6.1.2 Data Memory Models

The MSP430 compiler supports three different data memory models: small, restricted and large. The data model used is controlled by the --data_model option. The 16-bit MSP430 devices always use the small data memory model. The 20-bit MSP430X devices can use any data memory model and use the small data model by default.

• The small data model requires that all data sections be located in the low 64K of memory. Data pointers are 16-bits in size. This is the most efficient data model in terms of performance and application size.
• The restricted data model allows data to be located throughout the entire 1MB address space available on MSP430X devices with only a minimal efficiency penalty over the small data model. It is restricted because individual objects (structures, arrays, etc.) cannot be larger than 64K in size. Data pointers are 32-bits in size.
• The large data model also allows data to be located throughout the entire 1MB address space and also places no restriction on the maximum size of an individual object. Permitting individual objects to be greater than 64K in size causes code generated for the large data model to be less efficient than code generated for the restricted data model.

Data memory is also affected by the --near_data option as described in Section 6.1.3.

The maximum size of an object (size_t) and the maximum difference between two pointers (ptrdiff_t) are increased from 16-bits to 32-bits in the large data model. Applications that rely on size_t or ptrdiff_t to be a specific size may need to be updated.

Object files built with different data models are not compatible. All files in an application must be built with the same data model. Additionally, a run-time-support library matching that data model must be used. When using automatic library selection (the default), the linker will automatically select the correct library Section 4.3.1.1. If you select the library manually, you must select the matching library according to Section 7.1.8.