colin
/
onnx-mlir
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Bug fix, ensure krnl.iterate can lower in the degenerate case. (#78) 

* Bug fix, ensure krnl.iterate can lower in the degenerate case.

* Fix parser issue with degenerate iterate op.

* Add a test case.

* Remove dead code.

Co-authored-by: Gheorghe-Teodor Bercea <gt.bercea@gmail.com>
This commit is contained in:
Tian Jin 2020-04-10 23:27:00 +08:00 committed by GitHub
parent 7dba324404
commit caeaa390e2
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
3 changed files with 96 additions and 81 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

111
src/Dialect/Krnl/KrnlOps.cpp
View File

@ -46,13 +46,13 @@ KrnlOpsDialect::KrnlOpsDialect(MLIRContext *context)
// KrnlDefineLoopsOp
//===----------------------------------------------------------------------===//
void KrnlDefineLoopsOp::build(Builder *builder, OperationState &result,
int64_t num_loops) {
void KrnlDefineLoopsOp::build(
Builder *builder, OperationState &result, int64_t num_loops) {
// Create the same number of dimension handlers as the number of
// dimensions in the associated integer set.
result.types.append(num_loops, LoopType::get(builder->getContext()));
result.addAttribute(getNumLoopsAttrName(),
builder->getI32IntegerAttr(num_loops));
result.addAttribute(
getNumLoopsAttrName(), builder->getI32IntegerAttr(num_loops));
}
void print(OpAsmPrinter &p, KrnlDefineLoopsOp &op) {
@ -61,15 +61,14 @@ void print(OpAsmPrinter &p, KrnlDefineLoopsOp &op) {
p << "krnl.define_loops " << numLoopAttr.getValue().getSExtValue();
}
ParseResult parseKrnlDefineLoopsOp(OpAsmParser &parser,
OperationState &result) {
ParseResult parseKrnlDefineLoopsOp(
OpAsmParser &parser, OperationState &result) {
// Parse the attribute indicating number of loops defined.
IntegerAttr numLoops;
auto &builder = parser.getBuilder();
auto intType = builder.getIntegerType(64);
if (parser.parseAttribute(numLoops, intType,
KrnlDefineLoopsOp::getNumLoopsAttrName(),
result.attributes))
KrnlDefineLoopsOp::getNumLoopsAttrName(), result.attributes))
return failure();
auto loopTypes = llvm::SmallVector<Type, 4>(
@ -82,10 +81,10 @@ ParseResult parseKrnlDefineLoopsOp(OpAsmParser &parser,
// KrnlOptimizeLoopsOp
//===----------------------------------------------------------------------===//
void KrnlOptimizeLoopsOp::build(Builder *builder, OperationState &result,
int num_optimized_loops) {
result.types.append(num_optimized_loops,
LoopType::get(builder->getContext()));
void KrnlOptimizeLoopsOp::build(
Builder *builder, OperationState &result, int num_optimized_loops) {
result.types.append(
num_optimized_loops, LoopType::get(builder->getContext()));
// Create a region and a block for the body.
// Schedule intrinsics will be placed into this region.
Region *region = result.addRegion();
@ -96,13 +95,13 @@ void KrnlOptimizeLoopsOp::build(Builder *builder, OperationState &result,
void print(OpAsmPrinter &p, KrnlOptimizeLoopsOp &op) {
p << "krnl.optimize_loops ";
p.printRegion(op.region(), /*printEntryBlockArgs=*/false,
/*printBlockTerminators=*/true);
/*printBlockTerminators=*/true);
p << " : ";
p.printFunctionalType(op);
}
ParseResult parseKrnlOptimizeLoopsOp(OpAsmParser &parser,
OperationState &result) {
ParseResult parseKrnlOptimizeLoopsOp(
OpAsmParser &parser, OperationState &result) {
// Parse the schedule body region.
Region *region = result.addRegion();
if (parser.parseRegion(*region, llvm::None, llvm::None))
@ -146,14 +145,13 @@ ParseResult parseKrnlOptimizeLoopsOp(OpAsmParser &parser,
* %i0 = 10 to N : %i1 = M to 20
*/
void KrnlIterateOp::build(Builder *builder, OperationState &result,
KrnlIterateOperandPack operandPack) {
KrnlIterateOperandPack operandPack) {
// Record optimized loops and the number of such loops.
result.addOperands(operandPack.getOperands());
result.addAttribute(KrnlIterateOp::getBoundsAttrName(),
operandPack.getAttributes());
result.addAttribute(
getNumOptimizedLoopsAttrName(),
KrnlIterateOp::getBoundsAttrName(), operandPack.getAttributes());
result.addAttribute(getNumOptimizedLoopsAttrName(),
builder->getI64IntegerAttr(operandPack.getNumOptimizedLoops()));
// Create a region and a block for the body. The arguments of the region are
@ -190,15 +188,17 @@ void print(OpAsmPrinter &p, KrnlIterateOp &op) {
p << " -> ";
p.printOperand(var);
p << " = ";
onnx_mlir::printBound((*boundItr++).cast<AffineMapAttr>(), operandItr, "max", p);
onnx_mlir::printBound(
(*boundItr++).cast<AffineMapAttr>(), operandItr, "max", p);
p << " to ";
onnx_mlir::printBound((*boundItr++).cast<AffineMapAttr>(), operandItr, "min", p);
onnx_mlir::printBound(
(*boundItr++).cast<AffineMapAttr>(), operandItr, "min", p);
delimiter = ", ";
}
p << ")";
p.printRegion(op.bodyRegion(), /*printEntryBlockArgs=*/false,
/*printBlockTerminators=*/false);
/*printBlockTerminators=*/false);
}
ParseResult parseKrnlIterateOp(OpAsmParser &parser, OperationState &result) {
@ -208,16 +208,15 @@ ParseResult parseKrnlIterateOp(OpAsmParser &parser, OperationState &result) {
// Parse optimized loops:
SmallVector<OpAsmParser::OperandType, 4> optimizedLoopRefs;
if (parser.parseOperandList(optimizedLoopRefs,
OpAsmParser::Delimiter::Paren) ||
if (parser.parseOperandList(
optimizedLoopRefs, OpAsmParser::Delimiter::Paren) ||
parser.resolveOperands(optimizedLoopRefs,
LoopType::get(result.getContext()),
result.operands))
LoopType::get(result.getContext()), result.operands))
return failure();
// Record how many optimized loops did we parse.
result.addAttribute(KrnlIterateOp::getNumOptimizedLoopsAttrName(),
builder.getI64IntegerAttr(optimizedLoopRefs.size()));
builder.getI64IntegerAttr(optimizedLoopRefs.size()));
// Parse input loops and their lower and upper bounds.
SmallVector<OpAsmParser::OperandType, 4> inductionVarRefs;
@ -227,16 +226,16 @@ ParseResult parseKrnlIterateOp(OpAsmParser &parser, OperationState &result) {
return failure();
// A function to parse a lower or upper bound.
auto parseBound = [&result, &builder, &parser, &operandParser,
&boundMaps](bool isUpper) -> ParseResult {
auto parseBound = [&result, &builder, &parser, &operandParser, &boundMaps](
bool isUpper) -> ParseResult {
// 'min' / 'max' prefixes are generally syntactic sugar, but are required if
// the map has multiple results.
bool failedToParsedMinMax =
failed(parser.parseOptionalKeyword(isUpper ? "min" : "max"));
// Try parse an SSA operand.
if (succeeded(operandParser.ParseOptionalOperand(builder.getIndexType(),
result.operands))) {
if (succeeded(operandParser.ParseOptionalOperand(
builder.getIndexType(), result.operands))) {
AffineMap map = builder.getSymbolIdentityMap();
boundMaps.emplace_back(AffineMapAttr::get(map));
return success();
@ -248,8 +247,8 @@ ParseResult parseKrnlIterateOp(OpAsmParser &parser, OperationState &result) {
llvm::SMLoc attrLoc = parser.getCurrentLocation();
Attribute boundAttr;
llvm::SmallVector<NamedAttribute, 1> tempBoundAttrContainer;
if (parser.parseAttribute(boundAttr, builder.getIndexType(), "temp",
tempBoundAttrContainer))
if (parser.parseAttribute(
boundAttr, builder.getIndexType(), "temp", tempBoundAttrContainer))
return failure();
if (auto affineMapAttr = boundAttr.dyn_cast<AffineMapAttr>()) {
@ -260,15 +259,13 @@ ParseResult parseKrnlIterateOp(OpAsmParser &parser, OperationState &result) {
auto map = affineMapAttr.getValue();
if (map.getNumDims() != numDims)
return parser.emitError(
parser.getNameLoc(),
return parser.emitError(parser.getNameLoc(),
"dim operand count and integer set dim count must match");
unsigned numDimAndSymbolOperands =
result.operands.size() - currentNumOperands;
if (numDims + map.getNumSymbols() != numDimAndSymbolOperands)
return parser.emitError(
parser.getNameLoc(),
return parser.emitError(parser.getNameLoc(),
"symbol operand count and integer set symbol count must match");
// If the map has multiple results, make sure that we parsed the min/max
@ -276,11 +273,11 @@ ParseResult parseKrnlIterateOp(OpAsmParser &parser, OperationState &result) {
if (map.getNumResults() > 1 && failedToParsedMinMax) {
if (isUpper)
return parser.emitError(attrLoc,
"upper loop bound affine map with multiple "
"results requires 'min' prefix");
"upper loop bound affine map with multiple "
"results requires 'min' prefix");
return parser.emitError(attrLoc,
"lower loop bound affine mapwith "
"multiple results requires 'max' prefix");
"lower loop bound affine mapwith "
"multiple results requires 'max' prefix");
}
boundMaps.emplace_back(AffineMapAttr::get(map));
return success();
@ -293,8 +290,7 @@ ParseResult parseKrnlIterateOp(OpAsmParser &parser, OperationState &result) {
}
};
bool keepParsing; // Do we keep parsing loops/bounds?
do {
while (failed(parser.parseOptionalRParen())) {
// Parse an input loop operand;
operandParser.ParseOperand(LoopType::get(context), result.operands);
parser.parseArrow();
@ -315,26 +311,23 @@ ParseResult parseKrnlIterateOp(OpAsmParser &parser, OperationState &result) {
// the entire "{operand bound}, {input_loop_operand}" sequence will
// be parsed as an operand list.
parser.parseOptionalComma();
// If we don't see a RParen token, we keep parsing.
keepParsing = failed(parser.parseOptionalRParen());
} while (keepParsing);
}
// At this point, there shouldn't be any operands left to parse.
if (operandParser.hasOperandLeft())
return parser.emitError(parser.getCurrentLocation());
result.addAttribute(KrnlIterateOp::getBoundsAttrName(),
builder.getArrayAttr(boundMaps));
result.addAttribute(
KrnlIterateOp::getBoundsAttrName(), builder.getArrayAttr(boundMaps));
Region *region = result.addRegion();
SmallVector<Type, 4> inductionVarTypes(inductionVarRefs.size(),
builder.getIndexType());
SmallVector<Type, 4> inductionVarTypes(
inductionVarRefs.size(), builder.getIndexType());
if (parser.parseRegion(*region, inductionVarRefs, inductionVarTypes))
return failure();
// Ensure iterate region is closed off with krnl.terminate.
KrnlIterateOp::ensureTerminator(*region, parser.getBuilder(),
result.location);
KrnlIterateOp::ensureTerminator(
*region, parser.getBuilder(), result.location);
return success();
}
@ -353,22 +346,20 @@ void print(OpAsmPrinter &p, KrnlReturnLoopsOp &op) {
p.printOperands(op.operand_begin(), op.operand_end());
}
ParseResult parseKrnlReturnLoopsOp(OpAsmParser &parser,
OperationState &result) {
ParseResult parseKrnlReturnLoopsOp(
OpAsmParser &parser, OperationState &result) {
// Parse the loops to return.
SmallVector<OpAsmParser::OperandType, 4> timestamp_dim_handlers;
if (parser.parseOperandList(timestamp_dim_handlers) ||
parser.resolveOperands(timestamp_dim_handlers,
LoopType::get(result.getContext()),
result.operands))
LoopType::get(result.getContext()), result.operands))
return failure();
return success();
}
void KrnlEntryPointOp::build(mlir::Builder *builder, OperationState &state,
SymbolRefAttr funcAttr, IntegerAttr numInputs,
IntegerAttr numOutputs) {
SymbolRefAttr funcAttr, IntegerAttr numInputs, IntegerAttr numOutputs) {
state.addAttribute(KrnlEntryPointOp::getEntryPointFuncAttrName(), funcAttr);
state.addAttribute(KrnlEntryPointOp::getNumInputsAttrName(), numInputs);
state.addAttribute(KrnlEntryPointOp::getNumOutputsAttrName(), numOutputs);

51
src/Transform/LowerKrnl.cpp
View File

@ -27,8 +27,8 @@ namespace {
struct KrnlIterateOpLowering : public OpRewritePattern<KrnlIterateOp> {
using OpRewritePattern<KrnlIterateOp>::OpRewritePattern;
LogicalResult matchAndRewrite(KrnlIterateOp iterateOp,
PatternRewriter &rewriter) const override {
LogicalResult matchAndRewrite(
KrnlIterateOp iterateOp, PatternRewriter &rewriter) const override {
auto boundMapAttrs =
iterateOp.getAttrOfType<ArrayAttr>(KrnlIterateOp::getBoundsAttrName())
.getValue();
@ -48,21 +48,21 @@ struct KrnlIterateOpLowering : public OpRewritePattern<KrnlIterateOp> {
map = boundMapAttrs[boundIdx + boundType]
.cast<AffineMapAttr>()
.getValue();
operands.insert(operands.end(), operandItr,
operandItr + map.getNumInputs());
operands.insert(
operands.end(), operandItr, operandItr + map.getNumInputs());
std::advance(operandItr, map.getNumInputs());
}
nestedForOps.emplace_back(rewriter.create<AffineForOp>(
iterateOp.getLoc(), lbOperands, lbMap, ubOperands, ubMap));
rewriter.setInsertionPoint(nestedForOps.back().getBody(),
nestedForOps.back().getBody()->begin());
nestedForOps.back().getBody()->begin());
}
// Replace induction variable references from those introduced by a
// single krnl.iterate to those introduced by multiple affine.for
// operations.
for (size_t i = 0; i < nestedForOps.size() - 1; i++) {
for (int64_t i = 0; i < (int64_t)nestedForOps.size() - 1; i++) {
auto iterateIV = iterateOp.bodyRegion().front().getArgument(0);
auto forIV = nestedForOps[i].getBody()->getArgument(0);
iterateIV.replaceAllUsesWith(forIV);
@ -74,11 +74,21 @@ struct KrnlIterateOpLowering : public OpRewritePattern<KrnlIterateOp> {
while (iterateOp.bodyRegion().front().getNumArguments() > 1)
iterateOp.bodyRegion().front().eraseArgument(0);
// Transfer krnl.iterate region to innermost for op.
auto innermostForOp = nestedForOps.back();
innermostForOp.region().getBlocks().clear();
rewriter.inlineRegionBefore(iterateOp.bodyRegion(), innermostForOp.region(),
innermostForOp.region().end());
if (nestedForOps.empty()) {
// If no loops are involved, simply move operations from within iterateOp
// body region to the parent region of iterateOp.
rewriter.setInsertionPoint(iterateOp);
iterateOp.bodyRegion().walk([&](Operation *op) {
if (!op->isKnownTerminator())
op->replaceAllUsesWith(rewriter.clone(*op));
});
} else {
// Transfer krnl.iterate region to innermost for op.
auto innermostForOp = nestedForOps.back();
innermostForOp.region().getBlocks().clear();
rewriter.inlineRegionBefore(iterateOp.bodyRegion(),
innermostForOp.region(), innermostForOp.region().end());
}
rewriter.eraseOp(iterateOp);
return success();
@ -93,8 +103,8 @@ class KrnlTerminatorLowering : public OpRewritePattern<KrnlTerminatorOp> {
public:
using OpRewritePattern<KrnlTerminatorOp>::OpRewritePattern;
LogicalResult matchAndRewrite(KrnlTerminatorOp op,
PatternRewriter &rewriter) const override {
LogicalResult matchAndRewrite(
KrnlTerminatorOp op, PatternRewriter &rewriter) const override {
rewriter.replaceOpWithNewOp<AffineTerminatorOp>(op);
return success();
}
@ -108,8 +118,8 @@ class KrnlDefineLoopsLowering : public OpRewritePattern<KrnlDefineLoopsOp> {
public:
using OpRewritePattern<KrnlDefineLoopsOp>::OpRewritePattern;
LogicalResult matchAndRewrite(KrnlDefineLoopsOp op,
PatternRewriter &rewriter) const override {
LogicalResult matchAndRewrite(
KrnlDefineLoopsOp op, PatternRewriter &rewriter) const override {
rewriter.eraseOp(op);
return success();
}
@ -123,8 +133,8 @@ class KrnlOptimizeLoopsLowering : public OpRewritePattern<KrnlOptimizeLoopsOp> {
public:
using OpRewritePattern<KrnlOptimizeLoopsOp>::OpRewritePattern;
LogicalResult matchAndRewrite(KrnlOptimizeLoopsOp op,
PatternRewriter &rewriter) const override {
LogicalResult matchAndRewrite(
KrnlOptimizeLoopsOp op, PatternRewriter &rewriter) const override {
rewriter.eraseOp(op);
return success();
}
@ -158,8 +168,7 @@ void KrnlToAffineLoweringPass::runOnFunction() {
OwningRewritePatternList patterns;
patterns.insert<KrnlIterateOpLowering, KrnlTerminatorLowering,
KrnlDefineLoopsLowering, KrnlOptimizeLoopsLowering>(
&getContext());
KrnlDefineLoopsLowering, KrnlOptimizeLoopsLowering>(&getContext());
if (failed(applyPartialConversion(getFunction(), target, patterns))) {
signalPassFailure();
@ -172,5 +181,5 @@ std::unique_ptr<Pass> mlir::createLowerKrnlPass() {
return std::make_unique<KrnlToAffineLoweringPass>();
}
static PassRegistration<KrnlToAffineLoweringPass> pass("lower-krnl",
"Lower Krnl dialect.");
static PassRegistration<KrnlToAffineLoweringPass> pass(
"lower-krnl", "Lower Krnl dialect.");

15
test/mlir/conversion/krnl_to_affine.mlir Normal file
View File

@ -0,0 +1,15 @@
// RUN: onnx-mlir-opt --lower-krnl %s -split-input-file | FileCheck %s
func @test_lower_degenerate_iterate(%arg0: memref<f32>) -> memref<f32> {
%0 = alloc() : memref<f32>
krnl.iterate() with () {
%1 = load %arg0[] : memref<f32>
store %1, %0[] : memref<f32>
}
return %0 : memref<f32>
// CHECK-LABEL: test_lower_degenerate_iterate
// CHECK-NEXT: [[ALLOC:%.+]] = alloc() : memref<f32>
// CHECK-NEXT: [[LOAD:%.+]] = load %{{.*}}[] : memref<f32>
// CHECK-NEXT: store [[LOAD]], [[ALLOC]][] : memref<f32>
// CHECK-NEXT: return [[ALLOC]] : memref<f32>
}