cctw2 Activity
Brought to you by:
guyjennings
Activity for cctw2
-
Guy Jennings committed [e9d9c1] on Code
Parameters in hdf5 files largely working
-
Working on reading/writing parameters in hdf/nexus files
-
Tweaks to cuda build
-
Merging
-
Mods to allow building on polaris
-
More work on cctw2 file reading
-
Found bugs in cuda transform - rationalized to single transform cuda kernel
-
Work on cuda transform. Added lzf, lz4 and bz2 filters
-
Work on cctw2-threaded --transform
-
Working on cctw2 cuda transform
-
Starting testing cctw2-cuda --transform
-
Partial implementation of GPU transform
-
Working towards cctw2 transform alg0
-
Work on cct2wcudatransformalgorithm
-
Outline code for 2 different CUDA transform algorithms, start implementing algorithm 0
-
Correctly calculate number of output chunks that will fit in GPU
-
Calculate number of output chunks that will fit in GPU
-
Add some TODO reminders
-
Work on angle parameter parsing
-
Implement weights and masks
-
Implement basic skeleton for weights and mask processing
-
Implement cctw2-cuda input chunk reader
-
Work on enumeration of input block coords
-
Check transform input parameters
-
Check transform input parameters
-
Fill out basic skeleton of cctw2-cuda tranform command, complete with explanations of needed code
-
More work on Cctw2CudaTransform
-
Reduce c++ language level required, work on checking input dataset dimensions
-
Remove dependency on boost::trim_right
-
More cctw2 file checking
-
More cctw2 argument processing
-
Start work on cctw2-cuda transform command
-
Add cctw-util size argument
-
de-templatize parameters, command line parser and command file parser
-
Skeleton versions of transform, merge, range, subrange commands for cuda, single and threaded cctw2
-
Start work on transform commands
-
Working on file name decoding and validation
-
Rename of cctw2cuda files, removing dashes from file names
-
cctw2 files renamed, removing hyphens in file names
-
Work on input file name parsing
-
Work on input file name parsing
-
Comment out mpi-specific hdf5 calls
-
Working on cctw2 --range command
-
Working on Cctw2ChunkedArray3DDevice
-
Tweak to output data summarize command
-
Implement Cctw2ChunkedArray3DHost
-
Rework MPI/HDF5 interactions, more work on tests subdir
-
Update hostname test for weimar
-
Removed cctw2-tests-code submodule
-
Removed cctw2-tests submodule
-
Properly test for cctw2-tests-code/CMakeLists.txt
-
Added cctw2-tests-code subdirectory
-
Added cctw2-tests submodule
-
Added cctw2-tests submodule
-
Merged
-
Start work on cctw2 3d data array types, work on build files
-
Tweaks for cuda include paths
-
Working on subrange command - basic skeleton
-
Fix up cctw-mpi build
-
Include build of previous 'cctw' versions
-
More tweaks for cuda build
-
More tweaks to cuda detection
-
Added some config scripts - should be called from the cctw2-code directory
-
Fixes to compile on Polaris
-
Implement help output
-
Templatize much of the code so that float and double performance may be compared
-
Reorganize build system to build cctw2-libraries only once
-
Basic implementation of range command for single, threaded and cuda
-
Added skeleton code for merge, range and transform commands
-
Implement cctw2 --gentest command
-
Added --transform, --merge, --genpars, --gentest, --genmerge options
-
Try to make CUDA presence conditional
-
Added test parameters file, trim input line trailing white space
-
Fixed problems with cuda atomic min/max, implement parameter file reading
-
Naive implementation to find transformed range using atomic globals
-
Implement thread-safe Cctw2Transform::forward
-
Defined Cctw2Parameters and demonstrate passing it to cuda kernel
-
More work on cctw2 host and device library functions
-
More work on cctw2 host and device library functions
-
More work on cctw2 command argument decoding
-
More work on cctw2 command argument decoding
-
Implement running subtask for different types of cctw2-<xxx>
-
Added code in cctw2 to test for cuda
-
Initial Commit
1
Gen AI apps are built with MongoDB Atlas
Atlas offers built-in vector search and global availability across 125+ regions. Start building AI apps faster, all in one place.