ZigZag - Deep Learning Hardware Design Space Exploration: /home/runner/work/zigzag/zigzag/zigzag/hardware/architecture/get_cacti_cost.py File Reference

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.

Classes
class	CactiConfig
	Configuration for Cacti. More...

Namespaces
	architecture.get_cacti_cost

Functions
tuple[float, float, float, float]	get_cacti_cost (str cacti_path, float tech_node, str mem_type, float mem_size_in_byte, float bw, str hd_hash="a")
	extract time, area, r_energy, w_energy cost from cacti 7.0 :param cacti_path: the location of cacti :param tech_node: technology node (directly supported node by CACTI: 0.022, 0.032, 0.045, 0.065, 0.09, 0.18) :param mem_type: memory type (sram or dram) :param mem_size_in_byte: memory size (unit: byte) :param bw: memory IO bitwidth :param hd_hash: input file suffix when generating CACTI input file (useful and in avoid of file conflict for multi-processing simulation) Attention: for CACTI, the minimum mem_size=64B, minimum_rows=32 More...

float	get_w_cost_per_weight_from_cacti (str cacti_path, dict[str, float] tech_param, dict[str, Any] hd_param, dict[str, Any] dimensions)