5.14.3.10 The #pragma Shift Directive

The #pragma shift directive allows the programmer to control which encoding strategy is employed by the compiler for individual shift operations in project's source code. For more information on how shift operations can be encoded, see Shift Operators.

The shift pragma has the general form:

#pragma shift strategy

where strategy can be one of speed, space, or auto.

If this pragma is not in effect, the compiler will choose a strategy for all shift operations that will favor speed execution when -O3 level optimizations are selected, or favor code size for all other optimization levels. Choosing the auto strategy makes this choice explicit. The speed or space strategies can be specified with this pragma to encode any subsequent shift operations (up to the end of the translation unit or function, or the next shift pragma) so that they will execute in a shorter time or take up minimal code space, respectively. This, for example, would allow you to select -Os optimizations for the whole program to reduce the total code size, but still allow any number of individually selected shift operations to be executed as fast as possible.

Additionally, the current shift strategy can be pushed to an internal stack using the push pragma argument, and restored at a later point using the pop argument. For example, the code:

#pragma shift push
#pragma shift speed
    dispVal = reading << 5;
#pragma shift pop

will save the current shift strategy, then encode the shift operation favoring speed, then return the shift strategy to what was selected earlier.

At the end of the current translation unit, the compiler's internal stack is emptied and the current shift strategy will be restored to the default for the current optimization level.