Hi, I'm a MSc student interested in high-performance computing. This spring I will write my master thesis, and would like to work on code transformation in LLVM. I imagine that the community knows what might a good project, so I thought I'd ask you. I would like to work on a project that could be - of value for the LLVM project - of research value Thanks in advance, Karl Hylén -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20141126/54679ace/attachment.html>
On Wed, Nov 26, 2014 at 2:45 AM, Karl Hylén <karl.hylen at gmail.com> wrote:> Hi, > > I'm a MSc student interested in high-performance computing. This spring I > will write my master thesis, and would like to work on code transformation > in LLVM. I imagine that the community knows what might a good project, so I > thought I'd ask you. > > I would like to work on a project that could be > > - of value for the LLVM project > - of research value > > >If you have the theoretical chops, a formal characterization of the vector shuffle lowering problem would be useful. -- Sean Silva> Thanks in advance, > Karl Hylén > > > _______________________________________________ > LLVM Developers mailing list > LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu > http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev > >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20141126/e2722edf/attachment.html>
----- Original Message -----> From: "Karl Hylén" <karl.hylen at gmail.com> > To: llvmdev at cs.uiuc.edu > Sent: Wednesday, November 26, 2014 4:45:45 AM > Subject: [LLVMdev] Master thesis > > Hi, > > > I'm a MSc student interested in high-performance computing.HPC is a large, somewhat generic term. Do you have specific areas of interest within HPC?> This > spring I will write my master thesis, and would like to work on code > transformation in LLVM.How many months/years will you have to spend on the R&D part of your thesis work?> I imagine that the community knows what > might a good project, so I thought I'd ask you.There are lots of potential topics of interest, from higher-level things like loop transformations, memory hierarchy modeling, etc. to lower-level issues involving code generation (i.e. vectorization), non-traditional architectures (GPUs, etc.), etc. If you can give us a little more context, and your level of familiarity with LLVM, it will be easier to suggest some projects. -Hal> > > I would like to work on a project that could be > > > * of value for the LLVM project > * of research value > > > Thanks in advance, > Karl Hylén > > > _______________________________________________ > LLVM Developers mailing list > LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu > http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev >-- Hal Finkel Assistant Computational Scientist Leadership Computing Facility Argonne National Laboratory
On Wed, Nov 26, 2014 at 5:45 AM, Karl Hylén <karl.hylen at gmail.com> wrote:> Hi, > > I'm a MSc student interested in high-performance computing. This spring I > will write my master thesis, and would like to work on code transformation > in LLVM. >I am a Masters student. What kind of transformation you would like to do? As in CPU to GPU code transformation or GPU higher level to lower level or creating a new language or new backend for language or focusing on creating better algorithms for parallel compilation. Where does your interest lie? And, If you have only one semester master's thesis, it's really hard to make significant research. It takes atleast 3 months to get to the good stuff (assuming you have other courses to do).> I imagine that the community knows what might a good project, so I thought > I'd ask you. >If you are looking for a project, applying through GSOC 2015 would be good. (Can be used for thesis)> > I would like to work on a project that could be > > - of value for the LLVM project > > Decent patches can do (you can use these in thesis).> > - of research value > > Assuming you are focusing on parallelizing compilers, this is the topicwhere lots of research is done. But, we are unable to decrease the gap between hand written and compiler parallelized code (some times performance gets degraded by trying to parallelize from serial code).> > Thanks in advance, > Karl Hylén > >Finally what I want to say is, the time you got is not sufficient. If you are looking for research (writing papers), start reading papers on compilers. I have seen lots of people in this mailing list who have done amazing research work on compilers (Chris Lattners team at Apple). What I found is that, they implemented ideas using LLVM. You need to find a piece which you want to solve and use LLVM as a tool. Thank you!!!> _______________________________________________ > LLVM Developers mailing list > LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu > http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev > >-- Regards, *Aditya Atluri,* *USA.* -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20141130/65cb43cf/attachment.html>