After merging main part of the gather/buildvector code, CreateShuffle
lambda can removed and ShuffleBuilder add functions can be used instead.
Also, part of the code from CreateShuffle migrated to createShuffle of
the BaseShuffleAnalysis::createShuffle function for better code emission.
Differential Revision: https://reviews.llvm.org/D145988
operation for combined entries.
The vector factor after combining of the shuffle entries is defined by
the size of the mask, not by the vector factors of the original
entries. So, need to adjust it to emit correct code.
This slightly increases the costs of InsertElement instructions that are part
of a vector splat sequence, i.e. a load, InsertElement and a shuffle (load +
dup). The resulting LD1R is a high latency instruction, and this slight
increase in costs avoids SLP vectorisation for a couple of cases where this
isn't profitable.
Fixes: https://github.com/llvm/llvm-project/issues/61047
Differential Revision: https://reviews.llvm.org/D145578
Previously only the very first gather/buildvector node might be probed for reshuffling of other nodes.
But the compiler may do the same for other gather/buildvector nodes too, just need to check the
dependency and postpone the emission of the dependent nodes, if the origin nodes were not emitted yet.
Part of D110978
Differential Revision: https://reviews.llvm.org/D144958
This caused verifier errors:
Instruction does not dominate all uses!
%8 = insertelement <2 x i64> %7, i64 %pgocount1330, i64 1
%15 = shufflevector <2 x i64> %8, <2 x i64> poison, <2 x i32> <i32 1, i32 1>
in function ?NearestInclusiveAncestorAssignedToSlot@SlotScopedTraversal@blink@@SAPAVElement@2@ABV32@@Z
(or register allocator crash when the verifier was disabled).
See comment on the code review.
> Previously only the very first gather/buildvector node might be probed for reshuffling of other nodes.
> But the compiler may do the same for other gather/buildvector nodes too, just need to check the
> dependency and postpone the emission of the dependent nodes, if the origin nodes were not emitted yet.
>
> Part of D110978
>
> Differential Revision: https://reviews.llvm.org/D144958
This reverts commit a611b3f3059e4c3b9e7b914091c3edaef099fd5d.
It also reverts 7a4061ae372b3262703ffeea3b64db89187db611 which depended on the above.
RISCV has "vfabs.v" and "vfsqrt.v" so math functions abs and sqrt
can be SLP vectorized. But others exp/log/sin/asin/sinh/asinh/...
can not.
Reviewed By: reames
Differential Revision: https://reviews.llvm.org/D145562
Previously only the very first gather/buildvector node might be probed for reshuffling of other nodes.
But the compiler may do the same for other gather/buildvector nodes too, just need to check the
dependency and postpone the emission of the dependent nodes, if the origin nodes were not emitted yet.
Part of D110978
Differential Revision: https://reviews.llvm.org/D144958
IRBuilder in many cases is able to fold constant code automatically,
but in some cases (for some intrinsics) it cannot do it. Need to perform
manual calculation, if constant provided in these corner cases, to avoid
infinite loop.
When gathering the counter for the reused scalars, need to use reduced
value, not the original reduced value. Same values counter is gathered
for reduced values, not original ones.
of scalars."' failed.
Need to check for the reused indices when checking if 2 insertelement
instruction are from the same buildvector. If the inidices are reused,
better not to match buildvectors and consider them as differenet,
otherwise need to track the order of insertelement operations.
If have just one non-undef scalar in the buildvector/gather node, we try
to put it to be the very first element, which is profitable in most
cases. Do the preliminary estimation, if this more profitable during
graph rotation and do same for all elements, including extractelements.
Differential Revision: https://reviews.llvm.org/D144689
The target hook prefersVectorizedAddressing() already exists to check with
target if address computations should be vectorized, so it seems like this
should be used in SLPVectorizer as well.
Reviewed By: ABataev, RKSimon
Differential Revision: https://reviews.llvm.org/D144128
No need to swap extractelements, which were not excluded from the list
during cost analysis. It leads to incorrect cost calculation and make
vector code more profitable than it is actually is.
If the scalar must be extracted and then used in the gather node,
instead we can emit shuffle instruction to avoid those extra
extractelements and vector-to-scalar and back data movement.
Part of D110978
Differential Revision: https://reviews.llvm.org/D141940
This fixes an infinite loop if isa<T>(II->getOperand(1)) is true.
Update Base at the top of the loop, before the continue.
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D144292
Metric: size..text
size..text results results0 diff
SingleSource/Regression/C/gcc-c-torture/execute/GCC-C-execute-980605-1.test 445.00 461.00 3.6%
SingleSource/Benchmarks/Adobe-C++/loop_unroll.test 428477.00 428445.00 -0.0%
External/SPEC/CFP2006/447.dealII/447.dealII.test 618849.00 618785.00 -0.0%
For all tests some extra code was optimized, GCC-C-execute has some more
inlining after
Differential Revision: https://reviews.llvm.org/D132261
The compiler may produce better results if it does not look for
constants, uses an extra analysis of phi nodes, looks through all tree
nodes without skipping the cases, where the very first set of nodes is
empty. Also, it tries to reshufle the nodes if it is profitable for
sure, i.e. at least 2 scalars are used for single node permutation and at
least 3 scalars are used for the permutation of 2 nodes.
Part of D110978
Differential Revision: https://reviews.llvm.org/D141512
When rooted with PHIs, a vectorization tree may have another node with PHIs
which have roots as their operands. We cannot ignore ordering information
for root in such a case.
Differential Revision: https://reviews.llvm.org/D141309
The new mask represents the order, not the mask itself. At first, need
to treat as the order, convert to mask and only after that reorder
gathered scalars to build correct clustered order.
Differential Revision: https://reviews.llvm.org/D141161
Need to include the cost of the initial insertelement to the cost of the
broadcasts. Also, need to adjust the cost of the gather/buildvector if
the element is inserted into poison/undef vector.
Differential Revision: https://reviews.llvm.org/D140498
Make a separate routine for GEPs cost calculation and make
the approach uniform across load, store and GEP tree nodes.
Additional issue fixed is GEP cost savings were applied twice
for ScatterVectorize nodes (aka gather load) making them look
unrealistically profitable for vectorization.
Differential Revision: https://reviews.llvm.org/D140789
analysis.
Missed the analysis of the shuffle mask when trying to analyze the
operands of the shuffle instruction during peeking through shuffle
instructions.
We incorrectly assume intrinsic as a function call and it prevents us from
the opportunity to vectorize. On Aarch64 Cortex-A53 we think that
llvm.fmuladd.f64 is a function call which is wrong.
Differential Revision: https://reviews.llvm.org/D140392
SLP produces two gather loads (one feeds another).
For the first set of scalar loads GEP indices are all constant.
The result of the second load is then fed into reduction (as a seed).
Differential Revision: https://reviews.llvm.org/D140785
We do not need to emit many extractelements for each particular use, we
can reuse the only one, just need to adjust it to make it dominate on
all uses.
Differential Revision: https://reviews.llvm.org/D140580
We can use ShuffleInstructionBuilder now for shrinking shuffle emission.
It allows to remove extra shuffle from the emitted code and reuse
original vector.
Part of D110978
Differential Revision: https://reviews.llvm.org/D140499
of extractvalues.
No need to get the last instruction only for vectorized extractvalues,
for gathered(buildvector sequence) still need to get the insertion
point.
