This updatesCompiler::fgFindJumpTargets to support resolving call tokens when making inline observations for Tier 1.

A test ofcsc compiling a simpleHelloWorld program (csc.dll Program.cs /r:System.Runtime.dll /r:System.Console.dll /r:System.Private.Corelib.dll) withCOMPlus_ReadyToRun=0 andCOMPlus_TieredCompilation=0 resulted in the following impact:

• Compiler::fgFindJumpTarget goes from0.48% to0.63% of instructions retired (28.5m to39.25m)
• CEEInfo::resolveToken goes from0.32% to0.34% of instructions retired (19m to21m)
• CEEInfo::getCallInfo goes from0.36% to0.32% of instructions retired (21.25m to19.75m)

As per the flame graph (see below), the majority of time is still spent in the backend. There was no measured impact fortier 0 methods.

This change will allow us to make potential improvements to inlining for methods that involve many intrinsic calls. Some potential metrics we could use to influence inlining include:

• How many calls areintrinsic (Named or Otherwise)
• How many calls arehwintrinsic and therefore are not calls at all, but are effectively "simple ops"
• How many calls are CSE or constant folding candidates (such asMath calls)
• How many calls are "constants" likeIsa.IsSupported
• How many calls are potential constants, likestring.Length orType.op_Equality checks

We can place this behind aDOTNET_* (COMPlus_*) switch if there are any remaining concerns around impact.