Stack frames

Written by Julie Zelenski

Stack intuition

Functions often need space. For example, they allocate variables or store the return address of their caller before calling another function. There's nothing special about this space, and we could allocate it as we would any other memory. However, functions calls are frequent, so we want them as fast as possible. Fortunately, function calls have two important properties we can exploit for speed:

  1. When functions return, all allocated memory is no longer usable.
  2. Functions return in Last-In-First-Out (LIFO) order.

To optimize for speed, people have converged on using a contiguous region of memory (called a stack because it's LIFO, like a stack of plates). Roughly speaking it works as follows:

  • At program start, we reserve a fixed-sized region to hold the stack and set a pointer (the stack pointer) to the start of that region.
  • When entering a new function, the memory the function needs is allocated in a contingous single chunk by adjusting the stack pointer. This memory is referred to as the function's stack frame.
  • When exiting the function, the memory used for its stack frame is recycled by restoring the stack pointer to its position when the function was entered.

On the Mango Pi, the stack grows downward. When making a function call, the data for the callee function is stored at a lower address than the caller's data. It looks like this:

stack picture

This organization is so effective that compilers explicitly support management of the stack pointer and stack frames, architecture manuals provide the rules for how to do so, and an ISA may even include designated instructions for stack push and pop.

Managing a stack on RISC-V architecture

Register x2 (synonym sp) is designated for use as the stack pointer. sp points to the location of the last value added to the stack. This value is referred to as the "top" of the stack. (yes, it is confusing that the top is stored at the lowest address!)

The stack operates as a descending stack. This stack pointer is initialized to its highest address. Adding a new value to the stack moves the sp downward (to a lower address). Removing a value from the stack is accomplished by incrementing the sp.

When entering a function, the sp is adjusted downward to reserve space for the new stack frame.

addi sp,sp,-16

Registers that need to be saved store the value into the stack frame using a sp-relative offset.

sd ra,8(sp)
sd fp,0(sp)

When exiting the function, load the saved contents from the stack frame to restore the register's saved value.

ld ra,8(sp)
ld fp,0(sp)

And finally adjust the stack pointer upward to "remove" the stack frame:

addi sp,sp,16

The data being stored on the stack might be caller-owned registers or intermediate results. Large local variables (e.g. arrays and structs) that are too large to store in a register are also stored on the stack. The adjustment to the sp will be the total space needed for all the saved registers and local variables.

Use of frame pointer

The executing function can access data on the stack using relative offsets from the sp register. The compiler has latitude in what it stores on the stack; there is no guarantee about what values are stored in each frame and how values are laid out. This makes it challenging if trying to ascertain where each stack frame starts and stops.

Applying the compiler flag f-no-omit-frame-pointer tells the compiler to use a frame pointer when managing the stack. This enforces a particular order on the stack frame layout. The first two values pushed by a function will always be the saved values of the two registers ra and fp. In addition, the register s0 is used a dedicated "frame pointer" (fp). Whereas the sp points to the last word of the current frame, fp is set to point to the last word of the previous frame. Thus the range from fp to sp bookends the extent of the current stack frame. The fp tells you where to access the start of the frame of the currently executing function.

stack diagram

Above is a diagram taken from the lecture slides. Each stack frame begins by pushing saved values for the ra and fp registers. The saved fp within a stack frame points to the saved fp of the caller's frame. Following the sequence of fp links will lead back to the main function. This information is used by the debugger to produce a backtrace of the current call stack.