<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
</head>
<body text="#000000" bgcolor="#FFFFFF">
<p>There is less calculation, but the noise matrix takes up cache
space (about half a modern CPU's L1) and hence increases memory
pressure. The actual performance relationship is probably
hardware dependent, but on my laptop doing a quick check with
patched lowlevel-blt-bench:<br>
</p>
<p> - Bayer is ~8% faster than blue noise<br>
- Bayer is only ~5% faster if I use `dither_blue_noise_64x64[((y
& 0x3f) << 6) | (x & 0x3f)]` in the index
computation, which I would have expected GCC to do automatically
but oh well.<br>
- Both are about the same speed with uint16_t array for the blue
noise (halves memory consumption).<br>
</p>
<p>I will make those two changes (they are clearly desirable) and
re-post the blue noise patch. Since this is in a write-back loop
which should be memory-bound in general, I think it makes sense to
keep Bayer as FAST dithering -- but I'm not opposed to removing
it.</p>
<p>- Basile<br>
</p>
On 4/22/19 8:12 PM, Bill Spitzak wrote:<br>
<blockquote type="cite"
cite="mid:CAL-8oAgpn1EdGfGK4M1eUqH_uOWjsGRGe8jY1KOu0nQFP7cDRg@mail.gmail.com">
<meta http-equiv="content-type" content="text/html; charset=UTF-8">
<div dir="ltr">Is the "blue noise" version actually slower than
the Bayer? Seems to be doing a bit less calculation but it
reading the array of weights. The removal of the need for the
pattern offset seems like a win.</div>
<br>
<div class="gmail_quote">
<div dir="ltr" class="gmail_attr">On Mon, Apr 22, 2019 at 9:27
AM Matt Turner <<a href="mailto:mattst88@gmail.com"
moz-do-not-send="true">mattst88@gmail.com</a>> wrote:<br>
</div>
<blockquote class="gmail_quote" style="margin:0px 0px 0px
0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">On
Fri, Apr 19, 2019 at 4:52 PM Bryce Harrington<br>
<<a href="mailto:bryce@bryceharrington.org" target="_blank"
moz-do-not-send="true">bryce@bryceharrington.org</a>>
wrote:<br>
> Inkscape would love to see Basile's dithering patches
included. Our<br>
> testing shows that they make a huge quality difference
for our users;<br>
> this solves a critical need.<br>
><br>
> Mc and I have done some preliminary investigation into
how to plumb this<br>
> into Cairo, and would love to hear your review of
Basile's approach to<br>
> the problem.<br>
<br>
I don't feel like I'm experienced enough with that side of
pixman to<br>
offer meaningful comments. I've Cc'd Søren in the hopes that
he<br>
remains interested enough in the project to review the patches
that<br>
Basile says implement the approach Søren described.<br>
_______________________________________________<br>
Pixman mailing list<br>
<a href="mailto:Pixman@lists.freedesktop.org" target="_blank"
moz-do-not-send="true">Pixman@lists.freedesktop.org</a><br>
<a
href="https://lists.freedesktop.org/mailman/listinfo/pixman"
rel="noreferrer" target="_blank" moz-do-not-send="true">https://lists.freedesktop.org/mailman/listinfo/pixman</a></blockquote>
</div>
<br>
<fieldset class="mimeAttachmentHeader"></fieldset>
<pre class="moz-quote-pre" wrap="">_______________________________________________
Pixman mailing list
<a class="moz-txt-link-abbreviated" href="mailto:Pixman@lists.freedesktop.org">Pixman@lists.freedesktop.org</a>
<a class="moz-txt-link-freetext" href="https://lists.freedesktop.org/mailman/listinfo/pixman">https://lists.freedesktop.org/mailman/listinfo/pixman</a></pre>
</blockquote>
</body>
</html>