There are two versions that developers need to care about when developing Cuda applications: computing capability-describe product specifications and computing device features and Cuda driver API version-Describe the features supported by the driver API and runtime.
You can obtain the driver API version from the macro cuda_version in the driver header file. Developers can check whether their applications require a version later than the existing version. This is important because the driver API is backward compatible, meaning that applications, plug-ins, and libraries compiled for the special edition (including C Runtime) it can continue running on the device drivers of later releases, as shown in 11. The driver is not forward compatible, meaning that the applications, plug-ins, and libraries (including C Runtime) compiled for the special edition cannot run on the device driver of the later releases.
It is important to remember versions that do not support mixed match, especially:
- Because the system can only install one Cuda device driver, all applications, plug-ins and libraries in the system must use the same version of The Cuda driver API.
- The plug-ins and libraries used by the application must use the same version of runtime.
- The plug-ins and libraries used by applications must use any runtime libraries of the same version.
Figure 11 the API driver is backward-compatible, not forward-compatible.