<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
<br>
<div class="moz-cite-prefix">On 02/19/2014 02:02 PM, Chandler
Carruth wrote:<br>
</div>
<blockquote
cite="mid:CAGCO0Kj8PF2aWHF-f6ycVoZ-1N9wNdooF_1BpFhkOE1iUx4c1A@mail.gmail.com"
type="cite">
<div dir="ltr">
<div class="gmail_extra"><br>
<div class="gmail_quote">On Wed, Feb 19, 2014 at 1:43 PM, Hal
Finkel <span dir="ltr"><<a moz-do-not-send="true"
href="mailto:hfinkel@anl.gov" target="_blank">hfinkel@anl.gov</a>></span>
wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0
.8ex;border-left:1px #ccc solid;padding-left:1ex">It seems
to me that IR optimizations that depend directly on
TargetLowering belong in CodeGen. Why don't you just move
them all there?</blockquote>
</div>
<br>
Some of them *could* be profitably lifted to use a more
abstract interface like TTI. I'm pretty sure CGP is on the
"not worth it" side of this equation, and it may not be the
only thing on that side, but I'm trying to leave open the idea
that there is some inflection point here and we may have
passes on both sides of it that currently are still misusing
TargetLowering and in the IR pass libraries.</div>
</div>
</blockquote>
Well said.<br>
<br>
If we do move a bunch of things into CodeGen, should we create a
subdirectory?<br>
<br>
Philip<br>
</body>
</html>