seen as local to each other. Used for debugging on a single - machine. - **Default:** 1 MPIR\_CVAR\_CLIQUES\_BY\_BLOCK ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CLIQUES\_BY\_BLOCK - MPICH\_CLIQUES\_BY\_BLOCK - **Description:** - Specify to divide processes into cliques by uniform - blocks. The default is to divide in round-robin fashion. Used for - debugging on a single machine. - **Default:** 0 MPIR\_CVAR\_PMI\_VERSION ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_PMI\_VERSION - MPICH\_PMI\_VERSION - **Description:** - Variable to select runtime PMI version. - 1 - - PMI (default) - 2 - - PMI2 - x - - PMIx - **Default:** MPIR\_CVAR\_PMI\_VERSION\_1 MPIR\_CVAR\_COLL\_ALIAS\_CHECK ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COLL\_ALIAS\_CHECK - MPICH\_COLL\_ALIAS\_CHECK - **Description:** - Enable checking of aliasing in collective operations - **Default:** 1 MPIR\_CVAR\_ENABLE\_GPU ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ENABLE\_GPU - MPICH\_ENABLE\_GPU - **Description:** - Control MPICH GPU support. If set to 0, all GPU support is - disabled and we do not query the buffer type internally because we - assume no GPU buffer is use. - **Default:** 1 MPIR\_CVAR\_GPU\_HAS\_WAIT\_KERNEL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_GPU\_HAS\_WAIT\_KERNEL - MPICH\_GPU\_HAS\_WAIT\_KERNEL - **Description:** - If set to 1, avoid allocate allocating GPU registered host - buffers for temporary buffers. When stream workq and GPU wait - kernels are in use, access APIs for GPU registered memory may cause - deadlock. - **Default:** 0 MPIR\_CVAR\_ENABLE\_GPU\_REGISTER ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ENABLE\_GPU\_REGISTER - MPICH\_ENABLE\_GPU\_REGISTER - **Description:** - Control whether to actually register buffers with the GPU - runtime in MPIR\_gpu\_register\_host. This could lower the latency of - certain GPU communication at the cost of some amount of GPU memory - consumed by the MPI library. By default, registration is enabled. - **Default:** 1 MPIR\_CVAR\_LMEM\_POOL\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_LMEM\_POOL\_SIZE - MPICH\_LMEM\_POOL\_SIZE - **Description:** - Sets the amount of space allocated at init time to be used - by internal functions for local temp buffers. - **Default:** 32 * 1024 * 1024 MPIR\_CVAR\_POLLS\_BEFORE\_YIELD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_POLLS\_BEFORE\_YIELD - MPICH\_POLLS\_BEFORE\_YIELD - **Description:** - When MPICH is in a busy waiting loop, it will periodically - call a function to yield the processor. This cvar sets the number - of loops before the yield function is called. A value of 0 - disables yielding. - **Default:** 1000 MPIR\_CVAR\_CH3\_INTERFACE\_HOSTNAME ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH3\_INTERFACE\_HOSTNAME - MPICH\_CH3\_INTERFACE\_HOSTNAME - MPIR\_CVAR\_INTERFACE\_HOSTNAME - MVP\_INTERFACE\_HOSTNAME - MPICH\_INTERFACE\_HOSTNAME - **Description:** - If non-NULL, this cvar specifies the IP address that other - processes should use when connecting to this process. This cvar is - mutually exclusive with the MPIR\_CVAR\_CH3\_NETWORK\_IFACE cvar and it - is an error to set them both. - **Default:** NULL MPIR\_CVAR\_CH3\_PORT\_RANGE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH3\_PORT\_RANGE - MPICH\_CH3\_PORT\_RANGE - MPIR\_CVAR\_PORTRANGE - MPIR\_CVAR\_PORT\_RANGE - MVP\_PORTRANGE - MVP\_PORT\_RANGE - MPICH\_PORTRANGE - MPICH\_PORT\_RANGE - **Description:** - The MPIR\_CVAR\_CH3\_PORT\_RANGE environment variable allows - you to specify the range of TCP ports to be used by the process - manager and the MPICH library. The format of this variable is - :. To specify any available port, use 0:0. - **Default:** {0,0} MPIR\_CVAR\_NEMESIS\_TCP\_NETWORK\_IFACE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEMESIS\_TCP\_NETWORK\_IFACE - MPICH\_NEMESIS\_TCP\_NETWORK\_IFACE - MPIR\_CVAR\_NETWORK\_IFACE - MVP\_NETWORK\_IFACE - MPICH\_NETWORK\_IFACE - **Description:** - If non-NULL, this cvar specifies which pseudo-ethernet - interface the tcp netmod should use (e.g., "eth1", "ib0"). Note, - this is a Linux-specific cvar. This cvar is mutually exclusive with - the MPIR\_CVAR\_CH3\_INTERFACE\_HOSTNAME cvar and it is an error to set - them both. - **Default:** NULL MPIR\_CVAR\_NEMESIS\_TCP\_HOST\_LOOKUP\_RETRIES ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEMESIS\_TCP\_HOST\_LOOKUP\_RETRIES - MPICH\_NEMESIS\_TCP\_HOST\_LOOKUP\_RETRIES - **Description:** - This cvar controls the number of times to retry the - gethostbyname() function before giving up. - **Default:** 10 MPIR\_CVAR\_NEMESIS\_ENABLE\_CKPOINT ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEMESIS\_ENABLE\_CKPOINT - MPICH\_NEMESIS\_ENABLE\_CKPOINT - **Description:** - If true, enables checkpointing support and returns an - error if checkpointing library cannot be initialized. - **Default:** 0 MPIR\_CVAR\_NEMESIS\_SHM\_EAGER\_MAX\_SZ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEMESIS\_SHM\_EAGER\_MAX\_SZ - MPICH\_NEMESIS\_SHM\_EAGER\_MAX\_SZ - **Description:** - This cvar controls the message size at which Nemesis - switches from eager to rendezvous mode for shared memory. If this - cvar is set to -1, then Nemesis will choose an appropriate value. - **Default:** -1 MPIR\_CVAR\_NEMESIS\_SHM\_READY\_EAGER\_MAX\_SZ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEMESIS\_SHM\_READY\_EAGER\_MAX\_SZ - MPICH\_NEMESIS\_SHM\_READY\_EAGER\_MAX\_SZ - **Description:** - This cvar controls the message size at which Nemesis - switches from eager to rendezvous mode for ready-send messages. If - this cvar is set to -1, then ready messages will always be sent - eagerly. If this cvar is set to -2, then Nemesis will choose an - appropriate value. - **Default:** -2 MPIR\_CVAR\_ENABLE\_FT ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ENABLE\_FT - MPICH\_ENABLE\_FT - **Description:** - Enable fault tolerance functions - **Default:** 0 MPIR\_CVAR\_NEMESIS\_NETMOD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NEMESIS\_NETMOD - MPICH\_NEMESIS\_NETMOD - **Description:** - If non-empty, this cvar specifies which network module - should be used for communication. This variable is - case-insensitive. - **Default:** "" MPIR\_CVAR\_CH3\_ENABLE\_HCOLL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH3\_ENABLE\_HCOLL - MPICH\_CH3\_ENABLE\_HCOLL - **Description:** - If true, enable HCOLL collectives. - **Default:** 0 MPIR\_CVAR\_CH3\_COMM\_CONNECT\_TIMEOUT ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH3\_COMM\_CONNECT\_TIMEOUT - MPICH\_CH3\_COMM\_CONNECT\_TIMEOUT - **Description:** - The default time out period in seconds for a connection - attempt to the server communicator where the named port exists but - no pending accept. User can change the value for a specified - connection through its info argument. - **Default:** 180 MPIR\_CVAR\_CH3\_RMA\_OP\_PIGGYBACK\_LOCK\_DATA\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH3\_RMA\_OP\_PIGGYBACK\_LOCK\_DATA\_SIZE - MPICH\_CH3\_RMA\_OP\_PIGGYBACK\_LOCK\_DATA\_SIZE - **Description:** - Specify the threshold of data size of a RMA operation - which can be piggybacked with a LOCK message. It is always a - positive value and should not be smaller than - MPIDI\_RMA\_IMMED\_BYTES. If user sets it as a small value, for middle - and large data size, we will lose performance because of always - waiting for round-trip of LOCK synchronization; if user sets it as - a large value, we need to consume more memory on target side to - buffer this lock request when lock is not satisfied. - **Default:** 65536 MPIR\_CVAR\_CH3\_RMA\_ACTIVE\_REQ\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH3\_RMA\_ACTIVE\_REQ\_THRESHOLD - MPICH\_CH3\_RMA\_ACTIVE\_REQ\_THRESHOLD - **Description:** - Threshold of number of active requests to trigger blocking - waiting in operation routines. When the value is negative, we never - blockingly wait in operation routines. When the value is zero, we - always trigger blocking waiting in operation routines to wait until - no. of active requests becomes zero. When the value is positive, we - do blocking waiting in operation routines to wait until no. of - active requests being reduced to this value. - **Default:** 65536 MPIR\_CVAR\_CH3\_RMA\_POKE\_PROGRESS\_REQ\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH3\_RMA\_POKE\_PROGRESS\_REQ\_THRESHOLD - MPICH\_CH3\_RMA\_POKE\_PROGRESS\_REQ\_THRESHOLD - **Description:** - Threshold at which the RMA implementation attempts to - complete requests while completing RMA operations and while using - the lazy synchronization approach. Change this value if programs - fail because they run out of requests or other internal resources - **Default:** 128 MPIR\_CVAR\_CH3\_RMA\_SCALABLE\_FENCE\_PROCESS\_NUM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH3\_RMA\_SCALABLE\_FENCE\_PROCESS\_NUM - MPICH\_CH3\_RMA\_SCALABLE\_FENCE\_PROCESS\_NUM - **Description:** - Specify the threshold of switching the algorithm used in - FENCE from the basic algorithm to the scalable algorithm. The value - can be negative, zero or positive. When the number of processes is - larger than or equal to this value, FENCE will use a scalable - algorithm which do not use O(P) data structure; when the number of - processes is smaller than the value, FENCE will use a basic but - fast algorithm which requires an O(P) data structure. - **Default:** 1024 MPIR\_CVAR\_CH3\_RMA\_DELAY\_ISSUING\_FOR\_PIGGYBACKING ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH3\_RMA\_DELAY\_ISSUING\_FOR\_PIGGYBACKING - MPICH\_CH3\_RMA\_DELAY\_ISSUING\_FOR\_PIGGYBACKING - **Description:** - Specify if delay issuing of RMA operations for - piggybacking LOCK/UNLOCK/FLUSH is enabled. It can be either 0 or 1. - When it is set to 1, the issuing of LOCK message is delayed until - origin process see the first RMA operation and piggyback LOCK with - that operation, and the origin process always keeps the current - last operation until the ending synchronization call in order to - piggyback UNLOCK/FLUSH with that operation. When it is set to 0, in - WIN\_LOCK/UNLOCK case, the LOCK message is sent out as early as - possible, in WIN\_LOCK\_ALL/UNLOCK\_ALL case, the origin process still - tries to piggyback LOCK message with the first operation; for - UNLOCK/FLUSH message, the origin process no longer keeps the - current last operation but only piggyback UNLOCK/FLUSH if there is - an operation available in the ending synchronization call. - **Default:** 0 MPIR\_CVAR\_CH3\_RMA\_SLOTS\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH3\_RMA\_SLOTS\_SIZE - MPICH\_CH3\_RMA\_SLOTS\_SIZE - **Description:** - Number of RMA slots during window creation. Each slot - contains a linked list of target elements. The distribution of - ranks among slots follows a round-robin pattern. Requires a - positive value. - **Default:** 262144 MPIR\_CVAR\_CH3\_RMA\_TARGET\_LOCK\_DATA\_BYTES ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH3\_RMA\_TARGET\_LOCK\_DATA\_BYTES - MPICH\_CH3\_RMA\_TARGET\_LOCK\_DATA\_BYTES - **Description:** - Size (in bytes) of available lock data this window can - provided. If current buffered lock data is more than this value, - the process will drop the upcoming operation data. Requires a - positive value. - **Default:** 655360 MPIR\_CVAR\_CH3\_EAGER\_MAX\_MSG\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH3\_EAGER\_MAX\_MSG\_SIZE - MPICH\_CH3\_EAGER\_MAX\_MSG\_SIZE - **Description:** - This cvar controls the message size at which CH3 switches - from eager to rendezvous mode. - **Default:** 131072 MPIR\_CVAR\_CH3\_PG\_VERBOSE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH3\_PG\_VERBOSE - MPICH\_CH3\_PG\_VERBOSE - **Description:** - If set, print the PG state on finalize. - **Default:** 0 MPIR\_CVAR\_CH3\_RMA\_OP\_WIN\_POOL\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH3\_RMA\_OP\_WIN\_POOL\_SIZE - MPICH\_CH3\_RMA\_OP\_WIN\_POOL\_SIZE - **Description:** - Size of the window-private RMA operations pool (in number - of operations) that stores information about RMA operations that - could not be issued immediately. Requires a positive value. - **Default:** 256 MPIR\_CVAR\_CH3\_RMA\_OP\_GLOBAL\_POOL\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH3\_RMA\_OP\_GLOBAL\_POOL\_SIZE - MPICH\_CH3\_RMA\_OP\_GLOBAL\_POOL\_SIZE - **Description:** - Size of the Global RMA operations pool (in number of - operations) that stores information about RMA operations that could - not be issued immediately. Requires a positive value. - **Default:** 16384 MPIR\_CVAR\_CH3\_RMA\_TARGET\_WIN\_POOL\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH3\_RMA\_TARGET\_WIN\_POOL\_SIZE - MPICH\_CH3\_RMA\_TARGET\_WIN\_POOL\_SIZE - **Description:** - Size of the window-private RMA target pool (in number of - targets) that stores information about RMA targets that could not - be issued immediately. Requires a positive value. - **Default:** 256 MPIR\_CVAR\_CH3\_RMA\_TARGET\_GLOBAL\_POOL\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH3\_RMA\_TARGET\_GLOBAL\_POOL\_SIZE - MPICH\_CH3\_RMA\_TARGET\_GLOBAL\_POOL\_SIZE - **Description:** - Size of the Global RMA targets pool (in number of targets) - that stores information about RMA targets that could not be issued - immediately. Requires a positive value. - **Default:** 16384 MPIR\_CVAR\_CH3\_RMA\_TARGET\_LOCK\_ENTRY\_WIN\_POOL\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH3\_RMA\_TARGET\_LOCK\_ENTRY\_WIN\_POOL\_SIZE - MPICH\_CH3\_RMA\_TARGET\_LOCK\_ENTRY\_WIN\_POOL\_SIZE - **Description:** - Size of the window-private RMA lock entries pool (in - number of lock entries) that stores information about RMA lock - requests that could not be satisfied immediately. Requires a - positive value. - **Default:** 256 MPIR\_CVAR\_OFI\_USE\_PROVIDER ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_OFI\_USE\_PROVIDER - MPICH\_OFI\_USE\_PROVIDER - **Description:** - This variable is no longer supported. Use FI\_PROVIDER - instead to select libfabric providers. - **Default:** NULL MPIR\_CVAR\_SINGLE\_HOST\_ENABLED ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_SINGLE\_HOST\_ENABLED - MPICH\_SINGLE\_HOST\_ENABLED - **Description:** - Set this variable to true to indicate that processes are - launched on a single host. The current implication is to avoid the - cxi provider to prevent the use of scarce hardware resources. - **Default:** 1 MPIR\_CVAR\_CH4\_OFI\_AM\_LONG\_FORCE\_PIPELINE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_AM\_LONG\_FORCE\_PIPELINE - MPICH\_CH4\_OFI\_AM\_LONG\_FORCE\_PIPELINE - **Description:** - For long message to be sent using pipeline rather than - default RDMA read. - **Default:** 0 MPIR\_CVAR\_BCAST\_OFI\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_OFI\_INTRA\_ALGORITHM - MPICH\_BCAST\_OFI\_INTRA\_ALGORITHM - **Description:** - Variable to select algorithm for intra-node bcast - mpir - - - - Fallback to MPIR collectives - trigger\_tree\_tagged - - Force triggered ops based Tagged Tree - trigger\_tree\_rma - - Force triggered ops based RMA Tree - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_CH4\_OFI\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - **Default:** MPIR\_CVAR\_BCAST\_OFI\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_CH4\_OFI\_GPU\_RECEIVE\_ENGINE\_TYPE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_GPU\_RECEIVE\_ENGINE\_TYPE - MPICH\_CH4\_OFI\_GPU\_RECEIVE\_ENGINE\_TYPE - **Description:** - Specifies GPU engine type for GPU pt2pt on the receiver - side. - compute - use a compute engine - copy\_high\_bandwidth - use a high-bandwidth copy engine - copy\_low\_latency - use a low-latency copy engine - yaksa - use Yaksa - **Default:** MPIR\_CVAR\_CH4\_OFI\_GPU\_RECEIVE\_ENGINE\_TYPE\_copy\_low\_latency MPIR\_CVAR\_OFI\_SKIP\_IPV6 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_OFI\_SKIP\_IPV6 - MPICH\_OFI\_SKIP\_IPV6 - **Description:** - Skip IPv6 providers. - **Default:** 0 MPIR\_CVAR\_CH4\_OFI\_ENABLE\_DATA ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_ENABLE\_DATA - MPICH\_CH4\_OFI\_ENABLE\_DATA - **Description:** - Enable immediate data fields in OFI to transmit source - rank outside of the match bits - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_ENABLE\_AV\_TABLE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_ENABLE\_AV\_TABLE - MPICH\_CH4\_OFI\_ENABLE\_AV\_TABLE - **Description:** - If true, the OFI addressing information will be stored - with an FI\_AV\_TABLE. If false, an FI\_AV\_MAP will be used. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_ENABLE\_SCALABLE\_ENDPOINTS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_ENABLE\_SCALABLE\_ENDPOINTS - MPICH\_CH4\_OFI\_ENABLE\_SCALABLE\_ENDPOINTS - **Description:** - If true, use OFI scalable endpoints. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_ENABLE\_SHARED\_CONTEXTS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_ENABLE\_SHARED\_CONTEXTS - MPICH\_CH4\_OFI\_ENABLE\_SHARED\_CONTEXTS - **Description:** - If set to false (zero), MPICH does not use OFI shared - contexts. If set to -1, it is determined by the OFI capability sets - based on the provider. Otherwise, MPICH tries to use OFI shared - contexts. If they are unavailable, it'll fall back to the mode - without shared contexts. - **Default:** 0 MPIR\_CVAR\_CH4\_OFI\_ENABLE\_MR\_VIRT\_ADDRESS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_ENABLE\_MR\_VIRT\_ADDRESS - MPICH\_CH4\_OFI\_ENABLE\_MR\_VIRT\_ADDRESS - **Description:** - If true, enable virtual addressing for OFI memory regions. - This variable is only meaningful for OFI versions 1.5+. It is - equivalent to using FI\_MR\_BASIC in versions of OFI older than 1.5. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_ENABLE\_MR\_ALLOCATED ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_ENABLE\_MR\_ALLOCATED - MPICH\_CH4\_OFI\_ENABLE\_MR\_ALLOCATED - **Description:** - If true, require all OFI memory regions must be backed by - physical memory pages at the time the registration call is made. - This variable is only meaningful for OFI versions 1.5+. It is - equivalent to using FI\_MR\_BASIC in versions of OFI older than 1.5. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_ENABLE\_MR\_REGISTER\_NULL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_ENABLE\_MR\_REGISTER\_NULL - MPICH\_CH4\_OFI\_ENABLE\_MR\_REGISTER\_NULL - **Description:** - If true, memory registration call supports registering - with NULL addresses. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_ENABLE\_MR\_PROV\_KEY ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_ENABLE\_MR\_PROV\_KEY - MPICH\_CH4\_OFI\_ENABLE\_MR\_PROV\_KEY - **Description:** - If true, enable provider supplied key for OFI memory - regions. This variable is only meaningful for OFI versions 1.5+. It - is equivalent to using FI\_MR\_BASIC in versions of OFI older than - 1.5. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_ENABLE\_TAGGED ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_ENABLE\_TAGGED - MPICH\_CH4\_OFI\_ENABLE\_TAGGED - **Description:** - If true, use tagged message transmission functions in OFI. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_ENABLE\_AM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_ENABLE\_AM - MPICH\_CH4\_OFI\_ENABLE\_AM - **Description:** - If true, enable OFI active message support. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_ENABLE\_RMA ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_ENABLE\_RMA - MPICH\_CH4\_OFI\_ENABLE\_RMA - **Description:** - If true, enable OFI RMA support for MPI RMA operations. - OFI support for basic RMA is always required to implement large - messgage transfers in the active message code path. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_ENABLE\_ATOMICS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_ENABLE\_ATOMICS - MPICH\_CH4\_OFI\_ENABLE\_ATOMICS - **Description:** - If true, enable OFI Atomics support. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_FETCH\_ATOMIC\_IOVECS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_FETCH\_ATOMIC\_IOVECS - MPICH\_CH4\_OFI\_FETCH\_ATOMIC\_IOVECS - **Description:** - Specifies the maximum number of iovecs that can be used by - the OFI provider for fetch\_atomic operations. The default value is - -1, indicating that no value is set. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_ENABLE\_DATA\_AUTO\_PROGRESS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_ENABLE\_DATA\_AUTO\_PROGRESS - MPICH\_CH4\_OFI\_ENABLE\_DATA\_AUTO\_PROGRESS - **Description:** - If true, enable MPI data auto progress. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_ENABLE\_CONTROL\_AUTO\_PROGRESS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_ENABLE\_CONTROL\_AUTO\_PROGRESS - MPICH\_CH4\_OFI\_ENABLE\_CONTROL\_AUTO\_PROGRESS - **Description:** - If true, enable MPI control auto progress. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_ENABLE\_PT2PT\_NOPACK ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_ENABLE\_PT2PT\_NOPACK - MPICH\_CH4\_OFI\_ENABLE\_PT2PT\_NOPACK - **Description:** - If true, enable iovec for pt2pt. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_ENABLE\_HMEM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_ENABLE\_HMEM - MPICH\_CH4\_OFI\_ENABLE\_HMEM - **Description:** - If true, uses GPU direct RDMA support in the provider. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_ENABLE\_MR\_HMEM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_ENABLE\_MR\_HMEM - MPICH\_CH4\_OFI\_ENABLE\_MR\_HMEM - **Description:** - If true, need to register the buffer to use GPU direct - RDMA. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_GPU\_RDMA\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_GPU\_RDMA\_THRESHOLD - MPICH\_CH4\_OFI\_GPU\_RDMA\_THRESHOLD - **Description:** - The threshold to start using GPU direct RDMA. - **Default:** 0 MPIR\_CVAR\_CH4\_OFI\_CONTEXT\_ID\_BITS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_CONTEXT\_ID\_BITS - MPICH\_CH4\_OFI\_CONTEXT\_ID\_BITS - **Description:** - Specifies the number of bits that will be used for - matching the context ID. The default value is -1, indicating that - no value is set and that the default will be defined in the - ofi\_types.h file. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_RANK\_BITS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_RANK\_BITS - MPICH\_CH4\_OFI\_RANK\_BITS - **Description:** - Specifies the number of bits that will be used for - matching the MPI rank. The default value is -1, indicating that no - value is set and that the default will be defined in the - ofi\_types.h file. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_TAG\_BITS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_TAG\_BITS - MPICH\_CH4\_OFI\_TAG\_BITS - **Description:** - Specifies the number of bits that will be used for - matching the user tag. The default value is -1, indicating that no - value is set and that the default will be defined in the - ofi\_types.h file. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_MAJOR\_VERSION ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_MAJOR\_VERSION - MPICH\_CH4\_OFI\_MAJOR\_VERSION - **Description:** - Specifies the major version of the OFI library. The - default is the major version of the OFI library used with MPICH. If - using this CVAR, it is recommended that the user also specifies a - specific OFI provider. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_MINOR\_VERSION ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_MINOR\_VERSION - MPICH\_CH4\_OFI\_MINOR\_VERSION - **Description:** - Specifies the major version of the OFI library. The - default is the minor version of the OFI library used with MPICH. If - using this CVAR, it is recommended that the user also specifies a - specific OFI provider. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_MAX\_RMA\_SEP\_CTX ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_MAX\_RMA\_SEP\_CTX - MPICH\_CH4\_OFI\_MAX\_RMA\_SEP\_CTX - **Description:** - If set to positive, this CVAR specifies the maximum number - of transmit contexts RMA can utilize in a scalable endpoint. This - value is effective only when scalable endpoint is available, - otherwise it will be ignored. - **Default:** 0 MPIR\_CVAR\_CH4\_OFI\_MAX\_EAGAIN\_RETRY ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_MAX\_EAGAIN\_RETRY - MPICH\_CH4\_OFI\_MAX\_EAGAIN\_RETRY - **Description:** - If set to positive, this CVAR specifies the maximum number - of retries of an ofi operations before returning MPIX\_ERR\_EAGAIN. - This value is effective only when the communicator has the - MPI\_OFI\_set\_eagain info hint set to true. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_NUM\_AM\_BUFFERS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_NUM\_AM\_BUFFERS - MPICH\_CH4\_OFI\_NUM\_AM\_BUFFERS - **Description:** - Specifies the number of buffers for receiving active - messages. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_NUM\_OPTIMIZED\_MEMORY\_REGIONS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_NUM\_OPTIMIZED\_MEMORY\_REGIONS - MPICH\_CH4\_OFI\_NUM\_OPTIMIZED\_MEMORY\_REGIONS - **Description:** - Specifies the number of optimized memory regions supported - by the provider. An optimized memory region is used for - lower-overhead, unordered RMA operations. It uses a low-overhead RX - path and additionally, a low-overhead packet format may be used to - target an optimized memory region. - **Default:** 0 MPIR\_CVAR\_CH4\_OFI\_RMA\_PROGRESS\_INTERVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_RMA\_PROGRESS\_INTERVAL - MPICH\_CH4\_OFI\_RMA\_PROGRESS\_INTERVAL - **Description:** - Specifies the interval for manually flushing RMA - operations when automatic progress is not enabled. It the - underlying OFI provider supports auto data progress, this value is - ignored. If the value is -1, this optimization will be turned off. - **Default:** 100 MPIR\_CVAR\_CH4\_OFI\_RMA\_IOVEC\_MAX ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_RMA\_IOVEC\_MAX - MPICH\_CH4\_OFI\_RMA\_IOVEC\_MAX - **Description:** - Specifies the maximum number of iovecs to allocate for RMA - operations to/from noncontiguous buffers. - **Default:** 16384 MPIR\_CVAR\_CH4\_OFI\_EAGER\_MAX\_MSG\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_EAGER\_MAX\_MSG\_SIZE - MPICH\_CH4\_OFI\_EAGER\_MAX\_MSG\_SIZE - **Description:** - This cvar controls the message size at which OFI native - path switches from eager to rendezvous mode. It does not affect the - AM path eager limit. Having this gives a way to reliably test - native non-path. If the number is positive, OFI will init the - MPIDI\_OFI\_global.max\_msg\_size to the value of cvar. If the number - is negative, OFI will init the MPIDI\_OFI\_globa.max\_msg\_size using - whatever provider gives (which might be unlimited for socket - provider). - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_MAX\_NICS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_MAX\_NICS - MPICH\_CH4\_OFI\_MAX\_NICS - **Description:** - If set to positive number, this cvar determines the - maximum number of physical nics to use (if more than one is - available). If the number is -1, underlying netmod or shmmod - automatically uses an optimal number depending on what is detected - on the system up to the limit determined by MPIDI\_MAX\_NICS (in - ofi\_types.h). - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_ENABLE\_MULTI\_NIC\_STRIPING ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_ENABLE\_MULTI\_NIC\_STRIPING - MPICH\_CH4\_OFI\_ENABLE\_MULTI\_NIC\_STRIPING - **Description:** - If true, this cvar enables striping of large messages - across multiple NICs. - **Default:** 0 MPIR\_CVAR\_CH4\_OFI\_MULTI\_NIC\_STRIPING\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_MULTI\_NIC\_STRIPING\_THRESHOLD - MPICH\_CH4\_OFI\_MULTI\_NIC\_STRIPING\_THRESHOLD - **Description:** - Striping will happen for message sizes beyond this - threshold. - **Default:** 1048576 MPIR\_CVAR\_CH4\_OFI\_ENABLE\_MULTI\_NIC\_HASHING ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_ENABLE\_MULTI\_NIC\_HASHING - MPICH\_CH4\_OFI\_ENABLE\_MULTI\_NIC\_HASHING - **Description:** - Multi-NIC hashing means to use more than one NIC to send - and receive messages above a certain size. If set to positive - number, this feature will be turned on. If set to 0, this feature - will be turned off. If the number is -1, MPICH automatically - determines whether to use multi-nic hashing depending on what is - detected on the system (e.g., number of NICs available, number of - processes sharing the NICs). - **Default:** 0 MPIR\_CVAR\_CH4\_OFI\_MULTIRECV\_BUFFER\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_MULTIRECV\_BUFFER\_SIZE - MPICH\_CH4\_OFI\_MULTIRECV\_BUFFER\_SIZE - **Description:** - Controls the multirecv am buffer size. It is recommended - to match this to the hugepage size so that the buffer can be - allocated at the page boundary. - **Default:** 2097152 MPIR\_CVAR\_OFI\_USE\_MIN\_NICS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_OFI\_USE\_MIN\_NICS - MPICH\_OFI\_USE\_MIN\_NICS - **Description:** - If true and all nodes do not have the same number of NICs, - MPICH will fall back to using the fewest number of NICs instead of - returning an error. - **Default:** 1 MPIR\_CVAR\_CH4\_OFI\_ENABLE\_TRIGGERED ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_ENABLE\_TRIGGERED - MPICH\_CH4\_OFI\_ENABLE\_TRIGGERED - **Description:** - If true, enable OFI triggered ops for MPI collectives. - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_ENABLE\_GPU\_PIPELINE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_ENABLE\_GPU\_PIPELINE - MPICH\_CH4\_OFI\_ENABLE\_GPU\_PIPELINE - **Description:** - If true, enable pipeline for GPU data transfer. GPU - pipeline does not support non-contiguous datatypes or mixed buffer - types (i.e. GPU send buffer, host recv buffer). If GPU pipeline is - enabled, the unsupported scenarios will cause undefined behavior if - encountered. - **Default:** 0 MPIR\_CVAR\_CH4\_OFI\_GPU\_PIPELINE\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_GPU\_PIPELINE\_THRESHOLD - MPICH\_CH4\_OFI\_GPU\_PIPELINE\_THRESHOLD - **Description:** - This is the threshold to start using GPU pipeline. - **Default:** 131072 MPIR\_CVAR\_CH4\_OFI\_GPU\_PIPELINE\_BUFFER\_SZ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_GPU\_PIPELINE\_BUFFER\_SZ - MPICH\_CH4\_OFI\_GPU\_PIPELINE\_BUFFER\_SZ - **Description:** - Specifies the buffer size (in bytes) for GPU pipeline data - transfer. - **Default:** 1048576 MPIR\_CVAR\_CH4\_OFI\_GPU\_PIPELINE\_NUM\_BUFFERS\_PER\_CHUNK ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_GPU\_PIPELINE\_NUM\_BUFFERS\_PER\_CHUNK - MPICH\_CH4\_OFI\_GPU\_PIPELINE\_NUM\_BUFFERS\_PER\_CHUNK - **Description:** - Specifies the number of buffers for GPU pipeline data - transfer in each block/chunk of the pool. - **Default:** 32 MPIR\_CVAR\_CH4\_OFI\_GPU\_PIPELINE\_MAX\_NUM\_BUFFERS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_GPU\_PIPELINE\_MAX\_NUM\_BUFFERS - MPICH\_CH4\_OFI\_GPU\_PIPELINE\_MAX\_NUM\_BUFFERS - **Description:** - Specifies the total number of buffers for GPU pipeline - data transfer - **Default:** 32 MPIR\_CVAR\_CH4\_OFI\_GPU\_PIPELINE\_D2H\_ENGINE\_TYPE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_GPU\_PIPELINE\_D2H\_ENGINE\_TYPE - MPICH\_CH4\_OFI\_GPU\_PIPELINE\_D2H\_ENGINE\_TYPE - **Description:** - Specifies the GPU engine type for GPU pipeline on the - sender side, default is MPL\_GPU\_ENGINE\_TYPE\_COMPUTE - **Default:** 0 MPIR\_CVAR\_CH4\_OFI\_GPU\_PIPELINE\_H2D\_ENGINE\_TYPE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_GPU\_PIPELINE\_H2D\_ENGINE\_TYPE - MPICH\_CH4\_OFI\_GPU\_PIPELINE\_H2D\_ENGINE\_TYPE - **Description:** - Specifies the GPU engine type for GPU pipeline on the - receiver side, default is MPL\_GPU\_ENGINE\_TYPE\_COMPUTE - **Default:** 0 MPIR\_CVAR\_MALLOC\_ENABLE\_MMAP ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_MALLOC\_ENABLE\_MMAP - MPICH\_MALLOC\_ENABLE\_MMAP - **Description:** - Set to 0 to disable the use of MMAP by the standard memory - allocator. Doing so provides cache performance improvements for - most cases. Setting to 1 will return to the system default - behaviour. - **Default:** 0 MPIR\_CVAR\_MALLOC\_HEAP\_STEP ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_MALLOC\_HEAP\_STEP - MPICH\_MALLOC\_HEAP\_STEP - **Description:** - Sets the size by which the malloc program heap is expanded - by the memory allocator when new memory is required by the program. - A higher value will generally provide better cache performance, but - will result in a higher memory footprint. The program heap exists - on a per-processes basis. Setting this value to zero will use the - system default. - **Default:** 32 * 1024 * 1024 MPIR\_CVAR\_CH4\_OFI\_PREF\_NIC ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_PREF\_NIC - MPICH\_CH4\_OFI\_PREF\_NIC - **Description:** - Accept the NIC value from a user - **Default:** -1 MPIR\_CVAR\_CH4\_OFI\_DISABLE\_INJECT\_WRITE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_DISABLE\_INJECT\_WRITE - MPICH\_CH4\_OFI\_DISABLE\_INJECT\_WRITE - **Description:** - Avoid use fi\_inject\_write. For some provider, e.g. - tcp;ofi\_rxm, inject write may break the synchronization. - **Default:** 0 MPIR\_CVAR\_CH4\_OFI\_ENABLE\_INJECT ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_ENABLE\_INJECT - MPICH\_CH4\_OFI\_ENABLE\_INJECT - **Description:** - Set MPIR\_CVAR\_CH4\_OFI\_ENABLE\_INJECT=0 to disable buffered - send for small messages. This may help avoid hang due to lack of - global progress. - **Default:** 1 MPIR\_CVAR\_CH4\_OFI\_GPU\_SEND\_ENGINE\_TYPE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_GPU\_SEND\_ENGINE\_TYPE - MPICH\_CH4\_OFI\_GPU\_SEND\_ENGINE\_TYPE - **Description:** - Specifies GPU engine type for GPU pt2pt on the sender - side. - compute - use a compute engine - copy\_high\_bandwidth - use a high-bandwidth copy engine - copy\_low\_latency - use a low-latency copy engine - yaksa - use Yaksa - **Default:** MPIR\_CVAR\_CH4\_OFI\_GPU\_SEND\_ENGINE\_TYPE\_copy\_low\_latency MPIR\_CVAR\_CH4\_OFI\_EAGER\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_OFI\_EAGER\_THRESHOLD - MPICH\_CH4\_OFI\_EAGER\_THRESHOLD - **Description:** - Messages below MPIR\_CVAR\_CH4\_OFI\_EAGER\_THRESHOLD will be - sent eagerly using fi\_tagged interfaces. Messages above the - threshold will perform an MPICH-level rendezvous handshake before - sending the data. If set to -1, MPICH will only perform rendezvous - for messages larger than the provider max\_msg\_size. Note the MPICH - eager/rendezvous threshold is independent of any internal libfabric - provider threshold. - **Default:** -1 MPIR\_CVAR\_UCX\_DT\_RECV ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_UCX\_DT\_RECV - MPICH\_UCX\_DT\_RECV - **Description:** - Variable to select method for receiving noncontiguous data - true - - - Use UCX datatype with pack/unpack callbacks - false - - MPICH will decide to pack/unpack at - completion or use IOVs - based on the datatype - **Default:** 0 MPIR\_CVAR\_CH4\_CMA\_ENABLE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_CMA\_ENABLE - MPICH\_CH4\_CMA\_ENABLE - **Description:** - Set to 1 to manually enable CMA. It is disabled by default - because the CMA requires the ptrace\_scope permission, which is - often disabled. - **Default:** 1 MPIR\_CVAR\_CH4\_IPC\_CMA\_P2P\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_IPC\_CMA\_P2P\_THRESHOLD - MPICH\_CH4\_IPC\_CMA\_P2P\_THRESHOLD - **Description:** - If a send message size is greater than or equal to - MPIR\_CVAR\_CH4\_IPC\_CMA\_P2P\_THRESHOLD (in bytes), then enable - CMA-based single copy protocol for intranode communication. The - environment variable is valid only when the CMA submodule is - enabled. - **Default:** 16384 MPIR\_CVAR\_CH4\_IPC\_RNDV\_PROTOCOL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_IPC\_RNDV\_PROTOCOL - MPICH\_CH4\_IPC\_RNDV\_PROTOCOL - **Description:** - Variable to select the RNDV protocol. - auto - Default RNDV Protocol that enable COOP automatically - rget - RNDV RGET Protocol - rput - RNDV RPUT Protocol - coop - RNDV COOP Protocol - **Default:** MPIR\_CVAR\_CH4\_IPC\_RNDV\_PROTOCOL\_rget MPIR\_CVAR\_CH4\_IPC\_CMA\_COOP\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_IPC\_CMA\_COOP\_THRESHOLD - MPICH\_CH4\_IPC\_CMA\_COOP\_THRESHOLD - **Description:** - If a send message size is greater than or equal to - MPIR\_CVAR\_CH4\_CMA\_COOP\_THRESHOLD (in bytes), then enable CMA-based - COOP rndv protocol for intra-node communication. Messages below - this threshold will use either RPUT or RGET protocol as set by - MPIR\_CVAR\_CH4\_IPC\_RNDV\_PROTOCOL. - **Default:** 16384 MPIR\_CVAR\_CH4\_IPC\_GPU\_HANDLE\_CACHE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_IPC\_GPU\_HANDLE\_CACHE - MPICH\_CH4\_IPC\_GPU\_HANDLE\_CACHE - **Description:** - By default, we will cache ipc handles using the - specialized cache mechanism. If the - gpu-specific backend does not implement a specialized cache, then - we will fallback to - the generic cache mechanism. Users can optionally force the generic - cache mechanism or - disable ipc caching entirely. - generic - use the cache mechanism in the generic layer - specialized - use the cache mechanism in a gpu-specific mpl layer - (if applicable) - disabled - disable caching completely - **Default:** MPIR\_CVAR\_CH4\_IPC\_GPU\_HANDLE\_CACHE\_specialized MPIR\_CVAR\_CH4\_IPC\_GPU\_MAX\_CACHE\_ENTRIES ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_IPC\_GPU\_MAX\_CACHE\_ENTRIES - MPICH\_CH4\_IPC\_GPU\_MAX\_CACHE\_ENTRIES - **Description:** - The maximum number of entries to hold per device in the - cache containing IPC mapped buffers. When an entry is evicted, the - corresponding IPC handle is closed. This value is relevant only - when MPIR\_CVAR\_CH4\_IPC\_GPU\_CACHE\_SIZE=limited. - **Default:** 16 MPIR\_CVAR\_CH4\_IPC\_GPU\_CACHE\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_IPC\_GPU\_CACHE\_SIZE - MPICH\_CH4\_IPC\_GPU\_CACHE\_SIZE - **Description:** - The behavior of the cache containing IPC mapped buffers. - unlimited - don't restrict the cache size - limited - limit the cache size based on - MPIR\_CVAR\_CH4\_IPC\_GPU\_MAX\_CACHE\_ENTRIES - disabled - don't cache mapped IPC buffers - **Default:** MPIR\_CVAR\_CH4\_IPC\_GPU\_CACHE\_SIZE\_limited MPIR\_CVAR\_CH4\_IPC\_GPU\_P2P\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_IPC\_GPU\_P2P\_THRESHOLD - MPICH\_CH4\_IPC\_GPU\_P2P\_THRESHOLD - **Description:** - If a send message size is greater than or equal to - MPIR\_CVAR\_CH4\_IPC\_GPU\_P2P\_THRESHOLD (in bytes), then enable - GPU-based single copy protocol for intranode communication. The - environment variable is valid only when then GPU IPC shmmod is - enabled. - **Default:** (16 * 1024) MPIR\_CVAR\_CH4\_IPC\_ZE\_SHAREABLE\_HANDLE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_IPC\_ZE\_SHAREABLE\_HANDLE - MPICH\_CH4\_IPC\_ZE\_SHAREABLE\_HANDLE - **Description:** - Variable to select implementation for ZE shareable IPC - handle - pidfd - use pidfd\_getfd syscall to implement shareable IPC handle - drmfd - force to use device fd-based shareable IPC handle - **Default:** MPIR\_CVAR\_CH4\_IPC\_ZE\_SHAREABLE\_HANDLE\_drmfd MPIR\_CVAR\_CH4\_IPC\_GPU\_ENGINE\_TYPE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_IPC\_GPU\_ENGINE\_TYPE - MPICH\_CH4\_IPC\_GPU\_ENGINE\_TYPE - **Description:** - By default, select engine type automatically - auto - select automatically - compute - use compute engine - copy\_high\_bandwidth - use high-bandwidth copy engine - copy\_low\_latency - use low-latency copy engine - **Default:** MPIR\_CVAR\_CH4\_IPC\_GPU\_ENGINE\_TYPE\_auto MPIR\_CVAR\_CH4\_IPC\_GPU\_READ\_WRITE\_PROTOCOL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_IPC\_GPU\_READ\_WRITE\_PROTOCOL - MPICH\_CH4\_IPC\_GPU\_READ\_WRITE\_PROTOCOL - **Description:** - By default, use read protocol. - auto - select automatically - read - use read protocol - write - use write protocol if remote device is visible - **Default:** MPIR\_CVAR\_CH4\_IPC\_GPU\_READ\_WRITE\_PROTOCOL\_read MPIR\_CVAR\_CH4\_IPC\_GPU\_RMA\_ENGINE\_TYPE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_IPC\_GPU\_RMA\_ENGINE\_TYPE - MPICH\_CH4\_IPC\_GPU\_RMA\_ENGINE\_TYPE - **Description:** - By default, select engine type automatically - yaksa - don't select, use yaksa - auto - select automatically - compute - use compute engine - copy\_high\_bandwidth - use high-bandwidth copy engine - copy\_low\_latency - use low-latency copy engine - **Default:** MPIR\_CVAR\_CH4\_IPC\_GPU\_RMA\_ENGINE\_TYPE\_auto MPIR\_CVAR\_ALLREDUCE\_IPC\_MSG\_SIZE\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_IPC\_MSG\_SIZE\_THRESHOLD - MPICH\_ALLREDUCE\_IPC\_MSG\_SIZE\_THRESHOLD - **Description:** - Messages below this threshold will fall back to staging to - avoid expensive IPC overheads - **Default:** (16 * 1024) MPIR\_CVAR\_REDUCE\_IPC\_MSG\_SIZE\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_IPC\_MSG\_SIZE\_THRESHOLD - MPICH\_REDUCE\_IPC\_MSG\_SIZE\_THRESHOLD - **Description:** - Messages below this threshold will fall back to staging to - avoid expensive IPC overheads - **Default:** (16 * 1024) MPIR\_CVAR\_DYNAMIC\_GPU\_INIT ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_DYNAMIC\_GPU\_INIT - MPICH\_DYNAMIC\_GPU\_INIT - **Description:** - Enable the dynamic GPU initialization. - **Default:** 1 MPIR\_CVAR\_CH4\_IPC\_MAP\_REPEAT\_ADDR ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_IPC\_MAP\_REPEAT\_ADDR - MPICH\_CH4\_IPC\_MAP\_REPEAT\_ADDR - **Description:** - If an address is used more than once in the last ten send - operations, map it for IPC use even if it is below the IPC - threshold. - **Default:** 1 MPIR\_CVAR\_CH4\_XPMEM\_ENABLE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_XPMEM\_ENABLE - MPICH\_CH4\_XPMEM\_ENABLE - **Description:** - To manually disable XPMEM set to 0. The environment - variable is valid only when the XPMEM submodule is enabled. - **Default:** 1 MPIR\_CVAR\_CH4\_IPC\_XPMEM\_P2P\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_IPC\_XPMEM\_P2P\_THRESHOLD - MPICH\_CH4\_IPC\_XPMEM\_P2P\_THRESHOLD - **Description:** - If a send message size is greater than or equal to - MPIR\_CVAR\_CH4\_IPC\_XPMEM\_P2P\_THRESHOLD (in bytes), then enable - XPMEM-based single copy protocol for intranode communication. The - environment variable is valid only when the XPMEM submodule is - enabled. - **Default:** 1024 MPIR\_CVAR\_BCAST\_POSIX\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_POSIX\_INTRA\_ALGORITHM - MPICH\_BCAST\_POSIX\_INTRA\_ALGORITHM - **Description:** - Variable to select algorithm for intra-node bcast - mpir - - Fallback to MPIR collectives - release\_gather - Force shm optimized algo using release, gather - primitives - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_CH4\_POSIX\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - ipc\_read - Uses read-based collective with ipc - **Default:** MPIR\_CVAR\_BCAST\_POSIX\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_IBCAST\_POSIX\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IBCAST\_POSIX\_INTRA\_ALGORITHM - MPICH\_IBCAST\_POSIX\_INTRA\_ALGORITHM - **Description:** - Variable to select algorithm for intra-node bcast - mpir - - Fallback to MPIR collectives - release\_gather - Force shm optimized algo using release, gather - primitives - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_CH4\_POSIX\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - **Default:** MPIR\_CVAR\_IBCAST\_POSIX\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_REDUCE\_POSIX\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_POSIX\_INTRA\_ALGORITHM - MPICH\_REDUCE\_POSIX\_INTRA\_ALGORITHM - **Description:** - Variable to select algorithm for intra-node reduce - mpir - - Fallback to MPIR collectives - release\_gather - Force shm optimized algo using release, gather - primitives - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_CH4\_POSIX\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - **Default:** MPIR\_CVAR\_REDUCE\_POSIX\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_IREDUCE\_POSIX\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_IREDUCE\_POSIX\_INTRA\_ALGORITHM - MPICH\_IREDUCE\_POSIX\_INTRA\_ALGORITHM - **Description:** - Variable to select algorithm for intra-node reduce - mpir - - Fallback to MPIR collectives - release\_gather - Force shm optimized algo using release, gather - primitives - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_CH4\_POSIX\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - **Default:** MPIR\_CVAR\_IREDUCE\_POSIX\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_ALLREDUCE\_POSIX\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_POSIX\_INTRA\_ALGORITHM - MPICH\_ALLREDUCE\_POSIX\_INTRA\_ALGORITHM - **Description:** - Variable to select algorithm for intra-node allreduce - mpir - - Fallback to MPIR collectives - release\_gather - Force shm optimized algo using release, gather - primitives - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_CH4\_POSIX\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - **Default:** MPIR\_CVAR\_ALLREDUCE\_POSIX\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_BARRIER\_POSIX\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BARRIER\_POSIX\_INTRA\_ALGORITHM - MPICH\_BARRIER\_POSIX\_INTRA\_ALGORITHM - **Description:** - Variable to select algorithm for intra-node barrier - mpir - - Fallback to MPIR collectives - release\_gather - Force shm optimized algo using release, gather - primitives - auto - Internal algorithm selection (can be overridden with - MPIR\_CVAR\_CH4\_POSIX\_COLL\_SELECTION\_TUNING\_JSON\_FILE) - **Default:** MPIR\_CVAR\_BARRIER\_POSIX\_INTRA\_ALGORITHM\_auto MPIR\_CVAR\_ALLTOALL\_POSIX\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLTOALL\_POSIX\_INTRA\_ALGORITHM - MPICH\_ALLTOALL\_POSIX\_INTRA\_ALGORITHM - **Description:** - Variable to select algorithm for intra-node alltoall - mpir - - Fallback to MPIR collectives (default) - ipc\_read - Uses read-based collective with ipc - **Default:** MPIR\_CVAR\_ALLTOALL\_POSIX\_INTRA\_ALGORITHM\_mpir MPIR\_CVAR\_ALLGATHER\_POSIX\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLGATHER\_POSIX\_INTRA\_ALGORITHM - MPICH\_ALLGATHER\_POSIX\_INTRA\_ALGORITHM - **Description:** - Variable to select algorithm for intra-node allgather - mpir - - Fallback to MPIR collectives (default) - ipc\_read - Uses read-based collective with ipc - **Default:** MPIR\_CVAR\_ALLGATHER\_POSIX\_INTRA\_ALGORITHM\_mpir MPIR\_CVAR\_ALLGATHERV\_POSIX\_INTRA\_ALGORITHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLGATHERV\_POSIX\_INTRA\_ALGORITHM - MPICH\_ALLGATHERV\_POSIX\_INTRA\_ALGORITHM - **Description:** - Variable to select algorithm for intra-node allgatherv - mpir - - Fallback to MPIR collectives (default) - ipc\_read - Uses read-based collective with ipc - **Default:** MPIR\_CVAR\_ALLGATHERV\_POSIX\_INTRA\_ALGORITHM\_mpir MPIR\_CVAR\_POSIX\_POLL\_FREQUENCY ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_POSIX\_POLL\_FREQUENCY - MPICH\_POSIX\_POLL\_FREQUENCY - **Description:** - This cvar sets the number of loops before the yield - function is called. A value of 0 disables yielding. - **Default:** 1000 MPIR\_CVAR\_BCAST\_IPC\_READ\_MSG\_SIZE\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_IPC\_READ\_MSG\_SIZE\_THRESHOLD - MPICH\_BCAST\_IPC\_READ\_MSG\_SIZE\_THRESHOLD - **Description:** - Use gpu ipc read bcast only when the message size is - larger than this threshold. - **Default:** 256 MPIR\_CVAR\_ALLTOALL\_IPC\_READ\_MSG\_SIZE\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLTOALL\_IPC\_READ\_MSG\_SIZE\_THRESHOLD - MPICH\_ALLTOALL\_IPC\_READ\_MSG\_SIZE\_THRESHOLD - **Description:** - Use gpu ipc read alltoall only when the message size is - larger than this threshold. - **Default:** 256 MPIR\_CVAR\_ALLGATHER\_IPC\_READ\_MSG\_SIZE\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLGATHER\_IPC\_READ\_MSG\_SIZE\_THRESHOLD - MPICH\_ALLGATHER\_IPC\_READ\_MSG\_SIZE\_THRESHOLD - **Description:** - Use gpu ipc read allgather only when the message size is - larger than this threshold. - **Default:** 256 MPIR\_CVAR\_ALLGATHERV\_IPC\_READ\_MSG\_SIZE\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLGATHERV\_IPC\_READ\_MSG\_SIZE\_THRESHOLD - MPICH\_ALLGATHERV\_IPC\_READ\_MSG\_SIZE\_THRESHOLD - **Description:** - Use gpu ipc read allgatherv only when the message size is - larger than this threshold. - **Default:** 256 MPIR\_CVAR\_POSIX\_NUM\_COLLS\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_POSIX\_NUM\_COLLS\_THRESHOLD - MPICH\_POSIX\_NUM\_COLLS\_THRESHOLD - **Description:** - Use posix optimized collectives (release\_gather) only when - the total number of Bcast, Reduce, Barrier, and Allreduce calls on - the node level communicator is more than this threshold. - **Default:** 5 MPIR\_CVAR\_CH4\_SHM\_POSIX\_EAGER ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_SHM\_POSIX\_EAGER - MPICH\_CH4\_SHM\_POSIX\_EAGER - **Description:** - If non-empty, this cvar specifies which shm posix eager - module to use - **Default:** "" MPIR\_CVAR\_CH4\_POSIX\_COLL\_SELECTION\_TUNING\_JSON\_FILE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_POSIX\_COLL\_SELECTION\_TUNING\_JSON\_FILE - MPICH\_CH4\_POSIX\_COLL\_SELECTION\_TUNING\_JSON\_FILE - **Description:** - Defines the location of tuning file. - **Default:** "" MPIR\_CVAR\_CH4\_POSIX\_COLL\_SELECTION\_TUNING\_JSON\_FILE\_GPU ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_POSIX\_COLL\_SELECTION\_TUNING\_JSON\_FILE\_GPU - MPICH\_CH4\_POSIX\_COLL\_SELECTION\_TUNING\_JSON\_FILE\_GPU - **Description:** - Defines the location of tuning file for GPU. - **Default:** "" MPIR\_CVAR\_CH4\_SHM\_POSIX\_TOPO\_ENABLE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_SHM\_POSIX\_TOPO\_ENABLE - MPICH\_CH4\_SHM\_POSIX\_TOPO\_ENABLE - **Description:** - Controls topology-aware communication in POSIX. - **Default:** 0 MPIR\_CVAR\_CH4\_SHM\_POSIX\_IQUEUE\_NUM\_CELLS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_SHM\_POSIX\_IQUEUE\_NUM\_CELLS - MPICH\_CH4\_SHM\_POSIX\_IQUEUE\_NUM\_CELLS - **Description:** - The number of cells used for the depth of the iqueue. - **Default:** 64 MPIR\_CVAR\_CH4\_SHM\_POSIX\_IQUEUE\_CELL\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_SHM\_POSIX\_IQUEUE\_CELL\_SIZE - MPICH\_CH4\_SHM\_POSIX\_IQUEUE\_CELL\_SIZE - **Description:** - Size of each cell. - **Default:** 16384 MPIR\_CVAR\_CH4\_MVP\_SHM\_POSIX\_NUM\_CELLS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_MVP\_SHM\_POSIX\_NUM\_CELLS - MPICH\_CH4\_MVP\_SHM\_POSIX\_NUM\_CELLS - **Description:** - The number of cells used per sender - **Default:** 128 MPIR\_CVAR\_CH4\_MVP\_SHM\_POSIX\_CELL\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_MVP\_SHM\_POSIX\_CELL\_SIZE - MPICH\_CH4\_MVP\_SHM\_POSIX\_CELL\_SIZE - **Description:** - Size of each cell. - **Default:** 16384 MPIR\_CVAR\_COLL\_SHM\_LIMIT\_PER\_NODE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COLL\_SHM\_LIMIT\_PER\_NODE - MPICH\_COLL\_SHM\_LIMIT\_PER\_NODE - **Description:** - Maximum shared memory created per node for optimized - intra-node collectives (in KB) - **Default:** 65536 MPIR\_CVAR\_BCAST\_INTRANODE\_BUFFER\_TOTAL\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_INTRANODE\_BUFFER\_TOTAL\_SIZE - MPICH\_BCAST\_INTRANODE\_BUFFER\_TOTAL\_SIZE - **Description:** - Total size of the bcast buffer (in bytes) - **Default:** 32768 MPIR\_CVAR\_BCAST\_INTRANODE\_NUM\_CELLS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_INTRANODE\_NUM\_CELLS - MPICH\_BCAST\_INTRANODE\_NUM\_CELLS - **Description:** - Number of cells the bcast buffer is divided into - **Default:** 4 MPIR\_CVAR\_REDUCE\_INTRANODE\_BUFFER\_TOTAL\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_INTRANODE\_BUFFER\_TOTAL\_SIZE - MPICH\_REDUCE\_INTRANODE\_BUFFER\_TOTAL\_SIZE - **Description:** - Total size of the reduce buffer per rank (in bytes) - **Default:** 32768 MPIR\_CVAR\_REDUCE\_INTRANODE\_NUM\_CELLS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_INTRANODE\_NUM\_CELLS - MPICH\_REDUCE\_INTRANODE\_NUM\_CELLS - **Description:** - Number of cells the reduce buffer is divided into, for - each rank - **Default:** 4 MPIR\_CVAR\_BCAST\_INTRANODE\_TREE\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_INTRANODE\_TREE\_KVAL - MPICH\_BCAST\_INTRANODE\_TREE\_KVAL - **Description:** - K value for the kary/knomial tree for intra-node bcast - **Default:** 64 MPIR\_CVAR\_BCAST\_INTRANODE\_TREE\_TYPE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_INTRANODE\_TREE\_TYPE - MPICH\_BCAST\_INTRANODE\_TREE\_TYPE - **Description:** - Tree type for intra-node bcast tree kary - kary tree - type knomial\_1 - knomial\_1 tree type (ranks are added in order from - the left side) knomial\_2 - knomial\_2 tree type (ranks are added in - order from the right side) knomial\_2 is only supported with non - topology aware trees. - **Default:** "kary" MPIR\_CVAR\_REDUCE\_INTRANODE\_MSG\_SIZE\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_INTRANODE\_MSG\_SIZE\_THRESHOLD - MPICH\_REDUCE\_INTRANODE\_MSG\_SIZE\_THRESHOLD - **Description:** - MPIR\_CVAR\_REDUCE\_INTRANODE\_TREE\_KVAL and - MPIR\_CVAR\_REDUCE\_INTRANODE\_TREE\_TYPE are used when the message size - is smaller than or equal to this threshold; - MPIR\_CVAR\_REDUCE\_INTRANODE\_TREE\_KVAL\_LARGE and - MPIR\_CVAR\_REDUCE\_INTRANODE\_TREE\_TYPE\_LARGE are used when the - message size is larger than this threshold. - **Default:** 2048 MPIR\_CVAR\_REDUCE\_INTRANODE\_TREE\_KVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_INTRANODE\_TREE\_KVAL - MPICH\_REDUCE\_INTRANODE\_TREE\_KVAL - **Description:** - K value for the kary/knomial tree for intra-node reduce - **Default:** 4 MPIR\_CVAR\_REDUCE\_INTRANODE\_TREE\_KVAL\_LARGE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_INTRANODE\_TREE\_KVAL\_LARGE - MPICH\_REDUCE\_INTRANODE\_TREE\_KVAL\_LARGE - **Description:** - K value for the kary/knomial tree for intra-node reduce. - Used for large messages. - **Default:** 2 MPIR\_CVAR\_REDUCE\_INTRANODE\_TREE\_TYPE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_INTRANODE\_TREE\_TYPE - MPICH\_REDUCE\_INTRANODE\_TREE\_TYPE - **Description:** - Tree type for intra-node reduce tree kary - kary tree - type knomial\_1 - knomial\_1 tree type (ranks are added in order from - the left side) knomial\_2 - knomial\_2 tree type (ranks are added in - order from the right side) knomial\_2 is only supported with non - topology aware trees. - **Default:** "kary" MPIR\_CVAR\_REDUCE\_INTRANODE\_TREE\_TYPE\_LARGE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_INTRANODE\_TREE\_TYPE\_LARGE - MPICH\_REDUCE\_INTRANODE\_TREE\_TYPE\_LARGE - **Description:** - Tree type for intra-node reduce tree. Used for large - messages. kary - kary tree type knomial\_1 - knomial\_1 tree - type (ranks are added in order from the left side) knomial\_2 - - knomial\_2 tree type (ranks are added in order from the right side) - knomial\_2 is only supported with non topology aware trees. - **Default:** "kary" MPIR\_CVAR\_ENABLE\_INTRANODE\_TOPOLOGY\_AWARE\_TREES ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ENABLE\_INTRANODE\_TOPOLOGY\_AWARE\_TREES - MPICH\_ENABLE\_INTRANODE\_TOPOLOGY\_AWARE\_TREES - **Description:** - Enable collective specific intra-node trees which leverage - the memory hierarchy of a machine. Depends on hwloc to extract the - binding information of each rank. Pick a leader rank per package - (socket), then create a per\_package tree for ranks on a same - package, package leaders tree for package leaders. For Bcast - - Assemble the per\_package and package\_leaders tree in such a way - that leaders interact among themselves first before interacting - with package local ranks. Both the package\_leaders and per\_package - trees are left skewed (children are added from left to right, first - child to be added is the first one to be processed in traversal) - For Reduce - Assemble the per\_package and package\_leaders tree in - such a way that a leader rank interacts with its package local - ranks first, then with the other package leaders. Both the - per\_package and package\_leaders tree is right skewed (children are - added in reverse order, first child to be added is the last one to - be processed in traversal) The tree radix and tree type of - package\_leaders and per\_package tree is - MPIR\_CVAR\_BCAST{REDUCE}\_INTRANODE\_TREE\_KVAL and - MPIR\_CVAR\_BCAST{REDUCE}\_INTRANODE\_TREE\_TYPE respectively for bast - and reduce. But of as now topology aware trees are only kary and - knomial\_1. knomial\_2 is not implemented. - **Default:** 1 MPIR\_CVAR\_BARRIER\_COMPOSITION ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BARRIER\_COMPOSITION - MPICH\_BARRIER\_COMPOSITION - **Description:** - Select composition (inter\_node + intra\_node) for Barrier 0 - Auto selection 1 NM + SHM 2 NM only - **Default:** 0 MPIR\_CVAR\_BCAST\_COMPOSITION ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_BCAST\_COMPOSITION - MPICH\_BCAST\_COMPOSITION - **Description:** - Select composition (inter\_node + intra\_node) for Bcast 0 - Auto selection 1 NM + SHM with explicit send-recv between rank 0 - and root 2 NM + SHM without the explicit send-recv 3 NM only - **Default:** 0 MPIR\_CVAR\_ALLREDUCE\_COMPOSITION ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_COMPOSITION - MPICH\_ALLREDUCE\_COMPOSITION - **Description:** - Select composition (inter\_node + intra\_node) for Allreduce - 0 Auto selection 1 NM + SHM with reduce + bcast 2 NM only - composition 3 SHM only composition 4 Multi leaders based inter node - + intra node composition 5 Intra-node only (flat) IPC and kernel - supported allreduce 6 Two level IPC enabled allreduce using NM + - IPC reduce + IPC Bcast 7 Multi-leader two level IPC enabled - allreduce using reduce\_scatter + allgatherv 8 Pipelined - multi-leader two level IPC enabled allreduce - **Default:** 0 MPIR\_CVAR\_ALLGATHER\_COMPOSITION ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLGATHER\_COMPOSITION - MPICH\_ALLGATHER\_COMPOSITION - **Description:** - Select composition (inter\_node + intra\_node) for Allgather - 0 Auto selection 1 Multi leaders based inter node + intra node - composition 2 NM only composition - **Default:** 0 MPIR\_CVAR\_ALLTOALL\_COMPOSITION ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLTOALL\_COMPOSITION - MPICH\_ALLTOALL\_COMPOSITION - **Description:** - Select composition (inter\_node + intra\_node) for Alltoall - 0 Auto selection 1 Multi leaders based inter node + intra node - composition 2 NM only composition - **Default:** 0 MPIR\_CVAR\_REDUCE\_COMPOSITION ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_COMPOSITION - MPICH\_REDUCE\_COMPOSITION - **Description:** - Select composition (inter\_node + intra\_node) for Reduce 0 - Auto selection 1 NM + SHM with explicit send-recv between rank 0 - and root 2 NM + SHM without the explicit send-recv 3 NM only 4 IPC - and kernel assisted GPU enabled - **Default:** 0 MPIR\_CVAR\_REDUCE\_SCATTER\_COMPOSITION ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_SCATTER\_COMPOSITION - MPICH\_REDUCE\_SCATTER\_COMPOSITION - **Description:** - Select composition (inter\_node + intra\_node) for - Reduce\_scatter 0 Auto selection 1 NM only composition 2 NM only - composition with CPU staging 3 Intra-node only (flat) IPC and - kernel supported reduce\_scatter 4 Pipelined multi-leader two level - IPC enabled reduce\_scatter - **Default:** 0 MPIR\_CVAR\_ALLTOALL\_SHM\_PER\_RANK ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLTOALL\_SHM\_PER\_RANK - MPICH\_ALLTOALL\_SHM\_PER\_RANK - **Description:** - Shared memory region per rank for multi-leaders based - composition for MPI\_Alltoall (in bytes) - **Default:** 4096 MPIR\_CVAR\_ALLGATHER\_SHM\_PER\_RANK ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLGATHER\_SHM\_PER\_RANK - MPICH\_ALLGATHER\_SHM\_PER\_RANK - **Description:** - Shared memory region per rank for multi-leaders based - composition for MPI\_Allgather (in bytes) - **Default:** 4096 MPIR\_CVAR\_NUM\_MULTI\_LEADS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_NUM\_MULTI\_LEADS - MPICH\_NUM\_MULTI\_LEADS - **Description:** - Number of leader ranks per node to be used for - multi-leaders based collective algorithms - **Default:** 4 MPIR\_CVAR\_ALLREDUCE\_SHM\_PER\_LEADER ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_SHM\_PER\_LEADER - MPICH\_ALLREDUCE\_SHM\_PER\_LEADER - **Description:** - Shared memory region per node-leader for multi-leaders - based composition for MPI\_Allreduce (in bytes). By default the size - of the first call to the algorithm will be used. Max shared memory - size is limited to 4MB. - **Default:** -1 MPIR\_CVAR\_ALLREDUCE\_CACHE\_PER\_LEADER ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_CACHE\_PER\_LEADER - MPICH\_ALLREDUCE\_CACHE\_PER\_LEADER - **Description:** - Amount of data reduced in allreduce delta composition's - reduce local step (in bytes). Smaller per-leader message sizes can - avoid cache misses and improve performance. - **Default:** 512 MPIR\_CVAR\_ALLREDUCE\_LOCAL\_COPY\_OFFSETS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_LOCAL\_COPY\_OFFSETS - MPICH\_ALLREDUCE\_LOCAL\_COPY\_OFFSETS - **Description:** - number of offsets in the allreduce delta composition's - local copy The value of 2 performed the best in our 2 NIC test - cases. - **Default:** 2 MPIR\_CVAR\_CH4\_NETMOD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_NETMOD - MPICH\_CH4\_NETMOD - **Description:** - If non-empty, this cvar specifies which network module to - use - **Default:** "" MPIR\_CVAR\_CH4\_SHM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_SHM - MPICH\_CH4\_SHM - **Description:** - If non-empty, this cvar specifies which shm module to use - **Default:** "" MPIR\_CVAR\_CH4\_ROOTS\_ONLY\_PMI ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_ROOTS\_ONLY\_PMI - MPICH\_CH4\_ROOTS\_ONLY\_PMI - **Description:** - Enables an optimized business card exchange over PMI for - node root processes only. - **Default:** 1 MPIR\_CVAR\_CH4\_RUNTIME\_CONF\_DEBUG ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_RUNTIME\_CONF\_DEBUG - MPICH\_CH4\_RUNTIME\_CONF\_DEBUG - **Description:** - If enabled, CH4-level runtime configurations are printed - out - **Default:** 0 MPIR\_CVAR\_CH4\_MT\_MODEL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_MT\_MODEL - MPICH\_CH4\_MT\_MODEL - **Description:** - Specifies the CH4 multi-threading model. Possible values - are: direct (default) lockless - **Default:** "" MPIR\_CVAR\_CH4\_NUM\_VCIS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_NUM\_VCIS - MPICH\_CH4\_NUM\_VCIS - **Description:** - Sets the number of VCIs to be implicitly used (should be a - subset of MPIDI\_CH4\_MAX\_VCIS). - **Default:** 1 MPIR\_CVAR\_CH4\_RESERVE\_VCIS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_RESERVE\_VCIS - MPICH\_CH4\_RESERVE\_VCIS - **Description:** - Sets the number of VCIs that user can explicitly allocate - (should be a subset of MPIDI\_CH4\_MAX\_VCIS). - **Default:** 0 MPIR\_CVAR\_CH4\_COLL\_SELECTION\_TUNING\_JSON\_FILE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_COLL\_SELECTION\_TUNING\_JSON\_FILE - MPICH\_CH4\_COLL\_SELECTION\_TUNING\_JSON\_FILE - **Description:** - Defines the location of tuning file. - **Default:** "" MPIR\_CVAR\_CH4\_COLL\_SELECTION\_TUNING\_JSON\_FILE\_GPU ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_COLL\_SELECTION\_TUNING\_JSON\_FILE\_GPU - MPICH\_CH4\_COLL\_SELECTION\_TUNING\_JSON\_FILE\_GPU - **Description:** - Defines the location of tuning file for GPU. - **Default:** "" MPIR\_CVAR\_CH4\_IOV\_DENSITY\_MIN ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_IOV\_DENSITY\_MIN - MPICH\_CH4\_IOV\_DENSITY\_MIN - **Description:** - Defines the threshold of high-density datatype. The - density is calculated by (datatype\_size / - datatype\_num\_contig\_blocks). - **Default:** 16384 MPIR\_CVAR\_CH4\_PACK\_BUFFER\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_PACK\_BUFFER\_SIZE - MPICH\_CH4\_PACK\_BUFFER\_SIZE - **Description:** - Specifies the number of buffers for packing/unpacking - active messages in each block of the pool. The size here should be - greater or equal to the max of the eager buffer limit of SHM and - NETMOD. - **Default:** 16384 MPIR\_CVAR\_CH4\_NUM\_PACK\_BUFFERS\_PER\_CHUNK ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_NUM\_PACK\_BUFFERS\_PER\_CHUNK - MPICH\_CH4\_NUM\_PACK\_BUFFERS\_PER\_CHUNK - **Description:** - Specifies the number of buffers for packing/unpacking - active messages in each block of the pool. - **Default:** 64 MPIR\_CVAR\_CH4\_MAX\_NUM\_PACK\_BUFFERS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_MAX\_NUM\_PACK\_BUFFERS - MPICH\_CH4\_MAX\_NUM\_PACK\_BUFFERS - **Description:** - Specifies the max number of buffers for packing/unpacking - buffers in the pool. Use 0 for unlimited. - **Default:** 0 MPIR\_CVAR\_CH4\_GPU\_COLL\_SWAP\_BUFFER\_SZ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_GPU\_COLL\_SWAP\_BUFFER\_SZ - MPICH\_CH4\_GPU\_COLL\_SWAP\_BUFFER\_SZ - **Description:** - Specifies the buffer size (in bytes) for GPU collectives - data transfer. - **Default:** 1048576 MPIR\_CVAR\_CH4\_GPU\_COLL\_NUM\_BUFFERS\_PER\_CHUNK ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_GPU\_COLL\_NUM\_BUFFERS\_PER\_CHUNK - MPICH\_CH4\_GPU\_COLL\_NUM\_BUFFERS\_PER\_CHUNK - **Description:** - Specifies the number of buffers for GPU collectives data - transfer in each block/chunk of the pool. - **Default:** 1 MPIR\_CVAR\_CH4\_GPU\_COLL\_MAX\_NUM\_BUFFERS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_GPU\_COLL\_MAX\_NUM\_BUFFERS - MPICH\_CH4\_GPU\_COLL\_MAX\_NUM\_BUFFERS - **Description:** - Specifies the total number of buffers for GPU collectives - data transfer. - **Default:** 256 MPIR\_CVAR\_ALLREDUCE\_MULTI\_LEADER\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_MULTI\_LEADER\_THRESHOLD - MPICH\_ALLREDUCE\_MULTI\_LEADER\_THRESHOLD - **Description:** - The threshold to switch from single-leader 2-level - allreduce algorithm to multi-leader 2-level allreduce algorithm. - **Default:** (8 * 1024 * 1024) MPIR\_CVAR\_ALLREDUCE\_MULTI\_LEADER\_USE\_KERNEL\_ALLGATHERV ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_MULTI\_LEADER\_USE\_KERNEL\_ALLGATHERV - MPICH\_ALLREDUCE\_MULTI\_LEADER\_USE\_KERNEL\_ALLGATHERV - **Description:** - The flag to enable multi-leader 2-level allreduce - algorithm using kernel-based Allgatherv. - **Default:** 1 MPIR\_CVAR\_ALLREDUCE\_PERSISTENT\_GPU\_BUFFER\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_PERSISTENT\_GPU\_BUFFER\_SIZE - MPICH\_ALLREDUCE\_PERSISTENT\_GPU\_BUFFER\_SIZE - **Description:** - Size of pre-allocated persistent GPU buffer for advanced - inter-node allreduce algorithm. - **Default:** (1024 * 1024 * 1024) MPIR\_CVAR\_ALLREDUCE\_INTER\_PIPELINE\_NUM ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_INTER\_PIPELINE\_NUM - MPICH\_ALLREDUCE\_INTER\_PIPELINE\_NUM - **Description:** - The flag is the pipeline number for pipelined inter-node - allreduce algorithm. 0 is auto. - **Default:** 0 MPIR\_CVAR\_ALLREDUCE\_INTER\_PIPELINE\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_INTER\_PIPELINE\_THRESHOLD - MPICH\_ALLREDUCE\_INTER\_PIPELINE\_THRESHOLD - **Description:** - The flag is the pipeline threshold for pipelined - inter-node allreduce algorithm. - **Default:** (32 * 1024 * 1024) MPIR\_CVAR\_ALLTOALL\_NAIVE\_MSG\_SIZE\_LIMIT ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLTOALL\_NAIVE\_MSG\_SIZE\_LIMIT - MPICH\_ALLTOALL\_NAIVE\_MSG\_SIZE\_LIMIT - **Description:** - The staging limit for GPU Alltoall. - **Default:** (8 * 1024) MPIR\_CVAR\_ALLREDUCE\_PIPELINE\_2LVL\_GPU\_NUM\_CHUNKS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_PIPELINE\_2LVL\_GPU\_NUM\_CHUNKS - MPICH\_ALLREDUCE\_PIPELINE\_2LVL\_GPU\_NUM\_CHUNKS - **Description:** - The flag is the pipeline number for pipelined inter-node - allreduce algorithm. 0 is auto. - **Default:** 0 MPIR\_CVAR\_ALLREDUCE\_PIPELINE\_2LVL\_GPU\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_PIPELINE\_2LVL\_GPU\_THRESHOLD - MPICH\_ALLREDUCE\_PIPELINE\_2LVL\_GPU\_THRESHOLD - **Description:** - The flag is the pipeline threshold for pipelined - inter-node allreduce algorithm. - **Default:** (32 * 1024 * 1024) MPIR\_CVAR\_ALLREDUCE\_PIPELINE\_NUM\_CHUNKS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_PIPELINE\_NUM\_CHUNKS - MPICH\_ALLREDUCE\_PIPELINE\_NUM\_CHUNKS - **Description:** - The flag is the number of pipeline chunks for the - allreduce algorithm. 0 is auto. - **Default:** 0 MPIR\_CVAR\_ALLREDUCE\_PIPELINE\_SIZE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_PIPELINE\_SIZE - MPICH\_ALLREDUCE\_PIPELINE\_SIZE - **Description:** - The flag is the pipeline message size for the allreduce - algorithm. - **Default:** (8 * 1024 * 1024) MPIR\_CVAR\_ALLREDUCE\_PIPELINE\_2LVL\_RD\_LIMIT ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_PIPELINE\_2LVL\_RD\_LIMIT - MPICH\_ALLREDUCE\_PIPELINE\_2LVL\_RD\_LIMIT - **Description:** - The flag is the switching point from using RD algorithm to - other algorithms in the kernel-based two-level allreduce algorithm. - **Default:** (16 * 1024 * 1024) MPIR\_CVAR\_ALLREDUCE\_GPU\_RING\_THRESHOLD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_GPU\_RING\_THRESHOLD - MPICH\_ALLREDUCE\_GPU\_RING\_THRESHOLD - **Description:** - The flag is the switching point from using group Allreduce - algorithm to ring Allreduce algorithm in the pipelined kernel-based - two-level allreduce algorithm. - **Default:** INT\_MAX MPIR\_CVAR\_ALLREDUCE\_GPU\_KERNEL\_VECTORIZE\_LIMIT ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ALLREDUCE\_GPU\_KERNEL\_VECTORIZE\_LIMIT - MPICH\_ALLREDUCE\_GPU\_KERNEL\_VECTORIZE\_LIMIT - **Description:** - The flag determines the switching point between vectorized - and non-vectorized execution in the GPU kernel. Vectorized - processing typically offers better performance for small message - sizes. - **Default:** (128 * 1024 * 1024) MPIR\_CVAR\_REDUCE\_SCATTER\_NAIVE\_MSG\_SIZE\_LIMIT ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_REDUCE\_SCATTER\_NAIVE\_MSG\_SIZE\_LIMIT - MPICH\_REDUCE\_SCATTER\_NAIVE\_MSG\_SIZE\_LIMIT - **Description:** - The staging limit for GPU Reduce\_scatter. - **Default:** (8 * 1024) MPIR\_CVAR\_CH4\_GLOBAL\_PROGRESS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_GLOBAL\_PROGRESS - MPICH\_CH4\_GLOBAL\_PROGRESS - **Description:** - If on, poll global progress every once a while. With - per-vci configuration, turning global progress off may improve the - threading performance. - **Default:** 1 MPIR\_CVAR\_CH4\_COMM\_CONNECT\_TIMEOUT ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_COMM\_CONNECT\_TIMEOUT - MPICH\_CH4\_COMM\_CONNECT\_TIMEOUT - **Description:** - The default time out period in seconds for a connection - attempt to the server communicator where the named port exists but - no pending accept. User can change the value for a specified - connection through its info argument. - **Default:** 180 MPIR\_CVAR\_CH4\_ENABLE\_STREAM\_WORKQ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_ENABLE\_STREAM\_WORKQ - MPICH\_CH4\_ENABLE\_STREAM\_WORKQ - **Description:** - Enable stream enqueue operations via stream work queue. - Requires progress thread on the corresponding MPIX stream. - Reference: MPIX\_Stream\_progress and MPIX\_Start\_progress\_thread. - **Default:** 0 MPIR\_CVAR\_CH4\_RMA\_MEM\_EFFICIENT ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_RMA\_MEM\_EFFICIENT - MPICH\_CH4\_RMA\_MEM\_EFFICIENT - **Description:** - If true, memory-saving mode is on, per-target object is - released at the epoch end call. If false, performance-efficient - mode is on, all allocated target objects are cached and freed at - win\_finalize. - **Default:** 0 MPIR\_CVAR\_CH4\_RMA\_ENABLE\_DYNAMIC\_AM\_PROGRESS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_RMA\_ENABLE\_DYNAMIC\_AM\_PROGRESS - MPICH\_CH4\_RMA\_ENABLE\_DYNAMIC\_AM\_PROGRESS - **Description:** - If true, allows RMA synchronization calls to dynamically - reduce the frequency of internal progress polling for incoming RMA - active messages received on the target process. The RMA - synchronization call initially polls progress with a low frequency - (defined by MPIR\_CVAR\_CH4\_RMA\_AM\_PROGRESS\_LOW\_FREQ\_INTERVAL) to - reduce synchronization overhead. Once any RMA active message has - been received, it will always poll progress once at every - synchronization call to ensure prompt target-side progress. - Effective only for passive target synchronization - MPI\_Win\_flush{\_all} and MPI\_Win\_flush\_local{\_all}. - **Default:** 0 MPIR\_CVAR\_CH4\_RMA\_AM\_PROGRESS\_INTERVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_RMA\_AM\_PROGRESS\_INTERVAL - MPICH\_CH4\_RMA\_AM\_PROGRESS\_INTERVAL - **Description:** - Specifies a static interval of progress polling for - incoming RMA active messages received on the target process. - Effective only for passive-target synchronization - MPI\_Win\_flush{\_all} and MPI\_Win\_flush\_local{\_all}. Interval - indicates the number of performed flush calls before polling. It is - counted globally across all windows. Invalid when - MPIR\_CVAR\_CH4\_RMA\_ENABLE\_DYNAMIC\_AM\_PROGRESS is true. - **Default:** 1 MPIR\_CVAR\_CH4\_RMA\_AM\_PROGRESS\_LOW\_FREQ\_INTERVAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_CH4\_RMA\_AM\_PROGRESS\_LOW\_FREQ\_INTERVAL - MPICH\_CH4\_RMA\_AM\_PROGRESS\_LOW\_FREQ\_INTERVAL - **Description:** - Specifies the interval of progress polling with low - frequency for incoming RMA active message received on the target - process. Effective only for passive-target synchronization - MPI\_Win\_flush{\_all} and MPI\_Win\_flush\_local{\_all}. Interval - indicates the number of performed flush calls before polling. It is - counted globally across all windows. Used when - MPIR\_CVAR\_CH4\_RMA\_ENABLE\_DYNAMIC\_AM\_PROGRESS is true. - **Default:** 100 MPIR\_CVAR\_PT2PT\_JSON\_TUNING\_FILE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_PT2PT\_JSON\_TUNING\_FILE - MPICH\_PT2PT\_JSON\_TUNING\_FILE - **Description:** - If non-empty, this cvar specifies a file for - architecture-specific tuning - **Default:** "" MPIR\_CVAR\_GENQ\_SHMEM\_POOL\_FREE\_QUEUE\_SENDER\_SIDE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_GENQ\_SHMEM\_POOL\_FREE\_QUEUE\_SENDER\_SIDE - MPICH\_GENQ\_SHMEM\_POOL\_FREE\_QUEUE\_SENDER\_SIDE - **Description:** - The genq shmem code allocates pools of cells on each - process and, when needed, a cell is removed from the pool and - passed to another process. This can happen by either removing a - cell from the pool of the sending process or from the pool of the - receiving process. This CVAR determines which pool to use. If true, - the cell will come from the sender-side. If false, the cell will - com from the receiver-side. There are specific advantages of using - receiver-side cells when combined with the "avx" fast configure - option, which allows MPICH to use AVX streaming copy - intrintrinsics, when available, to avoid polluting the cache of the - sender with the data being copied to the receiver. Using - receiver-side cells does have the trade-off of requiring an MPMC - lock for the free queue rather than an MPSC lock, which is used for - sender-side cells. Initial performance analysis shows that using - the MPMC lock in this case had no significant performance loss. By - default, the queue will continue to use sender-side queues until - the performance impact is verified. - **Default:** 1 MPIR\_CVAR\_ENABLE\_HCOLL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ENABLE\_HCOLL - MPICH\_ENABLE\_HCOLL - **Description:** - Enable hcoll collective support. - **Default:** 0 MPIR\_CVAR\_COLL\_SCHED\_DUMP ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_COLL\_SCHED\_DUMP - MPICH\_COLL\_SCHED\_DUMP - **Description:** - Print schedule data for nonblocking collective operations. - **Default:** 0 MPIR\_CVAR\_SHM\_RANDOM\_ADDR\_RETRY ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_SHM\_RANDOM\_ADDR\_RETRY - MPICH\_SHM\_RANDOM\_ADDR\_RETRY - **Description:** - The default number of retries for generating a random - address. A retrying involves only local operations. - **Default:** 100 MPIR\_CVAR\_SHM\_SYMHEAP\_RETRY ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_SHM\_SYMHEAP\_RETRY - MPICH\_SHM\_SYMHEAP\_RETRY - **Description:** - The default number of retries for allocating a symmetric - heap in shared memory. A retrying involves collective communication - over the group in the shared memory. - **Default:** 100 MPIR\_CVAR\_ENABLE\_HEAVY\_YIELD ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - **Aliases:** - MVP\_ENABLE\_HEAVY\_YIELD - MPICH\_ENABLE\_HEAVY\_YIELD - **Description:** - If enabled, use nanosleep to ensure other threads have a - chance to grab the lock. Note: this may not work with some thread - runtimes, e.g. non-preemptive user-level threads. - **Default:** 0