Mostra el registre d'ítem simple
An energy-efficient and scalable eDRAM-based register file architecture for GPGPU
dc.contributor.author | Jing, Naifeng |
dc.contributor.author | Shen, Yao |
dc.contributor.author | Lu, Yao |
dc.contributor.author | Ganapathy, Shrikanth |
dc.contributor.author | Mao, Zhigang |
dc.contributor.author | Guo, Minyi |
dc.contributor.author | Canal Corretger, Ramon |
dc.contributor.author | Liang, Xiaoyao |
dc.contributor.other | Universitat Politècnica de Catalunya. Departament d'Arquitectura de Computadors |
dc.date.accessioned | 2013-10-01T13:21:20Z |
dc.date.created | 2013 |
dc.date.issued | 2013 |
dc.identifier.citation | Jing, N. [et al.]. An energy-efficient and scalable eDRAM-based register file architecture for GPGPU. A: Annual International Symposium on Computer Architecture. "ISCA 2013: the 40th Annual International Symposium on Computer Architecture: conference proceedings: June 23-27, 2013: Tel-Aviv, Israel". Tel-Aviv: ACM, 2013, p. 344-355. |
dc.identifier.isbn | 978-1-4503-2079-5 |
dc.identifier.uri | http://hdl.handle.net/2117/20247 |
dc.description.abstract | The heavily-threaded data processing demands of streaming multiprocessors (SM) in a GPGPU require a large register file (RF). The fast increasing size of the RF makes the area cost and power consumption unaffordable for traditional SRAM designs in the future technologies. In this paper, we propose to use embedded-DRAM (eDRAM) as an alternative in future GPGPUs. Compared with SRAM, eDRAM provides higher density and lower leakage power. However, the limited data retention time in eDRAM poses new challenges. Periodic refresh operations are needed to maintain data integrity. This is exacerbated with the scaling of eDRAM density, process variations and temperature. Unlike conventional CPUs which make use of multi-ported RF, most of the RFs in modern GPGPU are heavily banked but not multi-ported to reduce the hardware cost. This provides a unique opportunity to hide the refresh overhead. We propose two different eDRAM implementations based on 3T1D and 1T1C memory cells. To mitigate the impact of periodic refresh, we propose two novel refresh solutions using bank bubble and bank walk-through. Plus, for the 1T1C RF, we design an interleaved bank organization together with an intelligent warp scheduling strategy to reduce the impact of the destructive reads. The analysis shows that our schemes present better energy efficiency, scalability and variation tolerance than traditional SRAM-based designs |
dc.format.extent | 12 p. |
dc.language.iso | eng |
dc.publisher | ACM |
dc.rights | Attribution-NonCommercial-NoDerivs 3.0 Spain |
dc.rights.uri | http://creativecommons.org/licenses/by-nc-nd/3.0/es/ |
dc.subject | Àrees temàtiques de la UPC::Informàtica::Arquitectura de computadors |
dc.subject.lcsh | Computer architecture |
dc.subject.other | Program processors |
dc.subject.other | Computer architecture |
dc.subject.other | Data processing |
dc.subject.other | Energy efficiency |
dc.subject.other | Logic design |
dc.title | An energy-efficient and scalable eDRAM-based register file architecture for GPGPU |
dc.type | Conference report |
dc.subject.lemac | Arquitectura d'ordinadors |
dc.contributor.group | Universitat Politècnica de Catalunya. ARCO - Microarquitectura i Compiladors |
dc.identifier.doi | 10.1145/2485922.2485952 |
dc.description.peerreviewed | Peer Reviewed |
dc.relation.publisherversion | http://dl.acm.org/citation.cfm?doid=2485922.2485952 |
dc.rights.access | Restricted access - publisher's policy |
local.identifier.drac | 12772134 |
dc.description.version | Postprint (published version) |
dc.date.lift | 10000-01-01 |
local.citation.author | Jing, N.; Shen, Y.; Lu, Y.; Ganapathy, S.; Mao, Z.; Guo, M.; Canal, R.; Liang, X. |
local.citation.contributor | Annual International Symposium on Computer Architecture |
local.citation.pubplace | Tel-Aviv |
local.citation.publicationName | ISCA 2013: the 40th Annual International Symposium on Computer Architecture: conference proceedings: June 23-27, 2013: Tel-Aviv, Israel |
local.citation.startingPage | 344 |
local.citation.endingPage | 355 |