[ValueTracking] Lower dom-conditions-dom-blocks and dom-conditions-max-uses thresholds

[igor-laevsky]

Summary

On some of our benchmarks this change shows about 50% compile time improvement without any noticeable performance difference.

[reames]

To give some more context here, this appears to be related to the use of isKnownNonNullAt in InstCombine to tag parameters as nonnull so that the inliner can exploit that fact. Dropping the max-uses parameter addresses the compile time impact we've observed without detrimental effects being observed.

We're also re-evaluating whether the context sensitivity in InstCombine is necessary at all. Igor added support to CorrelatedValueProp to do the same thing and InstCombine is run far more frequently than CVP is. It may be worth removing the handling from InstCombine entirely (or at least going back to the non-context sensative verserion), but we're still evaluating the impact of that change on our local tree.

lib/Analysis/ValueTracking.cpp

 // (TODO: evaluate cost vs profit, consider higher thresholds)
 static cl::opt<unsigned> DomConditionsMaxDepth("dom-conditions-max-depth",
                                                cl::Hidden, cl::init(1));
 
 /// How many dominating blocks should be scanned looking for dominating
 /// conditions?
 static cl::opt<unsigned> DomConditionsMaxDomBlocks("dom-conditions-dom-blocks",
                                                    cl::Hidden,
-                                                   cl::init(20000));
+                                                   cl::init(20));
 
 // Controls the number of uses of the value searched for possible
 // dominating comparisons.
 static cl::opt<unsigned> DomConditionsMaxUses("dom-conditions-max-uses",
-                                              cl::Hidden, cl::init(2000));
+                                              cl::Hidden, cl::init(20));
 
 // If true, don't consider only compares whose only use is a branch.
 static cl::opt<bool> DomConditionsSingleCmpUse("dom-conditions-single-cmp-use",
                                                cl::Hidden, cl::init(false));
 
 /// Returns the bitwidth of the given scalar or pointer type (if unknown returns
 /// 0). For vector types, returns the element type's bitwidth.
 static unsigned getBitWidth(Type *Ty, const DataLayout &DL) {