A performance issue was reported due to the usage of NestedMatcher in
ComposeAffineMaps. The main culprit was the ubiquitous copies that were
occuring when appending even a single element in `matchOne`.
This CL generally simplifies the implementation and removes one level of indirection by getting rid of
auxiliary storage as well as simplifying the API.
The users of the API are updated accordingly.
The implementation was tested on a heavily unrolled example with
ComposeAffineMaps and is now close in performance with an implementation based
on stateless InstWalker.
As a reminder, the whole ComposeAffineMaps pass is slated to disappear but the bug report was very useful as a stress test for NestedMatchers.
Lastly, the following cleanups reported by @aminim were addressed:
1. make NestedPatternContext scoped within runFunction rather than at the Pass level. This was caused by a previous misunderstanding of Pass lifetime;
2. use defensive assertions in the constructor of NestedPatternContext to make it clear a unique such locally scoped context is allowed to exist.
PiperOrigin-RevId: 231781279
This CL addresses some cleanups that were leftover after an incorrect rebase:
1. use StringSwitch
2. use // NOLINTNEXTLINE
3. remove a dead line of code
PiperOrigin-RevId: 231726640
a trivial inst walker :-) (reduces pass time from several minutes non-terminating to 120ms) - (fixes b/123541184)
- use a simple 7-line inst walker to collect affine_apply op's instead of the nested
matcher; -compose-affine-maps pass runs in 120ms now instead of 5 minutes + (non-
terminating / out of memory) - on a realistic test case that is 20,000 lines 12-d
loop nest
- this CL is also pushing for simple existing/standard patterns unless there
is a real efficiency issue (OTOH, fixing nested matcher to address this issue requires
cl/231400521)
- the improvement is from swapping out the nested walker as opposed to from a bug
or anything else that this CL changes
- update stale comment
PiperOrigin-RevId: 231623619
This CL mandated TypeConstraint and Type to provide descriptions and fixed
various subclasses and definitions to provide so. The purpose is to enforce
good documentation; using empty string as the default just invites oversight.
PiperOrigin-RevId: 231579629
Similar to op operands and attributes, use DAG to specify operation's results.
This will allow us to provide names and matchers for outputs.
Also Defined `outs` as a marker to indicate the start of op result list.
PiperOrigin-RevId: 231422455
This CL also introduces a set of python bindings using pybind11. The bindings
are exercised using a `test_py2andpy3.py` test suite that works for both
python 2 and 3.
`test_py3.py` on the other hand uses the more idiomatic,
python 3 only "PEP 3132 -- Extended Iterable Unpacking" to implement a rank
and type-agnostic copy with transposition.
Because python assignment is by reference, we cannot easily make the
assignment operator use the same type of sugaring as in C++; i.e. the
following:
```cpp
Stmt block = edsc::Block({
For(ivs, zeros, shapeA, ones, {
C[ivs] = IA[ivs] + IB[ivs]
})});
```
has no equivalent in the native Python EDSCs at this time.
However, the sugaring can be built as a simple DSL in python and is left as
future work.
PiperOrigin-RevId: 231337667
This CL adds support for calling EDSCs from other languages than C++.
Following the LLVM convention this CL:
1. declares simple opaque types and a C API in mlir-c/Core.h;
2. defines the implementation directly in lib/EDSC/Types.cpp and
lib/EDSC/MLIREmitter.cpp.
Unlike LLVM however the nomenclature for these types and API functions is not
well-defined, naming suggestions are most welcome.
To avoid the need for conversion functions, Types.h and MLIREmitter.h include
mlir-c/Core.h and provide constructors and conversion operators between the
mlir::edsc type and the corresponding C type.
In this first commit, mlir-c/Core.h only contains the types for the C API
to allow EDSCs to work from Python. This includes both a minimal set of core
MLIR
types (mlir_context_t, mlir_type_t, mlir_func_t) as well as the EDSC types
(edsc_mlir_emitter_t, edsc_expr_t, edsc_stmt_t, edsc_indexed_t). This can be
restructured in the future as concrete needs arise.
For now, the API only supports:
1. scalar types;
2. memrefs of scalar types with static or symbolic shapes;
3. functions with input and output of these types.
The C API is not complete wrt ownership semantics. This is in large part due
to the fact that python bindings are written with Pybind11 which allows very
idiomatic C++ bindings. An effort is made to write a large chunk of these
bindings using the C API but some C++isms are used where the design benefits
from this simplication. A fully isolated C API will make more sense once we
also integrate with another language like Swift and have enough use cases to
drive the design.
Lastly, this CL also fixes a bug in mlir::ExecutionEngine were the order of
declaration of llvmContext and the JIT result in an improper order of
destructors (which used to crash before the fix).
PiperOrigin-RevId: 231290250
Similar to other tblgen:: abstractions, tblgen::Pattern hides the native TableGen
API and provides a nicer API that is more coherent with the TableGen definitions.
PiperOrigin-RevId: 231285143
* Matching an attribute and specifying a attribute constraint is the same thing executionally, so represent it such.
* Extract AttrConstraint helper to match TypeConstraint and use that where mAttr was previously used in RewriterGen.
PiperOrigin-RevId: 231213580
Cleanup a usage of functional::map that is deemed too obscure in
`reindexAffineIndices`. Also fix a stale comment in `reindexAffineIndices`.
PiperOrigin-RevId: 231211184
Addresses b/122486036
This CL addresses some leftover crumbs in AffineMap and IntegerSet by removing
the Null method and cleaning up the constructors.
As the ::Null uses were tracked down, opportunities appeared to untangle some
of the Parsing logic and make it explicit where AffineMap/IntegerSet have
ambiguous syntax. Previously, ambiguous cases were hidden behind the implicit
pointer values of AffineMap* and IntegerSet* that were passed as function
parameters. Depending the values of those pointers one of 3 behaviors could
occur.
This parsing logic convolution is one of the rare cases where I would advocate
for code duplication. The more proper fix would be to make the syntax
unambiguous or to allow some lookahead.
PiperOrigin-RevId: 231058512
This CL follows up on a memory leak issue related to SmallVector growth that
escapes the BumpPtrAllocator.
The fix is to properly use ArrayRef and placement new to define away the
issue.
The following renaming is also applied:
1. MLFunctionMatcher -> NestedPattern
2. MLFunctionMatches -> NestedMatch
As a consequence all allocations are now guaranteed to live on the BumpPtrAllocator.
PiperOrigin-RevId: 231047766
Example:
dma-generate options:
-dma-fast-mem-capacity - Set fast memory space ...
-dma-fast-mem-space=<uint> - Set fast memory space ...
loop-fusion options:
-fusion-compute-tolerance=<number> - Fractional increase in ...
-fusion-maximal - Enables maximal loop fusion
loop-tile options:
-tile-size=<uint> - Use this tile size for ...
loop-unroll options:
-unroll-factor=<uint> - Use this unroll factor ...
-unroll-full - Fully unroll loops
-unroll-full-threshold=<uint> - Unroll all loops with ...
-unroll-num-reps=<uint> - Unroll innermost loops ...
loop-unroll-jam options:
-unroll-jam-factor=<uint> - Use this unroll jam factor ...
PiperOrigin-RevId: 231019363
Use `-mlir-pretty-debuginfo` if the user wants line breaks between different callsite lines.
The print results before and after this CL are shown in the tests.
PiperOrigin-RevId: 231013812
index remapping
- generate a sequence of single result affine_apply's for the index remapping
(instead of one multi result affine_apply)
- update dma-generate and loop-fusion test cases; while on this, change test cases
to use single result affine apply ops
- some fusion comment fix/cleanup
PiperOrigin-RevId: 230985830
- Update createAffineComputationSlice to generate a sequence of single result
affine apply ops instead of one multi-result affine apply
- update pipeline-data-transfer test case; while on this, also update the test
case to use only single result affine maps, and make it more robust to
change.
PiperOrigin-RevId: 230965478
This commit introduces a generic dialect conversion/lowering/legalization pass
and illustrates it on StandardOps->LLVMIR conversion.
It partially reuses the PatternRewriter infrastructure and adds the following
functionality:
- an actual pass;
- non-default pattern constructors;
- one-to-many rewrites;
- rewriting terminators with successors;
- not applying patterns iteratively (unlike the existing greedy rewrite driver);
- ability to change function signature;
- ability to change basic block argument types.
The latter two things required, given the existing API, to create new functions
in the same module. Eventually, this should converge with the rest of
PatternRewriter. However, we may want to keep two pass versions: "heavy" with
function/block argument conversion and "light" that only touches operations.
This pass creates new functions within a module as a means to change function
signature, then creates new blocks with converted argument types in the new
function. Then, it traverses the CFG in DFS-preorder to make sure defs are
converted before uses in the dominated blocks. The generic pass has a minimal
interface with two hooks: one to fill in the set of patterns, and another one
to convert types for functions and blocks. The patterns are defined as
separate classes that can be table-generated in the future.
The LLVM IR lowering pass partially inherits from the existing LLVM IR
translator, in particular for type conversion. It defines a conversion pattern
template, instantiated for different operations, and is a good candidate for
tablegen. The lowering does not yet support loads and stores and is not
connected to the translator as it would have broken the existing flows. Future
patches will add missing support before switching the translator in a single
patch.
PiperOrigin-RevId: 230951202
This implements a simple CPU runner based on LLVM Orc JIT. The base
functionality is provided by the ExecutionEngine class that compiles and links
the module, and provides an interface for obtaining function pointers to the
JIT-compiled MLIR functions and for invoking those functions directly. Since
function pointers need to be casted to the correct pointer type, the
ExecutionEngine wraps LLVM IR functions obtained from MLIR into a helper
function with the common signature `void (void **)` where the single argument
is interpreted as a list of pointers to the actual arguments passed to the
function, eventually followed by a pointer to the result of the function.
Additionally, the ExecutionEngine is set up to resolve library functions to
those available in the current process, enabling support for, e.g., simple C
library calls.
For integration purposes, this also provides a simplistic runtime for memref
descriptors as expected by the LLVM IR code produced by MLIR translation. In
particular, memrefs are transformed into LLVM structs (can be mapped to C
structs) with a pointer to the data, followed by dynamic sizes. This
implementation only supports statically-shaped memrefs of type float, but can
be extened if necessary.
Provide a binary for the runner and a test that exercises it.
PiperOrigin-RevId: 230876363
- introduce a way to compute union using symbolic rectangular bounding boxes
- handle multiple load/store op's to the same memref by taking a union of the regions
- command-line argument to provide capacity of the fast memory space
- minor change to replaceAllMemRefUsesWith to not generate affine_apply if the
supplied index remap was identity
PiperOrigin-RevId: 230848185
canonicalizations of operations. The ultimate important user of this is
going to be a funcBuilder->foldOrCreate<YourOp>(...) API, but for now it
is just a more convenient way to write certain classes of canonicalizations
(see the change in StandardOps.cpp).
NFC.
PiperOrigin-RevId: 230770021
- switch some debug info to emitError
- use a single constant op for zero index to make it easier to write/update
test cases; avoid creating new constant op's for common zero index cases
- test case cleanup
This is in preparation for an upcoming major update to this pass.
PiperOrigin-RevId: 230728379
Example inline notation:
trailing-location ::= 'loc' '(' location ')'
// FileLineCol Location.
%1 = "foo"() : () -> i1 loc("mysource.cc":10:8)
// Name Location
return loc("foo")
// CallSite Location
return loc(callsite("foo" at "mysource.cc":19:9))
// Fused Location
/// Without metadata
func @inline_notation() loc(fused["foo", "mysource.cc":10:8])
/// With metadata
return loc(fused<"myPass">["foo", "foo2"])
// Unknown location.
return loc(unknown)
Locations are currently only printed with inline notation at the line of each instruction. Further work is needed to allow for reference notation, e.g:
...
return loc 1
}
...
loc 1 = "source.cc":10:1
PiperOrigin-RevId: 230587621
This CL just changes various docs and comments to use the term "generic" and
"custom" when mentioning assembly forms. To be consist, several methods are
also renamed:
* FunctionParser::parseVerboseOperation() -> parseGenericOperation()
* ModuleState::hasShorthandForm() -> hasCustomForm()
* OpAsmPrinter::printDefaultOp() -> printGenericOp()
PiperOrigin-RevId: 230568819
- update fusion cost model to fuse while tolerating a certain amount of redundant
computation; add cl option -fusion-compute-tolerance
evaluate memory footprint and intermediate memory reduction
- emit debug info from -loop-fusion showing what was fused and why
- introduce function to compute memory footprint for a loop nest
- getMemRefRegion readability update - NFC
PiperOrigin-RevId: 230541857
This CL adds the Return op to EDSCs types and emitter.
This allows generating full function bodies that can be compiled all the way
down to LLVMIR and executed on CPU.
At this point, the MLIR lacks the testing infrastructure to exercise this.
End-to-end testing of full functions written in EDSCs is left for a future CL.
PiperOrigin-RevId: 230527530
- unrolling a single iteration loop by a factor of one should promote its body
into its parent; this makes it consistent with the behavior/expectation that
unrolling a loop by a factor equal to its trip count makes the loop go away.
PiperOrigin-RevId: 230426499
- ForInst::walkOps will also be used in an upcoming CL (cl/229438679); better to have
this instead of deriving from the InstWalker
PiperOrigin-RevId: 230413820
- improve/fix doc comments for affine apply composition related methods.
- drop makeSingleValueComposedAffineApply - really redundant and out of line in
a public API; it's just returning the first result of the composed affine
apply op, and not making a single result affine map or an affine_apply op.
PiperOrigin-RevId: 230406169
- the size of the private memref created for the slice should be based on
the memref region accessed at the depth at which the slice is being
materialized, i.e., symbolic in the outer IVs up until that depth, as opposed
to the region accessed based on the entire domain.
- leads to a significant contraction of the temporary / intermediate memref
whenever the memref isn't reduced to a single scalar (through store fwd'ing).
Other changes
- update to promoteIfSingleIteration - avoid introducing unnecessary identity
map affine_apply from IV; makes it much easier to write and read test cases
and pass output for all passes that use promoteIfSingleIteration; loop-fusion
test cases become much simpler
- fix replaceAllMemrefUsesWith bug that was exposed by the above update -
'domInstFilter' could be one of the ops erased due to a memref replacement in
it.
- fix getConstantBoundOnDimSize bug: a division by the coefficient of the identifier was
missing (the latter need not always be 1); add lbFloorDivisors output argument
- rename getBoundingConstantSizeAndShape -> getConstantBoundingSizeAndShape
PiperOrigin-RevId: 230405218
*) Do not remove loop nests which write to memrefs which escape the function.
*) Do not remove memrefs which escape the function (e.g. are used in the return instruction).
PiperOrigin-RevId: 230398630
Add default values to attributes, to allow attribute being left unspecified. The attr getter will always return an attribute so callers need not check for it, if the attribute is not set then the default will be returned (at present the default will be constructed upon query but this will be changed).
Add op definition for tf.AvgPool in ops.td, rewrite matcher using pattern using attribute matching & transforms. Adding some helper functions to make it simpler.
Handle attributes with dialect prefix and map them to getter without dialect prefix.
Note: VerifyAvgPoolOp could probably be autogenerated by know given the predicate specification on attributes, but deferring that to a follow up.
PiperOrigin-RevId: 230364857
Start doc generation pass that generates simple markdown output. The output is formatted simply[1] in markdown, but this allows seeing what info we have, where we can refine the op description (e.g., the inputs is probably redundant), what info is missing (e.g., the attributes could probably have a description).
The formatting of the description is still left up to whatever was in the op definition (which luckily, due to the uniformity in the .td file, turned out well but relying on the indentation there is fragile). The mechanism to autogenerate these post changes has not been added yet either. The output file could be run through a markdown formatter too to remove extra spaces.
[1]. This is not proposal for final style :) There could also be a discussion around single doc vs multiple (per dialect, per op), whether we want a TOC, whether operands/attributes should be headings or just formatted differently ...
PiperOrigin-RevId: 230354538
1) Fix FloatAttr type inconsistency in conversion from tf.FusedBatchNorm to TFLite ops
We used to compose the splat tensor out of the scalar epsilon attribute by using the
type of the variance operand. However, the epsilon attribute may have a different
bitwidth than the one in the variance operand. So it ends up we were creating
inconsistent types within the FloatAttr itself.
2) Fix SplatElementsAttr type inconsistency in AnnotateInputArrays
We need to create the zero-valued attribute according to the type provided as the
command-line arguments.
3) Concretize the result type of tf.Shape constant folding test case
Currently the resultant constant is created by the constant folding harness, using
the result type of the original op as the constant's result type. That can be
a different type than the constant's internal DenseElementsAttr.
PiperOrigin-RevId: 230244665
- print multiplication by -1 as unary negate; expressions like s0 * -1, d0 * -1
+ d1 will now appear as -s0, -d0 + d1 resp.
- a minor cleanup while on printAffineExprInternal
PiperOrigin-RevId: 230222151
This CL also makes ScopedEDSCContexts to reset the Bindable numbering when
creating a new context.
This is useful to write minimal tests that don't use FileCheck pattern
captures for now.
PiperOrigin-RevId: 230079997
This CL performs a bunch of cleanups related to EDSCs that are generally
useful in the context of using them with a simple wrapping C API (not in this
CL) and with simple language bindings to Python and Swift.
PiperOrigin-RevId: 230066505
*) Enables reduction of private memref size based on MemRef region accessed by fused slice.
*) Enables maximal fusion by creating a private memref to break a fusion-preventing dependence.
*) Adds maximal fusion flag to enable fusing as much as possible (though it still fuses the minimum cost computation slice).
PiperOrigin-RevId: 229936698
This CL adds a test reported by andydavis@ and fixes the corner case that
appears when operands do not come from an AffineApply and no Dim composition
is needed.
In such cases, we would need to create an empty map which is disallowed.
The composition in such cases becomes trivial: there is no composition.
This CL also updates the name AffineNormalizer to AffineApplyNormalizer.
PiperOrigin-RevId: 229819234
Change MinMaxAttr to match hasValidMinMaxAttribute behavior. Post rewriting the other users of that function it could be removed too. The currently generated error message is:
error: 'tfl.fake_quant' op attribute 'minmax' failed to satisfy constraint of MinMaxAttr
PiperOrigin-RevId: 229775631
This CL fixes a misunderstanding in how to build DimOp which triggered
execution issues in the CPU path.
The problem is that, given a `memref<?x4x?x8x?xf32>`, the expressions to
construct the dynamic dimensions should be:
`dim %arg, 0 : memref<?x4x?x8x?xf32>`
`dim %arg, 2 : memref<?x4x?x8x?xf32>`
and
`dim %arg, 4 : memref<?x4x?x8x?xf32>`
Before this CL, we wold construct:
`dim %arg, 0 : memref<?x4x?x8x?xf32>`
`dim %arg, 1 : memref<?x4x?x8x?xf32>`
`dim %arg, 2 : memref<?x4x?x8x?xf32>`
and expect the other dimensions to be constants.
This assumption seems consistent at first glance with the syntax of alloc:
```
%tensor = alloc(%M, %N, %O) : memref<?x4x?x8x?xf32>
```
But this was actuallyincorrect.
This CL also makes the relevant functions available to EDSCs and removes
duplication of the incorrect function.
PiperOrigin-RevId: 229622766
The operand and result types of binary ops are not necessarily the
same. For those binary ops, we cannot print in the short-form assembly.
Enhance impl:::printBinaryOp to consider operand and result types
to select which assembly form to use.
PiperOrigin-RevId: 229608142
A recent change in TableGen definitions allowed arbitrary AND/OR predicate
compositions at the cost of removing known-true predicate simplification.
Introduce a more advanced simplification mechanism instead.
In particular, instead of folding predicate C++ expressions directly in
TableGen, keep them as is and build a predicate tree in TableGen C++ library.
The predicate expression-substitution mechanism, necessary to implement complex
predicates for nested classes such as `ContainerType`, is replaced by a
dedicated predicate. This predicate appears in the predicate tree and can be
used for tree matching and separation. More specifically, subtrees defined
below such predicate may be subject to different transformations than those
that appear above. For example, a subtree known to be true above the
substitution predicate is not necessarily true below it.
Use the predicate tree structure to eliminate known-true and known-false
predicates before code emission, as well as to collapse AND and OR predicates
if their value can be deduced based on the value of one child.
PiperOrigin-RevId: 229605997
Start simple with single predicate match & transform rules for attributes.
* Its unclear whether modelling Attr predicates will be needed so start with allowing matching attributes with a single predicate.
* The input and output attr type often differs and so add ability to specify a transform between the input and output format.
PiperOrigin-RevId: 229580879
*) Adds support for fusing into consumer loop nests with multiple loads from the same memref.
*) Adds support for reducing slice loop trip count by projecting out destination loop IVs greater than destination loop depth.
*) Removes dependence on src loop depth and simplifies cost model computation.
PiperOrigin-RevId: 229575126
This is mostly plumbing to start allowing testing EDSC lowering. Prototype specifying reference implementation using verbose format without any generation/binding support. Add test pass that dumps the constructed EDSC (of which there can only be one). The idea is to enable iterating from multiple sides, this is wrong on many dimensions at the moment.
PiperOrigin-RevId: 229570535
In TableGen definitions, the "Type" class has been used for types of things
that can be stored in Attributes, but not necessarily present in the MLIR type
system. As a consequence, records like "String" or "DerviedAttrBody" were of
class "Type", which can be confusing. Furthermore, the "builderCall" field of
the "Type" class serves only for attribute construction. Some TableGen "Type"
subclasses that correspond to MLIR kinds of types do not have a canonical way
of construction only from the data available in TableGen, e.g. MemRefType would
require the list of affine maps. This leads to a conclusion that the entities
that describe types of objects appearing in Attributes should be independent of
"Type": they have some properties "Type"s don't and vice versa.
Do not parameterize Tablegen "Attr" class by an instance of "Type". Instead,
provide a "constBuilderCall" field that can be used to build an attribute from
a constant value stored in TableGen instead of indirectly going through
Attribute.Type.builderCall. Some attributes still don't have a
"constBuilderCall" because they used to depend on types without a
"builderCall".
Drop definitions of class "Type" that don't correspond to MLIR Types. Provide
infrastructure to define type-dependent attributes and string-backed attributes
for convenience.
PiperOrigin-RevId: 229570087
We also need the broadcast logic in the TensorFlow dialect. Move it to a
Dialect/ directory for a broader scope. This Dialect/ directory is intended
for code not in core IR, but can potentially be shared by multiple dialects.
Apart from fixing TensorFlow op TableGen to use this trait, this CL only
contains mechanical code shuffling.
PiperOrigin-RevId: 229563911
The constant folding rules assumes value attributes of operands are already
verified to be in good standing.
For each op in the above, the constant folding rules support both integer and
floating point cases. Broadcast behavior is also supported as per the semantics
of TFLite ops.
This CL does not handle overflow/underflow cases yet.
PiperOrigin-RevId: 229441221
LLVM IR types are defined using MLIR's extendable type system. The dialect
provides the only type kind, LLVMType, that wraps an llvm::Type*. Since LLVM
IR types are pointer-unique, MLIR type systems relies on those pointers to
perform its own type unique'ing. Type parsing and printing is delegated to
LLVM libraries.
Define MLIR operations for the LLVM IR instructions currently used by the
translation to the LLVM IR Target to simplify eventual transition. Operations
classes are defined using TableGen. LLVM IR instruction operands that are only
allowed to take constant values are accepted as attributes instead. All
operations are using verbose form for printing and parsing.
PiperOrigin-RevId: 229400375
MLIR has support for type-polymorphic instructions, i.e. instructions that may
take arguments of different types. For example, standard arithmetic operands
take scalars, vectors or tensors. In order to express such instructions in
TableGen, we need to be able to verify that a type object satisfies certain
constraints, but we don't need to construct an instance of this type. The
existing TableGen definition of Type requires both. Extract out a
TypeConstraint TableGen class to define restrictions on types. Define the Type
TableGen class as a subclass of TypeConstraint for consistency. Accept records
of the TypeConstraint class instead of the Type class as values in the
Arguments class when defining operators.
Replace the predicate logic TableGen class based on conjunctive normal form
with the predicate logic classes allowing for abitrary combinations of
predicates using Boolean operators (AND/OR/NOT). The combination is
implemented using simple string rewriting of C++ expressions and, therefore,
respects the short-circuit evaluation order. No logic simplification is
performed at the TableGen level so all expressions must be valid C++.
Maintaining CNF using TableGen only would have been complicated when one needed
to introduce top-level disjunction. It is also unclear if it could lead to a
significantly simpler emitted C++ code. In the future, we may replace inplace
predicate string combination with a tree structure that can be simplified in
TableGen's C++ driver.
Combined, these changes allow one to express traits like ArgumentsAreFloatLike
directly in TableGen instead of relying on C++ trait classes.
PiperOrigin-RevId: 229398247
Nicolas Vasilache