<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 2020 Nov 17, at 13:40, Sean Silva <<a href="mailto:chisophugis@gmail.com" class="">chisophugis@gmail.com</a>> wrote:</div><br class="Apple-interchange-newline"><div class=""><div dir="ltr" class=""><div dir="ltr" class=""><br class=""></div><br class=""><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Tue, Nov 17, 2020 at 7:26 AM Chris Lattner <<a href="mailto:clattner@nondot.org" class="">clattner@nondot.org</a>> wrote:<br class=""></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div style="overflow-wrap: break-word;" class="">On Nov 13, 2020, at 2:06 PM, Sean Silva via llvm-dev <<a href="mailto:llvm-dev@lists.llvm.org" target="_blank" class="">llvm-dev@lists.llvm.org</a>> wrote:<div class=""><blockquote type="cite" class=""><div class=""><div dir="ltr" class="">We've pretty happy now with a patch that adds two wrappers around SmallVector that make it 1) more convenient to use and 2) will tend to mitigate misuse of SmallVector. We think it's ready for wider discussion: <a href="https://reviews.llvm.org/D90884" target="_blank" class="">https://reviews.llvm.org/D90884</a><div class=""><br class=""></div><div class="">SVec<T> is a convenience alias for SmallVector<T, N> with N chosen automatically to keep its size under 64 Bytes (that heuristic is easy to change though). The recommendation in the patch is to use this "on the stack, where a "small" number of elements are expected".<br class=""></div></div></div></blockquote><div class=""><br class=""></div><div class="">Hey Sean,</div><div class=""><br class=""></div><div class="">I agree with other comments that his approach unnecessarily fragments the API surface area for a core class. You’re doing two things here: 1) adding a new name for an existing thing, and 2) adding a default.</div><div class=""><br class=""></div><div class="">My major concern and objection is about #1, since the codebase will get to be a mishmash of the two. I’d rather keep the world consistent here and have an opinionated “one way to do it”.</div></div></div></blockquote><div class=""><br class=""></div><div class="">Thanks Chris. That was my original proposal when I first drafted the patch, and I'm happy to just add the default.</div><div class=""> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div style="overflow-wrap: break-word;" class=""><div class=""><div class=""><br class=""></div><div class="">Thoughts/suggestions:</div><div class=""> - Adding the default seems very reasonable to me, and I think that 64 bytes is a good default. I think you should change the behavior so that SmallVector<LargeThing> defaults to a single inline element instead of zero though. Perhaps generate a static_assert when it is crazy large.</div></div></div></blockquote><div class=""><br class=""></div><div class="">That would work for me.</div><div class=""><br class=""></div><div class="">Mehdi, would SmallVector<T>, defaulting to a minimum single inline element with static_assert(sizeof(T) < 512) address your concerns?</div></div></div></div></blockquote><div><br class=""></div><div>I'd be a bit concerned about this for two reasons:</div><div><br class=""></div><div>- Adding a large inlined element by-default may reinforce one of the problems with SmallVector, that people accidentally put big vectors in places that shouldn't have them. If we're going to make accidentally large inlined vectors more ergonomic, I think we ought to make 0-element versions more ergonomic too.</div><div class=""><br class=""></div><div>- The `static_assert` is going to fail differently on different platforms, since `T`'s size will tend to be platform-dependent for large `T`. I'm not sure it'll be predictable enough.</div><div><br class=""></div><blockquote type="cite" class=""><div class=""><div dir="ltr" class=""><div class="gmail_quote"><div class="">-- Sean Silva</div><div class=""> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div style="overflow-wrap: break-word;" class=""><div class=""><div class=""><br class=""></div><div class=""> - The name SmallVector has always been confusing, InlinedVector is more principled, but is even more verbose for such a common type. If you think it is worth shrinkifying the name SmallVector, then I think the best thing to do is *rename* SmallVector (perhaps to IVector?) everywhere in the codebase. I don’t think that introducing a redundant name is a good thing (except to smooth the transition).</div></div></div></blockquote></div></div></div></blockquote><div><br class=""></div><div>I like the idea of a minor rename, although a bit different:</div><div>- SmallVectorBase => VectorBase</div><div>- SmallVectorImpl => VectorImpl (we could keep both around for a transition period)</div><div>- (etc.)</div><div>- SmallVector => Vector, but give `Vector` a default small size of 0</div><div>- Add SmallVector as a wrapper around Vector that has a nice default for the small size:</div><div>```</div><div>template <class T></div><div>using SmallVector = Vector<T, DefaultInlinedElementSize<T>>;</div><div>```</div><br class=""><blockquote type="cite" class=""><div class=""><div dir="ltr" class=""><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div style="overflow-wrap: break-word;" class=""><div class=""><div class=""> - If people are concerned about the default being bad, then you could choose to make “SmallVector<T, -1>” be the thing that autosizes. I tend to agree with your viewpoint that the N chosen is arbitrary in almost all cases anyway though, so I wouldn’t go this way.</div><br class=""><blockquote type="cite" class=""><div class=""><div dir="ltr" class=""><div class=""><br class="">Vec<T> is a convenience alias for SmallVector<T, 0>. It lets us get the (little-known?) benefits of SmallVector even when it has no inline elements (see <a href="https://llvm.org/docs/ProgrammersManual.html#llvm-adt-smallvector-h" target="_blank" class="">https://llvm.org/docs/ProgrammersManual.html#llvm-adt-smallvector-h</a>). The recommendation in the patch is to use this when the SmallVector is on the heap.</div></div></div></blockquote><div class=""><br class=""></div><div class="">These benefits are tiny, I really don’t think it is worth introducing a new name for this. I think this is better handled with a change to CodingStandards (saying “don’t use std::vector”) and the programmers manual.</div><br class=""></div><div class="">-Chris</div><br class=""></div></blockquote></div></div>
</div></blockquote></div><br class=""></body></html>