About Cuda 1D convolution, How can I do this faster?

Question

• int threads = 32;
• dim3 blocks(250000/31,129,50);
• coefsize = 129;

global void D_Conv(float in, float coef, flo