Based on dicussion in
[this](https://llvm.discourse.group/t/remove-canonicalizer-for-memref-dim-via-shapedtypeopinterface/3641)
thread the pattern to resolve the `memref.dim` of a value that is a
result of an operation that implements the
`InferShapedTypeOpInterface` is moved to a separate pass instead of
running it as a canonicalization pass. This allows shape resolution to
happen when explicitly required, instead of automatically through a
canonicalization.
Differential Revision: https://reviews.llvm.org/D104321
Make store to load fwd condition for -memref-dataflow-opt less
conservative. Post dominance info is not really needed. Add additional
check for common cases.
Differential Revision: https://reviews.llvm.org/D104174
To control the number of outer parallel loops, we need to process the
outer loops first and hence pre-order walk fixes the issue.
Reviewed By: bondhugula
Differential Revision: https://reviews.llvm.org/D104361
We have several ways of introducing a scalar invariant value into
linalg generic ops (should we limit this somewhat?). This revision
makes sure we handle all of them correctly in the sparse compiler.
Reviewed By: gysit
Differential Revision: https://reviews.llvm.org/D104335
This is a very careful start with alllowing sparse tensors at the
left-hand-side of tensor index expressions (viz. sparse output).
Note that there is a subtle difference between non-annotated tensors
(dense, remain n-dim, handled by classic bufferization) and all-dense
annotated "sparse" tensors (linearized to 1-dim without overhead
storage, bufferized by sparse compiler, backed by runtime support library).
This revision gently introduces some new IR to facilitate annotated outputs,
to be generalized to truly sparse tensors in the future.
Reviewed By: gussmith23, bixia
Differential Revision: https://reviews.llvm.org/D104074
This doesn't add any canonicalizations, but executes the same
simplification on bufferSemantic linalg.generic ops by using
linalg::ReshapeOp instead of linalg::TensorReshapeOp.
Differential Revision: https://reviews.llvm.org/D103513
The parser of generic op did not recognize the output from mlir-opt when there
are multiple outputs. One would wrap the result types with braces, and one would
not. The patch makes the behavior the same.
Reviewed By: mravishankar
Differential Revision: https://reviews.llvm.org/D104256
Interface patterns are unique in that they get added to every operation that also implements that interface, given that they aren't tied to individual operations. When the same interface pattern gets added to multiple operations (such as the current behavior with Linalg), an reference to each of these patterns is added to every op (meaning that an operation will now have N references to effectively the same pattern). This revision fixes this problematic behavior in Linalg, and can bring upwards of a 25% reduction in compile time in Linalg based workloads.
Differential Revision: https://reviews.llvm.org/D104160
There's no need for `toSmallVector()` as `SmallVector.h` already provides a `to_vector` free function that takes a range.
Reviewed By: Quuxplusone
Differential Revision: https://reviews.llvm.org/D104024
Up to now all structured op operands are assumed to be shaped. The patch relaxes this assumption and allows scalar input operands. In contrast to shaped operands scalar operands are not indexed and directly forwarded to the body of the operation. As all other operands, scalar operands are associated to an indexing map that in case of a scalar or a 0D-operand has an empty range.
We will use scalar operands as a replacement for the capture mechanism. In contrast to captures, the approach ensures we can generate the function signature from the operand list and it prevents outdated capture values in case a transformation updates only the capture operand but not the hidden body of a named operation.
Removing captures and updating existing operations such as linalg.fill is left for a later patch.
The patch depends on https://reviews.llvm.org/D103891 and https://reviews.llvm.org/D103890.
Differential Revision: https://reviews.llvm.org/D104109
The padding of such ops is not generated in a vectorized way. Instead, emit a tensor::GenerateOp.
We may vectorize GenerateOps in the future.
Differential Revision: https://reviews.llvm.org/D103879
If the source operand of a linalg.pad_op operation has static shape, vectorize the copying of the source.
Differential Revision: https://reviews.llvm.org/D103747
Currently limited to constant pad values. Any combination of dynamic/static tensor sizes and padding sizes is supported.
Differential Revision: https://reviews.llvm.org/D103679
The generic vectorization pattern handles only those cases, where
low and high padding is zero. This is already handled by a
canonicalization pattern.
Also add a new canonicalization test case to ensure that tensor cast ops
are properly inserted.
A more general vectorization pattern will be added in a subsequent commit.
Differential Revision: https://reviews.llvm.org/D103590
Vectorize linalg.pad_tensor without generating a linalg.init_tensor when consumed by a transfer_write.
Differential Revision: https://reviews.llvm.org/D103137
Vectorize linalg.pad_tensor without generating a linalg.init_tensor when consumed by a subtensor_insert.
Differential Revision: https://reviews.llvm.org/D103780
Vectorize linalg.pad_tensor without generating a linalg.init_tensor when consumed by a transfer_read.
Differential Revision: https://reviews.llvm.org/D103735
* Add a helper function that returns the constant padding value (if applicable).
* Remove existing getConstantYieldValueFromBlock function, which does almost the same.
* Adapted from D103243.
Differential Revision: https://reviews.llvm.org/D104004
Add `tensor.insert` op to make `tensor.extract`/`tensor.insert` work in pairs
for `scalar` domain. Like `subtensor`/`subtensor_insert` work in pairs in
`tensor` domain, and `vector.transfer_read`/`vector.transfer_write` work in
pairs in `vector` domain.
Reviewed By: silvas
Differential Revision: https://reviews.llvm.org/D104139
The commit simplifies affine.if ops :
The affine if operation gets removed if the condition is universally true or false and then/else block is merged with the parent block.
Signed-off-by: Shashij Gupta shashij.gupta@polymagelabs.com
Reviewed By: bondhugula, pr4tgpt
Differential Revision: https://reviews.llvm.org/D104015
## Introduction
This proposal describes the new op to be added to the `std` (and later moved `memref`)
dialect called `alloca_scope`.
## Motivation
Alloca operations are easy to misuse, especially if one relies on it while doing
rewriting/conversion passes. For example let's consider a simple example of two
independent dialects, one defines an op that wants to allocate on-stack and
another defines a construct that corresponds to some form of looping:
```
dialect1.looping_op {
%x = dialect2.stack_allocating_op
}
```
Since the dialects might not know about each other they are going to define a
lowering to std/scf/etc independently:
```
scf.for … {
%x_temp = std.alloca …
… // do some domain-specific work using %x_temp buffer
… // and store the result into %result
%x = %result
}
```
Later on the scf and `std.alloca` is going to be lowered to llvm using a
combination of `llvm.alloca` and unstructured control flow.
At this point the use of `%x_temp` is bound to either be either optimized by
llvm (for example using mem2reg) or in the worst case: perform an independent
stack allocation on each iteration of the loop. While the llvm optimizations are
likely to succeed they are not guaranteed to do so, and they provide
opportunities for surprising issues with unexpected use of stack size.
## Proposal
We propose a new operation that defines a finer-grain allocation scope for the
alloca-allocated memory called `alloca_scope`:
```
alloca_scope {
%x_temp = alloca …
...
}
```
Here the lifetime of `%x_temp` is going to be bound to the narrow annotated
region within `alloca_scope`. Moreover, one can also return values out of the
alloca_scope with an accompanying `alloca_scope.return` op (that behaves
similarly to `scf.yield`):
```
%result = alloca_scope {
%x_temp = alloca …
…
alloca_scope.return %myvalue
}
```
Under the hood the `alloca_scope` is going to lowered to a combination of
`llvm.intr.stacksave` and `llvm.intr.strackrestore` that are going to be invoked
automatically as control-flow enters and leaves the body of the `alloca_scope`.
The key value of the new op is to allow deterministic guaranteed stack use
through an explicit annotation in the code which is finer-grain than the
function-level scope of `AutomaticAllocationScope` interface. `alloca_scope`
can be inserted at arbitrary locations and doesn’t require non-trivial
transformations such as outlining.
## Which dialect
Before memref dialect is split, `alloca_scope` can temporarily reside in `std`
dialect, and later on be moved to `memref` together with the rest of
memory-related operations.
## Implementation
An implementation of the op is available [here](https://reviews.llvm.org/D97768).
Original commits:
* Add initial scaffolding for alloca_scope op
* Add alloca_scope.return op
* Add no region arguments and variadic results
* Add op descriptions
* Add failing test case
* Add another failing test
* Initial implementation of lowering for std.alloca_scope
* Fix backticks
* Fix getSuccessorRegions implementation
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D97768
This is a roll forward of D102679.
This patch simplifies the implementation of Sequence and makes it compatible with llvm::reverse.
It exposes the reverse iterators through rbegin/rend which prevents a dangling reference in std::reverse_iterator::operator++().
Note: Compared to D102679, this patch introduces a `asSmallVector()` member function and fixes compilation issue with GCC 5.
Differential Revision: https://reviews.llvm.org/D103948
This brings us closer to replacing the LLVM data layout string with a
first-class layout modeling in MLIR.
Depends On D103945
Reviewed By: nicolasvasilache
Differential Revision: https://reviews.llvm.org/D103946
Allow gpu ops implementing the async interface to already be async when running the GpuAsyncRegionPass.
That pass threads a 'current token' through a block with ops implementing the gpu async interface.
After this change, existing async ops (returning a !gpu.async.token) set the current token.
Existing synchronous `gpu.wait` ops reset the current token.
Reviewed By: herhut
Differential Revision: https://reviews.llvm.org/D103396
This allows us to remove the `spv.mlir.endmodule` op and
all the code associated with it.
Along the way, tightened the APIs for `spv.module` a bit
by removing some aliases. Now we use `getRegion` to get
the only region, and `getBody` to get the region's only
block.
Reviewed By: mravishankar, hanchung
Differential Revision: https://reviews.llvm.org/D103265
ArmSVE-specific memory operations are needed to generate end-to-end
code for as long as MLIR core doesn't support scalable vectors. This
instructions will be eventually unnecessary, for now they're required
for more complex testing.
Differential Revision: https://reviews.llvm.org/D103535
These `arm_sve.cmp` functions are needed to generate scalable vector
masks as long as scalable vectors are not part of the standard types.
Once in standard, these can be removed and `std.cmp` can be used
instead.
Differential Revision: https://reviews.llvm.org/D103473
This reverts commit e772216e70
(and fixup 7f6c878a2c).
The build is broken with gcc5 host compiler:
In file included from
from mlir/lib/Dialect/Utils/StructuredOpsUtils.cpp:9:
tools/mlir/include/mlir/IR/BuiltinAttributes.h.inc:424:57: error: type/value mismatch at argument 1 in template parameter list for 'template<class ItTy, class FuncTy, class FuncReturnTy> class llvm::mapped_iterator'
std::function<T(ptrdiff_t)>>;
^
tools/mlir/include/mlir/IR/BuiltinAttributes.h.inc:424:57: note: expected a type, got 'decltype (seq<ptrdiff_t>(0, 0))::const_iterator'
This is both more efficient and more ergonomic than going
through an std::string, e.g. when using llvm::utostr and
in string concat cases.
Unfortunately we can't just overload ::get(). This causes an
ambiguity because both twine and stringref implicitly convert
from std::string.
Differential Revision: https://reviews.llvm.org/D103754
Currently canonicalizations of a store and a cast try to fold all casts into the store.
In the case where the operand being stored is itself a cast, this is illegal as the type of the value being stored
will change. This PR fixes this by not checking the value for folding with a cast.
Depends on https://reviews.llvm.org/D103828
Differential Revision: https://reviews.llvm.org/D103829
This patch simplifies the implementation of Sequence and makes it compatible with llvm::reverse.
It exposes the reverse iterators through rbegin/rend which prevents a dangling reference in std::reverse_iterator::operator++().
Differential Revision: https://reviews.llvm.org/D102679
These `arm_sve.cmp` functions are needed to generate scalable vector
masks as long as scalable vectors are not part of the standard types.
Once in standard, these can be removed and `std.cmp` can be used
instead.
Differential Revision: https://reviews.llvm.org/D103473
In an operation in the true/false dest of a branch,
one can assume that the operation itself was true/false if
only that edge can reach the operation.
Differential Revision: https://reviews.llvm.org/D101709
This patch add canonicalization for the standalone data operation with constant if condition.
It is extracted from this patch D103325.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D103712
Convert data operands from the acc.parallel operation using the same conversion pattern than D102170.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D103337
Implements better naming for results of spv.mlir.addressof ops by making it
inherit from OpAsmOpInterface and implementing the associated
getAsmResultName(...) hook.
Reviewed By: antiagainst
Differential Revision: https://reviews.llvm.org/D103594
* Add hasUnitStride and hasZeroOffset to OffsetSizeAndStrideOpInterface. These functions are useful for various patterns. E.g., some vectorization patterns apply only for tensor ops with zero offsets and/or unit stride.
* Add getConstantIntValue and isEqualConstantInt helper functions, which are useful for implementing the two above functions, as well as various patterns.
Differential Revision: https://reviews.llvm.org/D103763
Controlled by a compiler option, if 32-bit indices can be handled
with zero/sign-extention alike (viz. no worries on non-negative
indices), scatter/gather operations can use the more efficient
32-bit SIMD version.
Reviewed By: bixia
Differential Revision: https://reviews.llvm.org/D103632
* Rename PadTensorOpVectorizationPattern to GenericPadTensorOpVectorizationPattern.
* Make GenericPadTensorOpVectorizationPattern a private pattern, to be instantiated via populatePadTensorOpVectorizationPatterns.
* Factor out parts of PadTensorOpVectorizationPattern into helper functions.
This commit prepares PadTensorOpVectorizationPattern for a series of subsequent commits that add more specialized PadTensorOp vectorization patterns.
Differential Revision: https://reviews.llvm.org/D103681
Convert data operands from the acc.data operation using the same conversion pattern than D102170.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D103332
Introduces a test pass that rewrites PadTensorOps with static shapes as a sequence of:
```
linalg.init_tensor // to create output
linalg.fill // to initialize with padding value
linalg.generic // to copy the original contents to the padded tensor
```
The pass can be triggered with:
- `--test-linalg-transform-patterns="test-transform-pad-tensor"`
Differential Revision: https://reviews.llvm.org/D102804
Replace the uses of deprecated Structured Op Interface methods in TestLinalgElementwiseFusion.cpp, TestLinalgFusionTransforms.cpp, and Transforms.cpp. The patch is based on https://reviews.llvm.org/D103394.
Differential Revision: https://reviews.llvm.org/D103528
Adding methods to access operand properties via OpOperands and mark outdated methods as deprecated.
Differential Revision: https://reviews.llvm.org/D103394
Implements better naming for results of `spv.Constant` ops by making it
inherit from OpAsmOpInterface and implementing the associated
getAsmResultName(...) hook.
Reviewed By: antiagainst
Differential Revision: https://reviews.llvm.org/D103152
Depends On D103109
If any of the tokens/values added to the `!async.group` switches to the error state, than the group itself switches to the error state.
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D103203
Depends On D103102
Not yet implemented:
1. Error handling after synchronous await
2. Error handling for async groups
Will be addressed in the followup PRs
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D103109
Support reference counted values implicitly passed (live) only to some of the successors.
Example: if branched to ^bb2 token will leak, unless `drop_ref` operation is properly created
```
^entry:
%token = async.runtime.create : !async.token
cond_br %cond, ^bb1, ^bb2
^bb1:
async.runtime.await %token
async.runtime.drop_ref %token
br ^bb2
^bb2:
return
```
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D103102
In order to allow large matmul operations using the MMA ops we need to chain
operations this is not possible unless "DOp" and "COp" type have matching
layout so remove the "DOp" layout and force accumulator and result type to
match.
Added a test for the case where the MMA value is accumulated.
Differential Revision: https://reviews.llvm.org/D103023
This revision refactors and simplifies the pattern detection logic: thanks to SSA value properties, we can actually look at all the uses of a given value and avoid having to pattern-match specific chains of operations.
A bufferization pattern for subtensor is added and specific inplaceability analysis is implemented for the simple case of subtensor. More advanced use cases will follow.
Differential revision: https://reviews.llvm.org/D102512
Allow support for specifying empty IVs in an `affine.parallel`.
For example:
```
affine.parallel () = () to () {
affine.yield
}
```
Reviewed By: bondhugula, jbruestle
Differential Revision: https://reviews.llvm.org/D102895
Prevent users of `iter_args` of an affine for loop from being hoisted
out of it. Otherwise, LICM leads to a violation of the SSA dominance
(as demonstrated in the added test case).
Fixes: https://bugs.llvm.org/show_bug.cgi?id=50103
Reviewed By: bondhugula, ayzhuang
Differential Revision: https://reviews.llvm.org/D102984
This previously handled memref::SubviewOp, but this can be extended to
all ops implementing the interface.
Differential Revision: https://reviews.llvm.org/D103076
Fix inconsistent MLIR CMake variable names. Consistently name them as
MLIR_ENABLE_<feature>.
Eg: MLIR_CUDA_RUNNER_ENABLED -> MLIR_ENABLE_CUDA_RUNNER
MLIR follows (or has mostly followed) the convention of naming
cmake enabling variables in the from MLIR_ENABLE_... etc. Using a
convention here is easy and also important for convenience. A counter
pattern was started with variables named MLIR_..._ENABLED. This led to a
sequence of related counter patterns: MLIR_CUDA_RUNNER_ENABLED,
MLIR_ROCM_RUNNER_ENABLED, etc.. From a naming standpoint, the imperative
form is more meaningful. Additional discussion at:
https://llvm.discourse.group/t/mlir-cmake-enable-variable-naming-convention/3520
Switch all inconsistent ones to the ENABLE form. Keep the couple of old
mappings needed until buildbot config is migrated.
Differential Revision: https://reviews.llvm.org/D102976
This makes it possible for targets to define their own MCObjectFileInfo.
This MCObjectFileInfo is then used to determine things like section alignment.
This is a follow up to D101462 and prepares for the RISCV backend defining the
text section alignment depending on the enabled extensions.
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D101921
This revision completes the "dimension ordering" feature
of sparse tensor types that enables the programmer to
define a preferred order on dimension access (other than
the default left-to-right order). This enables e.g. selection
of column-major over row-major storage for sparse matrices,
but generalized to any rank, as in:
dimOrdering = affine_map<(i,j,k,l,m,n,o,p) -> (p,o,j,k,i,l,m,n)>
Reviewed By: bixia
Differential Revision: https://reviews.llvm.org/D102856
The previous implementation did not handle casting behavior properly and
did not consider aliases.
Differential Revision: https://reviews.llvm.org/D102785
This pattern inlines operands to a linalg.generic operation that use a constant
index and hence are loop-invariant scalars. This reduces the number of
linalg.generic operands and unlocks some canonicalizations that rely on seeing
an explicit tensor.extract.
Differential Revision: https://reviews.llvm.org/D102682
Skip the sparsification pass for Linalg ops without annotated tensors
(or cases that are not properly handled yet).
Reviewed By: bixia
Differential Revision: https://reviews.llvm.org/D102787
The patch extends the yaml code generation to support the following new OpDSL constructs:
- captures
- constants
- iteration index accesses
- predefined types
These changes have been introduced by revision
https://reviews.llvm.org/D101364.
Differential Revision: https://reviews.llvm.org/D102075
VectorTransferPermutationMapLoweringPatterns can be enabled via a pass option. These additional patterns lower permutation maps to minor identity maps with broadcasting, if possible, allowing for more efficient vector load/stores. The option is deactivated by default.
Differential Revision: https://reviews.llvm.org/D102593
LinalgOps that are all parallel do not use the value of `outs`
tensor. The semantics is that the `outs` tensor is fully
overwritten. Using anything other than `init_tensor` can add false
dependencies between operations, when the use is just for the shape of
the tensor. Adding a canonicalization to always use `init_tensor` in
such cases, breaks this dependence.
Differential Revision: https://reviews.llvm.org/D102561
Original interfaces are not safe to be called during dialect conversion.
This is because some ops (e.g. `dynamic_reshape(input, target_shape)`)
depend on the values of their operands to calculate the output shape.
However the operands may be out of reach during dialect conversion (e.g.
converting from tensor world to buffer world). This patch provides a new
kind of interface which accpets user-provided operands to solve this
problem.
Reviewed By: herhut
Differential Revision: https://reviews.llvm.org/D102317
- Enables inferring return type for ConstShape, takes into account valid return types;
- The compatible return type function could be reused, leaving that for next use refactoring;
Differential Revision: https://reviews.llvm.org/D102182
The experimental flag for "inplace" bufferization in the sparse
compiler can be replaced with the new inplace attribute. This gives
a uniform way of expressing the more efficient way of bufferization.
Reviewed By: bixia
Differential Revision: https://reviews.llvm.org/D102538
Broadcast dimensions of vector transfer ops are always in-bounds. This is consistent with the fact that the starting position of a transfer is always in-bounds.
Differential Revision: https://reviews.llvm.org/D102566
Splitting the memref dialect lead to an introduction of several dependencies
to avoid compilation issues. The canonicalize pass also depends on the
memref dialect, but it shouldn't. This patch resolves the dependencies
and the unintuitive includes are removed. However, the dependency moves
to the constructor of the std dialect.
Differential Revision: https://reviews.llvm.org/D102060
Replace the templated linalgLowerOpToLoops method by three specialized methods linalgOpToLoops, LinalgOpToParallelLoops, and linalgOpToAffineLoops.
Differential Revision: https://reviews.llvm.org/D102324
Add TransferWritePermutationLowering, which replaces permutation maps of TransferWriteOps with vector.transpose.
Differential Revision: https://reviews.llvm.org/D102548
We are moving from just dense/compressed to more general dim level
types, so we need more than just an "i1" array for annotations.
Reviewed By: bixia
Differential Revision: https://reviews.llvm.org/D102520
This change allows the SRC and DST of dma_start operations to be located in the
same memory space. This applies to both the Affine dialect and Memref dialect
versions of these Ops. The documention has been updated to reflect this by
explicitly stating overlapping memory locations are not supported (undefined
behavior).
Reviewed By: bondhugula
Differential Revision: https://reviews.llvm.org/D102274
This covers the extremely common case of replacing all uses of a Value
with a new op that is itself a user of the original Value.
This should also be a little bit more efficient than the
`SmallPtrSet<Operation *, 1>{op}` idiom that was being used before.
Differential Revision: https://reviews.llvm.org/D102373
Support OpImageQuerySize in spirv dialect
co-authored-by: Alan Liu <alanliu.yf@gmail.com>
Reviewed By: antiagainst
Differential Revision: https://reviews.llvm.org/D102029
Broadcast dimensions of a vector transfer op have no corresponding dimension in the mask vector. E.g., a 2-D TransferReadOp, where one dimension is a broadcast, can have a 1-D `mask` attribute.
This commit also adds a few additional transfer op integration tests for various combinations of broadcasts, masking, dim transposes, etc.
Differential Revision: https://reviews.llvm.org/D101745
Broadcast dimensions of a vector transfer op have no corresponding dimension in the mask vector. E.g., a 2-D TransferReadOp, where one dimension is a broadcast, can have a 1-D `mask` attribute.
This commit also adds a few additional transfer op integration tests for various combinations of broadcasts, masking, dim transposes, etc.
Differential Revision: https://reviews.llvm.org/D101745
The current static checker for linalg does not work on the decreasing
index cases well. So, this is to Update the current static bound checker
for linalg to cover decreasing index cases.
Reviewed By: hanchung
Differential Revision: https://reviews.llvm.org/D102302
Add a conversion pass to convert higher-level type before translation.
This conversion extract meangingful information and pack it into a struct that
the translation (D101504) will be able to understand.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D102170
First step in adding alignment as an attribute to MLIR global definitions. Alignment can be specified for global objects in LLVM IR. It can also be specified as a named attribute in the LLVMIR dialect of MLIR. However, this attribute has no standing and is discarded during translation from MLIR to LLVM IR. This patch does two things: First, it adds the attribute to the syntax of the llvm.mlir.global operation, and by doing this it also adds accessors and verifications. The syntax is "align=XX" (with XX being an integer), placed right after the value of the operation. Second, it allows transforming this operation to and from LLVM IR. It is checked whether the value is an integer power of 2.
Reviewed By: ftynse, mehdi_amini
Differential Revision: https://reviews.llvm.org/D101492
This is actually necessary for correctness, as memref.reinterpret_cast
doesn't verify if the output shape doesn't match the static sizes.
Differential Revision: https://reviews.llvm.org/D102232
VectorTransfer split previously only split read xfer ops. This adds
the same logic to write ops. The resulting code involves 2
conditionals for write ops while read ops only needed 1, but the created
ops are built upon the same patterns, so pattern matching/expectations
are all consistent other than in regards to the if/else ops.
Differential Revision: https://reviews.llvm.org/D102157
All glue and clutter in the linalg ops has been replaced by proper
sparse tensor type encoding. This code is no longer needed. Thanks
to ntv@ for giving us a temporary home in linalg.
So long, and thanks for all the fish.
Reviewed By: bixia
Differential Revision: https://reviews.llvm.org/D102098
A very elaborate, but also very fun revision because all
puzzle pieces are finally "falling in place".
1. replaces lingalg annotations + flags with proper sparse tensor types
2. add rigorous verification on sparse tensor type and sparse primitives
3. removes glue and clutter on opaque pointers in favor of sparse tensor types
4. migrates all tests to use sparse tensor types
NOTE: next CL will remove *all* obsoleted sparse code in Linalg
Reviewed By: bixia
Differential Revision: https://reviews.llvm.org/D102095
According to the API contract, LinalgLoopDistributionOptions
expects to work on parallel iterators. When getting processor
information, only loop ranges for parallel dimensions should
be fed in. But right now after generating scf.for loop nests,
we feed in *all* loops, including the ones materialized for
reduction iterators. This can cause unexpected distribution
of reduction dimensions. This commit fixes it.
Reviewed By: mravishankar
Differential Revision: https://reviews.llvm.org/D102079
In the buffer deallocation pass, unranked memref types are not properly supported.
After investigating this issue, it turns out that the Clone and Dealloc operation
does not support unranked memref types in the current implementation.
This patch adds the missing feature and enables the transformation of any memref
type.
This patch solves this bug: https://bugs.llvm.org/show_bug.cgi?id=48385
Differential Revision: https://reviews.llvm.org/D101760
The current design uses a unique entry for each argument/result attribute, with the name of the entry being something like "arg0". This provides for a somewhat sparse design, but ends up being much more expensive (from a runtime perspective) in-practice. The design requires building a string every time we lookup the dictionary for a specific arg/result, and also requires N attribute lookups when collecting all of the arg/result attribute dictionaries.
This revision restructures the design to instead have an ArrayAttr that contains all of the attribute dictionaries for arguments and another for results. This design reduces the number of attribute name lookups to 1, and allows for O(1) lookup for individual element dictionaries. The major downside is that we can end up with larger memory usage, as the ArrayAttr contains an entry for each element even if that element has no attributes. If the memory usage becomes too problematic, we can experiment with a more sparse structure that still provides a lot of the wins in this revision.
This dropped the compilation time of a somewhat large TensorFlow model from ~650 seconds to ~400 seconds.
Differential Revision: https://reviews.llvm.org/D102035
Replace all `linalg.indexed_generic` ops by `linalg.generic` ops that access the iteration indices using the `linalg.index` op.
Differential Revision: https://reviews.llvm.org/D101612
The pattern to convert subtensor ops to their rank-reduced versions
(by dropping unit-dims in the result) can also convert to a zero-rank
tensor. Handle that case.
This also fixes a OOB access bug in the existing pattern for such
cases.
Differential Revision: https://reviews.llvm.org/D101949
This expose a lambda control instead of just a boolean to control unit
dimension folding.
This however gives more control to user to pick a good heuristic.
Folding reshapes helps fusion opportunities but may generate sub-optimal
generic ops.
Differential Revision: https://reviews.llvm.org/D101917
Fixing a minor bug which lead to element type of the output being
modified when folding reshapes with generic op.
Differential Revision: https://reviews.llvm.org/D101942
This untangles the MCContext and the MCObjectFileInfo. There is a circular
dependency between MCContext and MCObjectFileInfo. Currently this dependency
also exists during construction: You can't contruct a MOFI without a MCContext
without constructing the MCContext with a dummy version of that MOFI first.
This removes this dependency during construction. In a perfect world,
MCObjectFileInfo wouldn't depend on MCContext at all, but only be stored in the
MCContext, like other MC information. This is future work.
This also shifts/adds more information to the MCContext making it more
available to the different targets. Namely:
- TargetTriple
- ObjectFileType
- SubtargetInfo
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D101462
These instructions map to SVE-specific instrinsics that accept a
predicate operand to support control flow in vector code.
Differential Revision: https://reviews.llvm.org/D100982
This patch adds support for vectorizing loops with 'iter_args'
implementing known reductions along the vector dimension. Comparing to
the non-vector-dimension case, two additional things are done during
vectorization of such loops:
- The resulting vector returned from the loop is reduced to a scalar
using `vector.reduce`.
- In some cases a mask is applied to the vector yielded at the end of
the loop to prevent garbage values from being written to the
accumulator.
Vectorization of reduction loops is disabled by default. To enable it, a
map from loops to array of reduction descriptors should be explicitly passed to
`vectorizeAffineLoops`, or `vectorize-reductions=true` should be passed
to the SuperVectorize pass.
Current limitations:
- Loops with a non-unit step size are not supported.
- n-D vectorization with n > 1 is not supported.
Reviewed By: nicolasvasilache
Differential Revision: https://reviews.llvm.org/D100694
The old index op handling let the new index operations point back to the
producer block. As a result, after fusion some index operations in the
fused block had back references to the old producer block resulting in
illegal IR. The patch now relies on a block and value mapping to avoid
such back references.
Differential Revision: https://reviews.llvm.org/D101887
While we figure out how to best add Standard support for scalable
vectors, these instructions provide a workaround for basic arithmetic
between scalable vectors.
Reviewed By: nicolasvasilache
Differential Revision: https://reviews.llvm.org/D100837
This revision migrates more code from Linalg into the new permanent home of
SparseTensor. It replaces the test passes with proper compiler passes.
NOTE: the actual removal of the last glue and clutter in Linalg will follow
Reviewed By: bixia
Differential Revision: https://reviews.llvm.org/D101811
TransferReadOps that are a scalar read + broadcast are handled by TransferReadToVectorLoadLowering.
Differential Revision: https://reviews.llvm.org/D101808
Given the source and destination shapes, if they are static, or if the
expanded/collapsed dimensions are unit-extent, it is possible to
compute the reassociation maps that can be used to reshape one type
into another. Add a utility method to return the reassociation maps
when possible.
This utility function can be used to fuse a sequence of reshape ops,
given the type of the source of the producer and the final result
type. This pattern supercedes a more constrained folding pattern added
to DropUnitDims pass.
Differential Revision: https://reviews.llvm.org/D101343
Convert subtensor and subtensor_insert operations to use their
rank-reduced versions to drop unit dimensions.
Differential Revision: https://reviews.llvm.org/D101495
The current implementation had a bug as it was relying on the target vector
dimension sizes to calculate where to insert broadcast. If several dimensions
have the same size we may insert the broadcast on the wrong dimension. The
correct broadcast cannot be inferred from the type of the source and
destination vector.
Instead when we want to extend transfer ops we calculate an "inverse" map to the
projected permutation and insert broadcast in place of the projected dimensions.
Differential Revision: https://reviews.llvm.org/D101738
Move TransposeOp lowering in its own populate function as in some cases
it is better to keep it during ContractOp lowering to better
canonicalize it rather than emiting scalar insert/extract.
Differential Revision: https://reviews.llvm.org/D101647
Added canonicalization for vector_load and vector_store. An existing
pattern SimplifyAffineOp can be reused to compose maps that supplies
result into them. Added AffineVectorStoreOp and AffineVectorLoadOp
into static_assert of SimplifyAffineOp to allow operation to use it.
This fixes the bug filed: https://bugs.llvm.org/show_bug.cgi?id=50058
Reviewed By: bondhugula
Differential Revision: https://reviews.llvm.org/D101691
(1) migrates the encoding from TensorDialect into the new SparseTensorDialect
(2) replaces dictionary-based storage and builders with struct-like data
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D101669
Three patterns are added to convert into vector.multi_reduction into a
sequence of vector.reduction as the following:
- Transpose the inputs so inner most dimensions are always reduction.
- Reduce rank of vector.multi_reduction into 2d with inner most
reduction dim (get the 2d canical form)
- 2D canonical form is converted into a sequence of vector.reduction.
There are two things we might worth in a follow up diff:
- An scf.for (maybe optionally) around vector.reduction instead of unrolling it.
- Breakdown the vector.reduction into a sequence of vector.reduction
(e.g tree-based reduction) instead of relying on how downstream dialects
handle it.
Note: this will requires passing target-vector-length
Differential Revision: https://reviews.llvm.org/D101570
This is the very first step toward removing the glue and clutter from linalg and
replace it with proper sparse tensor types. This revision migrates the LinalgSparseOps
into SparseTensorOps of a sparse tensor dialect. This also provides a new home for
sparse tensor related transformation.
NOTE: the actual replacement with sparse tensor types (and removal of linalg glue/clutter)
will follow but I am trying to keep the amount of changes per revision manageable.
Differential Revision: https://reviews.llvm.org/D101573
This is the very first step toward removing the glue and clutter from linalg and
replace it with proper sparse tensor types. This revision migrates the LinalgSparseOps
into SparseTensorOps of a sparse tensor dialect. This also provides a new home for
sparse tensor related transformation.
NOTE: the actual replacement with sparse tensor types (and removal of linalg glue/clutter)
will follow but I am trying to keep the amount of changes per revision manageable.
Reviewed By: bixia
Differential Revision: https://reviews.llvm.org/D101488
This enables to express more complex parallel loops in the affine framework,
for example, in cases of tiling by sizes not dividing loop trip counts perfectly
or inner wavefront parallelism, among others. One can't use affine.max/min
and supply values to the nested loop bounds since the results of such
affine.max/min operations aren't valid symbols. Making them valid symbols
isn't an option since they would introduce selection trees into memref
subscript arithmetic as an unintended and undesired consequence. Also
add support for converting such loops to SCF. Drop some API that isn't used in
the core repo from AffineParallelOp since its semantics becomes ambiguous in
presence of max/min bounds. Loop normalization is currently unavailable for
such loops.
Depends On D101171
Reviewed By: bondhugula
Differential Revision: https://reviews.llvm.org/D101172
Introduce a basic support for parallelizing affine loops with reductions
expressed using iteration arguments. Affine parallelism detector now has a flag
to assume such reductions are parallel. The transformation handles a subset of
parallel reductions that are can be expressed using affine.parallel:
integer/float addition and multiplication. This requires to detect the
reduction operation since affine.parallel only supports a fixed set of
reduction operators.
Reviewed By: chelini, kumasento, bondhugula
Differential Revision: https://reviews.llvm.org/D101171
FillOp allows complex ops, and filling a properly sized buffer with
a default zero complex number is implemented.
Differential Revision: https://reviews.llvm.org/D99939
This revision adds support for vectorizing more general linalg operations with projected permutation maps.
This is achieved by eagerly broadcasting the intermediate vector to the common size
of the iteration domain of the linalg op. This allows a much more natural expression of
generalized vectorization but may introduce additional computations until all the
proper canonicalizations are implemented.
This generalization modifies the vector.transfer_read/write permutation logic and
exposes the fact that the logic employed in vector.contract was too ad-hoc.
As a consequence, changes occur in the permutation / transposition logic for contraction. In turn this prompts supporting more cases in the lowering of contract
to matrix intrinsics, which is required to make the corresponding tests pass.
Differential revision: https://reviews.llvm.org/D101165
Canonicalizations for subtensor operations defaulted to use the
rank-reduced version of the operation, but the cast inserted to get
back the original type would be illegal if the rank was actually
reduced. Instead make the canonicalization not reduce the rank of the
operation.
Differential Revision: https://reviews.llvm.org/D101258
The current canonicalization did not remap operation results correctly
and attempted to erase tiledLoop, which is incorrect if not all tensor
results are folded.
Tensor inputs, if not used in the body of TiledLoopOp, can be removed.
memref::CastOp can be folded into TiledLoopOp as well.
Differential Revision: https://reviews.llvm.org/D101445
Both, `shape.broadcast` and `shape.cstr_broadcastable` accept dynamic and static
extent tensors. If their operands are casted, we can use the original value
instead.
Differential Revision: https://reviews.llvm.org/D101376
Empty extent tensor operands were only removed when they were defined as a
constant. Additionally, we can remove them if they are known to be empty by
their type `tensor<0xindex>`.
Differential Revision: https://reviews.llvm.org/D101351
Splat constant folding was limited to `std.constant` operations. Instead, use
the constant matcher and apply splat constant folding to any constant-like
operation that holds a splat attribute.
Differential Revision: https://reviews.llvm.org/D101301
The new "encoding" field in tensor types so far had no meaning. This revision introduces:
1. an encoding attribute interface in IR: for verification between tensors and encodings in general
2. an attribute in Tensor dialect; #tensor.sparse<dict> + concrete sparse tensors API
Active discussion:
https://llvm.discourse.group/t/rfc-introduce-a-sparse-tensor-type-to-core-mlir/2944/
Reviewed By: silvas, penpornk, bixia
Differential Revision: https://reviews.llvm.org/D101008
Add two canoncalizations for scf.if.
1) A canonicalization that allows users of a condition within an if to assume the condition
is true if in the true region, etc.
2) A canonicalization that removes yielded statements that are equivalent to the condition
or its negation
Differential Revision: https://reviews.llvm.org/D101012
Ensure to preserve the correct type during when folding and canonicalization.
`shape.broadcast` of of a single operand can only be folded away if the argument
type is correct.
Differential Revision: https://reviews.llvm.org/D101158
Eliminate empty shapes from the operands, partially fold all constant shape
operands, and fix normal folding.
Differential Revision: https://reviews.llvm.org/D100634
The interchange option attached to the linalg to loop lowering affects only the loops and does not update the memory accesses generated in to body of the operation. Instead of performing the interchange during the loop lowering use the interchange pattern.
Differential Revision: https://reviews.llvm.org/D100758
Example:
```
%0 = linalg.init_tensor : tensor<...>
%1 = linalg.generic ... outs(%0: tensor<...>)
%2 = linalg.generic ... outs(%0: tensor<...>)
```
Memref allocated as a result of `init_tensor` bufferization can be incorrectly overwritten by the second linalg.generic operation
Reviewed By: silvas
Differential Revision: https://reviews.llvm.org/D100921
This will prevent fusion that spains all dims and generates
(d0, d1, ...) -> () reshape that isn't legal
Differential Revision: https://reviews.llvm.org/D100805
Break up the dependency between SCF ops and substituteMin helper and make a
more generic version of AffineMinSCFCanonicalization. This reduce dependencies
between linalg and SCF and will allow the logic to be used with other kind of
ops. (Like ID ops).
Differential Revision: https://reviews.llvm.org/D100321
Previously, any terminator without ReturnLike and BranchOpInterface traits (e.g. scf.condition) were causing pass to fail.
Differential Revision: https://reviews.llvm.org/D100832
Instead of always running the region builder check if the generalized op has a region attached. If yes inline the existing region instead of calling the region builder. This change circumvents a problem with named operations that have a region builder taking captures and the generalization pass not knowing about this captures.
Differential Revision: https://reviews.llvm.org/D100880
The current implementation allows for TransferWriteOps with broadcasts that do not make sense. E.g., a broadcast could write a vector into a single (scalar) memory location, which is effectively the same as writing only the last element of the vector.
Differential Revision: https://reviews.llvm.org/D100842
The patch extends the vectorization pass to lower linalg index operations to vector code. It allocates constant 1d vectors that enumerate the indexes along the iteration dimensions and broadcasts/transposes these 1d vectors to the iteration space.
Differential Revision: https://reviews.llvm.org/D100373
This patch extends the control-flow cost-model for detensoring by
implementing a forward-looking pass on block arguments that should be
detensored. This makes sure that if a (to-be-detensored) block argument
"escapes" its block through the terminator, then the successor arguments
are also detensored.
Reviewed By: silvas
Differential Revision: https://reviews.llvm.org/D100457
The patch replaces the index operations in the body of fused producers and linearizes the indices after expansion.
Differential Revision: https://reviews.llvm.org/D100479
Update the dimensions of the index operations to account for dropped dimensions and replace the index operations of dropped dimensions by zero.
Differential Revision: https://reviews.llvm.org/D100395
This patch add the UnnamedAddr attribute for the GlobalOp in the LLVM
dialect. The attribute is also handled to and from LLVM IR.
This is meant to be used in a follow up patch to lower OpenACC/OpenMP ops to
call to kmp and tgt runtime calls (D100678).
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D100677
Instead of interchanging loops during the loop lowering this pass performs the interchange by permuting the indexing maps. It also updates the iterator types and the index accesses in the body of the operation.
Differential Revision: https://reviews.llvm.org/D100627
Move the existing optimization for transfer op on tensor to folder and
canonicalization. This handles the write after write case and read after write
and also add write after read case.
Differential Revision: https://reviews.llvm.org/D100597
ArmSVE dialect is behind the recent changes in how the Vector dialect
interacts with backend vector dialects and the MLIR -> LLVM IR
translation module. This patch cleans up ArmSVE initialization within
Vector and removes the need for an LLVMArmSVE dialect.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D100171
In the long run, we want to unify the dot product codegen solutions between
all target architectures, but this intrinsic enables experimenting with AVX
specific implementations in the meantime.
Reviewed By: nicolasvasilache
Differential Revision: https://reviews.llvm.org/D100593
Rationale:
Now that vector<?xindex> is allowed, the restriction on vectorization
of index types in the sparse compiler can be removed. Also needs
generalization of scatter/gather index types.
Reviewed By: gysit
Differential Revision: https://reviews.llvm.org/D100522
This patch collects operations that have users in a for loop and uses
them when loop invariant operations are detected and hoisted.
Reviewed By: bondhugula, vinayaka-polymage
Differential Revision: https://reviews.llvm.org/D99761
Per the SPIR-V spec "2.16.2. Validation Rules for Shader Capabilities":
Composite objects in the StorageBuffer, PhysicalStorageBuffer,
Uniform, and PushConstant Storage Classes must be explicitly
laid out.
For other cases we don't need to attach the struct offsets.
Reviewed By: hanchung
Differential Revision: https://reviews.llvm.org/D100386
The patch updates the tiling pass to add the tile offsets to the indices returned by the linalg operations.
Differential Revision: https://reviews.llvm.org/D100379
The patch extends the linalg to loop lowering pass to replace all linalg index operations by the induction variables of the generated loop nests.
Differential Revision: https://reviews.llvm.org/D100364
This patch introduces the neccessary infrastructure changes to implement
cost-modelling for detensoring. In particular, it introduces the
following changes:
- An extension to the dialect conversion framework to selectively
convert sub-set of non-entry BB arguments.
- An extension to branch conversion pattern to selectively convert
sub-set of a branche's operands.
- An interface for detensoring cost-modelling.
- 2 simple implementations of 2 different cost models.
This sets the stage to explose cost-modelling for detessoring in an
easier way. We still need to come up with better cost models.
Reviewed By: silvas
Differential Revision: https://reviews.llvm.org/D99945
Depends On D95311
Previous automatic-ref-counting pass worked with high level async operations (e.g. async.execute), however async values reference counting is a runtime implementation detail.
New pass mostly relies on the save liveness analysis to place drop_ref operations, and does better verification of CFG with different liveIn sets in block successors.
This is almost NFC change. No new reference counting ideas, just a cleanup of the previous version.
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D95390
This is similar to the definition of llvm.switch, providing
unstructured branch-based control flow. It differs from the LLVM
operation in that it accepts any signless integer (not only an i32),
takes no branch weights (the same as the Branch and CondBranch ops),
and has a slightly different syntax for the default case that includes
it in the list of cases with an explicit `default` keyword.
Also included are several canonicalizers.
See https://llvm.discourse.group/t/rfc-add-std-switch-and-scf-switch/3090
Reviewed By: rriddle, bondhugula
Differential Revision: https://reviews.llvm.org/D99925
The stride should be calculated with the converted array element
type, not the original input type.
Reviewed By: mravishankar
Differential Revision: https://reviews.llvm.org/D100337
These patterns have been used as a prerequisite step for lowering
to SPIR-V. But they don't involve SPIR-V dialect ops; they are
pure memref/vector op transformations. Given now we have a dedicated
MemRef dialect, moving them to Memref/Transforms/, which is a more
suitable place to host them, to allow used by others.
This commit just moves code around and renames patterns/passes
accordingly. CMakeLists.txt for existing MemRef libraries are
also improved along the way.
Reviewed By: mravishankar
Differential Revision: https://reviews.llvm.org/D100326
We will soon be adding non-AVX512 operations to MLIR, such as AVX's rsqrt. In https://reviews.llvm.org/D99818 several possibilities were discussed, namely to (1) add non-AVX512 ops to the AVX512 dialect, (2) add more dialects (e.g. AVX dialect for AVX rsqrt), and (3) expand the scope of the AVX512 to include these SIMD x86 ops, thereby renaming the dialect to something more accurate such as X86Vector.
Consensus was reached on option (3), which this patch implements.
Reviewed By: aartbik, ftynse, nicolasvasilache
Differential Revision: https://reviews.llvm.org/D100119
Fusing a constant with a linalg.generic operation can result in the
fused operation being illegal since the loop bound computation
fails. Avoid such fusions.
Differential Revision: https://reviews.llvm.org/D100272
The `linalg.index` operation provides access to the iteration indexes of immediately enclosing linalg operations. It takes a dimension `dim` attribute and returns the iteration index in the given dimension. Having `linalg.index` allows us to unify `linalg.generic` and `linalg.indexed_generic` and also enables index access in named operations.
Differential Revision: https://reviews.llvm.org/D100292
Recent change enable dropping unit-trip loops of "reduction" iterator
type as well. This is fine as long as there is one other "reduction"
iterator in the operation. Without this the initialized value (value
of `out`) is not read which leads to a correctness issue.
Also fix a bug in the `fill` -> `tensor_reshape` folding. The `out`
operand of the `fill` needs to be reshaped to get the `out` operand of
the generated `fill` operation.
Differential Revision: https://reviews.llvm.org/D100145
This patch doesn't support the optional operands of ImageDrefGather. The support of optional operands will be implemented later.
co-authered-by: Alan Liu <alanliu.yf@gmail.com>
Differential Revision: https://reviews.llvm.org/D100128
Alexander Belyaev