Changeset 237 for anr/section-3.1.tex
- Timestamp:
- Feb 16, 2010, 5:24:12 PM (14 years ago)
- File:
-
- 1 edited
Legend:
- Unmodified
- Added
- Removed
-
anr/section-3.1.tex
r235 r237 8 8 \subsubsection{High Performance Computing} 9 9 % Un marché bouffé par les archi GPGPU tel que le FERMI de NvidiaCUDA programming language 10 High-Performance Computing (HPC) world is composed of three main families of architectures:10 The High-Performance Computing (HPC) world is composed of three main families of architectures: 11 11 many-core, GPGPU (General Purpose computation on Graphics Unit Processing) and FPGA. 12 The two firstfamilies are dominating the market by taking benefit12 The first two families are dominating the market by taking benefit 13 13 of the strength and influence of mass-market leaders (Intel, Nvidia). 14 14 %such as Intel for many-core CPU and Nvidia for GPGPU. … … 51 51 \subsubsection{System Synthesis} 52 52 Today, several solutions for system design are proposed and commercialized. 53 The existing commercial or free tools do esnot53 The existing commercial or free tools do not 54 54 cover the whole system synthesis process in a full automatic way. Moreover, 55 55 they are bound to a particular device family and to IPs library. … … 84 84 set of constraints (area, power, frequency, ...) to a micro-architecture at 85 85 Register Transfer Level (RTL). 86 Several academic and commercial tools are today available. Most common86 Several academic and commercial tools are today available. The most common 87 87 tools are SPARK~\cite{spark04}, GAUT~\cite{gaut08}, UGH~\cite{ugh08} in the 88 88 academic world and CATAPULTC~\cite{catapult-c}, PICO~\cite{pico} and 89 CYNTHETIZER~\cite{cynthetizer} in commercial world. Despite their89 CYNTHETIZER~\cite{cynthetizer} in the commercial world. Despite their 90 90 maturity, their usage is restrained by \cite{IEEEDT} \cite{CATRENE} \cite{HLSBOOK}: 91 91 \begin{itemize} … … 98 98 designs are multi-constrained, 99 99 Moreover, low power consumption constraint is mandatory for embedded systems. 100 However, it is not yet well handled or not handle at all by the synthesis tools already available,100 However, it is not yet well handled or not handled at all by the synthesis tools already available, 101 101 \item The parallelism is extracted from initial algorithmic specification. 102 102 To get more parallelism or to reduce the amount of required memory in the SoC, the user … … 106 106 pipelining, current HLS tools do not provide support for design space exploration neither 107 107 through automatic loop transformations nor through memory mapping, 108 \item Despite they havethe same input language (C/C++), they are sensitive to the style in109 which the algorithm is written. Consequently, engineering work is required to swap from108 \item Despite having the same input language (C/C++), they are sensitive to the style in 109 which the algorithm dis written. Consequently, engineering work is required to swap from 110 110 a tool to another, 111 111 \item They do not respect accurately the frequency constraint when they target an FPGA device. 112 112 Their error is about 10 percent. This is annoying when the generated component is integrated 113 in a SoC since it will slow down the hole system.113 in a SoC since it will slow down the whole system. 114 114 \end{itemize} 115 115 Regarding these limitations, it is necessary to create a new tool generation reducing the gap
Note: See TracChangeset
for help on using the changeset viewer.