Boost logo

Boost :

Subject: Re: [boost] sorting floats by casting to integer
From: Steven Ross (spreadsort_at_[hidden])
Date: 2009-07-06 16:48:15


On Mon, Jul 6, 2009 at 11:07 AM, Vladimir Prus <vladimir_at_[hidden]>wrote:

> > Yes, I guess I was using denormalized numbers for the 23-bit case, but
> it's
> > interesting that there is such a huge speed difference, and I doubt it's
> > because they're denormalized; there is no reason to special-case
> > denormalized sorting.
>
> It is well possible that comparing denormalized numbers is slow.

 Taking 10M elements with my 23 bits of randomness case, and making the top
exponent bit 1, so that the data is normalized:
     if(!(float_mem_cast<float, int>(array[v]) & 0x7f800000)) {
       //Make the top exponent bit high
       CAST_TYPE temp = 0x80000000 | float_mem_cast<float, int>(array[v]);
       memcpy(&(array[v]), &temp, sizeof(DATATYPE));
       //printf("new %e\n", array[v]);
     }

I see std::sort take 83.527s, as compared to 1.226s when sorted as
integers. denormalized numbers are not the problem.

> Notably, normalized numbers also show the same problem if their exponent
> > range is small relative to the number of elements; this can be achieved
> by
> > using a small number of different exponents, or simply a large number of
> > floats (as shown with my large test). What do you consider a reasonable
> > range? I can easily make a test that excludes denormalized numbers.
> >
> > Even if the full exponent range is in use (which isn't realistic),
> wouldn't
> > it be better to use an algorithm that doesn't blow up when n gets much
> > larger than 10 million elements, and is still faster with a smaller
> number
> > of elements?
>
> I don't quite follow the above point. My argument is (and always was), that
> it's important to provide accurate measurements on the data that ordinary
> humans will use. Sorting of 200'000'000 floats does not seem a typical use
> case, for performance on that is probably less important. Just like sorting
> demormalized numbers.
>

Normal problems don't use the full space of possible exponents for
floating-point numbers.
Can you think of a benchmark that accurately reflects this reality?


Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk