From 5cba8f7b27a5d46b7189126d60406c9edc7672d3 Mon Sep 17 00:00:00 2001
From: Ryan Newton <rrnewton@gmail.com>
Date: Thu, 23 Oct 2014 23:03:00 -0400
Subject: [PATCH] minor: note about aforementioned benchmarks

---
 DEVLOG.md | 13 +++++++++++++
 1 file changed, 13 insertions(+)

diff --git a/DEVLOG.md b/DEVLOG.md
index a574295b..16bcaa11 100644
--- a/DEVLOG.md
+++ b/DEVLOG.md
@@ -817,7 +817,18 @@ But the MINIMUM doesn't go as low as -A20M, and it's USUALLY worse
 than the -A20M version.
 
 
+[2014.10.23] {Criterion Microbenchmarking}
+------------------------------------------
+
+Adding simple microbenchmarks for "return ()" and fork.  The NOOP
+benchmark takes 18.5 ns if the "benchPar" helper is not inlined.
+Adding inlining drops it down to one iteration/cycle, e.g. "283
+ps"/iter.
 
+Also dramatically, inlining benchPar drops the fork microbenchmork
+from 1.6 microseconds / 751 bytes, to 389ns/600-bytes.  That
+allocation is still high enough that it's not completely deforesting
+the Trace structure...
 
 
 TEMP / SCRAP:
@@ -830,3 +841,5 @@ TEMP / SCRAP:
 
 
 
+
+