Rename old function to explicitly show that it cares only about alignment.
The new allowsMemoryAccess call the function related to alignment by default
and can be overridden by target to inform whether the memory access is legal or
not.
Differential Revision: https://reviews.llvm.org/D67121
llvm-svn: 372935
We might be able to do better on the example in the test,
but in general, we should not scalarize a splatted vector
binop if there are other uses of the binop. Otherwise, we
can end up with code as we had - a scalar op that is
redundant with a vector op.
llvm-svn: 372886
This patch converts the DAGCombine isNegatibleForFree/GetNegatedExpression into overridable TLI hooks and includes a demonstration X86 implementation.
The intention is to let us extend existing FNEG combines to work more generally with negatible float ops, allowing it work with target specific combines and opcodes (e.g. X86's FMA variants).
Unlike the SimplifyDemandedBits, we can't just handle target nodes through a Target callback, we need to do this as an override to allow targets to handle generic opcodes as well. This does mean that the target implementations has to duplicate some checks (recursion depth etc.).
I've only begun to replace X86's FNEG handling here, handling FMADDSUB/FMSUBADD negation and some low impact codegen changes (some FMA negatation propagation). We can build on this in future patches.
Differential Revision: https://reviews.llvm.org/D67557
llvm-svn: 372333
As commented on D67557 we have a lot of uses of depth checks all using magic numbers.
This patch adds the SelectionDAG::MaxRecursionDepth constant and moves over some general cases to use this explicitly.
Differential Revision: https://reviews.llvm.org/D67711
llvm-svn: 372315
Summary:
`DAGCombiner::visitADDLikeCommutative()` already has a sibling fold:
`(add X, Carry) -> (addcarry X, 0, Carry)`
This fold, as suggested by @efriedma, helps recover from //some//
of the regressions of D62266
Reviewers: efriedma, deadalnix
Subscribers: javed.absar, kristof.beyls, llvm-commits, efriedma
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62392
llvm-svn: 372259
This is the first sweep of generic code to add isAtomic bailouts where appropriate. The intention here is to have the switch from AtomicSDNode to LoadSDNode/StoreSDNode be close to NFC; that is, I'm not looking to allow additional optimizations at this time. That will come later. See D66309 for context.
Differential Revision: https://reviews.llvm.org/D66318
llvm-svn: 371786
The X86 decision assumes the compare will produce a result in an XMM
register, but that can't happen for an fp128 compare since those
go to a libcall the returns an i32. Pass the VT so X86 can check
the type.
llvm-svn: 371775
Current implementation of estimating divisions loses precision since it
estimates reciprocal first and does multiplication. This patch is to re-order
arithmetic operations in the last iteration in DAGCombiner to improve the
accuracy.
Reviewed By: Sanjay Patel, Jinsong Ji
Differential Revision: https://reviews.llvm.org/D66050
llvm-svn: 371713
Summary:
Add an intrinsic that takes 2 unsigned integers with
the scale of them provided as the third argument and
performs fixed point multiplication on them. The
result is saturated and clamped between the largest and
smallest representable values of the first 2 operands.
This is a part of implementing fixed point arithmetic
in clang where some of the more complex operations
will be implemented as intrinsics.
Patch by: leonardchan, bjope
Reviewers: RKSimon, craig.topper, bevinh, leonardchan, lebedev.ri, spatel
Reviewed By: leonardchan
Subscribers: ychen, wuzish, nemanjai, MaskRay, jsji, jdoerfert, Ka-Ka, hiraditya, rjmccall, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D57836
llvm-svn: 371308
The motivating bugs are:
https://bugs.llvm.org/show_bug.cgi?id=41340https://bugs.llvm.org/show_bug.cgi?id=42697
As discussed there, we could view this as a failure of IR canonicalization,
but then we would need to implement a backend fixup with target overrides
to get this right in all cases. Instead, we can just view this as a codegen
opportunity. It's not even clear for x86 exactly when we should favor
test+set; some CPUs have better theoretical throughput for the ALU ops than
bt/test.
This patch is made more complicated than I expected because there's an early
DAGCombine for 'and' that can change types of the intermediate ops via
trunc+anyext.
Differential Revision: https://reviews.llvm.org/D66687
llvm-svn: 370668
The motivating case for this is a long way from here:
https://bugs.llvm.org/show_bug.cgi?id=43146
...but I think this is where we have to start.
We need to canonicalize/optimize sequences of shift and logic to ease
pattern matching for things like bswap and improve perf in general.
But without the artificial limit of '!LegalTypes' (early combining),
there are a lot of test diffs, and not all are good.
In the minimal tests added for this proposal, x86 should have better
throughput in all cases. AArch64 is neutral for scalar tests because
it can fold shifts into bitwise logic ops.
There are 3 shift opcodes and 3 logic opcodes for a total of 9 possible patterns:
https://rise4fun.com/Alive/VlIhttps://rise4fun.com/Alive/n1mhttps://rise4fun.com/Alive/1Vn
Differential Revision: https://reviews.llvm.org/D67021
llvm-svn: 370617
Narrowing stores when the target doesn't support the narrow version
forces the target to expand into a load-modify-store sequence, which
is highly suboptimal. The information narrowing throws away (legality
of the inverse transform) is hard to re-analyze. If the target doesn't
support a store of the narrow type, don't narrow even in pre-legalize
mode.
No test as this is DAGCombiner and depends on target bits.
llvm-svn: 370576
This is hidden behind a (scalar-only) isOneConstant(N1) check at the moment, but once we get around to adding vector support we need to ensure we're dealing with the scalar bitwidth, not the total.
llvm-svn: 370468
Return a proper zero vector, just in case some elements are undef.
Noticed by inspection after dealing with a similar issue in PR43159.
llvm-svn: 370460
Summary: This is beneficial when the shuffle is only used once and end up being generated in a few places when some node is combined into a shuffle.
Reviewers: craig.topper, efriedma, RKSimon, lebedev.ri
Subscribers: llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D66718
llvm-svn: 370326
Summary: There are at least 2 ways to express the same shuffle. Various pieces of code explicit check for both option, but other places do not when they would benefit from doing it. This patches refactor the codebase to use buildLegalVectorShuffle in order to make that behavior more consistent.
Reviewers: craig.topper, efriedma, RKSimon, lebedev.ri
Subscribers: javed.absar, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D66804
llvm-svn: 370190
(-X) * (-Y) + Z --> X * Y + Z
This is a missing optimization that shows up as a potential regression in D66050,
so we should solve it first. We appear to be partly missing this fold in IR as well.
We do handle the simpler case already:
(-X) * (-Y) --> X * Y
And it might be beneficial to make the constraint less conservative (eg, if both
operands are cheap, but not necessarily cheaper), but that causes infinite looping
for the existing fmul transform.
Differential Revision: https://reviews.llvm.org/D66755
llvm-svn: 370071
This change causes instrumented builds of Clang to have a fatal error in the
backend. https://reviews.llvm.org/D66537 has the details.
llvm-svn: 370006
This improves the combine I included in D66504 to handle constants in the upper operands of the concat. If we can constant fold them away we can pull the concat after the bin op. This helps with chains of madd reductions on X86 from loop unrolling. The loop madd reduction pattern creates pmaddwd with half the width of the add that follows it using zeroes to fill the upper bits. If we have two of these added together we can pull the zeroes through the accumulating add and then shrink it.
Differential Revision: https://reviews.llvm.org/D66680
llvm-svn: 369937
Summary:
This comes as a first step toward processing the DAG nodes in topological orders. Doing so ensure that arguments of a node are combined before the node itself is combined, which exposes ore opportunities for optimization and/or reduce the amount of patterns a node has to match for.
DAGCombiner adding nodes to the worklist is various places causes the nodes to be in a different order from what is expected. In addition, this is reduant because these nodes end up being added to the worklist anyways due to the machinery at line 1621.
Reviewers: craig.topper, efriedma, RKSimon, lebedev.ri
Subscribers: llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D66537
llvm-svn: 369927
Summary:
Concat_vectors is more canonical during early DAG combine. For example, its what's used by SelectionDAGBuilder when converting IR shuffles into SelectionDAG shuffles when element counts between inputs and mask don't match. We also have combines in DAGCombiner than can pull concat_vectors through a shuffle. See partitionShuffleOfConcats. So it seems like concat_vectors is a better operation to use here. I had to teach DAGCombiner's SimplifyVBinOp to also handle concat_vectors with undef. I haven't checked yet if we can remove the INSERT_SUBVECTOR version in there or not.
I didn't want to mess with the other caller of getShuffleHalfVectors that's used during shuffle lowering where insert_subvector probably is what we want to produce so I've enabled this via a boolean passed to the function.
Reviewers: spatel, RKSimon
Reviewed By: RKSimon
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D66504
llvm-svn: 369872
These can turn up during multiplication legalization. In principle
these should also apply to smul_lohi, but I wasn't able to figure
out how to produce those with the necessary operands.
Differential Revision: https://reviews.llvm.org/D66380
llvm-svn: 369864
If the accumulator and either of the multiply operands are negatable then we can we negate the entire expression.
Differential Revision: https://reviews.llvm.org/D63141
llvm-svn: 369746
Summary: These nodes end up being processed regardless due to DAGCombiner ensuring arguments are processed. This changes the order in which nodes are processed, which fixes an issue on PowerPC.
Reviewers: craig.topper, efriedma, RKSimon, lebedev.ri, mcberg2017, stefanp, hfinkel
Subscribers: nemanjai, MaskRay, jsji, steven.zhang, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D66548
llvm-svn: 369662
Summary:
These calls change the order in which some nodes are processed and so have an effect on codegen.
The change in fixup-bw-copy.ll is due to (and (load anyext)) gets transformed into (load zext) while previously the and was removed by SimplifyDemandedBits, so the (load anyext) remained.
Reviewers: craig.topper, efriedma, RKSimon, lebedev.ri
Subscribers: llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D66543
llvm-svn: 369561
I also had to add a new combine to X86's combineExtractSubvector to prevent a regression.
This helps our vXi1 code see the full concat operation and allow it optimize undef to a zero if there is already a zero in the concat. This helped us use a movzx instead of an AND in some of the tests. In those tests, one concat comes from SelectionDAGBuilder and the second comes from type legalization of v4i1->i4 bitcasts which uses an additional concat. Though these changes weren't my original motivation.
I'm looking at making X86ISelLowering's narrowShuffle emit a concat_vectors instead of an insert_subvector since concat_vectors is more canonical during early DAG combine. This patch helps prevent a regression from my experiments with that.
Differential Revision: https://reviews.llvm.org/D66456
llvm-svn: 369459
Set the StartIdx type to size_t so that it matches the StoreNodes SmallVector size() and index types.
Silences the MSVC analyzer warning that unsigned increment might overflow before exceeding size_t on 64-bit targets - this isn't likely to happen but it means we use consistent types and reduces the warning "noise" a little.
llvm-svn: 368998
This is the codegen part of fixing:
https://bugs.llvm.org/show_bug.cgi?id=32939
Even with the optimal/canonical IR that is ideally created by D65954,
we would reverse that transform in DAGCombiner and end up with the same
asm on AArch64 or x86.
I see 2 options for trying to correct this:
1. Limit isNegatibleForFree() by special-casing the fmul pattern (this patch).
2. Avoid creating (fmul X, 2.0) in the 1st place by adding a special-case
transform to SelectionDAG::getNode() and/or SelectionDAGBuilder::visitFMul()
that matches the transform done by DAGCombiner.
This seems like the less intrusive patch, but if there's some other reason to
prefer 1 option over the other, we can change to the other option.
Differential Revision: https://reviews.llvm.org/D66016
llvm-svn: 368490
We may be able to look to how VSELECT is handled to further
improve this, but this appears to be neutral or an improvement
on the test cases we have.
llvm-svn: 368344
Summary:
Before this patch MGATHER/MSCATTER is capable of representing all
common addressing modes, but only when illegal types are used.
This patch adds an IndexType property so more representations
are available when using legal types only.
Original modes:
vector of bases
base + vector of signed scaled offsets
New modes:
base + vector of signed unscaled offsets
base + vector of unsigned scaled offsets
base + vector of unsigned unscaled offsets
The current behaviour of addressing modes for gather/scatter remains
unchanged.
Patch by Paul Walker.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D65636
llvm-svn: 368008
The test case is based on the example from the post-commit thread for:
https://reviews.llvm.org/rGc9171bd0a955
This replaces the x86-specific simple-type check from:
rL367766
with a check in the DAGCombiner. Adding the check isn't
strictly necessary after the fix from:
rL367768
...but it seems likely that we're heading for trouble if
we are creating weird types in this transform.
I combined the earlier legality check into the initial
clause to simplify the code.
So we should only try the trunc/sext transform at the
earliest combine stage, but we limit the transform to
simple types anyway because the TLI hook is probably
too lax about what it considers a free truncate.
llvm-svn: 367834
This is further fix for PR42880.
Sanjay already disabled the X86 TLI hook for non-simple types,
but we should really call isTypeLegal here if we're after type
legalization.
llvm-svn: 367768
This reverses a questionable IR canonicalization when a truncate
is free:
sra (add (shl X, N1C), AddC), N1C -->
sext (add (trunc X to (width - N1C)), AddC')
https://rise4fun.com/Alive/slRC
More details in PR42644:
https://bugs.llvm.org/show_bug.cgi?id=42644
I limited this to pre-legalization for code simplicity because that
should be enough to reverse the IR patterns. I don't have any
evidence (no regression test diffs) that we need to try this later.
Differential Revision: https://reviews.llvm.org/D65607
llvm-svn: 367710
If a type is larger than a legal type and needs to be split, we would previously allow the multiply to be decomposed even if the split multiply is legal. Since the shift + add/sub code would also need to be split, its not any better to decompose it.
This patch figures out what type the mul will eventually be legalized to and then uses that type for the query. I tried just returning false illegal types and letting them get handled after type legalization, but then we can't recognize and i64 constant splat on 32-bit targets since will be destroyed by type legalization. We could special case vectors of i64 to avoid that...
Differential Revision: https://reviews.llvm.org/D65533
llvm-svn: 367601
Summary: Honoring no signed zeroes is also available as a user control through clang separately regardless of fastmath or UnsafeFPMath context, DAG guards should reflect this context.
Reviewers: spatel, arsenm, hfinkel, wristow, craig.topper
Reviewed By: spatel
Subscribers: rampitec, foad, nhaehnle, wuzish, nemanjai, jvesely, wdng, javed.absar, MaskRay, jsji
Differential Revision: https://reviews.llvm.org/D65170
llvm-svn: 367486
to bail out in store merging dependence check.
We run into a case where dependence check in store merging bail out many times
for the same store and root nodes in a huge basicblock. That increases compile
time by almost 100x. The patch add a map to track how many times the bailing
out happen for the same store and root, and if it is over a limit, stop
considering the store with the same root as a merging candidate.
Differential Revision: https://reviews.llvm.org/D65174
llvm-svn: 367472
Add an option to control whether or not to enable store merging in dag combiner
so we can workaround some bugs more easily.
Differential Revision: https://reviews.llvm.org/D65482
llvm-svn: 367365
The function was calling getNode() on an SDValue to return and the
caller turned the result back into a SDValue. So just return the
original SDValue to avoid this.
llvm-svn: 366779
Summary:
As per title. DAGCombiner only mathes the special case where b = 0, this patches extends the pattern to match any value of b.
Depends on D57302
Reviewers: hfinkel, RKSimon, craig.topper
Subscribers: llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D59208
llvm-svn: 366214
We already split extract_subvector(binop(insert_subvector(v,x),insert_subvector(w,y))) -> binop(x,y).
This patch adds support for extract_subvector(binop(concat_vectors(),concat_vectors())) cases as well.
In particular this means we don't have to wait for X86 lowering to convert concat_vectors to insert_subvector chains, which helps avoid some cases where demandedelts/combine calls occur too late to split large vector ops.
The fast-isel-store.ll load folding regression is annoying but I don't think is that critical.
Differential Revision: https://reviews.llvm.org/D63653
llvm-svn: 365785
Summary: Unsafe does not map well alone for each of these three cases as it is missing NoNan context when accessed directly with clang. I have migrated the fold guards to reflect the expectations of handing nan and zero contexts directly (NoNan, NSZ) and some tests with it. Unsafe does include NSZ, however there is already precedent for using the target option directly to reflect that context.
Reviewers: spatel, wristow, hfinkel, craig.topper, arsenm
Reviewed By: arsenm
Subscribers: michele.scandale, wdng, javed.absar
Differential Revision: https://reviews.llvm.org/D64450
llvm-svn: 365679
Basically the problem is that X86 doesn't set the Fast flag from
allowsMemoryAccess on certain CPUs due to slow unaligned memory
subtarget features. This prevents bitcasts from being folded into
loads and stores. But all vector loads and stores of the same width
are the same cost on X86.
This patch merges the allowsMemoryAccess call into isLoadBitCastBeneficial to allow X86 to skip it.
Differential Revision: https://reviews.llvm.org/D64295
llvm-svn: 365549
Summary:
The uaddo won't be removed and the addcarry will still be
dependent on the uaddo. So we'll just increase the use count
of X and Y and potentially require a COPY.
Reviewers: spatel, RKSimon, deadalnix
Reviewed By: RKSimon
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D64190
llvm-svn: 365149
Summary:
This diff improve the capability of DAGCOmbine to generate linear carries propagation in presence of a diamond pattern. It is now able to match a large variety of different patterns rather than some hardcoded one.
Arguably, the codegen in test cases is not better, but this is to be expected. The goal of this transformation is more about canonicalisation than actual optimisation.
Reviewers: hfinkel, RKSimon, craig.topper
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D57302
llvm-svn: 365051
Summary:
This is the backend part of [[ https://bugs.llvm.org/show_bug.cgi?id=42457 | PR42457 ]].
In middle-end, we'd want to prefer the form with two adds - D63992,
but as this diff shows, not every target will prefer that pattern.
Out of 4 targets for which i added tests all seem to be ok with inc-of-add for scalars,
but only X86 prefer that same pattern for vectors.
Here i'm adding a new TLI hook, always defaulting to the inc-of-add,
but adding AArch64,ARM,PowerPC overrides to prefer inc-of-add only for scalars.
Reviewers: spatel, RKSimon, efriedma, t.p.northover, hfinkel
Reviewed By: efriedma
Subscribers: nemanjai, javed.absar, kristof.beyls, kbarton, jsji, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D64090
llvm-svn: 365010
For a given floating point load / store pair, if the load value isn't used by any other operations,
then consider transforming the pair to integer load / store operations if the target deems the transformation profitable.
And we can exploiting much more when there are other operation nodes with chain operand between the load/store pair
so long as we keep the chain ordering original. We only replace the register used to load/store from float to integer.
I only add testcase in ARM because the TLI.isDesirableToTransformToIntegerOp hook is only enabled in ARM target.
Differential Revision: https://reviews.llvm.org/D60601
llvm-svn: 364883
We support 'big to little' (e.g. extract_subvector(v16i8 bitcast(v2i64))) but not 'little to big' cases (e.g. extract_subvector(v2i64 bitcast(v16i8)))
llvm-svn: 364405
This can occur under certain circumstances when undefs are created later on in the constant multipliers (e.g. in this case due to SimplifyDemandedVectorElts). Its better to let the shift by zero to occur and perform any cleanup afterward.
Fixes OSS Fuzz #15429
llvm-svn: 364179
The code divides the alignment by 2 if the original alignment is
equal to the original VT size. But this wouldn't be correct
if the alignment was larger than the VT size.
The memory operand object already takes care of calling MinAlign
on the base alignment and the memory pointer offset. So we don't
need any special code at all.
llvm-svn: 364151
We tend to only test for scalar/scalar consts when really we could support non-uniform vectors using ISD::matchUnaryPredicate/matchBinaryPredicate etc.
llvm-svn: 363924
Use getAPIntValue() in a few more places. Most of the time getZExtValue() is fine, but occasionally there's fuzzed code or someone decides to create i65536 or something.....
llvm-svn: 363887
Use matchBinaryPredicate instead of isConstOrConstSplat to let us handle non-uniform shift cases.
This requires us to tweak matchBinaryPredicate to allow it to (optionally) handle constants with different type widths.
llvm-svn: 363792
Some GEPs were not being split, presumably because that split would just be
undone by the DAGCombiner. Not performing those splits can prevent important
optimizations, such as preventing the element indices / member offsets from
being (partially) folded into load/store instruction immediates. This patch:
- Makes the splits also occur in the cases where the base address and the GEP
are in the same BB.
- Ensures that the DAGCombiner doesn't reassociate them back again.
Differential Revision: https://reviews.llvm.org/D60294
llvm-svn: 363544
This reverts rL363474. -debug-only=isel was added to some tests that
don't specify `REQUIRES: asserts`. This causes failures on
-DLLVM_ENABLE_ASSERTIONS=off builds.
I chose to revert instead of fixing the tests because I'm not sure
whether we should add `REQUIRES: asserts` to more tests.
llvm-svn: 363482
As discussed on D62910, we need to check whether particular types of memory access are allowed, not just their alignment/address-space.
This NFC patch adds a MachineMemOperand::Flags argument to allowsMemoryAccess and allowsMisalignedMemoryAccesses, and wires up calls to pass the relevant flags to them.
If people are happy with this approach I can then update X86TargetLowering::allowsMisalignedMemoryAccesses to handle misaligned NT load/stores.
Differential Revision: https://reviews.llvm.org/D63075
llvm-svn: 363179
As suggested by @arsenm on D63075 - this adds a TargetLowering::allowsMemoryAccess wrapper that takes a Load/Store node's MachineMemOperand to handle the AddressSpace/Alignment arguments and will also implicitly handle the MachineMemOperand::Flags change in D63075.
llvm-svn: 363048
This opportunity is found from spec 2017 557.xz_r. And it is used by the sha encrypt/decrypt. See sha-2/sha512.c
static void store64(u64 x, unsigned char* y)
{
for(int i = 0; i != 8; ++i)
y[i] = (x >> ((7-i) * 8)) & 255;
}
static u64 load64(const unsigned char* y)
{
u64 res = 0;
for(int i = 0; i != 8; ++i)
res |= (u64)(y[i]) << ((7-i) * 8);
return res;
}
The load64 has been implemented by https://reviews.llvm.org/D26149
This patch is trying to implement the store pattern.
Match a pattern where a wide type scalar value is stored by several narrow
stores. Fold it into a single store or a BSWAP and a store if the targets
supports it.
Assuming little endian target:
i8 *p = ...
i32 val = ...
p[0] = (val >> 0) & 0xFF;
p[1] = (val >> 8) & 0xFF;
p[2] = (val >> 16) & 0xFF;
p[3] = (val >> 24) & 0xFF;
>
*((i32)p) = val;
i8 *p = ...
i32 val = ...
p[0] = (val >> 24) & 0xFF;
p[1] = (val >> 16) & 0xFF;
p[2] = (val >> 8) & 0xFF;
p[3] = (val >> 0) & 0xFF;
>
*((i32)p) = BSWAP(val);
Differential Revision: https://reviews.llvm.org/D62897
llvm-svn: 362921
This patch is the first step towards ensuring MergeConsecutiveStores correctly handles non-temporal loads\stores:
1 - When merging load\stores we must ensure that they all have the same non-temporal flag. This is unlikely to occur, but can in strange cases where we're storing at the end of one page and the beginning of another.
2 - The merged load\store node must retain the non-temporal flag.
Differential Revision: https://reviews.llvm.org/D62910
llvm-svn: 362723
This is a special case of a more general transform (not (sub Y, X)) -> (add X, ~Y). InstCombine knows the general form. I've restricted to the special case to fix the motivating case PR42118. I tried handling any case where Y was constant, but got some changes on some Mips tests that I couldn't quickly prove where beneficial.
Fixes PR42118
Differential Revision: https://reviews.llvm.org/D62828
llvm-svn: 362533
The proposal in D62498 showed that x86 would benefit from vector
store splitting, but that may conflict with the generic DAG
combiner's store merging transforms.
Add memory type to the existing TLI hook that enables the merging
transforms, so we can limit those changes to scalars only for x86.
llvm-svn: 362507
Summary:
This *might* be the last fold for `sink-addsub-of-const.ll`, but i'm not sure yet.
As far as i can tell, there are no regressions here (ignoring x86-32),
all changes are either good or neutral.
This, almost surprisingly to me, fixes the motivational tests (in `shift-amount-mod.ll`)
`@reg32_lshr_by_sub_from_negated` from [[ https://bugs.llvm.org/show_bug.cgi?id=41952 | PR41952 ]].
https://rise4fun.com/Alive/vMd3
Reviewers: RKSimon, t.p.northover, craig.topper, spatel, efriedma
Reviewed By: RKSimon
Subscribers: sdardis, javed.absar, arichardson, kristof.beyls, jrtc27, atanasyan, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62774
llvm-svn: 362488
This opportunity is found from spec 2017 557.xz_r. And it is used by the sha encrypt/decrypt. See sha-2/sha512.c
static void store64(u64 x, unsigned char* y)
{
for(int i = 0; i != 8; ++i)
y[i] = (x >> ((7-i) * 8)) & 255;
}
static u64 load64(const unsigned char* y)
{
u64 res = 0;
for(int i = 0; i != 8; ++i)
res |= (u64)(y[i]) << ((7-i) * 8);
return res;
}
The load64 has been implemented by https://reviews.llvm.org/D26149
This patch is trying to implement the store pattern.
Match a pattern where a wide type scalar value is stored by several narrow
stores. Fold it into a single store or a BSWAP and a store if the targets
supports it.
Assuming little endian target:
i8 *p = ...
i32 val = ...
p[0] = (val >> 0) & 0xFF;
p[1] = (val >> 8) & 0xFF;
p[2] = (val >> 16) & 0xFF;
p[3] = (val >> 24) & 0xFF;
>
*((i32)p) = val;
i8 *p = ...
i32 val = ...
p[0] = (val >> 24) & 0xFF;
p[1] = (val >> 16) & 0xFF;
p[2] = (val >> 8) & 0xFF;
p[3] = (val >> 0) & 0xFF;
>
*((i32)p) = BSWAP(val);
Differential Revision: https://reviews.llvm.org/D61843
llvm-svn: 362472
Summary: This change facilitates propagating fmf which was placed on setcc from fcmp through folds with selects so that back ends can model this path for arithmetic folds on selects in SDAG.
Reviewers: qcolombet, spatel
Reviewed By: qcolombet
Subscribers: nemanjai, jsji
Differential Revision: https://reviews.llvm.org/D62552
llvm-svn: 362439
We were missing this fold in the DAG, which I've copied directly from llvm::ConstantFoldCastInstruction
Differential Revision: https://reviews.llvm.org/D62807
llvm-svn: 362397
If we hit the limit, we do expand the outstanding tokenfactors.
Otherwise, we might drop nodes with users in the unexpanded
tokenfactors. This fixes the crashes reported by Jordan Rupprecht.
Reviewers: niravd, spatel, craig.topper, rupprecht
Reviewed By: niravd
Differential Revision: https://reviews.llvm.org/D62633
llvm-svn: 362350
Move this combine from x86 into generic DAGCombine, which currently only manages cases where the bitcast is between types of the same scalarsize.
Differential Revision: https://reviews.llvm.org/D59188
llvm-svn: 362324
The results of the dyn_casts were immediately dereferenced on the next line
so they had better not be null.
I don't think there's any way for these dyn_casts to fail, so use a cast
of adding null check.
llvm-svn: 362315
I don't have a test case for these, but there is a test case for D62266
where, even after all the constant-folding patches, we still end up
with endless combine loop. Which makes sense, since we don't constant
fold for opaque constants.
llvm-svn: 362156
Summary:
Only vector tests are being affected here,
since subtraction by scalar constant is rewritten
as addition by negated constant.
No surprising test changes.
https://rise4fun.com/Alive/pbT
This is a recommit, originally committed in rL361852, but reverted
to investigate test-suite compile-time hangs.
Reviewers: RKSimon, craig.topper, spatel
Reviewed By: RKSimon
Subscribers: javed.absar, kristof.beyls, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62257
llvm-svn: 362146
Summary:
Again only vectors affected. Frustrating. Let me take a look into that..
https://rise4fun.com/Alive/AAq
This is a recommit, originally committed in rL361852, but reverted
to investigate test-suite compile-time hangs, and then reverted in
rL362109 to fix missing constant folds that were causing
endless combine loops.
Reviewers: RKSimon, craig.topper, spatel
Reviewed By: RKSimon
Subscribers: javed.absar, JDevlieghere, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62294
llvm-svn: 362145
Summary:
This prevents regressions in next patch,
and somewhat recovers from the regression to AMDGPU test in D62223.
It is indeed not great that we leave vector decrement,
don't transform it into vector add all-ones..
https://rise4fun.com/Alive/ZRl
This is a recommit, originally committed in rL361852, but reverted
to investigate test-suite compile-time hangs, and then reverted in
rL362109 to fix missing constant folds that were causing
endless combine loops.
Reviewers: RKSimon, craig.topper, spatel, arsenm
Reviewed By: RKSimon, arsenm
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, javed.absar, dstuttard, tpr, t-tye, kristof.beyls, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62263
llvm-svn: 362144
Summary:
Direct sibling of D62223 patch.
While i don't have a direct motivational pattern for this,
it would seem to make sense to handle both patterns (or none),
for symmetry?
The aarch64 changes look neutral;
sparc and systemz look like improvement (one less instruction each);
x86 changes - 32bit case improves, 64bit case shows that LEA no longer
gets constructed, which may be because that whole test is `-mattr=+slow-lea,+slow-3ops-lea`
https://rise4fun.com/Alive/ffh
This is a recommit, originally committed in rL361852, but reverted
to investigate test-suite compile-time hangs, and then reverted in
rL362109 to fix missing constant folds that were causing
endless combine loops.
Reviewers: RKSimon, craig.topper, spatel, t.p.northover
Reviewed By: t.p.northover
Subscribers: t.p.northover, jyknight, javed.absar, kristof.beyls, fedor.sergeev, jrtc27, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62252
llvm-svn: 362143
Summary:
The main motivation is shown by all these `neg` instructions that are now created.
In particular, the `@reg32_lshr_by_negated_unfolded_sub_b` test.
AArch64 test changes all look good (`neg` created), or neutral.
X86 changes look neutral (vectors), or good (`neg` / `xor eax, eax` created).
I'm not sure about `X86/ragreedy-hoist-spill.ll`, it looks like the spill
is now hoisted into preheader (which should still be good?),
2 4-byte reloads become 1 8-byte reload, and are elsewhere,
but i'm not sure how that affects that loop.
I'm unable to interpret AMDGPU change, looks neutral-ish?
This is hopefully a step towards solving [[ https://bugs.llvm.org/show_bug.cgi?id=41952 | PR41952 ]].
https://rise4fun.com/Alive/pkdq (we are missing more patterns, i'll submit them later)
This is a recommit, originally committed in rL361852, but reverted
to investigate test-suite compile-time hangs, and then reverted in
rL362109 to fix missing constant folds that were causing
endless combine loops.
Reviewers: craig.topper, RKSimon, spatel, arsenm
Reviewed By: RKSimon
Subscribers: bjope, qcolombet, kzhuravl, jvesely, wdng, nhaehnle, yaxunl, javed.absar, dstuttard, tpr, t-tye, kristof.beyls, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62223
llvm-svn: 362142
Summary:
Direct sibling of D62662, the root cause of the endless combine loop in D62257
https://rise4fun.com/Alive/d3W
Reviewers: RKSimon, craig.topper, spatel, t.p.northover
Reviewed By: t.p.northover
Subscribers: t.p.northover, javed.absar, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62664
llvm-svn: 362133
Summary:
No tests change, and i'm not sure how to test this, but it's better safe than sorry.
Reviewers: spatel, RKSimon, craig.topper, t.p.northover
Reviewed By: craig.topper
Subscribers: llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62663
llvm-svn: 362132
Summary:
This was the root cause of the endless combine loop in D62257
https://rise4fun.com/Alive/d3W
Reviewers: RKSimon, spatel, craig.topper, t.p.northover
Reviewed By: t.p.northover
Subscribers: t.p.northover, javed.absar, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62662
llvm-svn: 362131
Summary: No tests change, and i'm not sure how to test this, but it's better safe than sorry.
Reviewers: spatel, RKSimon, craig.topper, t.p.northover
Reviewed By: craig.topper
Subscribers: llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62661
llvm-svn: 362130
I was looking into an endless combine loop the uncommitted follow-up patch
was causing, and it appears even these patches can exibit such an
endless loop. The root cause is that we try to hoist one binop (add/sub) with
constant operand, and if we get two such binops both of which are
eligible for this hoisting, we get stuck.
Some cases may highlight missing constant-folds.
Reverts r361871,r361872,r361873,r361874.
llvm-svn: 362109
Summary:
Again only vectors affected. Frustrating. Let me take a look into that..
https://rise4fun.com/Alive/AAq
This is a recommit, originally committed in rL361856, but reverted
to investigate test-suite compile-time hangs.
Reviewers: RKSimon, craig.topper, spatel
Reviewed By: RKSimon
Subscribers: javed.absar, JDevlieghere, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62294
llvm-svn: 361874
Summary:
This prevents regressions in next patch,
and somewhat recovers from the regression to AMDGPU test in D62223.
It is indeed not great that we leave vector decrement,
don't transform it into vector add all-ones..
https://rise4fun.com/Alive/ZRl
This is a recommit, originally committed in rL361855, but reverted
to investigate test-suite compile-time hangs.
Reviewers: RKSimon, craig.topper, spatel, arsenm
Reviewed By: RKSimon, arsenm
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, javed.absar, dstuttard, tpr, t-tye, kristof.beyls, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62263
llvm-svn: 361873
Summary:
Direct sibling of D62223 patch.
While i don't have a direct motivational pattern for this,
it would seem to make sense to handle both patterns (or none),
for symmetry?
The aarch64 changes look neutral;
sparc and systemz look like improvement (one less instruction each);
x86 changes - 32bit case improves, 64bit case shows that LEA no longer
gets constructed, which may be because that whole test is `-mattr=+slow-lea,+slow-3ops-lea`
https://rise4fun.com/Alive/ffh
This is a recommit, originally committed in rL361853, but reverted
to investigate test-suite compile-time hangs.
Reviewers: RKSimon, craig.topper, spatel, t.p.northover
Reviewed By: t.p.northover
Subscribers: t.p.northover, jyknight, javed.absar, kristof.beyls, fedor.sergeev, jrtc27, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62252
llvm-svn: 361872
Summary:
The main motivation is shown by all these `neg` instructions that are now created.
In particular, the `@reg32_lshr_by_negated_unfolded_sub_b` test.
AArch64 test changes all look good (`neg` created), or neutral.
X86 changes look neutral (vectors), or good (`neg` / `xor eax, eax` created).
I'm not sure about `X86/ragreedy-hoist-spill.ll`, it looks like the spill
is now hoisted into preheader (which should still be good?),
2 4-byte reloads become 1 8-byte reload, and are elsewhere,
but i'm not sure how that affects that loop.
I'm unable to interpret AMDGPU change, looks neutral-ish?
This is hopefully a step towards solving [[ https://bugs.llvm.org/show_bug.cgi?id=41952 | PR41952 ]].
https://rise4fun.com/Alive/pkdq (we are missing more patterns, i'll submit them later)
This is a recommit, originally committed in rL361852, but reverted
to investigate test-suite compile-time hangs.
Reviewers: craig.topper, RKSimon, spatel, arsenm
Reviewed By: RKSimon
Subscribers: bjope, qcolombet, kzhuravl, jvesely, wdng, nhaehnle, yaxunl, javed.absar, dstuttard, tpr, t-tye, kristof.beyls, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62223
llvm-svn: 361871
Summary:
Again only vectors affected. Frustrating. Let me take a look into that..
https://rise4fun.com/Alive/AAq
Reviewers: RKSimon, craig.topper, spatel
Reviewed By: RKSimon
Subscribers: javed.absar, JDevlieghere, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62294
llvm-svn: 361856
Summary:
This prevents regressions in next patch,
and somewhat recovers from the regression to AMDGPU test in D62223.
It is indeed not great that we leave vector decrement,
don't transform it into vector add all-ones..
https://rise4fun.com/Alive/ZRl
Reviewers: RKSimon, craig.topper, spatel, arsenm
Reviewed By: RKSimon, arsenm
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, javed.absar, dstuttard, tpr, t-tye, kristof.beyls, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62263
llvm-svn: 361855
Summary:
Only vector tests are being affected here,
since subtraction by scalar constant is rewritten
as addition by negated constant.
No surprising test changes.
https://rise4fun.com/Alive/pbT
Reviewers: RKSimon, craig.topper, spatel
Reviewed By: RKSimon
Subscribers: javed.absar, kristof.beyls, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62257
llvm-svn: 361854
Summary:
Direct sibling of D62223 patch.
While i don't have a direct motivational pattern for this,
it would seem to make sense to handle both patterns (or none),
for symmetry?
The aarch64 changes look neutral;
sparc and systemz look like improvement (one less instruction each);
x86 changes - 32bit case improves, 64bit case shows that LEA no longer
gets constructed, which may be because that whole test is `-mattr=+slow-lea,+slow-3ops-lea`
https://rise4fun.com/Alive/ffh
Reviewers: RKSimon, craig.topper, spatel, t.p.northover
Reviewed By: t.p.northover
Subscribers: t.p.northover, jyknight, javed.absar, kristof.beyls, fedor.sergeev, jrtc27, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62252
llvm-svn: 361853
Summary:
The main motivation is shown by all these `neg` instructions that are now created.
In particular, the `@reg32_lshr_by_negated_unfolded_sub_b` test.
AArch64 test changes all look good (`neg` created), or neutral.
X86 changes look neutral (vectors), or good (`neg` / `xor eax, eax` created).
I'm not sure about `X86/ragreedy-hoist-spill.ll`, it looks like the spill
is now hoisted into preheader (which should still be good?),
2 4-byte reloads become 1 8-byte reload, and are elsewhere,
but i'm not sure how that affects that loop.
I'm unable to interpret AMDGPU change, looks neutral-ish?
This is hopefully a step towards solving [[ https://bugs.llvm.org/show_bug.cgi?id=41952 | PR41952 ]].
https://rise4fun.com/Alive/pkdq (we are missing more patterns, i'll submit them later)
Reviewers: craig.topper, RKSimon, spatel, arsenm
Reviewed By: RKSimon
Subscribers: bjope, qcolombet, kzhuravl, jvesely, wdng, nhaehnle, yaxunl, javed.absar, dstuttard, tpr, t-tye, kristof.beyls, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62223
llvm-svn: 361852
Details: To make instruction selection really divergence driven it is necessary to assign
the correct register classes to the cross block values beforehand. For the divergent targets
same value type requires different register classes dependent on the value divergence.
Reviewers: rampitec, nhaehnle
Differential Revision: https://reviews.llvm.org/D59990
This commit was reverted because of the build failure.
The reason was mlformed patch.
Build failure fixed.
llvm-svn: 361741
Details: To make instruction selection really divergence driven it is necessary to assign
the correct register classes to the cross block values beforehand. For the divergent targets
same value type requires different register classes dependent on the value divergence.
Reviewers: rampitec, nhaehnle
Differential Revision: https://reviews.llvm.org/D59990
llvm-svn: 361644
This patch adds the overridable TargetLowering::getTargetConstantFromLoad function which allows targets to return any constant value loaded by a LoadSDNode node - only X86 makes use of this so far but everything should be in place for other targets.
computeKnownBits then uses this function to improve codegen, notably vector code after legalization.
A future commit will do the same for ComputeNumSignBits but computeKnownBits sees the bigger benefit.
This required a couple of fixes:
* SimplifyDemandedBits must early-out for getTargetConstantFromLoad cases to prevent infinite loops of constant regeneration (similar to what we already do for BUILD_VECTOR).
* Fix a DAGCombiner::visitTRUNCATE issue as we had trunc(shl(v8i32),v8i16) <-> shl(trunc(v8i16),v8i32) infinite loops after legalization on AVX512 targets.
Differential Revision: https://reviews.llvm.org/D61887
llvm-svn: 361620
This is no-functional-change-intended currently because the definition
of isBinOp() only includes opcodes that produce 1 value. But if we
share that implementation with isCommutativeBinOp() as proposed in
D62191, then we need to make sure that the callers bail out for
opcodes that they are not prepared to handle correctly.
llvm-svn: 361547
There are no FP callers of DAGCombiner::reassociateOps() currently,
but we can add a fast-math check to make sure this API is not being
misused.
This was noted as a potential risk (and that risk might increase) with:
D62191
llvm-svn: 361268
This changes the isShift variable to include the constant operand
check that was previously in the if statement.
While there fix an 80 column violation and an unnecessary use of
getNode. Also fix variable name capitalization.
llvm-svn: 361168
Summary:
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?
This changes the testcase that was touched by @spatel in rL347478,
but i'm not sure that test tests anything particular?
Reviewers: RKSimon, spatel, craig.topper, jojo, rengolin
Reviewed By: spatel
Subscribers: javed.absar, llvm-commits, spatel
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D61918
llvm-svn: 361044
We catch most of these patterns (on x86 at least) by matching
a concat vectors opcode early in combining, but the pattern may
emerge later using insert subvector instead.
The AVX1 diffs for add/sub overflow show another missed narrowing
pattern. That one may be falling though the cracks because of
combine ordering and multiple uses.
llvm-svn: 360585
Summary:
When we know for sure whether two addresses do or do not alias, we
should immediately return from DAGCombiner::isAlias().
I think this comes from a bad copy/paste, Sorry for not catching that during the
code review.
Fixes PR41855.
Reviewers: niravd, gchatelet, EricWF
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D61846
llvm-svn: 360566
I noticed that we were failing to narrow an x86 ymm math op in a case similar
to the 'madd' test diff. That is because a bitcast is sitting between the math
and the extract subvector and thwarting our pattern matching for narrowing:
t56: v8i32 = add t59, t58
t68: v4i64 = bitcast t56
t73: v2i64 = extract_subvector t68, Constant:i64<2>
t96: v4i32 = bitcast t73
There are a few wins and neutral diffs in the other tests.
Differential Revision: https://reviews.llvm.org/D61806
llvm-svn: 360541
To find the candidates to merge stores we iterate over all nodes in a chain
for each store, which leads to quadratic compile times for large basic blocks
with a large number of stores.
Reviewers: niravd, spatel, craig.topper
Reviewed By: niravd
Differential Revision: https://reviews.llvm.org/D61511
llvm-svn: 360357
Add a new function to do the endian check, as I will commit another patch later, which will also need the endian check.
Differential Revision: https://reviews.llvm.org/D61236
llvm-svn: 360226
When simplifying TokenFactors, we potentially iterate over all
operands of a large number of TokenFactors. This causes quadratic
compile times in some cases and the large token factors cause additional
scalability problems elsewhere.
This patch adds some limits to the number of nodes explored for the
cases mentioned above.
Reviewers: niravd, spatel, craig.topper
Reviewed By: niravd
Differential Revision: https://reviews.llvm.org/D61397
llvm-svn: 360171
The problem was that we were creating a CMOV64rr <TargetFrameIndex>, <TargetFrameIndex>. The entire point of a TFI is that address code is not generated, so there's no way to legalize/lower this. Instead, simply prevent it's creation.
Arguably, we shouldn't be using *Target*FrameIndices in StatepointLowering at all, but that's a much deeper change.
llvm-svn: 360090
This addresses one half of https://bugs.llvm.org/show_bug.cgi?id=41635
by combining a VECREDUCE_AND/OR into VECREDUCE_UMIN/UMAX (if latter is
legal but former is not) for zero-or-all-ones boolean reductions (which
are detected based on sign bits).
Differential Revision: https://reviews.llvm.org/D61398
llvm-svn: 360054
The original patch was committed at rL359398 and reverted at rL359695 because of
infinite looping.
This includes a fix to check for a vector splat of "1.0" to avoid the infinite loop.
Original commit message:
This was originally part of D61028, but it's an independent diff.
If we try the repeated divisor reciprocal transform before producing an estimate sequence,
then we have an opportunity to use scalar fdiv. On x86, the trade-off is 1 divss vs. 5
vector FP ops in the default estimate sequence. On recent chips (Skylake, Ryzen), the
full-precision division is only 3 cycle throughput, so that's probably the better perf
default option and avoids problems from x86's inaccurate estimates.
The last 2 tests show that users still have the option to override the defaults by using
the function attributes for reciprocal estimates, but those patterns are potentially made
faster by converting the vector ops (including ymm ops) to scalar math.
Differential Revision: https://reviews.llvm.org/D61149
llvm-svn: 359793
Do not combine (trunc adde(X, Y, Carry)) into (adde trunc(X), trunc(Y), Carry),
if adde is not legal for the target. Even it's at type-legalize phase.
Because adde is special and will not be legalized at operation-legalize phase later.
This fixes: PR40922
https://bugs.llvm.org/show_bug.cgi?id=40922
Differential Revision: https://reviews.llvm.org//D60854
llvm-svn: 359532
Summary:
Extract the logic for doing reassociations
from DAGCombiner::reassociateOps into a helper
function DAGCombiner::reassociateOpsCommutative,
and use that helper to trigger reassociation
on the original operand order, or the commuted
operand order.
Codegen is not identical since the operand order will
be different when doing the reassociations for the
commuted case. That causes some unfortunate churn in
some test cases. Apart from that this should be NFC.
Reviewers: spatel, craig.topper, tstellar
Reviewed By: spatel
Subscribers: dmgreen, dschuff, jvesely, nhaehnle, javed.absar, sbc100, jgravelle-google, hiraditya, aheejin, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D61199
llvm-svn: 359476
This was originally part of D61028, but it's an independent diff.
If we try the repeated divisor reciprocal transform before producing an estimate sequence,
then we have an opportunity to use scalar fdiv. On x86, the trade-off is 1 divss vs. 5
vector FP ops in the default estimate sequence. On recent chips (Skylake, Ryzen), the
full-precision division is only 3 cycle throughput, so that's probably the better perf
default option and avoids problems from x86's inaccurate estimates.
The last 2 tests show that users still have the option to override the defaults by using
the function attributes for reciprocal estimates, but those patterns are potentially made
faster by converting the vector ops (including ymm ops) to scalar math.
Differential Revision: https://reviews.llvm.org/D61149
llvm-svn: 359398
As detailed on PR40758, Bobcat/Jaguar can perform vector immediate shifts on the same pipes as vector ANDs with the same latency - so it doesn't make sense to replace a shl+lshr with a shift+and pair as it requires an additional mask (with the extra constant pool, loading and register pressure costs).
Differential Revision: https://reviews.llvm.org/D61068
llvm-svn: 359293
If we have a vector FP division with a splatted divisor, use the existing transform
that converts 'x/y' into 'x * (1.0/y)' to allow more conversions. This can then
potentially be converted into a scalar FP division by existing combines (rL358984)
as seen in the tests here.
That can be a potentially big perf difference if scalar fdiv has better timing
(including avoiding possible frequency throttling for vector ops).
Differential Revision: https://reviews.llvm.org/D61028
llvm-svn: 359147
If we only match build vectors, we can miss some patterns
that use shuffles as seen in the affected tests.
Note that the underlying calls within getSplatSourceVector()
have the potential for compile-time explosion because of
exponential recursion looking through binop opcodes, but
currently the list of supported opcodes is very limited.
Both of those problems should be addressed in follow-up
patches.
llvm-svn: 358984
Summary:
The DAGCombiner is rewriting (canonicalizing) an ISD::ADD
with no common bits set in the operands as an ISD::OR node.
This could sometimes result in "missing out" on some
combines that normally are performed for ADD. To be more
specific this could happen if we already have rewritten an
ADD into OR, and later (after legalizations or combines)
we expose patterns that could have been optimized if we
had seen the OR as an ADD (e.g. reassociations based on ADD).
To make the DAG combiner less sensitive to if ADD or OR is
used for these "no common bits set" ADD/OR operations we
now apply most of the ADD combines also to an OR operation,
when value tracking indicates that the operands have no
common bits set.
Reviewers: spatel, RKSimon, craig.topper, kparzysz
Reviewed By: spatel
Subscribers: arsenm, rampitec, lebedev.ri, jvesely, nhaehnle, hiraditya, javed.absar, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D59758
llvm-svn: 358965
As discussed on PR41359, this patch renames the pair of shift-mask target feature functions to make their purposes more obvious.
shouldFoldShiftPairToMask -> shouldFoldConstantShiftPairToMask
preferShiftsToClearExtremeBits -> shouldFoldMaskToVariableShiftPair
llvm-svn: 358526
The checks in `canFoldInAddressingMode` tested for addressing modes that have a
base register but didn't set the `HasBaseReg` flag to true (it's false by
default). This patch fixes that. Although the omission of the flag was
technically incorrect it had no known observable impact, so no tests were
changed by this patch.
Differential Revision: https://reviews.llvm.org/D60314
llvm-svn: 358502
// shuffle (concat X, undef), (concat Y, undef), Mask -->
// concat (shuffle X, Y, Mask0), (shuffle X, Y, Mask1)
The ARM changes with 'vtrn' and narrowed 'vuzp' are improvements.
The x86 changes look neutral or better. There's one test with an
extra instruction, but that could be reversed for a subtarget with
the right attributes. But by default, we want to avoid the 256-bit
op when possible (in my motivating benchmark, a handful of ymm ops
sprinkled into a sequence of xmm ops are triggering frequency
throttling on Haswell resulting in significantly worse perf).
Differential Revision: https://reviews.llvm.org/D60545
llvm-svn: 358291
// bo (build_vec ...undef, x, undef...), (build_vec ...undef, y, undef...) -->
// build_vec ...undef, (bo x, y), undef...
The lifetime of the nodes in these examples is different for variables versus constants,
but they are all build vectors briefly, so I'm proposing to catch them in this form to
handle all of the leading examples in the motivating test file.
Before we have build vectors, we might have insert_vector_element. After that, we might
have scalar_to_vector and constant pool loads.
It's going to take more work to ensure that FP vector operands are getting simplified
with undef elements, so this transform can apply more widely. In a non-loose FP environment,
we are likely simplifying FP elements to NaN values rather than undefs.
We also need to allow more opcodes down this path. Eg, we don't handle FP min/max flavors
yet.
Differential Revision: https://reviews.llvm.org/D60514
llvm-svn: 358172
This lines up with what we do for regular subtract and it matches up better with X86 assumptions in isel patterns that add with immediate is more canonical than sub with immediate.
Differential Revision: https://reviews.llvm.org/D60020
llvm-svn: 358027
There are a variety of vector patterns that may be profitably reduced to a
scalar op when scalar ops are performed using a subset (typically, the
first lane) of the vector register file.
For x86, this is true for float/double ops and element 0 because
insert/extract is just a sub-register rename.
Other targets should likely enable the hook in a similar way.
Differential Revision: https://reviews.llvm.org/D60150
llvm-svn: 357760
There are 3 changes to make this correspond to the same transform in instcombine:
1. Remove the legality check - we can't create anything less legal than we started with.
2. Ease the use restriction, so we only bail out if both operands have >1 use.
3. Ease the use restriction for binops with a repeated operand (eg, mul x, x).
As discussed in D60150, there's a scalarization opportunity that will be made
easier by allowing this transform more generally.
llvm-svn: 357580
Summary:
Nodes that have no uses are eventually pruned when they are selected
from the worklist. Record nodes newly added to the worklist or DAG and
perform pruning after every combine attempt.
Reviewers: efriedma, RKSimon, craig.topper, spatel, jyknight
Reviewed By: jyknight
Subscribers: jdoerfert, jyknight, nemanjai, jvesely, nhaehnle, javed.absar, hiraditya, jsji, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D58070
llvm-svn: 357283
Summary:
Various SelectionDAG non-combine operations (e.g. the getNode smart
constructor and legalization) may leave dangling nodes by applying
optimizations without fully pruning unused result values. This results
in nodes that are never added to the worklist and therefore can not be
pruned.
Add a node inserter for the combiner to make sure such nodes have the
chance of being pruned. This allows a number of additional peephole
optimizations.
Reviewers: efriedma, RKSimon, craig.topper, jyknight
Reviewed By: jyknight
Subscribers: msearles, jyknight, sdardis, nemanjai, javed.absar, hiraditya, jrtc27, atanasyan, jsji, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D58068
llvm-svn: 357279
After investigating the examples from D59777 targeting an SSE4.1 machine,
it looks like a very different problem due to how we map illegal types (256-bit in these cases).
We're missing a shuffle simplification that maps elements of a vector back to a shuffled operand.
We have a more general version of this transform in DAGCombiner::visitVECTOR_SHUFFLE(), but that
generality means it is limited to patterns with a one-use constraint, and the examples here have
2 uses. We don't need any uses or legality limitations for a simplification (no new value is
created).
It looks like we miss this pattern in IR too.
In one of the zext examples here, we have shuffle masks like this:
Shuf0 = vector_shuffle<0,u,3,7,0,u,3,7>
Shuf = vector_shuffle<4,u,6,7,u,u,u,u>
...so that's moving the high half of the 1st vector into the low half. But the high half of the
1st vector is already identical to the low half.
Differential Revision: https://reviews.llvm.org/D59961
llvm-svn: 357258
This is a sibling to rL357178 that I noticed we'd hit if we chose
an alternate transform in D59818.
%z = zext i8 %x to i32
%dec = add i32 %z, -1
%r = sext i32 %dec to i64
=>
%z2 = zext i8 %x to i64
%r = add i64 %z2, -1
https://rise4fun.com/Alive/kPP
The x86 vector diffs show a slight regression, so there's a chance
that we should limit this and the previous transform to scalars.
But given that we allowed vectors before, I'm matching that behavior
here. We should change both transforms together if that's the right
thing to do.
llvm-svn: 357254
If scalar truncates are free, attempt to pre-truncate build_vectors source operands.
Only attempt to do this before legalization as we often end up with truncations/extensions during build_vector lowering.
Differential Revision: https://reviews.llvm.org/D59654
llvm-svn: 357161
Rework BaseIndexOffset and isAlias to fully work with lifetime nodes
and fold in lifetime alias analysis.
This is mostly NFC.
Reviewers: courbet
Reviewed By: courbet
Subscribers: hiraditya, jdoerfert, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D59794
llvm-svn: 357070
getAsCarry() checks that the input argument is a carry-producing node before
allowing a transformation to addcarry. This patch adds a check to make sure
that the carry-producing node is legal. If it is not, it may not remain in a
form that is manageable by the target backend. The test case caused a
compilation failure during instruction selection for this reason on SystemZ.
Patch by Ulrich Weigand.
Review: Sanjay Patel
https://reviews.llvm.org/D59822
llvm-svn: 357052
Various SelectionDAG non-combine operations (e.g. the getNode smart
constructor and legalization) may leave dangling nodes by applying
optimizations or not fully pruning unused result values. This can
result in nodes that are never added to the worklist and therefore can
not be pruned.
Add a node inserter as the current node deleter to make sure such
nodes have the chance of being pruned.
Many minor changes, mostly positive.
llvm-svn: 356996
This helps us relax the extension of a lot of scalar elements before they are inserted into a vector.
Its exposes an issue in DAGCombiner::convertBuildVecZextToZext as some/all the zero-extensions may be relaxed to ANY_EXTEND, so we need to handle that case to avoid a couple of AVX2 VPMOVZX test regressions.
Once this is in it should be easier to fix a number of remaining failures to fold loads into VBROADCAST nodes.
Differential Revision: https://reviews.llvm.org/D59484
llvm-svn: 356989
SDNodes can only have 64k operands and for some inputs (e.g. large
number of stores), we can reach this limit when creating TokenFactor
nodes. This patch is a follow up to D56740 and updates a few more places
that potentially can create TokenFactors with too many operands.
Reviewers: efriedma, craig.topper, aemerson, RKSimon
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D59156
llvm-svn: 356668
In r311255 we added a case where we split vectors whose elements are
all derived from the same input vector so that we could shuffle it
more efficiently. In doing so, createBuildVecShuffle was taught to
adjust for the fact that all indices would be based off of the first
vector when this happens, but it's possible for the code that checked
that to fire incorrectly if we happen to have a BUILD_VECTOR of
extracts from subvectors and don't hit this new optimization.
Instead of trying to detect if we've split the vector by checking if
we have extracts from the same base vector, we can just pass that
information into createBuildVecShuffle, avoiding the miscompile.
Differential Revision: https://reviews.llvm.org/D59507
llvm-svn: 356476
These changes are related to PR37743 and include:
SelectionDAGBuilder::visitSelect handles the unary SelectPatternFlavor::SPF_ABS case to build ABS node.
Delete the redundant recognizer of the integer ABS pattern from the DAGCombiner.
Add promoting the integer ABS node in the LegalizeIntegerType.
Expand-based legalization of integer result for the ABS nodes.
Expand-based legalization of ABS vector operations.
Add some integer abs testcases for different typesizes for Thumb arch
Add the custom ABS expanding and change the SAD pattern recognizer for X86 arch: The i64 result of the ABS is expanded to:
tmp = (SRA, Hi, 31)
Lo = (UADDO tmp, Lo)
Hi = (XOR tmp, (ADDCARRY tmp, hi, Lo:1))
Lo = (XOR tmp, Lo)
The "detectZextAbsDiff" function is changed for the recognition of pattern with the ABS node. Given a ABS node, detect the following pattern:
(ABS (SUB (ZERO_EXTEND a), (ZERO_EXTEND b))).
Change integer abs testcases for codegen with the ABS node support for AArch64.
Indicate that the ABS is legal for the i64 type when the NEON is supported.
Change the integer abs testcases to show changing of codegen.
Add combine and legalization of ABS nodes for Thumb arch.
Extend 'matchSelectPattern' to recognize the ABS patterns with ICMP_SGE condition.
For discussion, see https://bugs.llvm.org/show_bug.cgi?id=37743
Patch by: @ikulagin (Ivan Kulagin)
Differential Revision: https://reviews.llvm.org/D49837
llvm-svn: 356468
This allows better code size for aarch64 floating point materialization
in a future patch.
Reviewers: evandro
Differential Revision: https://reviews.llvm.org/D58690
llvm-svn: 356389
Delete temporarily constructed node uses for analysis after it's use,
holding onto original input nodes. Ideally this would be rewritten
without making nodes, but this appears relatively complex.
Reviewers: spatel, RKSimon, craig.topper
Subscribers: jdoerfert, hiraditya, deadalnix, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D57921
llvm-svn: 356382
Fold (x & ~y) | y and it's four commuted variants to x | y. This pattern
can in particular appear when a vselect c, x, -1 is expanded to
(x & ~c) | (-1 & c) and combined to (x & ~c) | c.
This change has some overlap with D59066, which avoids creating a
vselect of this form in the first place during uaddsat expansion.
Differential Revision: https://reviews.llvm.org/D59174
llvm-svn: 356333
rL356292 reduces the size of scalar_to_vector if we know the upper bits are undef - which means that shuffles may find they are suddenly referencing scalar_to_vector elements other than zero - so make sure we handle this as undef.
llvm-svn: 356327
Summary:
A number of optimizations are inhibited by single-use TokenFactors not
being merged into the TokenFactor using it. This makes we consider if
we can do the merge immediately.
Most tests changes here are due to the change in visitation causing
minor reorderings and associated reassociation of paired memory
operations.
CodeGen tests with non-reordering changes:
X86/aligned-variadic.ll -- memory-based add folded into stored leaq
value.
X86/constant-combiners.ll -- Optimizes out overlap between stores.
X86/pr40631_deadstore_elision -- folds constant byte store into
preceding quad word constant store.
Reviewers: RKSimon, craig.topper, spatel, efriedma, courbet
Reviewed By: courbet
Subscribers: dylanmckay, sdardis, nemanjai, jvesely, nhaehnle, javed.absar, eraman, hiraditya, kbarton, jrtc27, atanasyan, jsji, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D59260
llvm-svn: 356068
Fixes https://bugs.llvm.org/show_bug.cgi?id=36796.
Implement basic legalizations (PromoteIntRes, PromoteIntOp,
ExpandIntRes, ScalarizeVecOp, WidenVecOp) for VECREDUCE opcodes.
There are more legalizations missing (esp float legalizations),
but there's no way to test them right now, so I'm not adding them.
This also includes a few more changes to make this work somewhat
reasonably:
* Add support for expanding VECREDUCE in SDAG. Usually
experimental.vector.reduce is expanded prior to codegen, but if the
target does have native vector reduce, it may of course still be
necessary to expand due to legalization issues. This uses a shuffle
reduction if possible, followed by a naive scalar reduction.
* Allow the result type of integer VECREDUCE to be larger than the
vector element type. For example we need to be able to reduce a v8i8
into an (nominally) i32 result type on AArch64.
* Use the vector operand type rather than the scalar result type to
determine the action, so we can control exactly which vector types are
supported. Also change the legalize vector op code to handle
operations that only have vector operands, but no vector results, as
is the case for VECREDUCE.
* Default VECREDUCE to Expand. On AArch64 (only target using VECREDUCE),
explicitly specify for which vector types the reductions are supported.
This does not handle anything related to VECREDUCE_STRICT_*.
Differential Revision: https://reviews.llvm.org/D58015
llvm-svn: 355860
Move the x86 combine from D58974 into the DAGCombine VSELECT code and update the SELECT version to use the isBooleanFlip helper as well.
Requested by @spatel on D59006
llvm-svn: 355533
This patch enables combining integer bitcasts of integer build vectors when the new scalar type is legal. I've avoided floating point because the implementation bitcasts float to int along the way and we would need to check the intermediate types for legality
Differential Revision: https://reviews.llvm.org/D58884
llvm-svn: 355324
Support undef shuffle mask indices in the shuffle(concat_vectors, concat_vectors) -> concat_vectors fold
Differential Revision: https://reviews.llvm.org/D58585
llvm-svn: 354793
This fold can occur during legalization, so it can fight with promotion
to the larger type. It apparently takes a special sequence and subtarget
to avoid more basic simplifications that would hide the problem.
But there's a bigger question raised here: why does distributeTruncateThroughAnd()
even exist? It duplicates functionality from a more minimal pattern that we
already have. But getting rid of this function requires some preliminary steps.
https://bugs.llvm.org/show_bug.cgi?id=40793
llvm-svn: 354594
Summary:
A store to an object whose lifetime is about to end can be removed.
See PR40550 for motivation.
Reviewers: niravd
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D57541
llvm-svn: 354244
If we're comparing some value for equality against 2 constants
and those constants have an absolute difference of just 1 bit,
then we can offset and mask off that 1 bit and reduce to a single
compare against zero:
and/or (setcc X, C0, ne), (setcc X, C1, ne/eq) -->
setcc ((add X, -C1), ~(C0 - C1)), 0, ne/eq
https://rise4fun.com/Alive/XslKj
This transform is disabled by default using a TLI hook
("convertSetCCLogicToBitwiseLogic()").
That should be overridden for AArch64, MIPS, Sparc and possibly
others based on the asm shown in:
https://bugs.llvm.org/show_bug.cgi?id=40611
llvm-svn: 353859
Now that we have SimplifyDemandedBits support for funnel shifts (rL353539), we need to simplify funnel shifts back to bitshifts in cases where either argument has been folded to undef/zero.
Differential Revision: https://reviews.llvm.org/D58009
llvm-svn: 353645
The sqrt case is faster and we already do this for the case where
the exponent is 0.25. This adds the 0.75 case which is also not
sensitive to signed zeros.
Patch by Whitney Tsang (Whitney)
Differential revision: https://reviews.llvm.org/D57434
llvm-svn: 353557
Move the (add (umax X, C), -C) --> (usubsat X, C) X86 combine into generic DAGCombiner
First of a number of saturated arithmetic folds that can be moved out of X86-specific code for PR40111.
Differential Revision: https://reviews.llvm.org/D57754
llvm-svn: 353457
I noticed that we are missing this canonicalization in IR:
rL352515
...and then realized that we don't get this right in SDAG either,
so this has to be fixed first regardless of what we choose to do in IR.
The existing fold was limited to scalars and using the wrong predicate
to guard the transform. We have a boolean contents TLI query that can
be used to decide which direction to fold.
This may eventually lead back to the problems/question in:
https://bugs.llvm.org/show_bug.cgi?id=40486
...but it makes no difference to that yet.
Differential Revision: https://reviews.llvm.org/D57401
llvm-svn: 353433
Summary:
If the index isn't constant, this transform inserts a multiply and an add on the index to calculating the base pointer for a scalar load. But we still create a memory operand with an offset of 0 and the size of the scalar access. But the access is really to an unknown offset within the original access size.
This can cause the machine scheduler to incorrectly calculate dependencies between this load and other accesses. In the case we saw, there was a 32 byte vector store that was split into two 16 byte stores, one with offset 0 and one with offset 16. The size of the memory operand for both was 16. The scheduler correctly detected the alias with the offset 0 store, but not the offset 16 store.
This patch discards the pointer info so we don't incorrectly detect aliasing. I wasn't sure if we could keep using the original offset and size without risking some other transform on the load changing the size.
I tried to reduce a test case, but there's still a lot of memory operations needed to get the scheduler to do the bad reordering. So it looked pretty fragile to maintain.
Reviewers: efriedma
Reviewed By: efriedma
Subscribers: arphaman, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D57616
llvm-svn: 353124
Noticed while investigating PR40483, and fixes the basic test case from the bug - but not a more general case.
We're pretty weak at dealing with ADD/SUB combines compared to the SimplifyAssociativeOrCommutative/SimplifyUsingDistributiveLaws abilities that InstCombine can manage.
llvm-svn: 353044
We already have the getConstantOperandVal helper which returns a uint64_t, but along comes the fuzzer and inserts a i128 -1 constant or something and the whole thing asserts.......
I've updated a few obvious cases, and tried to make use of the const reference where possible, but there's more to do. A number of existing oss-fuzz tickets should be fixed if we start using APInt and perform value clamping where necessary.
llvm-svn: 352961
This patch fixes pr39098.
For the attached test case, CombineZExtLogicopShiftLoad can optimize it to
t25: i64 = Constant<1099511627775>
t35: i64 = Constant<0>
t0: ch = EntryToken
t57: i64,ch = load<(load 4 from `i40* undef`, align 8), zext from i32> t0, undef:i64, undef:i64
t58: i64 = srl t57, Constant:i8<1>
t60: i64 = and t58, Constant:i64<524287>
t29: ch = store<(store 5 into `i40* undef`, align 8), trunc to i40> t57:1, t60, undef:i64, undef:i64
But later visitANDLike transforms it to
t25: i64 = Constant<1099511627775>
t35: i64 = Constant<0>
t0: ch = EntryToken
t57: i64,ch = load<(load 4 from `i40* undef`, align 8), zext from i32> t0, undef:i64, undef:i64
t61: i32 = truncate t57
t63: i32 = srl t61, Constant:i8<1>
t64: i32 = and t63, Constant:i32<524287>
t65: i64 = zero_extend t64
t58: i64 = srl t57, Constant:i8<1>
t60: i64 = and t58, Constant:i64<524287>
t29: ch = store<(store 5 into `i40* undef`, align 8), trunc to i40> t57:1, t60, undef:i64, undef:i64
And it triggers CombineZExtLogicopShiftLoad again, causes a dead loop.
Both forms should generate same instructions, CombineZExtLogicopShiftLoad generated IR looks cleaner. But it looks more difficult to prevent visitANDLike to do the transform, so I prevent CombineZExtLogicopShiftLoad to do the transform if the ZExt is free.
Differential Revision: https://reviews.llvm.org/D57491
llvm-svn: 352792
While dangling nodes will eventually be pruned when they are
considered, leaving them disables combines requiring single-use.
Reviewers: Carrot, spatel, craig.topper, RKSimon, efriedma
Subscribers: hiraditya, llvm-commits
Differential Revision: https://reviews.llvm.org/D57520
llvm-svn: 352784
This extends the existing transform for:
add X, 0/1 --> sub X, 0/-1
...to allow the sibling subtraction fold.
This pattern could regress with the proposed change in D57401.
llvm-svn: 352680
This is the sibling fold for insert-of-insert that was added with D56604.
Now that we have x86 shuffle narrowing (D57156), this change shows improvements for
lots of AVX512 reduction code (not sure that we would ever expect extract-of-extract otherwise).
There's a small regression in some of the partial-permute tests (extracting followed by splat).
That is tracked by PR40500:
https://bugs.llvm.org/show_bug.cgi?id=40500
Differential Revision: https://reviews.llvm.org/D57336
llvm-svn: 352528
The current check in CombineToPreIndexedLoadStore is too
conversative, preventing a pre-indexed store when the base pointer
is a predecessor of the value being stored. Instead, we should check
the pointer operand of the store.
Differential Revision: https://reviews.llvm.org/D56719
llvm-svn: 351933
vecbo (insertsubv undef, X, Z), (insertsubv undef, Y, Z) --> insertsubv VecC, (vecbo X, Y), Z
This is another step in generic vector narrowing. It's also a step towards more horizontal op
formation specifically for x86 (although we still failed to match those in the affected tests).
The scalarization cases are also not optimal (we should be scalarizing those), but it's still
an improvement to use a narrower vector op when we know part of the result must be constant
because both inputs are undef in some vector lanes.
I think a similar match but checking for a constant operand might help some of the cases in
D51553.
Differential Revision: https://reviews.llvm.org/D56875
llvm-svn: 351825
The regression test is reduced from the example shown in D56281.
This does raise a question as noted in the test file: do we want
to handle this pattern? I don't have a motivating example for
that on x86 yet, but it seems like we could have that pattern
there too, so we could avoid the back-and-forth using a shuffle.
llvm-svn: 351753
to reflect the new license.
We understand that people may be surprised that we're moving the header
entirely to discuss the new license. We checked this carefully with the
Foundation's lawyer and we believe this is the correct approach.
Essentially, all code in the project is now made available by the LLVM
project under our new license, so you will see that the license headers
include that license only. Some of our contributors have contributed
code under our old license, and accordingly, we have retained a copy of
our old license notice in the top-level files in each project and
repository.
llvm-svn: 351636
Similar to D55073. Without this change, the DAG combiner crashes on code
with more than 64k of stores in a single basic block that form parallelizable
chains.
No test case, as it would be very IR file.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D56740
llvm-svn: 351571
ReduceLoadWidth can trigger using a shifted mask is used and this
requires that the function return a shl node to correct for the
offset. However, the way that this was implemented meant that the
returned result could be an existing node, which would be incorrect.
This fixes the method of inserting the new node and replacing uses.
Differential Revision: https://reviews.llvm.org/D50432
llvm-svn: 351310
The motivating case for this is shown in the first regression test. We are
transferring to scalar and back rather than just zero-extending with 'vpmovzxdq'.
That's a special-case for a more general pattern as shown here. In all tests,
we're avoiding the vector-scalar-vector moves in favor of vector ops.
We aren't producing optimal shuffle code in some cases though, so the patch is
limited to reduce regressions.
Differential Revision: https://reviews.llvm.org/D56281
llvm-svn: 351198
This pattern:
t33: v8i32 = insert_subvector undef:v8i32, t35, Constant:i64<0>
t21: v16i32 = insert_subvector undef:v16i32, t33, Constant:i64<0>
...shows up in PR33758:
https://bugs.llvm.org/show_bug.cgi?id=33758
...although this patch doesn't make any difference to the final result on that yet.
In the affected tests here, it looks like it just makes RA wiggle. But we might
as well squash this to prevent it interfering with other pattern-matching.
Differential Revision:
https://reviews.llvm.org/D56604
llvm-svn: 351008
As noted in PR39973 and D55558:
https://bugs.llvm.org/show_bug.cgi?id=39973
...this is a partial implementation of a fold that we do as an IR canonicalization in instcombine:
// extelt (binop X, Y), Index --> binop (extelt X, Index), (extelt Y, Index)
We want to have this in the DAG too because as we can see in some of the test diffs (reductions),
the pattern may not be visible in IR.
Given that this is already an IR canonicalization, any backend that would prefer a vector op over
a scalar op is expected to already have the reverse transform in DAG lowering (not sure if that's
a realistic expectation though). The transform is limited with a TLI hook because there's an
existing transform in CodeGenPrepare that tries to do the opposite transform.
Differential Revision: https://reviews.llvm.org/D55722
llvm-svn: 350354
Currently we expand the two nodes separately. This gives DAG combiner an opportunity to optimize the expanded sequence taking into account only one set of users. When we expand the other node we'll create the expansion again, but might not be able to optimize it the same way. So the nodes won't CSE and we'll have two similarish sequences in the same basic block. By expanding both nodes at the same time we'll avoid prematurely optimizing the expansion until both the division and remainder have been replaced.
Improves the test case from PR38217. There may be additional opportunities after this.
Differential Revision: https://reviews.llvm.org/D56145
llvm-svn: 350239
If x has multiple sign bits than it doesn't matter which one we extend from so we can sext from x's msb instead.
The X86 setcc-combine.ll changes are a little weird. It appears we ended up with a (sext_inreg (aext (trunc (extractelt)))) after type legalization. The sext_inreg+aext now gets optimized by this combine to leave (sext (trunc (extractelt))). Then we visit the trunc before we visit the sext. This ends up changing the truncate to an extractvectorelt from a bitcasted vector. I have a follow up patch to fix this.
Differential Revision: https://reviews.llvm.org/D56156
llvm-svn: 350235
It's dangerous to knowingly create an illegal vector type
no matter what stage of combining we're in.
This prevents the missed folding/scalarization seen in:
https://bugs.llvm.org/show_bug.cgi?id=40146
llvm-svn: 350034
trunc (add X, C ) --> add (trunc X), C'
If we're throwing away the top bits of an 'add' instruction, do it in the narrow destination type.
This makes the truncate-able opcode list identical to the sibling transform done in IR (in instcombine).
This change used to show regressions for x86, but those are gone after D55494.
This gets us closer to deleting the x86 custom function (combineTruncatedArithmetic)
that does almost the same thing.
Differential Revision: https://reviews.llvm.org/D55866
llvm-svn: 350006
This saves materializing the immediate. The additional forms are less
common (they don't usually show up for bitfield insert/extract), but
they're still relevant.
I had to add a new target hook to prevent DAGCombine from reversing the
transform. That isn't the only possible way to solve the conflict, but
it seems straightforward enough.
Differential Revision: https://reviews.llvm.org/D55630
llvm-svn: 349857
Now that SimplifyDemandedBits/SimplifyDemandedVectorElts is simplifying vector elements, we're seeing more constant BUILD_VECTOR containing undefs.
This patch provides opt-in support for UNDEF elements in matchBinaryPredicate, passing NULL instead of the result ConstantSDNode* argument.
I've updated the (or (and X, c1), c2) -> (and (or X, c2), c1|c2) fold to demonstrate its use, which I believe is safe for undef cases.
Differential Revision: https://reviews.llvm.org/D55822
llvm-svn: 349629
As described on PR40091, we have several places where zext (and zext_vector_inreg) fold an undef input into an undef output. For zero extensions this is incorrect as the output should guarantee to least have the new upper bits set to zero.
SimplifyDemandedVectorElts is the worst offender (and its the most likely to cause new undefs to appear) but DAGCombiner's tryToFoldExtendOfConstant has a similar issue.
Thanks to @dmgreen for catching this.
Differential Revision: https://reviews.llvm.org/D55883
llvm-svn: 349625
The transform performs a bitwise logic op in a wider type followed by
truncate when both inputs are truncated from the same source type:
logic_op (truncate x), (truncate y) --> truncate (logic_op x, y)
There are a bunch of other checks that should prevent doing this when
it might be harmful.
We already do this transform for scalars in this spot. The vector
limitation was shared with a check for the case when the operands are
extended. I'm not sure if that limit is needed either, but that would
be a separate patch.
Differential Revision: https://reviews.llvm.org/D55448
llvm-svn: 349303
Also exposes an issue in DAGCombiner::visitFunnelShift where we were assuming the shift amount had the result type (after legalization it'll have the targets shift amount type).
llvm-svn: 349298
Summary:
If the setcc already has the target desired type we can reach the getSetCC/getSExtOrTrunc after the MatchingVecType check with the exact same types as the nodes we started with. This causes those causes VsetCC to be CSEd to N0 and the getSExtOrTrunc will CSE to N. When we return N, the caller will think that meant we called CombineTo and did our own worklist management. But that's not what happened. This prevents target hooks from being called for the node.
To fix this, I've now returned SDValue if the setcc is already the desired type. But to avoid some regressions in X86 I've had to disable one of the target combines that wasn't being reached before in the case of a (sext (setcc)). If we get vector widening legalization enabled that entire function will be deleted anyway so hopefully this is only for the short term.
Reviewers: RKSimon, spatel
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D55459
llvm-svn: 349137
This isn't quite NFC, but I don't know how to expose
any outward diffs from these changes. Mostly, this
was confusing because it used 'VT' to refer to the
operand type rather the usual type of the input node.
There's also a large block at the end that is dedicated
solely to matching loads, but that wasn't obvious. This
could probably be split up into separate functions to
make it easier to see.
It's still not clear to me when we make certain transforms
because the legality and constant conditions are
intertwined in a way that might be improved.
llvm-svn: 349095
This is a retry of rL349051 (reverted at rL349056). I changed the check for dead-ness from
number of uses to an opcode test for DELETED_NODE based on existing similar code.
Differential Revision: https://reviews.llvm.org/D55655
llvm-svn: 349058
As discussed on D55511, this caused an issue if the inner node deletes a node that the outer node depends upon. As it doesn't affect any lit-tests and I've only been able to expose this with the D55511 change I'm committing this now.
llvm-svn: 348781
This triggers an assert when combining concat_vectors of a bitcast of
merge_values.
With asserts disabled, it fails to select:
fatal error: error in backend: Cannot select: 0x7ff19d000e90: i32 = any_extend 0x7ff19d000ae8
0x7ff19d000ae8: f64,ch = CopyFromReg 0x7ff19d000c20:1, Register:f64 %1
0x7ff19d000b50: f64 = Register %1
In function: d
Differential Revision: https://reviews.llvm.org/D55507
llvm-svn: 348759
This is effectively re-committing the changes from:
rL347917 (D54640)
rL348195 (D55126)
...which were effectively reverted here:
rL348604
...because the code had a bug that could induce infinite looping
or eventual out-of-memory compilation.
The bug was that this code did not guard against transforming
opaque constants. More details are in the post-commit mailing
list thread for r347917. A reduced test for that is included
in the x86 bool-math.ll file. (I wasn't able to reduce a PPC
backend test for this, but it was almost the same pattern.)
Original commit message for r347917:
The motivating case for this is shown in:
https://bugs.llvm.org/show_bug.cgi?id=32023
and the corresponding rot16.ll regression tests.
Because x86 scalar shift amounts are i8 values, we can end up with trunc-binop-trunc
sequences that don't get folded in IR.
As the TODO comments suggest, there will be regressions if we extend this (for x86,
we mostly seem to be missing LEA opportunities, but there are likely vector folds
missing too). I think those should be considered existing bugs because this is the
same transform that we do as an IR canonicalization in instcombine. We just need
more tests to make those visible independent of this patch.
llvm-svn: 348706
As discussed in the post-commit thread of r347917, this
transform is fighting with an existing transform causing
an infinite loop or out-of-memory, so this is effectively
reverting r347917 and its follow-up r348195 while we
investigate the bug.
llvm-svn: 348604
If this is not a valid way to assign an SDLoc, then we get this
wrong all over SDAG.
I don't know enough about the SDAG to explain this. IIUC, theoretically,
debug info is not supposed to affect codegen. But here it has clearly
affected 3 different targets, and the x86 change is an actual improvement.
llvm-svn: 348552
We shouldn't care about the debug location for a node that
we're creating, but attaching the root of the pattern should
be the best effort. (If this is not true, then we are doing
it wrong all over the SDAG).
This is no-functional-change-intended, and there are no
regression test diffs...and that's what I expected. But
there's a similar line above this diff, where those
assumptions apparently do not hold.
llvm-svn: 348550
This was probably organized as it was because bswap is a unary op.
But that's where the similarity to the other opcodes ends. We should
not limit this transform to scalars, and we should not try it if
either input has other uses. This is another step towards trying to
clean this whole function up to prevent it from causing infinite loops
and memory explosions.
Earlier commits in this series:
rL348501
rL348508
rL348518
llvm-svn: 348534
Unlike some of the folds in hoistLogicOpWithSameOpcodeHands()
above this shuffle transform, this has the expected hasOneUse()
checks in place.
llvm-svn: 348523
This patch introduces a new DAGCombiner rule to simplify concat_vectors nodes:
concat_vectors( bitcast (scalar_to_vector %A), UNDEF)
--> bitcast (scalar_to_vector %A)
This patch only partially addresses PR39257. In particular, it is enough to fix
one of the two problematic cases mentioned in PR39257. However, it is not enough
to fix the original test case posted by Craig; that particular case would
probably require a more complicated approach (and knowledge about used bits).
Before this patch, we used to generate the following code for function PR39257
(-mtriple=x86_64 , -mattr=+avx):
vmovsd (%rdi), %xmm0 # xmm0 = mem[0],zero
vxorps %xmm1, %xmm1, %xmm1
vblendps $3, %xmm0, %xmm1, %xmm0 # xmm0 = xmm0[0,1],xmm1[2,3]
vmovaps %ymm0, (%rsi)
vzeroupper
retq
Now we generate this:
vmovsd (%rdi), %xmm0 # xmm0 = mem[0],zero
vmovaps %ymm0, (%rsi)
vzeroupper
retq
As a side note: that VZEROUPPER is completely redundant...
I guess the vzeroupper insertion pass doesn't realize that the definition of
%xmm0 from vmovsd is already zeroing the upper half of %ymm0. Note that on
%-mcpu=btver2, we don't get that vzeroupper because pass vzeroupper insertion
%pass is disabled.
Differential Revision: https://reviews.llvm.org/D55274
llvm-svn: 348522
The PPC test with 2 extra uses seems clearly better by avoiding this transform.
With 1 extra use, we also prevent an extra register move (although that might
be an RA problem). The general rule should be to only make a change here if
it is always profitable. The x86 diffs are all neutral.
llvm-svn: 348518
The AVX512 diffs are neutral, but the bswap test shows a clear overreach in
hoistLogicOpWithSameOpcodeHands(). If we don't check for other uses, we can
increase the instruction count.
This could also fight with transforms trying to go in the opposite direction
and possibly blow up/infinite loop. This might be enough to solve the bug
noted here:
http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20181203/608593.html
I did not add the hasOneUse() checks to all opcodes because I see a perf
regression for at least one opcode. We may decide that's irrelevant in the
face of potential compiler crashing, but I'll see if I can salvage that first.
llvm-svn: 348508
Because we're potentially peeking through a bitcast in this transform,
we need to use overall bitwidths rather than number of elements to
determine when it's safe to proceed.
Should fix:
https://bugs.llvm.org/show_bug.cgi?id=39893
llvm-svn: 348383
This is an initial patch to add a minimum level of support for funnel shifts to the SelectionDAG and to begin wiring it up to the X86 SHLD/SHRD instructions.
Some partial legalization code has been added to handle the case for 'SlowSHLD' where we want to expand instead and I've added a few DAG combines so we don't get regressions from the existing DAG builder expansion code.
Differential Revision: https://reviews.llvm.org/D54698
llvm-svn: 348353
Add support for ISD::*_EXTEND and ISD::*_EXTEND_VECTOR_INREG opcodes.
The extra broadcast in trunc-subvector.ll will be fixed in an upcoming patch.
llvm-svn: 348246
This is the smallest vector enhancement I could find to D54640.
Here, we're allowing narrowing to only legal vector ops because we'll see
regressions without that. All of the test diffs are wins from what I can tell.
With AVX/AVX512, we can shrink ymm/zmm ops to xmm.
x86 vector multiplies are the problem case that we're avoiding due to the
patchwork ISA, and it's not clear to me if we can dance around those
regressions using TLI hooks or if we need preliminary patches to plug those
holes.
Differential Revision: https://reviews.llvm.org/D55126
llvm-svn: 348195
Summary:
Under -x86-experimental-vector-widening-legalization, fp_to_uint/fp_to_sint with a smaller than 128 bit vector type results are custom type legalized by promoting the result to a 128 bit vector by promoting the elements, inserting an assertzext/assertsext, then truncating back to original type. The truncate will be further legalizdd to a pack shuffle. In the case of a v8i8 result type, we'll end up with a v8i16 fp_to_sint. This will need to be further legalized during vector op legalization by promoting to v8i32 and then truncating again. Under avx2 this produces good code with two pack instructions, but Under avx512 this will result in a truncate instruction and a packuswb instruction. But we should be able to get away with a single truncate instruction.
The other option is to promote all the way to vXi32 result type during the first type legalization. But in some experimentation that seemed to require more work to produce good code for other configurations.
Reviewers: RKSimon, spatel
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D54836
llvm-svn: 348158
This change prevents the crash noted in the post-commit comments
for rL347478 :
http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20181119/605166.html
We can't guarantee that an oversized shift amount is folded away,
so we have to check for it.
Note that I committed an incomplete fix for that crash with:
rL347502
But as discussed here:
http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20181126/605679.html
...we have to try harder.
So I'm not sure how to expose the bug now (and apparently no fuzzers have found
a way yet either).
On the plus side, we have discovered that we're missing real optimizations by
not simplifying nodes sooner, so the earlier fix still has value, and there's
likely more value in extending that so we can simplify more opcodes and simplify
when doing RAUW and/or putting nodes on the combiner worklist.
Differential Revision: https://reviews.llvm.org/D54954
llvm-svn: 348089
The motivating case for this is shown in:
https://bugs.llvm.org/show_bug.cgi?id=32023
and the corresponding rot16.ll regression tests.
Because x86 scalar shift amounts are i8 values, we can end up with trunc-binop-trunc
sequences that don't get folded in IR.
As the TODO comments suggest, there will be regressions if we extend this (for x86,
we mostly seem to be missing LEA opportunities, but there are likely vector folds
missing too). I think those should be considered existing bugs because this is the
same transform that we do as an IR canonicalization in instcombine. We just need
more tests to make those visible independent of this patch.
Differential Revision: https://reviews.llvm.org/D54640
llvm-svn: 347917
This should likely be adjusted to limit this transform
further, but these diffs should be clear wins.
If we have blendv/conditional move, then we should assume
those are cheap ops. The loads become independent of the
compare, so those can be speculated before we need to use
the values in the blend/mov.
llvm-svn: 347526
rL347502 moved the null sibling, so we should group all of these
together. I'm not sure why these aren't methods of the SDValue
class itself, but that's another patch if that's possible.
llvm-svn: 347523
...and use them to avoid creating obviously undef values as
discussed in the post-commit thread for r347478.
The diffs in vector div/rem show that we were missing real
optimizations by creating bogus shift nodes.
llvm-svn: 347502
We fail to canonicalize IR this way (prefer 'not' ops to arbitrary 'xor'),
but that would not matter without this patch because DAGCombiner was
reversing that transform. I think we need this transform in the backend
regardless of what happens in IR to catch cases where the shift-xor
is formed late from GEP or other ops.
https://rise4fun.com/Alive/NC1
Name: shl
Pre: (-1 << C2) == C1
%shl = shl i8 %x, C2
%r = xor i8 %shl, C1
=>
%not = xor i8 %x, -1
%r = shl i8 %not, C2
Name: shr
Pre: (-1 u>> C2) == C1
%sh = lshr i8 %x, C2
%r = xor i8 %sh, C1
=>
%not = xor i8 %x, -1
%r = lshr i8 %not, C2
https://bugs.llvm.org/show_bug.cgi?id=39657
llvm-svn: 347478
This transform needs to be limited.
We are converting to a constant pool load very early, and we
are turning loads that are independent of the select condition
(and therefore speculatable) into a dependent non-speculatable
load.
We may also be transferring a condition code from an FP register
to integer to create that dependent load.
llvm-svn: 347424
This is another step in vector narrowing - a follow-up to D53784
(and hoping to eventually squash potential regressions seen in
D51553).
The x86 test diffs are wins, but the AArch64 diff is probably not.
That problem already exists independent of this patch (see PR39722), but it
went unnoticed in the previous patch because there were no regression tests
that showed the possibility.
The x86 diff in i64-mem-copy.ll is close. Given the frequency throttling
concerns with using wider vector ops, an extra extract to reduce vector
width is the right trade-off at this level of codegen.
Differential Revision: https://reviews.llvm.org/D54392
llvm-svn: 347356
This uncovered an off-by-one typo in SimplifyDemandedVectorElts's INSERT_SUBVECTOR handling as its bounds check was bailing on safe indices.
llvm-svn: 347313
Consistently use (!LegalOperations || isOperationLegalOrCustom) for all node pairs.
Differential Revision: https://reviews.llvm.org/D53478
llvm-svn: 347255
Sadly, this duplicates (twice) the logic from InstSimplify. There
might be some way to at least share the DAG versions of the code,
but copying the folds seems to be the standard method to ensure
that we don't miss these folds.
Unlike in IR, we don't run DAGCombiner to fixpoint, so there's no
way to ensure that we do these kinds of simplifications unless the
code is repeated at node creation time and during combines.
There were other tests that would become worthless with this
improvement that I changed as pre-commits:
rL347161
rL347164
rL347165
rL347166
rL347167
I'm not sure how to salvage the remaining tests (diffs in this patch).
So the x86 tests verify that the new code is working as intended.
The AMDGPU test is actually similar to my motivating case: we have
some undef value that has survived to machine IR in an x86 test, and
then it gets folded in some weird way, or we crash if we don't transfer
the undef flag. But we would have been better off never getting to that
point by doing these simplifications.
This will lead back to PR32023 someday...
https://bugs.llvm.org/show_bug.cgi?id=32023
llvm-svn: 347170
It should be ok to create a new build_vector after legal operations so long as it doesn't cause an infinite loop in DAG combiner.
Unfortunately, X86's custom constant folding in combineVSZext is hiding any test changes from this. But I'm trying to get to a point where that X86 specific code isn't necessary at all.
Differential Revision: https://reviews.llvm.org/D54285
llvm-svn: 346728
Summary:
Handle extra output from index loads in cases where we wish to
forward a load value directly from a preceeding store.
Fixes PR39571.
Reviewers: peter.smith, rengolin
Subscribers: javed.absar, hiraditya, arphaman, llvm-commits
Differential Revision: https://reviews.llvm.org/D54265
llvm-svn: 346654
This is a long-awaited follow-up suggested in D33578. Since then, we've picked up even more
opportunities for vector narrowing from changes like D53784, so there are a lot of test diffs.
Apart from 2-3 strange cases, these are all wins.
I've structured this to be no-functional-change-intended for any target except for x86
because I couldn't tell if AArch64, ARM, and AMDGPU would improve or not. All of those
targets have existing regression tests (4, 4, 10 files respectively) that would be
affected. Also, Hexagon overrides the shouldReduceLoadWidth() hook, but doesn't show
any regression test diffs. The trade-off is deciding if an extra vector load is better
than a single wide load + extract_subvector.
For x86, this is almost always better (on paper at least) because we often can fold
loads into subsequent ops and not increase the official instruction count. There's also
some unknown -- but potentially large -- benefit from using narrower vector ops if wide
ops are implemented with multiple uops and/or frequency throttling is avoided.
Differential Revision: https://reviews.llvm.org/D54073
llvm-svn: 346595
It's possible for vector op legalization to generate a shuffle. If that happens we should give a chance for DAG combine to combine that with a build_vector input.
I also fixed a bug in combineShuffleOfScalars that was considering the number of uses on a undef input to a shuffle. We don't care how many times undef is used.
Differential Revision: https://reviews.llvm.org/D54283
llvm-svn: 346530
The DAGCombiner tries to SimplifySelectCC as follows:
select_cc(x, y, 16, 0, cc) -> shl(zext(set_cc(x, y, cc)), 4)
It can't cope with the situation of reordered operands:
select_cc(x, y, 0, 16, cc)
In that case we just need to swap the operands and invert the Condition Code:
select_cc(x, y, 16, 0, ~cc)
Differential Revision: https://reviews.llvm.org/D53236
llvm-svn: 346484
FindBetterNeighborChains simulateanously improves the chain
dependencies of a chain of related stores avoiding the generation of
extra token factors. For chains longer than the GatherAllAliasDepths,
stores further down in the chain will necessarily fail, a potentially
significant waste and preventing otherwise trivial parallelization.
This patch directly parallelize the chains of stores before improving
each store. This generally improves DAG-level parallelism.
Reviewers: courbet, spatel, RKSimon, bogner, efriedma, craig.topper, rnk
Subscribers: sdardis, javed.absar, hiraditya, jrtc27, atanasyan, llvm-commits
Differential Revision: https://reviews.llvm.org/D53552
llvm-svn: 346432
The original code avoided creating a zero vector after type legalization, but if we're after type legalization the type we have is legal. The real hazard we need to avoid is creating a build vector after op legalization. tryFoldToZero takes care of checking for this.
llvm-svn: 346119
These methods were just wrappers around getNode with additional asserts (identical and repeated 3 times). But getNode already has a switch that can be used to hold these asserts that allows them to be shared for all 3 opcodes. This also enables checking on the places that create these nodes without using the wrappers.
The rest of the patch is just changing all callers to use getNode directly.
llvm-svn: 346087
We already have custom lowering for the AVX case in LegalizeVectorOps. So its better to keep the regular extend op around as long as possible.
I had to qualify one place in DAG combine that created illegal vector extending load operations. This change by itself had no effect on any tests which is why its included here.
I've made a few cleanups to the custom lowering. The sign extend code no longer creates an identity shuffle with undef elements. The zero extend code now emits a zero_extend_vector_inreg instead of an unpckl with a zero vector.
For the high half of the custom lowering of zero_extend/any_extend, we're now using an unpckh with a zero vector or undef. Previously we used used a pshufd to move the upper 64-bits to the lower 64-bits and then used a zero_extend_vector_inreg. I think the zero vector should require less execution resources and be smaller code size.
Differential Revision: https://reviews.llvm.org/D54024
llvm-svn: 346043
reduceBuildVecConvertToConvertBuildVec vectorizes int2float in the DAGCombiner, which means that even if the LV/SLP has decided to keep scalar code using the cost models, this will override this.
While there are cases where vectorization is necessary in the DAG (mainly due to legalization artefacts), I don't think this is the case here, we should assume that the vectorizers know what they are doing.
Differential Revision: https://reviews.llvm.org/D53712
llvm-svn: 345964
I'm having trouble creating a test case for the ISD::TRUNCATE part of this that shows any codegen differences. But I was able to test the setcc path which is what the test changes here cover.
llvm-svn: 345908
The test causes a crash because we were trying to extract v4f32 to v3f32, and the
narrowing factor was then 4/3 = 1 producing a bogus narrow type.
This should fix:
https://bugs.llvm.org/show_bug.cgi?id=39511
llvm-svn: 345842
Summary:
Normalize the offset for endianess before checking
if the store cover the load in ForwardStoreValueToDirectLoad.
Without this we missed out on some optimizations for big
endian targets. If for example having a 4 bytes store followed
by a 1 byte load, loading the least significant byte from the
store, the STCoversLD check would fail (see @test4 in
test/CodeGen/AArch64/load-store-forwarding.ll).
This patch also fixes a problem seen in an out-of-tree target.
The target has i40 as a legal type, it is big endian,
and the StoreSize for i40 is 48 bits. So when normalizing
the offset for endianess we need to take the StoreSize into
account (assuming that padding added when storing into
a larger StoreSize always is added at the most significant
end).
Reviewers: niravd
Reviewed By: niravd
Subscribers: javed.absar, kristof.beyls, llvm-commits, uabelho
Differential Revision: https://reviews.llvm.org/D53776
llvm-svn: 345636
Narrowing vector binops came up in the demanded bits discussion in D52912.
I don't think we're going to be able to do this transform in IR as a canonicalization
because of the risk of creating unsupported widths for vector ops, but we already have
a DAG TLI hook to allow what I was hoping for: isExtractSubvectorCheap(). This is
currently enabled for x86, ARM, and AArch64 (although only x86 has existing regression
test diffs).
This is artificially limited to not look through bitcasts because there are so many
test diffs already, but that's marked with a TODO and is a small follow-up.
Differential Revision: https://reviews.llvm.org/D53784
llvm-svn: 345602
Summary:
Changes all uses of minnan/maxnan to minimum/maximum
globally. These names emphasize that the semantic difference between
these operations is more than just NaN-propagation.
Reviewers: arsenm, aheejin, dschuff, javed.absar
Subscribers: jholewinski, sdardis, wdng, sbc100, jgravelle-google, jrtc27, atanasyan, llvm-commits
Differential Revision: https://reviews.llvm.org/D53112
llvm-svn: 345218
Until now, we've only checked whether merging stores would cause a cycle via
the value argument, but the address and indexed offset arguments are also
capable of creating cycles in some situations.
The addresses are all base+offset with notionally the same base, but the base
SDNode may still be different (e.g. via an indexed load in one case, and an
ISD::ADD elsewhere). This allows cycles to creep in if one of these sources
depends on another.
The indexed offset is usually undef (representing a non-indexed store), but on
some architectures (e.g. 32-bit ARM-mode ARM) it can be an arbitrary value,
again allowing dependency cycles to creep in.
llvm-svn: 345200
When implementing memset's today we often see this pattern:
$x0 = MOV 0xXYXYXYXYXYXYXYXY
store $x0, ...
$w1 = MOV 0xXYXYXYXY
store $w1, ...
We first create a 64bit constant in a 64bit register with all bytes the
same and then create a 32bit constant with all bytes the same in a 32bit
register. In many targets we could just access the lower byte of the
64bit register instead.
- Ideally this would be handled by the ConstantHoist pass but it runs
too early when memset isn't expanded yet.
- The memset expansion code already had this optimization implemented,
however SelectionDAG constantfolding would constantfold the
"trunc(bigconstnat)" pattern to "smallconstant".
- This patch makes the memset expansion mark the constant as Opaque and
stop DAGCombiner from constant folding in this situation. (Similar to
how ConstantHoisting marks things as Opaque to avoid folding
ADD/SUB/etc.)
Differential Revision: https://reviews.llvm.org/D53181
llvm-svn: 345102
I've included a fix to DAGCombiner::ForwardStoreValueToDirectLoad that I believe will prevent the previous miscompile.
Original commit message:
Theoretically this was done to simplify the amount of isel patterns that were needed. But it also meant a substantial number of our isel patterns have to match an explicit bitcast. By making the vXi32/vXi16/vXi8 types legal for loads, DAG combiner should be able to change the load type to rem
I had to add some additional plain load instruction patterns and a few other special cases, but overall the isel table has reduced in size by ~12000 bytes. So it looks like this promotion was hurting us more than helping.
I still have one crash in vector-trunc.ll that I'm hoping @RKSimon can help with. It seems to relate to using getTargetConstantFromNode on a load that was shrunk due to an extract_subvector combine after the constant pool entry was created. So we end up decoding more mask elements than the lo
I'm hoping this patch will simplify the number of patterns needed to remove the and/or/xor promotion.
Reviewers: RKSimon, spatel
Reviewed By: RKSimon
Subscribers: llvm-commits, RKSimon
Differential Revision: https://reviews.llvm.org/D53306
llvm-svn: 344965
Introduce new versions that follow the IEEE semantics
to help with legalization that may need quieted inputs.
There are some regressions from inserting unnecessary
canonicalizes when these are matched from fast math
fcmp + select which should be fixed in a future commit.
llvm-svn: 344914
This is a late backend subset of the IR transform added with:
D52439
We can confirm that the conversion to a 'trunc' is correct by running:
$ opt -instcombine -data-layout="e"
(assuming the IR transforms are correct; change "e" to "E" for big-endian)
As discussed in PR39016:
https://bugs.llvm.org/show_bug.cgi?id=39016
...the pattern may emerge during legalization, so that's we are waiting for an
insertelement to become a scalar_to_vector in the pattern matching here.
The DAG allows for fun variations that are not possible in IR. Result types for
extracts and scalar_to_vector don't necessarily match input types, so that means
we have to be a bit more careful in the transform (see code comments).
The tests show that we don't handle cases that require a shift (as we did in the
IR version). I've left that as a potential follow-up because I'm not sure if
that's a real concern at this late stage.
Differential Revision: https://reviews.llvm.org/D53201
llvm-svn: 344872
I want to add another pattern here that includes scalar_to_vector,
so this makes that patch smaller. I was hoping to remove the
hasOneUse() check because it shouldn't be necessary for common
codegen, but an AMDGPU test has a comment suggesting that the
extra check makes things better on one of those targets.
llvm-svn: 344320
Summary:
Extend analysis forwarding loads from preceeding stores to work with
extended loads and truncated stores to the same address so long as the
load is fully subsumed by the store.
Hexagon's swp-epilog-phis.ll and swp-memrefs-epilog1.ll test are
deleted as they've no longer seem to be relevant.
Reviewers: RKSimon, rnk, kparzysz, javed.absar
Subscribers: sdardis, nemanjai, hiraditya, atanasyan, llvm-commits
Differential Revision: https://reviews.llvm.org/D49200
llvm-svn: 344142
We already do the following combines:
(bitcast int (and (bitcast fp X to int), 0x7fff...) to fp) -> fabs X
(bitcast int (xor (bitcast fp X to int), 0x8000...) to fp) -> fneg X
When the target has "bit preserving fp logic". This patch just extends it
to also combine:
(bitcast int (or (bitcast fp X to int), 0x8000...) to fp) -> fneg (fabs X)
As some targets have fnabs and even those that don't can efficiently lower
both the fabs and the fneg.
Differential revision: https://reviews.llvm.org/D44548
llvm-svn: 344093
This change is proposed as a part of D44548, but we
need this independently to avoid regressions from improved
undef propagation in SimplifyDemandedVectorElts().
llvm-svn: 343940
And use that to transform fsub with zero constant operands.
The integer part isn't used yet, but it is proposed for use in
D44548, so adding both enhancements here makes that
patch simpler.
llvm-svn: 343865
This fixes a case of bad index calculation when merging mismatching
vector types. This changes the existing code to just use the existing
extract_{subvector|element} and a bitcast (instead of bitcast first and
then newly created extract_xxx) so we don't need to adjust any indices
in the first place.
rdar://44584718
Differential Revision: https://reviews.llvm.org/D52681
llvm-svn: 343493
The SINT_TO_FP<->UINT_TO_FP combines for non-negative integers should only occur for legal ops once LegalOperations = true
No test case to hand, noticed when investigating PR38226 + PR38970
llvm-svn: 343405
DAGCombine will try to fold two loads that feed a SELECT or SELECT_CC
after the select, resulting in a select of an address and a single
load after.
If either of the loads depend on the other, this is not legal as it
could introduce cycles. However, it only checked this if the opcode
was a SELECT, and not for a SELECT_CC.
Unfortunately, the only reproducer I have for this is for our
downstream target. I've tried getting it to trigger on an upstream one
but haven't been successful.
Patch thanks to Bevin Hansson.
llvm-svn: 342980
This is a preliminary step towards solving PR14613:
https://bugs.llvm.org/show_bug.cgi?id=14613
If we have an 'add' instruction that sets flags, we can use that to eliminate an
explicit compare instruction or some other instruction (cmn) that sets flags for
use in the later select.
As shown in the unchanged tests that use 'icmp ugt %x, %a', we're effectively
reversing an IR icmp canonicalization that replaces a variable operand with a
constant:
https://rise4fun.com/Alive/V1Q
But we're not using 'uaddo' in those cases via DAG transforms. This happens in
CGP after D8889 without checking target lowering to see if the op is supported.
So AArch already shows 'uaddo' codegen for the i8/i16/i32/i64 test variants with
"using_cmp_sum" in the title. That's the pattern that CGP matches as an unsigned
saturated add and converts to uaddo without checking target capabilities.
This patch is gated by isOperationLegalOrCustom(ISD::UADDO, VT), so we see only
see AArch diffs for i32/i64 in the tests with "using_cmp_notval" in the title
(unlike x86 which sees improvements for all sizes because all sizes are 'custom').
But the AArch code (like x86) looks better when translated to 'uaddo' in all cases.
So someone that is involved with AArch may want to set i8/i16 to 'custom' for UADDO,
so this patch will fire on those tests.
Another possibility given the existing behavior: we could remove the legal-or-custom
check altogether because we're assuming that a UADDO sequence is canonical/optimal
before we ever reach here. But that seems like a bug to me. If the target doesn't
have an add-with-flags op, then it's not likely that we'll get optimal DAG combining
using a UADDO node. This is similar justification for why we don't canonicalize IR to
the overflow math intrinsic sibling (llvm.uadd.with.overflow) for UADDO in the first
place.
Differential Revision: https://reviews.llvm.org/D51929
llvm-svn: 342886
This code handled SCALAR_TO_VECTOR being returned by the recursion, but the code that used to return SCALAR_TO_VECTOR was removed in 2015.
llvm-svn: 342856
This comment was misleading about why we were restricting to before legalize types. The reason given would only apply to before legalize ops. But there is a before legalize types reason that should also be listed.
llvm-svn: 342851
This is an alternative to https://reviews.llvm.org/D37896. We can't decompose
multiplies generically without a target hook to tell us when it's profitable.
ARM and AArch64 may be able to remove some existing code that overlaps with
this transform.
This extends D52195 and may resolve PR34474:
https://bugs.llvm.org/show_bug.cgi?id=34474
(still an open question about transforming legal vector multiplies, but we
could open another bug report for those)
llvm-svn: 342844
x86 had 2 versions of peekThroughBitcast. DAGCombiner had 1. Plus, it had a 1-off implementation for the one-use variant.
Move the x86 versions of the code to SelectionDAG, so we don't have different copies of the code.
No functional change intended.
I'm putting this next to isBitwiseNot() because I am planning to use it in there. Another option is next to the
helpers in the ISD namespace (eg, ISD::isConstantSplatVector()). But if there's no good reason for those to be
there, I'd prefer to pull other helpers over to SelectionDAG in follow-up steps.
Differential Revision: https://reviews.llvm.org/D52285
llvm-svn: 342669
The test diff in not-and-simplify.ll is from a use in SimplifyDemandedBits,
and the test diff in add.ll is from a DAGCombiner transform.
llvm-svn: 342594
This is an alternative to D37896. I don't see a way to decompose multiplies
generically without a target hook to tell us when it's profitable.
ARM and AArch64 may be able to remove some duplicate code that overlaps with
this transform.
As a first step, we're only getting the most clear wins on the vector examples
requested in PR34474:
https://bugs.llvm.org/show_bug.cgi?id=34474
As noted in the code comment, it's likely that the x86 constraints are tighter
than necessary, but it may not always be a win to replace a pmullw/pmulld.
Differential Revision: https://reviews.llvm.org/D52195
llvm-svn: 342554
This is a follow-up suggested in D51630 and originally proposed as an IR transform in D49040.
Copying the motivational statement by @evandro from that patch:
"This transformation helps some benchmarks in SPEC CPU2000 and CPU2006, such as 188.ammp,
447.dealII, 453.povray, and especially 300.twolf, as well as some proprietary benchmarks.
Otherwise, no regressions on x86-64 or A64."
I'm proposing to add only the minimum support for a DAG node here. Since we don't have an
LLVM IR intrinsic for cbrt, and there are no other DAG ways to create a FCBRT node yet, I
don't think we need to worry about DAG builder, legalization, a strict variant, etc. We
should be able to expand as needed when adding more functionality/transforms. For reference,
these are transform suggestions currently listed in SimplifyLibCalls.cpp:
// * cbrt(expN(X)) -> expN(x/3)
// * cbrt(sqrt(x)) -> pow(x,1/6)
// * cbrt(cbrt(x)) -> pow(x,1/9)
Also, given that we bail out on long double for now, there should not be any logical
differences between platforms (unless there's some platform out there that has pow()
but not cbrt()).
Differential Revision: https://reviews.llvm.org/D51753
llvm-svn: 342348
Add support for bitcasts from float type to an integer type of the same element bitwidth.
There maybe cases where we need to support different widths (e.g. as SSE __m128i is treated as v2i64) - but I haven't seen cases of this in the wild yet.
llvm-svn: 341652
This was proposed as an IR transform in D49306, but it was not clearly justifiable as a canonicalization.
Here, we only do the transform when the target tells us that sqrt can be lowered with inline code.
This is the basic case. Some potential enhancements are in the TODO comments:
1. Generalize the transform for other exponents (allow more than 2 sqrt calcs if that's really cheaper).
2. If we have less fast-math-flags, generate code to avoid -0.0 and/or INF.
3. Allow the transform when optimizing/minimizing size (might require a target hook to get that right).
Note that by default, x86 converts single-precision sqrt calcs into sqrt reciprocal estimate with
refinement. That codegen is controlled by CPU attributes and can be manually overridden. We have plenty
of test coverage for that already, so I didn't bother to include extra testing for that here. AArch uses
its full-precision ops in all cases (not sure if that's the intended behavior or not, but that should
also be covered by existing tests).
Differential Revision: https://reviews.llvm.org/D51630
llvm-svn: 341481
Summary:
I'm not sure if this patch is correct or if it needs more qualifying somehow. Bitcast shouldn't change the size of the load so it should be ok? We already do something similar for stores. We'll change the type of a volatile store if the resulting store is Legal or Custom. I'm not sure we should be allowing Custom there...
I was playing around with converting X86 atomic loads/stores(except seq_cst) into regular volatile loads and stores during lowering. This would allow some special RMW isel patterns in X86InstrCompiler.td to be removed. But there's some floating point patterns in there that didn't work because we don't fold (f64 (bitconvert (i64 volatile load))) or (f32 (bitconvert (i32 volatile load))).
Reviewers: efriedma, atanasyan, arsenm
Reviewed By: efriedma
Subscribers: jvesely, arsenm, sdardis, kzhuravl, wdng, yaxunl, dstuttard, tpr, t-tye, arichardson, jrtc27, atanasyan, jfb, llvm-commits
Differential Revision: https://reviews.llvm.org/D50491
llvm-svn: 340797
I noticed this along with the patterns in D51125, but when the index is variable,
we don't convert insertelement into a build_vector.
For x86, that means these get expanded at legalization time into the loading/spilling
code that we see in the tests. I think it's always better to avoid going to memory on
these, and we get the optimal 'broadcast' if it's available.
I suspect other targets may want to look at enabling the hook. AArch64 and AMDGPU have
regression tests that would be affected (although I did not check what would happen in
those cases). In the most basic cases shown here, AArch64 would probably do much
better with a splat.
Differential Revision: https://reviews.llvm.org/D51186
llvm-svn: 340705
Previously we allowed the store to be Custom. But without knowing for sure that the Custom handling won't split the store, we shouldn't convert a volatile store. We also probably shouldn't be creating a store the requires custom handling after LegalizeOps. This could lead to an infinite loop if the custom handling was to insert a bitcast. Though I guess isStoreBitCastBeneficial could be used to block such a loop.
The test changes here are due to the volatile part of this. The stores in the test are all volatile and i32 stores are marked custom, So we are no longer converting them
This is related to D50491 where I was trying to allow some bitcasting of volatile loads
Differential Revision: https://reviews.llvm.org/D50578
llvm-svn: 340626
During combining, ReduceLoadWdith is used to combine AND nodes that
mask loads into narrow loads. This patch allows the mask to be a
shifted constant. This results in a narrow load which is then left
shifted to compensate for the new offset.
Differential Revision: https://reviews.llvm.org/D50432
llvm-svn: 340261
Summary:
I believe this restores the behavior we had before r339147.
Fixes PR38622.
Reviewers: RKSimon, chandlerc, spatel
Reviewed By: chandlerc
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D50936
llvm-svn: 340120
Add support for cases where only some c1+c2 results exceed the max bitshift, clamping accordingly.
Differential Revision: https://reviews.llvm.org/D35722
llvm-svn: 340010
When nodes are reassociated the vector-reduction flag gets lost.
The test case is here is what would happen if you had a sum of absolute differences loop that started with a non-zero but contant sum and that loop was unrolled. The vectorizer will generate a constant vector for the initial value. And DAGCombiner reassociate tries to move it down the addition tree erasing the vector-reduction flag. Interestingly this moves constants the opposite direction of the reassociate IR pass.
I've chosen to just punt on the reassociate, but I suppose we could maybe preserve the flag if both nodes have it set.
Differential Revision: https://reviews.llvm.org/D50827
llvm-svn: 339946
Intentionally excluding nodes from the DAGCombine worklist is likely to
lead to weird optimizations and infinite loops, so it's generally a bad
idea.
To avoid the infinite loops, fix DAGCombine to use the
isDesirableToCommuteWithShift target hook before performing the
transforms in question, and implement the target hook in the ARM backend
disable the transforms in question.
Fixes https://bugs.llvm.org/show_bug.cgi?id=38530 . (I don't have a
reduced testcase for that bug. But we should have sufficient test
coverage for PerformSHLSimplify given that we're not playing weird
tricks with the worklist. I can try to bugpoint it if necessary,
though.)
Differential Revision: https://reviews.llvm.org/D50667
llvm-svn: 339734
Summary: This change provides a common optimization path for both Unsafe and FMF driven optimization for this fsub fold adding reassociation, as it the flag that most closely represents the translation
Reviewers: spatel, wristow, arsenm
Reviewed By: spatel
Subscribers: wdng
Differential Revision: https://reviews.llvm.org/D50195
llvm-svn: 339357
isNegatibleForFree() should not matter here (as the test diffs show)
because it's always a win to replace an fsub+fadd with fneg. The
problem in D50195 persists because either (1) we are doing these
folds in the wrong order or (2) we're missing another fold for fadd.
llvm-svn: 339299
I don't know if it's possible to expose this diff in a test,
but we should always try simplifications (no new nodes created)
before more complicated transforms for efficiency (similar to
what we do in IR).
llvm-svn: 339298
The isConstOrConstSplat result is only used in a ISD::matchUnaryPredicate call which can perform the equivalent iteration just as quickly.
llvm-svn: 339262
Scatter could have multiple identical indices. We need to maintain sequential order. We get this right in LegalizeVectorTypes, but not in this code.
Differential Revision: https://reviews.llvm.org/D50374
llvm-svn: 339157
This was missed in D50185.
NFC until we add actual non-uniform support to BuildSDIV (similar BuildUDIV support in D49248) - for now it just early outs.
llvm-svn: 339147
This patch refactors the existing TargetLowering::BuildUDIV base implementation to support non-uniform constant vector denominators.
It also includes a fold for MULHU by pow2 constants to SRL which can now more readily occur from BuildUDIV.
Differential Revision: https://reviews.llvm.org/D49248
llvm-svn: 339121
Src0 doesn't really convey any meaning to what the operand is. Passthru matches what's used in the documentation for the intrinsic this comes from.
llvm-svn: 339101
First step towards a BuildSDIV equivalent to D49248 for non-uniform vector support - this just pushes the splat detection down into TargetLowering::BuildSDIV where its still used.
Differential Revision: https://reviews.llvm.org/D50185
llvm-svn: 338838
The vector contains the SDNodes that these functions create. The number of nodes is always a small number so we should use SmallVector to avoid a heap allocation.
llvm-svn: 338329
This is exchanging a sub-of-1 with add-of-minus-1:
https://rise4fun.com/Alive/plKAH
This is another step towards improving select-of-constants codegen (see D48970).
x86 is the motivating target, and those diffs all appear to be wins. PPC and AArch64 look neutral.
I've limited this to early combining (!LegalOperations) in case a target wants to reverse it, but
I think canonicalizing to 'add' is more likely to produce further transforms because we have more
folds for 'add'.
Differential Revision: https://reviews.llvm.org/D49924
llvm-svn: 338317
Thinking about it more it might be possible for the later nodes to be folded in getNode in such a way that the other created nodes are left dead. This can cause use counts to be incorrect on nodes that aren't dead.
So its probably safer to leave this alone.
llvm-svn: 338298
Amaury Sechet