It looks as though Visual Studio 2008 does the 'right' thing, too, at 
least in some circumstances.  Here's some assembler output (MSVC Express 
Edition, 32-bit Windows XP / Macbook Pro).

; 3    : unsigned long long mul(unsigned long x, unsigned long y) {

	push	ebp
	mov	ebp, esp

; 4    :     return (unsigned long long)x * y;

	mov	eax, DWORD PTR _x$[ebp]
	mov	ecx, DWORD PTR _y$[ebp]
	mul	ecx

; 5    : }

	pop	ebp
	ret	0