file dependency resolution #5
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
- in some cases fi_trecvmsg(PEEK) returns -NOMSG error which should be interpreted as "normal", no error state, but it was processed as error. added special case to gracefully exit from function on NOMSG state Change-Id: I532d37f536117f3ebfca78cc36227c34a587cddf Signed-off-by: Ken Raffenetti <[email protected]>
Signed-off-by: Yanfei Guo <[email protected]>
Declare these functions statically at the top of the file. There is no need to expose further at this time. Signed-off-by: Yanfei Guo <[email protected]>
Signed-off-by: Yanfei Guo <[email protected]>
Missed this function in [345b175]. Signed-off-by: Ken Raffenetti <[email protected]>
Signed-off-by: Ken Raffenetti <[email protected]>
Signed-off-by: Ken Raffenetti <[email protected]>
Signed-off-by: Ken Raffenetti <[email protected]>
This is a follow up patch to 3baf0de. Signed-off-by: Ken Raffenetti <[email protected]>
No reviewer.
No reviewer.
Signed-off-by: Ken Raffenetti <[email protected]>
To allow lower layers to choose a specific collective implemenation, expose them as non-static functions. We start with four collectives - barrier, bcast, reduce and allreduce. Signed-off-by: Ken Raffenetti <[email protected]>
Extend device and nm/shm-level collective operations to accept a struct of parameters to facilitate optimal algorithm selection. Starting with barrier, bcast, reduce and allreduce - each operation is split into a "select" and "call" phases. Signed-off-by: Ken Raffenetti <[email protected]>
test/mpi/spawn/spaiccreate2 existed but not listed in testlist. Signed-off-by: Ken Raffenetti <[email protected]>
Signed-off-by: Ken Raffenetti <[email protected]>
No reviewer.
Fixes the bug that the startall function do not perform rank-address translation before calling NM_mpi_isend and NM_mpi_irecv. Signed-off-by: Ken Raffenetti <[email protected]>
By referring to MPIT_PVARs usage in CH3, adding corresponding parts into CH4 active message level. Signed-off-by: Ken Raffenetti <[email protected]>
These settings are outdated for modern networks like 10Gb ethernet. Even better, operating systems will now dynamically tune these for best performance, so we can avoid setting them altogether. Closes pmodels#2635 Signed-off-by: Rob Latham <[email protected]>
Current logic for determining the number of VNIs created may end up claiming all available contexts in the hardware. This patch changes the default number of VNIs to one. A user may set MPIR_CVAR_CH4_OFI_MAX_VNIS for more VNIs. Reviewed-by: Rubasri Kalidas <[email protected]> Reviewed-by: Chongxiao Cao <[email protected]> Signed-off-by: Ken Raffenetti <[email protected]>
There was some missing functionality, mismatched variable names, and outright wrong code in the Solaris thread support header. This should get things back to a buildable state. No reviewer.
In order to detect whether find and xargs are supported on the target
platform, autogen.sh runs the find command on the whole MPICH
directory ("."). This is overfill as we only want to detect that those
features exist. On some platforms (e.g., with NFS), this step can take
very a long time. This patch solves this issue by reducing the scope
of find to the ./maint directory.
Signed-off-by: Ken Raffenetti <[email protected]>
The precedence of [] is higher than *. I.e. *bitarray[i] means to first deref by the index i and the pointer, but needed here is to first deref the pointer and then the array. Signed-off-by: Ken Raffenetti <[email protected]>
We need to use FCFLAGS in order to test for "kind-ness" the way the rest of ROMIO will be built. Reported-by: Kurt Glaesemann <[email protected]> Signed-off-by: Ken Raffenetti <[email protected]>
The redscat algorithm in reduce and allreduce needs a load distribution when count is not a multiplier of pof2. This patch reduces the imbalance. See pmodels#2637. Reported-by: Mikhail Kurnosov [email protected] Signed-off-by: Ken Raffenetti <[email protected]>
AlexeyMalkhanov
force-pushed
the
coll-select
branch
from
93630ca to
e41a803
Compare
Signed-off-by: Ken Raffenetti <[email protected]>
Signed-off-by: Ken Raffenetti <[email protected]>
raffenet
force-pushed
the
coll-select
branch
from
e41a803 to
a2496cb
Compare
pmrsbot
pushed a commit
that referenced
this pull request
Jun 23, 2023
mpl/gpu: Implement fast memcpy for ze backend (PR #5)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
@raffenet I've created one more PR where I removed functions declarations in *_coll_select.h files and moved specific algorithms realization into *_coll_impl.h from *_coll.h