In dialect conversion infrastructure, source materialization applies as part of
the finalization procedure to results of the newly produced operations that
replace previously existing values with values having a different type.
However, such operations may be created to replace operations created in other
patterns. At this point, it is possible that the results of the _original_
operation are still in use and have mismatching types, but the results of the
_intermediate_ operation that performed the type change are not in use leading
to the absence of source materialization. For example,
%0 = dialect.produce : !dialect.A
dialect.use %0 : !dialect.A
can be replaced with
%0 = dialect.other : !dialect.A
%1 = dialect.produce : !dialect.A // replaced, scheduled for removal
dialect.use %1 : !dialect.A
and then with
%0 = dialect.final : !dialect.B
%1 = dialect.other : !dialect.A // replaced, scheduled for removal
%2 = dialect.produce : !dialect.A // replaced, scheduled for removal
dialect.use %2 : !dialect.A
in the same rewriting, but only the %1->%0 replacement is currently considered.
Change the logic in dialect conversion to look up all values that were replaced
by the given value and performing source materialization if any of those values
is still in use with mismatching types. This is performed by computing the
inverse value replacement mapping. This arguably expensive manipulation is
performed only if there were some type-changing replacements. An alternative
could be to consider all replaced operations and not only those that resulted
in type changes, but it would harm pattern-level composability: the pattern
that performed the non-type-changing replacement would have to be made aware of
the type converter in order to call the materialization hook.
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D95626
We could extend this with an interface to allow dialect to perform a type
conversion, but that would make the folder creating operation which isn't
the case at the moment, and isn't necessarily always desirable.
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D95991
In dialect conversion, signature conversions essentially perform block argument
replacement and are added to the general value remapping. However, the replaced
values were not tracked, so if a signature conversion was rolled back, the
construction of operand lists for the following patterns could have obtained
block arguments from the mapping and give them to the pattern leading to
use-after-free. Keep track of signature conversions similarly to normal block
argument replacement, and erase such replacements from the general mapping when
the conversion is rolled back.
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D95688
Currently, for a scf.parallel (i,j,k) after the loop collapsing to 1D is done, the
IVs would be traversed as for an scf.parallel(k,j,i).
Differential Revision: https://reviews.llvm.org/D95693
This patch adds support for producer-consumer fusion scenarios with
multiple producer stores to the AffineLoopFusion pass. The patch
introduces some changes to the producer-consumer algorithm, including:
* For a given consumer loop, producer-consumer fusion iterates over its
producer candidates until a fixed point is reached.
* Producer candidates are gathered beforehand for each iteration of the
consumer loop and visited in reverse program order (not strictly guaranteed)
to maximize the number of loops fused per iteration.
In general, these changes were needed to simplify the multi-store producer
support and remove some of the workarounds that were introduced in the past
to support more fusion cases under the single-store producer limitation.
This patch also preserves the existing functionality of AffineLoopFusion with
one minor change in behavior. Producer-consumer fusion didn't fuse scenarios
with escaping memrefs and multiple outgoing edges (from a single store).
Multi-store producer scenarios will usually (always?) have multiple outgoing
edges so we couldn't fuse any with escaping memrefs, which would greatly limit
the applicability of this new feature. Therefore, the patch enables fusion for
these scenarios. Please, see modified tests for specific details.
Reviewed By: andydavis1, bondhugula
Differential Revision: https://reviews.llvm.org/D92876
This patch adds support for producer-consumer fusion scenarios with
multiple producer stores to the AffineLoopFusion pass. The patch
introduces some changes to the producer-consumer algorithm, including:
* For a given consumer loop, producer-consumer fusion iterates over its
producer candidates until a fixed point is reached.
* Producer candidates are gathered beforehand for each iteration of the
consumer loop and visited in reverse program order (not strictly guaranteed)
to maximize the number of loops fused per iteration.
In general, these changes were needed to simplify the multi-store producer
support and remove some of the workarounds that were introduced in the past
to support more fusion cases under the single-store producer limitation.
This patch also preserves the existing functionality of AffineLoopFusion with
one minor change in behavior. Producer-consumer fusion didn't fuse scenarios
with escaping memrefs and multiple outgoing edges (from a single store).
Multi-store producer scenarios will usually (always?) have multiple outgoing
edges so we couldn't fuse any with escaping memrefs, which would greatly limit
the applicability of this new feature. Therefore, the patch enables fusion for
these scenarios. Please, see modified tests for specific details.
Reviewed By: andydavis1, bondhugula
Differential Revision: https://reviews.llvm.org/D92876
Add a check if regions do not implement the RegionBranchOpInterface. This is not
allowed in the current deallocation steps. Furthermore, we handle edge-cases,
where a single region is attached and the parent operation has no results.
This fixes: https://bugs.llvm.org/show_bug.cgi?id=48575
Differential Revision: https://reviews.llvm.org/D94586
The standard and gpu dialect both have `alloc` operations which use the
memory effect `MemAlloc`. In both cases, it is specified on both the
operation itself and on the result. This results in two memory effects
being created for these operations. When `MemAlloc` is defined on an
operation, it represents some background effect which the compiler
cannot reason about, and inhibits the ability of the compiler to
remove dead `std.alloc` operations. This change removes the uneeded
`MemAlloc` effect from these operations and leaves the effect on the
result, which allows dead allocs to be erased.
There is the same problem, but to a lesser extent, with MemFree, MemRead
and MemWrite. Over-specifying these traits is not currently inhibiting
any optimization.
Differential Revision: https://reviews.llvm.org/D94662
This revision adds a new `replaceOpWithIf` hook that replaces uses of an operation that satisfy a given functor. If all uses are replaced, the operation gets erased in a similar manner to `replaceOp`. DialectConversion support will be added in a followup as this requires adjusting how replacements are tracked there.
Differential Revision: https://reviews.llvm.org/D94632
In the overwhelmingly common case, enum attribute case strings represent valid identifiers in MLIR syntax. This revision updates the format generator to format as a keyword in these cases, removing the need to wrap values in a string. The parser still retains the ability to parse the string form, but the printer will use the keyword form when applicable.
Differential Revision: https://reviews.llvm.org/D94575
The LLVM dialect type system has been closed until now, i.e. did not support
types from other dialects inside containers. While this has had obvious
benefits of deriving from a common base class, it has led to some simple types
being almost identical with the built-in types, namely integer and floating
point types. This in turn has led to a lot of larger-scale complexity: simple
types must still be converted, numerous operations that correspond to LLVM IR
intrinsics are replicated to produce versions operating on either LLVM dialect
or built-in types leading to quasi-duplicate dialects, lowering to the LLVM
dialect is essentially required to be one-shot because of type conversion, etc.
In this light, it is reasonable to trade off some local complexity in the
internal implementation of LLVM dialect types for removing larger-scale system
complexity. Previous commits to the LLVM dialect type system have adapted the
API to support types from other dialects.
Replace LLVMIntegerType with the built-in IntegerType plus additional checks
that such types are signless (these are isolated in a utility function that
replaced `isa<LLVMType>` and in the parser). Temporarily keep the possibility
to parse `!llvm.i32` as a synonym for `i32`, but add a deprecation notice.
Reviewed By: mehdi_amini, silvas, antiagainst
Differential Revision: https://reviews.llvm.org/D94178
Now that passes have support for running nested pipelines, the inliner can now allow for users to provide proper nested pipelines to use for optimization during inlining. This revision also changes the behavior of optimization during inlining to optimize before attempting to inline, which should lead to a more accurate cost model and prevents the need for users to schedule additional duplicate cleanup passes before/after the inliner that would already be run during inlining.
Differential Revision: https://reviews.llvm.org/D91211
This reverts commit 0d48d265db.
This reapplies the following commit, with a fix for CAPI/ir.c:
[mlir] Start splitting the `tensor` dialect out of `std`.
This starts by moving `std.extract_element` to `tensor.extract` (this
mirrors the naming of `vector.extract`).
Curiously, `std.extract_element` supposedly works on vectors as well,
and this patch removes that functionality. I would tend to do that in
separate patch, but I couldn't find any downstream users relying on
this, and the fact that we have `vector.extract` made it seem safe
enough to lump in here.
This also sets up the `tensor` dialect as a dependency of the `std`
dialect, as some ops that currently live in `std` depend on
`tensor.extract` via their canonicalization patterns.
Part of RFC: https://llvm.discourse.group/t/rfc-split-the-tensor-dialect-from-std/2347/2
Differential Revision: https://reviews.llvm.org/D92991
This starts by moving `std.extract_element` to `tensor.extract` (this
mirrors the naming of `vector.extract`).
Curiously, `std.extract_element` supposedly works on vectors as well,
and this patch removes that functionality. I would tend to do that in
separate patch, but I couldn't find any downstream users relying on
this, and the fact that we have `vector.extract` made it seem safe
enough to lump in here.
This also sets up the `tensor` dialect as a dependency of the `std`
dialect, as some ops that currently live in `std` depend on
`tensor.extract` via their canonicalization patterns.
Part of RFC: https://llvm.discourse.group/t/rfc-split-the-tensor-dialect-from-std/2347/2
Differential Revision: https://reviews.llvm.org/D92991
This fixes a subtle bug where SCCP could incorrectly optimize a private callable while waiting for its arguments to be resolved.
Fixes PR#48457
Differential Revision: https://reviews.llvm.org/D92976
Memrefs with affine_map in the results of normalizable operation were
not normalized by `--normalize-memrefs` option. This patch normalizes
them.
Differential Revision: https://reviews.llvm.org/D88719
Extended promote buffers to stack pass to support dynamically shaped allocas.
The conversion is limited by the rank of the underlying tensor.
An option is added to the pass to adjust the given rank.
Differential Revision: https://reviews.llvm.org/D91969
- 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).
The rewrite logic has an optimization to drop a cast operation after
rewriting block arguments if the cast operation has no users. This is
unsafe as there might be a pending rewrite that replaced the cast operation
itself and hence would trigger a second free.
Instead, do not remove the casts and leave it up to a later canonicalization
to do so.
Differential Revision: https://reviews.llvm.org/D92184
Block merging in MLIR will incorrectly merge blocks with operations whose values are used outside of that block. This change forbids this behavior and provides a test where it is illegal to perform such a merge.
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D91745
This replaces the old type decomposition logic that was previously mixed
into bufferization, and makes it easily accessible.
This also deletes TestFinalizingBufferize, because after we remove the type
decomposition, it doesn't do anything that is not already provided by
func-bufferize.
Differential Revision: https://reviews.llvm.org/D90899
The index type does not have a bitsize and hence the size of corresponding allocations cannot be computed. Instead, the promotion pass now has an explicit option to specify the size of index.
Differential Revision: https://reviews.llvm.org/D91360
The previous logic for inlining a region A with N blocks into region B
would produce incorrect results on rollback for N greater than 1. This
rollback logic would leave blocks 1..N in region B and only move block 0
to region A.
The new inlining action recording stores the block move actions from N-1
to 0. Now on roll back, block 0 is moved to region A and then 1..N is
appended to the list of blocks in region A.
Differential Revision: https://reviews.llvm.org/D91185
Locations often get very long and clutter up operations when printed inline with them. This revision adds support for using aliases with trailing operation locations, and makes printing with aliases the default behavior. Aliases in the trailing location take the form `loc(<alias>)`, such as `loc(#loc0)`. As with all aliases, using `mlir-print-local-scope` can be used to disable them and get the inline behavior.
Differential Revision: https://reviews.llvm.org/D90652
This revision refactors the way that attributes/types are considered when generating aliases. Instead of considering all of the attributes/types of every operation, we perform a "fake" print step that prints the operations using a dummy printer to collect the attributes and types that would actually be printed during the real process. This removes a lot of attributes/types from consideration that generally won't end up in the final output, e.g. affine map attributes in an `affine.apply`/`affine.for`.
This resolves a long standing TODO w.r.t aliases, and helps to have a much cleaner textual output format. As a datapoint to the latter, as part of this change several tests were identified as testing for the presence of attributes aliases that weren't actually referenced by the custom form of any operation.
To ensure that this wouldn't cause a large degradation in compile time due to the second full print, I benchmarked this change on a very large module with a lot of operations(The file is ~673M/~4.7 million lines long). This file before this change take ~6.9 seconds to print in the custom form, and ~7 seconds after this change. In the custom assembly case, this added an average of a little over ~100 miliseconds to the compile time. This increase was due to the way that argument attributes on functions are structured and how they get printed; i.e. with a better representation the negative impact here can be greatly decreased. When printing in the generic form, this revision had no observable impact on the compile time. This benchmarking leads me to believe that the impact of this change on compile time w.r.t printing is closely related to `print` methods that perform a lot of additional/complex processing outside of the OpAsmPrinter.
Differential Revision: https://reviews.llvm.org/D90512
- Change syntax for FuncOp to be `func <visibility>? @name` instead of printing the
visibility in the attribute dictionary.
- Since printFunctionLikeOp() and parseFunctionLikeOp() are also used by other
operations, make the "inline visibility" an opt-in feature.
- Updated unit test to use and check the new syntax.
Differential Revision: https://reviews.llvm.org/D90859
There exists a generic folding facility that folds the operand of a memref_cast
into users of memref_cast that support this. However, it was not used for the
memref_cast itself. Fix it to enable elimination of memref_cast chains such as
%1 = memref_cast %0 : A to B
%2 = memref_cast %1 : B to A
that is achieved by combining the folding with the existing "A to A" cast
elimination.
Reviewed By: nicolasvasilache
Differential Revision: https://reviews.llvm.org/D90910
This functionality is superceded by BufferResultsToOutParams pass (see
https://reviews.llvm.org/D90071) for users the require buffers to be
out-params. That pass should be run immediately after all tensors are gone from
the program (before buffer optimizations and deallocation insertion), such as
immediately after a "finalizing" bufferize pass.
The -test-finalizing-bufferize pass now defaults to what used to be the
`allowMemrefFunctionResults=true` flag. and the
finalizing-bufferize-allowed-memref-results.mlir file is moved
to test/Transforms/finalizing-bufferize.mlir.
Differential Revision: https://reviews.llvm.org/D90778
The LinalgDependenceGraph and alias analysis provide the necessary analysis for the Linalg fusion on buffers case.
However this is not enough for linalg on tensors which require proper memory effects to play nicely with DCE and other transformations.
This revision adds side effects to Linalg ops that were previously missing and has 2 consequences:
1. one example in the copy removal pass now fails since the linalg.generic op has side effects and the pass does not perform alias analysis / distinguish between reads and writes.
2. a few examples in fusion-tensor.mlir need to return the resulting tensor otherwise DCE automatically kicks in as part of greedy pattern application.
Differential Revision: https://reviews.llvm.org/D90762
BufferizeTests.
Summary:
Added test operations to replace the LinalgDialect dependency in tests
which use the buffer-deallocation, buffer-hoisting,
buffer-loop-hoisting, promote-buffers-to-stack,
buffer-placement-preparation-allowed-memref-resutls and
buffer-placement-preparation pass. Adapted the corresponding tests cases
and TestBufferPlacement.cpp.
Differential Revision: https://reviews.llvm.org/D90037
BufferPlacement is no longer part of bufferization. However, this test
is an important test of "finalizing" bufferize passes.
A "finalizing" bufferize conversion is one that performs a "full"
conversion and expects all tensors to be gone from the program. This in
particular involves rewriting funcs (including block arguments of the
contained region), calls, and returns. The unique property of finalizing
bufferization passes is that they cannot be done via a local
transformation with suitable materializations to ensure composability
(as other bufferization passes do). For example, if a call is
rewritten, the callee needs to be rewritten otherwise the IR will end up
invalid. Thus, finalizing bufferization passes require an atomic change
to the entire program (e.g. the whole module).
This new designation makes it clear also that it shouldn't be testing
bufferization of linalg ops, so the tests have been updated to not use
linalg.generic ops. (linalg.copy is still used as the "copy" op for
copying into out-params)
Differential Revision: https://reviews.llvm.org/D89979
This pass allows removing getResultConversionKind from
BufferizeTypeConverter. This pass replaces the AppendToArgumentsList
functionality. As far as I could tell, the only use of this functionlity
is to perform the transformation that is implemented in this pass.
Future patches will remove the getResultConversionKind machinery from
BufferizeTypeConverter, but sending this patch for individual review for
clarity.
Differential Revision: https://reviews.llvm.org/D90071
Previously they were separated into "instance" and "kind" aliases, and also required that the dialect know ahead of time all of the instances that would have a corresponding alias. This approach was very clunky and not ergonomic to interact with. The new approach is to provide the dialect with an instance of an attribute/type to provide an alias for, fully replacing the original split approach.
Differential Revision: https://reviews.llvm.org/D89354
In certain situations it isn't legal to inline a call operation, but this isn't something that is possible(at least not easily) to prevent with the current hooks. This revision adds a new hook so that dialects with call operations that shouldn't be inlined can prevent it.
Differential Revision: https://reviews.llvm.org/D90359
Added optimization pass to convert heap-based allocs to stack-based allocas in
buffer placement. Added the corresponding test file.
Differential Revision: https://reviews.llvm.org/D89688
The current BufferPlacement transformation contains several concepts for
hoisting allocations. However, more advanced hoisting techniques should not be
integrated into the BufferPlacement transformation. Hence, this CL refactors the
current BufferPlacement pass into three separate pieces: BufferDeallocation and
BufferAllocation(Loop)Hoisting. Moreover, it extends the hoisting functionality
by allowing to move allocations out of loops.
Differential Revision: https://reviews.llvm.org/D87756
Parsing of a scalar subview did not create the required static_offsets attribute.
This also adds support for folding scalar subviews away.
Differential Revision: https://reviews.llvm.org/D89467
