calculation parallelization

[Leonid]

Hello,

Please help me with calculation parallelization . I have 2
processors DELL PRECISION 530 computer and I'd like
parallelize cycles like
for(int i = 0; i < last; ++i)
{
a[i] = b[i] + c[i];
}

I've read about '#pragma omp parallel for' instruction for
the future version of VC++(Visual C++ "Whidbey": Advanced
Code Generation ). What can I do with VC++6 and VC++ .NET
2003?

Regards

Leonid

[Mihajlo Cvetanovic]

Re: calculation parallelization

Leonid wrote:[color=blue]
> Please help me with calculation parallelization . I have 2
> processors DELL PRECISION 530 computer and I'd like
> parallelize cycles like
> for(int i = 0; i < last; ++i)
> {
> a[i] = b[i] + c[i];
> }[/color]

Just make two worker threads which would work on one half of your
data. One (and only) parameter of a worker thread should be pointer to
a struct embodying your input and output parameters, like:

class ThreadParams
{
public:
ThreadParams(in t* o, int* i1, int* i2, size_t l) : output(o),
input1(i1), input2(i2), length(l) {}

int* output;
int* input1;
int* input2;
size_t length;
}

Worker thread function should look like this:

unsigned int __stdcall ThreadProc(void * pParam)
{
ThreadParams* pParams = reinterpret_cas t<ThreadParams* >(pParam);
size_t i;

for (i = 0; i < pParams->length; ++i)
pParams->output[i] = pParams->input1[i] + pParams->input2[i];

return 0;
}

You create non-MFC threads with _beginthreadex and MFC threads with
AfxBeginThread. ThreadParams class should be created on heap like this:

ThreadParams* pTp = new ThreadParams(a+ last/2, b+last/2, c+last/2,
last/2);

Also consider the situation where <last> is odd number.