- Timestamp:
- Jan 26, 2011, 7:29:20 PM (14 years ago)
- Location:
- anr
- Files:
-
- 2 added
- 3 edited
Legend:
- Unmodified
- Added
- Removed
-
anr/section-consortium-people.tex
r327 r330 44 44 45 45 \peopletabularentry{\lip} 46 responsible & Alias & Christophe & junior reasearcher & HLS & ... & Memory management, HLS\\\hline47 contributor & Feautrier & Paul & emeritus professor & Compil . & ...& Process scheduling and building.46 responsible & Alias & Christophe & Permanent researcher at Inria & Compilers, HPC & 15 & Memory management and FIFO construction. Irregular extensions.\\\hline 47 contributor & Feautrier & Paul & emeritus professor & Compilers & 13 & Process scheduling and building. 48 48 Automatic Parallelization \\ \hline 49 49 -
anr/section-ressources.tex
r325 r330 75 75 workstations is evaluated to 4\% of the total requested ANR funding. 76 76 \item [Personnel costs] 77 The faculty members involved in the project are an emeritus 78 professor at ENS Lyon (Paul Feautrier) and a research associate 79 (CR2) at INRIA Rh\^one-Alpes (Christophe Alias). The non-permanent 80 personel required is a PhD student that will work on network process 81 generation from polyhedral loops, then on extensions to 82 non-polyhedral loops. We are looking for a student with both 83 theoretical and practical skills, that will be able to get a 84 sufficient understanding of the polyhedral techniques and to produce 85 a working implementation. 86 \parlf 87 The table below summarizes the \hommemois by 88 deliverables and tasks for both permanent and non-permanent 89 personnels. The non-permanent personnels costs represent 26\% of 90 the personnal costs. The requested funding for non 91 permanent personnels is 100\% of the total ANR requested funding. 92 \begin{center}\input{table_inria_compsys_full.tex}\end{center} 77 The faculty members involved in the project are Christophe Alias 78 (INRIA researcher) and Paul Feautrier (emeritus professor at 79 ENS-Lyon). The non-permanent personel required is a post-doc that 80 will work on FIFO construction, then on extensions of process 81 construction and memory optimization to non-polyhedral loops. We 82 are looking for a candidate with both theoretical and practical 83 skills, that will be able to get a sufficient understanding of the 84 polyhedral techniques to produce a working implementation. \parlf 85 The table below summarizes the \hommemois by deliverables and tasks 86 for both permanent and non-permanent personnels. The non-permanent 87 personnels costs represents ??\% of the personnal costs. The 88 requested funding for non permanent personnels is 100\% of the 89 total ANR requested 90 funding. \begin{center}\input{table_compsys_short.tex}\end{center} 93 91 \item [Subcontracting] 94 92 No subcontracting costs. … … 299 297 \end{figure} 300 298 \end{landscape} 299 % 300 \begin{landscape} 301 \begin{figure} 302 \begin{small} 303 \settowidth\desclen{XILINX RTL optimisation (5)} 304 \def\resstablestyletitle#1{\parbox{\desclen}{\textit{#1}}} 305 \hfill 306 \begin{minipage}[b]{.47\linewidth}\center 307 \input{table_compsys_full.tex}\vspace{.5ex}\\ \lip \vspace{2.5ex}\\ 308 \end{minipage}\hfill 309 \end{small} 310 \caption{\label{table-livrables-2} Man power in \hommemois for the deliverables (3)} 311 \end{figure} 312 \end{landscape} -
anr/task-frontend.tex
r328 r330 59 59 60 60 \subtask{Automatic parallelization and memory optimization} 61 62 61 This sub-task aims at providing a source-level optimizer in front the 63 HLS back-end tools. The optimization are twofold:62 HLS back-end tools. The optimizations are threefold: 64 63 \begin{itemize} 65 64 \item Extraction of parallelism in polyhedral loops and conversion … … 70 69 We will design these methods by using polyhedral techniques, as we did 71 70 in the past for pure HPC optimizations. The program model is typically 72 regular programs w ith affine loop bound conditions and array index73 functions. In a second part, we will extend the program model by using 74 conservative approximations.71 regular programs where loop bounds, conditions and array indices are 72 affine functions. In a second part, we will extend the program model 73 by using conservative approximations. 75 74 \begin{livrable} 76 75 \itemV{0}{6}{d}{\Slip}{Method, Preliminary Definition} … … 85 84 \itemV{6}{12}{d}{\Slip} {Arrays and FIFO} 86 85 Description and specification of the FIFO construction method 87 and local memory optimization algorithms.86 and the local memory optimization method. 88 87 \itemV{12}{18}{d+x}{\Slip} {Arrays and FIFO} 89 88 Preliminary implementation in the Bee framework. 90 89 Conversion of the input and output of Bee to the \xcoach format. 91 At this step, only memory optimizationswill be available.90 At this step, only local memory optimization will be available. 92 91 \itemL{18}{30}{d+x}{\Slip}{Arrays and FIFO}{10:0:9} 93 92 Final assessment of the method and improved version of the specification.
Note: See TracChangeset
for help on using the changeset viewer.