Commit Graph

88 Commits

Author SHA1 Message Date
Uday Bondhugula 18c8c934d8 [MLIR] Introduce scf.execute_region op
Introduce the execute_region op that is able to hold a region which it
executes exactly once. The op encapsulates a CFG within itself while
isolating it from the surrounding control flow. Proposal discussed here:
https://llvm.discourse.group/t/introduce-std-inlined-call-op-proposal/282

execute_region enables one to inline a function without lowering out all
other higher level control flow constructs (affine.for/if, scf.for/if)
to the flat list of blocks / CFG form. It thus allows the benefit of
transforms on higher level control flow ops available in the presence of
the inlined calls. The inlined calls continue to benefit from
propagation of SSA values across their top boundary. Functions won’t
have to remain outlined until later than desired.  Abstractions like
affine execute_regions, lambdas with implicit captures could be lowered
to this without first lowering out structured loops/ifs or outlining.
But two potential early use cases are of: (1) an early inliner (which
can inline functions by introducing execute_region ops), (2) lowering of
an affine.execute_region, which cleanly maps to an scf.execute_region
when going from the affine dialect to the scf dialect.

Differential Revision: https://reviews.llvm.org/D75837
2021-06-18 15:22:33 +05:30
MaheshRavishankar 621d93d263 [mlir][SCF] Remove empty else blocks of `scf.if` operations.
Differential Revision: https://reviews.llvm.org/D104273
2021-06-15 15:07:20 -07:00
Butygin 4184018253 [mlir][SCF] Canonicalize nested ParallelOp's
Differential Revision: https://reviews.llvm.org/D102799
2021-05-22 14:00:00 +03:00
Nicolas Vasilache 84a880e1e2 [mlir][SCF] NFC - Drop SCF EDSC usage
Drop the SCF dialect EDSC subdirectory and update all uses.

Differential Revision: https://reviews.llvm.org/D102780
2021-05-19 15:52:14 +00:00
Julian Gross 1fbb484ea4 [WIP][mlir] Resolve memref dependency in canonicalize pass.
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
2021-05-17 11:33:38 +02:00
Sean Silva 12874e93a1 [mlir][NFC] Add helper for common pattern of replaceAllUsesExcept
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
2021-05-13 12:42:10 -07:00
William S. Moses f4a2dbfe29 [MLIR][SCF] Combine adjacent scf.if with same condition
Differential Revision: https://reviews.llvm.org/D101798
2021-05-05 00:39:58 -04:00
William S. Moses 8e211bf1c8 [MLIR][SCF] Assume uses of condition in the body of scf.while is true
Differential Revision: https://reviews.llvm.org/D101801
2021-05-04 11:39:07 -04:00
Lorenzo Chelini de94b1855c [mlir] Fix top-level comments (NFC) 2021-04-29 13:06:40 +02:00
William S. Moses ca27260701 [MLIR] Add SCF.if Condition Canonicalizations
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
2021-04-26 20:13:08 -04:00
thomasraoux ded18708f9 [mlir][NFC] Refactor linalg substituteMin and AffineMinSCF canonizalizations
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
2021-04-21 07:19:36 -07:00
Nicolas Vasilache 843f1fc825 [mlir][scf] Add scf.for + tensor.cast canonicalization pattern
Fold scf.for iter_arg/result pairs that go through incoming/ougoing
a tensor.cast op pair so as to pull the tensor.cast inside the scf.for:

```
  %0 = tensor.cast %t0 : tensor<32x1024xf32> to tensor<?x?xf32>
  %1 = scf.for %i = %c0 to %c1024 step %c32 iter_args(%iter_t0 = %0)
     -> (tensor<?x?xf32>) {
    %2 = call @do(%iter_t0) : (tensor<?x?xf32>) -> tensor<?x?xf32>
    scf.yield %2 : tensor<?x?xf32>
  }
  %2 = tensor.cast %1 : tensor<?x?xf32> to tensor<32x1024xf32>
  use_of(%2)
```

folds into:

```
  %0 = scf.for %arg2 = %c0 to %c1024 step %c32 iter_args(%arg3 = %arg0)
      -> (tensor<32x1024xf32>) {
    %2 = tensor.cast %arg3 : tensor<32x1024xf32> to tensor<?x?xf32>
    %3 = call @do(%2) : (tensor<?x?xf32>) -> tensor<?x?xf32>
    %4 = tensor.cast %3 : tensor<?x?xf32> to tensor<32x1024xf32>
    scf.yield %4 : tensor<32x1024xf32>
  }
  use_of(%0)
```

Differential Revision: https://reviews.llvm.org/D100661
2021-04-16 16:50:21 +00:00
River Riddle 4efb7754e0 [mlir][NFC] Add a using directive for llvm::SetVector
Differential Revision: https://reviews.llvm.org/D100436
2021-04-15 16:09:34 -07:00
Butygin eb31540066 [mlir] Canonicalize single-iteration ParallelOp
Differential Revision: https://reviews.llvm.org/D100248
2021-04-13 13:42:19 +03:00
Chris Lattner dc4e913be9 [PatternMatch] Big mechanical rename OwningRewritePatternList -> RewritePatternSet and insert -> add. NFC
This doesn't change APIs, this just cleans up the many in-tree uses of these
names to use the new preferred names.  We'll keep the old names around for a
couple weeks to help transitions.

Differential Revision: https://reviews.llvm.org/D99127
2021-03-22 17:20:50 -07:00
Chris Lattner 3a506b31a3 Change OwningRewritePatternList to carry an MLIRContext with it.
This updates the codebase to pass the context when creating an instance of
OwningRewritePatternList, and starts removing extraneous MLIRContext
parameters.  There are many many more to be removed.

Differential Revision: https://reviews.llvm.org/D99028
2021-03-21 10:06:31 -07:00
Butygin 5657f93e78 [mlir] Canonicalize IfOp with trivial `then` and `else` bodies to list of SelectOp's
* Do we need a threshold on maximum number of Yeild arguments processed (maximum number of SelectOp's to be generated)?
* Had to modify some old IfOp tests to not get optimized by this pattern

Differential Revision: https://reviews.llvm.org/D98592
2021-03-20 12:18:49 +03:00
lorenzo chelini 4c782a24d9 [mlir] Fix typo in SCF.cpp (NFC) 2021-03-18 19:15:33 +01:00
lorenzo chelini 0a74a7161b [mlir] scf::ForOp: Drop iter arguments (and corresponding result) with no use
'ForOpIterArgsFolder' can now remove iterator arguments (and corresponding
results) with no use.

Example:

```
%cst = constant 32 : i32

%0:2 = scf.for %arg1 = %lb to %ub step %step iter_args(%arg2 = %arg0, %arg3 = %cst)
  -> (i32, i32) {
  %1 = addu %arg2, %cst : i32
  scf.yield %1, %1 : i32, i32
}

use(%0#0)

```

%arg3 is not used in the block, and its corresponding result `%0#1` has no use,
thus remove the iter argument.

Reviewed By: nicolasvasilache

Differential Revision: https://reviews.llvm.org/D98711
2021-03-17 12:06:17 +00:00
Lorenzo Chelini fd7eee64c5 scf::ForOp: Fold away iterator arguments with no use and for which the corresponding input is yielded
Enhance 'ForOpIterArgsFolder' to remove unused iteration arguments in a
scf::ForOp. If the block argument corresponding to the given iterator has no
use and the yielded value equals the input, we fold it away.

Reviewed By: nicolasvasilache

Differential Revision: https://reviews.llvm.org/D98503
2021-03-16 07:01:25 +00:00
Julian Gross e2310704d8 [MLIR] Create memref dialect and move dialect-specific ops from std.
Create the memref dialect and move dialect-specific ops
from std dialect to this dialect.

Moved ops:
AllocOp -> MemRef_AllocOp
AllocaOp -> MemRef_AllocaOp
AssumeAlignmentOp -> MemRef_AssumeAlignmentOp
DeallocOp -> MemRef_DeallocOp
DimOp -> MemRef_DimOp
MemRefCastOp -> MemRef_CastOp
MemRefReinterpretCastOp -> MemRef_ReinterpretCastOp
GetGlobalMemRefOp -> MemRef_GetGlobalOp
GlobalMemRefOp -> MemRef_GlobalOp
LoadOp -> MemRef_LoadOp
PrefetchOp -> MemRef_PrefetchOp
ReshapeOp -> MemRef_ReshapeOp
StoreOp -> MemRef_StoreOp
SubViewOp -> MemRef_SubViewOp
TransposeOp -> MemRef_TransposeOp
TensorLoadOp -> MemRef_TensorLoadOp
TensorStoreOp -> MemRef_TensorStoreOp
TensorToMemRefOp -> MemRef_BufferCastOp
ViewOp -> MemRef_ViewOp

The roadmap to split the memref dialect from std is discussed here:
https://llvm.discourse.group/t/rfc-split-the-memref-dialect-from-std/2667

Differential Revision: https://reviews.llvm.org/D98041
2021-03-15 11:14:09 +01:00
Nicolas Vasilache 35908406dc [mlir][scf] Canonicalize scf.for last tensor iteration result.
Canonicalize the iter_args of an scf::ForOp that involve a tensor_load and
for which only the last loop iteration is actually visible outside of the
loop. The canonicalization looks for a pattern such as:
```
   %t0 = ... : tensor_type
   %0 = scf.for ... iter_args(%bb0 : %t0) -> (tensor_type) {
     ...
     // %m is either tensor_to_memref(%bb00) or defined above the loop
     %m... : memref_type
     ... // uses of %m with potential inplace updates
     %new_tensor = tensor_load %m : memref_type
     ...
     scf.yield %new_tensor : tensor_type
   }
```

`%bb0` may have either 0 or 1 use. If it has 1 use it must be exactly a
`%m = tensor_to_memref %bb0` op that feeds into the yielded `tensor_load`
op.

If no aliasing write of `%new_tensor` occurs between tensor_load and yield
then the value %0 visible outside of the loop is the last `tensor_load`
produced in the loop.

For now, we approximate the absence of aliasing by only supporting the case
when the tensor_load is the operation immediately preceding the yield.

The canonicalization rewrites the pattern as:
```
   // %m is either a tensor_to_memref or defined above
   %m... : memref_type
   scf.for ... { // no iter_args
     ... // uses of %m with potential inplace updates
   }
   %0 = tensor_load %m : memref_type
```

Differential revision: https://reviews.llvm.org/D97953
2021-03-05 09:42:19 +00:00
Christian Sigg 8c074cb0b7 [mlir] Mark OpState::getAttrs() deprecated.
Fix call sites.

The method will be removed 2 weeks later.

Reviewed By: rriddle

Differential Revision: https://reviews.llvm.org/D97464
2021-02-25 20:54:42 +01:00
Nicolas Vasilache 551ba72760 [mlir] NFC - Use declarative assembly for scf::YieldOp 2021-02-23 11:17:30 +00:00
Alexander Belyaev a89035d750 Revert "[MLIR] Create memref dialect and move several dialect-specific ops from std."
This commit introduced a cyclic dependency:
Memref dialect depends on Standard because it used ConstantIndexOp.
Std depends on the MemRef dialect in its EDSC/Intrinsics.h

Working on a fix.

This reverts commit 8aa6c3765b.
2021-02-18 12:49:52 +01:00
Julian Gross 8aa6c3765b [MLIR] Create memref dialect and move several dialect-specific ops from std.
Create the memref dialect and move several dialect-specific ops without
dependencies to other ops from std dialect to this dialect.

Moved ops:
AllocOp -> MemRef_AllocOp
AllocaOp -> MemRef_AllocaOp
DeallocOp -> MemRef_DeallocOp
MemRefCastOp -> MemRef_CastOp
GetGlobalMemRefOp -> MemRef_GetGlobalOp
GlobalMemRefOp -> MemRef_GlobalOp
PrefetchOp -> MemRef_PrefetchOp
ReshapeOp -> MemRef_ReshapeOp
StoreOp -> MemRef_StoreOp
TransposeOp -> MemRef_TransposeOp
ViewOp -> MemRef_ViewOp

The roadmap to split the memref dialect from std is discussed here:
https://llvm.discourse.group/t/rfc-split-the-memref-dialect-from-std/2667

Differential Revision: https://reviews.llvm.org/D96425
2021-02-18 11:29:39 +01:00
Tres Popp 787d771dce [mlir] Don't return nullptrs from scf::IfOp::getSuccessorRegions
Previously this might happen if there was no elseRegion and the method
was asked for all successor regions.

Differential Revision: https://reviews.llvm.org/D96764
2021-02-16 12:06:30 +01:00
Alexander Belyaev 09c18a6606 [mlir] Return scf.parallel ops resulted from tiling.
Differential Revision: https://reviews.llvm.org/D96024
2021-02-04 14:47:14 +01:00
Lei Zhang 5b7619c90b [mlir] Fix scf.for single iteration canonicalization check
We should be check whether lb + step >= ub to determine
whether this is a single iteration. Previously we were
checking lb + lb >= ub.

Reviewed By: ftynse

Differential Revision: https://reviews.llvm.org/D95440
2021-02-02 18:30:50 -05:00
Lei Zhang a2e791e396 Revert "[mlir] Fix scf.for single iteration canonicalization check"
This reverts commit b2b35697dc.
It gotten accidentially landed before LGTM.
2021-02-02 11:13:39 -05:00
Lei Zhang b2b35697dc [mlir] Fix scf.for single iteration canonicalization check
We should be check whether lb + step >= ub to determine
whether this is a single iteration. Previously we were
checking lb + lb >= ub.

Differential Revision: https://reviews.llvm.org/D95440
2021-02-02 11:08:56 -05:00
Alexander Belyaev 80966447a2 [mlir][nfc] Move `getInnermostParallelLoops` to SCF/Transforms/Utils.h. 2021-01-26 17:00:15 +01:00
River Riddle 1b97cdf885 [mlir][IR][NFC] Move context/location parameters of builtin Type::get methods to the start of the parameter list
This better matches the rest of the infrastructure, is much simpler, and makes it easier to move these types to being declaratively specified.

Differential Revision: https://reviews.llvm.org/D93432
2020-12-17 13:01:36 -08:00
Christian Sigg 0bf4a82a5a [mlir] Use mlir::OpState::operator->() to get to methods of mlir::Operation. This is a preparation step to remove the corresponding methods from OpState.
Reviewed By: silvas, rriddle

Differential Revision: https://reviews.llvm.org/D92878
2020-12-09 12:11:32 +01:00
Christian Sigg c4a0405902 Add `Operation* OpState::operator->()` to provide more convenient access to members of Operation.
Given that OpState already implicit converts to Operator*, this seems reasonable.

The alternative would be to add more functions to OpState which forward to Operation.

Reviewed By: rriddle, ftynse

Differential Revision: https://reviews.llvm.org/D92266
2020-12-02 15:46:20 +01:00
Sean Silva 774f1d3ffd [mlir] Small cleanups to func-bufferize/finalizing-bufferize
- Address TODO in scf-bufferize: the argument materialization issue is
  now fixed and the code is now in Transforms/Bufferize.cpp
- Tighten up finalizing-bufferize to avoid creating invalid IR when
  operand types potentially change
- Tidy up the testing of func-bufferize, and move appropriate tests
  to a new finalizing-bufferize.mlir
- The new stricter checking in finalizing-bufferize revealed that we
  needed a DimOp conversion pattern (found when integrating into npcomp).
  Previously, the converion infrastructure was blindly changing the
  operand type during finalization, which happened to work due to
  DimOp's tensor/memref polymorphism, but is generally not encouraged
  (the new pattern is the way to tell the conversion infrastructure that
  it is legal to change that type).
2020-11-30 17:04:14 -08:00
Frederik Gossen 6484567f14 [MLIR][SCF] Find all innermost loops for parallel loop tiling
Overcome the assumption that parallel loops are only nested in other parallel
loops.

Differential Revision: https://reviews.llvm.org/D92188
2020-11-27 10:08:56 +01:00
Alex Zinenko 31a233d463 [mlir] canonicalize away zero-iteration SCF for loops
An SCF 'for' loop does not iterate if its lower bound is equal to its upper
bound. Remove loops where both bounds are the same SSA value as such bounds are
guaranteed to be equal. Similarly, remove 'parallel' loops where at least one
pair of respective lower/upper bounds is specified by the same SSA value.

Reviewed By: gysit

Differential Revision: https://reviews.llvm.org/D91880
2020-11-23 15:04:31 +01:00
Alex Zinenko 18d0f7d5c3 [mlir] add canonicalization patterns for trivial SCF 'for' and 'if'
Add canoncalization patterns to remove zero-iteration 'for' loops, replace
single-iteration 'for' loops with their bodies; remove known-false conditionals
with no 'else' branch and replace conditionals with known value by the
respective region. Although similar transformations are performed at the CFG
level, not all flows reach that level, e.g., the GPU flow may want to remove
single-iteration loops before deciding on loop mapping to thread dimensions.

Reviewed By: herhut

Differential Revision: https://reviews.llvm.org/D91865
2020-11-20 19:04:39 +01:00
River Riddle 65fcddff24 [mlir][BuiltinDialect] Resolve comments from D91571
* Move ops to a BuiltinOps.h
* Add file comments
2020-11-19 11:12:49 -08:00
River Riddle 73ca690df8 [mlir][NFC] Remove references to Module.h and Function.h
These includes have been deprecated in favor of BuiltinDialect.h, which contains the definitions of ModuleOp and FuncOp.

Differential Revision: https://reviews.llvm.org/D91572
2020-11-17 00:55:47 -08:00
Nicolas Vasilache 7625742237 [mlir][Linalg] Add support for tileAndDistribute on tensors.
scf.parallel is currently not a good fit for tiling on tensors.
Instead provide a path to parallelism directly through scf.for.
For now, this transformation ignores the distribution scheme and always does a block-cyclic mapping (where block is the tile size).

Differential revision: https://reviews.llvm.org/D90475
2020-11-16 11:12:50 +00:00
Eugene Zhulenev bb0d5f767d [mlir] Add NumberOfExecutions analysis + update RegionBranchOpInterface interface to query number of region invocations
Implements RFC discussed in: https://llvm.discourse.group/t/rfc-operationinstancesinterface-or-any-better-name/2158/10

Reviewed By: silvas, ftynse, rriddle

Differential Revision: https://reviews.llvm.org/D90922
2020-11-11 01:43:17 -08:00
Nicolas Vasilache f202d32216 [mlir][SCF] Add canonicalization pattern for scf::For to eliminate yields that just forward.
For instance:
```
func @for_yields_3(%lb : index, %ub : index, %step : index) -> (i32, i32, i32) {
  %a = call @make_i32() : () -> (i32)
  %b = call @make_i32() : () -> (i32)
  %r:3 = scf.for %i = %lb to %ub step %step iter_args(%0 = %a, %1 = %a, %2 = %b) -> (i32, i32, i32) {
    %c = call @make_i32() : () -> (i32)
    scf.yield %0, %c, %2 : i32, i32, i32
  }
  return %r#0, %r#1, %r#2 : i32, i32, i32
}
```

Canonicalizes as:
```
  func @for_yields_3(%arg0: index, %arg1: index, %arg2: index) -> (i32, i32, i32) {
    %0 = call @make_i32() : () -> i32
    %1 = call @make_i32() : () -> i32
    %2 = scf.for %arg3 = %arg0 to %arg1 step %arg2 iter_args(%arg4 = %0) -> (i32) {
      %3 = call @make_i32() : () -> i32
      scf.yield %3 : i32
    }
    return %0, %2, %1 : i32, i32, i32
  }
```

Differential Revision: https://reviews.llvm.org/D90745
2020-11-04 11:36:27 +00:00
Alex Zinenko 79716559b5 [mlir] Add a generic while/do-while loop to the SCF dialect
The new construct represents a generic loop with two regions: one executed
before the loop condition is verifier and another after that. This construct
can be used to express both a "while" loop and a "do-while" loop, depending on
where the main payload is located. It is intended as an intermediate
abstraction for lowering, which will be added later. This form is relatively
easy to target from higher-level abstractions and supports transformations such
as loop rotation and LICM.

Differential Revision: https://reviews.llvm.org/D90255
2020-11-04 09:43:13 +01:00
River Riddle fa4174792a [mlir][Inliner] Add a `wouldBeCloned` flag to each of the `isLegalToInline` hooks.
Often times the legality of inlining can change depending on if the callable is going to be inlined in-place, or cloned. For example, some operations are not allowed to be duplicated and can only be inlined if the original callable will cease to exist afterwards. The new `wouldBeCloned` flag allows for dialects to hook into this when determining legality.

Differential Revision: https://reviews.llvm.org/D90360
2020-10-28 21:49:28 -07:00
Sean Silva 1ce7040359 [mlir] Properly handle recursive bufferization for scf.for/scf.if
This fixes a subtle issue, described in the comment starting with
"Clone the op without the regions and inline the regions from the old op",
which prevented this conversion from working on non-trivial examples.

Differential Revision: https://reviews.llvm.org/D90203
2020-10-28 14:16:56 -07:00
River Riddle 3fffffa882 [mlir][Pattern] Add a new FrozenRewritePatternList class
This class represents a rewrite pattern list that has been frozen, and thus immutable. This replaces the uses of OwningRewritePatternList in pattern driver related API, such as dialect conversion. When PDL becomes more prevalent, this API will allow for optimizing a set of patterns once without the need to do this per run of a pass.

Differential Revision: https://reviews.llvm.org/D89104
2020-10-26 18:01:06 -07:00
Lei Zhang cb5ab3e90e [mlir] Add missing dependency for MLIRSCFTransforms
MLIRTransforms is needed to provide BufferizeTypeConverter
definitions.
2020-10-21 16:24:57 -04:00
Sean Silva f0292ede9b [mlir] Add structural type conversions for SCF dialect.
A "structural" type conversion is one where the underlying ops are
completely agnostic to the actual types involved and simply need to update
their types. An example of this is scf.if -- the scf.if op and the
corresponding scf.yield ops need to update their types accordingly to the
TypeConverter, but otherwise don't care what type conversions are happening.

To test the structural type conversions, it is convenient to define a
bufferize pass for a dialect, which exercises them nicely.

Differential Revision: https://reviews.llvm.org/D89757
2020-10-21 11:58:27 -07:00