********************************************************************************

                                CUDA Debugger API

                                   Version 4.2

                                  Release Notes

********************************************************************************

SUPPORT FOR SM_30 ARCHITECTURES (KEPLER)

The 4.2 CUDA Debugger API now supports the incoming Kepler architecture
(sm_30). All the existing API routines can be used with the new GPU
architecture. Using an older API version with Kepler is not supported and will
not probably work.


BINDLESS TEXTURES

Starting with sm_30, the textures are bindless. A new API call was added to
read the corresponding texture memory: readTextureMemoryBindless(). This new
API routine is only to be used for sm_30 GPUs. The existing readTextureMemory()
remains and must be used for all earlier architectures (sm_20 and lower).

The signature of both functions is identical. However, the expected texture id
argument value will differ. With readTextureMemory, the texture id is the value
read from EIATTR_IMAGE_SLOT in the ELF image. With readTextureMemoryBindless,
the texture id must be the index of the texture in the symbol table.
