public Grid2D add(OpenCLGrid2D image1, OpenCLGrid2D image2) {
// create context
if (context == null) {
context = OpenCLUtil.getStaticContext();
}
// select device
if (device == null) {
device = context.getMaxFlopsDevice();
}
// define local and global sizes
int width = Math.min(image1.getWidth(), image2.getWidth());
int height = Math.min(image1.getHeight(), image2.getHeight());
int imageSize = width * height;
int localWorkSize = Math.min(device.getMaxWorkGroupSize(), 8);
int globalWorkSizeW =
OpenCLUtil.roundUp(
localWorkSize, width); // rounded up to the nearest multiple of localWorkSize
int globalWorkSizeH = OpenCLUtil.roundUp(localWorkSize, height);
// load sources, create and build programm
if (program == null) {
try {
program =
context.createProgram(this.getClass().getResourceAsStream("exercise4.cl")).build();
} catch (IOException e) {
// TODO Auto-generated catch block
e.printStackTrace();
System.exit(-1);
}
}
// create output image
CLBuffer<FloatBuffer> output = context.createFloatBuffer(imageSize, Mem.WRITE_ONLY);
if (kernel == null) {
kernel = program.createCLKernel("addImages");
}
// createCommandQueue
CLCommandQueue queue = device.createCommandQueue();
image1.getDelegate().prepareForDeviceOperation();
image2.getDelegate().prepareForDeviceOperation();
// put memory on the graphics card
kernel
.putArg(image1.getDelegate().getCLBuffer())
.putArg(image2.getDelegate().getCLBuffer())
.putArg(output)
.putArg(width)
.putArg(height);
kernel.rewind();
queue
.put2DRangeKernel(
kernel, 0, 0, globalWorkSizeW, globalWorkSizeH, localWorkSize, localWorkSize)
.putBarrier()
// put memory from graphic card to host
.putReadBuffer(output, true)
.finish();
Grid2D result = new Grid2D(image1);
output.getBuffer().rewind();
for (int i = 0; i < result.getSize()[1]; ++i) {
for (int j = 0; j < result.getSize()[0]; j++) {
result.setAtIndex(j, i, output.getBuffer().get());
}
}
output.release();
queue.release();
return result;
}
public Grid2D openCLBackprojection(
OpenCLGrid2D filteredSinogramm,
int widthPhantom,
int heightPhantom,
int worksize,
float detectorSpacing,
int numberOfPixel,
int numberProjections,
float scanAngle,
double[] spacing,
double[] origin) {
// create context
CLContext context = OpenCLUtil.getStaticContext();
// select device
CLDevice device = context.getMaxFlopsDevice();
// define local and global sizes
double spacingAngle = (double) (scanAngle / numberProjections);
double originDetector = -(detectorSpacing * numberOfPixel) / 2.0;
int imageSize = widthPhantom * heightPhantom;
int localWorkSize = Math.min(device.getMaxWorkGroupSize(), worksize);
int globalWorkSizeW =
OpenCLUtil.roundUp(
localWorkSize, widthPhantom); // rounded up to the nearest multiple of localWorkSize
int globalWorkSizeH = OpenCLUtil.roundUp(localWorkSize, heightPhantom);
// load sources, create and build programm
try {
this.program =
context.createProgram(this.getClass().getResourceAsStream("exercise4.cl")).build();
} catch (IOException e) {
// TODO Auto-generated catch block
e.printStackTrace();
System.exit(-1);
}
// create image from input grid
// CLImageFormat format = new CLImageFormat(ChannelOrder.INTENSITY, ChannelType.FLOAT);
// create output image
CLBuffer<FloatBuffer> output = context.createFloatBuffer(imageSize, Mem.WRITE_ONLY);
if (kernel == null) {
kernel = program.createCLKernel("parallelBackProjection");
}
// createCommandQueue
CLCommandQueue queue = device.createCommandQueue();
filteredSinogramm.getDelegate().prepareForDeviceOperation();
// put memory on the graphics card
kernel
.putArg(filteredSinogramm.getDelegate().getCLBuffer())
.putArg(output)
.putArg(numberProjections)
.putArg(numberOfPixel)
.putArg(scanAngle)
.putArg(widthPhantom)
.putArg(heightPhantom)
.putArg(spacing[0])
.putArg(spacing[1])
.putArg(origin[0])
.putArg(origin[1])
.putArg(detectorSpacing)
.putArg(spacingAngle)
.putArg(originDetector)
.putArg(0.d);
kernel.rewind();
queue
.put2DRangeKernel(
kernel, 0, 0, globalWorkSizeW, globalWorkSizeH, localWorkSize, localWorkSize)
.putBarrier()
.finish();
// put memory from graphic card to host
queue.putReadBuffer(output, true).finish();
output.getBuffer().rewind();
for (int i = 0; i < image.getSize()[1]; ++i) {
for (int j = 0; j < image.getSize()[0]; j++) {
image.setAtIndex(j, i, output.getBuffer().get());
}
}
output.release();
queue.release();
return image;
}