Texas Instruments MSP50C614 Division, Function Calls, Stack frame has the following structure

Implementation Details

5.10.2 Division

The integer division currently requires the use of several accumulator pointers. We divide a 16 bit integer located in A0 by a 16 bit integer located in A0~. We return the quotient in A0~, and the remainder in A0. We make use of A3~ and A3 for scratch pads. We also set flag 1 if a division by zero is attempted, and zero out the quotient and the remainder in this case. We also use PH for temporary storage of the divisor.

5.10.3 Function Calls

Every function is associated with a stack frame. A regular C program is initially given control by a call to main(). A C± ± program starts with a jump to the _main symbol, which must therefore be present in the C± ± source code.

The stack frame has the following structure:

BP

SP

First Argument

•••

Last Argument

Return Address

Previous BP

Locals

Low Address

⇓

⇓

⇓

⇓

⇓

High Address

BP is the frame pointer (base pointer), SP the stack pointer.

We use R7 for stack pointer, and yet another register for BP, REG_BP (R5, because of its special arithmetic capabilities). Before a function is called, the arguments are pushed on the stack, first argument first. The function call automatically pushes the return address on the stack. Immediately upon entering the function body, the current BP is pushed on the stack to preserve it, so that the stack pointer now points to the next location. This location is copied to REG_BP, which becomes our fixed reference point for the current function. Locals are then allocated on the stack from this starting location.

When the function returns, SP is made to point to the return address, after the previous BP is popped. The return is performed by a RET instruction. The calling routine is then responsible for moving the stack pointer to its previous location, before the arguments were put on the stack. Because all functions return via A0, the only function return type allowed is integer. Our implementation of C± ± allows for function prototyping, and checks that prototyped functions are called with the correct number of arguments.

5-50