Tools for Reduced Precision Computation

Abstract

The use of reduced precision to improve performance metrics such as computation latency and power consumption is a common practice in the embedded systems field. This practice is emerging as a new trend in High Performance Computing (HPC), especially when new error-tolerant applications are considered. However, standard compiler frameworks do not support automated precision customization, and manual tuning and code transformation is the approach usually adopted in most domains. In recent years, research have been studying ways to improve the automation of this process. This article surveys this body of work, identifying the critical steps of this process, the most advanced tools available, and the open challenges in this research area. We conclude that, while several mature tools exist, there is still a gap to close, especially for tools based on static analysis rather than profiling, as well as for integration within mainstream, industry-strength compiler frameworks.

LORENA: Low memORy symmEtric-Key geNerAtion Method for Based on Group Cryptography Protocol Applied to the Internet of Healthcare Things

“…This section describes the Low memORy symmEtric-key geNerAtion method. The LORENA method follows a light algorithm, which reduces the consumption of hardware resources using approximate computing techniques [9], [10] to define how to save memory resources in wearable devices. The secret key generation employs physiological signs, particularly ECG, as input due to its biometric characteristics, allowing unique identification by the user.…”

“…The key generator module converts ECG signals from a time domain to a frequency domain, quantizing them into a single, highly user-linked secret key. This module applies concepts from the approximate computing technique [9], [10] to provide optimization on memory usage. ECG undergoes a transformation producing unique individual characteristics, which are later quantized to create a secret key.…”

TEXTAROSSA: Towards EXtreme scale Technologies and Accelerators for euROhpc hw/Sw Supercomputing Applications for exascale

“…Proposals have been made at the HW level to take advantage of inherent perceptual limitations, redundant data, or reduced precision input [15], as well as to reduce system costs or improve power efficiency [16]. At the same time, works on floating-point to fixed-point conversion tools [17], [18] allow us to trade-off the algorithm exactness for a more efficient implementation. Finally, new data types are emerging including BFloat16 [19] and Posit [20].…”

Dynamic Precision Autotuning with TAFFO

“…The interest for precision tuning is increasing in the recent years along with the increased popularity of error-tolerant applications, such as image and video processing, data mining, and artificial intelligence. For a comprehensive discussion of approaches to performance/energy optimization, including precision tuning and other approximate computing techniques, the reader is referred to other works [1,6].…”