Re: Stealing a Great Idea from the 6600

MitchAlsup1 <mitchalsup@aol.com> schrieb:

One of the things we found in Mc 88120 was that the compiler should
NEVER
be allowed to put unnecessary instructions in decode-execute slots that
were unused--and that almost invariable--the best code for the GBOoO
machine was almost invariably the one with the fewest instructions, and
if several sequences had equally few instructions, it basically did not
matter.
>
For example::
>
     for( i = 0; i < max, i++ )
          a[i] = b[i];
>
was invariably faster than::
>
     for( ap = &a[0], bp = & b[0];, i = 0; i < max; i++ )
          *ap++ = *bp++;
>
because the later has 3 ADDs in the loop wile the former has but 1.
Because of this, I altered my programming style and almost never end up
using ++ or -- anymore.

Interesting.

I looked at a variant of this,

void foo (int *a, int *b, int n)
{
  int i;
  for( i = 0; i < n; i++ )
    a[i] = b[i] + 1;
}

void bar (int *a, int *b, int n)
{

  int *ap, *bp;
  int i;

  for( ap = &a[0], bp = & b[0], i = 0; i < n; i++ )
    *ap++ = (*bp++) + 1;
}


I would expect ivopt to optimize these to the same assembly.
Gcc does so, at least for AMD64 and POWER, and so does clang for
the usual architectures if translated with -O2 -fno-unroll-loops
-fno-tree-vectorize.

But I think there are other reasons to chose the array version,
I think - clarity is one of them, the other being uses
of "restrict" on arguments.