Trade-offs between depth and width in parallel computation