CloudCutter is an independent project developed essentially by one person. Does software like this need serious performance testing? The answer is yes - perhaps even more than established software does.
Large software companies already possess something a new project does not: accumulated trust. An independent developer cannot ask for the same confidence without evidence. A performance statement is only a claim until other engineers can examine the dataset, hardware, method, measurements, and output verification.
1. Define the test before seeing the result
The methodology should not change depending on which result appears more attractive. Source data, requested operation, software version, and relevant conditions should be defined in advance. Better, similar, or worse performance are all valid outcomes.
2. Publish the environment
"This cloud loaded in 40 seconds" has little meaning without context. A useful report identifies CPU, RAM, GPU, storage, operating system, source data, software version, and test condition.
3. Separate machine time from operator time
Processing speed is only one part of an engineering workflow. CloudCutter reports distinguish:
- file reading and refinement time;
- active operator preparation time;
- processing and export time.
A workflow that saves computer time but requires substantially more engineering time is not necessarily more efficient.
4. Distinguish cold and repeated runs
One run is an observation, not a conclusion. Initial and repeated runs should remain distinguishable because operating-system cache, storage state, and prior file access can influence results.
5. Verify the output
A fast export is irrelevant if the output is incomplete or incorrect. Reports should include point counts, file sizes, validation status, and reproducibility checks where practical.
What the benchmarks are intended to prove
The objective is not to claim that CloudCutter is universally faster than established professional tools. The question is narrower: can independently developed software perform a real engineering task at a credible professional level?
If CloudCutter is faster in a specific workflow, that is useful. If performance is similar, that is already significant. If it is slower, the benchmark identifies where development should continue.
The benchmark series is now published at voxelstudio.app/cloudcutter/benchmarks. The archive should link this policy article directly to the measured results that followed it.