Designation of constant memory for device

Hello Forum,

For applications that leverage multiple GPU devices, how does one designate constant memory for a particular device? Or is that automatic?

I am currently using the function cudaMemcpyToSymbol to load constant memory from host memory. Do I do this from each host thread context if each GPU is to have the same constant memory content?

Thanks.

Hello Forum,

For applications that leverage multiple GPU devices, how does one designate constant memory for a particular device? Or is that automatic?

I am currently using the function cudaMemcpyToSymbol to load constant memory from host memory. Do I do this from each host thread context if each GPU is to have the same constant memory content?

Thanks.