Abstract: A fast parallel algorithm based on Bi-conjugate gradient stabilized(BiCGStab) was proposed to solve the simplified spherical harmonics(SP_N) equations to reducethe computational burden and improve computational efficiency for forward problem of bioluminescence tomography(BLT). In the algorithm, preconditioned method of incomplete-Cholesky factorization and sparse matrix's compressed row storage scheme (CSR) representing method were adopted. Furthermore, compute unified device architecture(CUDA) parallel programming model was used to implement parallel accelerating. The results of numerical simulation show that the proposed algorithm not only can ensure the solution accuracy of the forward problem but also greatly shorten the equation-solving time.