<html><head><meta http-equiv="Content-Type" content="text/html; charset=utf-8"></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;" class=""><br class=""><div><br class=""><blockquote type="cite" class=""><div class="">On 16 Nov 2020, at 16:37, Cameron McInally <<a href="mailto:cameron.mcinally@nyu.edu" class="">cameron.mcinally@nyu.edu</a>> wrote:</div><br class="Apple-interchange-newline"><div class=""><meta charset="UTF-8" class=""><span style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none; float: none; display: inline !important;" class="">On Fri, Nov 13, 2020 at 3:01 PM Florian Hahn <</span><a href="mailto:florian_hahn@apple.com" style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; orphans: auto; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: auto; word-spacing: 0px; -webkit-text-size-adjust: auto; -webkit-text-stroke-width: 0px;" class="">florian_hahn@apple.com</a><span style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none; float: none; display: inline !important;" class="">> wrote:</span><br style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;" class=""><blockquote type="cite" style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; orphans: auto; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: auto; word-spacing: 0px; -webkit-text-size-adjust: auto; -webkit-text-stroke-width: 0px; text-decoration: none;" class=""><br class=""><br class=""><br class="">On Nov 12, 2020, at 21:36, Cameron McInally <<a href="mailto:cameron.mcinally@nyu.edu" class="">cameron.mcinally@nyu.edu</a>> wrote:<br class=""><br class="">On Thu, Nov 12, 2020 at 2:47 PM Florian Hahn <<a href="mailto:florian_hahn@apple.com" class="">florian_hahn@apple.com</a>> wrote:<br class=""><br class="">There certainly is some pieces around the edges that will need adjusting or become obsolete, but I would hope/expect the majority of the work to be re-usable.<br class=""><br class="">As for getting the usual algebraic simplifications for free, even with a new type I suppose we would have to teach instcombine/InstructionSimplify about them. This is indeed an area where a dedicated type is probably quite a bit easier to deal with. But I think part of the vector-predication proposal includes some generic matchers for the different intrinsic, which should be relatively straight-forward to update as well.<br class=""><br class=""><br class="">Yes, the IR passes are where I see complex types win. For example,<br class="">pretty much all of InstCombine's visitFMul(...) transformations should<br class="">not depend on type.*** So there's no need to duplicate all that code<br class="">for @llvm.complex.multiply.<br class=""><br class="">It's true that teaching the pattern matcher to recognize intrinsics<br class="">gets us a long way. But we'd also need to update IR passes in places<br class="">where the Opcode is explicitly checked. E.g.:<br class=""><br class="">```<br class="">switch (Opcode) {<br class=""><snipped><br class="">case Instruction::FAdd: <== ***HERE***<br class=""> return SimplifyFAddInst(LHS, RHS, FastMathFlags(), Q, MaxRecurse);<br class="">```<br class=""><br class="">And when a transformation succeeds, we'd need to update how the result<br class="">is returned. E.g.:<br class=""><br class="">```<br class="">// (-X) + Y --> Y - X<br class="">Value *X, *Y;<br class="">if (match(&I, m_c_FAdd(m_FNeg(m_Value(X)), m_Value(Y))))<br class=""> return BinaryOperator::CreateFSubFMF(Y, X, &I); <== ***HERE***<br class="">```<br class=""><br class="">Both of those come for free with a complex type, along with the<br class="">pattern matching code.<br class=""><br class=""><br class="">Thanks for expanding on this! I think for InstructionSimplify it should be fairly easy to treat @llvm.complex.multiply & co as their instruction counter parts. InstCombine (or other passes that create new instructions as they combine) will require more additional changes.<br class=""><br class="">This is certainly a trade-off compared to adding dedicated types. But I given the additional complexity and overhead adding new types entails, I think the intrinsics are a good alternative that can get us quite far, especially if we can re-use/build on some of the techniques Simon mentioned.<br class=""><br class="">I'm not opposed to finding solutions for these problems, but this<br class="">pitch has been around for a while now, and it hasn't really taken off<br class="">yet. It's hard to rally behind it again.<br class=""><br class=""><br class="">If there’s been an earlier discussion about introducing a small set of intrinsics, I think I missed that. The references linked from the complex type proposal seem to all focus on adding a new type.<br class=""></blockquote><br style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;" class=""><span style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none; float: none; display: inline !important;" class="">Sorry, I wasn't clear. I meant the pitch to optimize experimental</span><br style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;" class=""><span style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none; float: none; display: inline !important;" class="">intrinsics in general. That general plan to optimize experimental</span><br style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;" class=""><span style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none; float: none; display: inline !important;" class="">intrinsics has been around since the constrained FP intrinsics project</span><br style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;" class=""><span style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none; float: none; display: inline !important;" class="">started. Probably 2016 or 2017.</span><br style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;" class=""></div></blockquote></div><br class=""><div class="">Oh I see, thanks for the pointers! Constraint FP support is another interesting example.</div><div class=""><br class=""></div><div class="">Cheers,</div><div class="">Florian</div></body></html>