Presentation | 2014-09-19 GRAPE9-MPX: A development of an accelerator dedicated for arbitrary-precision arithmetic by the FPGA boards Shinji MOTOKI, Hiroshi DAISAKA, Naohito NAKASATO, Tadashi ISHIKAWA, Fukuko YUASA, Toshiyuki FUKUSHIGE, Atsushi KAWAI, Jun'ichiro MAKINO, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | Higher order corrections in perturbative quantum field theory are required for precise theoretical analysis to investigate the new physics. Feynman loop diagrams which appear in the calculation of the corrections are taken into account, leading to the evaluation of loop integrals. We developed GRAPE9-MPX system consisting of multiple FPGA boards to accelerate parallel computation of loop integrals. In the system, we implemented Processor Elements (PE) to realize quadruple/hexuple-precision arithmetics in FPGA. The theoretical perk speed in the current implementation is about 6.6 Gflops (quadruple-precision) and 3.2 Gflops (hexuple-precision). With the compiler we are developing, the program code is generated automatically with only inserting directives in C/C++ code. We present performance results for the case of Feynman two-loop integrals with quadruple precision and achieve the effective performance of 2.4Gflops for 1 board, 4.7Gflops for 2 boards, and 9.1Gflops for 4 boards, respectively. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | multiple-precision arithmetic / quadruple precision / FPGA / Feynman amplitude |
Paper # | RECONF2014-29 |
Date of Issue |
Conference Information | |
Committee | RECONF |
---|---|
Conference Date | 2014/9/11(1days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | |
Chair | |
Vice Chair | |
Secretary | |
Assistant |
Paper Information | |
Registration To | Reconfigurable Systems (RECONF) |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | GRAPE9-MPX: A development of an accelerator dedicated for arbitrary-precision arithmetic by the FPGA boards |
Sub Title (in English) | |
Keyword(1) | multiple-precision arithmetic |
Keyword(2) | quadruple precision |
Keyword(3) | FPGA |
Keyword(4) | Feynman amplitude |
1st Author's Name | Shinji MOTOKI |
1st Author's Affiliation | High Energy Accelerator Research Organization() |
2nd Author's Name | Hiroshi DAISAKA |
2nd Author's Affiliation | Hitotsubashi Univerisity |
3rd Author's Name | Naohito NAKASATO |
3rd Author's Affiliation | Univerisity of Aizu |
4th Author's Name | Tadashi ISHIKAWA |
4th Author's Affiliation | High Energy Accelerator Research Organization |
5th Author's Name | Fukuko YUASA |
5th Author's Affiliation | High Energy Accelerator Research Organization |
6th Author's Name | Toshiyuki FUKUSHIGE |
6th Author's Affiliation | K&F Computing Research Co. |
7th Author's Name | Atsushi KAWAI |
7th Author's Affiliation | K&F Computing Research Co. |
8th Author's Name | Jun'ichiro MAKINO |
8th Author's Affiliation | RIKEN Advanced Institute for Computational Science |
Date | 2014-09-19 |
Paper # | RECONF2014-29 |
Volume (vol) | vol.114 |
Number (no) | 223 |
Page | pp.pp.- |
#Pages | 6 |
Date of Issue |