<html>
<head>
<meta content="text/html; charset=utf-8" http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
On 7/11/2014 11:20 AM, Patrick Cloke wrote:<br>
<span style="white-space: pre;">> Looking over posts on the forum
a bit it seems to be pretty much<br>
> Thunderbird:<br>
> <br>
> <a class="moz-txt-link-freetext" href="http://forum.palemoon.org/viewtopic.php?f=35&t=4130">http://forum.palemoon.org/viewtopic.php?f=35&t=4130</a>:
"Fossamail comes<br>
> with different defaults (both UI and under-the-hood) and has
been<br>
> rebuilt with a newer compiler and better optimizations than<br>
> Thunderbird. Otherwise, there aren't many differences. At
least not<br>
> at this time."<br>
> <br>
> <a class="moz-txt-link-freetext" href="http://forum.palemoon.org/viewtopic.php?f=35&t=3642">http://forum.palemoon.org/viewtopic.php?f=35&t=3642</a>
pretty much talks<br>
> in circles that it's "optimized" but won't say how.</span><br>
<br>
My recollection is that "optimized" in this builds usually means:<br>
* Use a different, newer compiler (maybe even ICC?)<br>
* 64-bit builds instead of 32-bit builds<br>
* Tuning options for relatively newer ISAs (e.g., SSE3 or SSSE3,
etc.)<br>
<br>
At least one the first one of these "optimized" rebuilds of Firefox
came out, it was very much well-known within Firefox that our 64-bit
JIT sucked more than the 32-bit JIT, which made the "optimized"
claim fairly laughable.<br>
<br>
My personal take on this sort of "optimization" is that it's
performance tuning by people who know nothing about performance
tuning and are just parroting effectively marketing material without
understanding it: amd64 isn't necessarily faster than ia32
(especially since increased pointer sizes stress caches a lot more),
and excessive flag combinations on your builds can very easily slow
your builds down because you lost the lottery. Seeing as how we
don't have any performance metrics, and I'm willing to bet whoever
built this didn't try to come up with one [let alone create one that
actually carries validity--not an easy task even if you're an
expert], I'm very skeptical that there's anything actually useful
here.<br>
<br>
-- <br>
Joshua Cranmer<br>
Thunderbird and DXR developer<br>
Source code archæologist<br>
<br>
</body>
</html>