After running into failures compiling our OpenCL 1.2 code; I did some research and found that NVidia has not updated their implementation sine 1.1.
We offload XML processing onto GPUs and have been wanting to try NVidia cards for this task. Unfortunately they are lagging far behind on this matter as compared with the AMD counterpart, while I understand NVidia has CUDA and it should be a simple mapping for features up towards 2.0. There are many performance and language improvements in 1.2 at minimum which are really valuable for scientific computing and general processing outside of the graphics environment.
Meanwhile in graphics, OpenGL has been increasingly built to work with OpenCL and while CUDA can somewhat mimic the functionality quite well, many of our clients would like to deploy for NVidia cards with the same standards-driven code they use for smart phone and WebGL development. So these sorts of customers are being kept outside unfortunately =[
When will NVidia update their driver support from legacy OpenCL to OpenCL 2 or even 1.2? Will we instead need to target PTX for compiling OpenCL ? (yes if this takes longer than 6 months)