[llvm-dev] GSOC 2016 Project proposal

[Sanjoy Das via llvm-dev]

Hi Shashidhar,

I've CC'ed some people who have a better understanding of alias
analyses than I do (so maybe they'll be able to suggest something),
but I have a couple of general questions below:

On Mon, Mar 14, 2016 at 5:27 AM, Shashidhar G via llvm-dev
<llvm-dev at lists.llvm.org> wrote:
> of Alias Analysis. I have implemented a simple version of Andersen's
> Analysis as LLVM Pass and have also worked on LLVM Code generator for a
> micro controller architecture.

Are any of these code bases open source?  A github link or something
like that demonstrating you can work effectively with the LLVM
code-base will really help in establishing credibility.

>    I am more interested to work on a new idea for improving the Alias
> analysis and the underlying optimizations. I am going through some research
> papers in this area.

Did you find anything specific that you want to implement in LLVM?  A
strong application should contain both what you're going to do, and
how you'll go about it.  For something like alias analysis, you
probably need strong motivation too -- i.e. "LLVM's current alias analysis
stack is unable to optimize ${example}, and this new alias analysis
will fix that".

Thanks,
-- Sanjoy