tvm_tirx.support.nvcc#

Utility to invoke nvcc compiler in the system

Functions#

`compile_cuda`(code, *[, backend_config, ...])	Compile source using the nested backend_config mapping and return its bytes.
`find_cuda_path`()	Utility function to find CUDA path
`get_cuda_version`([cuda_path])	Utility function to get CUDA version
`find_nvshmem_paths`()	Searches for the NVSHMEM include and library directories.
`tvm_callback_cuda_compile`	Callable wrapper around a TVM FFI function.
`find_libdevice_path`	Callable wrapper around a TVM FFI function.
`callback_libdevice_path`(arch)
`get_target_compute_version`	Callable wrapper around a TVM FFI function.
`parse_compute_version`(compute_version)	Parse compute capability string to divide major and minor version
`have_fp16`(compute_version)	Either fp16 support is provided in the compute capability or not
`have_int8`(compute_version)	Either int8 support is provided in the compute capability or not
`have_tensorcore`([compute_version, target])	Either TensorCore support is provided in the compute capability or not
`have_cudagraph`()	Either CUDA Graph support is provided
`have_bf16`	Callable wrapper around a TVM FFI function.
`have_fp8`	Callable wrapper around a TVM FFI function.
`have_fp4`	Callable wrapper around a TVM FFI function.