How to tune numerical D? (matrix multiplication is faster in g++ vs gdc)

[Manu]

On 4 March 2013 14:50, J <private at private-dont-email-dont-spam.com> wrote:

> On Monday, 4 March 2013 at 04:22:01 UTC, bearophile wrote:
>
>> So this should be better:
>>
>> http://codepad.org/B5b4uyBM
>>
>> Bye,
>> bearophile
>>
>
> @bearophile: Thank you!  Unfortunately the http://codepad.org/B5b4uyBMcode runs a bit *slower* than the original D code. Yikes!
>
> $  gdmd -O -inline -release -noboundscheck -m64 bear.d -ofdbear
> $ time ./dbear
> -1015380632 859379360 -367726792 -1548829944
>
> real    2m36.971s
> user    2m36.910s
> sys 0m0.030s
> $ time ./dbear
> -1015380632 859379360 -367726792 -1548829944
>
> real    2m34.425s
> user    2m34.370s
> sys 0m0.020s
>
>
> @John Colvin: here is the disassembly of mmult() in both languages.
> Unfortunately I'm not literate in x86_64 assembly.  Perhaps the problem is
> obvious to you?  All I can really tell is that the g++ version is shorter.
>
> The memory allocation, when timed separately (comment out mmult), is less
> than 60 msec for either version, so I don't think its a memory issue,
> although it could be caching issue since the matrix layouts are different.
>

Using dynamic arrays of dynamic arrays that way is pretty poor form
regardless of the language.

You should really use single dimensional array:
  int matrix[SIZE*SIZE];
And index via:
  matrix[y*SIZE+x]

(You can pretty this up in various ways, if this is too unsightly)


On 4 March 2013 14:02, John Colvin <john.loughran.colvin at gmail.com> wrote:

>     int[][] m = new int[][](rows, cols);
>

Does D support proper square array's this way? Or does it just automate
allocation of the inner arrays?
Does it allocate all the associated memory in one block?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.puremagic.com/pipermail/digitalmars-d/attachments/20130304/a36421ed/attachment.html>