Zobrazeno 1 - 10
of 35
pro vyhledávání: '"Angshuman Parashar"'
Autor:
Prasanth Chatarasi, Hyoukjun Kwon, Angshuman Parashar, Michael Pellauer, Tushar Krishna, Vivek Sarkar
Publikováno v:
ACM Transactions on Architecture and Code Optimization. 19:1-26
A spatial accelerator’s efficiency depends heavily on both its mapper and cost models to generate optimized mappings for various operators of DNN models. However, existing cost models lack a formal boundary over their input programs (operators) for
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_________::f7da9de5c3d61296d38a137951f4e6bc
https://doi.org/10.1145/3485137
https://doi.org/10.1145/3485137
Map Space Exploration is the problem of finding optimized mappings of a Deep Neural Network (DNN) model on an accelerator. It is known to be extremely computationally expensive, and there has been active research looking at both heuristics and learni
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_dedup___::50af22efb3a3e84ba2a8c4cfb814fc21
http://arxiv.org/abs/2210.03731
http://arxiv.org/abs/2210.03731
Publikováno v:
Abstract Proceedings of the 2022 ACM SIGMETRICS/IFIP PERFORMANCE Joint International Conference on Measurement and Modeling of Computer Systems.
The high efficiency of domain-specific hardware accelerators for machine learning (ML) has come fromspecialization, with the trade-off of less configurability/ flexibility. There is growing interest in developingflexible ML accelerators to make them
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_dedup___::8be6e54e710a0541979a1064591e6e72
https://doi.org/10.1145/3489048.3530974
https://doi.org/10.1145/3489048.3530974
Publikováno v:
2022 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS).
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_________::89354c0eebf2c789f1ff457a7694b6eb
https://doi.org/10.1109/ispass55109.2022.00039
https://doi.org/10.1109/ispass55109.2022.00039
Publikováno v:
IEEE Computer Architecture Letters. 20:1-4
Dataflow and tile size choices, which we collectively refer to as mappings, dictate the efficiency (i.e., latency and energy) of DNN accelerators. Rapidly evolving DNN models is one of the major challenges for DNN accelerators since the optimal mappi
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_________::fba7d8f523a540509c6484aa090a5d94
https://doi.org/10.1109/lca.2020.3044607
https://doi.org/10.1109/lca.2020.3044607
Publikováno v:
Synthesis Lectures on Computer Architecture. 15:1-164
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_________::2baa915894621b0bedd1bfa335695719
https://doi.org/10.2200/s01015ed1v01y202005cac052
https://doi.org/10.2200/s01015ed1v01y202005cac052
MAESTRO: A Data-Centric Approach to Understand Reuse, Performance, and Hardware Cost of DNN Mappings
Autor:
Michael Pellauer, Prasanth Chatarasi, Angshuman Parashar, Tushar Krishna, Vivek Sarkar, Hyoukjun Kwon
Publikováno v:
IEEE Micro. 40:20-29
The efficiency of an accelerator depends on three factors-mapping, deep neural network (DNN) layers, and hardware-constructing extremely complicated design space of DNN accelerators. To demystify such complicated design space and guide the DNN accele
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_________::64b70bbda8bd1e72eb67847fc0ecdb61
https://doi.org/10.1109/mm.2020.2985963
https://doi.org/10.1109/mm.2020.2985963
Autor:
Geonhwa Jeong, Tushar Krishna, Gokcen Kestor, Prasanth Chatarasi, Sivasankaran Rajamanickam, Roberto Gioiosa, Angshuman Parashar, Po-An Tsai
Publikováno v:
PACT
To meet the extreme compute demands for deep learning across commercial and scientific applications, dataflow accelerators are becoming increasingly popular. While these "domain-specific" accelerators are not fully programmable like CPUs and GPUs, th
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_dedup___::037907a57702d2dff1170b9d17127045
https://doi.org/10.2172/1890906
https://doi.org/10.2172/1890906
Autor:
Christopher W. Fletcher, Po-An Tsai, Sitao Huang, Vikas Chandra, Kartik Hegde, Angshuman Parashar
Publikováno v:
ASPLOS
Modern day computing increasingly relies on specialization to satiate growing performance and efficiency requirements. A core challenge in designing such specialized hardware architectures is how to perform mapping space search, i.e., search for an o
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_dedup___::ad14422e350832c22ca9f6fc0b6fc830
https://doi.org/10.1145/3445814.3446762
https://doi.org/10.1145/3445814.3446762
Publikováno v:
ISPASS
This paper presents Sparseloop, the first infrastructure that implements an analytical design space exploration methodology for sparse tensor accelerators. Sparseloop comprehends a wide set of architecture specifications including various sparse opti
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_________::9f893dd2f6f7dc7e3463f3ae84f61fca
https://doi.org/10.1109/ispass51385.2021.00043
https://doi.org/10.1109/ispass51385.2021.00043