[PATCH] D34458: [TTI] Refine the cost of EXT in getUserCost()

[Eli Friedman via Phabricator via llvm-commits]

efriedma added a comment.

> The regression in spec2006/h264ref is caused by a fully unrolled loop. The loop has an indrect call inside it. If transforming the indirect call to direct call, the performance can be improved by +10%, and then unrolling this loop can get additional +3% improvement. We are working on it.

You're planning a change to the unroller?  Or something else?

---

It looks like this has some impact on other targets; do we have any test coverage?



================
Comment at: include/llvm/Analysis/TargetTransformInfoImpl.h:720
+      if (isa<SExtInst>(CI) || isa<ZExtInst>(CI) || isa<FPExtInst>(CI))
+        return static_cast<T *>(this)->getExtCost(CI);
     }
----------------
Do you need to pass through Arguments here?  It looks like getExtCost depends on the operands.


Repository:
  rL LLVM

https://reviews.llvm.org/D34458