I wish to know how much time data transmission between CPU and GPU costs so that I can try to improve the overall WebGPU programs' performance. But till now I only know few ways to measure it and I don't really know are they correct.

• Use "timestamp-query" feature. It seems that it can only measure compute passes' cost as GPUComputePassTimestampWrites only has 2 time point for measurement (pass beginning and end). So it can't be used for instructions like queue.writeBuffer(). See also https://stackoverflow.com/questions/76713260/how-to-time-profile-see-when-tasks-are-complete-in-wgpu-targeting-web.
• Use queue.onSubmittedWorkDone()before and after the queue.writeBuffer() and calculate the time cost in JS promise callbacks. I don't know is it correct and it seems not precise intuitively.
• Use other debugger tools, like DevTools when running WebGPU programs in Chrome. In DevTools I can only get a very sketchy view about how much time it costs before and after GPU executing some tasks.

So I wonder to know how to measure the time cost of:

• Transmitting data from CPU to GPU, including createBuffer() and queue.writeBuffer();
• Transmitting data from GPU to CPU, including createBuffer() for staging buffer and buffer.mapAsync() when reading.

I've tried searched keywords like 'track''performace''writeBuffer' but can't get an answer. If any one can give me some ideas or figure out any mistakes in my thought I'll be much appreciated. Thank you!