cufile

The cuda.bindings.cufile Python module wraps the cuFile C APIs. Supported on Linux only.

Currently using this module requires NumPy to be present. Any recent NumPy 1.x or 2.x should work.

Functions

handle_register(intptr_t descr)	cuFileHandleRegister is required, and performs extra checking that is memoized to provide increased performance on later cuFile operations.
handle_deregister(intptr_t fh)	releases a registered filehandle from cuFile.
buf_register(intptr_t buf_ptr_base, ...)	register an existing cudaMalloced memory with cuFile to pin for GPUDirect Storage access or register host allocated memory with cuFile.
buf_deregister(intptr_t buf_ptr_base)	deregister an already registered device or host memory from cuFile.
read(intptr_t fh, intptr_t buf_ptr_base, ...)	read data from a registered file handle to a specified device or host memory.
write(intptr_t fh, intptr_t buf_ptr_base, ...)	write data from a specified device or host memory to a registered file handle.
driver_open()	Initialize the cuFile library and open the nvidia-fs driver.
use_count()	returns use count of cufile drivers at that moment by the process.
driver_get_properties(intptr_t props)	Gets the Driver session properties.
driver_set_poll_mode(bool poll, ...)	Sets whether the Read/Write APIs use polling to do IO operations.
driver_set_max_direct_io_size(...)	Control parameter to set max IO size(KB) used by the library to talk to nvidia-fs driver.
driver_set_max_cache_size(size_t max_cache_size)	Control parameter to set maximum GPU memory reserved per device by the library for internal buffering.
driver_set_max_pinned_mem_size(...)	Sets maximum buffer space that is pinned in KB for use by cuFileBufRegister.
batch_io_set_up(unsigned int nr)
batch_io_submit(intptr_t batch_idp, ...)
batch_io_get_status(intptr_t batch_idp, ...)
batch_io_cancel(intptr_t batch_idp)
batch_io_destroy(intptr_t batch_idp)
read_async(intptr_t fh, ...)
write_async(intptr_t fh, ...)
stream_register(intptr_t stream, ...)
stream_deregister(intptr_t stream)
get_version()
get_parameter_size_t(int param)
get_parameter_bool(int param)
get_parameter_string(int param, int len)
set_parameter_size_t(int param, size_t value)
set_parameter_bool(int param, bool value)
set_parameter_string(int param, ...)
op_status_error(int status)	cufileop status string.
driver_close()	reset the cuFile library and release the nvidia-fs driver

Types

IOEvents([size])	Empty-initialize an array of CUfileIOEvents_t.
Descr([size])	Empty-initialize an array of CUfileDescr_t.
IOParams([size])	Empty-initialize an array of CUfileIOParams_t.
OpError(value[, names, module, qualname, ...])	See CUfileOpError.
DriverStatusFlags(value[, names, module, ...])	See CUfileDriverStatusFlags_t.
DriverControlFlags(value[, names, module, ...])	See CUfileDriverControlFlags_t.
FeatureFlags(value[, names, module, ...])	See CUfileFeatureFlags_t.
FileHandleType(value[, names, module, ...])	See CUfileFileHandleType.
Opcode(value[, names, module, qualname, ...])	See CUfileOpcode_t.
Status(value[, names, module, qualname, ...])	See CUfileStatus_t.
BatchMode(value[, names, module, qualname, ...])	See CUfileBatchMode_t.
SizeTConfigParameter(value[, names, module, ...])	See CUFileSizeTConfigParameter_t.
BoolConfigParameter(value[, names, module, ...])	See CUFileBoolConfigParameter_t.
StringConfigParameter(value[, names, ...])	See CUFileStringConfigParameter_t.
cuFileError(status[, cu_err])