[fpc-pascal] code optimization

[Adrian Veith]

On 23.09.2010 17:03, Jonas Maebe wrote:
>
> On 23 Sep 2010, at 16:59, Adrian Veith wrote:
>
>> I analyzed your code - I think the problem is the array element address
>> calculation of the fpc compiler. You have a lot of code like
>> Bar[MinValley] etc. The delphi compile uses the lea assembler code for
>> this, whereas fpc calculates the address of one element with imul  which
>> is much slower.
>
> Please see the last paragraph of
> http://lists.freepascal.org/lists/fpc-pascal/2010-September/026510.html
>
>> Anyway you could speed up your code significantly if you
>> help the compiler by reducing the address calculations with the help of
>> pointers like this:
>
> It may help a lot, but only because it will reduce register pressure,
> not because the multiplications are gone.

It reduces the total number of multiplications about 70% - I gave the
code to one of my guys and he changed the code using pointers to
elements wherever possible. This are the differences:

fpc - original code: 17s
fpc - pointer to elements: 12 s
delphi - original code: 9s

tested on a AMD notebook.

We haven't tested the new code with delphi yet, but the benefits should
be marginal compared to fpc.

As a conclusion one can say, that fpc's array arithmetic is suboptimal.

Cheers,

Adrian