[llvm-dev] Load combine pass

Wed Sep 25 01:10:13 PDT 2019

On Wed, Sep 25, 2019 at 3:12 AM Philip Reames via llvm-dev
<llvm-dev at lists.llvm.org> wrote:
>
> If we do load combining at the IR level, one thing we'll need to give some thought to is atomicity.  Combining two atomic loads into a wider (legal) atomic load is not a reversible transformation given our current specification.
Can we not just not combine non-simple (atomic/volatile) loads?

> I've been thinking about a concept I've been tentatively calling "element wise atomicity" which would make this a reversible transform by representing the load as either a vector or struct load on which the atomicity requirement was clearly expressed on the individual elements, not the vector/struct itself.  (We have no way to spell that today.)
>
> Not having load combining be a reversible transformation is really unfortunate.  Consider this small example:
>
> store i32 0, a+4
> load i32, a
> load i32, a+4
>
> If you happen to visit the load pair first, you really want to be able to split them again once you discover the store.
>
> Note that the same theoretical problem exists in MI/SelectionDAG.  We just do so much less memory optimization there we get away with mostly not caring.  :)
>
> Philip
Roman

> On 9/24/2019 4:50 PM, Artur Pilipenko via llvm-dev wrote:
>
> Implementing load combine/widening on the IR level sounds like a reasonable enhancement to me. Although, we (Azul) don't have any plans of doing that in the near future.
>
> Artur
>
> On 12 Sep 2019, at 00:58, Paweł Bylica <chfast at gmail.com> wrote:
>
> Ok, thanks.
>
> Are there any plans to reintroduce it on the IR level? I'm not confident this is strictly necessary, but in some cases not having load widening ends up really bad.
> Like in the case where vectorizer tries to do something about it:
> https://godbolt.org/z/60RuEw
> https://bugs.llvm.org/show_bug.cgi?id=42708
>
> At the current state I'm forced to use memset() to express uint64 load from an array of bytes.
>
> // P.
>
> On Thu, Sep 12, 2019 at 4:22 AM Artur Pilipenko <apilipenko at azul.com> wrote:
>>
>> Hi,
>>
>> Load widening/combine for the original pattern was implemented in DAG combiner. See:
>> https://github.com/llvm/llvm-project/blob/23bbeb52f392d88bf0d0527392a7a11561ee09c0/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp#L6872
>>
>> Later a similar transform was introduced fro stores:
>> https://github.com/llvm/llvm-project/blob/23bbeb52f392d88bf0d0527392a7a11561ee09c0/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp#L6700
>>
>> There is no load widening on the IR level as of now.
>>
>> Artur
>>
>> On 11 Sep 2019, at 04:47, Paweł Bylica <chfast at gmail.com> wrote:
>>
>> Hi,
>>
>> Can I ask what is the status of load widening. It seems there is no load widening on IR at all.
>>
>> // Paweł
>>
>> On Wed, Oct 5, 2016 at 1:49 PM Artur Pilipenko via llvm-dev <llvm-dev at lists.llvm.org> wrote:
>>>
>>> Philip and I talked about this is person. Given the fact that load widening in presence of atomics is irreversible transformation we agreed that we don't want to do this early. For now it can be implemented as a peephole optimization over machine IR. MIR is preferred here because X86 backend does GEP reassociation at MIR level and it can make information about addresses being adjacent available.
>>>
>>> Inline cost of the original pattern is a valid concern and we might want to revisit our decision later. But in order to do widening earlier we need to have a way to undo this transformation.
>>>
>>> I’m going to try implementing MIR optimization but not in the immediate future.
>>>
>>> Artur
>>>
>>> > On 28 Sep 2016, at 19:32, Artur Pilipenko <apilipenko at azulsystems.com> wrote:
>>> >
>>> > One of the arguments for doing this earlier is inline cost perception of the original pattern. Reading i32/i64 by bytes look much more expensive than it is and can prevent inlining of interesting function.
>>> >
>>> > Inhibiting other optimizations concern can be addressed by careful selection of the pattern we’d like to match. I limit the transformation to the case when all the individual have no uses other than forming a wider load. In this case it’s less likely to loose information during this transformation.
>>> >
>>> > I didn’t think of atomicity aspect though.
>>> >
>>> > Artur
>>> >
>>> >> On 28 Sep 2016, at 18:50, Philip Reames <listmail at philipreames.com> wrote:
>>> >>
>>> >> There's a bit of additional context worth adding here...
>>> >>
>>> >> Up until very recently, we had a form of widening implemented in GVN.  We decided to remove this in https://reviews.llvm.org/D24096 precisely because its placement in the pass pipeline was inhibiting other optimizations. There's also a major problem with doing widening at the IR level which is that widening a pair of atomic loads into a single wider atomic load can not be undone. This creates a major pass ordering problem of its own.
>>> >>
>>> >> At this point, my general view is that widening transformations of any kind should be done very late.  Ideally, this is something the backend would do, but doing it as a CGP like fixup pass over the IR is also reasonable.
>>> >>
>>> >> With that in mind, I feel both the current placement of LoadCombine (within the inliner iteration) and the proposed InstCombine rule are undesirable.
>>> >>
>>> >> Philip
>>> >>
>>> >>
>>> >> On 09/28/2016 08:22 AM, Artur Pilipenko wrote:
>>> >>> Hi,
>>> >>>
>>> >>> I'm trying to optimize a pattern like this into a single i16 load:
>>> >>> %1 = bitcast i16* %pData to i8*
>>> >>> %2 = load i8, i8* %1, align 1
>>> >>> %3 = zext i8 %2 to i16
>>> >>> %4 = shl nuw i16 %3, 8
>>> >>> %5 = getelementptr inbounds i8, i8* %1, i16 1
>>> >>> %6 = load i8, i8* %5, align 1
>>> >>> %7 = zext i8 %6 to i16
>>> >>> %8 = shl nuw nsw i16 %7, 0
>>> >>> %9 = or i16 %8, %4
>>> >>>
>>> >>> I came across load combine pass which is motivated by virtualliy the same pattern. Load combine optimizes the pattern by combining adjacent loads into one load and lets the rest of the optimizer cleanup the rest. From what I see on the initial review for load combine (https://reviews.llvm.org/D3580) it was not enabled by default because it caused some performance regressions. It's not very surprising, I see how this type of widening can obscure some facts for the rest of the optimizer.
>>> >>>
>>> >>> I can't find any backstory for this pass, why was it chosen to optimize the pattern in question in this way? What is the current status of this pass?
>>> >>>
>>> >>> I have an alternative implementation for it locally. I implemented an instcombine rule similar to recognise bswap/bitreverse idiom. It relies on collectBitParts (Local.cpp) to determine the origin of the bits in a given or value. If all the bits are happen to be loaded from adjacent locations it replaces the or with a single load or a load plus bswap.
>>> >>>
>>> >>> If the alternative approach sounds reasonable I'll post my patches for review.
>>> >>>
>>> >>> Artur
>>> >>
>>> >
>>>
>>> _______________________________________________
>>> LLVM Developers mailing list
>>> llvm-dev at lists.llvm.org
>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>>
>>
>
>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev