CreateTensor with tensor size of bigger than 65536 causes BAD_ACCESS in Run on macOS

[jie-chen]

Describe the issue

When the size of Output Tensor is bigger than 65536 (e.g. 131024), the BAD_ACCESS exception will be raised in Run(), this only happens on macOS. Windows does not have such problem.

To reproduce

auto memory_info = Ort::MemoryInfo::CreateCpu(OrtDeviceAllocator, OrtMemTypeCPU);
auto input_tensor = Ort::Value::CreateTensor(memory_info, input.data(), input.size(), input_shape.data(), input_shape.size());
auto output_tensor = Ort::Value::CreateTensor(memory_info, output.data(), output.size(), //bigger than 65536
output_shape.data(), output_shape.size());
Ort::AllocatorWithDefaultOptions ort_alloc;
std::optionalOrt::AllocatedStringPtr input_name;
std::optionalOrt::AllocatedStringPtr output_name;
input_name.emplace(sessions[index]->GetInputNameAllocated(0, ort_alloc));
output_name.emplace(sessions[index]->GetOutputNameAllocated(0, ort_alloc));
const char* input_names[] = {input_name->get()};
char* output_names[] = {output_name->get()};
session->Run(Ort::RunOptions{nullptr}, input_names, &input_tensor, 1, output_names, &output_tensor, 1);

Urgency

No response

Platform

Mac

OS Version

14.0

ONNX Runtime Installation

Released Package

ONNX Runtime Version or Commit ID

1.14.0

ONNX Runtime API

C++

Architecture

ARM64

Execution Provider

Default CPU

Execution Provider Library Version

No response

[yuslepukhin]

Check whether the API expects size in bytes or in elements, and whether your shape matches it.

[jie-chen]

Check whether the API expects size in bytes or in elements, and whether your shape matches it.

Hi, thanks for replying. Given the parameters passed in, it should be the first one of CreateTensor whose size is in elements (https://onnxruntime.ai/docs/api/c/struct_ort_1_1_value.html#a3898146b5fc35b838bd48db807dd6c8e).

As I mentioned this only happens on macOS with model with bigger output tensor size. Lowering the model output tensor size or running the same model on Windows won't raise the exception. Please see the details of exception in screenshot

[github-actions]

This issue has been automatically marked as stale due to inactivity and will be closed in 7 days if no further activity occurs. If further support is needed, please provide an update and/or more details.