Summary: 62 instances, 55 unique

Text	Count
// TODO: find out which compiler we're using here and use the suppression.	1
// TODO: benchmark what's faster.	1
// in this bin. TODO abstract out this logic into some data structure	2
# TODO: enable material properties to also be saved.	1
// TODO: faster	1
// TODO: make more efficient.	1
// TODO: revisit performance of this kernel with shared memory usage	1
# TODO: replace with tutil.wmean once landed	1
// TODO: is 8 enough? Would increasing have performance considerations?	1
// TODO: Add support for double once we fix atomicAdd	1
// TODO: check if the data is null.	1
// TODO: determine if the value of MAX_TRIS is sufficient or	1
# TODO: Implement a C++ version of this function	1
// TODO: variable reuse?	1
// TODO: check if caching these value helps performance.	1
# TODO (nikhilar) Speed up this function.	1
# TODO: also support the case where RT is specified as one matrix	1
// TODO: benchmark parallelizing over faces_verts instead of over pixels.	1
// TODO: make sorting an option as only top k is needed, not sorted values.	2
// TODO: switch to Eigen if more functionality is needed.	2
# TODO better heuristics for bin size.	1
# TODO: inverse is bad! Solve a linear system instead	1
TODO: Add a different sampling mode which interpolates the barycentric	1
# TODO: Revisit whether or not to transform z coordinate to [-1, 1] or	1
make_float3( // TODO: make offset consistent with `get_screen_area`.	1
// TODO: change to kEpsilon to match triangles?	1
# TODO (nikhilar) remove following 4 lines when torch.unique has support	1
# TODO: should the texture map values be multiplied	1
// TODO: check for synchronization.	1
# TODO: Remove this line when the convention for the z coordinate in	1
# TODO: allow customization of this base color?	1
# TODO: Choose naive vs coarse-to-fine based on mesh size and image size.	1
// TODO: using shared memory for min_point_dist gives an ~2x speed up	1
# TODO: Can we do this all in one go somehow?	1
# TODO (nikhilar) Find a faster way of computing the weights for each edge	1
# This doesn't pass unit tests. TODO investigate further	1
# TODO: more stable when initialised with the center and eigenvectors!	1
# TODO make the epsilon user configurable	1
// TODO: fix index calculation for non square images.	1
# TODO (gkioxari) Confirm multinomial bug is not present with real data.	1
TODO: Add support for textures.	2
// TODO: rewrite using nested contexts instead of temporary names.	1
# TODO: attenuate based on inverse squared distance to the light source	1
# TODO: support providing K as a ratio of the total number of points instead of as an int	2
// TODO: Create wrapper function and merge kernel with other array kernel	2
# TODO: handle multiple directional lights per batch element.	1
# TODO: Add better error handling for when there are more than	1
# TODO: consider a better way of setting the bin size.	2
// TODO: optimize.	1
# TODO: there may still be some instability in the exponent calculation.	1
# TODO: handle multiple directional lights	1
// TODO: Better type checking.	1
// TODO: remove once https://github.com/NVlabs/cub/issues/172 is resolved.	1
# TODO: handle attenuation.	1
// TODO: put intrinsics here.	1