Is Intel planning full OpenCL 2.1+ with Spir-v support for their Xe Discrete Graphics Card?
Will there be a product that competes with the Radeon 7's 3.46 Tflops of fp64 performance?
How will native fp64 precision compare with Nvidia's and AMD's?
How will precision compare with the AVX512 ER extension for the relevant fp64 operations?
Will there be tight integration between opencl on the cpu, igpu and the Xe?
Will this info be available before 2020?
Should I hold off on a Radeon 7...?
Thanks for the general interest and the interest in heterogeneous programming support.
On this forum, we can look to discuss Intel product developer implications when they become available... Which is whenever there is a product release or more rarely with respect to specific details in a prerelease announcement. Unfortunately, we can't really comment on non-Intel products in this forum.
That being said, sharing your OpenCL or heterogeneous programming goals or priorities can help us shape future guidance. In particular, it could be useful to know more about what developers like you feel 'tight integration' means... and also, how that could improve your project. This is an open ended invitation for developer feedback, but I'll also look to contact you individually.