polygraphy.util.cuda.DeviceBuffer Class Reference

Inheritance diagram for polygraphy.util.cuda.DeviceBuffer:

Collaboration diagram for polygraphy.util.cuda.DeviceBuffer:

Public Member Functions
def	__init__ (self, shape=None, dtype=None)
def	address (self)
def	allocate (self, nbytes)
def	free (self)
def	resize (self, shape)
def	copy_from (self, host_buffer, stream=None)
def	copy_to (self, host_buffer, stream=None)
def	__str__ (self)

Public Attributes
	shape
	dtype
	allocated_nbytes

Private Member Functions
def	_check_dtype_matches (self, host_buffer)

Private Attributes
	_ptr

Constructor & Destructor Documentation

__init__()

def polygraphy.util.cuda.DeviceBuffer.__init__	(		self,
			shape = None,
			dtype = None
	)

Represents a buffer on the GPU.

Args:
    shape (Tuple[int]): The initial shape of the buffer.
    dtype (numpy.dtype): The data type of the buffer.

Member Function Documentation

address()

def polygraphy.util.cuda.DeviceBuffer.address

(

self

)

allocate()

def polygraphy.util.cuda.DeviceBuffer.allocate	(		self,
			nbytes
	)

Here is the call graph for this function:

Here is the caller graph for this function:

free()

def polygraphy.util.cuda.DeviceBuffer.free

(

self

)

Here is the call graph for this function:

Here is the caller graph for this function:

resize()

def polygraphy.util.cuda.DeviceBuffer.resize	(		self,
			shape
	)

Here is the call graph for this function:

Here is the caller graph for this function:

_check_dtype_matches()

def polygraphy.util.cuda.DeviceBuffer._check_dtype_matches (   self,   host_buffer  )

private

Here is the caller graph for this function:

copy_from()

def polygraphy.util.cuda.DeviceBuffer.copy_from	(		self,
			host_buffer,
			stream = None
	)

Here is the call graph for this function:

copy_to()

def polygraphy.util.cuda.DeviceBuffer.copy_to	(		self,
			host_buffer,
			stream = None
	)

Copies from this device buffer to the provided host buffer.
Host buffer must be contiguous in memory (see np.ascontiguousarray).

Args:
    host_buffer (numpy.ndarray): The host buffer to copy into.
    stream (Stream):
    A Stream instance (see util/cuda.py). Performs a synchronous copy if no stream is provided.

Returns:
    numpy.ndarray: The host buffer, possibly reallocated if the provided buffer was too small.

Here is the call graph for this function:

__str__()

def polygraphy.util.cuda.DeviceBuffer.__str__

(

self

)

Member Data Documentation

shape

polygraphy.util.cuda.DeviceBuffer.shape

dtype

polygraphy.util.cuda.DeviceBuffer.dtype

allocated_nbytes

polygraphy.util.cuda.DeviceBuffer.allocated_nbytes

_ptr

polygraphy.util.cuda.DeviceBuffer._ptr

private

---

The documentation for this class was generated from the following file:

• cuda.py