On systems where numTasks is not evenly divisible by 'tasksPerNode' we were
seeing some nodes reading multiple files while others read none after
reordering.
Commonly all nodes have the same number of tasks but there is nothing
requiring that to be the case. Imagine having 64 tasks running against 4
nodes which can run 20 tasks each. Here you get three groups of 20 and one
group of 4. On this sytem nodes running in the group of 4 were previously
getting tasksPerNode of 4 which meant they reordered tasks differently than
the nodes which got tasksPerNode of 20.
The key to fixing this is ensuring that every node reorders tasks the same
way, which means ensuring they all use the same input values. Obviously on
systems where the number of tasks per node is inconsistent the reordering will
also be inconsistent (some tasks may end up on the same node, or not as far
separated as desired, etc.) but at least this way you'll always end up with a
1:1 reordering.
- Renamed nodes/nodeCount to numNodes
- Renamed tasksPerNode to numTasksOnNode0
- Ensured that numTasksOnNode0 will always have the same value regardless of
which node you're on
- Removed inconsistently used globals numTasksWorld and tasksPerNode and
replaced with per-test params equivalents
- Added utility functions for setting these values:
- numNodes -> GetNumNodes
- numTasks -> GetNumTasks
- numTasksOnNode0 -> GetNumNodesOnTask0
- Improved MPI_VERSION < 3 logic for GetNumNodes so it works when numTasks is
not evenly divisible by numTasksOnNode0
- Left 'nodes' and 'tasksPerNode' in output alone to not break compatibility
- Allowed command-line params to override numTasks, numNodes, and
numTasksOnNode0 but default to using the MPI-calculated values
Context: Some backends may have used different names in
the past (like IME backend use to be IM). Legacy scripts
may break.
This patch adds a legacy name option in the aiori structure.
Both name and legacy name work to select the interface.
But the following warning is printed if the legacy name is used:
ior WARNING: [legacy name] backend is deprecated use [name] instead.
From Pnetcdf 1.7, the MPI datatype corresponding to NC_BYTE is change from MPI_BYTE to MPI_SIGNED_CHAR. If not change, running IOR with NCMPI will cause a fatal error as below:
ERROR in aiori-NCMPI.c (line 287): cannot write to data set.
ERROR: NetCDF: Not a valid data type or _FillValue type mismatch.
This commit makes changes to the AIORI backends to add support for
abstacting statfs, mkdir, rmdir, stat, and access. These new
abstractions are used by a modified mdtest. Some changes:
- Require C99. Its 2017 and most compilers now support C11. The
benefits of using C99 include subobject naming (for aiori backend
structs), and fixed size integers (uint64_t). There is no reason to
use the non-standard long long type.
- Moved some of the aiori code into aiori.c so it can be used by both
mdtest and ior.
- Code cleanup of mdtest. This is mostly due to the usage of the IOR
backends rather than a mess of #if code.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
Clean up the header files to only contain those things that
need to be shared between .c files.
Functions that are not shared are now declared static to
make their file scope explicit. Functions that ARE shared
are declared in appropriate headers.
I am not going to claim that I caugh everything, but at
least it is a good start.