Modularize POTRF, make DPLASMA optional. #222
Closed
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
therault
force-pushed
the
potrf-composition
branch
from
20dbd01 to
30a3b6b
Compare
… a deadlock (looking for it, probably a task that is partially triggered by mistake)
in the previous fence(), when inside ttg_finalize(), so that if we're using user-trigger termination detection the parsec_context_wait() on that context will complete during the call to destroy_worlds(). final_task() uses the parsec_taskpool_started member to remember if it is necessay to update the number of pending actions again, allowing the user to terminate the taskpool before, if they want to.
…bly because of porting to new master
continue to move things in appropriate .h file start to refactor profiling
…ributed yet. Test feature with dpotrf. Not integrated well at this time: should that be a functionality of world? Of world::impl? Or of something else?
…ommand of GraphViz to group TTs under the same TTG
… the user has enabled it via MCA.
therault
force-pushed
the
potrf-composition
branch
from
d3325f3 to
1d53756
Compare
…creation time, so we can capture the TTG hierarchy in the naming of events; use the actual task class id to store the event identifier.
…c_task_classes is not contiguous. Add some TTG-specific tracing in parsec traces to estimate amount of time spent in dependency management.
…ers of the world so it compiles with both parsec and madness backends.
…filing is enabled, parsec_task_t stores the hash of the task key in the locals, and the pointer to the key after that. Then, the hash is used to identify the task globally in the DOT file, and the key to provide the human-readable name of the task. If set_arg_local_impl is called from a dummy task, we don't output the dependency; we create the destination task (and destroy it) at the end of set_arg_impl, if the destination task is remote, just to be able to output the dependency.
9 tasks
…als from tts *before* they are moved into the ttg's vector of tt; improve passing arguments to testing_dpotri; make testing_dpotri silent for performance measurement; introduce a loop for performance measurement in testing_dpotri
…) we register Ops that can be used in multiple taskpools, and B) we destruct and re-create the taskpool at each fence()
…les, as they are needed for the non-sequential version -- Need to check with Joseph how many of the deep-copy are done, and why. Check validty of POTRF and POTRI calls.
|
Superseded by PR #238 |
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.
Make POTRF composable via make_ttg.
The goal is threefold: