Skip to content

Commit

Permalink
Fix max_iterations limit
Browse files Browse the repository at this point in the history
  • Loading branch information
@teunvw14
teunvw14 committed Jan 25, 2021
1 parent 6404ef3 commit 6ac9005
Showing 1 changed file with 14 additions and 14 deletions.
28 changes: 14 additions & 14 deletions GenericMandelBrotViewer/src/kernel.cu
View file
Original file line number Diff line number Diff line change
Expand Up @@ -17,7 +17,7 @@ int resolution_x = 1024;
int resolution_y = 1024;
double draw_radius = 2.5;
double escape_radius_squared = 4; // escape_radius = 2^7 = 256
int max_iterations = 64;
int max_iterations = 64000;

bool incremental_iteration = false;
int iterations_per_frame; // value set in main()
Expand All @@ -34,7 +34,7 @@ thrust::complex<double>* points;
thrust::complex<double>* iterated_points;
double* squared_absolute_values;
unsigned char* pixels_rgb;
unsigned short* iterationsArr; // TODO: Change to unsigned int. The program completely breaks if max_iterations >= 2^16 (= 65536) because of the unsigned short data type used here.
unsigned int* iterationsArr;


__global__ void build_complex_grid_cuda(
Expand Down Expand Up @@ -102,7 +102,7 @@ __global__ void mandelbrot_iterate_cuda(
thrust::complex<double>* points,
thrust::complex<double>* iterated_points,
double* squared_absolute_values,
unsigned short* iterationsArr
unsigned int* iterationsArr
)
{
int block_index_x = blockIdx.x;
Expand All @@ -126,7 +126,7 @@ __global__ void mandelbrot_iterate_cuda(
thrust::complex<double> c = points[index];
thrust::complex<double> it_point = iterated_points[index];
double sq_abs = squared_absolute_values[index];
unsigned short iterations_ = iterationsArr[index];
unsigned int iterations_ = iterationsArr[index];
while (iterations_ < max_iterations && sq_abs < escape_radius_squared) {
it_point = thrust::complex<double>(it_point.real() * it_point.real() - it_point.imag() * it_point.imag(), 2 * it_point.real() * it_point.imag()) + c; // z^2 + c
sq_abs = it_point.real() * it_point.real() + it_point.imag() * it_point.imag();
Expand All @@ -146,7 +146,7 @@ void mandelbrot_iterate_non_cuda(
thrust::complex<double>* points,
thrust::complex<double>* iterated_points,
double* squared_absolute_values,
unsigned short* iterationsArr
unsigned int* iterationsArr
)
{
int index;
Expand All @@ -161,7 +161,7 @@ void mandelbrot_iterate_non_cuda(
thrust::complex<double> c = points[index];
thrust::complex<double> it_point = iterated_points[index];
double sq_abs = squared_absolute_values[index];
unsigned short iterations_ = iterationsArr[index];
unsigned int iterations_ = iterationsArr[index];
while (iterations_ < max_iterations && sq_abs < escape_radius_squared) {
it_point = thrust::complex<double>(it_point.real() * it_point.real() - it_point.imag() * it_point.imag(), 2 * it_point.real() * it_point.imag()) + c; // z^2 + c
sq_abs = it_point.real() * it_point.real() + it_point.imag() * it_point.imag();
Expand All @@ -177,7 +177,7 @@ void mandelbrot_iterate_non_cuda(

__global__ void color_cuda(
int max_iterations,
unsigned short* iterationsArr,
unsigned int* iterationsArr,
double * squared_absolute_values,
int resolution_x,
int resolution_y,
Expand All @@ -197,7 +197,7 @@ __global__ void color_cuda(
//int thread_index_y = threadIdx.y;
//int thread_stride_y = blockDim.y;
int index;
int iterations;
unsigned int iterations;

//printf("thread_index_x: %i | block_index_x: %i | thread_stride_x: %i | block_stride_x: %i\n", thread_index_x, block_index_x, thread_stride_x, block_stride_x);
for (int pixel_y = block_index_x; pixel_y < resolution_y; pixel_y += block_stride_x)
Expand Down Expand Up @@ -302,7 +302,7 @@ __global__ void color_cuda(

void color_non_cuda(
int max_iterations,
unsigned short* iterationsArr,
unsigned int* iterationsArr,
double* squared_absolute_values,
int resolution_x,
int resolution_y,
Expand All @@ -312,7 +312,7 @@ void color_non_cuda(
{
// Do some coloring!
int index;
int iterations;
unsigned int iterations;

//printf("thread_index_x: %i | block_index_x: %i | thread_stride_x: %i | block_stride_x: %i\n", thread_index_x, block_index_x, thread_stride_x, block_stride_x);
for (int pixel_y = 1; pixel_y < resolution_y; pixel_y++)
Expand Down Expand Up @@ -499,15 +499,15 @@ void reset_render_objects()
cudaFree(squared_absolute_values);
cudaFree(iterationsArr);
cudaMallocManaged(&squared_absolute_values, resolution_x * resolution_y * sizeof(double));
cudaMallocManaged(&iterationsArr, resolution_x * resolution_y * sizeof(unsigned short));
cudaMallocManaged(&iterationsArr, resolution_x * resolution_y * sizeof(unsigned int));
// Synchronize the GPU so the whole thing doesn't crash.
cudaDeviceSynchronize();
}
else if (!(cuda_device_available)) {
free(squared_absolute_values);
free(iterationsArr);
squared_absolute_values = (double*)malloc(resolution_x * resolution_y * sizeof(double));
iterationsArr = (unsigned short*)malloc(resolution_x * resolution_y * sizeof(unsigned short));
iterationsArr = (unsigned int*)malloc(resolution_x * resolution_y * sizeof(unsigned int));
}

// Reset the amount of rendered iterations to 0.
Expand Down Expand Up @@ -629,14 +629,14 @@ void allocate_memory() {
cudaMallocManaged(&iterated_points, resolution_x * resolution_y * sizeof(thrust::complex<double>));
cudaMallocManaged(&squared_absolute_values, resolution_x * resolution_y * sizeof(double));
cudaMallocManaged(&pixels_rgb, resolution_x * resolution_y * 3 * sizeof(unsigned char));
cudaMallocManaged(&iterationsArr, resolution_x * resolution_y * sizeof(unsigned short));
cudaMallocManaged(&iterationsArr, resolution_x * resolution_y * sizeof(unsigned int));
}
else if (!(cuda_device_available)) {
points = (thrust::complex<double>*)malloc(resolution_x * resolution_y * sizeof(thrust::complex<double>));
iterated_points = (thrust::complex<double>*)malloc(resolution_x * resolution_y * sizeof(thrust::complex<double>));
squared_absolute_values = (double*)malloc(resolution_x * resolution_y * sizeof(double));
pixels_rgb = (unsigned char*)malloc(resolution_x * resolution_y * 3 * sizeof(unsigned char));
iterationsArr = (unsigned short*)malloc(resolution_x * resolution_y * sizeof(unsigned short));
iterationsArr = (unsigned int*)malloc(resolution_x * resolution_y * sizeof(unsigned int));
}
}

Expand Down

0 comments on commit 6ac9005

Please sign in to comment.