<div dir="ltr"><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"><div dir="ltr"><div>I've bisected this to PR24535, please try to take a look at it if you get a free moment.</div></div></blockquote><div><br></div><div>Looking...</div><div> </div><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"><div dir="ltr"><div>Given what Eli brought up, shouldn't we guard it's usage in 64-bit mode on a -mattr flag?  Materializing lahf or sahf without knowing apriori that a CPU can handle it seems problematic.<br></div></div></blockquote><div><br></div><div>See this discussion:</div></div></div><blockquote style="margin:0 0 0 40px;border:none;padding:0px"><div class="gmail_extra"><div class="gmail_quote"><div><a href="http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20150810/293071.html">http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20150810/293071.html</a></div></div></div></blockquote><div class="gmail_extra"><div class="gmail_quote"><div>We've been emitting lahf with impunity since 2012 without issues. We could fix it, but is it worth fixing at this point if it's never been a problem?</div></div></div></div>