nnicolosi/Raptor
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

automatic code reformat

Browse Source
This commit is contained in:
NiccoloN
2026-05-22 15:23:48 +02:00
parent d136136d22
commit 8337a11ce9
37 changed files with 312 additions and 354 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/PIM/Conversion/ONNXToSpatial/Patterns/Math/Gemm.cpp
+7 -4
View File
@@ -423,8 +423,11 @@ LogicalResult GemvToSpatialCompute::matchAndRewrite(ONNXGemmOp gemmOp,
SmallVector<Value> vmmOutputs;
vmmOutputs.reserve(aHSlices[coreId].size());
for (auto aHSliceId : llvm::seq<size_t>(0, aHSlices[coreId].size()))
vmmOutputs.push_back(spatial::SpatVMMOp::create(
rewriter, gemmLoc, currOutHSliceType, computeOp.getWeightArgument(aHSliceId), computeOp.getInputArgument(aHSliceId)));
vmmOutputs.push_back(spatial::SpatVMMOp::create(rewriter,
gemmLoc,
currOutHSliceType,
computeOp.getWeightArgument(aHSliceId),
computeOp.getInputArgument(aHSliceId)));
if (vmmOutputs.empty()) {
gemmOp.emitOpError("requires at least one non-empty slice when lowering tiled Gemm to Spatial VMMs");
return failure();
@@ -579,8 +582,8 @@ LogicalResult GemmToSpatialComputeBatch::matchAndRewrite(ONNXGemmOp gemmOp,
rewriter.setInsertionPointToStart(&inParallelOp.getRegion().front());
SmallVector<OpFoldResult> outputOffsets {lane, rewriter.getIndexAttr(0)};
SmallVector<OpFoldResult> outputSizes {rewriter.getIndexAttr(1), rewriter.getIndexAttr(outType.getDimSize(1))};
tensor::ParallelInsertSliceOp::create(rewriter, loc, laneResult, packedOutput, outputOffsets, outputSizes,
unitStrides);
tensor::ParallelInsertSliceOp::create(
rewriter, loc, laneResult, packedOutput, outputOffsets, outputSizes, unitStrides);
rewriter.setInsertionPointAfter(batchOp);
rewriter.replaceOp(gemmOp, batchOp.getResults());
src/PIM/Conversion/ONNXToSpatial/Patterns/Math/MatMul.cpp
+11 -23
View File
@@ -38,23 +38,16 @@ static FailureOr<SmallVector<int64_t>> inferSupportedBatchShape(ArrayRef<int64_t
return SmallVector<int64_t>(lhsBatchShape.begin(), lhsBatchShape.end());
}
static Value collapseBatchDims(Value value,
int64_t batchSize,
int64_t rows,
int64_t cols,
PatternRewriter& rewriter,
Location loc) {
static Value
collapseBatchDims(Value value, int64_t batchSize, int64_t rows, int64_t cols, PatternRewriter& rewriter, Location loc) {
auto type = cast<RankedTensorType>(value.getType());
if (type.getRank() == 2 || type.getRank() == 3)
return value;
auto collapsedType =
RankedTensorType::get({batchSize, rows, cols}, type.getElementType(), type.getEncoding());
SmallVector<ReassociationIndices> reassociation = {
ReassociationIndices {},
ReassociationIndices {static_cast<int64_t>(type.getRank() - 2)},
ReassociationIndices {static_cast<int64_t>(type.getRank() - 1)}
};
auto collapsedType = RankedTensorType::get({batchSize, rows, cols}, type.getElementType(), type.getEncoding());
SmallVector<ReassociationIndices> reassociation = {ReassociationIndices {},
ReassociationIndices {static_cast<int64_t>(type.getRank() - 2)},
ReassociationIndices {static_cast<int64_t>(type.getRank() - 1)}};
for (int64_t dim = 0; dim < type.getRank() - 2; ++dim)
reassociation.front().push_back(dim);
@@ -72,19 +65,14 @@ static Value collapseBatchDims(Value value,
return collapseCompute.getResult(0);
}
static Value expandBatchDims(Value value,
RankedTensorType outputType,
size_t batchRank,
PatternRewriter& rewriter,
Location loc) {
static Value
expandBatchDims(Value value, RankedTensorType outputType, size_t batchRank, PatternRewriter& rewriter, Location loc) {
if (cast<RankedTensorType>(value.getType()) == outputType)
return value;
SmallVector<ReassociationIndices> reassociation = {
ReassociationIndices {},
ReassociationIndices {static_cast<int64_t>(batchRank)},
ReassociationIndices {static_cast<int64_t>(batchRank + 1)}
};
SmallVector<ReassociationIndices> reassociation = {ReassociationIndices {},
ReassociationIndices {static_cast<int64_t>(batchRank)},
ReassociationIndices {static_cast<int64_t>(batchRank + 1)}};
for (size_t dim = 0; dim < batchRank; ++dim)
reassociation.front().push_back(static_cast<int64_t>(dim));