Dynamically Allocating 2D Arrays Efficiently (and Correctly!) in C 2.0

Introduction

In my article Dynamically Allocating 2D Arrays Efficiently (and Correctly!) in C, I showed how to do exactly that by allocating a buffer then using pointers within it to point to the start of rows later in the same buffer.

It was recently pointed out to me by a reader that there’s a much simpler way to do it.

The Simpler Way

If you want to allocate an m × n 2D matrix, you can much more simply do:

int (*const a)[m] = malloc( m * n * sizeof(int) );
a[i][j] = 42;  // works
free( a );

where a is a constant pointer to an array of m integers. Although it’s using variable length array (VLA) syntax, it’s not allocating a VLA on the stack since malloc is being used. Yet the compiler is still generating VLA-like code in order to calculate the offset of the ith row at runtime, that is multiplying m × sizeof(int) — a variable, not a constant integer expression.

The only caveat is that gcc with the -Wvla warning enabled complains that a VLA is being used. I mean, it is sort-of, but not the allocate-on-the-stack part which is the dangerous part. The above code is safe in that it can’t overflow the stack.

Note that if you want to allocate a triangular matrix, then you’ll still need to use the method I described in my previous article.