Who is interested in automatic C++ (or C#) code parallelizer? Should I start development?

I'm talking about smart automatic parallelizer (with ability to do parallel code even from linear) - not about simple handparallelizer like inside of Visual Studio 2011. And actually I don't care about supported languages - pluginable system. C++ is just because I'm more familiar to C++. But it also can be Java or Python, or something.

Of course I already have groundwork of ideas similar to things, that as I already know :) , called BNF (Backus–Naur Form) and AST (Abstract Syntax Tree). And even some pseudo-code. But all this things are 6-7 years old and so their implementation need time for think over and time for actual implementation.

So I wondering is it worth to spend my time for this. As result, I'm interested in real life usage of that parallelizer - not just in academic curiosity.

DAPS (Data Access Prevention System). Alpha version preview.

The very first DAPS alpha version preview video. DAPS (http://www.mihail-fieryice-butenko.in.ua/project-definition/daps) - is my protection system against theft and damage to confidential information.



(http://www.youtube.com/watch?v=Z6vwx-DrsjA)

Why do I need yet another license?

Summary: I want to make a profitable spread of commercial development in the form of free software, and at the same time - not burdensome use of these commercial developments, even for the poor developers. Of course there are more complex and interesting.

It's not dead

The project is not dead. At the moment, work continues on the license, which will be distributed FITaFS and its source code. The thing is that I have decided not to use the GNU/GPL. In addition I am not interested in using Permissive free software licences.

Theses

At conference participated, a diploma had protected. Freedom (almost :) )! As a result of appearance of additional spare time I, in the nearest terms, will set to work above realization of «Fitafs» under FUSE.

The impressions from a conference are positive both from the heard information and from circumstance that - the real participants was not hurries with a lecture there was in once or twice a less amount of requests. To lay out the article of lecture will not be while, but will lay out theses.

So...

Tag-based filesystems. The features of organization of data on the example of «Fitafs»

Classic FS (file systems) with the hard hierarchical structure of catalogues successfully used already more than 30 years. However now that the volumes of storable users information are calculated in tens and by hundreds gigabyte, arrangement and retrieval of the required data appears either strongly laboured or not comfortable. The matter is an all that hard hierarchy and impossibility of dynamic reflection of all information, proper the set criterion in one place, by facilities FS. Certainly, an user has a good mechanism of hard Refs on files, however and it far not panacea. For example, we want to represent all present on a computer voice files in the folder of «Music» – it is for this purpose necessary it will be to find all necessary files, create on them hard Refs.s inwardly «Music» (thus it will be necessary to change the original names in order to avoid a reiteration), and periodically to scan the system in the presence of new voice files. If inwardly «Music» must be reproduced part of original hierarchy, as that: « performer \ album \ file_name.mp3», the that amount of operations yet more will be increased. For the removal of similar problems the great number of realization of FS and buildings (control system by information and metadatas, being not the file systems) on, representing information as pages of DB (databases) was developed, some of which use SQL as the mother tongue of queries. To them behave: «WINFS» (Microsoft), «Hyppocampus». However and they are not deprived failings main which it is been from: 1) absence of complete reverse compatibility with classic FS and using them programs; 2) most similar systems are buildings on above existing FS and use external universal SUBD, that notedly increases loading on CPU and expense of memory (both operative and permanent). As appeared, there is another not obscure mechanism of reflection and arrangement of information – tags. Yes, certainly, systems, based on databases tags use also, but only in a role of additional properties of file. In a difference from them tag-based filesystems («Fitafs», «Tagsistant») make from tags way to the file. If file 1.mp3 tags [music, the classics]has, and file 2.mp3 - [music, fate], passing on any of ways [«music\fate» or «fate\music»] we will find a file 1.mp3. In also time in a catalogue «music» it is possible to find out both files - 1.mp3 and 2.mp3. Directory tree, in most cases is a list of all tags-catalogues, into each of which again list from all tags-catalogues (except for already chosen) together with the list of files, having set tags. Complete reverse compatibility is the same provided with classic FS. However, at plenty of tegov, an interface becomes uncomfortable, and the enormous inlaid can result in serious problems in the programs, trying to go (proskanirovat') round all tree folders. Developed by me the tag-based filesystem of «Fitafs» uses a few other approach, leveling the problem of over-surplus inlaid of tegov-catalogues. In a lecture the methods of removal of the above enumerated problems will be lighted up with an interface, management the volumes of information and increase fast-actings, used for planning of «Fitafs». And also comparing of «Fitafs» to the existent systems.

Butenko M.S. 2008

First note ))

First note that my lecture is accepted in the program of conference of SAIT-2008 (IPSA KPI). And it means that I for the first time will be a speaker at scientific conference.)) I will tell about the "tag-based filesystems" in general and about own similar development - "FITaFS" - in particular.