Overheads for using various emake-annodetail options
Hi,
Do we have and ballpark figures for the likely impact on build performance for using the _basic_, _file_, _history_, _lookup_, _waiting_ values on the emake-annodetail options against emake? Similarly, do we have relatively sizes (i.e. percentages) that each option will use in the resulting annotation files (i.e. something like 10%,20%,30%,20%, 20% respectively)
TIA
Kev
Best Answer
Nobody's done exhaustive testing, but as a general rule of thumb you can assume:
- basic: 1% performance impact, small size impact
- env: 2% - 3% performance impact, small size impact
- file: 2% - 5% performance impact, medium size impact
- history: 1% - 2% performance impact, small size impact
- lookup: 5% - 10% performance impact, large size impact
- registry: 5% - 10% performance impact, large size impact
- waiting: 2% - 3% performance impact, small size impact
For the most part you can figure out the total impact by adding together the impact for individual detail levels. For example, if you enable both history and waiting, the impact is probably between 3% and 8% total. However, that does not apply for file combined with lookup, since file is a subset of lookup; and the same is true for basic, which is a subset of everything.
Of course, your mileage may vary, and the specific impact depends on your particular build and environment. If you are really concerned about the performance the best thing to do is to benchmark the build both with and without annotation enabled and base your decisions on that, rather than some vague estimates here.
Keep in mind too that for lookup and registry annotation, a significant part of the performance impact comes from the cost of writing so much annotation to disk. lookup in particular can easily cause annotation files to run to a gigabyte or more, so make sure the disk storing the annotation file is able to keep up, or you may face additional performance penalties.
