Abstract
Low-precision numeric types offer significant efficiency and energy benefits for computing applications. Mixed-precision algorithms, combining low and high precision types, maintain accuracy while improving performance. Despite advantages, there exist challenges on adapting existing mixed-precision algorithms to new technologies, such as new hardware architectures and new low-precision data types. This paper presents current challenges and opportunities to advance science in this domain targetting more energy efficient solutions.
| Original language | American English |
|---|---|
| Number of pages | 5 |
| State | Published - 2024 |
| Event | ASCR Workshop on Energy-Efficient Computing for Science - Bethesda, MD Duration: 9 Sep 2024 → 12 Sep 2024 |
Conference
| Conference | ASCR Workshop on Energy-Efficient Computing for Science |
|---|---|
| City | Bethesda, MD |
| Period | 9/09/24 → 12/09/24 |
NLR Publication Number
- NREL/CP-2C00-90661
Keywords
- energy efficiency
- low-precision
- mixed-precision