<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">On 07/29/2013 09:24 AM, Eli Bendersky
wrote:<br>
</div>
<blockquote
cite="mid:CACLQwhGnPYCn=UbjZCbcdM33rQ8mnjkyQpNqk4T9tFmBPBc_dA@mail.gmail.com"
type="cite">
<meta http-equiv="Content-Type" content="text/html;
charset=ISO-8859-1">
<div dir="ltr"><br>
<div class="gmail_extra"><br>
<br>
<div class="gmail_quote">On Sun, Jul 28, 2013 at 3:02 PM, Reed
Kotler <span dir="ltr"><<a moz-do-not-send="true"
href="mailto:rkotler@mips.com" target="_blank">rkotler@mips.com</a>></span>
wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0
.8ex;border-left:1px #ccc solid;padding-left:1ex">Ooops...
Ignore this previous mail.<br>
<br>
The problem still exists with this change.
<div class="HOEnZb">
<div class="h5"><br>
<br>
On 07/28/2013 02:36 PM, reed kotler wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0
.8ex;border-left:1px #ccc solid;padding-left:1ex">
It seems that several intrinsics are missing from
this routine.<br>
<br>
In particular, floor, which was causing problems in
the mips16 port.<br>
<br>
Is there some reason to not add the ones that are
missing?<br>
<br>
For example, adding the following fixed my problem
with floor.<br>
<br>
case Intrinsic::floor:<br>
EnsureFPIntrinsicsExist(M, I, "floorf",
"floor", "floor");<br>
break;<br>
</blockquote>
<br>
</div>
</div>
</blockquote>
<div><br>
</div>
<div>Note that this code is being used mainly (only?) by the
Interpreter.</div>
<div><br>
</div>
<div>Eli</div>
<div> </div>
</div>
</div>
</div>
</blockquote>
Yes.<br>
<br>
I don't know the source of this problem with some intrinsics being
declared improperly.<br>
Sin, cos are okay but Floor and trunc not.<br>
<br>
I have a workaround that I will implement today.<br>
<br>
For mips16 pic mode I need to make calls to helper functions for
functions with certain prototype.<br>
<br>
I intend to move this logic to an earlier place where I am operating
on clang ir but for now it is being handled during call lowering and
I rely on Args and RetTy being correct.<br>
<br>
<br>
<br>
</body>
</html>