AMD Developer Forums - Framewave and APL http://forums.amd.com/forum/index.cfm?forumid=9 FuseTalk Hosting Executive Plan 90 degree rotation http://forums.amd.com/forum/messageview.cfm?catid=204&threadid=102342 2008-10-31T16:41:17 -05.00 rafajna 3

Hi,

I am trying to rotate an image 90 degrees using the Rotate function.  Here is the code

if( Rotate == 90 )

{

::FwiSize roiSize = {width, height};

::FwiRect srcR = {0, 0, width, height };

::FwiRect dstR = {0, 0, height, width };

::fwiRotate_8u_C3R( (Fw8u *)pSrc, roiSize, width * 3, srcR

, (Fw8u *)pDst, height * 3, dstR, 90, 0, width, FWI_INTER_CUBIC);

int temp = width;

width = height;

height = temp;

}// end if

However, the resulting image (pDst)  is scrambled.  Am I using the fwiRotate correctly or is there a problem with that function?

Thank you.

]]> APL is now Framewave http://forums.amd.com/forum/messageview.cfm?catid=204&threadid=92516 2008-02-20T16:05:28 -05.00 devcentral 1
The Framewave Project]]> APL JPEG vs. IPP JPEG http://forums.amd.com/forum/messageview.cfm?catid=204&threadid=92061 2008-02-06T15:10:55 -05.00 mrubelmann 3
Thanks,
Mark]]> New APL How-To Articles http://forums.amd.com/forum/messageview.cfm?catid=204&threadid=90499 2007-12-18T16:50:46 -05.00 devcentral 1
»AMD Performance Library on Linux®: Making the Penguin Fly

»Using AMD's Performance Library on Windows®: How To Get Smashing Performance]]> Wavelet Support? http://forums.amd.com/forum/messageview.cfm?catid=204&threadid=89788 2007-11-25T13:42:12 -05.00 daryll 2 I see some nice new features in the APL 1.1. Here's a suggestion for 1.2. It would be great to see some wavelet support for JPEG2000 and the new Microsoft/JPEG photo standards. Any chance that can get in for 1.2?]]> APL 1.1 is here! http://forums.amd.com/forum/messageview.cfm?catid=204&threadid=88125 2007-09-19T16:40:46 -05.00 devcentral 3
http://developer.amd.com/apl.jsp]]> APL function is much slower than C version http://forums.amd.com/forum/messageview.cfm?catid=204&threadid=87817 2007-09-10T20:15:40 -05.00 lha@sci.utah.edu 2 my C++ version , how ever the result really makes me disappointed the apl version
is much slower than C version (about 10 time slower or more)

This is the program


#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <aplBase.h>
#include <aplImage.h>
#include <math.h>

Apl64u start, end;

template <typename T> int test1D(T* src, T* dest, int count){
for (int i=0; i< count; ++i)
if (src!=dest)
return 0;
return 1;
};

template<typename T> int test2D(T* src, int sStride, T* dst, int dStride,
int w, int h){
T* pSrc = src,*pDst = dst;
for (int j=0; j< h; ++j){
pSrc =(T*) ((char*) src + j * sStride);
pDst =(T*) ((char*) dst + j * dStride);
for (int i=0; i< w; ++i, ++pSrc, ++pDst)
if (*pSrc != *pDst)
return 0;
}

return 1;
}

template < class T>
void addImages(T* ia, T* ib, T* result, int width, int height, int stride){
T* pA = ia,*pB = ib, *pR = result;
for (int j=0; j< height; ++j){
pA =(T*) ((char*) ia + j * stride);
pB =(T*) ((char*) ib + j * stride);
pR =(T*) ((char*) result + j * stride);

for (int i=0; i< width; ++i, ++pA, ++pB, ++pR)
(*pR) = (*pA) + (*pB);
}
}


void absImage(Apl32f* ia, Apl32f* result, int width, int height, int stride){
Apl32f* pA = ia,*pR = result;
for (int j=0; j< height; ++j){
pA =(Apl32f*) ((char*) ia + j * stride);
pR =(Apl32f*) ((char*) result + j * stride);
for (int i=0; i< width; ++i, ++pA, ++pR)
(*pR) = fabs((*pA));
}
}


void generateRandomInput(Apl32f* img, int width, int height, int stride){

Apl32f* pImg;

for (int j=0; j< height; ++j){
pImg = (Apl32f*) ((char*) img + j * stride);
for (int i=0; i< width; ++i, ++pImg)
*pImg = (float) rand() /RAND_MAX;
}
}
int main(int argc, char *argv[])
{

int width = 256;
int height = 256;
int stride = 100;
float cTime = 0.f, aplTime = 0.f;


Apl32f* srcImg1 = apliMalloc_32f_C1(width, height,&stride);
Apl32f* srcImg2 = apliMalloc_32f_C1(width, height,&stride);
Apl32f* dstImg = apliMalloc_32f_C1(width, height,&stride);
Apl32f* tmpImg = apliMalloc_32f_C1(width, height,&stride);
ApliSize size;

size.width = width;
size.height = height;

generateRandomInput(srcImg1, width , height, stride);
generateRandomInput(srcImg2, width , height, stride);

int numIter = 1000;

fprintf(stderr, "Perform APL function \n");

start = aplGetCpuClocks();
for (int i=0; i< 1000; i++)
apliAbs_32f_C1R(srcImg1, stride, tmpImg, stride, size);
end = aplGetCpuClocks();
aplTime = ((double)end - (double) start) / 1000.f;

fprintf(stderr, "Perform C function \n");

start = aplGetCpuClocks();
for (int i=0; i< numIter; i++)
absImage(srcImg1, dstImg, width , height , stride);
end = aplGetCpuClocks();
cTime = ((double)end - (double) start) / 1000.f;


if (!test2D(tmpImg, stride, dstImg, stride, width , height))
fprintf(stderr, "Wrong result ');


fprintf(stderr, "C time %g Apl time %g \n", cTime, aplTime);

addImages(srcImg1, srcImg2, dstImg, width, height, stride);

apliFree(srcImg1);
apliFree(srcImg2);
apliFree(dstImg);
apliFree(tmpImg);
return 0;
}


My compile options is
g++ -c -Wall -O3 -D_REENTRANT -DQT_THREAD_SUPPORT -DNO_DEBUG -march=opteron -m64. I link with the static libray.

My machine is Opteron 64 , 8GB memory. I use APL 1.0 linux 64 in Ubunto.

Can any one try it and say what happen. It seems that even Intel IPP still faster on
Opteron than AMP APL.

Any idea is appreciated.]]> What version of the MSVC C run time library should I use when linking to the static version of the APL libraries? http://forums.amd.com/forum/messageview.cfm?catid=204&threadid=86268 2007-06-26T17:31:41 -05.00 pkamd123 2 The static version of the APL library for windows has been compiled with support for the static version of MSVC CRT library. In order to avoid various linker errors and warnings, you should use the same version of the CRT library in your code that statically links to APL.]]> How do I install APL? http://forums.amd.com/forum/messageview.cfm?catid=204&threadid=86266 2007-06-26T16:39:06 -05.00 prashah 3 1) Unzip the downloaded file into the desired folder.
2) Include the APL library file in the compiler's path.]]>