As seen in #44790 and elsewhere, the crude model we use for the IL stack in fgFindJumpTargets can give false positives when assessing inline profitability.
It's possible that an improved model might (in part) pay for itself by eliminating some of these inlines with unrealistic profitability estimates. The big holdup in the past has been calls; we need a quick way to go from a method token to a method sig so we can assess the call's push/pop impact.
category:cq
theme:inlining
skill-level:expert
cost:large
As seen in #44790 and elsewhere, the crude model we use for the IL stack in
fgFindJumpTargetscan give false positives when assessing inline profitability.It's possible that an improved model might (in part) pay for itself by eliminating some of these inlines with unrealistic profitability estimates. The big holdup in the past has been calls; we need a quick way to go from a method token to a method sig so we can assess the call's push/pop impact.
category:cq
theme:inlining
skill-level:expert
cost:large