[Feature Request] Cooperative Vectors Support

[pow2clk]

Support for cooperative vectors allowing long vector operations that are distributed across a wave.

Is your feature request related to a problem? Please describe.
Enables efficient operations for larger matrices and vectors for operations useful to ML.

Describe the solution you'd like

Prereqs:

• DXIL Native vector support [Feature Request] Support Long Vectors in HLSL #7067
• Native vector execution tests [Feature Request] Execution Test for DXIL Native Vectors #7068
• HLSL Long vector support [Feature Request] Support Native Vectors In Dxil #7066
• Long vector execution tests [Feature Request] Execution Tests for Long Vectors #7069

More implementation details to come

[devshgraphicsprogramming]

But wouldn't this be smarter to be regular float or vector<float,N/SubgroupSize> private variables and use subgroup ops to get them?