symforce.codegen.backends.cuda package

*THIS MODULE IS EXPERIMENTAL*

Backend for CUDA. This generates CUDA __host__ __device__ functions that are intended to be called from kernels - e.g. to use one of these, you should write a kernel with your memory, threads, etc laid out how you’d like, and call the generated function from one thread / for one element of your input or output.

This currently only supports vector inputs and outputs, we do not have geo or cam types for CUDA yet.

Submodules

• symforce.codegen.backends.cuda.cuda_code_printer module
  • ScalarType
    • FLOAT
    • DOUBLE
  • CudaCodePrinter
• symforce.codegen.backends.cuda.cuda_config module
  • CudaConfig
    • doc_comment_line_prefix
    • line_length
    • use_eigen_types
    • override_methods
    • extra_imports
    • scalar_type
    • inline
    • builtin_vector_variables
    • backend_name()
    • template_dir()
    • templates_to_render()
    • printer()
    • format_matrix_accessor()
    • format_eigen_lcm_accessor()