Investigation 1: Throughput and efficiencyDwarves: Dwarves have default speed (900). They have no breathe, no eat, no drink tags, maybe a few others like that (shouldn't affect results, but full disclosure).
Population is 86 and 1 child. Sufficient haulers exist such that there are always idlers to haul.
Military population is 10. They're poorly trained. Several dwarves have wounds, and there are no uber-dwarves.
Fortress: Meeting hall, where idlers will mostly congregate, is about 40 tiles from the door to the undump chamber.
Objects: We're using bar/block stockpiles. My fortress contains 1991 bars or blocks that are unused and unforbidden.
Undump Facility: Here's a color-coded picture of it.
Beginning state is empty undumps and a full 5x5 bar/block stockpile in the center of it. At the beginning of summer (seasonal autosave + pause) the central stockpile was destroyed and 8 stockpiles were created: 4 undump stockpiles and 4 feeder stockpiles. All feeder stockpiles were created after the undump stockpiles, but during the same pause. It's color coded so you can see position and which stockpile feeds which undump. I've labelled the north one to avoid confusion-- you can see that dwarves enter the north undump from the south, while they enter the south undump from the north.
Each undump is set to take from the associated feeder stockpile.
At the end of summer, the number of objects hauled and the number of job cancellations were counted.
Perfect hauls: A perfect haul is a haul that delivers the object to the undump stockpile.
North: 46 perfect hauls
West: 31 perfect hauls
South: 42 perfect hauls
East: 45 perfect hauls
Adequate hauls: An adequate haul is a haul that delivers the object to the pressure plate of the associated undump. These are adequate for clearing or feeding workshops, but food will rot on these squares, so that's why I distinguish between these two kinds of hauls. Note that adequate hauls will be brought to the feeder stockpiles, so there may have been more adequate hauls than are documented. Although feeder stockpiles are probably necessary for labor efficiency, these undocumented adequate hauls would still be valid hauls for certain purposes-- for instance, for feeding raw materials into a active workshop, or for uncluttering an active workshop.
North: 7 adequate hauls
West: 5 adequate hauls
South: 5 adequate hauls
East: 7 adequate hauls
Average perfect hauls per undump: 41Average adequate hauls per undump: 6Adequate hauls/Total hauls: 12.8% This can be seen as potential food wastage.
Expected number of job cancellations due to drop-off inaccessible: 160 This requires a little explanation. With so many idlers, all 8 of our undump squares will generate hauling labors when we start the clock. We have 164 perfect hauls, but 4 of those hauling jobs won't create cancellations, because they are to accessible tiles. We expect this spam only when dwarves successfully bring an object to a stockpile square.
Actual number of job cancellations due to drop-off inaccessible: 188 Dwarf should only get this cancellation when dropping an item on the stockpile square. Not sure why it's higher than expected number.
Actual number of job cancellations due to path blocked: 144 This is tough to interpret as well. Sometimes, a dwarf will report this cancellation instead of drop-off inaccessible, even when dropping on to the stockpile square. I believe this is the spam when dropping onto the plate as well.
Expected cancellations as a percentage of total cancellations: 48.2% This is a lower-bound estimate of labor efficiency. It's extraordinarily rough right now, because I don't know where the cancellations are occurring-- if a dwarf cancels at the last minute with every cancellation, we have this number. If the dwarf cancels the job earlier, we have a higher level of efficiency, potentially up to 100%.
EDIT: Have corrected some numbers and calculations, as well as grammar-- reasoning goes slow for me
Peer review (criticism) welcome as always; I make dumb mistakes all the time.
A second investigation (single dwarf, single undump, single feeder stockpile, only 15 bars available) led to 26 cancellations, and no plated bars. It was using the northern undump only. Watching the dwarf, he appeared to only cancel either at the end or beginning of his run, meaning that while I can't explain the extra cancellations, they appear to be 0-time cancellations-- immediate upon receiving the job. That suggests good labor efficiency.
When I dumped one bar on to the pressure plate to test feeder functionality, the dwarf first moved it to the feeder stockpile, and from there back to the undump.
One dwarf tests are probably anecdotal at best. Still, I feel confident about the use of a enters-from-the-south undump fed by a nearby feeder stockpile (1 tile works) that was built subsequent to the undump stockpile. That's probably way more specific than I need to be, but just in case