chore: import upstream snapshot with attribution

2026-07-13 13:33:03 +08:00
commit 5b57521aa1
8226 changed files with 3425766 additions and 0 deletions
@@ -0,0 +1,75 @@
+//
+//  ShapeBroadcastTo.cpp
+//  MNN
+//
+//  Created by MNN on 2019/12/2.
+//  Copyright © 2018, Alibaba Group Holding Limited
+//
+
+#include "shape/SizeComputer.hpp"
+#include "core/Macro.h"
+#include "core/TensorUtils.hpp"
+
+namespace MNN {
+
+class ShapeBroadcastTo : public SizeComputer {
+    virtual bool onComputeSize(const MNN::Op* op, const std::vector<Tensor*>& inputs,
+                               const std::vector<Tensor*>& outputs) const override {
+        MNN_ASSERT(inputs.size() == 2);
+        MNN_ASSERT(outputs.size() == 1);
+        auto input  = inputs[0];
+        auto shape  = inputs[1];
+        auto output = outputs[0];
+        int inputDims = input->dimensions();
+        int shapeDims = shape->elementSize();
+        output->buffer().dimensions = inputDims > shapeDims ? inputDims : shapeDims;
+        const int dimension = output->dimensions();
+        const int* shapeData        = shape->host<int>();
+        if (op->main() && op->main_as_Axis() && op->main_as_Axis()->axis()) {
+            for (int i = 0; i < dimension; i++) {
+                output->setLength(i, shapeData[i]);
+            }
+        } else {
+            int offset;
+            int alignShape[MNN_MAX_TENSOR_DIM];
+            if (inputDims > shapeDims) {
+                for (int i = 0; i < input->dimensions(); ++i) {
+                    output->setLength(i, input->length(i));
+                }
+                offset = inputDims - shapeDims;
+                for (int i=0; i<shapeDims; ++i) {
+                    alignShape[i] = shapeData[i];
+                }
+            } else {
+                for (int i = 0; i < shapeDims; ++i) {
+                    output->setLength(i, shapeData[i]);
+                }
+                for (int i=0; i<input->dimensions(); ++i) {
+                    alignShape[i] = input->length(i);
+                }
+                offset = shapeDims - inputDims;
+            }
+            for (int i = offset; i < output->dimensions(); ++i) {
+                int dim1 = alignShape[i - offset];
+                int dim2 = output->length(i);
+                if (dim1 != dim2 && (dim1 != 1 && dim2 != 1)) {
+                    MNN_ERROR("Broad cast error, dim1 = %d, dim2 = %d\n", dim1, dim2);
+                    return false;
+                }
+                if (dim1 == dim2) {
+                    continue;
+                }
+                if (dim1 != 1) {
+                    output->setLength(i, dim1);
+                }
+            }
+        }
+        output->buffer().type                             = input->buffer().type;
+        TensorUtils::getDescribe(output)->dimensionFormat = TensorUtils::getDescribe(input)->dimensionFormat;
+        return true;
+    }
+};
+
+REGISTER_SHAPE_INPUTS(ShapeBroadcastTo, OpType_BroadcastTo, {1});
+
+} // namespace MNN