Extend OpDSL with a `defines` method that can set the `hasCanonicalizer` flag for an OpDSL operation. If the flag is set via `defines(Canonicalizer)` the operation needs to implement the `getCanonicalizationPatterns` method. The revision specifies the flag for linalg.fill_tensor and adds an empty `FillTensorOp::getCanonicalizationPatterns` implementation.
This revision is a preparation step to replace linalg.fill by its OpDSL counterpart linalg.fill_tensor. The two are only functionally equivalent if both specify the same canonicalization patterns. The revision is thus a prerequisite for the linalg.fill replacement.
Depends On D120725
Reviewed By: nicolasvasilache
Differential Revision: https://reviews.llvm.org/D120726
Add a FillOpInterface similar to the contraction and convolution op interfaces. The FillOpInterface is a preparation step to replace linalg.fill by its OpDSL version linalg.fill_tensor. The interface implements the `value()`, `output()`, and `result()` methods that by default are not available on linalg.fill_tensor.
Reviewed By: nicolasvasilache
Differential Revision: https://reviews.llvm.org/D120725
Currently, the transfer mask is materialized by generating the vector
comparison: [offset + 0, .., offset + length - 1] < [dim, .., dim]
A better alternative is to materialize the transfer mask by using the
operation: `vector.create_mask (dim - offset)`, which will generate
simpler code and compose better with scalable vectors.
Differential Revision: https://reviews.llvm.org/D120487
This is to align with the PyTACO API better.
Modify an existing unit test to test the new routines.
Reviewed By: aartbik
Differential Revision: https://reviews.llvm.org/D121083
In quantized comutation, there are casting ops around computation ops.
Reorder the ops to make reduce-to-contract actually work.
Reviewed By: ThomasRaoux
Differential Revision: https://reviews.llvm.org/D120760
The current StandardToLLVM conversion patterns only really handle
the Func dialect. The pass itself adds patterns for Arithmetic/CFToLLVM, but
those should be/will be split out in a followup. This commit focuses solely
on being an NFC rename.
Aside from the directory change, the pattern and pass creation API have been renamed:
* populateStdToLLVMFuncOpConversionPattern -> populateFuncToLLVMFuncOpConversionPattern
* populateStdToLLVMConversionPatterns -> populateFuncToLLVMConversionPatterns
* createLowerToLLVMPass -> createConvertFuncToLLVMPass
Differential Revision: https://reviews.llvm.org/D120778
These unit tests resides in an internal repository. Porting the tests to the
public repository.
Reviewed By: aartbik
Differential Revision: https://reviews.llvm.org/D121021
The default lowering of vector transpose operations generates a large sequence of
scalar extract/insert operations, one pair for each scalar element in the input tensor.
In other words, the vector transpose is scalarized. However, there are transpose
patterns where one or more adjacent high-order dimensions are not transposed (for
example, in the transpose pattern [1, 0, 2, 3], dimensions 2 and 3 are not transposed).
This patch improves the lowering of those cases by not scalarizing them and extracting/
inserting a full n-D vector, where 'n' is the number of adjacent high-order dimensions
not being transposed. By doing so, we prevent the scalarization of the code and generate a
more performant vector version.
Paradoxically, this patch shouldn't improve the performance of transpose operations if
we are using LLVM. The LLVM pipeline is able to optimize away some of the extract/insert
operations and the SLP vectorizer is converting the scalar operations back to its vector
form. However, scalarizing a vector version of the code in MLIR and relying on the SLP
vectorizer to reconstruct the vector code again is highly undesirable for several reasons.
Reviewed By: nicolasvasilache, ThomasRaoux
Differential Revision: https://reviews.llvm.org/D120601
This patch fixes the crash when printing some ops (like affine.for and
scf.for) when they are dumped in invalid state, e.g. during pattern
application. Now the AsmState constructor verifies the operation
first and switches to generic operation printing when the verification
fails. Also operations are now printed in generic form when emitting
diagnostics and the severity level is Error.
Reviewed By: rriddle, mehdi_amini
Differential Revision: https://reviews.llvm.org/D117834
The OpenMPIRBuilder has a bug. Specifically, suppose you have two nested openmp parallel regions (writing with MLIR for ease)
```
omp.parallel {
%a = ...
omp.parallel {
use(%a)
}
}
```
As OpenMP only permits pointer-like inputs, the builder will wrap all of the inputs into a stack allocation, and then pass this
allocation to the inner parallel. For example, we would want to get something like the following:
```
omp.parallel {
%a = ...
%tmp = alloc
store %tmp[] = %a
kmpc_fork(outlined, %tmp)
}
```
However, in practice, this is not what currently occurs in the context of nested parallel regions. Specifically to the OpenMPIRBuilder,
the entirety of the function (at the LLVM level) is currently inlined with blocks marking the corresponding start and end of each
region.
```
entry:
...
parallel1:
%a = ...
...
parallel2:
use(%a)
...
endparallel2:
...
endparallel1:
...
```
When the allocation is inserted, it presently inserted into the parent of the entire function (e.g. entry) rather than the parent
allocation scope to the function being outlined. If we were outlining parallel2, the corresponding alloca location would be parallel1.
This causes a variety of bugs, including https://github.com/llvm/llvm-project/issues/54165 as one example.
This PR allows the stack allocation to be created at the correct allocation block, and thus remedies such issues.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D121061
This commit adds a new hook Pass `bool canScheduleOn(RegisteredOperationName)` that
indicates if the given pass can be scheduled on operations of the given type. This makes it
easier to define constraints on generic passes without a) adding conditional checks to
the beginning of the `runOnOperation`, or b) defining a new pass type that forwards
from `runOnOperation` (after checking the invariants) to a new hook. This new hook is
used to implement an `InterfacePass` pass class, that represents a generic pass that
runs on operations of the given interface type.
The PassManager will also verify that passes added to a pass manager can actually be
scheduled on that pass manager, meaning that we will properly error when an Interface
is scheduled on an operation that doesn't actually implement that interface.
Differential Revision: https://reviews.llvm.org/D120791
RegionBranchOpInterface and BranchOpInterface are allowed to make implicit type conversions along control-flow edges. In effect, this adds an interface method, `areTypesCompatible`, to both interfaces, which should return whether the types of corresponding successor operands and block arguments are compatible. Users of the interfaces, here on forth, must be aware that types may mismatch, although current users (in MLIR core), are not affected by this change. By default, type equality is used.
`async.execute` already has unequal types along control-flow edges (`!async.value<f32>` vs. `f32`), but it opted out of calling `RegionBranchOpInterface::verifyTypes` in its verifier. That method has now been removed and `RegionBranchOpInterface` will verify types along control edges by default in its verifier.
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D120790
This patch extends the existing if combining canonicalization to also handle the case where a value returned by the first if is used within the body of the second if.
This patch also extends if combining to support if's whose conditions are logical negations of each other.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D120924
We can simplify an extractvalue of an insertvalue to extract out of the base of the insertvalue, if the insert and extract are at distinct and non-prefix'd indices
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D120915
Extend isLoopMemoryParallel check to include locally allocated memrefs.
This strengthens and also speeds up the dependence check used by the
utility by excluding locally allocated memrefs where appropriate.
Additional memref dialect ops can be supported exhaustively via proper
interfaces.
Reviewed By: dcaballe
Differential Revision: https://reviews.llvm.org/D120617
This commit adds support for processing tablegen include files, and importing
various information from ODS. This includes operations, attribute+type constraints,
attribute/operation/type interfaces, etc. This will allow for much more robust tooling,
and also allows for referencing ODS constructs directly within PDLL (imported interfaces
can be used as constraints, operation result names can be used for member access, etc).
Differential Revision: https://reviews.llvm.org/D119900
This commit deletes the old dialect conversion-based bufferization patterns, which are now obsolete.
Differential Revision: https://reviews.llvm.org/D120883
This reverts commit dbe9f0914fcfd8444fd9656821af0f1a34a27e7a.
The flang-x86_64-windows buildbot has been failing since this has been merged:
* https://lab.llvm.org/buildbot/#/builders/172/builds/9124
Similar failure was reported by the pre-commit CI.
The loopInfos gets invalidated after collapsing nested loops. Use the
saved afterIP since the returned afterIP by applyDynamicWorkshareLoop
may be not valid.
Reviewed By: shraiysh
Differential Revision: https://reviews.llvm.org/D120294
StandardToSPIRV currently contains an assortment of patterns converting from
different dialects to SPIRV. This commit splits up StandardToSPIRV into separate
conversions for each of the dialects involved (some of which already exist).
Differential Revision: https://reviews.llvm.org/D120767
Add support for extensible dialects, which are dialects that can be
extended at runtime with new operations and types.
These operations and types cannot at the moment implement traits
or interfaces.
Differential Revision: https://reviews.llvm.org/D104554
LLVM defines several default datalayouts for integer and floating point types that are not being considered when importing into MLIR. This patch remedies this.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D120832
https://reviews.llvm.org/D120423 replaced the use of stacksave/restore with memref.alloca_scope, but kept the save/restore at the same location. This PR places the allocation scope within the wsloop, thus keeping the same allocation scope as the original scf.parallel (e.g. no longer over stack allocating).
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D120772
Add support for translating data layout specifications for integer and float
types between MLIR and LLVM IR. This is a first step towards removing the
string-based LLVM dialect data layout attribute on modules. The latter is still
available and will remain so until the first-class MLIR modeling can fully
replace it.
Depends On D120739
Reviewed By: wsmoses
Differential Revision: https://reviews.llvm.org/D120740
Add support for integer and float types into the data layout subsystem with
default logic similar to LLVM IR. Given the flexibility of the sybsystem, the
logic can be easily overwritten by operations if necessary. This provides the
connection necessary, e.g., for the GPU target where alignment requirements for
integers and floats differ from those provided by default (although still
compatible with the LLVM IR model). Previously, it was impossible to use
non-default alignment requirements for integer and float types, which could
lead to incorrect address and size calculations when targeting GPUs.
Depends On D120737
Reviewed By: wsmoses
Differential Revision: https://reviews.llvm.org/D120739
This patch adds assemblyFormat for `omp.critical.declare`, `omp.atomic.read`,
`omp.atomic.write`, `omp.atomic.update` and `omp.atomic.capture`.
Also removing those clauses from `parseClauses` that aren't needed
anymore, thanks to the new assemblyFormats.
Reviewed By: NimishMishra, rriddle
Differential Revision: https://reviews.llvm.org/D120248
sparsity values.
Previously, we can't properly handle input tensors with a dimension
ordering that is different from the natural ordering or with a mixed of
compressed and dense dimensions. This change fixes the problems by
passing the dimension ordering and sparsity values to the runtime
routine.
Modify an existing test to test the situation.
Reviewed By: aartbik
Differential Revision: https://reviews.llvm.org/D120777
The remanants of Standard was renamed to Func, but the test directory
remained named as Standard. In adidition to fixing the name, this commit
also moves the tests for operations not in the Func dialect to the proper
parent dialect test directory.
The Func has a large number of legacy dependencies carried over from the old
Standard dialect, which was pervasive and contained a large number of varied
operations. With the split of the standard dialect and its demise, a lot of lingering
dead dependencies have survived to the Func dialect. This commit removes a
large majority of then, greatly reducing the dependence surface area of the
Func dialect.
The last remaining operations in the standard dialect all revolve around
FuncOp/function related constructs. This patch simply handles the initial
renaming (which by itself is already huge), but there are a large number
of cleanups unlocked/necessary afterwards:
* Removing a bunch of unnecessary dependencies on Func
* Cleaning up the From/ToStandard conversion passes
* Preparing for the move of FuncOp to the Func dialect
See the discussion at https://discourse.llvm.org/t/standard-dialect-the-final-chapter/6061
Differential Revision: https://reviews.llvm.org/D120624
Previously, convertToMLIRSparseTensor assumes identity storage ordering and all
compressed dimensions. This change extends the function with two parameters for
users to specify the storage ordering and the sparsity of each dimension.
Modify PyTACO to reflect this change.
Reviewed By: aartbik
Differential Revision: https://reviews.llvm.org/D120643
As discussed in https://reviews.llvm.org/D119743 scf.parallel would continuously stack allocate since the alloca op was placd in the wsloop rather than the omp.parallel. This PR is the second stage of the fix for that problem. Specifically, we now introduce an alloca scope around the inlined body of the scf.parallel and enable a canonicalization to hoist the allocations to the surrounding allocation scope (e.g. omp.parallel).
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D120423
The llvm.mlir.global operation accepts a region as initializer. This region
corresponds to an LLVM IR constant expression and therefore should not accept
operations with side effects. Add a corresponding verifier.
Reviewed By: wsmoses, bondhugula
Differential Revision: https://reviews.llvm.org/D120632
The revision renames the following OpDSL functions:
```
TypeFn.cast -> TypeFn.cast_signed
BinaryFn.min -> BinaryFn.min_signed
BinaryFn.max -> BinaryFn.max_signed
```
The corresponding enum values on the C++ side are renamed accordingly:
```
#linalg.type_fn<cast> -> #linalg.type_fn<cast_signed>
#linalg.binary_fn<min> -> #linalg.binary_fn<min_signed>
#linalg.binary_fn<max> -> #linalg.binary_fn<max_signed>
```
Depends On D120110
Reviewed By: aartbik
Differential Revision: https://reviews.llvm.org/D120562
The revision extends OpDSL with unary and binary function attributes. A function attribute, makes the operations used in the body of a structured operation configurable. For example, a pooling operation may take an aggregation function attribute that specifies if the op shall implement a min or a max pooling. The goal of this revision is to define less and more flexible operations.
We may thus for example define an element wise op:
```
linalg.elem(lhs, rhs, outs=[out], op=BinaryFn.mul)
```
If the op argument is not set the default operation is used.
Depends On D120109
Reviewed By: nicolasvasilache, aartbik
Differential Revision: https://reviews.llvm.org/D120110
Add applyStaticChunkedWorkshareLoop method implementing static schedule when chunk-size is specified. Unlike a static schedule without chunk-size (where chunk-size is chosen by the runtime such that each thread receives one chunk), we need two nested loops: one for looping over the iterations of a chunk, and a second for looping over all chunks assigned to the threads.
This patch includes the following related changes:
* Adapt applyWorkshareLoop to triage between the schedule types, now possible since all schedules have been implemented. The default schedule is assumed to be non-chunked static, as without OpenMPIRBuilder.
* Remove the chunk parameter from applyStaticWorkshareLoop, it is ignored by the runtime. Change the value for the value passed to the init function to 0, as without OpenMPIRBuilder.
* Refactor CanonicalLoopInfo::setTripCount and CanonicalLoopInfo::mapIndVar as used by both, applyStaticWorkshareLoop and applyStaticChunkedWorkshareLoop.
* Enable Clang to use the OpenMPIRBuilder in the presence of the schedule clause.
Differential Revision: https://reviews.llvm.org/D114413
Add a pattern matcher for ExtractSliceOp when its source is a constant.
The matching heuristics can be governed by the control function since
generating a new constant is not always beneficial.
Differential Revision: https://reviews.llvm.org/D119605
If we have a chain of `tensor.insert_slice` ops inserting some
`tensor.pad` op into a `linalg.fill` and ranges do not overlap,
we can also elide the `tensor.pad` later.
Reviewed By: ThomasRaoux
Differential Revision: https://reviews.llvm.org/D120446
