<html><head><meta http-equiv="Content-Type" content="text/html charset=windows-1252"></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;">Hi Elena,<div><br><div><div>On Sep 3, 2014, at 5:30 AM, Demikhovsky, Elena <<a href="mailto:elena.demikhovsky@intel.com">elena.demikhovsky@intel.com</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: 14px; font-style: normal; font-variant: normal; font-weight: normal; letter-spacing: normal; line-height: normal; orphans: auto; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: auto; word-spacing: 0px; -webkit-text-stroke-width: 0px;"><div class="WordSection1" style="page: WordSection1;"><div style="margin: 0cm 0cm 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 wrote one more implementation with binary search on a sorted static table.<o:p></o:p></span></div><div style="margin: 0cm 0cm 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);">Please take a look.</span></div></div></div></blockquote><div><br></div><div>LGTM with one change to consider in addition to Juergen’s comments:</div><div><br></div><div><div>+static const IntrinsicData* GetIntrinsicWithoutChain(unsigned IntNo) {</div><div>+ assert(std::is_sorted(std::begin(IntrinsicsWithoutChain),</div><div>+ std::end(IntrinsicsWithoutChain)) &&</div><div>+ "Intrinsics data array should be sorted before search”);</div><div><br></div><div>We may want to move the asserts to resetOperationActions.</div><div><br></div><div>Thanks,</div><div>Adam</div></div><br><blockquote type="cite"><div lang="EN-US" link="blue" vlink="purple" style="font-family: Helvetica; font-size: 14px; font-style: normal; font-variant: normal; font-weight: normal; letter-spacing: normal; line-height: normal; orphans: auto; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: auto; word-spacing: 0px; -webkit-text-stroke-width: 0px;"><div class="WordSection1" style="page: WordSection1;"><div style="margin: 0cm 0cm 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);"><o:p></o:p></span></div><div style="margin: 0cm 0cm 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: 0cm 0cm 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.<o:p></o:p></span></div><div style="margin: 0cm 0cm 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><div style="margin: 0cm 0cm 0.0001pt 36pt; font-size: 12pt; font-family: 'Times New Roman', serif; text-indent: -18pt;"><span style="font-family: Calibri, sans-serif; color: rgb(49, 132, 155);"><span>-<span style="font-style: normal; font-variant: normal; font-weight: normal; font-size: 7pt; line-height: normal; font-family: 'Times New Roman';"> <span class="Apple-converted-space"> </span></span></span></span><span dir="LTR"></span><b><i><span style="color: rgb(49, 132, 155);"> Elena<o:p></o:p></span></i></b></div></div><div style="margin: 0cm 0cm 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><div style="border-style: solid none none; border-top-color: rgb(181, 196, 223); border-top-width: 1pt; padding: 3pt 0cm 0cm;"><div style="margin: 0cm 0cm 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>Demikhovsky, Elena<span class="Apple-converted-space"> </span><br><b>Sent:</b><span class="Apple-converted-space"> </span>Wednesday, September 03, 2014 09:16<br><b>To:</b><span class="Apple-converted-space"> </span>Juergen Ributzka<br><b>Cc:</b><span class="Apple-converted-space"> </span>Sean Silva; David Blaikie; LLVM Commits<br><b>Subject:</b><span class="Apple-converted-space"> </span>RE: [llvm] r216345 - X86 intrinsics table - simplifies intrinsics lowering.<o:p></o:p></span></div></div></div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></div><div style="margin: 0cm 0cm 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);">But this patch solves the problem. I removed globals from the header file. Can I commit it or I should look for another solution?<o:p></o:p></span></div><div style="margin: 0cm 0cm 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><div style="margin: 0cm 0cm 0.0001pt 36pt; font-size: 12pt; font-family: 'Times New Roman', serif; text-indent: -18pt;"><span style="font-family: Calibri, sans-serif; color: rgb(49, 132, 155);"><span>-<span style="font-style: normal; font-variant: normal; font-weight: normal; font-size: 7pt; line-height: normal; font-family: 'Times New Roman';"> <span class="Apple-converted-space"> </span></span></span></span><span dir="LTR"></span><b><i><span style="color: rgb(49, 132, 155);"> Elena<o:p></o:p></span></i></b></div></div><div style="margin: 0cm 0cm 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><div style="border-style: solid none none; border-top-color: rgb(181, 196, 223); border-top-width: 1pt; padding: 3pt 0cm 0cm;"><div style="margin: 0cm 0cm 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>Juergen Ributzka [<a href="mailto:juergen@apple.com" style="color: purple; text-decoration: underline;">mailto:juergen@apple.com</a>]<span class="Apple-converted-space"> </span><br><b>Sent:</b><span class="Apple-converted-space"> </span>Tuesday, September 02, 2014 23:37<br><b>To:</b><span class="Apple-converted-space"> </span>Demikhovsky, Elena<br><b>Cc:</b><span class="Apple-converted-space"> </span>Sean Silva; David Blaikie; LLVM Commits<br><b>Subject:</b><span class="Apple-converted-space"> </span>Re: [llvm] r216345 - X86 intrinsics table - simplifies intrinsics lowering.<o:p></o:p></span></div></div></div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">Hi Elena,<o:p></o:p></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">this change broke concurrent compilation for us. I looks like the globals (in a header file???) IntrWithChainMap and IntrWithoutChainMap are not guarded against multiple and concurrent initialization.<o:p></o:p></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">Could you please take a look?<o:p></o:p></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">Thanks<o:p></o:p></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">Cheers,<o:p></o:p></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">Juergen<o:p></o:p></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></div><div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">On Aug 27, 2014, at 12:54 AM, Demikhovsky, Elena <<a href="mailto:elena.demikhovsky@intel.com" style="color: purple; text-decoration: underline;">elena.demikhovsky@intel.com</a>> wrote:<o:p></o:p></div></div><p class="MsoNormal" style="margin: 0cm 0cm 12pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></p><div><div><div style="margin: 0cm 0cm 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 put the intrinsics map as a static object inside static function that will be initialized on first call as Sean suggested.</span><o:p></o:p></div></div><div><div style="margin: 0cm 0cm 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><div style="margin: 0cm 0cm 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);">Please review.</span><o:p></o:p></div></div><div><div style="margin: 0cm 0cm 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="margin-left: 36pt;"><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif; text-indent: -18pt;"><span style="font-family: Calibri, sans-serif; color: rgb(49, 132, 155);">-</span><span style="font-size: 7pt; color: rgb(49, 132, 155);"> <span class="apple-converted-space"> </span></span><b><i><span style="color: rgb(49, 132, 155);"> Elena</span></i></b><o:p></o:p></div></div><div><div style="margin: 0cm 0cm 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><div style="margin: 0cm 0cm 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 class="apple-converted-space"><span style="font-size: 10pt; font-family: Tahoma, sans-serif;"> </span></span><span style="font-size: 10pt; font-family: Tahoma, sans-serif;">Sean Silva [<a href="mailto:chisophugis@gmail.com" style="color: purple; text-decoration: underline;"><span style="color: purple;">mailto:chisophugis@gmail.com</span></a>]<span class="apple-converted-space"> </span><br><b>Sent:</b><span class="apple-converted-space"> </span>Wednesday, August 27, 2014 03:18<br><b>To:</b><span class="apple-converted-space"> </span>David Blaikie<br><b>Cc:</b><span class="apple-converted-space"> </span>Demikhovsky, Elena;<span class="apple-converted-space"> </span><a href="mailto:llvm-commits@cs.uiuc.edu" style="color: purple; text-decoration: underline;"><span style="color: purple;">llvm-commits@cs.uiuc.edu</span></a><br><b>Subject:</b><span class="apple-converted-space"> </span>Re: [llvm] r216345 - X86 intrinsics table - simplifies intrinsics lowering.</span><o:p></o:p></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"> <o:p></o:p></div></div><div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"> <o:p></o:p></div></div><div><p class="MsoNormal" style="margin: 0cm 0cm 12pt; font-size: 12pt; font-family: 'Times New Roman', serif;"> <o:p></o:p></p><div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">On Tue, Aug 26, 2014 at 2:22 PM, David Blaikie <<a href="mailto:dblaikie@gmail.com" target="_blank" style="color: purple; text-decoration: underline;"><span style="color: purple;">dblaikie@gmail.com</span></a>> wrote:<o:p></o:p></div></div><div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"> <o:p></o:p></div></div><div><p class="MsoNormal" style="margin: 0cm 0cm 12pt; font-size: 12pt; font-family: 'Times New Roman', serif;"> <o:p></o:p></p><div><div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">On Tue, Aug 26, 2014 at 2:10 PM, Sean Silva <<a href="mailto:chisophugis@gmail.com" target="_blank" style="color: purple; text-decoration: underline;"><span style="color: purple;">chisophugis@gmail.com</span></a>> wrote:<o:p></o:p></div></div><div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"> <o:p></o:p></div></div><div><p class="MsoNormal" style="margin: 0cm 0cm 12pt; font-size: 12pt; font-family: 'Times New Roman', serif;"> <o:p></o:p></p><div><div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">On Tue, Aug 26, 2014 at 4:15 AM, Demikhovsky, Elena <<a href="mailto:elena.demikhovsky@intel.com" target="_blank" style="color: purple; text-decoration: underline;"><span style="color: purple;">elena.demikhovsky@intel.com</span></a>> wrote:<o:p></o:p></div></div><div><div><div style="margin: 0cm 0cm 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 initialize the map in static time, when X86Target object is created.</span><o:p></o:p></div></div><div><div style="margin: 0cm 0cm 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><div style="margin: 0cm 0cm 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 map initialization is supported in C++11, like</span><o:p></o:p></div></div><pre style="margin: 0cm 0cm 0.0001pt; font-size: 10pt; font-family: 'Courier New'; line-height: 13.35pt; background-color: rgb(238, 238, 238); vertical-align: baseline; background-position: initial initial; background-repeat: initial initial;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span><span style="font-size: 10.5pt; font-family: Consolas; color: rgb(43, 145, 175); border: 1pt none windowtext; padding: 0cm;">map</span><span style="font-size: 10.5pt; font-family: Consolas; border: 1pt none windowtext; padding: 0cm;"><<span style="color: rgb(43, 145, 175);">int</span>, <span style="color: rgb(0, 0, 139);">char</span>> m = {{<span style="color: maroon;">1</span>, <span style="color: maroon;">'a'</span>}, {<span style="color: maroon;">3</span>, <span style="color: maroon;">'b'</span>}, {<span style="color: maroon;">5</span>, <span style="color: maroon;">'c'</span>}, {<span style="color: maroon;">7</span>, <span style="color: maroon;">'d'</span>}};</span><o:p></o:p></pre><pre style="margin: 0cm 0cm 0.0001pt; font-size: 10pt; font-family: 'Courier New'; line-height: 13.35pt; background-color: rgb(238, 238, 238); vertical-align: baseline; background-position: initial initial; background-repeat: initial initial;"><span style="font-size: 10.5pt; font-family: Consolas; border: 1pt none windowtext; padding: 0cm;"> if LLVM strongly requires C++11, I can put static initialization.</span><o:p></o:p></pre><div><div style="margin: 0cm 0cm 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><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"> <o:p></o:p></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"> <o:p></o:p></div></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">I'm pretty sure that will still produce a static initializer called on startup. Please revert this until you can find a solution that doesn't require runtime initialization.<o:p></o:p></div></div></div></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"> <o:p></o:p></div></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">+1, this isn't cool.<br><br>I don't know if we're able to rely on multithreaded function-local static initialization (MSVC? - check the LLVM compiler feature compat page, perhaps) then you could just put this map in a static accessor function as a function-local static. That way it won't be a global constructor, just lazy initialized on first-call.<o:p></o:p></div></div></div></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"> <o:p></o:p></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">It looks like this is just constant integer data though. A binary-search static table seems preferable.<o:p></o:p></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"> <o:p></o:p></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">-- Sean Silva<o:p></o:p></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"> <o:p></o:p></div></div><blockquote style="border-style: none none none solid; border-left-color: rgb(204, 204, 204); border-left-width: 1pt; padding: 0cm 0cm 0cm 6pt; margin: 5pt 0cm 5pt 4.8pt;"><div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"> <o:p></o:p></div></div><blockquote style="border-style: none none none solid; border-left-color: rgb(204, 204, 204); border-left-width: 1pt; padding: 0cm 0cm 0cm 6pt; margin: 5pt 0cm 5pt 4.8pt;"><div><div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="color: rgb(136, 136, 136);"> </span><o:p></o:p></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="color: rgb(136, 136, 136);">-- Sean Silva</span><o:p></o:p></div></div><div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"> <o:p></o:p></div></div><blockquote style="border-style: none none none solid; border-left-color: rgb(204, 204, 204); border-left-width: 1pt; padding: 0cm 0cm 0cm 6pt; margin: 5pt 0cm 5pt 4.8pt;"><div><div><div style="margin: 0cm 0cm 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="margin-left: 36pt;"><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-family: Calibri, sans-serif; color: rgb(49, 132, 155);">-</span><span style="font-size: 7pt; color: rgb(49, 132, 155);"> <span class="apple-converted-space"> </span></span><b><i><span style="color: rgb(49, 132, 155);"> Elena</span></i></b><o:p></o:p></div></div><div><div style="margin: 0cm 0cm 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><div style="margin: 0cm 0cm 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 class="apple-converted-space"><span style="font-size: 10pt; font-family: Tahoma, sans-serif;"> </span></span><span style="font-size: 10pt; font-family: Tahoma, sans-serif;">Sean Silva [mailto:<a href="mailto:chisophugis@gmail.com" target="_blank" style="color: purple; text-decoration: underline;"><span style="color: purple;">chisophugis@gmail.com</span></a>]<span class="apple-converted-space"> </span><br><b>Sent:</b><span class="apple-converted-space"> </span>Monday, August 25, 2014 21:42<br><b>To:</b><span class="apple-converted-space"> </span>Demikhovsky, Elena<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;"><span style="color: purple;">llvm-commits@cs.uiuc.edu</span></a><br><b>Subject:</b><span class="apple-converted-space"> </span>Re: [llvm] r216345 - X86 intrinsics table - simplifies intrinsics lowering.</span><o:p></o:p></div></div><div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"> <o:p></o:p></div></div><div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 10pt; font-family: Arial, sans-serif;">+std::map < unsigned, IntrinsicData> IntrWithChainMap;</span><o:p></o:p></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"> <o:p></o:p></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">What's up with the random global (w/ static initializer)? We shouldn't have any of those in the library code.<o:p></o:p></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"> <o:p></o:p></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">-- Sean Silva<o:p></o:p></div></div></div><div><p class="MsoNormal" style="margin: 0cm 0cm 12pt; font-size: 12pt; font-family: 'Times New Roman', serif;"> <o:p></o:p></p><div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">On Sun, Aug 24, 2014 at 2:19 AM, Elena Demikhovsky <<a href="mailto:elena.demikhovsky@intel.com" target="_blank" style="color: purple; text-decoration: underline;"><span style="color: purple;">elena.demikhovsky@intel.com</span></a>> wrote:<o:p></o:p></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">Author: delena<br>Date: Sun Aug 24 04:19:56 2014<br>New Revision: 216345<br><br>URL:<span class="apple-converted-space"> </span><a href="http://llvm.org/viewvc/llvm-project?rev=216345&view=rev" target="_blank" style="color: purple; text-decoration: underline;"><span style="color: purple;">http://llvm.org/viewvc/llvm-project?rev=216345&view=rev</span></a><br>Log:<br>X86 intrinsics table - simplifies intrinsics lowering.<br>The tables are initialized when X86TargetLowering object is created.<br><br>Added:<br> llvm/trunk/lib/Target/X86/X86IntrinsicsInfo.h<br>Modified:<br> llvm/trunk/include/llvm/IR/IntrinsicsX86.td<br> llvm/trunk/lib/Target/X86/X86ISelLowering.cpp<br><br>Modified: llvm/trunk/include/llvm/IR/IntrinsicsX86.td<br>URL:<span class="apple-converted-space"> </span><a href="http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/IR/IntrinsicsX86.td?rev=216345&r1=216344&r2=216345&view=diff" target="_blank" style="color: purple; text-decoration: underline;"><span style="color: purple;">http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/IR/IntrinsicsX86.td?rev=216345&r1=216344&r2=216345&view=diff</span></a><br>==============================================================================<br>--- llvm/trunk/include/llvm/IR/IntrinsicsX86.td (original)<br>+++ llvm/trunk/include/llvm/IR/IntrinsicsX86.td Sun Aug 24 04:19:56 2014<br>@@ -1389,6 +1389,10 @@ let TargetPrefix = "x86" in { // All in<br> GCCBuiltin<"__builtin_ia32_storeupd512_mask">,<br> Intrinsic<[], [llvm_ptr_ty, llvm_v8f64_ty, llvm_i8_ty],<br> [IntrReadWriteArgMem]>;<br>+ def int_x86_avx512_mask_store_ss :<br>+ GCCBuiltin<"__builtin_ia32_storess_mask">,<br>+ Intrinsic<[], [llvm_ptr_ty, llvm_v4f32_ty, llvm_i8_ty],<br>+ [IntrReadWriteArgMem]>;<br> }<br><br> //===----------------------------------------------------------------------===//<br><br>Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp<br>URL:<span class="apple-converted-space"> </span><a href="http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=216345&r1=216344&r2=216345&view=diff" target="_blank" style="color: purple; text-decoration: underline;"><span style="color: purple;">http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=216345&r1=216344&r2=216345&view=diff</span></a><br>==============================================================================<br>--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)<br>+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Sun Aug 24 04:19:56 2014<br>@@ -49,6 +49,7 @@<br> #include "llvm/Support/ErrorHandling.h"<br> #include "llvm/Support/MathExtras.h"<br> #include "llvm/Target/TargetOptions.h"<br>+#include "X86IntrinsicsInfo.h"<br> #include <bitset><br> #include <numeric><br> #include <cctype><br>@@ -1642,6 +1643,8 @@ void X86TargetLowering::resetOperationAc<br> PredictableSelectIsExpensive = !Subtarget->isAtom();<br><br> setPrefFunctionAlignment(4); // 2^4 bytes.<br>+<br>+ InitIntrinsicTables();<br> }<br><br> // This has so far only been implemented for 64-bit MachO.<br>@@ -14488,109 +14491,40 @@ static unsigned getOpcodeForFMAIntrinsic<br> static SDValue LowerINTRINSIC_WO_CHAIN(SDValue Op, SelectionDAG &DAG) {<br> SDLoc dl(Op);<br> unsigned IntNo = cast<ConstantSDNode>(Op.getOperand(0))->getZExtValue();<br>- switch (IntNo) {<br>- default: return SDValue(); // Don't custom lower most intrinsics.<br>- // Comparison intrinsics.<br>- case Intrinsic::x86_sse_comieq_ss:<br>- case Intrinsic::x86_sse_comilt_ss:<br>- case Intrinsic::x86_sse_comile_ss:<br>- case Intrinsic::x86_sse_comigt_ss:<br>- case Intrinsic::x86_sse_comige_ss:<br>- case Intrinsic::x86_sse_comineq_ss:<br>- case Intrinsic::x86_sse_ucomieq_ss:<br>- case Intrinsic::x86_sse_ucomilt_ss:<br>- case Intrinsic::x86_sse_ucomile_ss:<br>- case Intrinsic::x86_sse_ucomigt_ss:<br>- case Intrinsic::x86_sse_ucomige_ss:<br>- case Intrinsic::x86_sse_ucomineq_ss:<br>- case Intrinsic::x86_sse2_comieq_sd:<br>- case Intrinsic::x86_sse2_comilt_sd:<br>- case Intrinsic::x86_sse2_comile_sd:<br>- case Intrinsic::x86_sse2_comigt_sd:<br>- case Intrinsic::x86_sse2_comige_sd:<br>- case Intrinsic::x86_sse2_comineq_sd:<br>- case Intrinsic::x86_sse2_ucomieq_sd:<br>- case Intrinsic::x86_sse2_ucomilt_sd:<br>- case Intrinsic::x86_sse2_ucomile_sd:<br>- case Intrinsic::x86_sse2_ucomigt_sd:<br>- case Intrinsic::x86_sse2_ucomige_sd:<br>- case Intrinsic::x86_sse2_ucomineq_sd: {<br>- unsigned Opc;<br>- ISD::CondCode CC;<br>- switch (IntNo) {<br>- default: llvm_unreachable("Impossible intrinsic"); // Can't reach here.<br>- case Intrinsic::x86_sse_comieq_ss:<br>- case Intrinsic::x86_sse2_comieq_sd:<br>- Opc = X86ISD::COMI;<br>- CC = ISD::SETEQ;<br>- break;<br>- case Intrinsic::x86_sse_comilt_ss:<br>- case Intrinsic::x86_sse2_comilt_sd:<br>- Opc = X86ISD::COMI;<br>- CC = ISD::SETLT;<br>- break;<br>- case Intrinsic::x86_sse_comile_ss:<br>- case Intrinsic::x86_sse2_comile_sd:<br>- Opc = X86ISD::COMI;<br>- CC = ISD::SETLE;<br>- break;<br>- case Intrinsic::x86_sse_comigt_ss:<br>- case Intrinsic::x86_sse2_comigt_sd:<br>- Opc = X86ISD::COMI;<br>- CC = ISD::SETGT;<br>- break;<br>- case Intrinsic::x86_sse_comige_ss:<br>- case Intrinsic::x86_sse2_comige_sd:<br>- Opc = X86ISD::COMI;<br>- CC = ISD::SETGE;<br>- break;<br>- case Intrinsic::x86_sse_comineq_ss:<br>- case Intrinsic::x86_sse2_comineq_sd:<br>- Opc = X86ISD::COMI;<br>- CC = ISD::SETNE;<br>- break;<br>- case Intrinsic::x86_sse_ucomieq_ss:<br>- case Intrinsic::x86_sse2_ucomieq_sd:<br>- Opc = X86ISD::UCOMI;<br>- CC = ISD::SETEQ;<br>- break;<br>- case Intrinsic::x86_sse_ucomilt_ss:<br>- case Intrinsic::x86_sse2_ucomilt_sd:<br>- Opc = X86ISD::UCOMI;<br>- CC = ISD::SETLT;<br>- break;<br>- case Intrinsic::x86_sse_ucomile_ss:<br>- case Intrinsic::x86_sse2_ucomile_sd:<br>- Opc = X86ISD::UCOMI;<br>- CC = ISD::SETLE;<br>- break;<br>- case Intrinsic::x86_sse_ucomigt_ss:<br>- case Intrinsic::x86_sse2_ucomigt_sd:<br>- Opc = X86ISD::UCOMI;<br>- CC = ISD::SETGT;<br>- break;<br>- case Intrinsic::x86_sse_ucomige_ss:<br>- case Intrinsic::x86_sse2_ucomige_sd:<br>- Opc = X86ISD::UCOMI;<br>- CC = ISD::SETGE;<br>- break;<br>- case Intrinsic::x86_sse_ucomineq_ss:<br>- case Intrinsic::x86_sse2_ucomineq_sd:<br>- Opc = X86ISD::UCOMI;<br>- CC = ISD::SETNE;<br>+<br>+ const IntrinsicData* IntrData = GetIntrinsicWithoutChain(IntNo);<br>+ if (IntrData) {<br>+ switch(IntrData->Type) {<br>+ case INTR_TYPE_1OP:<br>+ return DAG.getNode(IntrData->Opc0, dl, Op.getValueType(), Op.getOperand(1));<br>+ case INTR_TYPE_2OP:<br>+ return DAG.getNode(IntrData->Opc0, dl, Op.getValueType(), Op.getOperand(1),<br>+ Op.getOperand(2));<br>+ case INTR_TYPE_3OP:<br>+ return DAG.getNode(IntrData->Opc0, dl, Op.getValueType(), Op.getOperand(1),<br>+ Op.getOperand(2), Op.getOperand(3));<br>+ case COMI: { // Comparison intrinsics<br>+ ISD::CondCode CC = (ISD::CondCode)IntrData->Opc1;<br>+ SDValue LHS = Op.getOperand(1);<br>+ SDValue RHS = Op.getOperand(2);<br>+ unsigned X86CC = TranslateX86CC(CC, true, LHS, RHS, DAG);<br>+ assert(X86CC != X86::COND_INVALID && "Unexpected illegal condition!");<br>+ SDValue Cond = DAG.getNode(IntrData->Opc0, dl, MVT::i32, LHS, RHS);<br>+ SDValue SetCC = DAG.getNode(X86ISD::SETCC, dl, MVT::i8,<br>+ DAG.getConstant(X86CC, MVT::i8), Cond);<br>+ return DAG.getNode(ISD::ZERO_EXTEND, dl, MVT::i32, SetCC);<br>+ }<br>+ case VSHIFT:<br>+ return getTargetVShiftNode(IntrData->Opc0, dl, Op.getSimpleValueType(),<br>+ Op.getOperand(1), Op.getOperand(2), DAG);<br>+ default:<br> break;<br> }<br>-<br>- SDValue LHS = Op.getOperand(1);<br>- SDValue RHS = Op.getOperand(2);<br>- unsigned X86CC = TranslateX86CC(CC, true, LHS, RHS, DAG);<br>- assert(X86CC != X86::COND_INVALID && "Unexpected illegal condition!");<br>- SDValue Cond = DAG.getNode(Opc, dl, MVT::i32, LHS, RHS);<br>- SDValue SetCC = DAG.getNode(X86ISD::SETCC, dl, MVT::i8,<br>- DAG.getConstant(X86CC, MVT::i8), Cond);<br>- return DAG.getNode(ISD::ZERO_EXTEND, dl, MVT::i32, SetCC);<br> }<br><br>+ switch (IntNo) {<br>+ default: return SDValue(); // Don't custom lower most intrinsics.<br>+<br> // Arithmetic intrinsics.<br> case Intrinsic::x86_sse2_pmulu_dq:<br> case Intrinsic::x86_avx2_pmulu_dq:<br>@@ -14612,128 +14546,6 @@ static SDValue LowerINTRINSIC_WO_CHAIN(S<br> return DAG.getNode(ISD::MULHS, dl, Op.getValueType(),<br> Op.getOperand(1), Op.getOperand(2));<br><br>- // SSE2/AVX2 sub with unsigned saturation intrinsics<br>- case Intrinsic::x86_sse2_psubus_b:<br>- case Intrinsic::x86_sse2_psubus_w:<br>- case Intrinsic::x86_avx2_psubus_b:<br>- case Intrinsic::x86_avx2_psubus_w:<br>- return DAG.getNode(X86ISD::SUBUS, dl, Op.getValueType(),<br>- Op.getOperand(1), Op.getOperand(2));<br>-<br>- // SSE3/AVX horizontal add/sub intrinsics<br>- case Intrinsic::x86_sse3_hadd_ps:<br>- case Intrinsic::x86_sse3_hadd_pd:<br>- case Intrinsic::x86_avx_hadd_ps_256:<br>- case Intrinsic::x86_avx_hadd_pd_256:<br>- case Intrinsic::x86_sse3_hsub_ps:<br>- case Intrinsic::x86_sse3_hsub_pd:<br>- case Intrinsic::x86_avx_hsub_ps_256:<br>- case Intrinsic::x86_avx_hsub_pd_256:<br>- case Intrinsic::x86_ssse3_phadd_w_128:<br>- case Intrinsic::x86_ssse3_phadd_d_128:<br>- case Intrinsic::x86_avx2_phadd_w:<br>- case Intrinsic::x86_avx2_phadd_d:<br>- case Intrinsic::x86_ssse3_phsub_w_128:<br>- case Intrinsic::x86_ssse3_phsub_d_128:<br>- case Intrinsic::x86_avx2_phsub_w:<br>- case Intrinsic::x86_avx2_phsub_d: {<br>- unsigned Opcode;<br>- switch (IntNo) {<br>- default: llvm_unreachable("Impossible intrinsic"); // Can't reach here.<br>- case Intrinsic::x86_sse3_hadd_ps:<br>- case Intrinsic::x86_sse3_hadd_pd:<br>- case Intrinsic::x86_avx_hadd_ps_256:<br>- case Intrinsic::x86_avx_hadd_pd_256:<br>- Opcode = X86ISD::FHADD;<br>- break;<br>- case Intrinsic::x86_sse3_hsub_ps:<br>- case Intrinsic::x86_sse3_hsub_pd:<br>- case Intrinsic::x86_avx_hsub_ps_256:<br>- case Intrinsic::x86_avx_hsub_pd_256:<br>- Opcode = X86ISD::FHSUB;<br>- break;<br>- case Intrinsic::x86_ssse3_phadd_w_128:<br>- case Intrinsic::x86_ssse3_phadd_d_128:<br>- case Intrinsic::x86_avx2_phadd_w:<br>- case Intrinsic::x86_avx2_phadd_d:<br>- Opcode = X86ISD::HADD;<br>- break;<br>- case Intrinsic::x86_ssse3_phsub_w_128:<br>- case Intrinsic::x86_ssse3_phsub_d_128:<br>- case Intrinsic::x86_avx2_phsub_w:<br>- case Intrinsic::x86_avx2_phsub_d:<br>- Opcode = X86ISD::HSUB;<br>- break;<br>- }<br>- return DAG.getNode(Opcode, dl, Op.getValueType(),<br>- Op.getOperand(1), Op.getOperand(2));<br>- }<br>-<br>- // SSE2/SSE41/AVX2 integer max/min intrinsics.<br>- case Intrinsic::x86_sse2_pmaxu_b:<br>- case Intrinsic::x86_sse41_pmaxuw:<br>- case Intrinsic::x86_sse41_pmaxud:<br>- case Intrinsic::x86_avx2_pmaxu_b:<br>- case Intrinsic::x86_avx2_pmaxu_w:<br>- case Intrinsic::x86_avx2_pmaxu_d:<br>- case Intrinsic::x86_sse2_pminu_b:<br>- case Intrinsic::x86_sse41_pminuw:<br>- case Intrinsic::x86_sse41_pminud:<br>- case Intrinsic::x86_avx2_pminu_b:<br>- case Intrinsic::x86_avx2_pminu_w:<br>- case Intrinsic::x86_avx2_pminu_d:<br>- case Intrinsic::x86_sse41_pmaxsb:<br>- case Intrinsic::x86_sse2_pmaxs_w:<br>- case Intrinsic::x86_sse41_pmaxsd:<br>- case Intrinsic::x86_avx2_pmaxs_b:<br>- case Intrinsic::x86_avx2_pmaxs_w:<br>- case Intrinsic::x86_avx2_pmaxs_d:<br>- case Intrinsic::x86_sse41_pminsb:<br>- case Intrinsic::x86_sse2_pmins_w:<br>- case Intrinsic::x86_sse41_pminsd:<br>- case Intrinsic::x86_avx2_pmins_b:<br>- case Intrinsic::x86_avx2_pmins_w:<br>- case Intrinsic::x86_avx2_pmins_d: {<br>- unsigned Opcode;<br>- switch (IntNo) {<br>- default: llvm_unreachable("Impossible intrinsic"); // Can't reach here.<br>- case Intrinsic::x86_sse2_pmaxu_b:<br>- case Intrinsic::x86_sse41_pmaxuw:<br>- case Intrinsic::x86_sse41_pmaxud:<br>- case Intrinsic::x86_avx2_pmaxu_b:<br>- case Intrinsic::x86_avx2_pmaxu_w:<br>- case Intrinsic::x86_avx2_pmaxu_d:<br>- Opcode = X86ISD::UMAX;<br>- break;<br>- case Intrinsic::x86_sse2_pminu_b:<br>- case Intrinsic::x86_sse41_pminuw:<br>- case Intrinsic::x86_sse41_pminud:<br>- case Intrinsic::x86_avx2_pminu_b:<br>- case Intrinsic::x86_avx2_pminu_w:<br>- case Intrinsic::x86_avx2_pminu_d:<br>- Opcode = X86ISD::UMIN;<br>- break;<br>- case Intrinsic::x86_sse41_pmaxsb:<br>- case Intrinsic::x86_sse2_pmaxs_w:<br>- case Intrinsic::x86_sse41_pmaxsd:<br>- case Intrinsic::x86_avx2_pmaxs_b:<br>- case Intrinsic::x86_avx2_pmaxs_w:<br>- case Intrinsic::x86_avx2_pmaxs_d:<br>- Opcode = X86ISD::SMAX;<br>- break;<br>- case Intrinsic::x86_sse41_pminsb:<br>- case Intrinsic::x86_sse2_pmins_w:<br>- case Intrinsic::x86_sse41_pminsd:<br>- case Intrinsic::x86_avx2_pmins_b:<br>- case Intrinsic::x86_avx2_pmins_w:<br>- case Intrinsic::x86_avx2_pmins_d:<br>- Opcode = X86ISD::SMIN;<br>- break;<br>- }<br>- return DAG.getNode(Opcode, dl, Op.getValueType(),<br>- Op.getOperand(1), Op.getOperand(2));<br>- }<br>-<br> // SSE/SSE2/AVX floating point max/min intrinsics.<br> case Intrinsic::x86_sse_max_ps:<br> case Intrinsic::x86_sse2_max_pd:<br>@@ -14838,17 +14650,6 @@ static SDValue LowerINTRINSIC_WO_CHAIN(S<br> return DAG.getNode(X86ISD::PSIGN, dl, Op.getValueType(),<br> Op.getOperand(1), Op.getOperand(2));<br><br>- case Intrinsic::x86_sse41_insertps:<br>- return DAG.getNode(X86ISD::INSERTPS, dl, Op.getValueType(),<br>- Op.getOperand(1), Op.getOperand(2), Op.getOperand(3));<br>-<br>- case Intrinsic::x86_avx_vperm2f128_ps_256:<br>- case Intrinsic::x86_avx_vperm2f128_pd_256:<br>- case Intrinsic::x86_avx_vperm2f128_si_256:<br>- case Intrinsic::x86_avx2_vperm2i128:<br>- return DAG.getNode(X86ISD::VPERM2X128, dl, Op.getValueType(),<br>- Op.getOperand(1), Op.getOperand(2), Op.getOperand(3));<br>-<br> case Intrinsic::x86_avx2_permd:<br> case Intrinsic::x86_avx2_permps:<br> // Operands intentionally swapped. Mask is last operand to intrinsic,<br>@@ -14856,12 +14657,6 @@ static SDValue LowerINTRINSIC_WO_CHAIN(S<br> return DAG.getNode(X86ISD::VPERMV, dl, Op.getValueType(),<br> Op.getOperand(2), Op.getOperand(1));<br><br>- case Intrinsic::x86_sse_sqrt_ps:<br>- case Intrinsic::x86_sse2_sqrt_pd:<br>- case Intrinsic::x86_avx_sqrt_ps_256:<br>- case Intrinsic::x86_avx_sqrt_pd_256:<br>- return DAG.getNode(ISD::FSQRT, dl, Op.getValueType(), Op.getOperand(1));<br>-<br> case Intrinsic::x86_avx512_mask_valign_q_512:<br> case Intrinsic::x86_avx512_mask_valign_d_512:<br> // Vector source operands are swapped.<br>@@ -14947,100 +14742,6 @@ static SDValue LowerINTRINSIC_WO_CHAIN(S<br> return DAG.getNode(ISD::ZERO_EXTEND, dl, MVT::i32, SetCC);<br> }<br><br>- // SSE/AVX shift intrinsics<br>- case Intrinsic::x86_sse2_psll_w:<br>- case Intrinsic::x86_sse2_psll_d:<br>- case Intrinsic::x86_sse2_psll_q:<br>- case Intrinsic::x86_avx2_psll_w:<br>- case Intrinsic::x86_avx2_psll_d:<br>- case Intrinsic::x86_avx2_psll_q:<br>- case Intrinsic::x86_sse2_psrl_w:<br>- case Intrinsic::x86_sse2_psrl_d:<br>- case Intrinsic::x86_sse2_psrl_q:<br>- case Intrinsic::x86_avx2_psrl_w:<br>- case Intrinsic::x86_avx2_psrl_d:<br>- case Intrinsic::x86_avx2_psrl_q:<br>- case Intrinsic::x86_sse2_psra_w:<br>- case Intrinsic::x86_sse2_psra_d:<br>- case Intrinsic::x86_avx2_psra_w:<br>- case Intrinsic::x86_avx2_psra_d: {<br>- unsigned Opcode;<br>- switch (IntNo) {<br>- default: llvm_unreachable("Impossible intrinsic"); // Can't reach here.<br>- case Intrinsic::x86_sse2_psll_w:<br>- case Intrinsic::x86_sse2_psll_d:<br>- case Intrinsic::x86_sse2_psll_q:<br>- case Intrinsic::x86_avx2_psll_w:<br>- case Intrinsic::x86_avx2_psll_d:<br>- case Intrinsic::x86_avx2_psll_q:<br>- Opcode = X86ISD::VSHL;<br>- break;<br>- case Intrinsic::x86_sse2_psrl_w:<br>- case Intrinsic::x86_sse2_psrl_d:<br>- case Intrinsic::x86_sse2_psrl_q:<br>- case Intrinsic::x86_avx2_psrl_w:<br>- case Intrinsic::x86_avx2_psrl_d:<br>- case Intrinsic::x86_avx2_psrl_q:<br>- Opcode = X86ISD::VSRL;<br>- break;<br>- case Intrinsic::x86_sse2_psra_w:<br>- case Intrinsic::x86_sse2_psra_d:<br>- case Intrinsic::x86_avx2_psra_w:<br>- case Intrinsic::x86_avx2_psra_d:<br>- Opcode = X86ISD::VSRA;<br>- break;<br>- }<br>- return DAG.getNode(Opcode, dl, Op.getValueType(),<br>- Op.getOperand(1), Op.getOperand(2));<br>- }<br>-<br>- // SSE/AVX immediate shift intrinsics<br>- case Intrinsic::x86_sse2_pslli_w:<br>- case Intrinsic::x86_sse2_pslli_d:<br>- case Intrinsic::x86_sse2_pslli_q:<br>- case Intrinsic::x86_avx2_pslli_w:<br>- case Intrinsic::x86_avx2_pslli_d:<br>- case Intrinsic::x86_avx2_pslli_q:<br>- case Intrinsic::x86_sse2_psrli_w:<br>- case Intrinsic::x86_sse2_psrli_d:<br>- case Intrinsic::x86_sse2_psrli_q:<br>- case Intrinsic::x86_avx2_psrli_w:<br>- case Intrinsic::x86_avx2_psrli_d:<br>- case Intrinsic::x86_avx2_psrli_q:<br>- case Intrinsic::x86_sse2_psrai_w:<br>- case Intrinsic::x86_sse2_psrai_d:<br>- case Intrinsic::x86_avx2_psrai_w:<br>- case Intrinsic::x86_avx2_psrai_d: {<br>- unsigned Opcode;<br>- switch (IntNo) {<br>- default: llvm_unreachable("Impossible intrinsic"); // Can't reach here.<br>- case Intrinsic::x86_sse2_pslli_w:<br>- case Intrinsic::x86_sse2_pslli_d:<br>- case Intrinsic::x86_sse2_pslli_q:<br>- case Intrinsic::x86_avx2_pslli_w:<br>- case Intrinsic::x86_avx2_pslli_d:<br>- case Intrinsic::x86_avx2_pslli_q:<br>- Opcode = X86ISD::VSHLI;<br>- break;<br>- case Intrinsic::x86_sse2_psrli_w:<br>- case Intrinsic::x86_sse2_psrli_d:<br>- case Intrinsic::x86_sse2_psrli_q:<br>- case Intrinsic::x86_avx2_psrli_w:<br>- case Intrinsic::x86_avx2_psrli_d:<br>- case Intrinsic::x86_avx2_psrli_q:<br>- Opcode = X86ISD::VSRLI;<br>- break;<br>- case Intrinsic::x86_sse2_psrai_w:<br>- case Intrinsic::x86_sse2_psrai_d:<br>- case Intrinsic::x86_avx2_psrai_w:<br>- case Intrinsic::x86_avx2_psrai_d:<br>- Opcode = X86ISD::VSRAI;<br>- break;<br>- }<br>- return getTargetVShiftNode(Opcode, dl, Op.getSimpleValueType(),<br>- Op.getOperand(1), Op.getOperand(2), DAG);<br>- }<br>-<br> case Intrinsic::x86_sse42_pcmpistria128:<br> case Intrinsic::x86_sse42_pcmpestria128:<br> case Intrinsic::x86_sse42_pcmpistric128:<br>@@ -15352,134 +15053,25 @@ static SDValue LowerREADCYCLECOUNTER(SDV<br> return DAG.getMergeValues(Results, DL);<br> }<br><br>-enum IntrinsicType {<br>- GATHER, SCATTER, PREFETCH, RDSEED, RDRAND, RDPMC, RDTSC, XTEST, ADX<br>-};<br>-<br>-struct IntrinsicData {<br>- IntrinsicData(IntrinsicType IType, unsigned IOpc0, unsigned IOpc1)<br>- :Type(IType), Opc0(IOpc0), Opc1(IOpc1) {}<br>- IntrinsicType Type;<br>- unsigned Opc0;<br>- unsigned Opc1;<br>-};<br>-<br>-std::map < unsigned, IntrinsicData> IntrMap;<br>-static void InitIntinsicsMap() {<br>- static bool Initialized = false;<br>- if (Initialized)<br>- return;<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_gather_qps_512,<br>- IntrinsicData(GATHER, X86::VGATHERQPSZrm, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_gather_qps_512,<br>- IntrinsicData(GATHER, X86::VGATHERQPSZrm, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_gather_qpd_512,<br>- IntrinsicData(GATHER, X86::VGATHERQPDZrm, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_gather_dpd_512,<br>- IntrinsicData(GATHER, X86::VGATHERDPDZrm, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_gather_dps_512,<br>- IntrinsicData(GATHER, X86::VGATHERDPSZrm, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_gather_qpi_512,<br>- IntrinsicData(GATHER, X86::VPGATHERQDZrm, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_gather_qpq_512,<br>- IntrinsicData(GATHER, X86::VPGATHERQQZrm, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_gather_dpi_512,<br>- IntrinsicData(GATHER, X86::VPGATHERDDZrm, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_gather_dpq_512,<br>- IntrinsicData(GATHER, X86::VPGATHERDQZrm, 0)));<br>-<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_scatter_qps_512,<br>- IntrinsicData(SCATTER, X86::VSCATTERQPSZmr, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_scatter_qpd_512,<br>- IntrinsicData(SCATTER, X86::VSCATTERQPDZmr, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_scatter_dpd_512,<br>- IntrinsicData(SCATTER, X86::VSCATTERDPDZmr, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_scatter_dps_512,<br>- IntrinsicData(SCATTER, X86::VSCATTERDPSZmr, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_scatter_qpi_512,<br>- IntrinsicData(SCATTER, X86::VPSCATTERQDZmr, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_scatter_qpq_512,<br>- IntrinsicData(SCATTER, X86::VPSCATTERQQZmr, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_scatter_dpi_512,<br>- IntrinsicData(SCATTER, X86::VPSCATTERDDZmr, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_scatter_dpq_512,<br>- IntrinsicData(SCATTER, X86::VPSCATTERDQZmr, 0)));<br>-<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_gatherpf_qps_512,<br>- IntrinsicData(PREFETCH, X86::VGATHERPF0QPSm,<br>- X86::VGATHERPF1QPSm)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_gatherpf_qpd_512,<br>- IntrinsicData(PREFETCH, X86::VGATHERPF0QPDm,<br>- X86::VGATHERPF1QPDm)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_gatherpf_dpd_512,<br>- IntrinsicData(PREFETCH, X86::VGATHERPF0DPDm,<br>- X86::VGATHERPF1DPDm)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_gatherpf_dps_512,<br>- IntrinsicData(PREFETCH, X86::VGATHERPF0DPSm,<br>- X86::VGATHERPF1DPSm)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_scatterpf_qps_512,<br>- IntrinsicData(PREFETCH, X86::VSCATTERPF0QPSm,<br>- X86::VSCATTERPF1QPSm)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_scatterpf_qpd_512,<br>- IntrinsicData(PREFETCH, X86::VSCATTERPF0QPDm,<br>- X86::VSCATTERPF1QPDm)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_scatterpf_dpd_512,<br>- IntrinsicData(PREFETCH, X86::VSCATTERPF0DPDm,<br>- X86::VSCATTERPF1DPDm)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_avx512_scatterpf_dps_512,<br>- IntrinsicData(PREFETCH, X86::VSCATTERPF0DPSm,<br>- X86::VSCATTERPF1DPSm)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_rdrand_16,<br>- IntrinsicData(RDRAND, X86ISD::RDRAND, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_rdrand_32,<br>- IntrinsicData(RDRAND, X86ISD::RDRAND, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_rdrand_64,<br>- IntrinsicData(RDRAND, X86ISD::RDRAND, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_rdseed_16,<br>- IntrinsicData(RDSEED, X86ISD::RDSEED, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_rdseed_32,<br>- IntrinsicData(RDSEED, X86ISD::RDSEED, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_rdseed_64,<br>- IntrinsicData(RDSEED, X86ISD::RDSEED, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_xtest,<br>- IntrinsicData(XTEST, X86ISD::XTEST, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_rdtsc,<br>- IntrinsicData(RDTSC, X86ISD::RDTSC_DAG, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_rdtscp,<br>- IntrinsicData(RDTSC, X86ISD::RDTSCP_DAG, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_rdpmc,<br>- IntrinsicData(RDPMC, X86ISD::RDPMC_DAG, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_addcarryx_u32,<br>- IntrinsicData(ADX, X86ISD::ADC, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_addcarryx_u64,<br>- IntrinsicData(ADX, X86ISD::ADC, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_addcarry_u32,<br>- IntrinsicData(ADX, X86ISD::ADC, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_addcarry_u64,<br>- IntrinsicData(ADX, X86ISD::ADC, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_subborrow_u32,<br>- IntrinsicData(ADX, X86ISD::SBB, 0)));<br>- IntrMap.insert(std::make_pair(Intrinsic::x86_subborrow_u64,<br>- IntrinsicData(ADX, X86ISD::SBB, 0)));<br>- Initialized = true;<br>-}<br><br> static SDValue LowerINTRINSIC_W_CHAIN(SDValue Op, const X86Subtarget *Subtarget,<br> SelectionDAG &DAG) {<br>- InitIntinsicsMap();<br> unsigned IntNo = cast<ConstantSDNode>(Op.getOperand(1))->getZExtValue();<br>- std::map < unsigned, IntrinsicData>::const_iterator itr = IntrMap.find(IntNo);<br>- if (itr == IntrMap.end())<br>+<br>+ const IntrinsicData* IntrData = GetIntrinsicWithChain(IntNo);<br>+ if (!IntrData)<br> return SDValue();<br><br> SDLoc dl(Op);<br>- IntrinsicData Intr = itr->second;<br>- switch(Intr.Type) {<br>+ switch(IntrData->Type) {<br>+ default:<br>+ llvm_unreachable("Unknown Intrinsic Type");<br>+ break;<br> case RDSEED:<br> case RDRAND: {<br> // Emit the node with the right value type.<br> SDVTList VTs = DAG.getVTList(Op->getValueType(0), MVT::Glue, MVT::Other);<br>- SDValue Result = DAG.getNode(Intr.Opc0, dl, VTs, Op.getOperand(0));<br>+ SDValue Result = DAG.getNode(IntrData->Opc0, dl, VTs, Op.getOperand(0));<br><br> // If the value returned by RDRAND/RDSEED was valid (CF=1), return 1.<br> // Otherwise return the value from Rand, which is always 0, casted to i32.<br>@@ -15503,7 +15095,7 @@ static SDValue LowerINTRINSIC_W_CHAIN(SD<br> SDValue Index = Op.getOperand(4);<br> SDValue Mask = Op.getOperand(5);<br> SDValue Scale = Op.getOperand(6);<br>- return getGatherNode(Intr.Opc0, Op, DAG, Src, Mask, Base, Index, Scale, Chain,<br>+ return getGatherNode(IntrData->Opc0, Op, DAG, Src, Mask, Base, Index, Scale, Chain,<br> Subtarget);<br> }<br> case SCATTER: {<br>@@ -15514,7 +15106,7 @@ static SDValue LowerINTRINSIC_W_CHAIN(SD<br> SDValue Index = Op.getOperand(4);<br> SDValue Src = Op.getOperand(5);<br> SDValue Scale = Op.getOperand(6);<br>- return getScatterNode(Intr.Opc0, Op, DAG, Src, Mask, Base, Index, Scale, Chain);<br>+ return getScatterNode(IntrData->Opc0, Op, DAG, Src, Mask, Base, Index, Scale, Chain);<br> }<br> case PREFETCH: {<br> SDValue Hint = Op.getOperand(6);<br>@@ -15522,7 +15114,7 @@ static SDValue LowerINTRINSIC_W_CHAIN(SD<br> if (dyn_cast<ConstantSDNode> (Hint) == nullptr ||<br> (HintVal = dyn_cast<ConstantSDNode> (Hint)->getZExtValue()) > 1)<br> llvm_unreachable("Wrong prefetch hint in intrinsic: should be 0 or 1");<br>- unsigned Opcode = (HintVal ? Intr.Opc1 : Intr.Opc0);<br>+ unsigned Opcode = (HintVal ? IntrData->Opc1 : IntrData->Opc0);<br> SDValue Chain = Op.getOperand(0);<br> SDValue Mask = Op.getOperand(2);<br> SDValue Index = Op.getOperand(3);<br>@@ -15533,7 +15125,7 @@ static SDValue LowerINTRINSIC_W_CHAIN(SD<br> // Read Time Stamp Counter (RDTSC) and Processor ID (RDTSCP).<br> case RDTSC: {<br> SmallVector<SDValue, 2> Results;<br>- getReadTimeStampCounter(Op.getNode(), dl, Intr.Opc0, DAG, Subtarget, Results);<br>+ getReadTimeStampCounter(Op.getNode(), dl, IntrData->Opc0, DAG, Subtarget, Results);<br> return DAG.getMergeValues(Results, dl);<br> }<br> // Read Performance Monitoring Counters.<br>@@ -15545,7 +15137,7 @@ static SDValue LowerINTRINSIC_W_CHAIN(SD<br> // XTEST intrinsics.<br> case XTEST: {<br> SDVTList VTs = DAG.getVTList(Op->getValueType(0), MVT::Other);<br>- SDValue InTrans = DAG.getNode(X86ISD::XTEST, dl, VTs, Op.getOperand(0));<br>+ SDValue InTrans = DAG.getNode(IntrData->Opc0, dl, VTs, Op.getOperand(0));<br> SDValue SetCC = DAG.getNode(X86ISD::SETCC, dl, MVT::i8,<br> DAG.getConstant(X86::COND_NE, MVT::i8),<br> InTrans);<br>@@ -15560,7 +15152,7 @@ static SDValue LowerINTRINSIC_W_CHAIN(SD<br> SDVTList VTs = DAG.getVTList(Op.getOperand(3)->getValueType(0), MVT::Other);<br> SDValue GenCF = DAG.getNode(X86ISD::ADD, dl, CFVTs, Op.getOperand(2),<br> DAG.getConstant(-1, MVT::i8));<br>- SDValue Res = DAG.getNode(Intr.Opc0, dl, VTs, Op.getOperand(3),<br>+ SDValue Res = DAG.getNode(IntrData->Opc0, dl, VTs, Op.getOperand(3),<br> Op.getOperand(4), GenCF.getValue(1));<br> SDValue Store = DAG.getStore(Op.getOperand(0), dl, Res.getValue(0),<br> Op.getOperand(5), MachinePointerInfo(),<br>@@ -15573,7 +15165,6 @@ static SDValue LowerINTRINSIC_W_CHAIN(SD<br> return DAG.getMergeValues(Results, dl);<br> }<br> }<br>- llvm_unreachable("Unknown Intrinsic Type");<br> }<br><br> SDValue X86TargetLowering::LowerRETURNADDR(SDValue Op,<br><br>Added: llvm/trunk/lib/Target/X86/X86IntrinsicsInfo.h<br>URL:<span class="apple-converted-space"> </span><a href="http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86IntrinsicsInfo.h?rev=216345&view=auto" target="_blank" style="color: purple; text-decoration: underline;"><span style="color: purple;">http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86IntrinsicsInfo.h?rev=216345&view=auto</span></a><br>==============================================================================<br>--- llvm/trunk/lib/Target/X86/X86IntrinsicsInfo.h (added)<br>+++ llvm/trunk/lib/Target/X86/X86IntrinsicsInfo.h Sun Aug 24 04:19:56 2014<br>@@ -0,0 +1,241 @@<br>+//===-- X86IntinsicsInfo.h - X86 Instrinsics ------------*- C++ -*-===//<br>+//<br>+// The LLVM Compiler Infrastructure<br>+//<br>+// This file is distributed under the University of Illinois Open Source<br>+// License. See LICENSE.TXT for details.<br>+//<br>+//===----------------------------------------------------------------------===//<br>+//<br>+// This file contains the details for lowering X86 intrinsics<br>+//<br>+//===----------------------------------------------------------------------===//<br>+<br>+#ifndef LLVM_LIB_TARGET_X86_X86INTRINSICSINFO_H<br>+#define LLVM_LIB_TARGET_X86_X86INTRINSICSINFO_H<br>+<br>+using namespace llvm;<br>+<br>+enum IntrinsicType {<br>+ GATHER, SCATTER, PREFETCH, RDSEED, RDRAND, RDPMC, RDTSC, XTEST, ADX,<br>+ INTR_TYPE_1OP, INTR_TYPE_2OP, INTR_TYPE_3OP, VSHIFT,<br>+ COMI<br>+};<br>+<br>+struct IntrinsicData {<br>+ IntrinsicData(IntrinsicType IType, unsigned IOpc0, unsigned IOpc1)<br>+ :Type(IType), Opc0(IOpc0), Opc1(IOpc1) {}<br>+ IntrinsicType Type;<br>+ unsigned Opc0;<br>+ unsigned Opc1;<br>+};<br>+<br>+#define INTRINSIC_WITH_CHAIN(id, type, op0, op1) \<br>+ IntrWithChainMap.insert(std::make_pair(Intrinsic::id, \<br>+ IntrinsicData(type, op0, op1)))<br>+<br>+std::map < unsigned, IntrinsicData> IntrWithChainMap;<br>+void InitIntrinsicsWithChain() {<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_gather_qps_512, GATHER, X86::VGATHERQPSZrm, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_gather_qpd_512, GATHER, X86::VGATHERQPDZrm, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_gather_dps_512, GATHER, X86::VGATHERDPSZrm, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_gather_dpd_512, GATHER, X86::VGATHERDPDZrm, 0);<br>+<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_gather_qpi_512, GATHER, X86::VPGATHERQDZrm, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_gather_qpq_512, GATHER, X86::VPGATHERQQZrm, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_gather_dpi_512, GATHER, X86::VPGATHERDDZrm, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_gather_dpq_512, GATHER, X86::VPGATHERDQZrm, 0);<br>+<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_scatter_qps_512, SCATTER, X86::VSCATTERQPSZmr, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_scatter_qpd_512, SCATTER, X86::VSCATTERQPDZmr, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_scatter_dps_512, SCATTER, X86::VSCATTERDPSZmr, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_scatter_dpd_512, SCATTER, X86::VSCATTERDPDZmr, 0);<br>+<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_scatter_qpi_512, SCATTER, X86::VPSCATTERQDZmr, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_scatter_qpq_512, SCATTER, X86::VPSCATTERQQZmr, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_scatter_dpi_512, SCATTER, X86::VPSCATTERDDZmr, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_scatter_dpq_512, SCATTER, X86::VPSCATTERDQZmr, 0);<br>+<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_gatherpf_qps_512, PREFETCH, X86::VGATHERPF0QPSm, X86::VGATHERPF1QPSm);<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_gatherpf_qpd_512, PREFETCH, X86::VGATHERPF0QPDm, X86::VGATHERPF1QPDm);<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_gatherpf_dps_512, PREFETCH, X86::VGATHERPF0DPSm, X86::VGATHERPF1DPSm);<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_gatherpf_dpd_512, PREFETCH, X86::VGATHERPF0DPDm, X86::VGATHERPF1DPDm);<br>+<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_scatterpf_qps_512, PREFETCH, X86::VSCATTERPF0QPSm, X86::VSCATTERPF1QPSm);<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_scatterpf_qpd_512, PREFETCH, X86::VSCATTERPF0QPDm, X86::VSCATTERPF1QPDm);<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_scatterpf_dps_512, PREFETCH, X86::VSCATTERPF0DPSm, X86::VSCATTERPF1DPSm);<br>+ INTRINSIC_WITH_CHAIN(x86_avx512_scatterpf_dpd_512, PREFETCH, X86::VSCATTERPF0DPDm, X86::VSCATTERPF1DPDm);<br>+<br>+ INTRINSIC_WITH_CHAIN(x86_rdrand_16, RDRAND, X86ISD::RDRAND, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_rdrand_32, RDRAND, X86ISD::RDRAND, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_rdrand_64, RDRAND, X86ISD::RDRAND, 0);<br>+<br>+ INTRINSIC_WITH_CHAIN(x86_rdseed_16, RDSEED, X86ISD::RDSEED, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_rdseed_32, RDSEED, X86ISD::RDSEED, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_rdseed_64, RDSEED, X86ISD::RDSEED, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_xtest, XTEST, X86ISD::XTEST, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_rdtsc, RDTSC, X86ISD::RDTSC_DAG, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_rdtscp, RDTSC, X86ISD::RDTSCP_DAG, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_rdpmc, RDPMC, X86ISD::RDPMC_DAG, 0);<br>+<br>+ INTRINSIC_WITH_CHAIN(x86_addcarryx_u32, ADX, X86ISD::ADC, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_addcarryx_u64, ADX, X86ISD::ADC, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_addcarry_u32, ADX, X86ISD::ADC, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_addcarry_u64, ADX, X86ISD::ADC, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_subborrow_u32, ADX, X86ISD::SBB, 0);<br>+ INTRINSIC_WITH_CHAIN(x86_subborrow_u64, ADX, X86ISD::SBB, 0);<br>+<br>+}<br>+<br>+const IntrinsicData* GetIntrinsicWithChain(unsigned IntNo) {<br>+ std::map < unsigned, IntrinsicData>::const_iterator itr =<br>+ IntrWithChainMap.find(IntNo);<br>+ if (itr == IntrWithChainMap.end())<br>+ return NULL;<br>+ return &(itr->second);<br>+}<br>+<br>+#define INTRINSIC_WO_CHAIN(id, type, op0, op1) \<br>+ IntrWithoutChainMap.insert(std::make_pair(Intrinsic::id, \<br>+ IntrinsicData(type, op0, op1)))<br>+<br>+<br>+std::map < unsigned, IntrinsicData> IntrWithoutChainMap;<br>+<br>+void InitIntrinsicsWithoutChain() {<br>+ INTRINSIC_WO_CHAIN(x86_sse_sqrt_ps, INTR_TYPE_1OP, ISD::FSQRT, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_sqrt_pd, INTR_TYPE_1OP, ISD::FSQRT, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx_sqrt_ps_256, INTR_TYPE_1OP, ISD::FSQRT, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx_sqrt_pd_256, INTR_TYPE_1OP, ISD::FSQRT, 0);<br>+<br>+ INTRINSIC_WO_CHAIN(x86_sse2_psubus_b, INTR_TYPE_2OP, X86ISD::SUBUS, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_psubus_w, INTR_TYPE_2OP, X86ISD::SUBUS, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_psubus_b, INTR_TYPE_2OP, X86ISD::SUBUS, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_psubus_w, INTR_TYPE_2OP, X86ISD::SUBUS, 0);<br>+<br>+ INTRINSIC_WO_CHAIN(x86_sse3_hadd_ps, INTR_TYPE_2OP, X86ISD::FHADD, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse3_hadd_pd, INTR_TYPE_2OP, X86ISD::FHADD, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx_hadd_ps_256, INTR_TYPE_2OP, X86ISD::FHADD, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx_hadd_pd_256, INTR_TYPE_2OP, X86ISD::FHADD, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse3_hsub_ps, INTR_TYPE_2OP, X86ISD::FHSUB, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse3_hsub_pd, INTR_TYPE_2OP, X86ISD::FHSUB, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx_hsub_ps_256, INTR_TYPE_2OP, X86ISD::FHSUB, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx_hsub_pd_256, INTR_TYPE_2OP, X86ISD::FHSUB, 0);<br>+ INTRINSIC_WO_CHAIN(x86_ssse3_phadd_w_128, INTR_TYPE_2OP, X86ISD::HADD, 0);<br>+ INTRINSIC_WO_CHAIN(x86_ssse3_phadd_d_128, INTR_TYPE_2OP, X86ISD::HADD, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_phadd_w, INTR_TYPE_2OP, X86ISD::HADD, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_phadd_d, INTR_TYPE_2OP, X86ISD::HADD, 0);<br>+ INTRINSIC_WO_CHAIN(x86_ssse3_phsub_w_128, INTR_TYPE_2OP, X86ISD::HSUB, 0);<br>+ INTRINSIC_WO_CHAIN(x86_ssse3_phsub_d_128, INTR_TYPE_2OP, X86ISD::HSUB, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_phsub_w, INTR_TYPE_2OP, X86ISD::HSUB, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_phsub_d, INTR_TYPE_2OP, X86ISD::HSUB, 0);<br>+<br>+ INTRINSIC_WO_CHAIN(x86_sse2_pmaxu_b, INTR_TYPE_2OP, X86ISD::UMAX, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse41_pmaxuw, INTR_TYPE_2OP, X86ISD::UMAX, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse41_pmaxud, INTR_TYPE_2OP, X86ISD::UMAX, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_pmaxu_b, INTR_TYPE_2OP, X86ISD::UMAX, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_pmaxu_w, INTR_TYPE_2OP, X86ISD::UMAX, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_pmaxu_d, INTR_TYPE_2OP, X86ISD::UMAX, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_pminu_b, INTR_TYPE_2OP, X86ISD::UMIN, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse41_pminuw, INTR_TYPE_2OP, X86ISD::UMIN, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse41_pminud, INTR_TYPE_2OP, X86ISD::UMIN, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_pminu_b, INTR_TYPE_2OP, X86ISD::UMIN, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_pminu_w, INTR_TYPE_2OP, X86ISD::UMIN, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_pminu_d, INTR_TYPE_2OP, X86ISD::UMIN, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse41_pmaxsb, INTR_TYPE_2OP, X86ISD::SMAX, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_pmaxs_w, INTR_TYPE_2OP, X86ISD::SMAX, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse41_pmaxsd, INTR_TYPE_2OP, X86ISD::SMAX, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_pmaxs_b, INTR_TYPE_2OP, X86ISD::SMAX, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_pmaxs_w, INTR_TYPE_2OP, X86ISD::SMAX, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_pmaxs_d, INTR_TYPE_2OP, X86ISD::SMAX, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse41_pminsb, INTR_TYPE_2OP, X86ISD::SMIN, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_pmins_w, INTR_TYPE_2OP, X86ISD::SMIN, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse41_pminsd, INTR_TYPE_2OP, X86ISD::SMIN, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_pmins_b, INTR_TYPE_2OP, X86ISD::SMIN, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_pmins_w, INTR_TYPE_2OP, X86ISD::SMIN, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_pmins_d, INTR_TYPE_2OP, X86ISD::SMIN, 0);<br>+<br>+ INTRINSIC_WO_CHAIN(x86_sse2_psll_w, INTR_TYPE_2OP, X86ISD::VSHL, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_psll_d, INTR_TYPE_2OP, X86ISD::VSHL, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_psll_q, INTR_TYPE_2OP, X86ISD::VSHL, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_psll_w, INTR_TYPE_2OP, X86ISD::VSHL, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_psll_d, INTR_TYPE_2OP, X86ISD::VSHL, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_psll_q, INTR_TYPE_2OP, X86ISD::VSHL, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_psrl_w, INTR_TYPE_2OP, X86ISD::VSRL, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_psrl_d, INTR_TYPE_2OP, X86ISD::VSRL, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_psrl_q, INTR_TYPE_2OP, X86ISD::VSRL, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_psrl_w, INTR_TYPE_2OP, X86ISD::VSRL, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_psrl_d, INTR_TYPE_2OP, X86ISD::VSRL, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_psrl_q, INTR_TYPE_2OP, X86ISD::VSRL, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_psra_w, INTR_TYPE_2OP, X86ISD::VSRA, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_psra_d, INTR_TYPE_2OP, X86ISD::VSRA, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_psra_w, INTR_TYPE_2OP, X86ISD::VSRA, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_psra_d, INTR_TYPE_2OP, X86ISD::VSRA, 0);<br>+<br>+ INTRINSIC_WO_CHAIN(x86_sse2_pslli_w, VSHIFT, X86ISD::VSHLI, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_pslli_d, VSHIFT, X86ISD::VSHLI, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_pslli_q, VSHIFT, X86ISD::VSHLI, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_pslli_w, VSHIFT, X86ISD::VSHLI, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_pslli_d, VSHIFT, X86ISD::VSHLI, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_pslli_q, VSHIFT, X86ISD::VSHLI, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_psrli_w, VSHIFT, X86ISD::VSRLI, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_psrli_d, VSHIFT, X86ISD::VSRLI, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_psrli_q, VSHIFT, X86ISD::VSRLI, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_psrli_w, VSHIFT, X86ISD::VSRLI, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_psrli_d, VSHIFT, X86ISD::VSRLI, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_psrli_q, VSHIFT, X86ISD::VSRLI, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_psrai_w, VSHIFT, X86ISD::VSRAI, 0);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_psrai_d, VSHIFT, X86ISD::VSRAI, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_psrai_w, VSHIFT, X86ISD::VSRAI, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_psrai_d, VSHIFT, X86ISD::VSRAI, 0);<br>+<br>+ INTRINSIC_WO_CHAIN(x86_avx_vperm2f128_ps_256, INTR_TYPE_3OP, X86ISD::VPERM2X128, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx_vperm2f128_pd_256, INTR_TYPE_3OP, X86ISD::VPERM2X128, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx_vperm2f128_si_256, INTR_TYPE_3OP, X86ISD::VPERM2X128, 0);<br>+ INTRINSIC_WO_CHAIN(x86_avx2_vperm2i128, INTR_TYPE_3OP, X86ISD::VPERM2X128, 0);<br>+<br>+ INTRINSIC_WO_CHAIN(x86_sse41_insertps, INTR_TYPE_3OP, X86ISD::INSERTPS, 0);<br>+<br>+ INTRINSIC_WO_CHAIN(x86_sse_comieq_ss, COMI, X86ISD::COMI, ISD::SETEQ);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_comieq_sd, COMI, X86ISD::COMI, ISD::SETEQ);<br>+ INTRINSIC_WO_CHAIN(x86_sse_comilt_ss, COMI, X86ISD::COMI, ISD::SETLT);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_comilt_sd, COMI, X86ISD::COMI, ISD::SETLT);<br>+ INTRINSIC_WO_CHAIN(x86_sse_comile_ss, COMI, X86ISD::COMI, ISD::SETLE);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_comile_sd, COMI, X86ISD::COMI, ISD::SETLE);<br>+ INTRINSIC_WO_CHAIN(x86_sse_comigt_ss, COMI, X86ISD::COMI, ISD::SETGT);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_comigt_sd, COMI, X86ISD::COMI, ISD::SETGT);<br>+ INTRINSIC_WO_CHAIN(x86_sse_comige_ss, COMI, X86ISD::COMI, ISD::SETGE);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_comige_sd, COMI, X86ISD::COMI, ISD::SETGE);<br>+ INTRINSIC_WO_CHAIN(x86_sse_comineq_ss, COMI, X86ISD::COMI, ISD::SETNE);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_comineq_sd,COMI, X86ISD::COMI, ISD::SETNE);<br>+<br>+ INTRINSIC_WO_CHAIN(x86_sse_ucomieq_ss, COMI, X86ISD::UCOMI, ISD::SETEQ);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_ucomieq_sd, COMI, X86ISD::UCOMI, ISD::SETEQ);<br>+ INTRINSIC_WO_CHAIN(x86_sse_ucomilt_ss, COMI, X86ISD::UCOMI, ISD::SETLT);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_ucomilt_sd, COMI, X86ISD::UCOMI, ISD::SETLT);<br>+ INTRINSIC_WO_CHAIN(x86_sse_ucomile_ss, COMI, X86ISD::UCOMI, ISD::SETLE);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_ucomile_sd, COMI, X86ISD::UCOMI, ISD::SETLE);<br>+ INTRINSIC_WO_CHAIN(x86_sse_ucomigt_ss, COMI, X86ISD::UCOMI, ISD::SETGT);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_ucomigt_sd, COMI, X86ISD::UCOMI, ISD::SETGT);<br>+ INTRINSIC_WO_CHAIN(x86_sse_ucomige_ss, COMI, X86ISD::UCOMI, ISD::SETGE);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_ucomige_sd, COMI, X86ISD::UCOMI, ISD::SETGE);<br>+ INTRINSIC_WO_CHAIN(x86_sse_ucomineq_ss, COMI, X86ISD::UCOMI, ISD::SETNE);<br>+ INTRINSIC_WO_CHAIN(x86_sse2_ucomineq_sd,COMI, X86ISD::UCOMI, ISD::SETNE);<br>+}<br>+<br>+const IntrinsicData* GetIntrinsicWithoutChain(unsigned IntNo) {<br>+ std::map < unsigned, IntrinsicData>::const_iterator itr =<br>+ IntrWithoutChainMap.find(IntNo);<br>+ if (itr == IntrWithoutChainMap.end())<br>+ return NULL;<br>+ return &(itr->second);<br>+}<br>+<br>+// Initialize intrinsics data<br>+void InitIntrinsicTables() {<br>+ InitIntrinsicsWithChain();<br>+ InitIntrinsicsWithoutChain();<br>+}<br>+<br>+<br>+#endif<br><br><br>_______________________________________________<br>llvm-commits mailing list<br><a href="mailto:llvm-commits@cs.uiuc.edu" target="_blank" style="color: purple; text-decoration: underline;"><span style="color: purple;">llvm-commits@cs.uiuc.edu</span></a><br><a href="http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits" target="_blank" style="color: purple; text-decoration: underline;"><span style="color: purple;">http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits</span></a><o:p></o:p></div></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"> <o:p></o:p></div></div></div></div></div><p style="margin-right: 0cm; margin-left: 0cm; font-size: 12pt; font-family: 'Times New Roman', serif;">---------------------------------------------------------------------<br>Intel Israel (74) Limited<o:p></o:p></p><p style="margin-right: 0cm; margin-left: 0cm; font-size: 12pt; font-family: 'Times New Roman', serif;">This e-mail and any attachments may contain confidential material for<br>the sole use of the intended recipient(s). Any review or distribution<br>by others is strictly prohibited. If you are not the intended<br>recipient, please contact the sender and delete all copies.<o:p></o:p></p></blockquote></div></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"> <o:p></o:p></div></div></div><p class="MsoNormal" style="margin: 0cm 0cm 12pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><br>_______________________________________________<br>llvm-commits mailing list<br><a href="mailto:llvm-commits@cs.uiuc.edu" target="_blank" style="color: purple; text-decoration: underline;"><span style="color: purple;">llvm-commits@cs.uiuc.edu</span></a><br><a href="http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits" target="_blank" style="color: purple; text-decoration: underline;"><span style="color: purple;">http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits</span></a><o:p></o:p></p></blockquote></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"> <o:p></o:p></div></div></blockquote></div><div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"> <o:p></o:p></div></div></div></div><p style="margin-right: 0cm; margin-left: 0cm; font-size: 12pt; font-family: 'Times New Roman', serif;">---------------------------------------------------------------------<br>Intel Israel (74) Limited<o:p></o:p></p><p style="margin-right: 0cm; margin-left: 0cm; font-size: 12pt; font-family: 'Times New Roman', serif;">This e-mail and any attachments may contain confidential material for<br>the sole use of the intended recipient(s). Any review or distribution<br>by others is strictly prohibited. If you are not the intended<br>recipient, please contact the sender and delete all copies.<o:p></o:p></p><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 9pt; font-family: Helvetica, sans-serif;"><intrinsics_map_2.diff>_______________________________________________<br>llvm-commits mailing list<br><a href="mailto:llvm-commits@cs.uiuc.edu" style="color: purple; text-decoration: underline;"><span style="color: purple;">llvm-commits@cs.uiuc.edu</span></a><br><a href="http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits" style="color: purple; text-decoration: underline;"><span style="color: purple;">http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits</span></a><o:p></o:p></span></div></div></div><div style="margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></div></div></div><p style="margin-right: 0cm; margin-left: 0cm; font-size: 12pt; font-family: 'Times New Roman', serif;">---------------------------------------------------------------------<br>Intel Israel (74) Limited</p><p style="margin-right: 0cm; margin-left: 0cm; font-size: 12pt; font-family: 'Times New Roman', serif;">This e-mail and any attachments may contain confidential material for<br>the sole use of the intended recipient(s). Any review or distribution<br>by others is strictly prohibited. If you are not the intended<br>recipient, please contact the sender and delete all copies.</p></div></blockquote></div></div></body></html>