Is there any way to get annotations of OpenCL code in the generated assembly code... either when we dump the kernel code, or in the kernel analyzer? Some of my kernels become quite huge, and it would be nice if there was some way to tell what parts of the assembly correspond to what parts of the original C code...
AFAIK you can see the ISA or IL code in the SKA as well as in profiler.
also there is a environment variable GPU_DUMP_DEVICE_KERNEL which can be used to dump ISA or IL.
Detailed description about using these is given in ATI Stream SDK OpenCL Programming Guide.