Currently, the import of LLVMIR fails if the program contains debug
intrinsics. The revision adds support to import debug intrinsics that
have no debug expression attached and drops all debug intrinsics with a
non-empty debug expression. It also moves the existing debug intrinsics
into the "intr" namespace by deriving from LLVM_IntrOp.
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D138405
Instead of importing constant expressions recursively, the revision
walks all dependencies of an LLVM constant iteratively. The actual
conversion then iterates over a list of constants and all intermediate
constant values are added to the value mapping. As a result, an LLVM IR
constant maps to exactly one MLIR operation per function. The revision
adapts the existing tests since the constant ordering changed for
aggregate types. Additionally, it adds extra tests that mix aggregate
constants and constant expressions.
Depends on D137416
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D137559
Add a DebugImporter to convert LLVMIR debug metadata into
MLIR debug attributes. It is the counterpart to the
DebugTranslation class and supports the same attributes.
The revision only supports the translation of instruction,
function, and module debug information. The import of
intrinsics is left to a later revision.
Depends on D138206
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D138209
Insert constants and globals in order by maintaining the position
of the constant and global inserted last. Update the tests
to reflect the updated insertion order. Also make sure functions
are always inserted at the end of the module instead of at
the second last position and delete a spurious function in
the intrinsic.ll that seems to exist to avoid the first
function under test ends up at the end of the module.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D136679
This adds a '--no-implicit-module' option, which disables the insertion
of a top-level 'builtin.module' during parsing.
The translation APIs are also updated to take/return 'Operation*'
instead of 'ModuleOp', to allow other operation types to be used. To
simplify translations which are restricted to specific operation types,
'TranslateFromMLIRRegistration' has an overload which performs the
necessary cast and error checking.
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D134237
Rename single letter member variables and function arguments to use
longer names in ConvertFromLLVMIR.cpp. Also drop some uses of auto in
favor our spelling out the type and refactor some llvm::enumerate loops.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D136246
The revision performs a topological sort of the blocks to
ensure the operations are processed in dominance order.
After the change, we do not need to introduce dummy
instructions if an operand has not yet been processed.
Additionally, the revision also moves and simplifies the
control-flow related tests to a separate test file.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D136230
The revision adds support for importing the masked load/store and
gather/scatter intrinsics from LLVM IR. To enable the import, the
revision also includes an extension of the mlirBuilder code generation
to support variadic arguments.
Depends on D136057
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D136058
This allows for setting an attribute using the underlying C++ type,
which is generally much nicer to interact with than the attribute type.
Differential Revision: https://reviews.llvm.org/D135838
The revision imports the atomic operations using
tablegen generated builders. Additionally, it moves their tests to
the instructions.ll test file.
Depends on D135880
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D135944
The revision imports the shuffle vector operation using
tablegen generated builders. Additionally, it moves its test to
the instructions.ll test file.
Depends on D135874
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D135880
The revision imports the extract and insert value operations using
tablegen generated builders. Additionally, it moves the tests to
the instructions.ll test file.
Reviewed By: ftynse, dcaballe
Differential Revision: https://reviews.llvm.org/D135874
The revision imports compare operations using TableGen generated
builders, instead of using the special handlers defined by the Importer.
It therefore adds a new llvmArgIndexes field that allows to specify
a mapping between MLIR argument and LLVM IR operand indexes if they do
not match. Additionally, the FCmp op is extended with an additional
builder and all compare operations are extended with verification
traits to ensure the operands types match. These extensions simplify
the logic of the newly introduced builders and are in line with the
compare operations define by the arithmetic dialect.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D135855
The revision uses tablegen generated builders to convert the most common
LLVM IR instructions to MLIR LLVM dialect operations. All instructions
with special handlers, except for alloca and fence, still use manual
handlers. The revision also introduces an additional "instructions.ll"
test file to test the import of instructions that have tablegen builders
(except for the resume instruction whose test remains untouched). A part
of the test cases are new, for example the integer instruction test,
while others are migrated from the "basic.ll" test file.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D135709
The revision adds a mapValue function to the Importer, which can be used
in the MLIR builders to provide controlled accesses to the result
mapping of the imported instructions. Additionally, the change allows us
to avoid accessing a private member variable of the Importer class,
which simplifies future refactorings that aim at factoring out a
conversion interface (similar to the MLIR to LLVM translation). The
revision also renames the variables used when emitting the MLIR builders
to prepare the generalization to non-intrinsic instructions. In
particular, it renames callInst to inst and it passes in the instruction
arguments using an llvmOperands array rather than accessing the call
arguments directly.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D135645
The revision adds support to specify custom import functions for
LLVM IR intrinsics with immediate arguments that translate to MLIR
attributes. It takes an approach similar to the MLIR to LLVM translation
that uses a tablegen defined build method. The default implementation
of this newly introduced "mlirBuilder" assumes all intrinsic arguments
translate to operands. Specific intrinsics, such as
llvm.lifetime.start/stop then define a custom builder that converts
their immediate arguments to MLIR attributes.
Depends on D135349
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D135350
The revision renames some methods of the Importer and changes
the error handling to be closer the ModuleTranslation. In particular,
processValue -> lookupValue and processType -> convertType
now fail if the translation fails (instead of returning an error),
which simplifies the error handling.
The revision prepares a follow up commit that will import
LLVMIR intrinsics using tablegen.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D135349
The revision enriches the debug locations generated during LLVMIR to MLIR translation with file name information and adds a separate test to exercise the debug location translation.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D135069
If any of the operands for FCmpOp is a vector, returns a vector<Nxi1>,
rather than an i1 type result.
Differential Revision: https://reviews.llvm.org/D134449
Previously in mlir-translate, if debug info was absent in a
llvm::Instruction, we tried to create one using the name of its defined
value in a textual LLVM IR file as the (pseudo) debug file name.
However, in order to get that name, we need to call out to LLVM's
SlotTracker, which, surprisingly, took a lot of time. Judging from
the usefulness of such pseudo debug file name and the performance penalty
during translation, this patch simply use "imported-bitcode" as the
debug file name in these case. Eliminating the need of using (expensive)
LLVM value numbering.
Differential Revision: https://reviews.llvm.org/D134305
With the transition to opaque pointers, type information has been
transferred to function parameter attributes. This patch adds correct
parsing for some of those arguments and fixes some tests, that
previously used UnitAttr for those.
Differential Revision: https://reviews.llvm.org/D132366
The attribute is translated into LLVM's function attribute 'readnone'.
There is no explicit verification regarding conflicting 'readnone'
and function attributes from 'passthrough', though, LLVM would assert
if they are incompatible during LLVM IR creation.
Differential Revision: https://reviews.llvm.org/D131457
This patch moves `LLVM::ShuffleVectorOp` to assembly format and in the
process drops the extra type that can be inferred (both operand types
are required to be the same) and switches to a dense integer array.
The syntax change:
```
// Before
%0 = llvm.shufflevector %0, %1 [0 : i32, 0 : i32, 0 : i32, 0 : i32] : vector<4xf32>, vector<4xf32>
// After
%0 = llvm.shufflevector %0, %1 [0, 0, 0, 0] : vector<4xf32>
```
Reviewed By: dcaballe
Differential Revision: https://reviews.llvm.org/D132038
This reland includes changes to the Python bindings.
Switch variadic operand and result segment size attributes to use the
dense i32 array. Dense integer arrays were introduced primarily to
represent index lists. They are a better fit for segment sizes than
dense elements attrs.
Depends on D131801
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D131803
Switch variadic operand and result segment size attributes to use the
dense i32 array. Dense integer arrays were introduced primarily to
represent index lists. They are a better fit for segment sizes than
dense elements attrs.
Depends on D131738
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D131702
This patch "modernizes" the LLVM `insertvalue` and `extractvalue`
operations to use DenseI64ArrayAttr, since they only require an array of
indices and previously there was confusion about whether to use i32 or
i64 arrays, and to use assembly format.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D131537
The implementation and API of GEP Op has gotten a bit convoluted over the time. Issues with it are:
* Misleading naming: `indices` actually only contains the dynamic indices, not all of them. To get the amount of indices you need to query the size of `structIndices`
* Very difficult to iterate over all indices properly: One had to iterate over `structIndices`, check whether it contains the magic constant `kDynamicIndex`, if it does, access the next value in `index` etc.
* Inconvenient to build: One either has create lots of constant ops for every index or have an odd split of passing both a `ValueRange` as well as a `ArrayRef<int32_t>` filled with `kDynamicIndex` at the correct places.
* Implementation doing verification in the build method
and more.
This patch attempts to address all these issues via convenience classes and reworking the way GEP Op works:
* Adds `GEPArg` class which is a sum type of a `int32_t` and `Value` and is used to have a single convenient easy to use `ArrayRef<GEPArg>` in the builders instead of the previous `ValueRange` + `ArrayRef<int32_t>` builders.
* Adds `GEPIndicesAdapter` which is a class used for easy random access and iteration over the indices of a GEP. It is generic and flexible enough to also instead return eg. a corresponding `Attribute` for an operand inside of `fold`.
* Rename `structIndices` to `rawConstantIndices` and `indices` to `dynamicIndices`: `rawConstantIndices` signifies one shouldn't access it directly as it is encoded, and `dynamicIndices` is more accurate and also frees up the `indices` name.
* Add `getIndices` returning a `GEPIndicesAdapter` to easily iterate over the GEP Ops indices.
* Move the verification/asserts out of the build method and into the `verify` method emitting op error messages.
* Add convenient builder methods making use of `GEPArg`.
* Add canonicalizer turning dynamic indices with constant values into constant indices to have a canonical representation.
The only breaking change is for any users building GEPOps that have so far used the old `ValueRange` + `ArrayRef<int32_t>` builder as well as those using the generic syntax.
Another follow up patch then goes through upstream and makes use of the new `ArrayRef<GEPArg>` builder to remove a lot of code building constants for GEP indices.
Differential Revision: https://reviews.llvm.org/D130730
When translating from a llvm::ConstantAggregate with vector type, we
should lower to insertelement operations (if needed) rather than using
insertvalue.
Differential Revision: https://reviews.llvm.org/D127534
This patch supports to convert the llvm intrinsic to the corresponding op. It still leaves some intrinsics to be handled specially.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D126639
This patch adds support for Calling Convention attribute in LLVM
dialect, including enums, custom syntax and import from LLVM IR.
Additionally fix import of dso_local attribute.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D126161
Add support for translating from llvm::Select, llvm::FNeg, and llvm::Unreachable.
This patch also cleans up (NFC) the opcode map for simple instructions and
adds `// clang-format off/on` comments to prevent those lines from being
churned by clang-format between commits.
Differential Revision: https://reviews.llvm.org/D125817
Previously, GEPOp relies on `findKnownStructIndices` to check if a GEP
index should be static. The truth is, `findKnownStructIndices` can only
tell you a GEP index _might_ be indexing into a struct (which should use
a static GEP index). But GEPOp::build and GEPOp::verify are falsely
taking this information as a certain answer, which creates many false
alarms like the one depicted in
`test/Target/LLVMIR/Import/dynamic-gep-index.ll`.
The solution presented here adopts a new verification scheme: When we're
recursively checking the child element types of a struct type, instead
of checking every child types, we only check the one dictated by the
(static) GEP index value. We also combine "refinement" logics --
refine/promote struct index mlir::Value into constants -- into the very
verification process since they have lots of logics in common. The
resulting code is more concise and less brittle.
We also hide GEPOp::findKnownStructIndices since most of the
aforementioned logics are already encapsulated within GEPOp::build and
GEPOp::verify, we found little reason for findKnownStructIndices (or the
new findStructIndices) to be public.
Differential Revision: https://reviews.llvm.org/D124935
Inside processInstruction, we assign the translated mlir::Value to a
reference previously taken from the corresponding entry in instMap.
However, instMap (a DenseMap) might resize after the entry reference was
taken, rendering the assignment useless since it's assigning to a
dangling reference. Here is a (pseudo) snippet that shows the concept:
```
// inst has type llvm::Instruction *
Value &v = instMap[inst];
...
// op is one of the operands of inst, has type llvm::Value *
processValue(op);
// instMap resizes inside processValue
...
translatedValue = b.createOp<Foo>(...);
// v is already a dangling reference at this point!
// The following assignment is bogus.
v = translatedValue;
```
Nevertheless, after we stop caching llvm::Constant into instMap, there
is only one case that can cause processValue to resize instMap: If the
operand is a llvm::ConstantExpr. In which case we will insert the
derived llvm::Instruction into instMap.
To trigger instMap to resize, which is a DenseMap, the threshold depends
on the ratio between # of map entries and # of (hash) buckets. More specifically,
it resizes if (# of map entries / # of buckets) >= 0.75.
In this case # of map entries is equal to # of LLVM instructions, and # of
buckets is the power-of-two upperbound of # of map entries. Thus, eventually
in the attaching test case (test/Target/LLVMIR/Import/incorrect-instmap-assignment.ll),
we picked 96 and 128 for the # of map entries and # of buckets, respectively.
(We can't pick numbers that are too small since DenseMap used inlined
storage for small number of entries). Therefore, the ConstantExpr in the
said test case (i.e. a GEP) is the 96-th llvm::Value cached into the
instMap, triggering the issue we're discussing here on its enclosing
instruction (i.e. a load).
This patch fixes this issue by calling `operator[]` everytime we need to
update an entry.
Differential Revision: https://reviews.llvm.org/D124627
This patch add supports for translating FCmp and more kinds of FP
constants in addition to 32 & 64-bit ones. However, we can't express
ppc_fp128 constants right now because the semantics for its underlying
APFloat is `S_PPCDoubleDouble` but mlir::FloatType doesn't support such
semantics right now.
Differential Revision: https://reviews.llvm.org/D124630
Constants in MLIR are not globally unique, unlike that in LLVM IR.
Therefore, reusing previous-translated constants might cause the user
operations not being dominated by the constant (because the
previous-translated ones can be placed in arbitrary place)
This indeed misses some opportunities where we actually can reuse a
previous-translated constants, but verbosity is not our first priority
here.
Differential Revision: https://reviews.llvm.org/D124404
More specifically, the llvm::Instruction generated by
llvm::ConstantExpr::getAsInstruction. Such Instruction will be deleted
right away, but it's possible that when getAsInstruction is called
again, it will create a new Instruction that has the same address with
the one we just deleted. Thus, we shouldn't keep it in the `instMap` to
avoid a conflicting index that triggers an assertion in
processInstruction.
Differential Revision: https://reviews.llvm.org/D124402
And move importer test files from `test/Target/LLVMIR` into
`test/Target/LLVMIR/Import`.
We simply translate struct-type ConstantAggregate(Zero) into a
serious of `llvm.insertvalue` operations against a `llvm.undef` root.
Note that this doesn't affect the original logics on translating
vector/array-type ConstantAggregate values.
Differential Revision: https://reviews.llvm.org/D124399
LLVM IR is moving towards adoption of opaque pointer types. These require extra
information to be passed when constructing some operations, in particular GEP
and Alloca. Adapt the builders of said operations and modify the translation
code to handle both opaque and non-opaque pointers.
This incidentally adds the translation for Alloca alignment and fixes the translation
of struct-related GEP indices that must be constant.
Reviewed By: wsmoses
Differential Revision: https://reviews.llvm.org/D123792
