Attendees: Ken Moreland (SNL), Abhishek Yenpure (UO), Rob Maynard (Kitware), Utkarsh Ayachit (Kitware), Allie Vacanti (Kitware), Dave Pugmire (ORNL)
Last week Rob was scrambling to get vtkm::cont::ColorTable and associated filters into VTK-m for quarterly milestones.
The default color map is now set to Viridis in vtkm::cont::ColorTable. This means that when you create a rendering scene, these are the default colors that will be used. No one seems to have any heartache about these colors.
Utkarsh has changed the filter API to completely remove vtkm::filter::Result. Instead, you return the output data set directly (or throw an exception if an error occurs). This was to simplify filter code where a filter takes a multiblock and returns a dataset or vice versa.
Utkarsh is working on a distributed histogram filter (using DIY). Initial implementation working. Working through some issues with CUDA and windows. Right now DIY is only being used for the reduction.
Rob added a change to ArrayHandle that allowed you to get a free function that you could use when you steal memory from it.
ArrayHandle now better supports CUDA universal virtual memory.
There is a new proposed constructor for the basic storage that takes a free function so that the ArrayHandle can own the data.
Revisiting the CMake infrastructure for VTK-m.
There are outstanding issues with ArrayHandleVirtual. The issue seems to be ArrayHandleCompositeVec.
Dave P. requests that VTK-m write out images in a more widely accepted format. Currently, Canvas can write out a .ppm (simply because .ppm is dirt simple to write). However, not a whole lot of readers understand this format. It would be nice if we could easily write out a more widely accepted image format. It was noted that there exist some very small implementations of png encode/decode that we could embed in the VTK-m source. In fact, there already exists a version of the decoder in the rendering library to decode font bitmaps. As it is acceptable to everyone, Dave will look into implementing this.