llvm dev - Mar 2018 - Externally loadable Alias Analysis pass

Hello,

I am currently working with the implementation of CFL Andersen algorithmand
I am trying to do some improvements to the Andersen algorithm.

I would like to create loadable module .so and run clang with my e.g.
"cfl-anders-improved.so" to avoid the "long" way -
generating IR, running
opt with my improved algorithm to generate optimalized code and then
building it.

Is a such thing even possible for Alias analyses in Clang?

Thanks
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20180308/72f51c8f/attachment.html>

Hi, David,

We do have an infrastructure for incorporating custom AA providers into
the pipeline. I've never tried it, but take a look at
unittests/Analysis/AliasAnalysisTest.cpp (that may provide some of what
you need).

 -Hal

On 03/08/2018 08:52 AM, Dávid Bolvanský via llvm-dev
wrote:
> Hello,
>
> I am currently working with the implementation of CFL Andersen
> algorithmand I am trying to do some improvements to the Andersen
> algorithm.
>
> I would like to create loadable module .so and run clang with my e.g.
> "cfl-anders-improved.so" to avoid the "long" way -
generating IR,
> running opt with my improved algorithm to generate optimalized code
> and then building it.
>
> Is a such thing even possible for Alias analyses in Clang?
>
> Thanks
>
>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
-- 
Hal Finkel
Lead, Compiler Technology and Programming Languages
Leadership Computing Facility
Argonne National Laboratory

-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20180308/f209b258/attachment.html>

I am currectly experimenting with -Xclang option.. I have a simple pass (to
just print function names).

I ran: clang -Xclang -load -Xclang LLVMHello.so t.c

but nothing was printed, no output from the pass.

when I run: opt -load LLVMHello.so -hello t.ll

I get the output from pass (i.e. list of function names)

Part of the source ode of the Hello pass:
  struct Hello : public FunctionPass {
    static char ID; // Pass identification, replacement for typeid
    Hello() : FunctionPass(ID) {}

    bool runOnFunction(Function &F) override {
      errs().write_escaped(F.getName()) << '\n';
      return false;
    }
  };
}

2018-03-09 14:44 GMT+01:00 Dávid Bolvanský <david.bolvansky at gmail.com>:
> I am currectly experimenting with -Xclang option.. I have a simple pass
> (to just print function names).
>
> I ran: clang -Xclang -load -Xclang LLVMHello.so t.c
>
> but nothing was printed, no output from the pass.
>
> when I run: opt -load LLVMHello.so -hello t.ll
>
> I get the output from pass (i.e. list of function names)
>
> Part of the source ode of the Hello pass:
>   struct Hello : public FunctionPass {
>     static char ID; // Pass identification, replacement for typeid
>     Hello() : FunctionPass(ID) {}
>
>     bool runOnFunction(Function &F) override {
>       errs().write_escaped(F.getName()) << '\n';
>       return false;
>     }
>   };
> }
>
> 2018-03-09 5:38 GMT+01:00 Hal Finkel <hfinkel at anl.gov>:
>
>> Hi, David,
>>
>> We do have an infrastructure for incorporating custom AA providers into
>> the pipeline. I've never tried it, but take a look at
>> unittests/Analysis/AliasAnalysisTest.cpp (that may provide some of what
>> you need).
>>
>>  -Hal
>> On 03/08/2018 08:52 AM, Dávid Bolvanský via llvm-dev wrote:
>>
>> Hello,
>>
>> I am currently working with the implementation of CFL Andersen
>> algorithmand I am trying to do some improvements to the Andersen
algorithm.
>>
>> I would like to create loadable module .so and run clang with my e.g.
>> "cfl-anders-improved.so" to avoid the "long" way -
generating IR, running
>> opt with my improved algorithm to generate optimalized code and then
>> building it.
>>
>> Is a such thing even possible for Alias analyses in Clang?
>>
>> Thanks
>>
>>
>> _______________________________________________
>> LLVM Developers mailing listllvm-dev at
lists.llvm.orghttp://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>>
>>
>> --
>> Hal Finkel
>> Lead, Compiler Technology and Programming Languages
>> Leadership Computing Facility
>> Argonne National Laboratory
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20180309/fe00fcf1/attachment-0001.html>