On Thu, Mar 30, 2017 at 4:04 AM, Hartmut Kaiser <[email protected]> wrote:
> > Hey Abhi, > > > It sounds like you've got a good grasp of the problem and the right > > skill set for this project. > > > > It sounds to me like you've done plenty of research into the project, > > so my only suggestion would be to spend a little bit of time looking > > at HPX (documentation, examples, etc). If you haven't done so already, > > it might be good to take an hour or two and try building HPX and > > playing around with some of the examples. > > > > Thanks Bryce, > > > > I received your mail and I am glad that you find this projects > > interesting, and your warm reply already makes me feel home with the > > community. I have already built it on my host, and done some sample > thread > > programs as well. > > > > I have been digging around the problem and been though 100's of resources > > from HN (HackerNews) to StackOverflow (the website) and been through > > Stellar Org's. previous projects, its very facinating work. Also been > > through the guide to contribute, from commit messages to coding > > conventions, all coverd up. > > > > Besides as the date for submitting the proposal is closing in I have a > few > > questions about that too. > > > > 1. How do you want me to submit a proposal, the level of granulity i.e. > > detail, I should write. > > We have a suggested template for proposals here: > https://github.com/STEllAR-GROUP/hpx/wiki/GSoC-Submission-Template. Also > a uideline for writing successful proposals here: > https://github.com/STEllAR-GROUP/hpx/wiki/Hints-for-Successful-Proposals. > You can also google around to find more suggestions from other > organizations. Those usually apply just fine. > > The bottom line is that you want to convince us that you'll be able to > deliver something nice and functioning in the end. That would include a > description of what you propose to do, what alternatives you'd like to try, > what is your proposed time line, etc. > > Ok My current approach revolves around 1. Using libsigsegv with the threads working via HPX,The core functionality of libsigsegv let us jump/exit and clean up ( in the case of a space efficient program) or make inconsistent data structures becomes consistent. First point, it falls under GPL lisence so it can only be used for testing and some benchmarking work. 2. About the legacy stackoverflow support in HPX, they are lightweight and efficient in problems where not so much dependency is involved, like we use Guard pages. Now combining my above two points, how the design will respond if we introduce flag based scheme and one default. Suppose, most of thread based programs are not writing any global variables or stateful changes, so we can keep Guard pages ( kind of what I propose to integrate/develop) as default, and for stateful threads we can use flags and include some methos for cleaning up and handler classes ( for this lets actively seeks for soultion). But having this inital support will have a great push to get ball rolling and actually having something which can tackle atleast 3/5th of the problem domain. I have seen a few canary value based methods which I can intergrate in HPX as an intial part of my work and then extending the support for further components in the problem domain. > 2. I have a few approaches around but I need your go ahead to affirm which > > ones you guys want, or shall I mentions a few and then later decide as > > mentioned on the website that even after proposal submission we should > > stay sync > > Yes, that sounds ok. > > > 3. Is there any thing I can start doing right now to increase my chances > > to be accepted? write a patch ? doing some contribution or something? > > Start discussing your idea either here on the mailing list or on the IRC > channel (#ste||ar on freenode). > > > Last but not the least, as it's just 5 days left to submit a proposal, so > > I would need a lot of your guidance to make a promising proposal. Kindly > > bear with me for writing late to you guys and making this such a hurry in > > last minute. I am very excited about the work and can't wait to get > > started this summer. > > Sure, you're not the only one being late, we're used to that. Just get > things going we'll comment as you go. > > Regards Hartmut > --------------- > http://boost-spirit.com > http://stellar.cct.lsu.edu > > > > > > Cheers, > > Abhi > > > > > > Given that this is an ambitious project, it might be helpful for you > > to send a proposed work plan to this mailing list before you submit so > > we can provide feedback. > > > > Overall, sounds good! > > > > On Tue, Mar 28, 2017 at 2:08 PM, Abhimanyu Rawat > > <[email protected]> wrote: > > > Hello Bryce, > > > > > > I am Abhimanyu Rawat, final year computer science student at Birla > > Institute > > > of Technology and Science (BITS, Pilani Campus). I hope I am not too > > late > > > showing an interest to you guys regarding the Stack Overflow Detection > > on > > > Linux project. > > > > > > Earlier I was guiding my junior Arpan for the same, but due to some > > > unavoidable issue, he won't be taking part this time. Currently, I am > an > > > intern at EMC^2 for a couple of more weeks, I am working on operating > > system > > > tools for the company. I have ample experience in operating system and > > > design, I have been using C and C++ for more than 5 years now. In > > college, I > > > have taught certified course on Operating system( where Arpan was > > enrolled), > > > spearheaded an inter-university project with Stanford along with Head > of > > > Computer Science Department of BITS Pilani, Lab instructor of Computer > > > Network session and in charge of HCI lab, developed various course > > projects > > > among which an OpenMP related where I used parallel threads to exchange > > the > > > data between processes resulting in the weighted graph showing file > > > association and more. I am also a fast sudoku solver, loves to teach > > kids > > > and helps my friends with their university projects as well. > > > > > > I have already been though Gaurd pages, GNU libsigsegv and done a lot > of > > > research ( all legacy techniques - stackGuard, PointGuard, canary etc) > > on > > > various components that take part in the existing implementation on > > Windows > > > platform, also how libsigsegv can help if we go for the integration > with > > HPX > > > ( which obviously comes with some tradeoff's if space and time are > > factors), > > > or we can develop a hybrid unit and see how it can support HPX. > > > > > > As HPX widely uses threading processes, so libsigsegv will also not be > a > > bad > > > choice, from issue tracker I figured out that using libsigsegv we can > > also > > > support flag based options, and provide a generic method for every > other > > > general segmentation fault and stack overflow (it can be done using > > legacy > > > techniques). > > > > > > It would be great if you could review this above abstract of what I > have > > > accumulated and throw some pointers at me so that I can get back to you > > guys > > > with my proposal in no time ( --t tick tock tick tock). > > > > > > Overall I am very excited to work on the project, with the community > and > > > improving my skills by actively contributing to the project(in long run > > > also). > > > > > > Closing with thank you and warm Regards, > > > > > > Abhimanyu Rawat > > > M.E. Computer Science, > > > CS/IS Department, BITS Pilani, Pilani Campus > > > Email - [email protected] / [email protected] > > > Phone. 08930399302 (call/Whatsapp), 09466899302 > > > > > > ????????????????????????????????????????????? > > > > > > > > -- > > Bryce Adelstein Lelbach aka wash > > Lawrence Berkeley National Laboratory > > ISO C++ Committee Member > > CppCon and C++Now Program Chair > > > > Replies are often throughput optimized at the expense of latency! > > -- > >
_______________________________________________ hpx-users mailing list [email protected] https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
