It is not clear what you are asking here. The Khronos-defined CL / GL switch is still limited to shared buffers / textures and some event processing materials, none of which are tessellation-related. Since tessellation is a specific task for rasterizing polygon-based models, it is unlikely to ever be relevant enough to guarantee a mention in the CL specification. If you are interested in whether the tessellation hardware will be used by the CL implementation, then it depends entirely on the specific hardware capabilities and CL implementation.
Update:
There are two types of tessellation shaders. Tessellation control shaders take a patch (a collection of triangles or ATVs) and calculate some attributes on top, most importantly, a number that controls the number of units to be executed.
After starting the tessellation control shader, the resulting patch is transferred to the hardware block of a fixed function, which performs the actual division to create a new patch with a large number of polygons.
After dividing, the Tedellation Evaluation Shader can calculate the attributes for each vertex in the patch generated by the separation process. This is a step in the process when, for example, you find the search information on an offset map stored as a texture.
Since interop CL / GL allows only buffer-based data exchange in order to use CL for tessellation, you will need to implement the entire tessellation process as essentially a preprocessor for your vertex data, and you will not be able to access from CL any dedicated equipment for the division step. Since current equipment (as far as I know) includes a unit with a fixed function instead of using general purpose computing units, the tessellator implemented in CL will almost certainly have a performance disadvantage.