PWR049

PWR049: Move iterator-dependent condition outside of the loop

Issue

A condition that depends only on the iterator variable can be moved outside of the loop.

Actions

Move iterator-dependent condition outside of the loop.

Relevance

A condition that depends only on the iterator is predictable: we know exactly at which iteration of the loop it is going to be true. Nevertheless, it is evaluated in each iteration of the loop.

Moving the iterator-dependent condition outside of the loop will result in fewer instructions executed in the loop. This transformation can occasionally enable vectorization, and for the loops that are already vectorized, it can increase vectorization efficiency.

Note

Moving an iterator-dependent condition outside of the loop is a creative process. Depending on the type of condition, it can involve loop peeling, loop fission or loop unrolling.

Code examples

C

Example 1

for (int i = 0; i < n; ++i) {
  if (i = 0) {
    a[i] = 0;
  } else {
    a[i] = 1;
  }
}

The condition on line 2 depends on the iterator i and can be removed by computing the first array element a[0] outside the loop. Thus, the loop iterator starts in 1 and the loop initializes the remaining array elements without computing any conditional statement:

a[0] = 0;

for (int i = 1; i < n; ++i) {
  a[i] = 1;
}

The iterator-dependent condition can appear in more complicated loops as well. For illustrative purposes, an example code with a loop nest is shown below:

for (int i = 0; i < n; ++i) {
  for (int j = 0; j < n; ++j) {
    if (j == 0) {
      a[i][j] = 0;
    } else {
      a[i][j] = a[i][j - 1] + b[i][j];
    }
  }
}

The condition on line 3 depends on the iterator j of the inner loop and can be removed as follows:

for (int i = 0; i < n; ++i) {
  a[i][0] = 0;

  for (int j = 1; j < n; ++j) {
    a[i][j] = a[i][j - 1] + b[i][j];
  }
}

In the example codes shown above, the resulting loops are branchless, avoiding redundant computations of predictable conditional instructions.

Example 2: Loop fission

for (int i = 0; i < n; ++i) {
  if (i < 10) {
    a[i] = 0;
  } else {
    a[i] = 1;
  }
}

The condition on line 2 depends on the iterator i and can be removed by splitting the loop over i into two loops:

for (int i = 0; i < 10; ++i) {
  a[i] = 0;
}

for (int i = 10; i < n; ++i) {
  a[i] = 1;
}

The first loop iterates from 0 to 9, and the second loop iterates from 10 until n - 1. The condition is removed from the loop.

Example 3: Loop unrolling

Here is another example of a iterator-dependent condition in the loop body:

for (int i = 0; i < n; ++i) {
  if (i % 2 == 0) {
    a[i] = 1;
  } else {
    a[i] = 0;
  }
}

The iterator-dependent condition is on line 2, and can be removed through loop unrolling:

for (int i = 0; i < n; i += 2) {
  a[i] = 1;
  a[i + 1] = 0;
}

Loop unrolling changes the increment of iterator variable i, so now it is 2 (see loop header at line 1). The condition is gone after this modification.

Fortran

Example 1

do i = 1, size(a, 1)
  if (i == 1) then
    a(i) = 0
  else
    a(i) = 1
  end if
end do

The condition on line 2 depends on the iterator i and can be removed by computing the first array element a(1) outside the loop. Thus, the loop iterator starts in 2 and the loop initializes the remaining array elements without computing any conditional statement:

a(1) = 0

do i = 2, size(a, 1)
  a(i) = 1
end do

The iterator-dependent condition can appear in more complicated loops as well. For illustrative purposes, an example code with a loop nest is shown below:

do j = 1, size(a, 2)
  do i = 1, size(a, 1)
    if (i == 1) then
      a(i, j) = 0
    else
      a(i, j) = a(i - 1, j) + b(i, j)
    end if
  end do
end do

The condition on line 3 depends on the iterator i of the inner loop and can be removed as follows:

do j = 1, size(a, 2)
  a(1, j) = 0

  do i = 2, size(a, 1)
    a(i, j) = a(i - 1, j) + b(i, j)
  end do
end do

In the example codes shown above, the resulting loops are branchless, avoiding redundant computations of predictable conditional instructions.

Example 2: Loop fission

do i = 1, size(a, 1)
  if (i < 10) then
    a(i) = 0
  else
    a(i) = 1
end do

The condition on line 2 depends on the iterator i and can be removed by splitting the loop over i into two loops:

do i = 1, 9
  a(i) = 0
end do

do i = 10, size(a, 1)
  a(i) = 1
end do

The first loop iterates from 1 to 9, and the second loop iterates from 10 until size(a, 1). The condition is removed from the loop.

Example 3: Loop unrolling

Here is another example of a iterator-dependent condition in the loop body:

do i = 1, size(a, 1)
  if (modulo(i, 2) == 0) then
    a(i) = 1
  else
    a(i) = 0
  end if
end do

The iterator-dependent condition is on line 2, and can be removed through loop unrolling:

do i = 1, size(a, 1), 2
    a(i) = 0
    a(i + 1) = 1
end do

Loop unrolling changes the increment of iterator variable i, so now it is 2 (see loop header at line 1). The condition is gone after this modification.

Related resources

• PWR049 examples

References

• Loop unswitching

• Loop fission