11.9 Memory Models

The compiler supports several memory models. Command-line options are available for selecting the optimum memory model for your application, based on the specific device that you are using and the type of memory usage.

Table 11-1. Memory Model Command-Line Options
Option	Memory Definition	Description
-msmall-data	Place all data in Near Data Space (Notes 1, 2).	Permits the use of direct operations on memory, like PIC18 devices.
-msmall-scalar	Place `scalar` variables into Near Data Space. This is the default (Notes 1, 3).	Scalar variables, single object entities such as `int` or `float` variables, are allocated into Near Data Space.
-msmall-code	Use short-calls and branches to transfer control. This is the default (Note 4).	Function calls use the `RCALL` instruction.
-mlarge-code	Use long-calls and goto instructions, where possible.	Function calls use the `CALL` instruction.
-mconst-in-data	`const` qualified objects are allocated in data memory.	Values copied from program memory by startup code.
-mconst-in-code	`const` qualified objects located in program memory.	Values accessed are directly from program Flash.
Note: The extent of Near Data Space is Device Dependent; check the data sheet of the device you are using. If the memory space is fully consumed, the linker will issue an error message. The language tool will default to small data memory models when the selected device memory does not exceed the near data size. The small scalar memory model can be combined with the large data model to effectively place all scalars in the near Data Space and all aggregates (structures, unions, arrays) outside of the Near Data Space. The linker can place such objects in the Near Data Space if it is not fully consumed. Unlike the data memory models, the code size is not related to the total size of the program text but rather the distance between branches and calls. Note that a conditional branch instruction is always relative, so a single function may have a size limit.

The command-line options apply globally to the modules being compiled. Individual variables and functions can be declared as near or far to better control the code generation. For information on setting individual variable or function attributes, see Variable Attributes and Function Specifiers.