void
benchmark(int iterations)
{
// allocate memory for result
unsigned int *d_result;
unsigned int size = width * height * sizeof(unsigned int);
checkCudaErrors(cudaMalloc((void **) &d_result, size));
// warm-up
gaussianFilterRGBA(d_img, d_result, d_temp, width, height, sigma, order, nthreads);
checkCudaErrors(cudaDeviceSynchronize());
sdkStartTimer(&timer);
// execute the kernel
for (int i = 0; i < iterations; i++)
{
gaussianFilterRGBA(d_img, d_result, d_temp, width, height, sigma, order, nthreads);
}
checkCudaErrors(cudaDeviceSynchronize());
sdkStopTimer(&timer);
// check if kernel execution generated an error
getLastCudaError("Kernel execution failed");
printf("Processing time: %f (ms)\n", sdkGetTimerValue(&timer));
printf("%.2f Mpixels/sec\n", (width*height*iterations / (sdkGetTimerValue(&timer) / 1000.0f)) / 1e6);
checkCudaErrors(cudaFree(d_result));
}
bool
runSingleTest(const char *ref_file, const char *exec_path)
{
// allocate memory for result
int nTotalErrors = 0;
unsigned int *d_result;
unsigned int size = width * height * sizeof(unsigned int);
checkCudaErrors(cudaMalloc((void **) &d_result, size));
// warm-up
gaussianFilterRGBA(d_img, d_result, d_temp, width, height, sigma, order, nthreads);
checkCudaErrors(cudaDeviceSynchronize());
sdkStartTimer(&timer);
gaussianFilterRGBA(d_img, d_result, d_temp, width, height, sigma, order, nthreads);
checkCudaErrors(cudaDeviceSynchronize());
getLastCudaError("Kernel execution failed");
sdkStopTimer(&timer);
unsigned char *h_result = (unsigned char *)malloc(width*height*4);
checkCudaErrors(cudaMemcpy(h_result, d_result, width*height*4, cudaMemcpyDeviceToHost));
char dump_file[1024];
sprintf(dump_file, "lena_%02d.ppm", (int)sigma);
sdkSavePPM4ub(dump_file, h_result, width, height);
if (!sdkComparePPM(dump_file, sdkFindFilePath(ref_file, exec_path), MAX_EPSILON_ERROR, THRESHOLD, false))
{
nTotalErrors++;
}
printf("Processing time: %f (ms)\n", sdkGetTimerValue(&timer));
printf("%.2f Mpixels/sec\n", (width*height / (sdkGetTimerValue(&timer) / 1000.0f)) / 1e6);
checkCudaErrors(cudaFree(d_result));
free(h_result);
printf("Summary: %d errors!\n", nTotalErrors);
printf(nTotalErrors == 0 ? "Test passed\n": "Test failed!\n");
return (nTotalErrors == 0);
}
void
runAutoTest(int argc, char **argv)
{
// allocate memory for result
unsigned int *d_result;
unsigned int size = width * height * sizeof(unsigned int);
cutilSafeCall( cudaMalloc( (void**) &d_result, size));
// warm-up
gaussianFilterRGBA(d_img, d_result, d_temp, width, height, sigma, order, nthreads);
cutilSafeCall( cudaThreadSynchronize() );
cutilCheckError( cutStartTimer( timer));
while (sigma <= 22) {
gaussianFilterRGBA(d_img, d_result, d_temp, width, height, sigma, order, nthreads);
cutilSafeCall( cudaThreadSynchronize() );
// check if kernel execution generated an error
cutilCheckMsg("Kernel execution failed");
cudaMemcpy(g_CheckRender->imageData(), d_result, width*height*4, cudaMemcpyDeviceToHost);
g_CheckRender->savePPM(sOriginal[g_Index], false, NULL);
if (!g_CheckRender->PPMvsPPM(sOriginal[g_Index], sReference[g_Index], MAX_EPSILON_ERROR, 0.50f)) {
g_TotalErrors++;
}
g_Index++;
sigma += 4;
}
cutilCheckError( cutStopTimer( timer));
printf("Processing time: %f (ms)\n", cutGetTimerValue( timer));
printf("%.2f Mpixels/sec\n", (width*height*g_Index / (cutGetTimerValue( timer) / 1000.0f)) / 1e6);
printf("Summary: %d errors!\n", g_TotalErrors);
printf("Test %s!\n", (g_TotalErrors==0) ? "PASSED" : "FAILED");
cutilSafeCall(cudaFree(d_result));
}
// display results using OpenGL
void display()
{
cutilCheckError(cutStartTimer(timer));
// execute filter, writing results to pbo
unsigned int *d_result;
cutilSafeCall(cudaGLMapBufferObject((void**)&d_result, pbo));
gaussianFilterRGBA(d_img, d_result, d_temp, width, height, sigma, order, nthreads);
cutilSafeCall(cudaGLUnmapBufferObject(pbo));
// load texture from pbo
glBindBufferARB(GL_PIXEL_UNPACK_BUFFER_ARB, pbo);
glBindTexture(GL_TEXTURE_2D, texid);
glPixelStorei(GL_UNPACK_ALIGNMENT, 1);
glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, width, height, GL_RGBA, GL_UNSIGNED_BYTE, 0);
glBindBufferARB(GL_PIXEL_UNPACK_BUFFER_ARB, 0);
// display results
glClear(GL_COLOR_BUFFER_BIT);
glEnable(GL_TEXTURE_2D);
glDisable(GL_DEPTH_TEST);
glBegin(GL_QUADS);
glTexCoord2f(0, 1); glVertex2f(0, 0);
glTexCoord2f(1, 1); glVertex2f(1, 0);
glTexCoord2f(1, 0); glVertex2f(1, 1);
glTexCoord2f(0, 0); glVertex2f(0, 1);
glEnd();
glDisable(GL_TEXTURE_2D);
if (g_CheckRender && g_CheckRender->IsQAReadback() && g_Verify) {
// readback for QA testing
printf("> (Frame %d) Readback BackBuffer\n", frameCount);
g_CheckRender->readback( width, height );
g_CheckRender->savePPM(sOriginal[g_Index], true, NULL);
if (!g_CheckRender->PPMvsPPM(sOriginal[g_Index], sReference[g_Index], MAX_EPSILON_ERROR, 0.50f )) {
g_TotalErrors++;
}
g_Verify = false;
}
glutSwapBuffers();
cutilCheckError(cutStopTimer(timer));
computeFPS();
}
// display results using OpenGL
void display()
{
sdkStartTimer(&timer);
// execute filter, writing results to pbo
unsigned int *d_result;
checkCudaErrors(cudaGLMapBufferObject((void **)&d_result, pbo));
gaussianFilterRGBA(d_img, d_result, d_temp, width, height, sigma, order, nthreads);
checkCudaErrors(cudaGLUnmapBufferObject(pbo));
// load texture from pbo
glBindBufferARB(GL_PIXEL_UNPACK_BUFFER_ARB, pbo);
glBindTexture(GL_TEXTURE_2D, texid);
glPixelStorei(GL_UNPACK_ALIGNMENT, 1);
glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, width, height, GL_RGBA, GL_UNSIGNED_BYTE, 0);
glBindBufferARB(GL_PIXEL_UNPACK_BUFFER_ARB, 0);
// display results
glClear(GL_COLOR_BUFFER_BIT);
glEnable(GL_TEXTURE_2D);
glDisable(GL_DEPTH_TEST);
glBegin(GL_QUADS);
glTexCoord2f(0, 1);
glVertex2f(0, 0);
glTexCoord2f(1, 1);
glVertex2f(1, 0);
glTexCoord2f(1, 0);
glVertex2f(1, 1);
glTexCoord2f(0, 0);
glVertex2f(0, 1);
glEnd();
glDisable(GL_TEXTURE_2D);
glutSwapBuffers();
sdkStopTimer(&timer);
computeFPS();
}