Sparse Matrix-Vector Multiplication Optimizations based on Matrix Bandwidth Reduction using NVIDIA CUDA | IEEE Conference Publication | IEEE Xplore