'''
'''

import pycuda.driver as cuda
import pycuda.autoinit

from pycuda.compiler import SourceModule


class SignalCalculator():
    '''
    '''


    def __init__(self):
        '''
        '''
        self.generators = []
        
    def add_signal_generator(self, gen):
        self.generators.append(gen)
        
    def make_cuda_kernel(self):
        src = """
        #include <curand_kernel.h>

        const int nstates = %(NGENERATORS)s;
        __device__ curandState_t* states[nstates];
    
        __global__ void initkernel(int seed)
        {
            int tidx = threadIdx.x + blockIdx.x * blockDim.x;
    
            if (tidx < nstates) {
                curandState_t* s = new curandState_t;
                if (s != 0) {
                    curand_init(seed, tidx, 0, s);
                }
    
                states[tidx] = s;
            }
        }

        __global__ void calc_signal(float* open, float* high, float* low, float* close, size_t points, int8_t* signals, char* sig_names, float* scratch)
        {
            int tidx = threadIdx.x + blockIdx.x * blockDim.x;
            curandState_t s = *states[tidx];
            int signals = 5 * curand_uniform(&s);
            float* tScratch = &scratch[tidx * points];
            int8_t* tSignals = &signals[tidx * points];
            char* tSigName = &signames[tidx * 256];
            
            for(int i = 0; i < signals; i++)
            {
                int sigtype = curand_uniform(&s);
                switch(sigtype)
                {
        """
        
        i = 0
        for gen in self.generators:
            src += gen.make_signal_kernel(i)
            i += 1
            
        src += """
                }
            }
        }
        """