Add a folder to the affine.parallel op so that loop bounds expressions are canonicalized.
Additionally, a new AffineParallelNormalizePass is added to adjust affine.parallel ops so that the lower bound is always 0 and the upper bound always represents a range with a step size of 1.
Differential Revision: https://reviews.llvm.org/D84998
PDL presents a high level abstraction for the rewrite pattern infrastructure available in MLIR. This abstraction allows for representing patterns transforming MLIR, as MLIR. This allows for applying all of the benefits that the general MLIR infrastructure provides, to the infrastructure itself. This means that pattern matching can be more easily verified for correctness, targeted by frontends, and optimized.
PDL abstracts over various different aspects of patterns and core MLIR data structures. Patterns are specified via a `pdl.pattern` operation. These operations contain a region body for the "matcher" code, and terminate with a `pdl.rewrite` that either dispatches to an external rewriter or contains a region for the rewrite specified via `pdl`. The types of values in `pdl` are handle types to MLIR C++ types, with `!pdl.attribute`, `!pdl.operation`, and `!pdl.type` directly mapping to `mlir::Attribute`, `mlir::Operation*`, and `mlir::Value` respectively.
An example pattern is shown below:
```mlir
// pdl.pattern contains metadata similarly to a `RewritePattern`.
pdl.pattern : benefit(1) {
// External input operand values are specified via `pdl.input` operations.
// Result types are constrainted via `pdl.type` operations.
%resultType = pdl.type
%inputOperand = pdl.input
%root, %results = pdl.operation "foo.op"(%inputOperand) -> %resultType
pdl.rewrite(%root) {
pdl.replace %root with (%inputOperand)
}
}
```
This is a culmination of the work originally discussed here: https://groups.google.com/a/tensorflow.org/g/mlir/c/j_bn74ByxlQ
Differential Revision: https://reviews.llvm.org/D84578
LinalgDistribution options to allow more general distributions.
Changing the signature of the callback to send in the ranges for all
the parallel loops and expect a vector with the Value to use for the
processor-id and number-of-processors for each of the parallel loops.
Differential Revision: https://reviews.llvm.org/D86095
When the operand to the linalg.tensor_reshape op is a splat constant,
the result can be replaced with a splat constant of the same value but
different type.
Differential Revision: https://reviews.llvm.org/D86117
According to the LLVM Language Reference, 'cmpxchg' accepts integer or pointer
types. Several MLIR tests were using it with floats as it appears possible to
programmatically construct and print such an instruction, but it cannot be
parsed back. Use integers instead.
Depends On D85899
Reviewed By: flaub, rriddle
Differential Revision: https://reviews.llvm.org/D85900
Legacy implementation of the LLVM dialect in MLIR contained an instance of
llvm::Module as it was required to parse LLVM IR types. The access to the data
layout of this module was exposed to the users for convenience, but in practice
this layout has always been the default one obtained by parsing an empty layout
description string. Current implementation of the dialect no longer relies on
wrapping LLVM IR types, but it kept an instance of DataLayout for
compatibility. This effectively forces a single data layout to be used across
all modules in a given MLIR context, which is not desirable. Remove DataLayout
from the LLVM dialect and attach it as a module attribute instead. Since MLIR
does not yet have support for data layouts, use the LLVM DataLayout in string
form with verification inside MLIR. Introduce the layout when converting a
module to the LLVM dialect and keep the default "" description for
compatibility.
This approach should be replaced with a proper MLIR-based data layout when it
becomes available, but provides an immediate solution to compiling modules with
different layouts, e.g. for GPUs.
This removes the need for LLVMDialectImpl, which is also removed.
Depends On D85650
Reviewed By: aartbik
Differential Revision: https://reviews.llvm.org/D85652
Masked loading/storing in various forms can be optimized
into simpler memory operations when the mask is all true
or all false. Note that the backend does similar optimizations
but doing this early may expose more opportunities for further
optimizations. This further prepares progressively lowering
transfer read and write into 1-D memory operations.
Reviewed By: ThomasRaoux
Differential Revision: https://reviews.llvm.org/D85769
Now that LLVM dialect types are implemented directly in the dialect, we can use
MLIR hooks for verifying type construction invariants. Implement the verifiers
and use them in the parser.
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D85663
Linalg to processors.
This changes adds infrastructure to distribute the loops generated in
Linalg to processors at the time of generation. This addresses use
case where the instantiation of loop is done just to distribute
them. The option to distribute is added to TilingOptions for now and
will allow specifying the distribution as a transformation option,
just like tiling and promotion are specified as options.
Differential Revision: https://reviews.llvm.org/D85147
This patch also fixes a minor issue that shape.rank should allow
returning !shape.size. The dialect doc has such an example for
shape.rank.
Differential Revision: https://reviews.llvm.org/D85556
This revision aims to provide a new API, `checkTilingLegality`, to
verify that the loop tiling result still satisifes the dependence
constraints of the original loop nest.
Previously, there was no check for the validity of tiling. For instance:
```
func @diagonal_dependence() {
%A = alloc() : memref<64x64xf32>
affine.for %i = 0 to 64 {
affine.for %j = 0 to 64 {
%0 = affine.load %A[%j, %i] : memref<64x64xf32>
%1 = affine.load %A[%i, %j - 1] : memref<64x64xf32>
%2 = addf %0, %1 : f32
affine.store %2, %A[%i, %j] : memref<64x64xf32>
}
}
return
}
```
You can find more information about this example from the Section 3.11
of [1].
In general, there are three types of dependences here: two flow
dependences, one in direction `(i, j) = (0, 1)` (notation that depicts a
vector in the 2D iteration space), one in `(i, j) = (1, -1)`; and one
anti dependence in the direction `(-1, 1)`.
Since two of them are along the diagonal in opposite directions, the
default tiling method in `affine`, which tiles the iteration space into
rectangles, will violate the legality condition proposed by Irigoin and
Triolet [2]. [2] implies two tiles cannot depend on each other, while in
the `affine` tiling case, two rectangles along the same diagonal are
indeed dependent, which simply violates the rule.
This diff attempts to put together a validator that checks whether the
rule from [2] is violated or not when applying the default tiling method
in `affine`.
The canonical way to perform such validation is by examining the effect
from adding the constraint from Irigoin and Triolet to the existing
dependence constraints.
Since we already have the prior knowlegde that `affine` tiles in a
hyper-rectangular way, and the resulting tiles will be scheduled in the
same order as their respective loop indices, we can simplify the
solution to just checking whether all dependence components are
non-negative along the tiling dimensions.
We put this algorithm into a new API called `checkTilingLegality` under
`LoopTiling.cpp`. This function iterates every `load`/`store` pair, and
if there is any dependence between them, we get the dependence component
and check whether it has any negative component. This function returns
`failure` if the legality condition is violated.
[1]. Bondhugula, Uday. Effective Automatic parallelization and locality optimization using the Polyhedral model. https://dl.acm.org/doi/book/10.5555/1559029
[2]. Irigoin, F. and Triolet, R. Supernode Partitioning. https://dl.acm.org/doi/10.1145/73560.73588
Differential Revision: https://reviews.llvm.org/D84882
This simple patch translates the num_threads and if clauses of the parallel
operation. Also includes test cases.
A minor change was made to parsing of the if clause to parse AnyType and
return the parsed type. Updates to test cases also.
Reviewed by: SouraVX
Differential Revision: https://reviews.llvm.org/D84798
This change adds initial support needed to generate OpenCL compliant SPIRV.
If Kernel capability is declared then memory model becomes OpenCL.
If Addresses capability is declared then addressing model becomes Physical64.
Additionally for Kernel capability interface variable ABI attributes are not
generated as entry point function is expected to have normal arguments.
Differential Revision: https://reviews.llvm.org/D85196
This revision adds a folding pattern to replace affine.min ops by the actual min value, when it can be determined statically from the strides and bounds of enclosing scf loop .
This matches the type of expressions that Linalg produces during tiling and simplifies boundary checks. For now Linalg depends both on Affine and SCF but they do not depend on each other, so the pattern is added there.
In the future this will move to a more appropriate place when it is determined.
The canonicalization of AffineMinOp operations in the context of enclosing scf.for and scf.parallel proceeds by:
1. building an affine map where uses of the induction variable of a loop
are replaced by `%lb + %step * floordiv(%iv - %lb, %step)` expressions.
2. checking if any of the results of this affine map divides all the other
results (in which case it is also guaranteed to be the min).
3. replacing the AffineMinOp by the result of (2).
The algorithm is functional in simple parametric tiling cases by using semi-affine maps. However simplifications of such semi-affine maps are not yet available and the canonicalization does not succeed yet.
Differential Revision: https://reviews.llvm.org/D82009
This new pattern mixes vector.transpose and direct lowering to vector.reduce.
This allows more progressive lowering than immediately going to insert/extract and
composes more nicely with other canonicalizations.
This has 2 use cases:
1. for very wide vectors the generated IR may be much smaller
2. when we have a custom lowering for transpose ops we can target it directly
rather than rely LLVM
Differential Revision: https://reviews.llvm.org/D85428
When any of the memrefs in a structured linalg op has a zero dimension, it becomes dead.
This is consistent with the fact that linalg ops deduce their loop bounds from their operands.
Note however that this is not the case for the `tensor<0xelt_type>` which is a special convention
that must be lowered away into either `memref<elt_type>` or just `elt_type` before this
canonicalization can kick in.
Differential Revision: https://reviews.llvm.org/D85413
The intrinsics were already supported and vector.transfer_read/write lowered
direclty into these operations. By providing them as individual ops, however,
clients can used them directly, and it opens up progressively lowering transfer
operations at higher levels (rather than direct lowering to LLVM IR as done now).
Reviewed By: bkramer
Differential Revision: https://reviews.llvm.org/D85357
This dialect was introduced during the bring-up of the new LLVM dialect type
system for testing purposes. The main LLVM dialect now uses the new type system
and the test dialect is no longer necessary, so remove it.
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D85224
Introduces the expand and compress operations to the Vector dialect
(important memory operations for sparse computations), together
with a first reference implementation that lowers to the LLVM IR
dialect to enable running on CPU (and other targets that support
the corresponding LLVM IR intrinsics).
Reviewed By: reidtatge
Differential Revision: https://reviews.llvm.org/D84888
Simplify semi-affine expression for the operations like ceildiv,
floordiv and modulo by any given symbol by checking divisibilty by that
symbol.
Some properties used in simplification are:
1) Commutative property of the floordiv and ceildiv:
((expr1 floordiv expr2) floordiv expr3 ) = ((expr1 floordiv expr3) floordiv expr2)
((expr1 ceildiv expr2) ceildiv expr3 ) = ((expr1 ceildiv expr3) ceildiv expr2)
While simplification if operations are different no simplification is
possible as there is no property that simplify expressions like these:
((expr1 ceildiv expr2) floordiv expr3) or ((expr1 floordiv expr2)
ceildiv expr3).
2) If both expr1 and expr2 are divisible by the expr3 then:
(expr1 % expr2) / expr3 = ((expr1 / expr3) % (expr2 / expr3))
where / is divide symbol.
3) If expr1 is divisible by expr2 then expr1 % expr2 = 0.
Signed-off-by: Yash Jain <yash.jain@polymagelabs.com>
Differential Revision: https://reviews.llvm.org/D84920
This revision adds a transformation and a pattern that rewrites a "maybe masked" `vector.transfer_read %view[...], %pad `into a pattern resembling:
```
%1:3 = scf.if (%inBounds) {
scf.yield %view : memref<A...>, index, index
} else {
%2 = linalg.fill(%extra_alloc, %pad)
%3 = subview %view [...][...][...]
linalg.copy(%3, %alloc)
memref_cast %extra_alloc: memref<B...> to memref<A...>
scf.yield %4 : memref<A...>, index, index
}
%res= vector.transfer_read %1#0[%1#1, %1#2] {masked = [false ... false]}
```
where `extra_alloc` is a top of the function alloca'ed buffer of one vector.
This rewrite makes it possible to realize the "always full tile" abstraction where vector.transfer_read operations are guaranteed to read from a padded full buffer.
The extra work only occurs on the boundary tiles.
A new first-party modeling for LLVM IR types in the LLVM dialect has been
developed in parallel to the existing modeling based on wrapping LLVM `Type *`
instances. It resolves the long-standing problem of modeling identified
structure types, including recursive structures, and enables future removal of
LLVMContext and related locking mechanisms from LLVMDialect.
This commit only switches the modeling by (a) renaming LLVMTypeNew to LLVMType,
(b) removing the old implementaiton of LLVMType, and (c) updating the tests. It
is intentionally minimal. Separate commits will remove the infrastructure built
for the transition and update API uses where appropriate.
Depends On D85020
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D85021
This revision adds a transformation and a pattern that rewrites a "maybe masked" `vector.transfer_read %view[...], %pad `into a pattern resembling:
```
%1:3 = scf.if (%inBounds) {
scf.yield %view : memref<A...>, index, index
} else {
%2 = vector.transfer_read %view[...], %pad : memref<A...>, vector<...>
%3 = vector.type_cast %extra_alloc : memref<...> to
memref<vector<...>> store %2, %3[] : memref<vector<...>> %4 =
memref_cast %extra_alloc: memref<B...> to memref<A...> scf.yield %4 :
memref<A...>, index, index
}
%res= vector.transfer_read %1#0[%1#1, %1#2] {masked = [false ... false]}
```
where `extra_alloc` is a top of the function alloca'ed buffer of one vector.
This rewrite makes it possible to realize the "always full tile" abstraction where vector.transfer_read operations are guaranteed to read from a padded full buffer.
The extra work only occurs on the boundary tiles.
Differential Revision: https://reviews.llvm.org/D84631
This reverts commit 35b65be041.
Build is broken with -DBUILD_SHARED_LIBS=ON with some undefined
references like:
VectorTransforms.cpp:(.text._ZN4llvm12function_refIFvllEE11callback_fnIZL24createScopedInBoundsCondN4mlir25VectorTransferOpInterfaceEE3$_8EEvlll+0xa5): undefined reference to `mlir::edsc::op::operator+(mlir::Value, mlir::Value)'
The current modeling of LLVM IR types in MLIR is based on the LLVMType class
that wraps a raw `llvm::Type *` and delegates uniquing, printing and parsing to
LLVM itself. This model makes thread-safe type manipulation hard and is being
progressively replaced with a cleaner MLIR model that replicates the type
system. Introduce a set of classes reflecting the LLVM IR type system in MLIR
instead of wrapping the existing types. These are currently introduced as
separate classes without affecting the dialect flow, and are exercised through
a test dialect. Once feature parity is reached, the old implementation will be
gradually substituted with the new one.
Depends On D84171
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D84339
This revision adds a transformation and a pattern that rewrites a "maybe masked" `vector.transfer_read %view[...], %pad `into a pattern resembling:
```
%1:3 = scf.if (%inBounds) {
scf.yield %view : memref<A...>, index, index
} else {
%2 = vector.transfer_read %view[...], %pad : memref<A...>, vector<...>
%3 = vector.type_cast %extra_alloc : memref<...> to
memref<vector<...>> store %2, %3[] : memref<vector<...>> %4 =
memref_cast %extra_alloc: memref<B...> to memref<A...> scf.yield %4 :
memref<A...>, index, index
}
%res= vector.transfer_read %1#0[%1#1, %1#2] {masked = [false ... false]}
```
where `extra_alloc` is a top of the function alloca'ed buffer of one vector.
This rewrite makes it possible to realize the "always full tile" abstraction where vector.transfer_read operations are guaranteed to read from a padded full buffer.
The extra work only occurs on the boundary tiles.
Differential Revision: https://reviews.llvm.org/D84631
This patch handles loopControl and selectionControl in parsing and
printing. In order to reuse the functionality, and avoid handling cases when
`{` of the region is parsed as a dictionary attribute, `control` keyword was
introduced.`None` is a default control attribute. This functionality can be
later extended to `spv.func`.
Also, loopControl and selectionControl can now be (de)serialized.
Reviewed By: antiagainst
Differential Revision: https://reviews.llvm.org/D84175
This is an operation that can returns a new ValueShape with a different shape. Useful for composing shape function calls and reusing existing shape transfer functions.
Just adding the op in this change.
Differential Revision: https://reviews.llvm.org/D84217
This change allow CooperativeMatrix Load/Store operations to use pointer type
that may not match the matrix element type. This allow us to declare buffer
with a larger type size than the matrix element type. This follows SPIR-V spec
and this is needed to be able to use cooperative matrix in combination with
shared local memory efficiently.
Differential Revision: https://reviews.llvm.org/D84993
In a context in which `shape.broadcast` is known not to produce an error value,
we want it to operate solely on extent tensors. The operation's behavior is
then undefined in the error case as the result type cannot hold this value.
Differential Revision: https://reviews.llvm.org/D84933
Replaced definition of named ND ConvOps with tensor comprehension
syntax which reduces boilerplate code significantly. Furthermore,
new ops to support TF convolutions added (without strides and dilations).
Reviewed By: nicolasvasilache
Differential Revision: https://reviews.llvm.org/D84628
This commit is part of a greater project which aims to add
full end-to-end support for convolutions inside mlir. The
reason behind having conv ops for each rank rather than
having one generic ConvOp is to enable better optimizations
for every N-D case which reflects memory layout of input/kernel
buffers better and simplifies code as well. We expect plain linalg.conv
to be progressively retired.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D83879
This adds conversions for const_size and to_extent_tensor. Also, cast-like operations are now folded away if the source and target types are the same.
Differential Revision: https://reviews.llvm.org/D84745
Added a check for 'Function' storage class in `spv.globalVariable`
verifier since it only can be used with `spv.Variable`.
Reviewed By: antiagainst
Differential Revision: https://reviews.llvm.org/D84731
The current transformation to shape.reduce does not support tensor values.
This adds the required changes to make that work, including fixing the builder
for shape.reduce.
Differential Revision: https://reviews.llvm.org/D84744
- replace DotOp, now that DRR rules have been dropped.
- Capture arguments mismatch in the parser. The number of parsed arguments must
equal the number of expected arguments.
Reviewed By: ftynse, nicolasvasilache
Differential Revision: https://reviews.llvm.org/D82952
linalg.indexed_generic (consumer) with tensor arguments.
The implementation of fusing std.constant producer with a
linalg.indexed_generic consumer was already in place. It is exposed
with this change. Also cleaning up some of the patterns that implement
the fusion to not be templated, thereby avoiding lot of conditional
checks for calling the right instantiation.
Differential Revision: https://reviews.llvm.org/D84566
This diff provides a concrete test case for the error that will be raised when the iteration space is non hyper-rectangular.
The corresponding emission method for this error message has been changed as well.
Differential Revision: https://reviews.llvm.org/D84531
Previous changes generalized some of the operands and results. Complete
a larger group of those to simplify progressive lowering. Also update
some of the declarative asm form due to generalization. Tried to keep it
mostly mechanical.
Based on https://reviews.llvm.org/D84439 but less restrictive, else we
don't allow shape_of to be able to produce a ranked output and doesn't
allow for iterative refinement here. We can consider making it more
restrictive later.