// C99
double trace(size_t n, double mat[][n])
{
double sum = 0;
for(size_t i = 0; i < n; ++i)
sum += mat[i][i];
return sum;
}
/* C90 */
double trace(size_t n, double mat[])
{
double sum = 0;
size_t i = 0;
for(; i < n; ++i)
sum += mat[i * n + i];
return sum;
}
While this may not look like much of an improvement in simple cases, not having to manually emulate array subscription is quite convenient in more complex ones.
2. Allocation of variably-sized objects with automatic storage duration. Some libc implementations provide alloca() for that purpose -- unfortunately, it has issues (see eg http://c-faq.com/malloc/alloca.glb.html ).
In fact, we've de-facto had such a mechanism for decades
Unfortunately, that's not the case. There are actually two primary use cases for VLAs:
1. Variable-length multi-dimensional array parameters:
While this may not look like much of an improvement in simple cases, not having to manually emulate array subscription is quite convenient in more complex ones.2. Allocation of variably-sized objects with automatic storage duration. Some libc implementations provide alloca() for that purpose -- unfortunately, it has issues (see eg http://c-faq.com/malloc/alloca.glb.html ).