🚀 go-pugleaf

In article <14251@lambda.UUCP> jlg@lambda.UUCP (Jim Giles) writes:
 >            In fact, for programs of any really useful size, passing
 > array arguments is vital.  In this respect, at least, C really _DOESN'T_
 > have arrays.
 > 
But in this aspect Fortran (at least upto and including 77) does not have
arrays either.  Let's compare some languages (for a matrix-vector product
routine*):

Algol 68:
	.proc matvec = (.ref [,] .real m, .ref [] .real v) .ref [] .real:
	(	[1 .lwb m:1 .upb m] .real result;
		.co assert correct bounds .co
		.if 2 .lwb m < .lwb v .or 2 .upb m > .upb v
		.then
			.skip .co error detected! .co
		.fi
		.for i .from 1 .lwb m .to 1 .upb m
		.do
			.ref .real sum = result[i];
			sum:= 0.0;
			.for j .from 2 .lwb m .to 2 .upb m
			.do
				sum += m[i,j] * v[j]
			.od
		.od;
		result
	)

Ada:
	with LA_PACK;  -- defines types MATRIX, VECTOR etc.
	use LA_PACK;
	function MATVEC(M: MATRIX; V: VECTOR) return VECTOR is
		RESULT: VECTOR(M'FIRST(1) .. M'LAST(1));
		SUM: FLOAT;
	begin -- assert correct bounds
		if M'FIRST(2) < V'FIRST or else M'LAST(2) > V'LAST then
			raise ARRAY_BOUNDS_ERROR;
		end if;
		for I in M'RANGE(1) loop
			SUM = 0.0;
			for J in M'RANGE(2) loop
				SUM = M(I,J) * V(J) + SUM;
			end loop;
			RESULT(I) = SUM;
		end loop;
		return RESULT;
	end MATVEC;

Fortran:
	      SUBROUTINE MATVEC(M, V, W, L1M, U1M, L2M, U2M)
	C     NO FUNCTION, FORTRAN CANNOT RETURN ARRAYS
	      INTEGER L1M, U1M, L2M, U2M, I, J
	      REAL M(L1M:U1M,L2M:U2M), V(L2M:U2M), W(L1M:U1M), SUM
	C     NO BOUND CHECKING POSSIBLE
	      DO 20 I = L1M, U1M
	        SUM = 0.0
	        DO 10 J = L2M, U2M
	          SUM = M(I,J) * V(J) + SUM
	10      CONTINUE
	        W(I) = SUM
	20    CONTINUE
	      RETURN
	      END

C (Fortran style (you can write Fortran in any language!)):
	void matvec(m, v, w, l1m, u1m, l2m, u2m)
	/* No function, C cannot return arrays.  It can return pointers,
	   but then we need malloc, and how to free the stuff?
	   (Garbage collection of course!) */
		int l1m, u1m, l2m, u2m;
		float *m, *v, *w;
	#define M(i,j) m[(u2m - l2m + 1) * i + j]
	#define V(i) v[i]
	#define W(i) w[i]
	{
	int i, j;
	float sum;
		/* No bound checking possible. */
		for(i = l1m; i <= u1m; i++) {
			sum = 0.0;
			for(j = l2m; j <= u2m; j++) {
				sum += M(i,j) * V(j);
			}
			W(i) = sum;
		}
	}

You can write it in C in C style of course; it looks a bit different.
Anyhow, I think I made my point.  C and Fortran are not very different
with respect to arrays.  Except of course that Fortran disallows aliasing,
while C does not (hence the Fortran compiler can vectorize the above text,
while a C compiler can not).  Clearly Algol 68 and Ada, that supports arrays
in their fulll glory, are superior with the possibility of array
bound checking etc..  Both in Fortran and C a routine receives an address
and defines its own indexing function, so both in Fortran and C it is
possible to give the array a different shape in the callee when compared
to the caller.  This is impossible in languages that have complete
support for arrays.  (And note that one of the most important packages
in Fortran, the BLAS, makes much use of it!)  But I do not think that is
a feature, it looks more like a bug.

You might argue that the C define for M(i,j) is complicated.  I would agree,
Fortran has a simple syntactic description for (essentially) the same
definition.  If you continue to say that in Fortran it is more clear to
the compiler that rows of a matrix are accessed, I am going to scream.
Everybody will agree that a good optimizing C compiler will compile the
C source to essentially the same code as a good optimizing Fortran compiler
does with the Fortran source (everybody, except Piercarlo Grandi of course).
--
(*)  Yes, I know, this is not the best (vectorizing) code for a matrix-
vector product.
-- 
dik t. winter, cwi, amsterdam, nederland
dik@cwi.nl