Early CSE clobbering llvm.assume

Project Infrastructure LLVM Dev List Archives
1

As of llvm 3.8, the early CSE pass seems to remove llvm.assume intrinsics. Is this the expected behavior?

I’ve attached as small-ish example of this happening in my production code.

$ opt -early-cse before-early-cse.ll -S > after-early-cse.ll

Note the use of the assume intrinsic indicating that the loaded value %channels equals 3. In a later pass I replace the load instruction with the constant value. This approach worked in llvm 3.7. I can solve the issue by moving my pass before early CSE if need be.

v/r,
Josh

after-early-cse.ll (3.81 KB)

before-early-cse.ll (4.51 KB)

2

On second thought, I don’t have a workaround. I had been relying on EarlyCSE + GVN to combine %channels and %channels4, which I would then replace with the constant 3. If I run my optimization before these passes then I have no way of knowing %channels4 can be replaced with the constant 3.

3

Thanks for the lead Kevin. Unfortunately when I updated to ToT the problem persists. Will put together a minimum reproducing example.

4

Here’s a minimum reproducing example (attached and inline below):

define i32 @foo(i32*) {
entry:
%1 = load i32, i32* %0
%2 = icmp eq i32 %1, 3
call void @llvm.assume(i1 %2)
%3 = load i32, i32* %0

ret i32 %3
}

Both -early-cse and -gvn independently simplify this IR to:

define i32 @foo(i32*) {
entry:
%1 = load i32, i32* %0
ret i32 %1
}

Whereas I think it would be more correct to simplify it to:

define i32 @foo(i32*) {
entry:
%1 = load i32, i32* %0
%2 = icmp eq i32 %1, 3
call void @llvm.assume(i1 %2)
ret i32 %1
}

And then:

define i32 @foo(i32*) {
entry:

ret i32 3
}

Both -early-cse and -gvn appear capable of this final optimization, but get stuck doing the wrong thing when given the initial code with the redundant load.

v/r,
Josh

assume_constant.ll (318 Bytes)

5

Yeah, that change is completely unrelated, that is about correctness, this is about optimization.
I’m working on a proposal to just fix assume at some point to deal with the former issue.

The problem with this testcase is that all the ways assume is propagate expect the variable in the assume to later be used.

bool GVN::processAssumeIntrinsic(IntrinsicInst *Inst) {

…

// We can replace assume value with true, which covers cases like this:
// call void @llvm.assume(i1 %cmp)
// br i1 %cmp, label %bb1, label %bb2 ; will change %cmp to true
ReplaceWithConstMap[V] = True;

…
}

…
bool GVN::processBlock(BasicBlock *BB) {

…

// Clearing map before every BB because it can be used only for single BB.
ReplaceWithConstMap.clear();

…

}

So it’s going to go through the rest of the bb, see nothing with %2, do nothing, and then next iteration, clear the constant map. It’s not valid to avoid clearing the constant map, and in fact, what is happening here is a pretty complicated to help with.
There is no easy way to see that the load at %3 is affected at all by the assume.

It’s possible to make this work using the predication/value inference algorithms in the paper newgvn is based on, but it’s not even implemented there.

Short answer, without special casing this in magic ways, i wouldn’t expect this to get fixed anytime soon.

If we fixed assume in one of the ways i thought about, like bodik’s extended ssa:
http://homepages.dcc.ufmg.br/~fernando/classes/dcc888/ementa/slides/RangeAnalysis.pdf

You would at least see that the load result is used by an assume, and could go look at that assume and so something with it. Currently, it’s a few steps away.

In the current scheme, assumes just float in air, and so it can be hard to see what their effects touch
:slight_smile:

6

Thanks Daniel, with that knowledge I think I can at least work around the issue in my frontend.

Ignoring GVN for a second though, and just looking at Early CSE, it seems to me that at least in this pass that there is the potential for an easy fix. Here the issue appears to be that we hit

if (Value *V = SimplifyInstruction(Inst, DL, &TLI, &DT, &AC))

immediately replacing %1 with 3 before we even reach %3. If we were to record this replacement in EarlyCSE::AvailableValues, wouldn’t that address the issue? I’ll try this out and see.

v/r,
Josh

7

Maybe. It may not fix it directly because you never use %1 or %2 again. I haven’t looked to see how good the lookup is.

8

(and you pointed out one of the other problems with current assume, which is that by not attaching the equality to the assume, and saying nothing should touch it, they can get messed up indiscriminately by random things. So we often can lose valuable info for no good reason)

9

Turns out the lookup isn’t good enough, the SimpleValue key type for the AvailableValues map doesn’t support LoadInst.

10

:frowning:
This, at least, is something newgvn fixes.

Loads, stores, scalars, insertvalues, etc, are all just expression types. They all go into the value table. any required memory state, etc, that goes along with it is just a thing that is part of that expression.

But even that won’t help in this case. It will only notice that

In any case, for now, you should file a bug pointing out this use of assume can’t be used to optimize, and assign/cc me.
It’ll be a while, and it’ll like get reassigned, but it’s good for us to not lose track of this.

(Of course, if you have time to help with the more complex problems here, happy to try to set you on the right path :P)

FWIW: Newgvn comes up with:

define i32 @foo(i32*) {
entry:
%1 = load i32, i32* %0
%2 = icmp eq i32 %1, 3
call void @llvm.assume(i1 %2)
ret i32 %1
}

It has no assume handling, it’s still a todo.

Humorously, gvn will already turn the above into

define i32 @foo(i32*) {
entry:
ret i32 3
}

11

https://llvm.org/bugs/show_bug.cgi?id=28086

Thank you so much for the helpful responses. Is the new GVN algorithm available in trunk? If so I couldn’t find it, and I’m curious to try the NewGVN-before-GVN workaround.