ZigZag - Deep Learning Hardware Design Space Exploration
This repository presents the novel version of our tried-and-tested hardware Architecture-Mapping Design Space Exploration (DSE) Framework for Deep Learning (DL) accelerators. ZigZag bridges the gap between algorithmic DL decisions and their acceleration cost on specialized accelerators through a fast and accurate hardware cost estimation.
|
Represents the memory access energy in four directions. More...
Public Member Functions | |
"AccessEnergy" | __add__ (self, "FourWayDataMoving[float]" other) |
Element-wise addition of two AccessEnergy instances. More... | |
"AccessEnergy" | __mul__ (self, float scalar) |
Element-wise multiplication by a scalar. More... | |
![]() | |
def | __init__ (self, dict[DataDirection, T]|None data=None) |
Initialize with a dictionary containing all four DataDirection values, defaulting to zero. More... | |
T | get (self, DataDirection direction) |
Retrieve the value associated with a specific data direction. More... | |
def | set (self, DataDirection direction, T value) |
Update the value of a specific data direction. More... | |
"FourWayDataMoving[T]" | __add__ (self, "FourWayDataMoving[T]" other) |
Element-wise addition of two FourWayDataMoving instances. More... | |
"FourWayDataMoving[T]" | __mul__ (self, T scalar) |
Element-wise multiplication by a scalar. More... | |
def | __repr__ (self) |
Readable string representation of the class. More... | |
def | __jsonrepr__ (self) |
JSON-friendly representation. More... | |
Represents the memory access energy in four directions.
"AccessEnergy" __add__ | ( | self, | |
"FourWayDataMoving[float]" | other | ||
) |
Element-wise addition of two AccessEnergy instances.
"AccessEnergy" __mul__ | ( | self, | |
float | scalar | ||
) |
Element-wise multiplication by a scalar.