For a multi-device rendering setup using Shader Model 2.0:
1 LPDIRECT3D9 with 2 or more LPDIRECT3DDEVICE9 objects.
Should I load/compile my shaders (HLSL) based on number of LPDIRECT3DDEVICE9 objects. Then provide each shader to its respective device object. Alternatively, is it expected to load/compile the shaders once and the GPU will provide the bytecode to each device auto-magically?