197 bool floorPrepare(const Shape& input, Shape* output) {  in floorPrepare()  argument
198     return SetShape(input, output);  in floorPrepare()
201 bool dequantizePrepare(const Shape& input, Shape* output) {  in dequantizePrepare()  argument
202     if (input.type != OperandType::TENSOR_QUANT8_ASYMM ||  in dequantizePrepare()
207     if (input.dimensions.size() != output->dimensions.size()) {  in dequantizePrepare()
211     output->dimensions = input.dimensions;  in dequantizePrepare()
215 bool convPrepare(const Shape& input,  in convPrepare()  argument
222     NN_OPS_CHECK(input.type == filter.type);  in convPrepare()
223     if (input.type == OperandType::TENSOR_QUANT8_ASYMM) {  in convPrepare()
226         NN_OPS_CHECK(input.type == bias.type);  in convPrepare()
228     NN_OPS_CHECK(getNumberOfDimensions(input) == 4);  in convPrepare()
233     NN_OPS_CHECK(getSizeOfDimension(filter, 3) == getSizeOfDimension(input, 3));  in convPrepare()
236     uint32_t width        = getSizeOfDimension(input, 2);  in convPrepare()
237     uint32_t height       = getSizeOfDimension(input, 1);  in convPrepare()
240     uint32_t batches      = getSizeOfDimension(input, 0);  in convPrepare()
247     output->type = input.type;  in convPrepare()
252 bool depthwiseConvPrepare(const Shape& input,  in depthwiseConvPrepare()  argument
259     NN_OPS_CHECK(input.type == filter.type);  in depthwiseConvPrepare()
260     if (input.type == OperandType::TENSOR_QUANT8_ASYMM) {  in depthwiseConvPrepare()
263         NN_OPS_CHECK(input.type == bias.type);  in depthwiseConvPrepare()
265     NN_OPS_CHECK(getNumberOfDimensions(input) == 4);  in depthwiseConvPrepare()
272     uint32_t width        = getSizeOfDimension(input, 2);  in depthwiseConvPrepare()
273     uint32_t height       = getSizeOfDimension(input, 1);  in depthwiseConvPrepare()
276     uint32_t batches      = getSizeOfDimension(input, 0);  in depthwiseConvPrepare()
283     output->type = input.type;  in depthwiseConvPrepare()
289 bool genericPoolingPrepare(const Shape& input,  in genericPoolingPrepare()  argument
295     NN_OPS_CHECK(getNumberOfDimensions(input) == 4);  in genericPoolingPrepare()
297     uint32_t batches      = getSizeOfDimension(input, 0);  in genericPoolingPrepare()
298     uint32_t width        = getSizeOfDimension(input, 2);  in genericPoolingPrepare()
299     uint32_t height       = getSizeOfDimension(input, 1);  in genericPoolingPrepare()
300     uint32_t channels_out = getSizeOfDimension(input, 3);  in genericPoolingPrepare()
307     output->type = input.type;  in genericPoolingPrepare()
313 bool genericActivationPrepare(const Shape& input,  in genericActivationPrepare()  argument
315     NN_OPS_CHECK(getNumberOfDimensions(input) <= 4);  in genericActivationPrepare()
316     return SetShape(input, output);  in genericActivationPrepare()
319 bool fullyConnectedPrepare(const Shape& input,  in fullyConnectedPrepare()  argument
325     NN_OPS_CHECK(input.type == weights.type);  in fullyConnectedPrepare()
326     if (input.type == OperandType::TENSOR_QUANT8_ASYMM) {  in fullyConnectedPrepare()
329         NN_OPS_CHECK(input.type == bias.type);  in fullyConnectedPrepare()
331     NN_OPS_CHECK(getNumberOfDimensions(input) >= 2);  in fullyConnectedPrepare()
332     uint32_t input_size = getNumberOfElements(input);  in fullyConnectedPrepare()
340     output->type = input.type;  in fullyConnectedPrepare()
388 bool genericNormalizationPrepare(const Shape& input, Shape* output) {  in genericNormalizationPrepare()  argument
389     NN_OPS_CHECK(getNumberOfDimensions(input) == 4);  in genericNormalizationPrepare()
390     return SetShape(input, output);  in genericNormalizationPrepare()
393 bool reshapePrepare(const Shape& input,  in reshapePrepare()  argument
401     int32_t numInputElements = (int32_t) getNumberOfElements(input);  in reshapePrepare()
424     output->type = input.type;  in reshapePrepare()
426     output->offset = input.offset;  in reshapePrepare()
427     output->scale = input.scale;  in reshapePrepare()
432 bool resizeBilinearPrepare(const Shape& input,  in resizeBilinearPrepare()  argument
436     NN_OPS_CHECK(getNumberOfDimensions(input) == 4);  in resizeBilinearPrepare()
437     uint32_t batches  = getSizeOfDimension(input, 0);  in resizeBilinearPrepare()
438     uint32_t channels = getSizeOfDimension(input, 3);  in resizeBilinearPrepare()
440     output->type = input.type;  in resizeBilinearPrepare()
446 bool depthToSpacePrepare(const Shape& input,  in depthToSpacePrepare()  argument
449     NN_OPS_CHECK(getNumberOfDimensions(input) == 4);  in depthToSpacePrepare()
452     uint32_t batches  = getSizeOfDimension(input, 0);  in depthToSpacePrepare()
453     uint32_t height   = getSizeOfDimension(input, 1);  in depthToSpacePrepare()
454     uint32_t width    = getSizeOfDimension(input, 2);  in depthToSpacePrepare()
455     uint32_t channels = getSizeOfDimension(input, 3);  in depthToSpacePrepare()
458     output->type = input.type;  in depthToSpacePrepare()
463     output->offset = input.offset;  in depthToSpacePrepare()
464     output->scale = input.scale;  in depthToSpacePrepare()
469 bool spaceToDepthPrepare(const Shape& input,  in spaceToDepthPrepare()  argument
472     NN_OPS_CHECK(getNumberOfDimensions(input) == 4);  in spaceToDepthPrepare()
475     uint32_t batches  = getSizeOfDimension(input, 0);  in spaceToDepthPrepare()
476     uint32_t height   = getSizeOfDimension(input, 1);  in spaceToDepthPrepare()
477     uint32_t width    = getSizeOfDimension(input, 2);  in spaceToDepthPrepare()
478     uint32_t channels = getSizeOfDimension(input, 3);  in spaceToDepthPrepare()
483     output->type = input.type;  in spaceToDepthPrepare()
488     output->offset = input.offset;  in spaceToDepthPrepare()
489     output->scale = input.scale;  in spaceToDepthPrepare()