From emaste at freebsd.org  Thu Nov 20 23:32:42 2014
From: emaste at freebsd.org (Ed Maste)
Date: Thu, 20 Nov 2014 18:32:20 -0500
Subject: pmcstat -z32 -G truncates callgraph to 8
In-Reply-To: <CAE40pdeQvmEDQRZo=v06Dzg7n0+gHNqPNvWp3R7cAQ2bBd3yoA@mail.gmail.com>
References: <CAE40pdf5AzR4tuGZUQugoK6jfbvd_dLLBtdfv2FPQYY3Lb5PrA@mail.gmail.com>
 <20141028114537.83f9f0a3c730949f453c3819@systemdatarecorder.org>
 <CAE40pdf6jUGEse0Mh+gV48VnwLrWS+fhO59gfzY=ocRauK43MA@mail.gmail.com>
 <CAPyFy2Df6NO4x-NG7eUiCbN8_9+vNAdtSnTg++PSnnoYeiEStg@mail.gmail.com>
 <CAE40pdeQvmEDQRZo=v06Dzg7n0+gHNqPNvWp3R7cAQ2bBd3yoA@mail.gmail.com>
Message-ID: <CAPyFy2D4_puQPg1zvE4-Eyaa9jR_v7xc988qGOygConA-aLHUA@mail.gmail.com>

On 31 October 2014 01:26, Brendan Gregg <brendan.d.gregg at gmail.com> wrote:
>
> Yes. When I'm using profiling data, I like full stacks for making
> flame graphs. For the FreeBSD kernel, 32 frames should usually be
> enough (I have a flame graph that reaches 24 frames for the kernel,
> but no more). For user-level, I'd probably need ~100. So making the
> compiled-in default to 32 would hopefully be sufficient for most
> kernel profiling, and one would need to bump that up for deep
> user-level stacks. I guess this would also need
> PMC_CALLCHAIN_DEPTH_MAX = 128 to work.

I have a review up to bump the default to 32 and the maximum to 128:
https://reviews.freebsd.org/D1203