X-Authentication-Warning: delorie.com: mail set sender to geda-user-bounces using -f
X-Recipient: geda-user AT delorie DOT com
Date: Thu, 17 May 2018 09:01:37 +0200 (CEST)
X-X-Sender: igor2 AT igor2priv
To: "michalwd1979 (michalwd1979 AT o2 DOT pl) [via geda-user AT delorie DOT com]" <geda-user AT delorie DOT com>
X-Debug: to=geda-user AT delorie DOT com from="gedau AT igor2 DOT repo DOT hu"
From: gedau AT igor2 DOT repo DOT hu
Subject: Re: Odp: Re: [geda-user] Opengl PCB and mainline PCB - pcb-rnd
 aspects
In-Reply-To: <bb691d68061d461287579b59351bf976@grupawp.pl>
Message-ID: <alpine.DEB.2.00.1805170850570.8169@igor2priv>
References: <ca+qhd=-z1tvzehozmmi+8eduqnwmaeeeodg=uk3p6mb9tj57dq AT mail DOT gmail DOT com> <647dca2ad89a4415ad980da6e5cdc701 AT grupawp DOT pl> <cajzxidbrzvoaivzxdhdcehdmubu-a5c8ueaeud6q-r0prvibya AT mail DOT gmail DOT com> <d44c1d9475c440a09121d3247c43b1d1 AT grupawp DOT pl>
 <alpine DOT deb DOT 2 DOT 00 DOT 1804301857010 DOT 19825 AT igor2priv> <7da892c189bd49838d6ce6eb2c2628e4 AT grupawp DOT pl> <alpine DOT deb DOT 2 DOT 00 DOT 1805042003061 DOT 19825 AT igor2priv> <b84c48bc6c46413e809346a1a0baad2c AT grupawp DOT pl> <cam2rghspcci+g9vxjvtjwtilbj9ebegskaz8spcn=l5jz-zqtg AT mail DOT gmail DOT com>
 <7e30777e38284644814271a68f2c2119 AT grupawp DOT pl> <alpine DOT deb DOT 2 DOT 00 DOT 1805080715580 DOT 8169 AT igor2priv> <00430c5cbe794d57918e5e9c532d436f AT grupawp DOT pl> <alpine DOT deb DOT 2 DOT 00 DOT 1805110914410 DOT 8169 AT igor2priv> <e537627331744537b6e5743d98f907a2 AT grupawp DOT pl> <alpine DOT DEB DOT 2 DOT 00 DOT 1805151534520 DOT 8169 AT igor2priv>
 <bb691d68061d461287579b59351bf976 AT grupawp DOT pl>
User-Agent: Alpine 2.00 (DEB 1167 2008-08-23)
MIME-Version: 1.0
Content-Type: MULTIPART/MIXED; BOUNDARY="0-619124716-1526540497=:8169"
Reply-To: geda-user AT delorie DOT com
Errors-To: nobody AT delorie DOT com
X-Mailing-List: geda-user AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

  This message is in MIME format.  The first part should be readable text,
  while the remaining parts are likely unreadable without MIME-aware tools.

--0-619124716-1526540497=:8169
Content-Type: TEXT/PLAIN; charset=UTF-8; format=flowed
Content-Transfer-Encoding: QUOTED-PRINTABLE

Hello Michael,

On Wed, 16 May 2018, michalwd1979 (michalwd1979 AT o2 DOT pl) [via geda-user AT delor=
ie.com] wrote:

>Hello Igor,
>
>You were right, somehow Yours mails end up in "spam" - I'm trying to fix
>"poczta.o2"=C2=A0 filters now, but it seems that spam is worth checking fo=
rm time
>to time. Anyway:

Modern day email and spam filtering, lol.

>I'm using svn myself, so working with it is not a problem. With fresh
>compiled pcb-rnd, 1024 resistors board and full screen window I have:
>dash_freq=3D0, benchmark 15.6 (15.2 to 16.1)
>dash_freq=3D1, benchmark 15.0 (14.3 to 15.6)
>dash_freq=3D5, benchmark 14.3 (13.5 to 14.9)
>dash_freq=3D-1, benchmark 16.5 (15.8 to 17)
>It seems that GUI does not let me set dash_freq to -1 (resets it to 0 -
>maybe wrong limits?), I had to use -c appearance/.... to get it right.

Thanks, this means the main problem is not the dashed line or the number=20
of lines drawn by that mechanism.

<snip>

>When comparing Yours 1024 resistors board to my (with only ~320 elements)
>I've noticed that Yours board do not have refdes numbers printed. Maybe th=
is
>is the main factor? On by board turning subcircuits ON/OFF and also

Yup, that's the next we are going to test. I deliberately omitted them so=
=20
we could get a clean measure of the subc dashed line bbox performance=20
without text render interfering.

>Subc.parts means massive difference in speed. It seems that these two laye=
rs
>also control displaying element names (black and red). If=C2=A0 I change
>view->Displayed subcircuits ID from "refdes" to "refdes+value" speed drops
>about 2x, so it seems that amount of printed text is really important.

Thanks, that also suggests it's the text rendering that makes it slow. To=
=20
get a clean measurement, please use the following test file:

http://igor2.repo.hu/tmp/refdes.lht.gz

Please turn off the padstack "layer" in the layer selector so we don't=20
render them, then do the usual fps test (big window, zoom to extent,=20
benchmark) with two different setup:

1. subcircuits "layer"  off, but top silk on - this way we measure how=20
much does it cost to render all the refdes text objects on silk

2. subcircuits "layer" on, but top silk off - this way we render about the=
=20
same thing, just by a different part of the code

I crafted this file to have the dash freq set to -1 so we don't measure=20
the dashed line rendering in (would be noise on the results).

If these results will show what I expect, I know where to poke the code.

Meanwhile I did some more testing with sw render and Ade did some more=20
testing with opengl render. We both identified parts of the rendering code=
=20
that could be optimized. We'll release pcb-rnd 2.0.0 in about 3 weeks, so=
=20
we don't start messing with the rendering now, but after the release we=20
will speed some parts up a bit.

Thanks for testing!

Best regards,

Igor2
--0-619124716-1526540497=:8169--