On Tue, 30 Jun 2020 at 22:30, Paul E. McKenney <paulmck@xxxxxxxxxx> wrote: > On Tue, Jun 30, 2020 at 10:12:43PM +0200, Peter Zijlstra wrote: > > On Tue, Jun 30, 2020 at 09:19:31PM +0200, Marco Elver wrote: > > > First of all, I agree with the concerns, but not because of LTO. > > > > > > To set the stage better, and summarize the fundamental problem again: > > > we're in the unfortunate situation that no compiler today has a way to > > > _efficiently_ deal with C11's memory_order_consume > > > [https://lwn.net/Articles/588300/]. If we did, we could just use that > > > and be done with it. But, sadly, that doesn't seem possible right now -- > > > compilers just say consume==acquire. > > > > I'm not convinced C11 memory_order_consume would actually work for us, > > even if it would work. That is, given: > > > > https://lore.kernel.org/lkml/20150520005510.GA23559@xxxxxxxxxxxxxxxxxx/ > > > > only pointers can have consume, but like I pointed out, we have code > > that relies on dependent loads from integers. > > I agree that C11 memory_order_consume is not normally what we want, > given that it is universally promoted to memory_order_acquire. > > However, dependent loads from integers are, if anything, more difficult > to defend from the compiler than are control dependencies. This applies > doubly to integers that are used to index two-element arrays, in which > case you are just asking the compiler to destroy your dependent loads > by converting them into control dependencies. > > > > Will suggests doing the same in the > > > kernel: https://lkml.kernel.org/r/20200630173734.14057-19-will@xxxxxxxxxx > > > > PowerPC would need a similar thing, it too will not preserve causality > > for control dependecies. > > > > > What we're most worried about right now is the existence of compiler > > > transformations that could break data dependencies by e.g. turning them > > > into control dependencies. > > > > Correct. > > > > > If this is a real worry, I don't think LTO is the magical feature that > > > will uncover those optimizations. If these compiler transformations are > > > real, they also exist in a normal build! > > > > Agreed, _however_ with the caveat that LTO could make them more common. > > > > After all, with whole program analysis, the compiler might be able to > > more easily determine that our pointer @ptr is only ever assigned the > > values of &A, &B or &C, while without that visibility it would not be > > able to determine this. > > > > Once it knows @ptr has a limited number of determined values, the > > conversion into control dependencies becomes much more likely. > > Which would of course break dependent loads. > > > > And if we are worried about them, we need to stop relying on dependent > > > load ordering across the board; or switch to -O0 for everything. > > > Clearly, we don't want either. > > > > Agreed. > > > > > Why do we think LTO is special? > > > > As argued above, whole-program analysis would make it more likely. But I > > agree the fundamental problem exists independent from LTO. > > > > > But as far as we can tell, there is no evidence of the dreaded "data > > > dependency to control dependency" conversion with LTO that isn't there > > > in non-LTO builds, if it's even there at all. Has the data to control > > > dependency conversion been encountered in the wild? If not, is the > > > resulting reaction an overreaction? If so, we need to be careful blaming > > > LTO for something that it isn't even guilty of. > > > > It is mostly paranoia; in a large part driven by the fact that even if > > such a conversion were to be done, it could go a very long time without > > actually causing problems, and longer still for such problems to be > > traced back to such an 'optimization'. > > > > That is, the collective hurt from debugging too many ordering issues. > > > > > So, we are probably better off untangling LTO from the story: > > > > > > 1. LTO or no LTO does not matter. The LTO series should not get tangled > > > up with memory model issues. > > > > > > 2. The memory model question and problems need to be answered and > > > addressed separately. > > > > > > Thoughts? > > > > How hard would it be to creates something that analyzes a build and > > looks for all 'dependent load -> control dependency' transformations > > headed by a volatile (and/or from asm) load and issues a warning for > > them? I was thinking about this, but in the context of the "auto-promote to acquire" which you didn't like. Issuing a warning should certainly be simpler. I think there is no one place where we know these transformations happen, but rather, need to analyze the IR before transformations, take note of all the dependent loads headed by volatile+asm, and then run an analysis after optimizations checking the dependencies are still there. > > This would give us an indication of how valuable this transformation is > > for the kernel. I'm hoping/expecting it's vanishingly rare, but what do > > I know. > > This could be quite useful! We might then even be able to say, "if you get this warning, turn on CONFIG_ACQUIRE_READ_DEPENDENCIES" (or however the option will be named). Or some other tricks, like automatically recompile the TU where this happens with the option. But again, this is not something that should specifically block LTO, because if we have this, we'll need to turn it on for everything. Thanks, -- Marco
