Directory CCA/Components/Schedulers/

Lines of Code

Developers

Most Recent Commits

ahumphrey 2012-12-13 17:28 Rev.: 49697

Merge changes in commit on branches/arches_alcc_dev r49696 to HEAD.
We now have GPU-RMCRT operational within Arches.

3 lines of code changed in 2 files:

• CCA/Components/Schedulers: UnifiedScheduler.cc (new)

ahumphrey 2012-12-11 17:12 Rev.: 49681

Fix warning about redeclared __host__ __device__ function. Some UnifiedScheduler formatting.

2 lines of code changed in 1 file:

• CCA/Components/Schedulers: UnifiedScheduler.cc (+2 -2)

ahumphrey 2012-11-01 14:03 Rev.: 49516

Report correct info when running UnifiedScheduler in SingleProcessor mode.

7 lines of code changed in 1 file:

• CCA/Components/Schedulers: UnifiedScheduler.cc (+7 -3)

qymeng 2012-10-31 15:48 Rev.: 49512

fix a subscheduling related timing bug

0 lines of code changed in 2 files:

• CCA/Components/Schedulers: MPIScheduler.cc (changed)

jas 2012-10-22 23:47 Rev.: 49484

Update the copyright notice to include the MIT License and fix the formatting
issues with the spurious "\".

245 lines of code changed in 96 files:

• CCA/Components/Schedulers: BatchReceiveHandler.h (+5 -3), CommRecMPI.cc (+5 -3), CommRecMPI.h (new), DWDatabase.h (+5 -3), DependencyException.cc (new), DependencyException.h (new), DetailedTasks.cc (new), DetailedTasks.h (+5 -3), DynamicMPIScheduler.cc (new), DynamicMPIScheduler.h (new), GPUThreadedMPIScheduler.cc (+5 -3), GPUThreadedMPIScheduler.h (+5 -3), IncorrectAllocation.cc (+5 -3), IncorrectAllocation.h (new), MPIScheduler.cc (+5 -3), MPIScheduler.h (+5 -3), MemoryLog.cc (+5 -3), MemoryLog.h (+5 -3), MessageLog.cc (+5 -3), MessageLog.h (new), MixedScheduler.cc (+5 -3), MixedScheduler.h (new), NullScheduler.cc (+5 -3), OnDemandDataWarehouse.cc (+5 -3), OnDemandDataWarehouse.h (+5 -3), OnDemandDataWarehouseP.h (new), Relocate.cc (+5 -3), Relocate.h (+5 -3), SchedulerCommon.cc (new), SchedulerCommon.h (new), SchedulerFactory.cc (+5 -3), SchedulerFactory.h (+5 -3), SendState.cc (+5 -3), SendState.h (new), SingleProcessorScheduler.cc (+5 -3), SingleProcessorScheduler.h (new), TaskGraph.cc (+5 -3), TaskGraph.h (+5 -3), ThreadPool.cc (new), ThreadPool.h (+5 -3), ThreadedMPIScheduler.cc (new), ThreadedMPIScheduler.h (+5 -3), ThreadedMPIScheduler2.cc (+5 -3), ThreadedMPIScheduler2.h (new), UnifiedScheduler.cc (+5 -3), UnifiedScheduler.h (new), Util.cc (+5 -3), sub.mk (+5 -3), templates.cc (+5 -3)

jas 2012-10-12 16:59 Rev.: 49413

Squashed commit of the following:

commit a1e8c675d139003247378067ec0431806029cb14
Merge: 5e29398 49ca967
Author: jas <jas@aee3fe44-4ef4-0310-ac56-cc817e9d0e43>
Date: Fri Oct 12 23:36:18 2012 +0000

Update branch with latest changes from trunk.

git-svn-id: https://gforge.sci.utah.edu/svn/uintah/branches/UintahRelease_1_5_0@49412 aee3fe44-4ef4-0310-ac56-cc817e9d0e43

commit 5e293985e95fc70fd682fbf31fb93d7b72f1108e
Author: jas <jas@aee3fe44-4ef4-0310-ac56-cc817e9d0e43>
Date: Fri Oct 12 23:36:07 2012 +0000

Fix missing slash in installation guide.

Refactor the boost discovery mechanism in configure.ac.

git-svn-id: https://gforge.sci.utah.edu/svn/uintah/branches/UintahRelease_1_5_0@49411 aee3fe44-4ef4-0310-ac56-cc817e9d0e43

commit dd2caee9921c7d89791d866bccc1adbddf967f04
Author: jas <jas@aee3fe44-4ef4-0310-ac56-cc817e9d0e43>
Date: Fri Oct 12 23:35:56 2012 +0000

Update build instructions for petsc and hypre for Fedora and CentOS.

Add BASE_LIB_DIR to the Boost section in configure.ac.

git-svn-id: https://gforge.sci.utah.edu/svn/uintah/branches/UintahRelease_1_5_0@49410 aee3fe44-4ef4-0310-ac56-cc817e9d0e43

commit c211e5394202174e69c4781b1bfd937e6a4e9e70
Merge: 425f4ca 6255204
Author: jas <jas@aee3fe44-4ef4-0310-ac56-cc817e9d0e43>
Date: Fri Oct 12 23:35:14 2012 +0000

Update the release branch with changes from master.

git-svn-id: https://gforge.sci.utah.edu/svn/uintah/branches/UintahRelease_1_5_0@49409 aee3fe44-4ef4-0310-ac56-cc817e9d0e43

commit 425f4caed59a4864374dbed64d46ade4d5fa5eee
Author: jas <jas@aee3fe44-4ef4-0310-ac56-cc817e9d0e43>
Date: Fri Oct 12 23:33:39 2012 +0000

Remove all remaining pieces of the Angio component.

git-svn-id: https://gforge.sci.utah.edu/svn/uintah/branches/UintahRelease_1_5_0@49408 aee3fe44-4ef4-0310-ac56-cc817e9d0e43

commit 5a92abece778059b6a1765ebd971b71744f64793
Author: jas <jas@aee3fe44-4ef4-0310-ac56-cc817e9d0e43>
Date: Fri Oct 12 23:33:24 2012 +0000

Remove the Angio component.

git-svn-id: https://gforge.sci.utah.edu/svn/uintah/branches/UintahRelease_1_5_0@49407 aee3fe44-4ef4-0310-ac56-cc817e9d0e43

commit 78e24afb3a187a45cf3669ff4588d096c8d38b8f
Author: jas <jas@aee3fe44-4ef4-0310-ac56-cc817e9d0e43>
Date: Fri Oct 12 23:33:11 2012 +0000

Remove the Dan_files/ from the StandAlone/inputs/UintahRelease/.

git-svn-id: https://gforge.sci.utah.edu/svn/uintah/branches/UintahRelease_1_5_0@49406 aee3fe44-4ef4-0310-ac56-cc817e9d0e43

commit a938560edb0607e2e4aa63cb742b4b55b7f62cd7
Author: jas <jas@aee3fe44-4ef4-0310-ac56-cc817e9d0e43>
Date: Fri Oct 12 23:33:01 2012 +0000

Notes about compiling boost and cmake and changes to the Centos section.

git-svn-id: https://gforge.sci.utah.edu/svn/uintah/branches/UintahRelease_1_5_0@49405 aee3fe44-4ef4-0310-ac56-cc817e9d0e43

commit 0c7aa437777c336f92dd8be971c1e1a1cf01c367
Author: jas <jas@aee3fe44-4ef4-0310-ac56-cc817e9d0e43>
Date: Fri Oct 12 23:32:53 2012 +0000

Add wasatch build dependences for opensuse.

git-svn-id: https://gforge.sci.utah.edu/svn/uintah/branches/UintahRelease_1_5_0@49404 aee3fe44-4ef4-0310-ac56-cc817e9d0e43

commit 23b71be7d9ffd177b81f64498411584c055504e1
Author: jas <jas@aee3fe44-4ef4-0310-ac56-cc817e9d0e43>
Date: Fri Oct 12 23:32:46 2012 +0000

Update dependencies for CentOS for building Wasatch.

git-svn-id: https://gforge.sci.utah.edu/svn/uintah/branches/UintahRelease_1_5_0@49403 aee3fe44-4ef4-0310-ac56-cc817e9d0e43

commit 5b99e25a6306b02cc8ecb9baad84fa3d470e194f
Author: jas <jas@aee3fe44-4ef4-0310-ac56-cc817e9d0e43>
Date: Fri Oct 12 23:32:38 2012 +0000

Update packages necessary for compiling Wasatch.

git-svn-id: https://gforge.sci.utah.edu/svn/uintah/branches/UintahRelease_1_5_0@49402 aee3fe44-4ef4-0310-ac56-cc817e9d0e43

commit add8b7cbd6668abbe4f57f584dab24c8552b6da8
Author: tsaad <tsaad@aee3fe44-4ef4-0310-ac56-cc817e9d0e43>
Date: Tue Oct 2 20:36:32 2012 +0000

remove obsolete input files.

git-svn-id: https://gforge.sci.utah.edu/svn/uintah/branches/UintahRelease_1_5_0@49330 aee3fe44-4ef4-0310-ac56-cc817e9d0e43

commit f7ac87d4fd61d896eb1ba9f52a0b6cd2970783e5
Author: jas <jas@aee3fe44-4ef4-0310-ac56-cc817e9d0e43>
Date: Sun Sep 30 02:11:34 2012 +0000

Use the 1.5.0 git branch for the various Wasatch3P libraries.

git-svn-id: https://gforge.sci.utah.edu/svn/uintah/branches/UintahRelease_1_5_0@49311 aee3fe44-4ef4-0310-ac56-cc817e9d0e43

commit 25156c6ba612634f5d66a5fb1be670fd5a82a23b
Author: jas <jas@aee3fe44-4ef4-0310-ac56-cc817e9d0e43>
Date: Sat Sep 29 03:41:39 2012 +0000

Remove the release file in the build_scripts/.

git-svn-id: https://gforge.sci.utah.edu/svn/uintah/branches/UintahRelease_1_5_0@49308 aee3fe44-4ef4-0310-ac56-cc817e9d0e43

commit 94b484f3b209b8b5d8ee73bae0f9ba54c91592ef
Author: jas <jas@aee3fe44-4ef4-0310-ac56-cc817e9d0e43>
Date: Sat Sep 29 03:41:31 2012 +0000

Change == to = in all "test statements".

git-svn-id: https://gforge.sci.utah.edu/svn/uintah/branches/UintahRelease_1_5_0@49307 aee3fe44-4ef4-0310-ac56-cc817e9d0e43

commit daeab4316362d0081fb8a66832734a3ffdfbff1f
Author: jas <jas@aee3fe44-4ef4-0310-ac56-cc817e9d0e43>
Date: Sat Sep 29 03:41:23 2012 +0000

Update to version 1.5.

git-svn-id: https://gforge.sci.utah.edu/svn/uintah/branches/UintahRelease_1_5_0@49306 aee3fe44-4ef4-0310-ac56-cc817e9d0e43

commit e3d51281308a5fbbde3f097e5826e85686fe9254
Author: jas <jas@aee3fe44-4ef4-0310-ac56-cc817e9d0e43>
Date: Sat Sep 29 03:41:15 2012 +0000

Update release files to include Wasatch component.

Add option to build Wasatch3P libraries for a particular hash tag.

git-svn-id: https://gforge.sci.utah.edu/svn/uintah/branches/UintahRelease_1_5_0@49305 aee3fe44-4ef4-0310-ac56-cc817e9d0e43

commit 9a8451d8a5b1fbf36626b3727dad48d7184b5e6d
Merge: 984d356 0a25aa2
Author: jas <jas@aee3fe44-4ef4-0310-ac56-cc817e9d0e43>
Date: Sat Sep 29 03:38:16 2012 +0000

Update copyright notice.

git-svn-id: https://gforge.sci.utah.edu/svn/uintah/branches/UintahRelease_1_5_0@49304 aee3fe44-4ef4-0310-ac56-cc817e9d0e43

commit 984d3566f78e7e1c2b2c29ea0368451ae768119a
Author: jas <jas@aee3fe44-4ef4-0310-ac56-cc817e9d0e43>
Date: Fri Sep 28 18:39:29 2012 +0000

Uintah 1.5.0 Branch.

git-svn-id: https://gforge.sci.utah.edu/svn/uintah/branches/UintahRelease_1_5_0@49299 aee3fe44-4ef4-0310-ac56-cc817e9d0e43

commit 0a25aa2731b9c75fcf609f70093484af5be5f23a
Author: John Schmidt <john.schmidt@utah.edu>
Date: Thu Sep 27 01:50:47 2012 -0600

Updated copyright header.

983 lines of code changed in 49 files:

• CCA/Components/Schedulers: BatchReceiveHandler.h (+20 -28), CommRecMPI.cc (+20 -28), CommRecMPI.h (+20 -28), DWDatabase.h (+21 -30), DependencyException.cc (+20 -28), DependencyException.h (+20 -29), DetailedTasks.cc (+19 -27), DetailedTasks.h (+20 -28), DynamicMPIScheduler.cc (+20 -28), DynamicMPIScheduler.h (+21 -30), GPUThreadedMPIScheduler.cc (+19 -27), GPUThreadedMPIScheduler.h (+21 -30), IncorrectAllocation.cc (+20 -28), IncorrectAllocation.h (+20 -29), MPIScheduler.cc (+20 -28), MPIScheduler.h (+20 -29), MemoryLog.cc (+20 -28), MemoryLog.h (+20 -28), MessageLog.cc (+20 -28), MessageLog.h (+20 -29), MixedScheduler.cc (+20 -28), MixedScheduler.h (+20 -29), NullScheduler.cc (+20 -28), OnDemandDataWarehouse.cc (+20 -28), OnDemandDataWarehouse.h (+21 -30), OnDemandDataWarehouseP.h (+20 -28), Relocate.cc (+20 -28), Relocate.h (+20 -29), SchedulerCommon.cc (+20 -28), SchedulerCommon.h (+20 -29), SchedulerFactory.cc (+19 -27), SchedulerFactory.h (+20 -28), SendState.cc (+20 -28), SendState.h (+21 -30), SingleProcessorScheduler.cc (+20 -28), SingleProcessorScheduler.h (+20 -29), TaskGraph.cc (+20 -28), TaskGraph.h (+20 -29), ThreadPool.cc (+20 -28), ThreadPool.h (+20 -28), ThreadedMPIScheduler.cc (+20 -28), ThreadedMPIScheduler.h (+20 -29), ThreadedMPIScheduler2.cc (+20 -28), ThreadedMPIScheduler2.h (+21 -30), UnifiedScheduler.cc (+19 -27), UnifiedScheduler.h (+21 -30), Util.cc (+20 -28), sub.mk (+20 -23), templates.cc (+20 -28)

qymeng 2012-10-09 11:01 Rev.: 49390

Bug fix: passing coarsestLevelwithParticles from scheduler instead of getting from local patches.

31 lines of code changed in 4 files:

• CCA/Components/Schedulers: Relocate.cc (+24 -28), Relocate.h (+4 -3), SchedulerCommon.cc (+2 -2), SchedulerCommon.h (+1 -1)

harman 2012-10-03 14:23 Rev.: 49352

Backed out previous commit which was causing explode2D_amr to hang.

5 lines of code changed in 1 file:

• CCA/Components/Schedulers: Relocate.cc (+5 -7)

harman 2012-10-01 14:36 Rev.: 49325

exchangeParticles() & relocateParticles()
patches->get(0)->getLevel() does not return the coarsest level!
Now using grid->getLevel(0). This fixes a multi-processor, multipatch, disappearing
particles bug

* This should not change the answers on 1L problems.

8 lines of code changed in 1 file:

• CCA/Components/Schedulers: Relocate.cc (+8 -4)

ahumphrey 2012-10-01 10:55 Rev.: 49321

Run with same # threads as decentralized CPU scheduler.

1 lines of code changed in 1 file:

• CCA/Components/Schedulers: UnifiedScheduler.cc (+1 -1)

ahumphrey 2012-09-30 13:49 Rev.: 49313

Add extra info to GPUStats DebugStream for multi-GPU nodes with non-uniform GPU models.
Also added some comments, formatting.

81 lines of code changed in 2 files:

• CCA/Components/Schedulers: UnifiedScheduler.cc (+74 -59), UnifiedScheduler.h (+7 -6)

ahumphrey 2012-09-30 10:24 Rev.: 49312

Fix output for thread DebugStream in Unified Scheduler.

5 lines of code changed in 1 file:

• CCA/Components/Schedulers: UnifiedScheduler.cc (+5 -1)

ahumphrey 2012-09-21 14:42 Rev.: 49256

Allow all schedulers to override the default via SCI_DEBUG env var or via XML in input files.
Also document these new debug vars in environmentalFLags.txt.
Fix two spots where only 1 thread was disallowed when using Unified Scheduler.

51 lines of code changed in 2 files:

• CCA/Components/Schedulers: SchedulerFactory.cc (+49 -22), UnifiedScheduler.cc (+2 -2)

ahumphrey 2012-09-17 08:07 Rev.: 49235

Change the default scheduler to be the Unified Scheduler.
All previous schedulers may still be invoked explicitly via command line.

55 lines of code changed in 2 files:

• CCA/Components/Schedulers: SchedulerFactory.cc (+37 -44), UnifiedScheduler.cc (+18 -16)

ahumphrey 2012-09-11 15:48 Rev.: 49205

Fix a deadlock.
Turns out that: task->getTask()->usesMPI() doesn't seem to ever return true, although it's defined and used throughout the framework.
Need to find out if d_usesMPI is ever set for a task.

2 lines of code changed in 1 file:

• CCA/Components/Schedulers: UnifiedScheduler.cc (+2 -2)

ahumphrey 2012-09-11 15:05 Rev.: 49204

Add SingleProcessor Scheduler functionality to Unified Scheduler.

56 lines of code changed in 2 files:

• CCA/Components/Schedulers: UnifiedScheduler.cc (+52 -8), UnifiedScheduler.h (+4)

ahumphrey 2012-09-11 12:49 Rev.: 49200

Consolidate threaded task execution logic in Unified Scheduler.
* Added some nice comments to explain what's going on here
Update input file for UnifiedSchedulerTest component.

207 lines of code changed in 2 files:

• CCA/Components/Schedulers: UnifiedScheduler.cc (+207 -232), UnifiedScheduler.h (-2)

ahumphrey 2012-09-10 16:53 Rev.: 49197

Some last refinements to the original GPU scheduler. This is necessary to compare results when using the Unified scheduler

13 lines of code changed in 1 file:

• CCA/Components/Schedulers: GPUThreadedMPIScheduler.cc (+13 -8)

ahumphrey 2012-09-10 12:08 Rev.: 49194

Work on memory leak in GPU Scheduler. Also do some formatting / update debug output.

259 lines of code changed in 1 file:

• CCA/Components/Schedulers: GPUThreadedMPIScheduler.cc (+259 -199)

ahumphrey 2012-09-10 10:14 Rev.: 49192

Unified Scheduler:
* Do some memory management on CUDA Stream and Event queues.
* Enhance GPUStats debug stream.

21 lines of code changed in 3 files:

• CCA/Components/Schedulers: DetailedTasks.h (+1 -1), UnifiedScheduler.cc (+18 -14), UnifiedScheduler.h (+2 -2)

ahumphrey 2012-09-07 17:33 Rev.: 49190

Make CUDA stream and event containers thread safe.
Cleanup some dead code.
Further refinement on GPUStats debug stream.

10 lines of code changed in 3 files:

• CCA/Components/Schedulers: DetailedTasks.cc (-9), DetailedTasks.h (-1), UnifiedScheduler.cc (+10 -4)

ahumphrey 2012-09-07 17:12 Rev.: 49189

Fix non-CUDA configuration compiler error.

6 lines of code changed in 2 files:

• CCA/Components/Schedulers: UnifiedScheduler.cc (+5 -6), UnifiedScheduler.h (+1 -1)

ahumphrey 2012-09-07 16:49 Rev.: 49188

Clear GPU-DB maps, etc after each timestep, this fixes D2H copy bug.
Also add more detailed GPU-related debug info.

This is largely working now.

20 lines of code changed in 1 file:

• CCA/Components/Schedulers: UnifiedScheduler.cc (+20 -4)

ahumphrey 2012-09-07 16:09 Rev.: 49187

Unified Scheduler: protect some critical sections, lock some GPU-related data structures.

62 lines of code changed in 2 files:

• CCA/Components/Schedulers: UnifiedScheduler.cc (+60 -32), UnifiedScheduler.h (+2 -3)

ahumphrey 2012-09-07 12:03 Rev.: 49186

Fix GPU logic in Unified Scheduler. Also some other small formatting/style mods.

28 lines of code changed in 1 file:

• CCA/Components/Schedulers: UnifiedScheduler.cc (+28 -17)

ahumphrey 2012-09-06 20:14 Rev.: 49183

Fix Unified scheduler compiler error.

8 lines of code changed in 1 file:

• CCA/Components/Schedulers: UnifiedScheduler.cc (+8)

ahumphrey 2012-09-06 17:53 Rev.: 49182

Unified Scheduler: Add logic to run with or without GPU support via command line switch.

23 lines of code changed in 1 file:

• CCA/Components/Schedulers: UnifiedScheduler.cc (+23 -12)

ahumphrey 2012-09-06 11:24 Rev.: 49176

Add GPU related logic and locks to Unified scheduler.

234 lines of code changed in 5 files:

• CCA/Components/Schedulers: DetailedTasks.cc (+5 -3), DetailedTasks.h (+4 -2), SchedulerFactory.cc (+21 -20), UnifiedScheduler.cc (+187 -12), UnifiedScheduler.h (+17 -10)

tsaad 2012-09-05 13:24 Rev.: 49172

Task checksum reports the wrong number of processors since process rank is counted from 0 not 1. fix that.

1 lines of code changed in 1 file:

• CCA/Components/Schedulers: MPIScheduler.cc (+1 -1)

ahumphrey 2012-09-04 08:42 Rev.: 49166

Some formatting.

944 lines of code changed in 3 files:

• CCA/Components/Schedulers: DetailedTasks.cc (+541 -569), UnifiedScheduler.cc (+398 -366), UnifiedScheduler.h (+5 -1)

ahumphrey 2012-08-31 16:47 Rev.: 49160

Start adding GPU task queue locks in Unified Scheduler.

31 lines of code changed in 2 files:

• CCA/Components/Schedulers: DetailedTasks.cc (+29 -7), DetailedTasks.h (+2)

ahumphrey 2012-08-31 16:10 Rev.: 49159

Fix SchedulerFactory so the Unified scheduler can be created (Corresponding info in ups_spec).
Also fix a debug stream issue in BoundCondReader (change std::cout to BCR_dbg).

5 lines of code changed in 3 files:

• CCA/Components/Schedulers: MPIScheduler.cc (+2 -1), SchedulerFactory.cc (+1 -1), UnifiedScheduler.cc (+2)

jas 2012-08-24 14:27 Rev.: 49128

Use the DW to keep track of the HypreSolver data structures from one time
step to the next to avoid the setup costs that occur for doing very large
core count runs.

For smaller core counts, this does not affect the efficiency of the solver.

Answers should not change.

22 lines of code changed in 2 files:

• CCA/Components/Schedulers: OnDemandDataWarehouse.cc (+21 -2), OnDemandDataWarehouse.h (+1)

qymeng 2012-08-02 09:54 Rev.: 49011

Skip MPI Sends/Recvs for notCopyData Variables.

16 lines of code changed in 5 files:

• CCA/Components/Schedulers: GPUThreadedMPIScheduler.cc (+2 -1), MPIScheduler.cc (+4 -2), ThreadedMPIScheduler.cc (+2 -1), ThreadedMPIScheduler2.cc (+4 -2), UnifiedScheduler.cc (+4 -2)

harman 2012-07-30 14:34 Rev.: 49000

SchedulerCommon & Scheduler:
- added containeers for variables that a component does not want to checkpoint
or copied to a new AMR grid, notCopyData_ & notCheckpointVars. The components
can access these via

scheduler::overrideVariableBehavior()

DataArchiver:
- do not checkpoint any variable in the notCheckpointVars container

Other files:
- updated calls to overrideVariableBehavior()


This is commit is ground work for updating how I/O is handled with
OnTheFly data analysis tasks.

*This commit should not change the answers.

41 lines of code changed in 2 files:

• CCA/Components/Schedulers: SchedulerCommon.cc (+30 -3), SchedulerCommon.h (+11 -1)

ahumphrey 2012-07-06 17:21 Rev.: 48930

Initial implementation of the Unified multi-threaded CPU/GPU scheduler.

TODO:
1.) make GPU related data structures thread-safe
2.) weave in GPU task queue logic

2024 lines of code changed in 4 files:

• CCA/Components/Schedulers: SchedulerFactory.cc (+12 -4), UnifiedScheduler.cc (+1739), UnifiedScheduler.h (+272), sub.mk (+1)

ahumphrey 2012-06-27 17:30 Rev.: 48903

For now, don't use cudaMemcpyDefault when doing CUDA memcopies.
Explicitly use cudaMemcpyHostToDevice and cudaMemcpyDeviceToHost for non-UVA environments.

3 lines of code changed in 1 file:

• CCA/Components/Schedulers: GPUThreadedMPIScheduler.cc (+3 -3)

ahumphrey 2012-06-06 08:53 Rev.: 48840

Re-enable cudaHostUnregister() for host memory that was page-locked with cudaHostRegister().
Also add a little documentation.

11 lines of code changed in 2 files:

• CCA/Components/Schedulers: GPUThreadedMPIScheduler.cc (+1 -1), GPUThreadedMPIScheduler.h (+10 -7)

ahumphrey 2012-06-04 11:40 Rev.: 48832

Enable cudaHostRegister() code in GPU scheduler. This works well with CUDA 4.1. Should be a slight performance boost.
The issue with 4.0 was that the allocated host memory needed to be page aligned to work with cudaHostRegister()
Also eliminated explicit linking of libcuda until we're using driver API. Still need libcudart link.
- New configure script generated.

22 lines of code changed in 1 file:

• CCA/Components/Schedulers: GPUThreadedMPIScheduler.cc (+22 -22)

ahumphrey 2012-05-31 14:58 Rev.: 48827

Removed dependence on CUDA SDK (cutil headers and libs).

CUDA support is now enabled with a single option:
--with-cuda=DIR (CUDA installation is in DIR)

Also generated new configure script.

29 lines of code changed in 2 files:

• CCA/Components/Schedulers: DetailedTasks.cc (+2 -2), GPUThreadedMPIScheduler.cc (+27 -27)

(36 more)