def run_OpenCL(oclu, ctx, queue, scene, leaf_array, fluence_data, intensities, settingsList, optimizationParameters):
debugOpenCL = Debug()
X_size = int(optimizationParameters[1][1])
Y_size = int(optimizationParameters[2][1])
Z_size = int(optimizationParameters[3][1])
if Settings.AUTOTUNE == 1:
Settings.PIECES = int(optimizationParameters[4][1])
[scene, collimators, leaf_array] = init_scene(Settings.PIECES)
settingsList = define_settings(scene, leaf_array)
settingsString = Settings.macroString(settingsList)
optParametersString = Settings.macroString(optimizationParameters)
Settings.WG_LIGHT_SAMPLING_SIZE = X_size*Y_size*Z_size
optParametersString += " -D WG_LIGHT_SAMPLING_SIZE=" + str(X_size*Y_size*Z_size)
#program = oclu.loadProgram(ctx, Settings.PATH_OPENCL + "RayTracingGPU.cl", "-cl-nv-verbose " + settingsString)
#program = oclu.loadProgram(ctx, Settings.PATH_OPENCL + "RayTracingGPU.cl", "-cl-auto-vectorize-disable " + settingsString)
#program = oclu.loadProgram(ctx, Settings.PATH_OPENCL + "RayTracingGPU.cl", " " + settingsString + " " + optParametersString)
program = oclu.loadCachedProgram(ctx, Settings.PATH_OPENCL + "RayTracing.cl", " " + settingsString + " " + optParametersString)
mf = cl.mem_flags
time0 = time()
scene_buf = cl.Buffer(ctx, mf.READ_ONLY | mf.COPY_HOST_PTR, hostbuf=scene)
#render_buf = cl.Buffer(ctx, mf.READ_ONLY | mf.COPY_HOST_PTR, hostbuf=render)
leaf_array_buf = cl.Buffer(ctx, mf.READ_ONLY | mf.COPY_HOST_PTR, hostbuf=leaf_array)
intensities_buf = cl.Buffer(ctx, mf.READ_WRITE | mf.ALLOC_HOST_PTR, size=intensities.nbytes)
fluence_data_buf = cl.Buffer(ctx, mf.READ_WRITE | mf.ALLOC_HOST_PTR, size=fluence_data.nbytes)
debugOpenCL_buf = cl.Buffer(ctx, mf.WRITE_ONLY, sizeof(debugOpenCL))
time1 = time()
program.flatLightSourceSampling(queue, intensities.shape, (X_size, Y_size, Z_size), scene_buf, leaf_array_buf, intensities_buf, debugOpenCL_buf).wait()
#program.flatLightSourceSampling(queue, intensities.shape, None, scene_buf, leaf_array_buf, intensities_buf, debugOpenCL_buf).wait()
time2 = time()
program.calculateIntensityDecreaseWithDistance(queue, fluence_data.shape, None, scene_buf, fluence_data_buf, debugOpenCL_buf).wait()
time3 = time()
program.calcFluenceElement(queue, fluence_data.shape, None, scene_buf, intensities_buf, fluence_data_buf, debugOpenCL_buf).wait()
time4 = time()
cl.enqueue_read_buffer(queue, fluence_data_buf, fluence_data)
cl.enqueue_read_buffer(queue, debugOpenCL_buf, debugOpenCL).wait()
totalTime = time()-time0
calculationTime = time4-time1
#print "flatLightSourceSampling(): ", time2 - time1, ", calculateIntensityDecreaseWithDistance():", time3 - time2, ", calcFluenceElement():", time4 - time3
samplesPerSecondOpenCL = Settings.FLX*Settings.FLY*Settings.LSAMPLESSQR/totalTime
#print "Time OpenCL: ", timeOpenCL, " Samples per second: ", samplesPerSecondOpenCL
#print fluence_data
return [fluence_data, totalTime, calculationTime, samplesPerSecondOpenCL]
def define_settings(scene, leaf_array):
# Settings
Settings.XSTEP = float(length(scene.fluenceMap.rectangle.p1 - scene.fluenceMap.rectangle.p0)/Settings.FLX) # Length in x / x resolution
Settings.YSTEP = float(length(scene.fluenceMap.rectangle.p3 - scene.fluenceMap.rectangle.p0)/Settings.FLY) # Length in y / y resolution
Settings.XOFFSET = float(Settings.XSTEP/2.0)
Settings.YOFFSET = float(Settings.YSTEP/2.0)
Settings.LSTEP = float(scene.raySource.radius*2/(Settings.LSAMPLES-1))
Settings.LEAF_DATA_SIZE = len(leaf_array)
settingsList = Settings.getDefaultSettingsList()
settingsList.append(("XSTEP", str(Settings.XSTEP)+'f', True))
settingsList.append(("YSTEP", str(Settings.YSTEP)+'f', True))
settingsList.append(("XOFFSET", str(Settings.XOFFSET)+'f', True))
settingsList.append(("YOFFSET", str(Settings.YOFFSET)+'f', True))
settingsList.append(("LSTEP", str(Settings.LSTEP)+'f', True))
settingsList.append(("LEAF_DATA_SIZE", str(Settings.LEAF_DATA_SIZE), True))
return settingsList
def main():
setDefaultSettings()
#select_excecution_environment()
if Settings.OPENCL == 1:
[ctx, queue] = init_OpenCL()
#test()
[scene, collimators, leaf_array] = init_scene(Settings.PIECES)
settingsList = define_settings(scene, leaf_array)
oclu = OpenCLUtility.OpenCLUtility()
if Settings.AUTOTUNE == 1:
list = []
list.append(Parameter("LINE_TRIANGLE_INTERSECTION_ALGORITHM", [2], True))
list.append(Parameter("WG_LIGHT_SAMPLING_X", [1,2,4,8,16,32,64,128], False))
list.append(Parameter("WG_LIGHT_SAMPLING_Y", [1,2,4,8,16,32,64,128], False))
list.append(Parameter("WG_LIGHT_SAMPLING_Z", [1,2,4,8,16], False))
list.append(Parameter("PIECES", [1,2,4,10,20], False))
list.append(Parameter("RAY_AS", [0], True))
list.append(Parameter("LEAF_AS", [1], True))
list.append(Parameter("LEAF_DATA_AS", [2], True))
list.append(Parameter("SCENE_AS", [2], True))
list.append(Parameter("DEPTH_FIRST", [0], True))
fluence_data = numpy.zeros(shape=(Settings.FLX,Settings.FLY), dtype=numpy.float32)
intensities = numpy.zeros(shape=(Settings.FLX,Settings.FLY,Settings.LSAMPLESSQR), dtype=numpy.float32)
at = Autotune(ParameterSet(list), run_OpenCL, (oclu, ctx, queue, scene, leaf_array, fluence_data, intensities, settingsList))
at.findOptimizationParameters()
#print at.getTable()
at.saveCSV()
[fluence_data_OpenCL, total_time_OpenCL, calculation_time_OpenCL, samplesPerSecond_OpenCL] = run_OpenCL(oclu, ctx, queue, scene, leaf_array, fluence_data, intensities, settingsList, at.best_parameters)
else:
fluence_data = numpy.zeros(shape=(Settings.FLX,Settings.FLY), dtype=numpy.float32)
intensities = numpy.zeros(shape=(Settings.FLX,Settings.FLY,Settings.LSAMPLESSQR), dtype=numpy.float32)
[fluence_data_OpenCL, total_time_OpenCL, calculation_time_OpenCL, samplesPerSecond_OpenCL] = run_OpenCL(oclu, ctx, queue, scene, leaf_array, fluence_data, intensities, settingsList, Settings.getDefaultOptimizationParameterList())
if Settings.SHOW_PLOT == 1:
if Settings.PYTHON == 1:
show_plot(scene, fluence_data_Python, time_Python, samplesPerSecond_Python)
if Settings.OPENCL == 1:
show_plot(scene, fluence_data_OpenCL, total_time_OpenCL, samplesPerSecond_OpenCL)
if Settings.SHOW_3D_SCENE == 1:
show_3D_scene(scene, leaf_array, collimators)
