<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body text="#000000" bgcolor="#FFFFFF">
<div class="moz-cite-prefix">Am 26.02.2013 03:33, schrieb Sean
Silva:<br>
</div>
<blockquote
cite="mid:CAHnXoa=pk7MisjB6fyW6wjq=hCv1TEz8bJ0VnQu3rxV-4sVE=A@mail.gmail.com"
type="cite">
<div dir="ltr">On Mon, Feb 25, 2013 at 3:00 PM, Tom Stellard <span
dir="ltr"><<a moz-do-not-send="true"
href="mailto:tom@stellard.net" target="_blank">tom@stellard.net</a>></span>
wrote:<br>
<div class="gmail_extra">
<div class="gmail_quote">
<blockquote class="gmail_quote" style="margin:0px 0px 0px
0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex">Can
you clarify this? Do you mean one email with with all the
patches<br>
as attachments?</blockquote>
</div>
<br>
</div>
<div class="gmail_extra" style="">That would be preferable. Note
that gmail (which many of us use) doesn't respect threading
(instead it arbitrarily threads based on subject for some
reason), so having them all as replies will still flood the
inbox of gmail users.</div>
<div class="gmail_extra" style=""><br>
</div>
</div>
</blockquote>
<br>
That's the exactly opposite policy as many other open source
development list use, the argument pro in-lining patches is usually
that the code is easier to comment on. But personally I usually also
prefer some form of one mail (or one thread) per patchset for
review. <br>
<br>
<blockquote
cite="mid:CAHnXoa=pk7MisjB6fyW6wjq=hCv1TEz8bJ0VnQu3rxV-4sVE=A@mail.gmail.com"
type="cite">
<div dir="ltr">
<div class="gmail_extra" style="">However, it is much more in
line with LLVM development style to send in patches
incrementally for review rather than dumping a whole branch;
that ensures focused review.</div>
</div>
</blockquote>
<br>
Those patches belong together and either implement new features or
fix bugs that were triggered/found while implementing those new
features. Reviewing them separately would just increase the chance
of missing something regarding inter patch dependency.<br>
<br>
<blockquote
cite="mid:CAHnXoa=pk7MisjB6fyW6wjq=hCv1TEz8bJ0VnQu3rxV-4sVE=A@mail.gmail.com"
type="cite">
<div dir="ltr">
<div class="gmail_extra" style="">(I'll also reiterate how
troubling it is that none of these R600 changes have unit
tests).<br>
</div>
</div>
</blockquote>
<br>
Totally agree on that, the major problem seems to be that currently
the instruction scheduler and/or register allocator produce
different results when compiling the same code twice. The results
seems to be always correct, it's just not very well comparable. That
fact makes it quite difficult for me to write profound unit tests.
Any ideas on that?<br>
<br>
Christian.<br>
</body>
</html>