llvm-project/mlir/lib/IR/Builders.cpp

347 lines
13 KiB
C++
Raw Normal View History

//===- Builders.cpp - Helpers for constructing MLIR Classes ---------------===//
//
// Copyright 2019 The MLIR Authors.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
// =============================================================================
#include "mlir/IR/Builders.h"
#include "mlir/IR/AffineExpr.h"
#include "mlir/IR/AffineMap.h"
#include "mlir/IR/Attributes.h"
#include "mlir/IR/IntegerSet.h"
#include "mlir/IR/Location.h"
#include "mlir/IR/Module.h"
#include "mlir/IR/Types.h"
using namespace mlir;
Builder::Builder(Module *module) : context(module->getContext()) {}
Identifier Builder::getIdentifier(StringRef str) {
return Identifier::get(str, context);
}
Module *Builder::createModule() { return new Module(context); }
//===----------------------------------------------------------------------===//
// Locations.
//===----------------------------------------------------------------------===//
UnknownLoc *Builder::getUnknownLoc() { return UnknownLoc::get(context); }
UniquedFilename Builder::getUniquedFilename(StringRef filename) {
return UniquedFilename::get(filename, context);
}
FileLineColLoc *Builder::getFileLineColLoc(UniquedFilename filename,
unsigned line, unsigned column) {
return FileLineColLoc::get(filename, line, column, context);
}
//===----------------------------------------------------------------------===//
// Types.
//===----------------------------------------------------------------------===//
FloatType *Builder::getBF16Type() { return Type::getBF16(context); }
FloatType *Builder::getF16Type() { return Type::getF16(context); }
FloatType *Builder::getF32Type() { return Type::getF32(context); }
FloatType *Builder::getF64Type() { return Type::getF64(context); }
OtherType *Builder::getAffineIntType() { return Type::getAffineInt(context); }
OtherType *Builder::getTFControlType() { return Type::getTFControl(context); }
OtherType *Builder::getTFResourceType() { return Type::getTFResource(context); }
OtherType *Builder::getTFVariantType() { return Type::getTFVariant(context); }
OtherType *Builder::getTFStringType() { return Type::getTFString(context); }
IntegerType *Builder::getIntegerType(unsigned width) {
return Type::getInteger(width, context);
}
FunctionType *Builder::getFunctionType(ArrayRef<Type *> inputs,
ArrayRef<Type *> results) {
return FunctionType::get(inputs, results, context);
}
MemRefType *Builder::getMemRefType(ArrayRef<int> shape, Type *elementType,
ArrayRef<AffineMap *> affineMapComposition,
unsigned memorySpace) {
return MemRefType::get(shape, elementType, affineMapComposition, memorySpace);
}
VectorType *Builder::getVectorType(ArrayRef<unsigned> shape,
Type *elementType) {
return VectorType::get(shape, elementType);
}
RankedTensorType *Builder::getTensorType(ArrayRef<int> shape,
Type *elementType) {
return RankedTensorType::get(shape, elementType);
}
UnrankedTensorType *Builder::getTensorType(Type *elementType) {
return UnrankedTensorType::get(elementType);
}
//===----------------------------------------------------------------------===//
// Attributes.
//===----------------------------------------------------------------------===//
BoolAttr *Builder::getBoolAttr(bool value) {
return BoolAttr::get(value, context);
}
IntegerAttr *Builder::getIntegerAttr(int64_t value) {
return IntegerAttr::get(value, context);
}
FloatAttr *Builder::getFloatAttr(double value) {
return FloatAttr::get(value, context);
}
StringAttr *Builder::getStringAttr(StringRef bytes) {
return StringAttr::get(bytes, context);
}
ArrayAttr *Builder::getArrayAttr(ArrayRef<Attribute *> value) {
return ArrayAttr::get(value, context);
}
AffineMapAttr *Builder::getAffineMapAttr(AffineMap *map) {
return AffineMapAttr::get(map, context);
}
TypeAttr *Builder::getTypeAttr(Type *type) {
return TypeAttr::get(type, context);
}
FunctionAttr *Builder::getFunctionAttr(const Function *value) {
return FunctionAttr::get(value, context);
}
//===----------------------------------------------------------------------===//
// Affine Expressions, Affine Maps, and Integet Sets.
//===----------------------------------------------------------------------===//
AffineMap *Builder::getAffineMap(unsigned dimCount, unsigned symbolCount,
ArrayRef<AffineExpr *> results,
ArrayRef<AffineExpr *> rangeSizes) {
return AffineMap::get(dimCount, symbolCount, results, rangeSizes, context);
}
AffineDimExpr *Builder::getDimExpr(unsigned position) {
return AffineDimExpr::get(position, context);
}
AffineSymbolExpr *Builder::getSymbolExpr(unsigned position) {
return AffineSymbolExpr::get(position, context);
}
AffineConstantExpr *Builder::getConstantExpr(int64_t constant) {
return AffineConstantExpr::get(constant, context);
}
AffineExpr *Builder::getAddExpr(AffineExpr *lhs, AffineExpr *rhs) {
return AffineBinaryOpExpr::get(AffineExpr::Kind::Add, lhs, rhs, context);
}
AffineExpr *Builder::getAddExpr(AffineExpr *lhs, int64_t rhs) {
return AffineBinaryOpExpr::getAdd(lhs, rhs, context);
}
AffineExpr *Builder::getMulExpr(AffineExpr *lhs, AffineExpr *rhs) {
return AffineBinaryOpExpr::get(AffineExpr::Kind::Mul, lhs, rhs, context);
}
Extend getConstantTripCount to deal with a larger subset of loop bounds; make loop unroll/unroll-and-jam more powerful; add additional affine expr builder methods - use previously added analysis/simplification to infer multiple of unroll factor trip counts, making loop unroll/unroll-and-jam more general. - for loop unroll, support bounds that are single result affine map's with the same set of operands. For unknown loop bounds, loop unroll will now work as long as trip count can be determined to be a multiple of unroll factor. - extend getConstantTripCount to deal with single result affine map's with the same operands. move it to mlir/Analysis/LoopAnalysis.cpp - add additional builder utility methods for affine expr arithmetic (difference, mod/floordiv/ceildiv w.r.t postitive constant). simplify code to use the utility methods. - move affine analysis routines to AffineAnalysis.cpp/.h from AffineStructures.cpp/.h. - Rename LoopUnrollJam to LoopUnrollAndJam to match class name. - add an additional simplification for simplifyFloorDiv, simplifyCeilDiv - Rename AffineMap::getNumOperands() getNumInputs: an affine map by itself does not have operands. Operands are passed to it through affine_apply, from loop bounds/if condition's, etc., operands are stored in the latter. This should be sufficiently powerful for now as far as unroll/unroll-and-jam go for TPU code generation, and can move to other analyses/transformations. Loop nests like these are now unrolled without any cleanup loop being generated. for %i = 1 to 100 { // unroll factor 4: no cleanup loop will be generated. for %j = (d0) -> (d0) (%i) to (d0) -> (5*d0 + 3) (%i) { %x = "foo"(%j) : (affineint) -> i32 } } for %i = 1 to 100 { // unroll factor 4: no cleanup loop will be generated. for %j = (d0) -> (d0) (%i) to (d0) -> (d0 - d mod 4 - 1) (%i) { %y = "foo"(%j) : (affineint) -> i32 } } for %i = 1 to 100 { for %j = (d0) -> (d0) (%i) to (d0) -> (d0 + 128) (%i) { %x = "foo"() : () -> i32 } } TODO(bondhugula): extend this to LoopUnrollAndJam as well in the next CL (with minor changes). PiperOrigin-RevId: 212661212
2018-09-12 10:21:23 -07:00
// Most multiply expressions are pure affine (rhs is a constant).
AffineExpr *Builder::getMulExpr(AffineExpr *lhs, int64_t rhs) {
return AffineBinaryOpExpr::get(AffineExpr::Kind::Mul, lhs,
getConstantExpr(rhs), context);
}
AffineExpr *Builder::getSubExpr(AffineExpr *lhs, AffineExpr *rhs) {
return getAddExpr(lhs, getMulExpr(rhs, getConstantExpr(-1)));
}
AffineExpr *Builder::getSubExpr(AffineExpr *lhs, int64_t rhs) {
return AffineBinaryOpExpr::getAdd(lhs, -rhs, context);
}
AffineExpr *Builder::getModExpr(AffineExpr *lhs, AffineExpr *rhs) {
return AffineBinaryOpExpr::get(AffineExpr::Kind::Mod, lhs, rhs, context);
}
Extend getConstantTripCount to deal with a larger subset of loop bounds; make loop unroll/unroll-and-jam more powerful; add additional affine expr builder methods - use previously added analysis/simplification to infer multiple of unroll factor trip counts, making loop unroll/unroll-and-jam more general. - for loop unroll, support bounds that are single result affine map's with the same set of operands. For unknown loop bounds, loop unroll will now work as long as trip count can be determined to be a multiple of unroll factor. - extend getConstantTripCount to deal with single result affine map's with the same operands. move it to mlir/Analysis/LoopAnalysis.cpp - add additional builder utility methods for affine expr arithmetic (difference, mod/floordiv/ceildiv w.r.t postitive constant). simplify code to use the utility methods. - move affine analysis routines to AffineAnalysis.cpp/.h from AffineStructures.cpp/.h. - Rename LoopUnrollJam to LoopUnrollAndJam to match class name. - add an additional simplification for simplifyFloorDiv, simplifyCeilDiv - Rename AffineMap::getNumOperands() getNumInputs: an affine map by itself does not have operands. Operands are passed to it through affine_apply, from loop bounds/if condition's, etc., operands are stored in the latter. This should be sufficiently powerful for now as far as unroll/unroll-and-jam go for TPU code generation, and can move to other analyses/transformations. Loop nests like these are now unrolled without any cleanup loop being generated. for %i = 1 to 100 { // unroll factor 4: no cleanup loop will be generated. for %j = (d0) -> (d0) (%i) to (d0) -> (5*d0 + 3) (%i) { %x = "foo"(%j) : (affineint) -> i32 } } for %i = 1 to 100 { // unroll factor 4: no cleanup loop will be generated. for %j = (d0) -> (d0) (%i) to (d0) -> (d0 - d mod 4 - 1) (%i) { %y = "foo"(%j) : (affineint) -> i32 } } for %i = 1 to 100 { for %j = (d0) -> (d0) (%i) to (d0) -> (d0 + 128) (%i) { %x = "foo"() : () -> i32 } } TODO(bondhugula): extend this to LoopUnrollAndJam as well in the next CL (with minor changes). PiperOrigin-RevId: 212661212
2018-09-12 10:21:23 -07:00
// Most modulo expressions are pure affine.
AffineExpr *Builder::getModExpr(AffineExpr *lhs, uint64_t rhs) {
return AffineBinaryOpExpr::get(AffineExpr::Kind::Mod, lhs,
getConstantExpr(rhs), context);
}
AffineExpr *Builder::getFloorDivExpr(AffineExpr *lhs, AffineExpr *rhs) {
return AffineBinaryOpExpr::get(AffineExpr::Kind::FloorDiv, lhs, rhs, context);
}
Extend getConstantTripCount to deal with a larger subset of loop bounds; make loop unroll/unroll-and-jam more powerful; add additional affine expr builder methods - use previously added analysis/simplification to infer multiple of unroll factor trip counts, making loop unroll/unroll-and-jam more general. - for loop unroll, support bounds that are single result affine map's with the same set of operands. For unknown loop bounds, loop unroll will now work as long as trip count can be determined to be a multiple of unroll factor. - extend getConstantTripCount to deal with single result affine map's with the same operands. move it to mlir/Analysis/LoopAnalysis.cpp - add additional builder utility methods for affine expr arithmetic (difference, mod/floordiv/ceildiv w.r.t postitive constant). simplify code to use the utility methods. - move affine analysis routines to AffineAnalysis.cpp/.h from AffineStructures.cpp/.h. - Rename LoopUnrollJam to LoopUnrollAndJam to match class name. - add an additional simplification for simplifyFloorDiv, simplifyCeilDiv - Rename AffineMap::getNumOperands() getNumInputs: an affine map by itself does not have operands. Operands are passed to it through affine_apply, from loop bounds/if condition's, etc., operands are stored in the latter. This should be sufficiently powerful for now as far as unroll/unroll-and-jam go for TPU code generation, and can move to other analyses/transformations. Loop nests like these are now unrolled without any cleanup loop being generated. for %i = 1 to 100 { // unroll factor 4: no cleanup loop will be generated. for %j = (d0) -> (d0) (%i) to (d0) -> (5*d0 + 3) (%i) { %x = "foo"(%j) : (affineint) -> i32 } } for %i = 1 to 100 { // unroll factor 4: no cleanup loop will be generated. for %j = (d0) -> (d0) (%i) to (d0) -> (d0 - d mod 4 - 1) (%i) { %y = "foo"(%j) : (affineint) -> i32 } } for %i = 1 to 100 { for %j = (d0) -> (d0) (%i) to (d0) -> (d0 + 128) (%i) { %x = "foo"() : () -> i32 } } TODO(bondhugula): extend this to LoopUnrollAndJam as well in the next CL (with minor changes). PiperOrigin-RevId: 212661212
2018-09-12 10:21:23 -07:00
// Most floordiv expressions are pure affine.
AffineExpr *Builder::getFloorDivExpr(AffineExpr *lhs, uint64_t rhs) {
return AffineBinaryOpExpr::get(AffineExpr::Kind::FloorDiv, lhs,
getConstantExpr(rhs), context);
}
AffineExpr *Builder::getCeilDivExpr(AffineExpr *lhs, AffineExpr *rhs) {
return AffineBinaryOpExpr::get(AffineExpr::Kind::CeilDiv, lhs, rhs, context);
}
Extend getConstantTripCount to deal with a larger subset of loop bounds; make loop unroll/unroll-and-jam more powerful; add additional affine expr builder methods - use previously added analysis/simplification to infer multiple of unroll factor trip counts, making loop unroll/unroll-and-jam more general. - for loop unroll, support bounds that are single result affine map's with the same set of operands. For unknown loop bounds, loop unroll will now work as long as trip count can be determined to be a multiple of unroll factor. - extend getConstantTripCount to deal with single result affine map's with the same operands. move it to mlir/Analysis/LoopAnalysis.cpp - add additional builder utility methods for affine expr arithmetic (difference, mod/floordiv/ceildiv w.r.t postitive constant). simplify code to use the utility methods. - move affine analysis routines to AffineAnalysis.cpp/.h from AffineStructures.cpp/.h. - Rename LoopUnrollJam to LoopUnrollAndJam to match class name. - add an additional simplification for simplifyFloorDiv, simplifyCeilDiv - Rename AffineMap::getNumOperands() getNumInputs: an affine map by itself does not have operands. Operands are passed to it through affine_apply, from loop bounds/if condition's, etc., operands are stored in the latter. This should be sufficiently powerful for now as far as unroll/unroll-and-jam go for TPU code generation, and can move to other analyses/transformations. Loop nests like these are now unrolled without any cleanup loop being generated. for %i = 1 to 100 { // unroll factor 4: no cleanup loop will be generated. for %j = (d0) -> (d0) (%i) to (d0) -> (5*d0 + 3) (%i) { %x = "foo"(%j) : (affineint) -> i32 } } for %i = 1 to 100 { // unroll factor 4: no cleanup loop will be generated. for %j = (d0) -> (d0) (%i) to (d0) -> (d0 - d mod 4 - 1) (%i) { %y = "foo"(%j) : (affineint) -> i32 } } for %i = 1 to 100 { for %j = (d0) -> (d0) (%i) to (d0) -> (d0 + 128) (%i) { %x = "foo"() : () -> i32 } } TODO(bondhugula): extend this to LoopUnrollAndJam as well in the next CL (with minor changes). PiperOrigin-RevId: 212661212
2018-09-12 10:21:23 -07:00
// Most ceildiv expressions are pure affine.
AffineExpr *Builder::getCeilDivExpr(AffineExpr *lhs, uint64_t rhs) {
return AffineBinaryOpExpr::get(AffineExpr::Kind::CeilDiv, lhs,
getConstantExpr(rhs), context);
}
/// Creates a sum of products affine expression from constant coefficients.
/// If c_0, c_1, ... are the coefficients in the order corresponding to
/// dimensions, symbols, and constant term, create the affine expression:
/// expr = c_0*d0 + c_1*d1 + ... + c_{ndims-1}*d_{ndims-1} + c_{..}*s0 +
/// c_{..}*s1 + ... + const
AffineExpr *Builder::getAddMulPureAffineExpr(unsigned numDims,
unsigned numSymbols,
ArrayRef<int64_t> coeffs) {
assert(coeffs.size() == numDims + numSymbols + 1);
AffineExpr *expr = AffineConstantExpr::get(0, context);
// Dimensions.
for (unsigned j = 0; j < numDims; j++) {
if (coeffs[j] == 0)
continue;
AffineExpr *id = AffineDimExpr::get(j, context);
auto *term = AffineBinaryOpExpr::getMul(id, coeffs[j], context);
expr = AffineBinaryOpExpr::getAdd(expr, term, context);
}
// Symbols.
for (unsigned j = numDims; j < numDims + numSymbols; j++) {
if (coeffs[j] == 0)
continue;
AffineExpr *id = AffineSymbolExpr::get(j - numDims, context);
auto *term = AffineBinaryOpExpr::getMul(id, coeffs[j], context);
expr = AffineBinaryOpExpr::getAdd(expr, term, context);
}
// Constant term.
unsigned constTerm = coeffs[coeffs.size() - 1];
if (constTerm != 0)
expr = AffineBinaryOpExpr::getAdd(expr, constTerm, context);
return expr;
}
IntegerSet *Builder::getIntegerSet(unsigned dimCount, unsigned symbolCount,
ArrayRef<AffineExpr *> constraints,
ArrayRef<bool> isEq) {
return IntegerSet::get(dimCount, symbolCount, constraints, isEq, context);
}
AffineMap *Builder::getConstantAffineMap(int64_t val) {
return AffineMap::get(/*dimCount=*/0, /*symbolCount=*/0, getConstantExpr(val),
{}, context);
}
AffineMap *Builder::getDimIdentityMap() {
return AffineMap::get(/*dimCount=*/1, /*symbolCount=*/0, getDimExpr(0), {},
context);
}
AffineMap *Builder::getSymbolIdentityMap() {
return AffineMap::get(/*dimCount=*/0, /*symbolCount=*/1, getSymbolExpr(0), {},
context);
}
//===----------------------------------------------------------------------===//
// CFG function elements.
//===----------------------------------------------------------------------===//
/// Add new basic block and set the insertion point to the end of it. If an
/// 'insertBefore' basic block is passed, the block will be placed before the
/// specified block. If not, the block will be appended to the end of the
/// current function.
BasicBlock *CFGFuncBuilder::createBlock(BasicBlock *insertBefore) {
BasicBlock *b = new BasicBlock();
// If we are supposed to insert before a specific block, do so, otherwise add
// the block to the end of the function.
if (insertBefore)
function->getBlocks().insert(CFGFunction::iterator(insertBefore), b);
else
function->push_back(b);
setInsertionPoint(b);
return b;
}
/// Create an operation given the fields represented as an OperationState.
OperationInst *CFGFuncBuilder::createOperation(const OperationState &state) {
SmallVector<CFGValue *, 8> operands;
operands.reserve(state.operands.size());
for (auto elt : state.operands)
operands.push_back(cast<CFGValue>(elt));
auto *op = OperationInst::create(state.location, state.name, operands,
state.types, state.attributes, context);
block->getOperations().insert(insertPoint, op);
return op;
}
//===----------------------------------------------------------------------===//
// Statements.
//===----------------------------------------------------------------------===//
/// Create an operation given the fields represented as an OperationState.
OperationStmt *MLFuncBuilder::createOperation(const OperationState &state) {
SmallVector<MLValue *, 8> operands;
operands.reserve(state.operands.size());
for (auto elt : state.operands)
operands.push_back(cast<MLValue>(elt));
auto *op = OperationStmt::create(state.location, state.name, operands,
state.types, state.attributes, context);
block->getStatements().insert(insertPoint, op);
return op;
}
ForStmt *MLFuncBuilder::createFor(Location *location,
ArrayRef<MLValue *> lbOperands,
AffineMap *lbMap,
ArrayRef<MLValue *> ubOperands,
AffineMap *ubMap, int64_t step) {
auto *stmt = ForStmt::create(location, lbOperands, lbMap, ubOperands, ubMap,
step, context);
block->getStatements().insert(insertPoint, stmt);
return stmt;
}
ForStmt *MLFuncBuilder::createFor(Location *location, int64_t lb, int64_t ub,
int64_t step) {
auto *lbMap = AffineMap::getConstantMap(lb, context);
auto *ubMap = AffineMap::getConstantMap(ub, context);
return createFor(location, {}, lbMap, {}, ubMap, step);
}
IfStmt *MLFuncBuilder::createIf(Location *location,
ArrayRef<MLValue *> operands, IntegerSet *set) {
auto *stmt = IfStmt::create(location, operands, set);
block->getStatements().insert(insertPoint, stmt);
return stmt;
}