| Directory | Changes | Lines of Code | Lines per Change |
|---|---|---|---|
| Totals | 161 (100.0%) | 1943 (100.0%) | 12.0 |
| CCA/Components/Schedulers/ | 73 (45.3%) | 914 (47.0%) | 12.5 |
| Core/Grid/ | 10 (6.2%) | 232 (11.9%) | 23.2 |
| CCA/Components/Examples/ | 12 (7.5%) | 170 (8.7%) | 14.1 |
| Core/Geometry/ | 3 (1.9%) | 142 (7.3%) | 47.3 |
| Core/Grid/Variables/ | 5 (3.1%) | 139 (7.2%) | 27.8 |
| CCA/Components/SimulationController/ | 12 (7.5%) | 132 (6.8%) | 11.0 |
| CCA/Components/Models/Radiation/RMCRT/ | 13 (8.1%) | 46 (2.4%) | 3.5 |
| CCA/Components/DataArchiver/ | 6 (3.7%) | 37 (1.9%) | 6.1 |
| CCA/Ports/ | 5 (3.1%) | 35 (1.8%) | 7.0 |
| include/sci_defs/ | 3 (1.9%) | 33 (1.7%) | 11.0 |
| / | 6 (3.7%) | 33 (1.7%) | 5.5 |
| Core/Parallel/ | 7 (4.3%) | 19 (1.0%) | 2.7 |
| include/ | 2 (1.2%) | 11 (0.6%) | 5.5 |
| StandAlone/ | 2 (1.2%) | 0 (0.0%) | 0.0 |
| CCA/Components/Parent/ | 2 (1.2%) | 0 (0.0%) | 0.0 |
schedule ray CPU task if -gpu argument not present, instead of depends on HAVE_CUDA macro.
17 lines of code changed in 5 files:
Add support for post GPU kernel processing callback.
A GPU task will be called multipule times and each time an event argument will be passed.
The task function can then use the event to decide what to run when
1) Task::CPU task is scheduled to run on CPU
2) Task::GPU task is scheduled to run on GPU
3) Task::postGPU task is finished execution on GPU, DW has been copied back.
CPU only task interface is not unchanged.
282 lines of code changed in 11 files:
fix build on blue waters due to new mpi libary changes.
0 lines of code changed in 2 files:
fix a bug in find_sendset from last commit.
0 lines of code changed in 2 files:
fix a particle sending state thread safety issue.
16 lines of code changed in 3 files:
default constructor for GPU vector
8 lines of code changed in 1 file:
Use derived class instead of union to support dimensional index.
This has been verfied to produce the same assumbly code as the orignal struct when using dimensional index.
40 lines of code changed in 1 file:
add new union type Int3 uInt3 Double3 Float3, so that it can be accessed by dimension index.
95 lines of code changed in 2 files:
add const GPU Grid Variable support
70 lines of code changed in 8 files:
use d_allmpi to check if running under MPI
0 lines of code changed in 2 files:
Do not call MPI_Comm_dup if not running under MPI/threads.
1 lines of code changed in 1 file:
Allocate MPI communicators based on taskgraph.
This fixed a hung when there are large number of reduction tasks in the taskgraph.
18 lines of code changed in 4 files:
add new operator to GPUGridVariable.
GPUGridVaraible can be accessed as var(x,y,z)
28 lines of code changed in 3 files:
Added array boundary checking for GPU varaible.
Fixed multi-GPU issue.
82 lines of code changed in 6 files:
detect and report GPU task kernel error
56 lines of code changed in 5 files:
better multi-GPU debug output
19 lines of code changed in 3 files:
Implemented a lightweight version of GPUDataWarehouse, GPUGridVariable and GPUArray3D inside GPU.
This commit included a new set of interfaces for better supporting GPU task.
See example in Examples/UnifiedSchedulerTest
691 lines of code changed in 27 files:
support C++11 style hash map
42 lines of code changed in 5 files:
skip first timestep for checking output/checkpoint interval changes.
4 lines of code changed in 1 file:
For debug stream sendtiming recvtiming taskorder, only output on where rank id is half of the world size.
47 lines of code changed in 9 files:
(25 more)