Not necessarily. GPUs simply aren't optimized around branch-heavy or pointer-chasey code. If that describes the inner loop of your workload, it just doesn't mat

Not necessarily. GPUs simply aren't optimized around branch-heavy or pointer-chasey code. If that describes the inner loop of your workload, it just doesn't matter how well you can parallelize it at a higher level, CPU cores are going to be better than GPU cores at it.