Comment by sjf

They ignored a key detail in their benchmark. They did not include the time it took to transfer the input file from memory to the GPU, (the GPU cores can only operate on graphics memory). This is easily going to be the most time consuming part. They also didn't include any IO operations in their timing, so instead of a 10x speedup, it is more likely to be a couple of percent faster as grep is going to be IO bound. This is a neat project, but there is a reason we are not using the GPU for regex matching.