[vox-tech] benchmarking (again)

[Kenneth Herron]

Peter Jay Salzman wrote:
> I was under the impression that this is what computer scientists call
> "locality of data" -- if data is used once, it'll be used again soon.  The
> strategy that attempts to exploit locality of data is to hold recent data in
> L1 cache.

The cache isn't under the programmer's direct control. It's a function 
of the hardware. The best the programmer can usually do is to try to 
access memory in ways that lets the cache work well.  For example:

	int array1[10000], array2[10000], array3[10000];

	/* Version 1 */
	for (i = 0; i < 10000; i++) {
		sum1 += array1[i];
		sum2 += array2[i];
		sum3 += array3[i];
	}

	/* Version 2 */
	for (i = 0; i < 10000; i++)
		sum1 += array1[i];
	for (i = 0; i < 10000; i++)
		sum2 += array2[i];
	for (i = 0; i < 10000; i++)
		sum3 += array3[i];

You might expect version 1 to run faster because it has only one set of 
loop increment operations. But version 1 jumps among three widely 
separated memory blocks on each iteration, while version 2 accesses 
memory in a simple linear fashion. The latter is more cache-friendly. It 
might even run faster.


> I see what you mean.  Yeah, but the question still remains -- what's going
> on with the user time in java apps, and why does the program run faster in
> Java than C++?  Certainly C++ compilers know how to do all that stuff too.

The code you're benchmarking doesn't have any side effects. Maybe the 
java compiler was smart enough to optimize the whole thing down to an 
empty function :-)