This was part of InstCombine, but it's better placed in
InstSimplify. InstCombine also had an unreachable but weaker
fold for insertelement with undef index, so that is deleted.
llvm-svn: 361559
This is reduced from a fuzzer test:
https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=14890
Usually, demanded elements should be able to simplify shuffle
mask elements that are pointing to undef elements of its source
operands, but that doesn't happen in the test case.
llvm-svn: 361533
We were turning roundss/sd/ps/pd intrinsics with immediates of 1 or 2 into
llvm.floor/ceil. The llvm.ceil/floor intrinsics are supposed to correspond
to the libm functions. For the libm functions we need to disable the
precision exception so the llvm.floor/ceil functions should always map to
encodings 0x9 and 0xA.
We had a mix of isel patterns where some used 0x9 and 0xA and others used
0x1 and 0x2. We need to be consistent and always use 0x9 and 0xA.
Since we have no way in isel of knowing where the llvm.ceil/floor came
from, we can't map X86 specific intrinsics with encodings 1 or 2 to it.
We could map 0x9 and 0xA to llvm.ceil/floor instead, but I'd really like
to see a use case and optimization advantage first.
I've left the backend test cases to show the blend we now emit without
the extra isel patterns. But I've removed the InstCombine tests completely.
llvm-svn: 361425
This should be a valid exception to the general rule of not creating new shuffle masks in IR...
because we already do it. :)
Also, DAG combining/legalization will undo this by widening the shuffle back out if needed.
Explanation for how we already do this: SLP or vector source can create chains of insert/extract
as shown in 1 of the examples from PR16739:
https://godbolt.org/z/NlK7rAhttps://bugs.llvm.org/show_bug.cgi?id=16739
And we expect instcombine or DAGCombine to clean that up by creating relatively simple shuffles.
Differential Revision: https://reviews.llvm.org/D62024
llvm-svn: 361338
Also, break out a helper function, namely foldFNegIntoConstant(...), which performs transforms common between visitFNeg(...) and visitFSub(...).
Differential Revision: https://reviews.llvm.org/D61693
llvm-svn: 361188
Summary:
In D61918 i was looking at dropping it in DAGCombiner `visitShiftByConstant()`,
but as @craig.topper pointed out, it was copied from here.
That check claims that the transform is illegal otherwise.
That isn't true:
1. For `ISD::ADD`, we only process `ISD::SHL` outer shift => sign bit does not matter
https://rise4fun.com/Alive/K4A
2. For `ISD::AND`, there is no restriction on constants:
https://rise4fun.com/Alive/Wy3
3. For `ISD::OR`, there is no restriction on constants:
https://rise4fun.com/Alive/GOH
3. For `ISD::XOR`, there is no restriction on constants:
https://rise4fun.com/Alive/ml6
So, why is it there then?
As far as i can tell, it dates all the way back to original check-in rL7793.
I think we should just drop it.
Reviewers: spatel, craig.topper, efriedma, majnemer
Reviewed By: spatel
Subscribers: llvm-commits, craig.topper
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D61938
llvm-svn: 361043
We have a similar match for patterns ending in a truncate. This
should be ok for all targets because the default expansion would
still likely be better from replacing 2 'and' ops with 1.
Attempt to show the logic equivalence in Alive (which doesn't
currently have funnel-shift in its vocabulary AFAICT):
%shamt = zext i8 %i to i32
%m = and i32 %shamt, 31
%neg = sub i32 0, %shamt
%and4 = and i32 %neg, 31
%shl = shl i32 %v, %m
%shr = lshr i32 %v, %and4
%or = or i32 %shr, %shl
=>
%a = and i8 %i, 31
%shamt2 = zext i8 %a to i32
%neg2 = sub i32 0, %shamt2
%and4 = and i32 %neg2, 31
%shl = shl i32 %v, %shamt2
%shr = lshr i32 %v, %and4
%or = or i32 %shr, %shl
https://rise4fun.com/Alive/V9r
llvm-svn: 360605
The worklist loop that we're returning back to should be able to do the repacement itself. This is how we normally do replacements.
My main motivation was that I observed that we weren't preserving the name of the result when we do this transform. The replacement code in the worklist loop will call takeName as part of the replacement.
Differential Revision: https://reviews.llvm.org/D61695
llvm-svn: 360284
(X | C1) + C2 --> (X | C1) ^ C1 iff (C1 == -C2)
I verified the correctness using Alive:
https://rise4fun.com/Alive/YNV
This transform enables the following transform that already exists in
instcombine:
(X | Y) ^ Y --> X & ~Y
As a result, the full expected transform is:
(X | C1) + C2 --> X & ~C1 iff (C1 == -C2)
There already exists the transform in the sub case:
(X | Y) - Y --> X & ~Y
However this does not trigger in the case where Y is constant due to an earlier
transform:
X - (-C) --> X + C
With this new add fold, both the add and sub constant cases are handled.
Patch by Chris Dawson.
Differential Revision: https://reviews.llvm.org/D61517
llvm-svn: 360185
Fundamentally/generally, we should not have to rely on bailouts/crippling of
folds. In this particular case, I think we always recognize the inverted
predicate min/max pattern, so there should not be any loss of optimization.
Codegen looks better because we are eliminating an fneg.
llvm-svn: 360180
We don't always get this:
Cond ? -X : -Y --> -(Cond ? X : Y)
...even with the legacy IR form of fneg in the case with extra uses,
and we miss matching with the newer 'fneg' instruction because we
are expecting binops through the rest of the path.
Differential Revision: https://reviews.llvm.org/D61604
llvm-svn: 360075
The demanded elts rules introduced for GEPs in https://reviews.llvm.org/rL356293 replaced vector constants with undefs (by design). It turns out that the LangRef disallows such cases when indexing structs. The right fix is probably to relax the langref requirement, and update other passes to expect the result, but for the moment, limit the transform to avoid compiler crashes.
This should fix https://bugs.llvm.org/show_bug.cgi?id=41624.
llvm-svn: 359633
Follow-up to:
rL359482
Avoid this potential problem throughout by giving the type a name
and verifying the assumption that both operands are the same type.
llvm-svn: 359485
PVS Studio's copy+paste recognizer was seeing this as a typo, technically Op0/Op1 in a fcmp should always be the same type, but we might as well avoid the issue.
Reported in https://www.viva64.com/en/b/0629/
llvm-svn: 359482
This patch rewrites the existing PACKSS/PACKUS constant folding code to expand as a generic expansion.
This is a first NFCI step toward expanding PACKSS/PACKUS intrinsics which are acting as non-saturating truncations (although technically the expansion could be used in all cases - but we'll probably want to be conservative).
llvm-svn: 359111
If we have a masked.load from a location we know to be dereferenceable, we can simply issue a speculative unconditional load against that address. The key advantage is that it produces IR which is well understood by the optimizer. The select (cnd, load, passthrough) form produced should be pattern matchable back to hardware predication if profitable.
Differential Revision: https://reviews.llvm.org/D59703
llvm-svn: 359000
If we have a store to a piece of memory which is known constant, then we know the store must be storing back the same value. As a result, the store (or memset, or memmove) must either be down a dead path, or a noop. In either case, it is valid to simply remove the store.
The motivating case for this involves a memmove to a buffer which is constant down a path which is dynamically dead.
Note that I'm choosing to implement the less aggressive of two possible semantics here. We could simply say that the store *is undefined*, and prune the path. Consensus in the review was that the more aggressive form might be a good follow on change at a later date.
Differential Revision: https://reviews.llvm.org/D60659
llvm-svn: 358919
In the process, use the existing masked.load combine which is slightly stronger, and handles a mix of zero and undef elements in the mask.
llvm-svn: 358913
In InstCombine, we use an idiom of "store i1 true, i1 undef" to indicate we've found a path which we've proven unreachable. We can't actually insert the unreachable instruction since that would require changing the CFG. We leave that to simplifycfg later.
This just factors out that idiom creation so we don't duplicate the same mostly undocument idiom creation in multiple places.
llvm-svn: 358600
If a constant shift amount is used, then only some of the LHS/RHS
operand bits are demanded and we may be able to simplify based on
that. InstCombineSimplifyDemanded already had the necessary support
for that, we just weren't calling it with fshl/fshr as root.
In particular, this allows us to relax some masked funnel shifts
into simple shifts, as shown in the tests.
Patch by Shawn Landden.
Differential Revision: https://reviews.llvm.org/D60660
llvm-svn: 358515
This adds a WithOverflowInst class with a few helper methods to get
the underlying binop, signedness and nowrap type and makes use of it
where sensible. There will be two more uses in D60650/D60656.
The refactorings are all NFC, though I left some TODOs where things
could be improved. In particular we have two places where add/sub are
handled but mul isn't.
Differential Revision: https://reviews.llvm.org/D60668
llvm-svn: 358512
Summary:
Enable some of the existing size optimizations for cold code under PGO.
A ~5% code size saving in big internal app under PGO.
The way it gets BFI/PSI is discussed in the RFC thread
http://lists.llvm.org/pipermail/llvm-dev/2019-March/130894.html
Note it doesn't currently touch loop passes.
Reviewers: davidxl, eraman
Reviewed By: eraman
Subscribers: mgorny, javed.absar, smeenai, mehdi_amini, eraman, zzheng, steven_wu, dexonsmith, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D59514
llvm-svn: 358422
As pointed out in D60518 folding mulo(%x, undef) to {undef, undef}
isn't correct. As a correct version of this already exists in
InstructionSimplify (bd8056ef32/lib/Analysis/InstructionSimplify.cpp (L4750-L4757)) this is just
dead code though. Drop it together with the mul(%x, 0) -> {0, false}
fold that is also already handled by InstSimplify.
Differential Revision: https://reviews.llvm.org/D60649
llvm-svn: 358339
This fixes a miscompile which was introduced in r356510 (https://reviews.llvm.org/D57372).
The problem is that the original patch removed pointer operands where the load results we're demanded, but without considering the legality of the load itself. If the masked.gather had active, but undemanded, lanes, then we could end up creating a load which loaded from an undef address. The result could be a segfault, or, in theory, an arbitrary read from a random memory location into an used register.
llvm-svn: 358299
Following D60483 and D60497, this adds support for AlwaysOverflows
handling for ssubo. This is the last case we can handle right now.
Differential Revision: https://reviews.llvm.org/D60518
llvm-svn: 358100
ssubo X, C is equivalent to saddo X, -C. Make the transformation in
InstCombine and allow the logic implemented for saddo to fold prior
usages of add nsw or sub nsw with constants.
Patch by Dan Robertson.
Differential Revision: https://reviews.llvm.org/D60061
llvm-svn: 358099
Check AlwaysOverflow condition for usubo. The implementation is the
same as the existing handling for uaddo and umulo. Handling for saddo
and ssubo will follow (smulo doesn't have the necessary ValueTracking
support).
Differential Revision: https://reviews.llvm.org/D60483
llvm-svn: 358052
This reverts commit 1383a91689.
sdiv-canonicalize.ll fails after this revision. The fold needs to be
moved outside the branch handling constant operands. However when this
is done there are further test changes, so I'm reverting this in the
meantime.
llvm-svn: 358026
Change the code to always handle the unsigned+signed cases together
with the same basic structure for add/sub/mul. The simple folds are
always handled first and then the ValueTracking overflow checks are
used.
llvm-svn: 358025
Similar to:
rL358005
Forego folding arbitrary vector constants to fix a possible miscompile bug.
We can enhance the transform if we do want to handle the more complicated
vector case.
llvm-svn: 358013
// 0 - (X sdiv C) -> (X sdiv -C) provided the negation doesn't overflow.
This fold has been around for many years and nobody noticed the potential
vector miscompile from overflow until recently...
So it seems unlikely that there's much demand for a vector sdiv optimization
on arbitrary vector constants, so just limit the matching to splat constants
to avoid the possible bug.
Differential Revision: https://reviews.llvm.org/D60426
llvm-svn: 358005
A more general canonicalization between fdiv and fmul would not
handle this case because that would have to be limited by uses
to prevent 2 values from becoming 3 values:
(x/y) * (x/y) --> (x*x) / (y*y)
(But we probably should still have that limited -- but more general --
canonicalization independently of this change.)
llvm-svn: 357943
First step towards removing the MOVMSK intrinsics completely - this patch expands MOVMSK to the pattern:
e.g. PMOVMSKB(v16i8 x):
%cmp = icmp slt <16 x i8> %x, zeroinitializer
%int = bitcast <16 x i8> %cmp to i16
%res = zext i16 %int to i32
Which is correctly handled by ISel and FastIsel (give or take an annoying movzx move....): https://godbolt.org/z/rkrSFW
Differential Revision: https://reviews.llvm.org/D60256
llvm-svn: 357909
Summary:
This fixes PR41270.
The recursive function evaluateInDifferentElementOrder expects to be called
on a vector Value, so when we call it on a vector GEP's arguments, we must
first check that the argument is indeed a vector.
Reviewers: reames, spatel
Reviewed By: spatel
Subscribers: llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D60058
llvm-svn: 357389
This reverts commit 75216a6dbcfe5fb55039ef06a07e419fa875f4a5.
I'll recommit with a better commit message with reference to the
phabricator review.
llvm-svn: 357387
This fixes PR41270.
The recursive function evaluateInDifferentElementOrder expects to be called
on a vector Value, so when we call it on a vector GEP's arguments, we must
first check that the argument is indeed a vector.
llvm-svn: 357385
If we have a commutable vector binop with inverted select-shuffles,
we don't care about the order of the operands in each vector lane:
LHS = shuffle V1, V2, <0, 5, 6, 3>
RHS = shuffle V2, V1, <0, 5, 6, 3>
LHS + RHS --> <V1[0]+V2[0], V2[1]+V1[1], V2[2]+V1[2], V1[3]+V2[3]> --> V1 + V2
PR41304:
https://bugs.llvm.org/show_bug.cgi?id=41304
...is currently titled as an SLP enhancement, but at least for the
given example, we can reduce that in instcombine because we are just
eliminating shuffles.
As noted in the TODO, this could be generalized, but I haven't thought
through those patterns completely, so this is limited to what appears
to be always safe.
Differential Revision: https://reviews.llvm.org/D60048
llvm-svn: 357382
In PR41304:
https://bugs.llvm.org/show_bug.cgi?id=41304
...we have a case where we want to fold a binop of select-shuffle (blended) values.
Rather than try to match commuted variants of the pattern, we can canonicalize the
shuffles and check for mask equality with commuted operands.
We don't produce arbitrary shuffle masks in instcombine, but select-shuffles are a
special case that the backend is required to handle because we already canonicalize
vector select to this shuffle form.
So there should be no codegen difference from this change. It's possible that this
improves CSE in IR though.
Differential Revision: https://reviews.llvm.org/D60016
llvm-svn: 357366
This may not be NFC, but I'm not sure how to expose any diffs in
tests. In theory, it should be slightly more efficient and possibly
more profitable to do the canonicalizations (which can increase the
undef elements in the mask) ahead of SimplifyDemandedVectorElts().
llvm-svn: 357272
Start using the uadd.sat and usub.sat intrinsics for the existing
canonicalizations. These intrinsics should optimize better than
expanded IR, have better handling in the X86 backend and should
be no worse than expanded IR in other backends, as far as we know.
rL357012 already introduced use of uadd.sat for the add+umin pattern.
Differential Revision: https://reviews.llvm.org/D58872
llvm-svn: 357103
This is the last step towards solving the examples shown in:
https://bugs.llvm.org/show_bug.cgi?id=14613
With this change, x86 should end up with psubus instructions
when those are available.
All known codegen issues with expanding the saturating intrinsics
were resolved with:
D59006 / rL356855
We also have some early evidence in D58872 that using the intrinsics
will lead to better perf. If some target regresses from this, custom
lowering of the intrinsics (as in the above for x86) may be needed.
llvm-svn: 357012
This helps to avoid the situation where RA spots that only 3 of the
v4f32 result of a load are used, and immediately reallocates the 4th
register for something else, requiring a stall waiting for the load.
Differential Revision: https://reviews.llvm.org/D58906
Change-Id: I947661edfd5715f62361a02b100f14aeeada29aa
llvm-svn: 356768
If they have other users we'll just end up increasing the instruction count.
We might be able to weaken this to only one of them having a single use if we can prove that the and will be removed.
Fixes PR41164.
Differential Revision: https://reviews.llvm.org/D59630
llvm-svn: 356690
If we know we're not storing a lane, we don't need to compute the lane. This could be improved by using the undef element result to further prune the mask, but I want to separate that into its own change since it's relatively likely to expose other problems.
Differential Revision: https://reviews.llvm.org/D57247
llvm-svn: 356590
Teach instcombine to propagate demanded elements through a masked load or masked gather instruction. This is in the broader context of improving vector pointer instcombine under https://reviews.llvm.org/D57140.
Differential Revision: https://reviews.llvm.org/D57372
llvm-svn: 356510
Combine 2 fcmps that are checking for nan-ness:
and (fcmp ord X, 0), (and (fcmp ord Y, 0), Z) --> and (fcmp ord X, Y), Z
or (fcmp uno X, 0), (or (fcmp uno Y, 0), Z) --> or (fcmp uno X, Y), Z
This is an exact match for a minimal reassociation pattern.
If we want to handle this more generally that should go in
the reassociate pass and allow removing this code.
This should fix:
https://bugs.llvm.org/show_bug.cgi?id=41069
llvm-svn: 356471
Follow-up to:
rL356338
rL356369
We can calculate an arbitrary vector constant minus the bitwidth, so there's
no need to limit this transform to scalars and splats.
llvm-svn: 356372
Follow-up to:
rL356338
Rotates are a special case of funnel shift where the 2 input operands
are the same value, but that does not need to be a restriction for the
canonicalization when the shift amount is a constant.
llvm-svn: 356369
This was noted as a backend problem:
https://bugs.llvm.org/show_bug.cgi?id=41057
...and subsequently fixed for x86:
rL356121
But we should canonicalize these in IR for the benefit of all targets
and improve IR analysis such as CSE.
llvm-svn: 356338
A change of two parts:
1) A generic enhancement for all callers of SDVE to exploit the fact that if all lanes are undef, the result is undef.
2) A GEP specific piece to strengthen/fix the vector index undef element handling, and call into the generic infrastructure when visiting the GEP.
The result is that we replace a vector gep with at least one undef in each lane with a undef. We can also do the same for vector intrinsics. Once the masked.load patch (D57372) has landed, I'll update to include call tests as well.
Differential Revision: https://reviews.llvm.org/D57468
llvm-svn: 356293
Before r355981, this was under LLVM_DEBUG. I don't think the assert is
quite right, but this really should be a verifier check. Instcombine
should not be asserting on this sort of thing.
llvm-svn: 356219
The shift argument is defined to be modulo the bitwidth, so if that argument
is a constant, we can always reduce the constant to its minimal form to allow
better CSE and other follow-on transforms.
We need to be careful to ignore constant expressions here, or we will likely
infinite loop. I'm adding a general vector constant query for that case.
Differential Revision: https://reviews.llvm.org/D59374
llvm-svn: 356192
This indicates an intrinsic parameter is required to be a constant,
and should not be replaced with a non-constant value.
Add the attribute to all AMDGPU and generic intrinsics that comments
indicate it should apply to. I scanned other target intrinsics, but I
don't see any obvious comments indicating which arguments are intended
to be only immediates.
This breaks one questionable testcase for the autoupgrade. I'm unclear
on whether the autoupgrade is supposed to really handle declarations
which were never valid. The verifier fails because the attributes now
refer to a parameter past the end of the argument list.
llvm-svn: 355981
I'm assuming that the nan propogation logic for InstructonSimplify's handling of fadd and fsub is correct, and applying the same to atomicrmw.
Differential Revision: https://reviews.llvm.org/D58836
llvm-svn: 355222
An idempotent atomicrmw is one that does not change memory in the process of execution. We have already added handling for the various integer operations; this patch extends the same handling to floating point operations which were recently added to IR.
Note: At the moment, we canonicalize idempotent fsub to fadd when ordering requirements prevent us from using a load. As discussed in the review, I will be replacing this with canonicalizing both floating point ops to integer ops in the near future.
Differential Revision: https://reviews.llvm.org/D58251
llvm-svn: 355210
This is part of a transform that may be done in the backend:
D13757
...but it should always be beneficial to fold this sooner in IR
for all targets.
https://rise4fun.com/Alive/vaiW
Name: sext add nsw
%add = add nsw i8 %i, C0
%ext = sext i8 %add to i32
%r = add i32 %ext, C1
=>
%s = sext i8 %i to i32
%r = add i32 %s, sext(C0)+C1
Name: zext add nuw
%add = add nuw i8 %i, C0
%ext = zext i8 %add to i16
%r = add i16 %ext, C1
=>
%s = zext i8 %i to i16
%r = add i16 %s, zext(C0)+C1
llvm-svn: 355118
Summary:
The description of KnownBits::zext() and
KnownBits::zextOrTrunc() has confusingly been telling
that the operation is equivalent to zero extending the
value we're tracking. That has not been true, instead
the user has been forced to explicitly set the extended
bits as known zero afterwards.
This patch adds a second argument to KnownBits::zext()
and KnownBits::zextOrTrunc() to control if the extended
bits should be considered as known zero or as unknown.
Reviewers: craig.topper, RKSimon
Reviewed By: RKSimon
Subscribers: javed.absar, hiraditya, jdoerfert, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D58650
llvm-svn: 355099
add A, sext(B) --> sub A, zext(B)
We have to choose 1 of these forms, so I'm opting for the
zext because that's easier for value tracking.
The backend should be prepared for this change after:
D57401
rL353433
This is also a preliminary step towards reducing the amount
of bit hackery that we do in IR to optimize icmp/select.
That should be waiting to happen at a later optimization stage.
The seeming regression in the fuzzer test was discussed in:
D58359
We were only managing that fold in instcombine by luck, and
other passes should be able to deal with that better anyway.
llvm-svn: 354748
This is no-functional-change-intended, but that was also
true when it was part of rL354276, and I managed to lose
2 predicates for the fold with constant...causing much bot
distress. So this time I'm adding a couple of negative tests
to avoid that.
llvm-svn: 354384
We want to use the sum in the icmp to allow matching with
m_UAddWithOverflow and eliminate the 'not'. This is discussed
in D51929 and is another step towards solving PR14613:
https://bugs.llvm.org/show_bug.cgi?id=14613
(The matching here is incomplete. Trying to take minimal steps
to make sure we don't induce infinite looping from existing
canonicalizations of the 'select'.)
llvm-svn: 354221
Implement two more transforms of atomicrmw:
1) We can convert an atomicrmw which produces a known value in memory into an xchg instead.
2) We can convert an atomicrmw xchg w/o users into a store for some orderings.
Differential Revision: https://reviews.llvm.org/D58290
llvm-svn: 354170
For "idempotent" atomicrmw instructions which we can't simply turn into load, canonicalize the operation and constant. This reduces the matching needed elsewhere in the optimizer, but doesn't directly impact codegen.
For any architecture where OR/Zero is not a good default choice, you can extend the AtomicExpand lowerIdempotentRMWIntoFencedLoad mechanism. I reviewed X86 to make sure this works well, haven't audited other backends.
Differential Revision: https://reviews.llvm.org/D58244
llvm-svn: 354058
Expand on Quentin's r353471 patch which converts some atomicrmws into loads. Handle remaining operation types, and fix a slight bug. Atomic loads are required to have alignment. Since this was within the InstCombine fixed point, somewhere else in InstCombine was adding alignment before the verifier saw it, but still, we should fix.
Terminology wise, I'm using the "idempotent" naming that is used for the same operations in AtomicExpand and X86ISelLoweringInfo. Once this lands, I'll add similar tests for AtomicExpand, and move the pattern match function to a common location. In the review, there was seemingly consensus that "idempotent" was slightly incorrect for this context. Once we setle on a better name, I'll update all uses at once.
Differential Revision: https://reviews.llvm.org/D58242
llvm-svn: 354046
Sanjay Patel