ISO/IEC 9899:2011 6.7.6.3 语义 7 是什么意思？

Question

first at all, here is the original(ISO/IEC 9899:2011 6.7.6.3 Semantics 7):

A declaration of a parameter as ''array of type'' shall be adjusted to ''qualified pointer to type'', where the type qualifiers (if any) are those specified within the [ and ] of the array type derivation. If the keyword static also appears within the [ and ] of the array type derivation, then for each call to the function, the value of the corresponding actual argument shall provide access to the first element of an array with at least as many elements as specified by the size expression.

In order to better understand this problem, I write some code:

#include <stdio.h>

void f(int a[static 10][10])
{
    printf("0x%p\n", a);
}

int main(int argc, char *argv[])
{
    int a[][10] = {[20] = {1, 2}};
    f(a);
    int b[][10] = {{3, 4}};
    f(b);
    return 0;
}

what is that static (int a[static 10][10]) stand for?
Is the behavior of that code undefined? why?
Why do we only need to provide access to the first element of an array not all of it?

Answer 1

what is that static(int a[static 10][10]) stand for?

As it is written in the quote provided by you:

... If the keyword static also appears within the [ and ] of the array type derivation, then for each call to the function, the value of the corresponding actual argument shall provide access to the first element of an array with at least as many elements as specified by the size expression .

Is the behavior of that code undefined? why?

Yes, the behavior of a function will be undefined if it expects an array with the specified number of elements in the size expression but the corresponding argument of the array type has fewer elements, because the function will try to access memory beyond the number of elements in the array used as an argument.

Why do we only need to provide access to the first element of an array not all of it?

We need to provide access at least to the number of elements specified in the parameter declaration in the size expression because it is a contract between the function and the user of the function. The function expects at least the number of elements specified in the size expression.

Answer 2

The purpose of the qualifier is to let a compiler know that it may, at its leisure, generate code which reads any portion of the array without regard for whether or not it will be needed. As an example of when that might be useful, consider something like:

int get_first_positive_of_four(int arr[static 4])
{
  if (arr[0] > 0) return arr[0];
  if (arr[1] > 0) return arr[1];
  if (arr[2] > 0) return arr[2];
  if (arr[3] > 0) return arr[3];
  return 0;
}

Some processors have an instruction that will read multiple words of memory into registers in a manner that's faster than repeated single reads. On the ARM7-TDMI, for example (somewhat obsolete architecture, but its instruction timing is easy to reason about), an instruction that loads one word into a register takes 3 cycles, and one that loads N words takes 2+N cycles. Thus, loading all four words with one instruction will take the same amount of time as performing two separate loads. If a compiler were to unconditionally load all four words, this code would take longer than a version which uses individual loads in the case where only the first value gets used, the same amount of time if two values get used, and less time if three or four values get used.

In the event that the function was passed a two-item array which was located right at the end of accessible memory, and whose second value is positive, behavior would be defined in the absence of the static qualifier, but the version of the code which preloads all four values might trap with a memory fault. Thus, that optimization wouldn't be valid without the static qualifier, but would be valid with it.

I don't know to what extent compilers exploit this qualifier, but it allows compilers to generate code that speculatively reads data that may or may not end up being required, which may allow faster execution than would be possible if the speculative reads of such data were forbidden.