In sycl/dpc++ malloc_shared I am aware that it is possible to create a buffer which could be shared between the host and a single gpu with the below function.
void* malloc_shared(size_t num_bytes,
const sycl::device& dev,
const sycl::context& ctxt);
or
int *data = malloc_shared<int>(N, q);
I wanted to know if something existed that could possibly share the same data/buffer across multiple gpus? Something like the below
int *data = malloc_shared<int>(N, q1,q2);