292 bool reshapePrepare(const Shape& input, const int32_t* targetDims, const int32_t targetDimsSize,  in reshapePrepare()  argument
298     int32_t numInputElements = (int32_t)getNumberOfElements(input);  in reshapePrepare()
321     output->type = input.type;  in reshapePrepare()
323     output->offset = input.offset;  in reshapePrepare()
324     output->scale = input.scale;  in reshapePrepare()
329 bool depthToSpacePrepare(const Shape& input, int32_t blockSize, Shape* output) {  in depthToSpacePrepare()  argument
330     NN_OPS_CHECK(getNumberOfDimensions(input) == 4);  in depthToSpacePrepare()
333     uint32_t batches = getSizeOfDimension(input, 0);  in depthToSpacePrepare()
334     uint32_t height = getSizeOfDimension(input, 1);  in depthToSpacePrepare()
335     uint32_t width = getSizeOfDimension(input, 2);  in depthToSpacePrepare()
336     uint32_t channels = getSizeOfDimension(input, 3);  in depthToSpacePrepare()
339     output->type = input.type;  in depthToSpacePrepare()
342     output->offset = input.offset;  in depthToSpacePrepare()
343     output->scale = input.scale;  in depthToSpacePrepare()
348 bool spaceToDepthPrepare(const Shape& input, int32_t blockSize, Shape* output) {  in spaceToDepthPrepare()  argument
349     NN_OPS_CHECK(getNumberOfDimensions(input) == 4);  in spaceToDepthPrepare()
352     uint32_t batches = getSizeOfDimension(input, 0);  in spaceToDepthPrepare()
353     uint32_t height = getSizeOfDimension(input, 1);  in spaceToDepthPrepare()
354     uint32_t width = getSizeOfDimension(input, 2);  in spaceToDepthPrepare()
355     uint32_t channels = getSizeOfDimension(input, 3);  in spaceToDepthPrepare()
360     output->type = input.type;  in spaceToDepthPrepare()
363     output->offset = input.offset;  in spaceToDepthPrepare()
364     output->scale = input.scale;  in spaceToDepthPrepare()
410 bool padPrepare(const Shape& input, const int32_t* paddingsData, const Shape& paddingsShape,  in padPrepare()  argument
412     uint32_t numInputDims = getNumberOfDimensions(input);  in padPrepare()
426         outDims[i] = beforePadding + getSizeOfDimension(input, i) + afterPadding;  in padPrepare()
428     output->type = input.type;  in padPrepare()
430     output->offset = input.offset;  in padPrepare()
431     output->scale = input.scale;  in padPrepare()
436 bool batchToSpacePrepare(const Shape& input, const int32_t* blockSizeData,  in batchToSpacePrepare()  argument
439     NN_OPS_CHECK(getNumberOfDimensions(input) == 4);  in batchToSpacePrepare()
447     uint32_t batches = getSizeOfDimension(input, 0);  in batchToSpacePrepare()
448     uint32_t height = getSizeOfDimension(input, 1);  in batchToSpacePrepare()
449     uint32_t width = getSizeOfDimension(input, 2);  in batchToSpacePrepare()
450     uint32_t channels = getSizeOfDimension(input, 3);  in batchToSpacePrepare()
453     output->type = input.type;  in batchToSpacePrepare()
456     output->offset = input.offset;  in batchToSpacePrepare()
457     output->scale = input.scale;  in batchToSpacePrepare()
462 bool spaceToBatchPrepare(const Shape& input, const int32_t* blockSizeData,  in spaceToBatchPrepare()  argument
466     NN_OPS_CHECK(getNumberOfDimensions(input) == 4);  in spaceToBatchPrepare()
480     uint32_t batches = getSizeOfDimension(input, 0);  in spaceToBatchPrepare()
481     uint32_t height = getSizeOfDimension(input, 1);  in spaceToBatchPrepare()
482     uint32_t width = getSizeOfDimension(input, 2);  in spaceToBatchPrepare()
483     uint32_t channels = getSizeOfDimension(input, 3);  in spaceToBatchPrepare()
491     output->type = input.type;  in spaceToBatchPrepare()
495     output->offset = input.offset;  in spaceToBatchPrepare()
496     output->scale = input.scale;  in spaceToBatchPrepare()
501 bool meanPrepare(const Shape& input, const int32_t* axisData, const Shape& axisShape, bool keepDims,  in meanPrepare()  argument
507     int32_t numInputDims = static_cast<int32_t>(getNumberOfDimensions(input));  in meanPrepare()
524                 outDims[idx] = getSizeOfDimension(input, idx);  in meanPrepare()
561                 outDims[idx - numSkipAxis] = getSizeOfDimension(input, idx);  in meanPrepare()
571     output->type = input.type;  in meanPrepare()
572     output->offset = input.offset;  in meanPrepare()
573     output->scale = input.scale;  in meanPrepare()
578 bool argMinMaxPrepare(const Shape& input, int32_t axis, Shape* output) {  in argMinMaxPrepare()  argument
579     NN_CHECK(handleNegativeAxis(input, &axis));  in argMinMaxPrepare()
585     if (getNumberOfDimensions(input) > 1) {  in argMinMaxPrepare()
586         output->dimensions.reserve(getNumberOfDimensions(input) - 1);  in argMinMaxPrepare()
587         output->dimensions.insert(output->dimensions.end(), input.dimensions.begin(),  in argMinMaxPrepare()
588                                   input.dimensions.begin() + axis);  in argMinMaxPrepare()
589         output->dimensions.insert(output->dimensions.end(), input.dimensions.begin() + axis + 1,  in argMinMaxPrepare()
590                                   input.dimensions.end());  in argMinMaxPrepare()
598 bool splitPrepare(const Shape& input, int32_t axis, int32_t numOutputs,  in splitPrepare()  argument
600     NN_CHECK(handleNegativeAxis(input, &axis));  in splitPrepare()
602     const int32_t sizeOfAxisToSplit = input.dimensions[axis];  in splitPrepare()
607         output->at(i).type = input.type;  in splitPrepare()
608         output->at(i).dimensions = input.dimensions;  in splitPrepare()
610         output->at(i).offset = input.offset;  in splitPrepare()
611         output->at(i).scale = input.scale;  in splitPrepare()
616 bool groupedConvPrepare(const Shape& input, const Shape& filter, const Shape& bias,  in groupedConvPrepare()  argument
621         NN_OPS_CHECK(input.type == OperandType::TENSOR_QUANT8_ASYMM ||  in groupedConvPrepare()
622                      input.type == OperandType::TENSOR_QUANT8_ASYMM_SIGNED);  in groupedConvPrepare()
624         NN_OPS_CHECK(input.type == filter.type);  in groupedConvPrepare()
626     if (input.type == OperandType::TENSOR_QUANT8_ASYMM ||  in groupedConvPrepare()
627         input.type == OperandType::TENSOR_QUANT8_ASYMM_SIGNED) {  in groupedConvPrepare()
630         NN_OPS_CHECK(input.type == bias.type);  in groupedConvPrepare()
632     NN_OPS_CHECK(getNumberOfDimensions(input) == 4);  in groupedConvPrepare()
638     NN_OPS_CHECK(getSizeOfDimension(filter, 3) * numGroups == getSizeOfDimension(input, 3));  in groupedConvPrepare()
642     uint32_t width = getSizeOfDimension(input, 2);  in groupedConvPrepare()
643     uint32_t height = getSizeOfDimension(input, 1);  in groupedConvPrepare()
646     uint32_t batches = getSizeOfDimension(input, 0);  in groupedConvPrepare()
658     output->type = input.type;  in groupedConvPrepare()