<div dir="ltr">Can someone fill me in on the issue with the dominator tree, precisely, during inlining?<div>We now have the capability of quickly keeping it up to date without too much trouble (it may require pushing it through a bunch of places, but the actual changes to do should be easy).</div><div><br></div><div>Was the original issue cost of rebuilding it repeatedly during inlining, or what?<br><br></div></div><div class="gmail_extra"><br><div class="gmail_quote">On Mon, Aug 7, 2017 at 7:56 AM, Sander De Smalen via llvm-dev <span dir="ltr"><<a href="mailto:llvm-dev@lists.llvm.org" target="_blank">llvm-dev@lists.llvm.org</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div bgcolor="white" lang="EN-GB" link="blue" vlink="purple">
<div class="m_-8006846865748462121WordSection1">
<p class="MsoNormal"><span>Hi,<u></u><u></u></span></p>
<p class="MsoNormal"><span><u></u> <u></u></span></p>
<p class="MsoNormal"><span>Coincidentally I've been working to optimize this same case last week. I was struggling a bit to determine where to put this functionality and eventually went for the pragmatic approach of creating
an experimental pass. Probably not the eventual solution, but it may provide some useful input to the discussion here.<u></u><u></u></span></p>
<p class="MsoNormal"><span><u></u> <u></u></span></p>
<p class="MsoNormal"><span>Basically, I experimented with a 'pre-inlining-transform' pass that clones the callsite if it can prove that one of the arguments to the call can be replaced by a constant, based on dominating conditions,
e.g.:<u></u><u></u></span></p>
<p class="MsoNormal"><span><u></u> <u></u></span></p>
<p class="MsoNormal"><span style="font-family:"Andale Mono",sans-serif"> if (!ptr || ptr && ptr->val)<u></u><u></u></span></p>
<p class="MsoNormal"><span style="font-family:"Andale Mono",sans-serif"> foo(ptr, ...)<u></u><u></u></span></p>
<p class="MsoNormal"><span style="font-family:"Andale Mono",sans-serif"><u></u> <u></u></span></p>
<p class="MsoNormal"><span style="font-family:"Andale Mono",sans-serif"> =><u></u><u></u></span></p>
<p class="MsoNormal"><span style="font-family:"Andale Mono",sans-serif"> <u></u><u></u></span></p>
<p class="MsoNormal"><span style="font-family:"Andale Mono",sans-serif"> if (!ptr)<u></u><u></u></span></p>
<p class="MsoNormal"><span style="font-family:"Andale Mono",sans-serif"> foo(nullptr, ...)<u></u><u></u></span></p>
<p class="MsoNormal"><span style="font-family:"Andale Mono",sans-serif"> else if (ptr && ptr->val)<u></u><u></u></span></p>
<p class="MsoNormal"><span style="font-family:"Andale Mono",sans-serif"> foo(ptr /*knownNonNull*/, ...)<u></u><u></u></span></p>
<p class="MsoNormal"><span><u></u> <u></u></span></p>
<p class="MsoNormal"><span>Here the first argument becomes constant for the first callsite and a further analysis pass sets the KnownNonNull attribute on the first argument in the second callsite. The InlinerCost algorithm
can then determine it is cheap enough to inline both cases, because it knows the callee (foo) distinguishes between the two cases. In the callee, the check for '(ptrA == ptrB)' in function foo becomes constant for the first callsite because it knows ptrA is
nullptr.<u></u><u></u></span></p>
<p class="MsoNormal"><span><u></u> <u></u></span></p>
<p class="MsoNormal"><span>To keep compile-time down, it doesn’t use the dominator tree as it seemed sufficient to look at the direct predecessors of the block containing the call (and their single-predecessors, for 'and'ed
conditions). To keep the cost of duplicating code down, it only clones the block upto the call if the number of instructions stays below some threshold, which in practice reduces the cases to simple callsites with directly dominating conditions. The reasoning
behind this is that duplicating the callsite is probably ‘cheap’ if it can eliminate an argument with a constant, since this is often a pointer and is likely to be checked for ‘nullptr’ in the callee which will improve the inline cost.<u></u><u></u></span></p>
<p class="MsoNormal"><span><u></u> <u></u></span></p>
<p class="MsoNormal"><span>We’re still collecting numbers to check there are no regressions from cloning the callsite and the pass is still a bit work in progress, but I should be able to share this work if anyone is interested
(if only just for reference).<u></u><u></u></span></p>
<p class="MsoNormal"><span><u></u> <u></u></span></p>
<p class="MsoNormal"><span>Sander<u></u><u></u></span></p>
<p class="MsoNormal"><span><u></u> <u></u></span></p>
<div style="border:none;border-top:solid #b5c4df 1.0pt;padding:3.0pt 0in 0in 0in">
<p class="MsoNormal"><b><span style="font-size:12.0pt;color:black">From: </span></b><span style="font-size:12.0pt;color:black">llvm-dev <<a href="mailto:llvm-dev-bounces@lists.llvm.org" target="_blank">llvm-dev-bounces@lists.llvm.<wbr>org</a>> on behalf of Sean Silva via llvm-dev <<a href="mailto:llvm-dev@lists.llvm.org" target="_blank">llvm-dev@lists.llvm.org</a>><br>
<b>Reply-To: </b>Sean Silva <<a href="mailto:chisophugis@gmail.com" target="_blank">chisophugis@gmail.com</a>><br>
<b>Date: </b>Friday, 4 August 2017 at 23:07<span class=""><br>
<b>To: </b>Chad Rosier <<a href="mailto:mcrosier@codeaurora.org" target="_blank">mcrosier@codeaurora.org</a>><br>
<b>Cc: </b>llvm-dev <<a href="mailto:llvm-dev@lists.llvm.org" target="_blank">llvm-dev@lists.llvm.org</a>><br>
<b>Subject: </b>Re: [llvm-dev] [RFC][InlineCost] Modeling JumpThreading (or similar) in inline cost model<u></u><u></u></span></span></p>
</div>
<div>
<p class="MsoNormal"><u></u> <u></u></p>
</div>
<div>
<p class="MsoNormal"><u></u> <u></u></p>
<div>
<p class="MsoNormal"><u></u> <u></u></p>
<div>
<p class="MsoNormal">On Fri, Aug 4, 2017 at 11:41 AM, Chad Rosier via llvm-dev <<a href="mailto:llvm-dev@lists.llvm.org" target="_blank">llvm-dev@lists.llvm.org</a>> wrote:<u></u><u></u></p><div><div class="h5">
<blockquote style="border:none;border-left:solid #cccccc 1.0pt;padding:0in 0in 0in 6.0pt;margin-left:4.8pt;margin-right:0in">
<div>
<p><u></u> <u></u></p>
<p class="MsoNormal"><u></u> <u></u></p>
<div>
<p class="MsoNormal">On 8/4/2017 2:06 PM, Daniel Berlin wrote:<u></u><u></u></p>
</div>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<div>
<p class="MsoNormal">A few notes:<br>
I'm a bit surprised IPO copy/constant propagation doesn't get this case, but i didn't look if the lattice supports variables.
<u></u><u></u></p>
<div>
<p class="MsoNormal">In particular, in your example, given no other call sites, it should eliminate the dead code.<u></u><u></u></p>
</div>
<div>
<p class="MsoNormal">(In a real program, it may require cloning).<u></u><u></u></p>
</div>
</div>
</blockquote>
<p class="MsoNormal"><br>
In the actual program (SPEC2017/gcc, ironically), there are multiple calls to fn2 and only one of them has the property that the 1st and 2nd argument are the same (as is shown in my pseudo code). Internally, we have another developer, Matt Simpson, working
on a function specialization patch that might be of value here. Specifically, you could clone fn2 based on the fact that a_ptr == dst_ptr and then simplify a great deal of the function. However, that patch is still a WIP.<br>
<br>
<br>
<u></u><u></u></p>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<div>
<div>
<p class="MsoNormal">GCC will do IPA-CP/const-prop with cloning, and i'm wildly curious if new GCC's catch this case for you at higher optimization levels?<u></u><u></u></p>
</div>
</div>
</blockquote>
<p class="MsoNormal"><br>
GCC does inline fn2 into fn1 in this particular case, but I'm not exactly sure how GCC accomplishes this. I'm guessing GCC is just more aggressive with its inlining (fn2 is also marked with the inline keyword, which I assume GCC uses as a hint). I'm speculating
here and I've never worked on GCC, so unfortunately I have little to go on.<br>
<br>
<br>
<u></u><u></u></p>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<div>
<div>
<p class="MsoNormal">If so, it may be worth not looking at this as an inlining problem, but as an area we need IPO infrastructure improvement
<u></u><u></u></p>
</div>
</div>
</blockquote>
<p class="MsoNormal"><br>
Because of the multiple callsites with varying characteristics I'm not sure this can be solved in this way.<br>
<br>
<br>
<u></u><u></u></p>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<div>
<div>
<div>
<p class="MsoNormal"><u></u> <u></u></p>
</div>
<div>
<p class="MsoNormal">Otherwise, a couple things: <u></u><u></u></p>
<div>
<div>
<p class="MsoNormal">Approximate dominators (for example, semi-dominators) can be computed fast (a DFS walk of the CFG with no real additional computation)<u></u><u></u></p>
</div>
<div>
<p class="MsoNormal">Except in strange CFGs that jump around a lot, they are the dominators.<u></u><u></u></p>
</div>
<div>
<p class="MsoNormal"><u></u> <u></u></p>
</div>
<div>
<p class="MsoNormal">More importantly, the dominator is either the sdom or a proper ancestor of the sdom.<u></u><u></u></p>
</div>
<div>
<p class="MsoNormal"><u></u> <u></u></p>
</div>
<div>
<p class="MsoNormal">The practical impact of this is that if you use them as if they were dominators, the set of conditions you discover will not be "too wrong". Occasionally wrong, but mostly not.<u></u><u></u></p>
</div>
<div>
<p class="MsoNormal"><u></u> <u></u></p>
</div>
<div>
<p class="MsoNormal">My guess is the cost of doing approximate dominators is ~50-60% of the cost of doing dominators. Nowadays, about half the time was in the DFS walk, the other half in the computation. At best, it would be 2-3x faster.<u></u><u></u></p>
</div>
<div>
<p class="MsoNormal">I've no idea if this changes whether we'd want dominators, approximate dominators, or stick with nothing.<u></u><u></u></p>
</div>
</div>
</div>
</div>
</div>
</blockquote>
<p class="MsoNormal"><br>
Right, this is kinda one of the bigger questions I'm trying to figure out. My proposed solution doesn't use the dominator tree in order to minimize the impact on compile-time. <u></u><u></u></p>
</div>
</blockquote>
<div>
<p class="MsoNormal"><u></u> <u></u></p>
</div>
<div>
<p class="MsoNormal">In the new PM, it's possible for a CGSCC pass to get a cached function analysis from functions that have been visited previously. This requires the CGSCC iteration on SCC's we visited earlier to keep an up to date domtree, and I don't know
if that's the case (or how much work it would be to make it the case).<u></u><u></u></p>
</div>
<div>
<p class="MsoNormal"><u></u> <u></u></p>
</div>
<div>
<p class="MsoNormal">-- Sean Silva<u></u><u></u></p>
</div>
<div>
<p class="MsoNormal"> <u></u><u></u></p>
</div>
<blockquote style="border:none;border-left:solid #cccccc 1.0pt;padding:0in 0in 0in 6.0pt;margin-left:4.8pt;margin-right:0in">
<div>
<p class="MsoNormal">However, I'd guess the ROI is going to be much smaller because of the limited scope. On the other end of the spectrum I'd fear the full dominator tree would be too computationally expensive (but of course some of that could be mitigated
by the ability to do incremental updates to the dominator tree).<br>
<br>
<br>
<u></u><u></u></p>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<div>
<div>
<div>
<div>
<div>
<p class="MsoNormal"><u></u> <u></u></p>
</div>
<div>
<p class="MsoNormal">If you have some sort of dominatorish tree could then just use earlycse's method of dominating condition finding:<u></u><u></u></p>
</div>
<div>
<p class="MsoNormal">Process in "dom tree" top-down order, push the equivalences you see, pop the relevant ones when you exit the relevant dom tree scope.<u></u><u></u></p>
</div>
<div>
<p class="MsoNormal"><u></u> <u></u></p>
</div>
<div>
<p class="MsoNormal">In practice, you'd only check comparisons against the hash table.<u></u><u></u></p>
</div>
</div>
</div>
</div>
</div>
</blockquote>
<p class="MsoNormal"><br>
Humm.. I'll have to think about it for a bit. I'm thinking this might be a good compromise for my needs.<br>
<br>
<br>
<u></u><u></u></p>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<div>
<div>
<div>
<div>
<div>
<p class="MsoNormal"><u></u> <u></u></p>
</div>
<div>
<p class="MsoNormal">The other option is PredicateInfo, but it requires dominators and modifies the IR.<u></u><u></u></p>
</div>
<div>
<p class="MsoNormal"><u></u> <u></u></p>
</div>
<div>
<p class="MsoNormal">My guess is this is undesired/too heavyweight for inline cost analysis, however the basic principle on how it renames things could also be applied without IR changing for this specific case. Unlike the EarlyCSE method, which is O(all instructons)
PredicateInfo is O(branches + number of uses of variables affected by conditions) Without going into futher details, if all you care about is "for each condition, give me the set of possibly affected variables" (so you can see if they may simplify), we could
do that very very quickly (as fast as we can sort a vector). But it does require dominators.<u></u><u></u></p>
</div>
</div>
</div>
</div>
</div>
</blockquote>
<p class="MsoNormal"><br>
For my particular problem, I think PredicateInfo would be sufficient IIUYC. But as you suggest, I'm thinking people aren't going to be fond of using the full dominators.<br>
<br>
Lots of great feedback. Thanks, Danny. <u></u><u></u></p>
<div>
<div>
<p class="MsoNormal"><br>
<br>
<br>
<u></u><u></u></p>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<div>
<div>
<div>
<div>
<div>
<p class="MsoNormal"><u></u> <u></u></p>
</div>
<div>
<p class="MsoNormal"> <u></u><u></u></p>
</div>
</div>
</div>
</div>
</div>
<div>
<p class="MsoNormal"><u></u> <u></u></p>
<div>
<p class="MsoNormal">On Fri, Aug 4, 2017 at 8:56 AM, Chad Rosier <<a href="mailto:mcrosier@codeaurora.org" target="_blank">mcrosier@codeaurora.org</a>> wrote:<u></u><u></u></p>
<blockquote style="border:none;border-left:solid #cccccc 1.0pt;padding:0in 0in 0in 6.0pt;margin-left:4.8pt;margin-right:0in">
<div>
<p><tt><span style="font-size:10.0pt">All,</span></tt><u></u><u></u></p>
<p><tt><span style="font-size:10.0pt">I'm working on an improvement to the inline cost model, but I'm unsure how to proceed. Let me begin by first describing the problem I'm trying to solve. Consider the following pseudo C code:</span></tt><u></u><u></u></p>
<p><b><span style="font-family:"Courier New",serif">typedef struct element {<br>
unsigned idx;<br>
} element_t;</span></b><u></u><u></u></p>
<p><b><span style="font-family:"Courier New",serif">static inline<br>
unsigned char fn2 (element_t *dst_ptr, const element_t *a_ptr,<br>
const element_t *b_ptr, unsigned char changed) {<br>
if (a_ptr && b_ptr && a_ptr->idx == b_ptr->idx) {<br>
if (!changed && dst_ptr && dst_ptr->idx == a_ptr->idx) {<br>
/* Do something */<br>
} else {<br>
changed = 1;<br>
if (!dst_ptr)<br>
dst_ptr = fn3();<br>
else<br>
dst_ptr->idx = a_ptr->idx;<br>
/* Do something. */<br>
}<br>
} else {<br>
changed = fn4();<br>
}<br>
return changed;<br>
}<br>
<br>
unsigned char fn1 (element_t *a_ptr, element_t *b_ptr) {<br>
unsigned char changed = 0;<br>
while (b_ptr) {<br>
if (!a_ptr || a_ptr->idx == b_ptr->idx) {<br>
changed = fn2 (a_ptr, a_ptr, b_ptr, changed);<br>
b_ptr = b_ptr->next;<br>
}<br>
}<br>
return changed;<br>
}</span></b><u></u><u></u></p>
<p><tt><span style="font-size:10.0pt">When the inline cost model computes the inline cost of fn2 it ends up being much higher than the inline threshold. A fair amount of the cost is due to the inlining of fn3 into fn2. However, if fn2 had been inlined into
fn1 the code from fn3 would have been removed as dead/unreachable.</span></tt><u></u><u></u></p>
<p><tt><span style="font-size:10.0pt">At the fn2 call site notice the first two arguments are equal. Thus, in the context of fn2 dst_ptr and a_ptr are equal. The call site of fn3 is predicated on dst_ptr being null (i.e., if (!dst_ptr) dst_ptr = fn3()), but
that code is predicated on a_ptr being non-null. Therefore, we know the condition !dst_ptr is false (because a_ptr == dst_ptr and a_ptr is non-null) and the call to fn3 is dead. I suspect one of JumpThreading, EarlyCSE, or GVN does the elimination after
inlining, so that's what I'd like to try and model in the inline cost model. (Note fn2 has multiple call sides and the property that the first and second arguments are equal isn't true for each call site, so something like IPSCCP doesn't actually help, AFAICT).</span></tt><u></u><u></u></p>
<p><tt><span style="font-size:10.0pt">My first attempt at solving this problem did something similar to what is done in JumpThreadingPass::<wbr>ProcessImpliedCondition(). Specifically, it tried to prove that dst_ptr was non-null based on a dominating condition.
The only tricky parts were to deal with hammocks/diamonds when walking up the CFG (See:
<a href="https://reviews.llvm.org/D36287" target="_blank">https://reviews.llvm.org/<wbr>D36287</a> as a concrete example of how I proposed to get an immediate dominator without the domtree) and to account for the fact that dst_ptr and a_ptr are equal.</span></tt><u></u><u></u></p>
<p><tt><span style="font-size:10.0pt">I'm pretty sure I can get this approach to work, however, I'm not convinced it's really extensible or general. Should we consider using the full dominator tree in the inline cost model to capture this?</span></tt><u></u><u></u></p>
<p><tt><span style="font-size:10.0pt">If you have any thoughts on how to tackle the problem, I would love to hear your feedback!</span></tt><u></u><u></u></p>
<p><tt><span style="font-size:10.0pt;color:#888888"> Chad</span></tt><span style="color:#888888"><u></u><u></u></span></p>
</div>
</blockquote>
</div>
<p class="MsoNormal"><u></u> <u></u></p>
</div>
</blockquote>
<p class="MsoNormal"><u></u> <u></u></p>
</div>
</div>
</div>
<p class="MsoNormal" style="margin-bottom:12.0pt"><br>
______________________________<wbr>_________________<br>
LLVM Developers mailing list<br>
<a href="mailto:llvm-dev@lists.llvm.org" target="_blank">llvm-dev@lists.llvm.org</a><br>
<a href="http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev" target="_blank">http://lists.llvm.org/cgi-bin/<wbr>mailman/listinfo/llvm-dev</a><u></u><u></u></p>
</blockquote>
</div></div></div>
<p class="MsoNormal"><u></u> <u></u></p>
</div>
</div>
</div>
IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose,
or store or copy the information in any medium. Thank you.
</div>
<br>______________________________<wbr>_________________<br>
LLVM Developers mailing list<br>
<a href="mailto:llvm-dev@lists.llvm.org">llvm-dev@lists.llvm.org</a><br>
<a href="http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev" rel="noreferrer" target="_blank">http://lists.llvm.org/cgi-bin/<wbr>mailman/listinfo/llvm-dev</a><br>
<br></blockquote></div><br></div>