Date: Wed, 9 Mar 2022 14:37:31 -0600
From: Justin Pryzby <pryzby@telsasoft.com>
To: Andres Freund <andres@anarazel.de>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, Robert Haas <robertmhaas@gmail.com>,
	Andrew Dunstan <andrew@dunslane.net>, pgsql-hackers@postgresql.org,
	Thomas Munro <thomas.munro@gmail.com>,
	Melanie Plageman <melanieplageman@gmail.com>,
	Peter Eisentraut <peter.eisentraut@enterprisedb.com>,
	Daniel Gustafsson <daniel@yesql.se>
Subject: Re: Adding CI to our tree
Message-ID: <20220309203731.GB28503@telsasoft.com>
References: <20220226025116.GJ9008@telsasoft.com>
 <20220227010908.vz2a7dmfzgwg742w@alap3.anarazel.de>
 <20220227024352.GC25269@telsasoft.com>
 <20220227025000.mjmfimls5gnz7em7@alap3.anarazel.de>
 <20220227031057.GD25269@telsasoft.com>
 <20220227040838.3z4gqihimyerxe3h@alap3.anarazel.de>
 <20220228205802.GE25269@telsasoft.com>
 <20220302205058.GJ15744@telsasoft.com>
 <20220309174723.GA28503@telsasoft.com>
 <20220309181254.ax4kkklkabj4ag2c@alap3.anarazel.de>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20220309181254.ax4kkklkabj4ag2c@alap3.anarazel.de>
User-Agent: Mutt/1.9.4 (2018-02-28)
Archived-At: 
 <https://www.postgresql.org/message-id/20220309203731.GB28503%40telsasoft.com>
Precedence: bulk

On Wed, Mar 09, 2022 at 10:12:54AM -0800, Andres Freund wrote:
> On 2022-03-09 11:47:23 -0600, Justin Pryzby wrote:
> > I'm curious what you think of this patch.
> > 
> > It makes check-world on freebsd over 30% faster - saving 5min.
> 
> That's nice! While -Og makes interactive debugging noticeably harder IME, it's
> not likely to be a large enough difference just for backtraces etc.

Yeah.  gcc(1) claims that -Og can improve debugging.

I should've mentioned that this seems to mitigate the performance effect of
--coverage on linux, too.

> I'm far less convinced that using "MaxSpeed" for the msvc build is a good
> idea. I've looked at one or two backtraces of optimized msvc builds and
> backtraces were quite a bit worse - and they're not great to start with.  What
> was the win there?

Did you compare FULL optimization or "favor speed/size" or "default"
optimization ?

It's worth trading some some build time (especially with a compiler cache) for
test time (especially with alltaptests).  But I didn't check backtraces, and I
didn't compare the various optimization options.  The argument may not be as
strong for windows, since it has no build cache (and it has no -Og).  We'd save
a bit more when also running the other tap tests.

CI runs are probably not very consistent, but I've just run
https://cirrus-ci.com/task/5236145167532032
master is the average of 4 patches at the top of cfbot.

                / master / patched / change
subscription    / 197s   / 195s    / +2s
recovery        / 234s   / 212s    / -22s
bin             / 383s   / 373s    / -10s

-- 
Justin