<html><head><meta http-equiv="Content-Type" content="text/html charset=us-ascii"></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;" class="">On 01 Jan 2016, at 00:07, James Y Knight <<a href="mailto:jyknight@google.com" class="">jyknight@google.com</a>> wrote:<div><blockquote type="cite" class=""><div class=""><meta http-equiv="Content-Type" content="text/html charset=us-ascii" class=""><div style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;" class=""><div class=""><div class="">On Dec 31, 2015, at 5:03 AM, Dimitry Andric via cfe-dev <<a href="mailto:cfe-dev@lists.llvm.org" class="">cfe-dev@lists.llvm.org</a>> wrote:</div><blockquote type="cite" class=""><div class="">This thread unfortunately died out without any resolution, so we are now<br class="">nearing 3.8 branching, and the problem is still not fixed. E.g., the<br class="">only custom patch we have now left in FreeBSD's version of llvm (which<br class="">we would really like to get rid of!) is the reversal of r240144: it is a<br class="">pretty gross workaround, but it works for now.<br class=""><br class="">But is there any way that we can get this resolved before 3.8.0 ships,<br class="">please? :-)<br class=""></div></blockquote></div><div class=""><br class=""></div><div class="">I believe the summary of the thread is that LLVM has a fundamentally unsound transformation: automatically increasing the required alignment of an externally visible global variable.</div></div></div></blockquote><div><br class=""></div><div>Yes, and also that it ignores (or silently drops) an attribute added by the developer, e,g.:</div><div><br class=""></div><div><font face="Menlo" class="">__attribute__((__aligned__(8))) char cout[sizeof(ostream)];</font></div><div><br class=""></div><div>gets transformed into:</div><div><br class=""></div><div><font face="Menlo" class="">@cout = global [24 x i8] zeroinitializer, align 16</font></div><div><br class=""></div><div><br class=""></div><blockquote type="cite" class=""><div style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;" class=""><div class="">In enforceKnownAlignment in lib/Transforms/Utils/Local.cpp, it checks isStrongDefinitionForLinker() purportedly to ensure that the "memory we set aside for the global" is "the memory used by the final program.". However, even if said predicate returns true, that condition is NOT guaranteed, at least on ELF platforms, because of copy relocations.</div></div></blockquote><div><br class=""></div><div>Ah yes, I get what you mean. When isStrongDefinitionForLinker() returns false in this case, the code falls through to the default global object handling code below there, which attempts to increase the alignment the the preferred alignment. One thing I don't completely understand though, is the comment slightly below it:</div><div><br class=""></div><div><div><font face="Menlo" class=""> // We can only increase the alignment of the global if it has no alignment</font></div><div><font face="Menlo" class=""> // specified or if it is not assigned a section. If it is assigned a</font></div><div><font face="Menlo" class=""> // section, the global could be densely packed with other objects in the</font></div><div><font face="Menlo" class=""> // section, increasing the alignment could cause padding issues.</font></div><div><font face="Menlo" class=""> if (!GO->hasSection() || GO->getAlignment() == 0) {</font></div><div><br class=""></div><div>However, the object certainly does have a specified alignment. So how is that supposed to work, then? :)</div><div><br class=""></div><div><br class=""></div></div><blockquote type="cite" class=""><div style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;" class=""><div class="">(There's also very similar-looking alignment-fixing code in CodeGenPrepare::optimizeCallInst, too, I'm not sure why it's duplicated there.)</div><div class=""><br class=""></div><div class="">Basically, we need to not modify the alignment if:</div><div style="margin: 0px; font-size: 11px; font-family: Menlo;" class=""><div style="margin: 0px; font-size: 11px;" class=""><span style="font-variant-ligatures: no-common-ligatures; color: #34bbc7" class=""><b class="">if</b></span> (Subtarget->isTargetELF() &&</div><div style="margin: 0px; font-size: 11px;" class=""> TM.getRelocationModel() == <span style="font-variant-ligatures: no-common-ligatures; color: #d53bd3" class="">Reloc</span>::PIC_ &&</div><div style="margin: 0px; font-size: 11px;" class=""> GV->hasDefaultVisibility() && !GV->hasLocalLinkage())</div><div style="margin: 0px; font-size: 11px;" class=""><br class=""></div><div class=""><span style="font-family: Helvetica;" class="">The above code snippet appears in X86FastISel.cpp, and equivalent-but-not-identical code appears in several other places, to determine whether a PLT load for the variable is required.</span></div><div class=""><span style="font-family: Helvetica;" class=""><br class=""></span></div><div class=""><span style="font-family: Helvetica;" class="">I'm not sure if isStrongDefinitionForLinker itself ought to be taking that into account, or if there should be a different new predicate for that check. isReallyActuallyTrulyStrongDefinitionForLinker()? :)</span></div></div></div></blockquote></div><div><br class=""></div><div>It is probably safest to use a new predicate, since isStrongDefinitionForLinker() is called all over the place, also in target-dependent code. Or does this really apply to everything with ELF? I mean, is the above set of conditions callable in the context of isStrongDefinitionForLinker()?</div><div><br class=""></div><div>-Dimitry</div><div><br class=""></div></body></html>