Commit Graph

1409 Commits

Author SHA1 Message Date
Stephan Herhut 2125c0e3a8 Extend conversion of SubViewOp to llvm to also support cases where size and stride
are constant (i.e., there are no size and stride operands).

We recently added canonicalization that rewrites constant size and stride operands to
SubViewOp into static information in the type, so these patterns now occur during code
generation.

PiperOrigin-RevId: 283524688
2019-12-03 05:11:49 -08:00
Lei Zhang 1af9633d85 [spirv] Add spv.SubgroupBallotKHROp
PiperOrigin-RevId: 283522284
2019-12-03 04:49:56 -08:00
Alex Zinenko fdbb99cd62 Add linkage support to LLVMFuncOp
A recent commit introduced the Linkage attribute to the LLVM dialect and used
it in the Global Op. Also use it in LLVMFuncOp. As per LLVM Language Reference,
if the linkage attribute is omitted, the function is assumed to have external
linkage.

PiperOrigin-RevId: 283493299
2019-12-03 00:26:44 -08:00
Aart Bik 3126004a5a [VectorOps] Add legality rules to broadcast
PiperOrigin-RevId: 283360101
2019-12-02 09:57:27 -08:00
Lei Zhang b41162b3af [ODS] Generate builders taking unwrapped value and defaults for attributes
Existing builders generated by ODS require attributes to be passed
in as mlir::Attribute or its subclasses. This is okay foraggregate-
parameter builders, which is primarily to be used by programmatic
C++ code generation; it is inconvenient for separate-parameter
builders meant to be called in manually written C++ code because
it requires developers to wrap raw values into mlir::Attribute by
themselves.

This CL extends to generate additional builder methods that
take raw values for attributes and handles the wrapping in the
builder implementation. Additionally, if an attribute appears
late in the arguments list and has a default value, the default
value is supplied in the declaration if possible.

PiperOrigin-RevId: 283355919
2019-12-02 09:33:57 -08:00
Lei Zhang 4982eaf87c [DRR] Introduce `$_` to ignore op argument match
Right now op argument matching in DRR is position-based, meaning we need to
specify N arguments for an op with N ODS-declared argument. This can be annoying
when we don't want to capture all the arguments. `$_` is to remedy the situation.

PiperOrigin-RevId: 283339992
2019-12-02 07:54:50 -08:00
Alexander Belyaev 9630fcbc52 Lower linalg.indexed_generic with libcall to LLVM.
PiperOrigin-RevId: 283328994
2019-12-02 06:30:52 -08:00
Alex Zinenko d5e627f84b Introduce Linkage attribute to the LLVM dialect
LLVM IR supports linkage on global objects such as global variables and
functions. Introduce the Linkage attribute into the LLVM dialect, backed by an
integer storage. Use this attribute on LLVM::GlobalOp and make it mandatory.
Implement parsing/printing of the attribute and conversion to LLVM IR.

See .

PiperOrigin-RevId: 283309328
2019-12-02 03:28:10 -08:00
Denis Khalikov cd556f25de [spirv] Check that operand of `spirv::CompositeExtractOp` is constant while folding.
Closes 

COPYBARA_INTEGRATE_REVIEW=https://github.com/tensorflow/mlir/pull/281 from denis0x0D:sandbox/composite_ex_fold d02d73658bd1b9eaa515eb4e0aee34bc41d4252b
PiperOrigin-RevId: 282971563
2019-11-28 13:27:56 -08:00
Jose Ignacio Gomez 0494ef60f7 [Linalg] Change attribute n_loop_types to iterator
This addresses issue . Linalg is updated to take the same form
of iterator_types than vector contraction.

Closes 

COPYBARA_INTEGRATE_REVIEW=https://github.com/tensorflow/mlir/pull/280 from tetuante:PRissue270 d26d88d090d3765d3b9884bfabdd023143f27287
PiperOrigin-RevId: 282905396
2019-11-28 01:59:55 -08:00
Lei Zhang d4e4387fbf [spirv] Add folders for spv.IAdd and spv.IMul
Adding zero and multiplying one can be common when generating code
for index calculation.

This CL also sorted canonicalize.mlir to alphabetical order.

PiperOrigin-RevId: 282828055
2019-11-27 13:46:52 -08:00
Aart Bik 9f89c34f4b Fixed typo in Toy tutorial (second var e -> var f)
PiperOrigin-RevId: 282810649
2019-11-27 11:58:45 -08:00
Nicolas Vasilache 1fa8c8070b Implement Linalg to loops lowering as a pattern
This CL rewrites the linalg ops to loops transformations as patterns that can be targeted directly from Tablegen. Reliance on OpFolder is removed and to cope with it we introduce local folding patterns that are applied greedily.

PiperOrigin-RevId: 282765550
2019-11-27 07:32:13 -08:00
Aart Bik e2232fbcee [VectorOps] Refine BroadcastOp in VectorOps dialect
Since second argument is always fully overwritten and
shape is define in "to" clause, it is not needed.
Also renamed "into" to "to" now that arg is dropped.

PiperOrigin-RevId: 282686475
2019-11-26 19:52:38 -08:00
Aart Bik cf97263cb8 [VectorOps] Add a BroadcastOp to the VectorOps dialect
PiperOrigin-RevId: 282643305
2019-11-26 14:43:31 -08:00
Mahesh Ravishankar 03620fa70a Misc changes to lowering to SPIR-V.
These changes to SPIR-V lowering while adding support for lowering
SUbViewOp, but are not directly related.
- Change the lowering of MemRefType to
  !spv.ptr<!spv.struct<!spv.array<...>[offset]>, ..>
  This is consistent with the Vulkan spec.
- To enable testing a simple pattern of lowering functions is added to
  ConvertStandardToSPIRVPass. This is just used to convert the type of
  the arguments of the function. The added function lowering itself is
  not meant to be the way functions are eventually lowered into SPIR-V
  dialect.

PiperOrigin-RevId: 282589644
2019-11-26 10:11:34 -08:00
Nicolas Vasilache 9059cf392d Automated rollback of commit d60133f89b
PiperOrigin-RevId: 282574110
2019-11-26 08:47:48 -08:00
Nicolas Vasilache 109338085d Relax restriction on affine_apply dim and symbol operands
The affine_apply operation is currently "doubly" affine and conflates two things:
1. it applies an affine map to a list of values of type `index` that are defined as either dim or symbol
2. it restricts (and propagates constraints on) the provenance of dims and symbols to a small subset of ops for which more restrictive polyhedral constraints apply.

Point 2. is related to the ability to form so-called static control parts and is related to dependence analysis and legality of transformations.

Point 1. however is completely independent, the only local implication of dims and symbol for affine_apply is that dims compose while symbols concatenate as well as the structural constraint that dims may not be multiplied.

The properties of composition and canonicalization in affine_apply are more generally useful. This CL relaxes the verifier on affine_apply so it can be used more generally.

The relevant affine.for/if/load/store op verifiers already implement the dim and symbol checking.

See this thread for the related discussion: https://groups.google.com/a/tensorflow.org/g/mlir/c/HkwCbV8D9N0/m/8srUNrX6CAAJ

PiperOrigin-RevId: 282562517
2019-11-26 07:39:05 -08:00
Lei Zhang 13c6e419ca Add support for AttrSizedOperandSegments/AttrSizedResultSegments
Certain operations can have multiple variadic operands and their size
relationship is not always known statically. For such cases, we need
a per-op-instance specification to divide the operands into logical
groups or segments. This can be modeled by attributes.

This CL introduces C++ trait AttrSizedOperandSegments for operands and
AttrSizedResultSegments for results. The C++ trait just guarantees
such size attribute has the correct type (1D vector) and values
(non-negative), etc. It serves as the basis for ODS sugaring that
with ODS argument declarations we can further verify the number of
elements match the number of ODS-declared operands and we can generate
handy getter methods.

PiperOrigin-RevId: 282467075
2019-11-25 17:26:50 -08:00
Nicolas Vasilache 174076a157 Use vector.InsertStridedSlice in Vector -> Vector unrolling
This CL uses the recently added op to finish the implementation of Vector -> Vector unrolling by replacing the "fake join op" by a series of InsertStridedSliceOp.

Test is updated accordingly

PiperOrigin-RevId: 282451126
2019-11-25 15:56:37 -08:00
Nicolas Vasilache 36469f7d2a Add a vector.InsertStridedSliceOp
This new op is the counterpart of vector.StridedSliceOp and will be used for in the pattern rewrites for vector unrolling.

PiperOrigin-RevId: 282447414
2019-11-25 15:37:13 -08:00
MLIR Team 1012c492f0 Allow LLVM::ExtractElementOp to have non-i32 indices.
Also change the text format a bit, so that indices are braced by squares.

PiperOrigin-RevId: 282437095
2019-11-25 14:44:52 -08:00
Ben Vanik 38d7870ee5 Make std.divis and std.diviu support ElementsAttr folding.
PiperOrigin-RevId: 282434465
2019-11-25 14:31:43 -08:00
Andy Davis 8fc44a4d13 Update VectorContractionOp to take iterator types and index mapping attributes compatible with linalg ops.
PiperOrigin-RevId: 282412311
2019-11-25 12:40:00 -08:00
Christian Sigg d60133f89b Changing directory shortcut for CPU/GPU runner utils.
Moving cuda-runtime-wrappers.so into subdirectory to match libmlir_runner_utils.so.
Provide parent directory when running test and load .so from subdirectory.

PiperOrigin-RevId: 282410749
2019-11-25 12:30:54 -08:00
Lei Zhang 9b6e6cef68 De-duplicate EnumAttr overrides by defining defaults
EnumAttr should provide meaningful defaults so concrete instances
do not need to duplicate the fields.

PiperOrigin-RevId: 282398431
2019-11-25 11:29:55 -08:00
Mahesh Ravishankar bd485afda0 Introduce attributes that specify the final ABI for a spirv::ModuleOp.
To simplify the lowering into SPIR-V, while still respecting the ABI
requirements of SPIR-V/Vulkan, split the process into two
1) While lowering a function to SPIR-V (when the function is an entry
   point function), allow specifying attributes on arguments and
   function itself that describe the ABI of the function.
2) Add a pass that materializes the ABI described in the function.

Two attributes are needed.
1) Attribute on arguments of the entry point function that describe
   the descriptor_set, binding, storage class, etc, of the
   spv.globalVariable this argument will be replaced by
2) Attribute on function that specifies workgroup size, etc. (for now
   only workgroup size).

Add the pass -spirv-lower-abi-attrs to materialize the ABI described
by the attributes.

This change makes the SPIRVBasicTypeConverter class unnecessary and is
removed, further simplifying the SPIR-V lowering path.

PiperOrigin-RevId: 282387587
2019-11-25 11:19:56 -08:00
Mahesh Ravishankar 1ea231bd39 Allow memref_cast from static strides to dynamic strides.
Memref_cast supports cast from static shape to dynamic shape
memrefs. The same should be true for strides as well, i.e a memref
with static strides can be casted to a memref with dynamic strides.

PiperOrigin-RevId: 282381862
2019-11-25 11:08:56 -08:00
Nicolas Vasilache 01145544aa Add vector.insertelement op
This is the counterpart of vector.extractelement op and has the same
limitations at the moment (static I64IntegerArrayAttr to express position).
This restriction will be filterd in the future.
LLVM lowering will be added in a subsequent commit.

PiperOrigin-RevId: 282365760
2019-11-25 08:47:15 -08:00
Alex Zinenko bf4692dc49 Introduce gpu.func
Introduce a new function-like operation to the GPU dialect to provide a
placeholder for the execution semantic description and to add support for GPU
memory hierarchy.  This aligns with the overall goal of the dialect to expose
the common abstraction layer for GPU devices, in particular by providing an
MLIR unit of semantics (i.e. an operation) for memory modeling.

This proposal has been discussed in the mailing list:
https://groups.google.com/a/tensorflow.org/d/msg/mlir/RfXNP7Hklsc/MBNN7KhjAgAJ
As decided, the "convergence" aspect of the execution model will be factored
out into a new discussion and therefore is not included in this commit. This
commit only introduces the operation but does not hook it up with the remaining
flow. The intention is to develop the new flow while keeping the old flow
operational and do the switch in a simple, separately reversible commit.

PiperOrigin-RevId: 282357599
2019-11-25 08:10:37 -08:00
Ben Vanik d2284f1f0b Support folding of StandardOps with DenseElementsAttr.
PiperOrigin-RevId: 282270243
2019-11-24 19:23:38 -08:00
Lei Zhang aaafeac89b [spirv] NFC: rename test files and sort tests inside
PiperOrigin-RevId: 282132339
2019-11-23 06:58:38 -08:00
Uday Bondhugula 6a101671b0 Make isValidSymbol more powerful
The check in isValidSymbol, as far as a DimOp result went, checked if
the dim op was on a top-level memref. However, any alloc'ed, view, or
subview memref would be fine as long as the corresponding dimension of
that memref is either a static one or was in turn created using a valid
symbol in the case of dynamic dimensions.

Reported-by: Jose Gomez

Signed-off-by: Uday Bondhugula <uday@polymagelabs.com>

Closes 

COPYBARA_INTEGRATE_REVIEW=https://github.com/tensorflow/mlir/pull/252 from bondhugula:symbol 7b57dc394df9375e651f497231c6e4525a32a662
PiperOrigin-RevId: 282097114
2019-11-22 22:09:31 -08:00
River Riddle b8ee563449 NFC: Remove unnecessarily guarded tablegen includes.
Support for including a file multiple times was added in tablegen, removing the need for these extra guards. This is because we already insert c/c++ style header guards within each of the specific .td files.

PiperOrigin-RevId: 282076728
2019-11-22 18:01:57 -08:00
Denis Khalikov a5cda4763f [spirv] Add a canonicalizer for `spirv::LogicalNotOp`.
Add a canonicalizer for `spirv::LogicalNotOp`.
Converts:
* spv.LogicalNot(spv.IEqual(...)) -> spv.INotEqual(...)
* spv.LogicalNot(spv.INotEqual(...)) -> spv.IEqual(...)
* spv.LogicalNot(spv.LogicalEqual(...)) -> spv.LogicalNotEqual(...)
* spv.LogicalNot(spv.LogicalNotEqual(...)) -> spv.LogicalEqual(...)

Also moved the test for spv.IMul to arithemtic tests.

Closes 

COPYBARA_INTEGRATE_REVIEW=https://github.com/tensorflow/mlir/pull/256 from denis0x0D:sandbox/canon_logical_not 76ab5787b2c777f948c8978db061d99e76453d44
PiperOrigin-RevId: 282012356
2019-11-22 12:25:52 -08:00
Mahesh Ravishankar 6db8530c26 Add more canonicalizations for SubViewOp.
Depending on which of the offsets, sizes, or strides are constant, the
subview op can be canonicalized in different ways. Add such
canonicalizations, which generalize the existing approach of
canonicalizing subview op only if all of offsets, sizes and shapes are
constants.

PiperOrigin-RevId: 282010703
2019-11-22 12:14:18 -08:00
River Riddle c35378003c Add support for using the ODS result names as the Asm result names for multi-result operations.
This changes changes the OpDefinitionsGen to automatically add the OpAsmOpInterface for operations with multiple result groups using the provided ODS names. We currently just limit the generation to multi-result ops as most single result operations don't have an interesting name(result/output/etc.). An example is shown below:
// The following operation:
def MyOp : ... {
  let results = (outs AnyType:$first, Variadic<AnyType>:$middle, AnyType);
}

// May now be printed as:
%first, %middle:2, %0 = "my.op" ...

PiperOrigin-RevId: 281834156
2019-11-21 14:55:46 -08:00
Christian Sigg d7c17195a4 Change CUDA tests to use print_memref.
Swap dimensions in all-reduce-op test.

PiperOrigin-RevId: 281791744
2019-11-21 11:26:36 -08:00
Nicolas Vasilache 2c4985816f Split Linalg declarative patterns from specific test patterns - NFC
This will make it easier to scale out test patterns and build specific passes that do not interfere with independent testing.

PiperOrigin-RevId: 281736335
2019-11-21 06:40:17 -08:00
Alex Zinenko b5af3784a6 Don't force newline before function attributes
Due to legacy reasons, a newline character followed by two spaces was always
inserted before the attributes of the function Op in pretty form. This breaks
formatting when functions are nested in some other operations. Don't print the
newline and just put the attributes on the same line, which is also more
consistent with module Op. Line breaking aware of indentation can be introduced
separately into the parser if deemed useful.

PiperOrigin-RevId: 281721793
2019-11-21 05:08:19 -08:00
MLIR Team 75379a684f Correctly parse empty affine maps.
Previously the test case crashes / produces an error.

PiperOrigin-RevId: 281630540
2019-11-20 18:30:15 -08:00
River Riddle fafb708b9a Merge DCE and unreachable block elimination into a new utility 'simplifyRegions'.
This moves the different canonicalizations of regions into one place and invokes them in the fixed-point iteration of the canonicalizer.

PiperOrigin-RevId: 281617072
2019-11-20 15:53:19 -08:00
Andy Davis d6a70b31be Add VectorContractionOp to the VectorOps dialect.
PiperOrigin-RevId: 281605471
2019-11-20 14:53:57 -08:00
Mahesh Ravishankar 1145cebdab Verify subview op result has dynamic shape, when sizes are specified.
If the sizes are specified as arguments to the subview op, then the
shape must be dynamic as well.

PiperOrigin-RevId: 281591608
2019-11-20 14:16:05 -08:00
Sean Silva e4f83c6c26 Add multi-level DCE pass.
This is a simple multi-level DCE pass that operates pretty generically on
the IR. Its key feature compared to the existing peephole dead op folding
that happens during canonicalization is being able to delete recursively
dead cycles of the use-def graph, including block arguments.

PiperOrigin-RevId: 281568202
2019-11-20 12:55:10 -08:00
Mahesh Ravishankar 19212105dd Changes to SubViewOp to make it more amenable to canonicalization.
The current SubViewOp specification allows for either all offsets,
shape and stride to be dynamic or all of them to be static. There are
opportunities for more fine-grained canonicalization based on which of
these are static. For example, if the sizes are static, the result
memref is of static shape. The specification of SubViewOp is modified
to allow on or more of offsets, shapes and strides to be statically
specified. The verification is updated to ensure that the result type
of the subview op is consistent with which of these are static and
which are dynamic.

PiperOrigin-RevId: 281560457
2019-11-20 12:32:51 -08:00
Nicolas Vasilache fa14d4f6ab Implement unrolling of vector ops to finer-grained vector ops as a pattern.
This CL uses the pattern rewrite infrastructure to implement a simple VectorOps -> VectorOps legalization strategy to unroll coarse-grained vector operations into finer grained ones.
The transformation is written using local pattern rewrites to allow composition with other rewrites. It proceeds by iteratively introducing fake cast ops and cleaning canonicalizing or lowering them away where appropriate.

This is an example of writing transformations as compositions of local pattern rewrites that should enable us to make them significantly more declarative.

PiperOrigin-RevId: 281555100
2019-11-20 11:49:36 -08:00
River Riddle eb418559ef Add a new OpAsmOpInterface to allow for ops to directly hook into the AsmPrinter.
This interface provides more fine-grained hooks into the AsmPrinter than the dialect interface, allowing for operations to define the asm name to use for results directly on the operations themselves. The hook is also expanded to enable defining named result "groups". Get a special name to use when printing the results of this operation.
The given callback is invoked with a specific result value that starts a
result "pack", and the name to give this result pack. To signal that a
result pack should use the default naming scheme, a None can be passed
in instead of the name.

For example, if you have an operation that has four results and you want
to split these into three distinct groups you could do the following:

  setNameFn(getResult(0), "first_result");
  setNameFn(getResult(1), "middle_results");
  setNameFn(getResult(3), ""); // use the default numbering.

This would print the operation as follows:

  %first_result, %middle_results:2, %0 = "my.op" ...

PiperOrigin-RevId: 281546873
2019-11-20 10:45:45 -08:00
Nicolas Vasilache 3c055957de Add StridedMemRef<>::operator[] - NFC
This operator is used for internal debugging purposes.

PiperOrigin-RevId: 281544152
2019-11-20 10:17:13 -08:00
Alexander Belyaev e50261657f Fix 'the the' typo.
PiperOrigin-RevId: 281501234
2019-11-20 05:38:14 -08:00