This PR adds "value casting", i.e., a mechanism to wrap `ir.Value` in a
proxy class that overloads dunders such as `__add__`, `__sub__`, and
`__mul__` for fun and great profit.
This is thematically similar to
bfb1ba7526
and
9566ee2806.
The example in the test demonstrates the value of the feature (no pun
intended):
```python
@register_value_caster(F16Type.static_typeid)
@register_value_caster(F32Type.static_typeid)
@register_value_caster(F64Type.static_typeid)
@register_value_caster(IntegerType.static_typeid)
class ArithValue(Value):
__add__ = partialmethod(_binary_op, op="add")
__sub__ = partialmethod(_binary_op, op="sub")
__mul__ = partialmethod(_binary_op, op="mul")
a = arith.constant(value=FloatAttr.get(f16_t, 42.42))
b = a + a
# CHECK: ArithValue(%0 = arith.addf %cst, %cst : f16)
print(b)
a = arith.constant(value=FloatAttr.get(f32_t, 42.42))
b = a - a
# CHECK: ArithValue(%1 = arith.subf %cst_0, %cst_0 : f32)
print(b)
a = arith.constant(value=FloatAttr.get(f64_t, 42.42))
b = a * a
# CHECK: ArithValue(%2 = arith.mulf %cst_1, %cst_1 : f64)
print(b)
```
**EDIT**: this now goes through the bindings and thus supports automatic
casting of `OpResult` (including as an element of `OpResultList`),
`BlockArgument` (including as an element of `BlockArgumentList`), as
well as `Value`.
The scalable dimension functionality was added to the vector type after
the bindings for it were defined, without the bindings being ever
updated. Fix that.
<img
src="https://github.com/llvm/llvm-project/assets/5657668/443852b6-ac25-45bb-a38b-5dfbda09d5a7"
height="400" />
<p></p>
So turns out that none of the `replace=True` things actually work
because of the map caches (except for
`register_attribute_builder(replace=True)`, which doesn't use such a
cache). This was hidden by a series of unfortunate events:
1. `register_type_caster` failure was hidden because it was the same
`TestIntegerRankedTensorType` being replaced with itself (d'oh).
2. `register_operation` failure was hidden behind the "order of events"
in the lifecycle of typical extension import/use. Since extensions are
loaded/registered almost immediately after generated builders are
registered, there is no opportunity for the `operationClassMapCache` to
be populated (through e.g., `module.body.operations[2]` or
`module.body.operations[2].opview` or something). Of course as soon as
you as actually do "late-bind/late-register" the extension, you see it's
not successfully replacing the stale one in `operationClassMapCache`.
I'll take this opportunity to propose we ditch the caches all together.
I've been cargo-culting them but I really don't understand how they
work. There's this comment above `operationClassMapCache`
```cpp
/// Cache of operation name to external operation class object. This is
/// maintained on lookup as a shadow of operationClassMap in order for repeat
/// lookups of the classes to only incur the cost of one hashtable lookup.
llvm::StringMap<pybind11::object> operationClassMapCache;
```
But I don't understand how that's true given that the canonical thing
`operationClassMap` is already a map:
```cpp
/// Map of full operation name to external operation class object.
llvm::StringMap<pybind11::object> operationClassMap;
```
Maybe it wasn't always the case? Anyway things work now but it seems
like an unnecessary layer of complexity for not much gain? But maybe I'm
wrong.
Added missing register_translations in python to replicate the same in
the C-API
Cleaned up the current calls to register passes where the other calls
are already embedded in the mlirRegisterAllPasses.
found here,
https://discourse.llvm.org/t/opencl-example/74187
Changes:
1. For both dimToLvl and lvlToDim, always returns the actual map instead
of AffineMap() for identity map.
2. Updated custom builder for encoding to have default values.
3. Non-inferable lvlToDim will still return AffineMap() during
inference, so it will be caught by verifier.
`PyOperations` are Python-level handles to `Operation *` instances. When
the latter are modified by C++, the former need to be invalidated.
#69746 implements such invalidation mechanism by setting all
`PyReferences` to `invalid`. However, that is not enough: they also need
to be removed from the `liveOperations` map since other parts of the
code (such as `PyOperation::createDetached`) assume that that map only
contains valid refs.
This is required to actually solve the issue in #69730.
Fixes https://github.com/llvm/llvm-project/issues/69730 (also see
https://reviews.llvm.org/D155543).
There are two things outstanding (why I didn't land before):
1. add some C API tests for `mlirOperationWalk`;
2. potentially refactor how the invalidation in `run` works; the first
version of the code looked like this:
```cpp
if (invalidateOps) {
auto *context = op.getOperation().getContext().get();
MlirOperationWalkCallback invalidatingCallback =
[](MlirOperation op, void *userData) {
PyMlirContext *context =
static_cast<PyMlirContext *>(userData);
context->setOperationInvalid(op);
};
auto numRegions =
mlirOperationGetNumRegions(op.getOperation().get());
for (int i = 0; i < numRegions; ++i) {
MlirRegion region =
mlirOperationGetRegion(op.getOperation().get(), i);
for (MlirBlock block = mlirRegionGetFirstBlock(region);
!mlirBlockIsNull(block);
block = mlirBlockGetNextInRegion(block))
for (MlirOperation childOp =
mlirBlockGetFirstOperation(block);
!mlirOperationIsNull(childOp);
childOp = mlirOperationGetNextInBlock(childOp))
mlirOperationWalk(childOp, invalidatingCallback, context,
MlirWalkPostOrder);
}
}
```
This is verbose and ugly but it has the important benefit of not
executing `mlirOperationEqual(rootOp->get(), op)` for every op
underneath the root op.
Supposing there's no desire for the slightly more efficient but highly
convoluted approach, I can land this "posthaste".
But, since we have eyes on this now, any suggestions or approaches (or
needs/concerns) are welcome.
Currently, `linalg.transpose` and `linalg.broadcast` can't be emitted
through either the C API or the python bindings (which of course go
through the C API). See
https://discourse.llvm.org/t/how-to-build-linalg-transposeop-in-mlir-pybind/73989/10.
The reason is even though they're named ops, there is no opdsl
`@linalg_structured_op` for them and thus while they can be instantiated
they cannot be passed to
[`mlirLinalgFillBuiltinNamedOpRegion`](a7cccb9cbb/mlir/lib/CAPI/Dialect/Linalg.cpp (L18)).
I believe the issue is they both take a `IndexAttrDef` but
`IndexAttrDef` cannot represent dynamic rank. Note, if I'm mistaken and
there is a way to write the `@linalg_structured_op` let me know.
The solution here simply implements the `regionBuilder` interface which
is then picked up by
[`LinalgDialect::addNamedOpBuilders`](7557530f42/mlir/lib/Dialect/Linalg/IR/LinalgDialect.cpp (L116)).
Extension classes are added "by hand" that mirror the API of the
`@linalg_structured_op`s. Note, the extension classes are added to to
`dialects/linalg/__init__.py` instead of
`dialects/linalg/opdsl/ops/core_named_ops.py` in order that they're not
confused for opdsl generators/emitters.
This PR replaces the mixin `OpView` extension mechanism with the
standard inheritance mechanism.
Why? Firstly, mixins are not very pythonic (inheritance is usually used
for this), a little convoluted, and too "tight" (can only be used in the
immediately adjacent `_ext.py`). Secondly, it (mixins) are now blocking
are correct implementation of "value builders" (see
[here](https://github.com/llvm/llvm-project/pull/68764)) where the
problem becomes how to choose the correct base class that the value
builder should call.
This PR looks big/complicated but appearances are deceiving; 4 things
were needed to make this work:
1. Drop `skipDefaultBuilders` in
`OpPythonBindingGen::emitDefaultOpBuilders`
2. Former mixin extension classes are converted to inherit from the
generated `OpView` instead of being "mixins"
a. extension classes that simply were calling into an already generated
`super().__init__` continue to do so
b. (almost all) extension classes that were calling `self.build_generic`
because of a lack of default builder being generated can now also just
call `super().__init__`
3. To handle the [lone single
use-case](https://sourcegraph.com/search?q=context%3Aglobal+select_opview_mixin&patternType=standard&sm=1&groupBy=repo)
of `select_opview_mixin`, namely
[linalg](https://github.com/llvm/llvm-project/blob/main/mlir/python/mlir/dialects/_linalg_ops_ext.py#L38),
only a small change was necessary in `opdsl/lang/emitter.py` (thanks to
the emission/generation of default builders/`__init__`s)
4. since the `extend_opview_class` decorator is removed, we need a way
to register extension classes as the desired `OpView` that `op.opview`
conjures into existence; so we do the standard thing and just enable
replacing the existing registered `OpView` i.e.,
`register_operation(_Dialect, replace=True)`.
Note, the upgrade path for the common case is to change an extension to
inherit from the generated builder and decorate it with
`register_operation(_Dialect, replace=True)`. In the slightly more
complicated case where `super().__init(self.build_generic(...))` is
called in the extension's `__init__`, this needs to be updated to call
`__init__` in `OpView`, i.e., the grandparent (see updated docs).
Note, also `<DIALECT>_ext.py` files/modules will no longer be automatically loaded.
Note, the PR has 3 base commits that look funny but this was done for
the purpose of tracking the line history of moving the
`<DIALECT>_ops_ext.py` class into `<DIALECT>.py` and updating (commit
labeled "fix").
The reason I want this is that I am writing my own Python bindings and
would like to use the insertion point from
`PyThreadContextEntry::getDefaultInsertionPoint()` to call C++ functions
that take an `OpBuilder` (I don't need to expose it in Python but it
also seems appropriate). AFAICT, there is currently no way to translate
a `PyInsertionPoint` into an `OpBuilder` because the operation is
inaccessible.
Updates:
1. Infer lvlToDim from dimToLvl
2. Add more tests for block sparsity
3. Finish TODOs related to lvlToDim, including adding lvlToDim to python
binding
Verification of lvlToDim that user provides will be implemented in the
next PR.
This makes these match the behaviour of optional attributes (which are
omitted when they are their default value of none). This allows for
concise assembly formats without a custom printer.
An extra print of " " is also removed, this does change any existing
uses of oilists, but if the parameter before the oilist is optional,
that would previously add an extra space.
This #68694 + some fixes for the MLIR Python tests, unfortunately GitHub
does not allow re-opening PRs 😕
This PR creates the necessary files to support bindings for operations
in the affine dialect.
This is the first of many PRs which will progressively introduce
affine.load, affine.for, etc operations. I would like to
acknowledge the work by Nelli's author @makslevental :
https://github.com/makslevental/nelli/blob/main/nelli/mlir/affine/affine.py
which jump-starts the work.
This function has several overloads that allow to specify the symbol
that should be renamed and the scope for that renaming in different
ways. The overloads were inconsistent in the following way (quoted
strings are `StringAttr`s, other variables are `Operation *`):
* `replaceAllSymbolUses(symbolOp, "new_symbol", scopeOp)` would traverse
into the nested regions of `scopeOp` and hence rename the symbol inside
of `scopeOp`.
* `replaceAllSymbolUses("symbol", "new_symbol", scopeOp)` would *not*
traverse into the nested regions of `scopeOp` and hence *not* rename the
symbol.
The underlying behavior was spread over different places and is somewhat
hard to understand. The two overloads above mainly differed by what
`collectSymbolScopes` computed, which is itself overloaded. If `scopeOp`
is a top-level module, then the overload on `(Operation *, Operation
*)`, which is used in the first of the above cases, computes a scope
where the body region of the module is the `limit`; however, the
overload on `(StringAttr, Operation *)` computed the module op itself as
the `limit`. Later, `walkSymbolTable` would walk the body of the module
if it was given as a region but it would *not* enter the regions of the
module op because that op has a symbol table (which was assumed to be a
*different* scope).
The fix in this commit is change the behavior of `collectSymbolScopes`
such that the `(StringAttr, Operation *)` overload returns a scope for
each region in the `limit` argument.
This PR adds the additional generation of what I'm calling "value
builders" (a term I'm not married to) that look like this:
```python
def empty(sizes, element_type, *, loc=None, ip=None):
return get_result_or_results(tensor.EmptyOp(sizes=sizes, element_type=element_type, loc=loc, ip=ip))
```
which instantiates a `tensor.EmptyOp` and then immediately grabs the
result (`OpResult`) and then returns that *instead of a handle to the
op*.
What's the point of adding these when `EmptyOp.result` already exists?
My claim/feeling/intuition is that eDSL users are more comfortable with
a value centric programming model (i.e., passing values as operands) as
opposed to an operator instantiation programming model. Thus this change
enables (or at least goes towards) the bindings supporting such a user
and use case. For example,
```python
i32 = IntegerType.get_signless(32)
...
ten1 = tensor.empty((10, 10), i32)
ten2 = tensor.empty((10, 10), i32)
ten3 = arith.addi(ten1, ten2)
```
Note, in order to present a "pythonic" API and enable "pythonic" eDSLs,
the generated identifiers (op names and operand names) are snake case
instead of camel case and thus `llvm::convertToSnakeFromCamelCase`
needed a small fix. Thus this PR is stacked on top of
https://github.com/llvm/llvm-project/pull/68375.
In addition, as a kind of victory lap, this PR adds a "rangefor" that
looks and acts exactly like python's `range` but emits `scf.for`.
This patch updates `transform.loop.peel` so that this Op returns two
rather than one handle:
* one for the peeled loop, and
* one for the remainder loop.
Also, following this change this Op will fail if peeling fails. This is
consistent with other similar Ops that also fail if no transformation
takes place.
Relands #67482 with an extra fix for transform_loop_ext.py
Rename and restructure tiling-related transform ops from the structured
extension to be more homogeneous. In particular, all ops now follow a
consistent naming scheme:
- `transform.structured.tile_using_for`;
- `transform.structured.tile_using_forall`;
- `transform.structured.tile_reduction_using_for`;
- `transform.structured.tile_reduction_using_forall`.
This drops the "_op" naming artifact from `tile_to_forall_op` that
shouldn't have been included in the first place, consistently specifies
the name of the control flow op to be produced for loops (instead of
`tile_reduction_using_scf` since `scf.forall` also belongs to `scf`),
and opts for the `using` connector to avoid ambiguity.
The loops produced by tiling are now systematically placed as *trailing*
results of the transform op. While this required changing 3 out of 4 ops
(except for `tile_using_for`), this is the only choice that makes sense
when producing multiple `scf.for` ops that can be associated with a
variadic number of handles. This choice is also most consistent with
*other* transform ops from the structured extension, in particular with
fusion ops, that produce the structured op as the leading result and the
loop as the trailing result.
This PR adds a new transform op that replaces `memref.alloca`s with
`memref.get_global`s to newly inserted `memref.global`s. This is useful,
for example, for allocations that should reside in the shared memory of
a GPU, which have to be declared as globals.
This PR renames the vectorization transform ops as follows:
* `structured.masked_vectorize` => `structured.vectorize`. This reflects
the fact that since [recently](https://reviews.llvm.org/D157774) the op
can also handle the unmasked case.
* `structured.vectorize` =>
`structured.vectorize_children_and_applies_patterns`. This reflects the
fact that the op does not just vectorize the given payload op but all
vectorizable children contained in it, and applies patterns before and
after for preparation and clean-up.
This rename was discussed first
[here](https://reviews.llvm.org/D157774).
The PR also adapts and cleans ups the tablegen description of the
`VectorizeChildrenAndApplyPatternsOp` (formerly `VectorizeOp`).
This does basic plumbing, ideally want a context approach to reduce
needing to thread these manually, but the current is useful even in that
state.
Made Value.get_name change backwards compatible, so one could either set
a field or create a state to pass in.
This commit removes the deallocation capabilities of
one-shot-bufferization. One-shot-bufferization should never deallocate
any memrefs as this should be entirely handled by the
ownership-based-buffer-deallocation pass going forward. This means the
`allow-return-allocs` pass option will default to true now,
`create-deallocs` defaults to false and they, as well as the escape
attribute indicating whether a memref escapes the current region, will
be removed. A new `allow-return-allocs-from-loops` option is added as a
temporary workaround for some bufferization limitations.
This PR cleans up the test of the mix-ins of this dialect. Most of the
character diff is due to factoring out the creation of the the top-level
sequence into a decorator. This decorator siginficantly shortens the
definition of the individual tests and can be used in all but one test,
where the top-level op is a PDL op. The only functional diff is due to
the fact that the decator uses `transform.any_op` instead of
`pdl.operation` for the type of the root handle. The only remaining
usages of the PDL dialects is now in the test a PDL-related op.
Only construction and type casting are implemented. The method to create
is explicitly named "unsafe" and the documentation calls out what the
caller is responsible for. There really isn't a better way to do this
and retain the power-user feature this represents.
This is the first commit in a series with the goal to rework the
BufferDeallocation pass. Currently, this pass heavily relies on copies
to perform correct deallocations, which leads to very slow code and
potentially high memory usage. Additionally, there are unsupported cases
such as returning memrefs which this series of commits aims to add
support for as well.
This first commit removes the deallocation capabilities of
one-shot-bufferization.One-shot-bufferization should never deallocate any
memrefs as this should be entirely handled by the buffer-deallocation pass
going forward. This means the allow-return-allocs pass option will
default to true now, create-deallocs defaults to false and they, as well
as the escape attribute indicating whether a memref escapes the current region,
will be removed.
The documentation should w.r.t. these pass option changes should also be
updated in this commit.
Reviewed By: springerm
Differential Revision: https://reviews.llvm.org/D156662
This patch is part of a larger initiative aimed at fixing floating-point `max` and `min` operations in MLIR: https://discourse.llvm.org/t/rfc-fix-floating-point-max-and-min-operations-in-mlir/72671.
This commit addresses Task 1.2 of the mentioned RFC. By renaming these operations, we align their names with LLVM intrinsics that have corresponding semantics.
This patch adds attribute builders for all buildable attributes from the
builtin dialect that did not previously have any. These builders can be
used to construct attributes of a particular type identified by a string
from a Python argument without knowing the details of how to pass that
Python argument to the attribute constructor. This is used, for example,
in the generated code of the Python bindings of ops.
The list of "all" attributes was produced with:
(
grep -h "ods_ir.AttrBuilder.get" $(find ../build/ -name "*_ops_gen.py") \
| cut -f2 -d"'"
git grep -ho "^def [a-zA-Z0-9_]*" -- include/mlir/IR/CommonAttrConstraints.td \
| cut -f2 -d" "
) | sort -u
Then, I only retained those that had an occurence in
`mlir/include/mlir/IR`. In particular, this drops many dialect-specific
attributes; registering those builders is something that those dialects
should do. Finally, I removed those attrbiutes that had a match in
`mlir/python/mlir/ir.py` already and implemented the remaining ones. The
only ones that still miss a builder now are the following:
* Represent more than one possible attribute type:
- `Any.*Attr` (9x)
- `IntNonNegative`
- `IntPositive`
- `IsNullAttr`
- `ElementsAttr`
* I am not sure what "constant attributes" are:
- `ConstBoolAttrFalse`
- `ConstBoolAttrTrue`
- `ConstUnitAttr`
* `Location` not exposed by Python bindings:
- `LocationArrayAttr`
- `LocationAttr`
* `get` function not implemented in Python bindings:
- `StringElementsAttr`
This patch also fixes a compilation problem with
`I64SmallVectorArrayAttr`.
Reviewed By: makslevental, rkayaith
Differential Revision: https://reviews.llvm.org/D159403
Memref descriptors contain an `offset` field that denotes the start of
the content of the memref relative to the `alignedPtr`. This offset is
not considered when converting a memref descriptor to a np.array in the
Python runtime library, essentially treating all memrefs as if they had
an offset of zero. This patch introduces the necessary pointer arithmetic
to find the actual beginning of the memref contents to the memref->numpy
conversion functions.
There is an ongoing discussion about whether the `offset` field is needed
at all in the memref descriptor.
Until that is decided, the Python runtime and CRunnerUtils should
still correctly implement the offset handling.
Related: https://reviews.llvm.org/D157008
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D158494
The mix-in of the `MultiTileSizesOp` set the default value of its
`divisor` argument. This repeats information from the tablegen
defintion, is not necessary (since the generic code deals with `None`
and default values), and has the risk of running out of sync without
people noticing. This patch removes the setting of the value and forward
`None` to the generic constructor instead.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D159416
This patch simplifies and improves the mix-in of the `TileOp`. In
particular:
* Accept all types of sizes (static, dynamic, scalable) in a single
argument `sizes`.
* Use the existing convenience function to dispatch different types of
sizes instead of repeating the implementation in the mix-in.
* Pass on `None` values as is of optional arguments to the init function
of the super class.
* Reformat with default indentation width (4 spaces vs 2 spaces).
* Add a a test for providing scalable sizes.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D159417
That commit changed the mix-ins for the Python bindings of the PadOp
including some tests, but did not change the corresponding `CHECK`
statements. This patch does that.
The mix-in did not allow to *not* set many of the arguments, even though
they represent optional attributes. Instead, it set default values,
which have different semantics in some cases. In other cases, setting
the default values is already done by the C++ layer, in which case they
are currently redundant and may be wrong in some potential future change
in the TD or C++ files. With this patch, `None` is preserved until the
generated binding, which handles them as desired.
Reviewed By: springerm
Differential Revision: https://reviews.llvm.org/D158844
