This is analogous to what LLVM's PatternMatch.h supports,
but LLVM calls it m_Value for both the binding and
nonbinding versions.
This is an upstream from CIRCT and is used there.
Differential Revision: https://reviews.llvm.org/D113905
This patch changes it to ignorelist and contains a filename change for the
.txt file that's called.
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D113689
The first try at this patch ( bf5748a1af ) was reverted ( 5be64d4164 )
because it could crash. The cause of that problem was failing to account
for the optional peek-through-bitcast in the enclosing function.
This version of the patch adds a clause to avoid the fold in case of
bitcasts because it is unlikely to be profitable in that scenario.
A test case based on https://llvm.org/PR52504 was added to make sure
we don't have that problem again.
Original commit message:
and (pcmpgt X, -1), Y --> pandn (vsrai X, BitWidth-1), Y
This avoids the -1 constant vector in favor of an arithmetic shift
instruction if it exists (the ISA is still not complete after all
these years...).
We catch this pattern late in combining by matching PCMPGT, so it
should not interfere with more general folds.
Differential Revision: https://reviews.llvm.org/D113603
Currently `X86TTIImpl::getInterleavedMemoryOpCostAVX512()` asks about i8 elt type,
so this change does affect vectorization. In the end, it will ask about i1.
We should also try to promote to i16 if we have AVX512BW, i'll do that in a follow-up.
All costs here look good, i've added the missing truncation costs in preparatory patches.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D113853
Some of the costs get larger here,
but i suppose that makes sense since we'd previously query
scalarization costs that may not be really representative of the reality.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D113852
While this one is trivial and identical to the previous patch,
there is a weird cost change in a follow-up patch that i'm not sure about.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D113851
While this one is trivial and identical to the previous patch,
there is a weird cost change in a follow-up patch that i'm not sure about.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D113850
Add a fir.llvm_ptr type to allow any level of indirections
Currently, fir pointer types (fir.ref, fir.ptr, and fir.heap) carry
a special Fortran semantics, and cannot be freely combined/nested.
When implementing some features, lowering sometimes needs more liberty
regarding the number of indirection levels. Add a fir.llvm_ptr that has
no constraints.
Allow its usage in fir.coordinate_op, fir.load, and fir.store.
Convert the FIR LLVMPointer to an LLVMPointer in the LLVM dialect.
Reviewed By: clementval
Differential Revision: https://reviews.llvm.org/D113755
Co-authored-by: Jean Perier <jperier@nvidia.com>
If an operand is a bitcasted or widended constant, try to more aggressively create broadcastable constants for folding, which in particular helps non-VLX modes.
I've refactored getAVX512Node so that VLX targets can make better use of this as well.
NOTE: In the future, I think we should consider removing the broadcast of constant data from DAG entirely and move this to either X86InstrInfo::foldMemoryOperand or a new pass - AVX1/2 targets has similar problems with missed (whole vector) folds that need to be improved as well.
Differential Revision: https://reviews.llvm.org/D113845
A bunch of scalars can be treated as a splat not only if all elements
are the same but also if some of them are undefvalues.
Differential Revision: https://reviews.llvm.org/D113774
Required by https://stackoverflow.com/questions/58073606
As the output argument is stripped out in the clang-check tool, it seems impossible for clang-check users to customize the output file name, even with -extra-args and -extra-arg-before.
This patch adds the -analyzer-output-path argument to allow users to adjust the output name. And if the argument is not set or the analyzer is not enabled, the original strip output adjuster will remove the output arguments.
Differential Revision: https://reviews.llvm.org/D97265
As for D113662, this patch just add a place holder for
the fir.global_len operation conversion. This operation
is part of F20xx and is not implemented yet.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: awarzynski
Differential Revision: https://reviews.llvm.org/D113887
This patch extends the `FIRToLLVMLowering` pass in Flang by adding a
hook to transform `fir.unboxchar` to a sequence of LLVM MLIR
instructions.
This is part of the upstreaming effort from the `fir-dev` branch in [1].
[1] https://github.com/flang-compiler/f18-llvm-project
Differential Revision: https://reviews.llvm.org/D113747
Originally written by:
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Co-authored-by: Jean Perier <jperier@nvidia.com>
This reverts commit c35e8185d8.
mstorsjo reported in the revision thread that one VNCoercion assertion
is violated and seemly in relate to this commit. As per "If a test case
that demonstrates a problem is reported in the commit thread, please
revert and investigate offline", this commit is reverted.
If the vector intrinsic has scalar argument, we currently still create
a tree entry for this argument. This entry is not used, just consumes
resources and increases the cost of the tree.
Differential Revision: https://reviews.llvm.org/D113806
When cross-compiling LLVM in an environment where there //is// an
objdump binary available but it does not understand the target
platform's object file format, extract_symbols.py fails, because its
initial check for tool availability decides that the existence of
objdump at all is good enough to settle on it as the tool of choice.
In such an environment it's useful to work around this by telling
extract_symbols.py to use llvm-readobj instead. The script itself has
an option for that, but its invocation in AddLLVM.cmake wasn't
providing a mechanism to add extra options passed through for the
cmake command line.
Reviewed By: DavidSpickett
Differential Revision: https://reviews.llvm.org/D113557
The value type can be a typedef of a reference (e.g. `typedef int& myint`).
In this case `GetQualType(type)` will return `clang::Typedef`, which cannot
be casted to `clang::ReferenceType`.
Fix a regression introduced in https://reviews.llvm.org/D103532.
Reviewed By: teemperor
Differential Revision: https://reviews.llvm.org/D113673
As for D113662, this patch just add a place holder for
the `fir.select_type` operation conversion. This operation
is part of F20xx and is not implemented yet.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: kiranchandramohan
Differential Revision: https://reviews.llvm.org/D113878
At the moment, computeRecurrenceType does not include any sign bits in
the maximum bit width. If the value can be negative, this means the sign
bit will be missing and the sext won't properly extend the value.
If the value can be negative, increment the bitwidth by one to make sure
there is at least one sign bit in the result value.
Note that the increment is also needed *if* the value is *known* to be
negative, as a sign bit needs to be preserved for the sext to work.
Note that this at the moment prevents vectorization, because the
analysis computes i1 as type for the recurrence when looking through the
AND in lookThroughAnd.
Fixes PR51794, PR52485.
Reviewed By: spatel
Differential Revision: https://reviews.llvm.org/D113056
The template std::is_assignable<T, U> checks that T is assignable from
U. Hence, the order of operands in the instantiation of
std::is_assignable in the std::reverse_iterator::operator= condition
should be reversed.
This issue remained unnoticed because std::reverse_iterator has an
implicit conversion constructor. This patch adds a test to check that
the assignment operator is used directly, without any implicit
conversions. The patch also adds a similar test for
std::move_iterator.
Reviewed By: Quuxplusone, ldionne, #libc
Differential Revision: https://reviews.llvm.org/D113417
The file seems to have been put in the wrong place in its original
commit. This had the effect of marking all llvm-nm tests as unsupported,
unless X86 was enabled, even for tests that weren't X86 specific.
Fixes https://bugs.llvm.org/show_bug.cgi?id=52506.
Reviewed by: mstorsjo
Differential Revision: https://reviews.llvm.org/D113882
When trying to connect the vectorization of depthwise convolutions to e2e execution
a number of problems surfaced.
Fix an off-by-one error on the size of the input vector (similary to what was previously done for regular conv).
Rewrite the lowering to vector.fma instead of vector.contract: the KW reduction dimension has already been unrolled and vector.contract requires a reduction dimension to be valid.
Differential Revision: https://reviews.llvm.org/D113884
FMAOp -> LLVM conversion is done progressively by peeling off 1 dimension from FMAOp at each pattern iteration. Add the recursively bounded property declaration to the pattern so that the rewriter can apply it multiple times.
Without this, FMAOps with 3+D do not lower to LLVM.
Differential Revision: https://reviews.llvm.org/D113886
modernize-loop-convert checks and fixes when a loop that iterates over the
elements of a container can be rewritten from a for(...; ...; ...) style into
the "new" C++11 for-range format. For that, it needs to parse the elements of
that loop, like its init-statement, such as ItType it = cont.begin().
modernize-loop-convert checks whether the loop variable is initialized by a
begin() member function.
When an iterator is initialized with a conversion operator (e.g. for
(const_iterator it = non_const_container.begin(); ...), attempts to retrieve the
name of the initializer expression resulted in an assert, as conversion
operators don't have a valid IdentifierInfo.
I fixed this by making digThroughConstructors dig through conversion operators
as well.
Differential Revision: https://reviews.llvm.org/D113201
OperationLegalizer::isIllegal returns false if operation legality wasn't
registered by user and we expect same behaviour when dynamic legality
callback return None, but instead true was returned.
Differential Revision: https://reviews.llvm.org/D113267
This casued assertion failures:
llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp:9446:
void llvm::SelectionDAG::ReplaceAllUsesWith(llvm::SDNode *, llvm::SDNode *):
Assertion `(!From->hasAnyUseOfValue(i) || From->getValueType(i) == To->getValueType(i))
&& "Cannot use this version of ReplaceAllUsesWith!"' failed.
See comment on the code review.
(Had to update some expectations in test/CodeGen/X86/vselect-zero.ll
manually due to other changes having landed after the reverted one.)
> and (pcmpgt X, -1), Y --> pandn (vsrai X, BitWidth-1), Y
>
> This avoids the -1 constant vector in favor of an arithmetic shift
> instruction if it exists (the ISA is still not complete after all
> these years...).
>
> We catch this pattern late in combining by matching PCMPGT, so it
> should not interfere with more general folds.
>
> Differential Revision: https://reviews.llvm.org/D113603
This reverts commit bf5748a1af.
This patch extends the `FIRToLLVMLowering` pass in Flang by adding a
hook to transform `fir.emboxchar` to a sequence of LLVM MLIR
instructions.
This is part of the upstreaming effort from the `fir-dev` branch in [1].
[1] https://github.com/flang-compiler/f18-llvm-project
Differential Revision: https://reviews.llvm.org/D113666
Patch originally written by:
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
This change makes it possible to set up custom mappings in a PostAnalysisStep. Some users of Comprehensive Bufferize have custom tensor types and it is most convenient to just reuse the same bvm.
Also add some more assertions.
Differential Revision: https://reviews.llvm.org/D113726
MachineVerifier verified the subranges of a live interval if
they existed, but did not complain if they did not exist.
This patch changes the verifier to complain if there are no
subranges in the live interval for a subreg operand (so long
as MachineRegisterInfo says we should be tracking subreg
liveness for that register). This matches the conditions for
LiveIntervalCalc to create subranges in the first place.
Differential Revision: https://reviews.llvm.org/D112556
D112976 changed the layout and 0d62e31c45 andjusted the test
expectations to match.
This patch changes the tests to expect both versions, so that one can
run the test suite against older libc++ versions as well.
Simon Pilgrim