<html><head><meta http-equiv="Content-Type" content="text/html charset=windows-1252"><base href="x-msg://1188/"></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space; ">The patch looks fine to me. However, I don't like the generic names of the functions AddRequiredExtension, SkipLoadExtension, and SkipExtension. These functions are only used for VMULL lowering and should be named as such.<div><br></div><div>Thanks,</div><div><br></div><div>Evan<div><br><div><div>On Nov 15, 2012, at 10:47 AM, David Peixotto <<a href="mailto:dpeixott@codeaurora.org">dpeixott@codeaurora.org</a>> wrote:</div><br class="Apple-interchange-newline"><blockquote type="cite"><div lang="EN-US" link="blue" vlink="purple" style="font-family: Helvetica; font-size: medium; font-style: normal; font-variant: normal; font-weight: normal; letter-spacing: normal; line-height: normal; orphans: 2; text-align: -webkit-auto; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-size-adjust: auto; -webkit-text-stroke-width: 0px; "><div class="WordSection1" style="page: WordSection1; "><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">Thanks for the review, Pete. I updated the patch based on your comments and attached the new version.<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">Reviewers, please send any other feedback.<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">--<span class="Apple-converted-space"> </span></span><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span></div><div style="border-style: none none none solid; border-left-width: 1.5pt; border-left-color: blue; padding: 0in 0in 0in 4pt; "><div><div style="border-style: solid none none; border-top-width: 1pt; border-top-color: rgb(181, 196, 223); padding: 3pt 0in 0in; "><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><b><span style="font-size: 10pt; font-family: Tahoma, sans-serif; ">From:</span></b><span style="font-size: 10pt; font-family: Tahoma, sans-serif; "><span class="Apple-converted-space"> </span>Pete Couperus [mailto:pjcoup@<a href="http://gmail.com" style="color: purple; text-decoration: underline; ">gmail.com</a>]<span class="Apple-converted-space"> </span><br><b>Sent:</b><span class="Apple-converted-space"> </span>Wednesday, November 14, 2012 10:21 PM<br><b>To:</b><span class="Apple-converted-space"> </span>David Peixotto<br><b>Cc:</b><span class="Apple-converted-space"> </span><a href="mailto:llvm-commits@cs.uiuc.edu" style="color: purple; text-decoration: underline; ">llvm-commits@cs.uiuc.edu</a><br><b>Subject:</b><span class="Apple-converted-space"> </span>Re: [llvm-commits] [PATCH] ARM: Allow long extensions to combine into vmull<o:p></o:p></span></div></div></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><o:p> </o:p></div><p class="MsoNormal" style="margin: 0in 0in 12pt; font-size: 12pt; font-family: 'Times New Roman', serif; ">Hi David,<br><br>This mostly looks fine to me.<br>Minor comments:<br>1. The body in your SkipLoadExtension is overindented (4 spaces, should be 2).<br>2. I'd reverse the conditional in AddRequiredExtension from:<br> if (OrigTy.getSizeInBits() < 64) {<br> [stuff]<br> }<br> return N;<br>to:<br> if (OrigTy.is64BitVector())<br> return N;<br> [stuff]<br><br>Maybe 2 is personal preference, but I think it reads easier this way.<br>Otherwise, I think this is good.<br><br>Pete<o:p></o:p></p><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; ">On Wed, Nov 14, 2012 at 5:06 PM, David Peixotto <<a href="mailto:dpeixott@codeaurora.org" target="_blank" style="color: purple; text-decoration: underline; ">dpeixott@codeaurora.org</a>> wrote:<o:p></o:p></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">Hi Pete,</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">As long as we are fixing it I think it makes sense to handle both cases (bare sign/zero extend and sign/zero load extend). If you have no objections, let’s move ahead with my patch since it is already factored that way.</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">I’ve attached an updated version that I rebased against the current head (contents are the same as my original patch). Please take a look and let me know what you think and we can work together to get it merged.</span><o:p></o:p></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">-David</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">-- Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div></div><div style="border-style: none none none solid; border-left-width: 1.5pt; border-left-color: blue; padding: 0in 0in 0in 4pt; "><div><div style="border-style: solid none none; border-top-width: 1pt; border-top-color: rgb(181, 196, 223); padding: 3pt 0in 0in; "><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><b><span style="font-size: 10pt; font-family: Tahoma, sans-serif; ">From:</span></b><span style="font-size: 10pt; font-family: Tahoma, sans-serif; "><span class="Apple-converted-space"> </span>Pete Couperus [mailto:<a href="mailto:pjcoup@gmail.com" target="_blank" style="color: purple; text-decoration: underline; ">pjcoup@gmail.com</a>]<span class="Apple-converted-space"> </span><br><b>Sent:</b><span class="Apple-converted-space"> </span>Tuesday, November 13, 2012 10:09 PM</span><o:p></o:p></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><br><b>To:</b><span class="Apple-converted-space"> </span>David Peixotto<br><b>Cc:</b><span class="Apple-converted-space"> </span><a href="mailto:llvm-commits@cs.uiuc.edu" target="_blank" style="color: purple; text-decoration: underline; ">llvm-commits@cs.uiuc.edu</a><br><b>Subject:</b><span class="Apple-converted-space"> </span>Re: [llvm-commits] [PATCH] ARM: Allow long extensions to combine into vmull<o:p></o:p></div></div></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "> <o:p></o:p></div><p class="MsoNormal" style="margin: 0in 0in 12pt; font-size: 12pt; font-family: 'Times New Roman', serif; ">Hi David,<br><br>I had basically aimed to minimize the changes, to address one problem. I agree with you, and think there are potential problems in this area...zext vs. sext, and SIGN/ZERO_EXTEND vs. [SZ]EXTLOAD. Without the long sign/zero extension change, we'll still trigger the outer assertion...but just writing that makes me feel that we might as well put that logic in as well. Since that is essentially your patch, want to have another go? Or, would you like me to fold it into mine? I'll try and be pleasantly persistent to get either of them in :).<br><br>Pete<o:p></o:p></p><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; ">On Tue, Nov 13, 2012 at 1:33 PM, David Peixotto <<a href="mailto:dpeixott@codeaurora.org" target="_blank" style="color: purple; text-decoration: underline; ">dpeixott@codeaurora.org</a>> wrote:<o:p></o:p></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">Hi Pete,</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">Thanks for updating your patch.</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">I don’t think I was able to trigger the long sign/zero extension with a simple test case either, although I don’t know of a reason why it would not be possible to trigger that path. If we don’t want to put the full logic in to support the possibility how about we add a comment and an assertion to verify that the underlying node is a 64-bit vector. That way if our assumption is wrong we can quickly identify and fix the issue.</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">-David</span><o:p></o:p></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">-- Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div></div><div style="border-style: none none none solid; border-left-width: 1.5pt; border-left-color: blue; padding: 0in 0in 0in 4pt; "><div><div style="border-style: solid none none; border-top-width: 1pt; border-top-color: rgb(181, 196, 223); padding: 3pt 0in 0in; "><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><b><span style="font-size: 10pt; font-family: Tahoma, sans-serif; ">From:</span></b><span style="font-size: 10pt; font-family: Tahoma, sans-serif; "><span class="Apple-converted-space"> </span>Pete Couperus [mailto:<a href="mailto:pjcoup@gmail.com" target="_blank" style="color: purple; text-decoration: underline; ">pjcoup@gmail.com</a>]<span class="Apple-converted-space"> </span><br><b>Sent:</b><span class="Apple-converted-space"> </span>Monday, November 12, 2012 10:02 PM<br><b>To:</b><span class="Apple-converted-space"> </span>David Peixotto<br><b>Cc:</b><span class="Apple-converted-space"> </span><a href="mailto:llvm-commits@cs.uiuc.edu" target="_blank" style="color: purple; text-decoration: underline; ">llvm-commits@cs.uiuc.edu</a><br><b>Subject:</b><span class="Apple-converted-space"> </span>Re: [llvm-commits] [PATCH] ARM: Allow long extensions to combine into vmull</span><o:p></o:p></div></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "> <o:p></o:p></div><p class="MsoNormal" style="margin: 0in 0in 12pt; font-size: 12pt; font-family: 'Times New Roman', serif; ">Hi David,<br><br>Thanks for taking a look.<br>My apologies, I hadn't seen your patch when I'd written this up.<br>And, you're right, I neglected the v2i8 case.<br>I originally had logic in for handling long sign/zero extension as well, but I couldn't trigger that with any of the basic test cases I thought of.<br>Curiously, the unsigned/long zero extension cases work in 3.1, but the codegen for unsigned v2i16->v2i64 followed by mul is substantially worse than for the signed case.<br>All the test cases that you sent pass with this new patch (allowing the v2i8->v2i64 extension). <span class="Apple-converted-space"> </span><br>Perhaps someone else wants to take a look :).<br><br>Pete<o:p></o:p></p><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; ">On Mon, Nov 12, 2012 at 10:00 AM, David Peixotto <<a href="mailto:dpeixott@codeaurora.org" target="_blank" style="color: purple; text-decoration: underline; ">dpeixott@codeaurora.org</a>> wrote:<o:p></o:p></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">Hi Pete,</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">I sent a patch for this a while back, but never got any feedback and failed to push it all the way through. I’ve attached my patch for reference.</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">It looks like the patches are fairly similar in their intent. I have a few more test cases, so please include those in your patch as well. It would also be good to include some of the comments, particularly the SkipExtension comment that explains the motivation for that function.</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">The main differences between the patches that I see are:</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div><p style="margin-right: 0in; margin-left: 0in; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">1)</span><span style="font-size: 7pt; color: rgb(31, 73, 125); "> <span class="Apple-converted-space"> </span></span><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">Handling a sign extend that is not large enough</span><o:p></o:p></p><p style="margin-right: 0in; margin-left: 0in; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">2)</span><span style="font-size: 7pt; color: rgb(31, 73, 125); "> <span class="Apple-converted-space"> </span></span><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">Handling v2i8 vector size</span><o:p></o:p></p><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">I’ll expand on these points below.</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">Index: lib/Target/ARM/ARMISelLowering.cpp</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">===================================================================</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">--- lib/Target/ARM/ARMISelLowering.cpp (revision 167684)</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+++ lib/Target/ARM/ARMISelLowering.cpp (working copy)</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">@@ -4922,11 +4922,30 @@</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">static SDValue SkipExtension(SDNode *N, SelectionDAG &DAG) {</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> if (N->getOpcode() == ISD::SIGN_EXTEND || N->getOpcode() == ISD::ZERO_EXTEND)</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> return N->getOperand(0);</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">What it N is a small vector here? Don’t we need to make sure that we have a sign/zero extend that will extend the size to 64-bit vector?</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">- if (LoadSDNode *LD = dyn_cast<LoadSDNode>(N))</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">- return DAG.getLoad(LD->getMemoryVT(), N->getDebugLoc(), LD->getChain(),</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">- LD->getBasePtr(), LD->getPointerInfo(), LD->isVolatile(),</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">- LD->isNonTemporal(), LD->isInvariant(),</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">- LD->getAlignment());</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+ if (LoadSDNode *LD = dyn_cast<LoadSDNode>(N)) {</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+ SDValue BareLD = DAG.getLoad(LD->getMemoryVT(), N->getDebugLoc(),</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+ LD->getChain(), LD->getBasePtr(),</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+ LD->getPointerInfo(), LD->isVolatile(),</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+ LD->isNonTemporal(), LD->isInvariant(),</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+ LD->getAlignment());</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+ if (BareLD.getValueType().is64BitVector())</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+ return BareLD;</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+ // This is a long extension, v4i8 -> v4i32 or v2i16 -> v2i64</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+ MVT ExtVT;</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+ if (LD->getMemoryVT() == MVT::v4i8)</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+ ExtVT = MVT::v4i16;</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+ else if (LD->getMemoryVT() == MVT::v2i16)</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+ ExtVT = MVT::v2i32;</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">What about MVT::v2i8. We can handle that case similar to v2i16 by extending the size to v2i32.</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+ if (LD->getExtensionType() == ISD::SEXTLOAD)</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+ return DAG.getNode(ISD::SIGN_EXTEND, BareLD.getDebugLoc(), ExtVT, BareLD);</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+ else if (LD->getExtensionType() == ISD::ZEXTLOAD)</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+ return DAG.getNode(ISD::ZERO_EXTEND, BareLD.getDebugLoc(), ExtVT, BareLD);</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+ llvm_unreachable("Unexpected extended load type for SkipExtension.");</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">+ }</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> // Otherwise, the value must be a BUILD_VECTOR. For v2i64, it will</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> // have been legalized as a BITCAST from v4i32.</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> if (N->getOpcode() == ISD::BITCAST) {</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); ">-- Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation</span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125); "> </span><o:p></o:p></div><div style="border-style: none none none solid; border-left-width: 1.5pt; border-left-color: blue; padding: 0in 0in 0in 4pt; "><div><div style="border-style: solid none none; border-top-width: 1pt; border-top-color: rgb(181, 196, 223); padding: 3pt 0in 0in; "><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><b><span style="font-size: 10pt; font-family: Tahoma, sans-serif; ">From:</span></b><span style="font-size: 10pt; font-family: Tahoma, sans-serif; "><span class="Apple-converted-space"> </span><a href="mailto:llvm-commits-bounces@cs.uiuc.edu" target="_blank" style="color: purple; text-decoration: underline; ">llvm-commits-bounces@cs.uiuc.edu</a><span class="Apple-converted-space"> </span>[mailto:<a href="mailto:llvm-commits-bounces@cs.uiuc.edu" target="_blank" style="color: purple; text-decoration: underline; ">llvm-commits-bounces@cs.uiuc.edu</a>]<span class="Apple-converted-space"> </span><b>On Behalf Of<span class="Apple-converted-space"> </span></b>Pete Couperus<br><b>Sent:</b><span class="Apple-converted-space"> </span>Saturday, November 10, 2012 11:48 PM<br><b>To:</b><span class="Apple-converted-space"> </span><a href="mailto:llvm-commits@cs.uiuc.edu" target="_blank" style="color: purple; text-decoration: underline; ">llvm-commits@cs.uiuc.edu</a><br><b>Subject:</b><span class="Apple-converted-space"> </span>[llvm-commits] [PATCH] ARM: Allow long extensions to combine into vmull</span><o:p></o:p></div></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "> <o:p></o:p></div><p class="MsoNormal" style="margin: 0in 0in 12pt; font-size: 12pt; font-family: 'Times New Roman', serif; ">Hello,<br><br>The ARM backend currently crashes when trying to combine:<br> %tmp1 = load <4 x i8>* %A<br> %tmp2 = load <4 x i8>* %B<br> %tmp3 = sext <4 x i8> %tmp1 to <4 x i32><br> %tmp4 = sext <4 x i8> %tmp2 to <4 x i32><br> %tmp5 = mul <4 x i32> %tmp3, %tmp4<br><br>into a vmull. This was reported as PR 12281.<br>This patch fixes this problem.<br>Please review!<br>Thanks,<br><br>Pete<o:p></o:p></p></div></div></div></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "> <o:p></o:p></div></div></div></div></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "> <o:p></o:p></div></div></div></div></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; "><o:p> </o:p></div></div></div><span><0001-Codegen-failure-for-vmull-with-small-vectors.patch></span>_______________________________________________<br>llvm-commits mailing list<br><a href="mailto:llvm-commits@cs.uiuc.edu" style="color: purple; text-decoration: underline; ">llvm-commits@cs.uiuc.edu</a><br><a href="http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits" style="color: purple; text-decoration: underline; ">http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits</a></div></blockquote></div><br></div></div></body></html>