OpenCL can be used for anything really, but interop with OpenGL is still problematic. Sharing buffers is easy enough, but when it’s time to synchronize CL computation with GL rendering, we’re forced to use clFinish/glFinish, sacrificing performance. It’s been almost 3 years and no vendor supports ARB_cl_event or KHR_gl_event yet.
ARB_compute_shader, which is compute via OpenGL/GLSL, might be a more approachable (and obviously less powerful) solution. There are no synchronization issues and you won’t have to learn another API. Only Nvidia supports this currently but we should see support from other vendors soon. Unfortunately the spec has bumped the required GL version to 4.2 (was 3.0 originally), so hardware support will be limited.