From d48762fa1dd32c8f99ea527d9655f460e08ba08b Mon Sep 17 00:00:00 2001 From: Vincent St-Amour Date: Fri, 29 Jun 2012 17:11:03 -0400 Subject: [PATCH] Comment on tradeoff. --- collects/typed-racket/optimizer/tool/mzc.rkt | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/collects/typed-racket/optimizer/tool/mzc.rkt b/collects/typed-racket/optimizer/tool/mzc.rkt index 5dbd00530f..3fdfdd47e9 100644 --- a/collects/typed-racket/optimizer/tool/mzc.rkt +++ b/collects/typed-racket/optimizer/tool/mzc.rkt @@ -169,7 +169,8 @@ ;; For instance, if `f' is a loop, and gets inlined in `g' multiple ;; times, it's likely to be unrolling. Same for out-of-fuels in `g'. ;; Therefore, we don't want to report these as inlinings (or failed - ;; inlinings). + ;; inlinings). If `g' has multiple call sites for `f', we lose + ;; precision, and may discard actual inlinings. ;; However, we care about `f' being unrolled at least once in `g'. ;; If we run out of fuel trying to inline `f' in `g' for the first ;; time, we report. The reason for this is that it's possible to