1. 14

How Does Energy, Time, and Memory Relate?

One strange result I noticed is that Swift often does well on energy usage even where doesn’t do so well on execution time. Maybe that’s a result of Swift being targeted towards mobile?

  1.  

  2. 1

    does the benchmark is related to any realistic workload ?

    1. 2

      No. It uses some benchmark game programs. These are not a realistic workload today.

      1. 2

        Would you think substring matching and binary tree search are unrealistic workloads?

        1. 1

          unrealistic workloads

          Are you seriously expecting different energy use profiles in realistic benchmarks?

          1. 2

            I am not. Why? I think these are “realistic” to begin with, if it wasn’t clear.

          2. 1

            No they aren’t: It is a very small subset of the realistic workloads. More realistic workload would be real programs doing the same functionality. For example, in python those are typically done using C/C++/fortran library, so it should be in similar efficiency as in C. It is clearly not in the same order of efficiency, so it is very likely the condition are unrealistic. (I don’t say that python is as efficient as C)

            1. 2

              So what you are saying is, CPU bound/algorithmic workloads in Python are unrealistic? OK then. They are certainly realistic in a bunch of other languages on the list which don’t outsource real work to FFI.

              1. 1

                I just say that they are not realists by all mean of measure. The python case was only an example of the unrealistic thing