I'm familiar with CUDA, but new to Intel OpenCL programming. I'm wondering if there is a document where I could find the warp size, and shared memory size for Intel HD graphics 4000 in Ivy Brdige. Thanks!
The Intel® SDK for OpenCL (both the 2012 version and the 2013 beta) come with a very good optimization guide. I recommend you check that out. There is an online version too.