Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

395 lines
14 KiB
C++
Raw Normal View History

//===- IR.cpp - C Interface for Core MLIR APIs ----------------------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
#include "mlir-c/IR.h"
#include "mlir/CAPI/IR.h"
#include "mlir/IR/Attributes.h"
Separate the Registration from Loading dialects in the Context This changes the behavior of constructing MLIRContext to no longer load globally registered dialects on construction. Instead Dialects are only loaded explicitly on demand: - the Parser is lazily loading Dialects in the context as it encounters them during parsing. This is the only purpose for registering dialects and not load them in the context. - Passes are expected to declare the dialects they will create entity from (Operations, Attributes, or Types), and the PassManager is loading Dialects into the Context when starting a pipeline. This changes simplifies the configuration of the registration: a compiler only need to load the dialect for the IR it will emit, and the optimizer is self-contained and load the required Dialects. For example in the Toy tutorial, the compiler only needs to load the Toy dialect in the Context, all the others (linalg, affine, std, LLVM, ...) are automatically loaded depending on the optimization pipeline enabled. To adjust to this change, stop using the existing dialect registration: the global registry will be removed soon. 1) For passes, you need to override the method: virtual void getDependentDialects(DialectRegistry &registry) const {} and registery on the provided registry any dialect that this pass can produce. Passes defined in TableGen can provide this list in the dependentDialects list field. 2) For dialects, on construction you can register dependent dialects using the provided MLIRContext: `context.getOrLoadDialect<DialectName>()` This is useful if a dialect may canonicalize or have interfaces involving another dialect. 3) For loading IR, dialect that can be in the input file must be explicitly registered with the context. `MlirOptMain()` is taking an explicit registry for this purpose. See how the standalone-opt.cpp example is setup: mlir::DialectRegistry registry; registry.insert<mlir::standalone::StandaloneDialect>(); registry.insert<mlir::StandardOpsDialect>(); Only operations from these two dialects can be in the input file. To include all of the dialects in MLIR Core, you can populate the registry this way: mlir::registerAllDialects(registry); 4) For `mlir-translate` callback, as well as frontend, Dialects can be loaded in the context before emitting the IR: context.getOrLoadDialect<ToyDialect>() Differential Revision: https://reviews.llvm.org/D85622
2020-08-18 20:01:19 +00:00
#include "mlir/IR/Dialect.h"
#include "mlir/IR/Module.h"
#include "mlir/IR/Operation.h"
#include "mlir/IR/Types.h"
#include "mlir/Parser.h"
#include "llvm/Support/raw_ostream.h"
using namespace mlir;
/* ========================================================================== */
/* Printing helper. */
/* ========================================================================== */
namespace {
/// A simple raw ostream subclass that forwards write_impl calls to the
/// user-supplied callback together with opaque user-supplied data.
class CallbackOstream : public llvm::raw_ostream {
public:
CallbackOstream(std::function<void(const char *, intptr_t, void *)> callback,
void *opaqueData)
: callback(callback), opaqueData(opaqueData), pos(0u) {}
void write_impl(const char *ptr, size_t size) override {
callback(ptr, size, opaqueData);
pos += size;
}
uint64_t current_pos() const override { return pos; }
private:
std::function<void(const char *, intptr_t, void *)> callback;
void *opaqueData;
uint64_t pos;
};
} // end namespace
/* ========================================================================== */
/* Context API. */
/* ========================================================================== */
MlirContext mlirContextCreate() {
Separate the Registration from Loading dialects in the Context This changes the behavior of constructing MLIRContext to no longer load globally registered dialects on construction. Instead Dialects are only loaded explicitly on demand: - the Parser is lazily loading Dialects in the context as it encounters them during parsing. This is the only purpose for registering dialects and not load them in the context. - Passes are expected to declare the dialects they will create entity from (Operations, Attributes, or Types), and the PassManager is loading Dialects into the Context when starting a pipeline. This changes simplifies the configuration of the registration: a compiler only need to load the dialect for the IR it will emit, and the optimizer is self-contained and load the required Dialects. For example in the Toy tutorial, the compiler only needs to load the Toy dialect in the Context, all the others (linalg, affine, std, LLVM, ...) are automatically loaded depending on the optimization pipeline enabled. To adjust to this change, stop using the existing dialect registration: the global registry will be removed soon. 1) For passes, you need to override the method: virtual void getDependentDialects(DialectRegistry &registry) const {} and registery on the provided registry any dialect that this pass can produce. Passes defined in TableGen can provide this list in the dependentDialects list field. 2) For dialects, on construction you can register dependent dialects using the provided MLIRContext: `context.getOrLoadDialect<DialectName>()` This is useful if a dialect may canonicalize or have interfaces involving another dialect. 3) For loading IR, dialect that can be in the input file must be explicitly registered with the context. `MlirOptMain()` is taking an explicit registry for this purpose. See how the standalone-opt.cpp example is setup: mlir::DialectRegistry registry; registry.insert<mlir::standalone::StandaloneDialect>(); registry.insert<mlir::StandardOpsDialect>(); Only operations from these two dialects can be in the input file. To include all of the dialects in MLIR Core, you can populate the registry this way: mlir::registerAllDialects(registry); 4) For `mlir-translate` callback, as well as frontend, Dialects can be loaded in the context before emitting the IR: context.getOrLoadDialect<ToyDialect>() Differential Revision: https://reviews.llvm.org/D85622
2020-08-18 20:01:19 +00:00
auto *context = new MLIRContext(/*loadAllDialects=*/false);
return wrap(context);
}
int mlirContextEqual(MlirContext ctx1, MlirContext ctx2) {
return unwrap(ctx1) == unwrap(ctx2);
}
void mlirContextDestroy(MlirContext context) { delete unwrap(context); }
/* ========================================================================== */
/* Location API. */
/* ========================================================================== */
MlirLocation mlirLocationFileLineColGet(MlirContext context,
const char *filename, unsigned line,
unsigned col) {
return wrap(FileLineColLoc::get(filename, line, col, unwrap(context)));
}
MlirLocation mlirLocationUnknownGet(MlirContext context) {
return wrap(UnknownLoc::get(unwrap(context)));
}
void mlirLocationPrint(MlirLocation location, MlirStringCallback callback,
void *userData) {
CallbackOstream stream(callback, userData);
unwrap(location).print(stream);
stream.flush();
}
/* ========================================================================== */
/* Module API. */
/* ========================================================================== */
MlirModule mlirModuleCreateEmpty(MlirLocation location) {
return wrap(ModuleOp::create(unwrap(location)));
}
MlirModule mlirModuleCreateParse(MlirContext context, const char *module) {
OwningModuleRef owning = parseSourceString(module, unwrap(context));
if (!owning)
return MlirModule{nullptr};
return MlirModule{owning.release().getOperation()};
}
void mlirModuleDestroy(MlirModule module) {
// Transfer ownership to an OwningModuleRef so that its destructor is called.
OwningModuleRef(unwrap(module));
}
MlirOperation mlirModuleGetOperation(MlirModule module) {
return wrap(unwrap(module).getOperation());
}
/* ========================================================================== */
/* Operation state API. */
/* ========================================================================== */
MlirOperationState mlirOperationStateGet(const char *name, MlirLocation loc) {
MlirOperationState state;
state.name = name;
state.location = loc;
state.nResults = 0;
state.results = nullptr;
state.nOperands = 0;
state.operands = nullptr;
state.nRegions = 0;
state.regions = nullptr;
state.nSuccessors = 0;
state.successors = nullptr;
state.nAttributes = 0;
state.attributes = nullptr;
return state;
}
#define APPEND_ELEMS(type, sizeName, elemName) \
state->elemName = \
(type *)realloc(state->elemName, (state->sizeName + n) * sizeof(type)); \
memcpy(state->elemName + state->sizeName, elemName, n * sizeof(type)); \
state->sizeName += n;
void mlirOperationStateAddResults(MlirOperationState *state, intptr_t n,
MlirType *results) {
APPEND_ELEMS(MlirType, nResults, results);
}
void mlirOperationStateAddOperands(MlirOperationState *state, intptr_t n,
MlirValue *operands) {
APPEND_ELEMS(MlirValue, nOperands, operands);
}
void mlirOperationStateAddOwnedRegions(MlirOperationState *state, intptr_t n,
MlirRegion *regions) {
APPEND_ELEMS(MlirRegion, nRegions, regions);
}
void mlirOperationStateAddSuccessors(MlirOperationState *state, intptr_t n,
MlirBlock *successors) {
APPEND_ELEMS(MlirBlock, nSuccessors, successors);
}
void mlirOperationStateAddAttributes(MlirOperationState *state, intptr_t n,
MlirNamedAttribute *attributes) {
APPEND_ELEMS(MlirNamedAttribute, nAttributes, attributes);
}
/* ========================================================================== */
/* Operation API. */
/* ========================================================================== */
MlirOperation mlirOperationCreate(const MlirOperationState *state) {
assert(state);
OperationState cppState(unwrap(state->location), state->name);
SmallVector<Type, 4> resultStorage;
SmallVector<Value, 8> operandStorage;
SmallVector<Block *, 2> successorStorage;
cppState.addTypes(unwrapList(state->nResults, state->results, resultStorage));
cppState.addOperands(
unwrapList(state->nOperands, state->operands, operandStorage));
cppState.addSuccessors(
unwrapList(state->nSuccessors, state->successors, successorStorage));
cppState.attributes.reserve(state->nAttributes);
for (intptr_t i = 0; i < state->nAttributes; ++i)
cppState.addAttribute(state->attributes[i].name,
unwrap(state->attributes[i].attribute));
for (intptr_t i = 0; i < state->nRegions; ++i)
cppState.addRegion(std::unique_ptr<Region>(unwrap(state->regions[i])));
MlirOperation result = wrap(Operation::create(cppState));
free(state->results);
free(state->operands);
free(state->successors);
free(state->regions);
free(state->attributes);
return result;
}
void mlirOperationDestroy(MlirOperation op) { unwrap(op)->erase(); }
int mlirOperationIsNull(MlirOperation op) { return unwrap(op) == nullptr; }
intptr_t mlirOperationGetNumRegions(MlirOperation op) {
return static_cast<intptr_t>(unwrap(op)->getNumRegions());
}
MlirRegion mlirOperationGetRegion(MlirOperation op, intptr_t pos) {
return wrap(&unwrap(op)->getRegion(static_cast<unsigned>(pos)));
}
MlirOperation mlirOperationGetNextInBlock(MlirOperation op) {
return wrap(unwrap(op)->getNextNode());
}
intptr_t mlirOperationGetNumOperands(MlirOperation op) {
return static_cast<intptr_t>(unwrap(op)->getNumOperands());
}
MlirValue mlirOperationGetOperand(MlirOperation op, intptr_t pos) {
return wrap(unwrap(op)->getOperand(static_cast<unsigned>(pos)));
}
intptr_t mlirOperationGetNumResults(MlirOperation op) {
return static_cast<intptr_t>(unwrap(op)->getNumResults());
}
MlirValue mlirOperationGetResult(MlirOperation op, intptr_t pos) {
return wrap(unwrap(op)->getResult(static_cast<unsigned>(pos)));
}
intptr_t mlirOperationGetNumSuccessors(MlirOperation op) {
return static_cast<intptr_t>(unwrap(op)->getNumSuccessors());
}
MlirBlock mlirOperationGetSuccessor(MlirOperation op, intptr_t pos) {
return wrap(unwrap(op)->getSuccessor(static_cast<unsigned>(pos)));
}
intptr_t mlirOperationGetNumAttributes(MlirOperation op) {
return static_cast<intptr_t>(unwrap(op)->getAttrs().size());
}
MlirNamedAttribute mlirOperationGetAttribute(MlirOperation op, intptr_t pos) {
NamedAttribute attr = unwrap(op)->getAttrs()[pos];
return MlirNamedAttribute{attr.first.c_str(), wrap(attr.second)};
}
MlirAttribute mlirOperationGetAttributeByName(MlirOperation op,
const char *name) {
return wrap(unwrap(op)->getAttr(name));
}
void mlirOperationPrint(MlirOperation op, MlirStringCallback callback,
void *userData) {
CallbackOstream stream(callback, userData);
unwrap(op)->print(stream);
stream.flush();
}
void mlirOperationDump(MlirOperation op) { return unwrap(op)->dump(); }
/* ========================================================================== */
/* Region API. */
/* ========================================================================== */
MlirRegion mlirRegionCreate() { return wrap(new Region); }
MlirBlock mlirRegionGetFirstBlock(MlirRegion region) {
Region *cppRegion = unwrap(region);
if (cppRegion->empty())
return wrap(static_cast<Block *>(nullptr));
return wrap(&cppRegion->front());
}
void mlirRegionAppendOwnedBlock(MlirRegion region, MlirBlock block) {
unwrap(region)->push_back(unwrap(block));
}
void mlirRegionInsertOwnedBlock(MlirRegion region, intptr_t pos,
MlirBlock block) {
auto &blockList = unwrap(region)->getBlocks();
blockList.insert(std::next(blockList.begin(), pos), unwrap(block));
}
void mlirRegionDestroy(MlirRegion region) {
delete static_cast<Region *>(region.ptr);
}
int mlirRegionIsNull(MlirRegion region) { return unwrap(region) == nullptr; }
/* ========================================================================== */
/* Block API. */
/* ========================================================================== */
MlirBlock mlirBlockCreate(intptr_t nArgs, MlirType *args) {
Block *b = new Block;
for (intptr_t i = 0; i < nArgs; ++i)
b->addArgument(unwrap(args[i]));
return wrap(b);
}
MlirBlock mlirBlockGetNextInRegion(MlirBlock block) {
return wrap(unwrap(block)->getNextNode());
}
MlirOperation mlirBlockGetFirstOperation(MlirBlock block) {
Block *cppBlock = unwrap(block);
if (cppBlock->empty())
return wrap(static_cast<Operation *>(nullptr));
return wrap(&cppBlock->front());
}
void mlirBlockAppendOwnedOperation(MlirBlock block, MlirOperation operation) {
unwrap(block)->push_back(unwrap(operation));
}
void mlirBlockInsertOwnedOperation(MlirBlock block, intptr_t pos,
MlirOperation operation) {
auto &opList = unwrap(block)->getOperations();
opList.insert(std::next(opList.begin(), pos), unwrap(operation));
}
void mlirBlockDestroy(MlirBlock block) { delete unwrap(block); }
int mlirBlockIsNull(MlirBlock block) { return unwrap(block) == nullptr; }
intptr_t mlirBlockGetNumArguments(MlirBlock block) {
return static_cast<intptr_t>(unwrap(block)->getNumArguments());
}
MlirValue mlirBlockGetArgument(MlirBlock block, intptr_t pos) {
return wrap(unwrap(block)->getArgument(static_cast<unsigned>(pos)));
}
void mlirBlockPrint(MlirBlock block, MlirStringCallback callback,
void *userData) {
CallbackOstream stream(callback, userData);
unwrap(block)->print(stream);
stream.flush();
}
/* ========================================================================== */
/* Value API. */
/* ========================================================================== */
MlirType mlirValueGetType(MlirValue value) {
return wrap(unwrap(value).getType());
}
void mlirValuePrint(MlirValue value, MlirStringCallback callback,
void *userData) {
CallbackOstream stream(callback, userData);
unwrap(value).print(stream);
stream.flush();
}
/* ========================================================================== */
/* Type API. */
/* ========================================================================== */
MlirType mlirTypeParseGet(MlirContext context, const char *type) {
return wrap(mlir::parseType(type, unwrap(context)));
}
MlirContext mlirTypeGetContext(MlirType type) {
return wrap(unwrap(type).getContext());
}
int mlirTypeEqual(MlirType t1, MlirType t2) { return unwrap(t1) == unwrap(t2); }
void mlirTypePrint(MlirType type, MlirStringCallback callback, void *userData) {
CallbackOstream stream(callback, userData);
unwrap(type).print(stream);
stream.flush();
}
void mlirTypeDump(MlirType type) { unwrap(type).dump(); }
/* ========================================================================== */
/* Attribute API. */
/* ========================================================================== */
MlirAttribute mlirAttributeParseGet(MlirContext context, const char *attr) {
return wrap(mlir::parseAttribute(attr, unwrap(context)));
}
int mlirAttributeEqual(MlirAttribute a1, MlirAttribute a2) {
return unwrap(a1) == unwrap(a2);
}
void mlirAttributePrint(MlirAttribute attr, MlirStringCallback callback,
void *userData) {
CallbackOstream stream(callback, userData);
unwrap(attr).print(stream);
stream.flush();
}
void mlirAttributeDump(MlirAttribute attr) { unwrap(attr).dump(); }
MlirNamedAttribute mlirNamedAttributeGet(const char *name, MlirAttribute attr) {
return MlirNamedAttribute{name, attr};
}