[LLVMdev] Data dependence analysis

[Tanya M. Lattner]

> Data Dependence Analysis is great. I am sure it would help the loop
> optimizations effort that Dr. Adve mentioned last month. After we have
> DDA we can implement software pipelining* (modulo scheduler). Chris,
> would you accept a software pipelining transformation as a pass or would
> you want it as part of the different backends ?

For my masters thesis I implemented Modulo Scheduling in the original LLVM 
Sparc backend (since removed though). To me, its seems very specific to a 
certain target and should be part of specific backend as machine function 
pass (ie. sparc). How do you plan to make it target independent?

-Tanya


>
> Nadav Rotem
>
> * http://en.wikipedia.org/wiki/Software_pipelining
>
>
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>