From e72595ee0f97746be3ce594834a7003d5e804795 Mon Sep 17 00:00:00 2001 From: Austin Clements Date: Tue, 25 Jul 2017 15:03:44 -0400 Subject: [PATCH] cmd/trace: notes on MMU view improvements Change-Id: Ib9dcdc76095f6718f1cdc83349503f52567c76d4 Reviewed-on: https://go-review.googlesource.com/c/60801 Run-TryBot: Austin Clements TryBot-Result: Gobot Gobot Reviewed-by: Hyang-Ah Hana Kim --- src/cmd/trace/mmu.go | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+) diff --git a/src/cmd/trace/mmu.go b/src/cmd/trace/mmu.go index 062e5ad2ca..6a7d28e61d 100644 --- a/src/cmd/trace/mmu.go +++ b/src/cmd/trace/mmu.go @@ -4,6 +4,25 @@ // Minimum mutator utilization (MMU) graphing. +// TODO: +// +// In worst window list, show break-down of GC utilization sources +// (STW, assist, etc). Probably requires a different MutatorUtil +// representation. +// +// When a window size is selected, show a second plot of the mutator +// utilization distribution for that window size. +// +// Render plot progressively so rough outline is visible quickly even +// for very complex MUTs. Start by computing just a few window sizes +// and then add more window sizes. +// +// Consider using sampling to compute an approximate MUT. This would +// work by sampling the mutator utilization at randomly selected +// points in time in the trace to build an empirical distribution. We +// could potentially put confidence intervals on these estimates and +// render this progressively as we refine the distributions. + package main import ( -- 2.50.0