This topic contains 14 replies, has 5 voices, and was last updated by 
Miqobot 7 years, 10 months ago.
-
Posts
-
At the moment Crafting Solver requires about 22 hours of extensive calculations in order to solve a 60** recipe. We are working on optimizations and optional parameters to bring the time down to usable range.
Unfortunately the optimization process is very unpredictable, so we can not give you an exact date. But we are doing our best to deliver the new features as fast as possible 🙂
We also have a special intermediate release being prepared. This will not be a beta, but a bonus feature. It was originally planned after Scenarios, but we think that changing priorities a little should not hurt, considering certain announcements in the game.
This feature will be revealed in a few days.Thank you for your patience very much!
Technically speaking, 22 hours is required only when CP chunks are calculated as 1:1. But even when calculated as 8:1, it still takes almost 3 hours so not an acceptable result.
We have implemented a special AI which is able to search for patterns within a calculated solution and convert them into clusters, which in turn can be implemented as formulas. This way, the solution stays mathematically correct, since Miqobot herself finds the proper way to optimize it.
So yes, at the moment she behaves somewhat like you described 🙂
But with v1.2.5 release, the clusters will be baked in Asm code and calculated at full speed.At the moment Crafting Solver requires about 22 hours of extensive calculations in order to solve a 60** recipe
Should the reduction of these times not pan out, could I recommend that Miqobot store these solved rotations in a database online? I realize gear is going to make this difficult. But what could be done is miqobot could check to see if the players gear is within a range where others are all getting the same recipe.
For example say player A first did the solver and player B with lower stats ran it later, received the same rotation. Both are now stored in the database when Player C comes along. He trys to run the solver but Miqobot sees his stats fall between the two, obviously he’d get the same rotation if he ran the solver, instead of waiting 10min-20min he could just click a button that says “Proceed with Community Recipe”…or some such.
There may be a level of risk involved, but everyone running the same computations seems wasteful to me…especially when you’d have to run it again everytime you change gear as well.
Yes, this a good idea, and we indeed consider adding an online database eventually 🙂
But time is not the only obstacle here, there are also memory requirements. For 60** recipe Miqobot needs 33.6 GB of data. It is compressible to some point, but it would still impose a significant strain on the network channel.
The ongoing clusterization process will address both time and memory issues.
Miqo, what I mean is more like storing the finished rotation and corresponding player stats. Data like:
Recipe:
70 dur, 1100 difficultyPlayer:
410 cp, 698 craft, 678 controlRotation:
Basic Touch, Steady Hand, Observe, etcI don’t see any need to store the proof of it, which I assume is what the 30+gb is. No need or benefit really, so long as its only stored if its generated by the bot…digital signature or some security method.
I think it makes sense at least?
Crafting Solver doesn’t use any rotations.
It calculates the required crafting ability on the fly.
This is 100% pure math, and it can not be expressed in static rotations.Otherwise, this would be a simple macro generator.
Crafting Solver is NOT a macro generator. It analyzes the current situation, predicts all possible outcomes, and calculates the best skill to apply at the current moment.
If you are interested, we can provide more details on what exactly is stored within these 30+ GB.A crafting state is described by a set of parameters:
- Condition (Normal, Good, Excellent, Poor)
- Progress (0-1436 for 60**)
- Quality (0-9430 for 60**)
- Durability (0-70 for 60**, divided by 5)
- CP (around 0-450)
- Stack of Inner Quiet (0-11)
- Stack of Steady Hand (0-5)
- Stack of Great Strides (0-3)
- Stack of Comfort Zone (0-10)
- Stack of Waste Not (0-8)
- Stack of Whistle While You Work (0-11 + 1 hidden stack for Finishing Touches)
There are other buffs with stacks, but this should be enough to give a general idea.
Next, in order for Crafting Solver to work, it has to calculate every possible state. Which means a total of
4 * 1436 * 9430 * 70/5 * 450 * 12 * 6 * 4 * 11 * 9 * 13 = 126 * 1015(126 quadrillions) states.
Every state requires 8 bytes, so this results in about 895 PB (Petabytes) of RAM.At this point, we have managed to completely remove the Quality bar from this equation and perform numerous other optimizations, which brought the number down to 30+ GB.
But this information must be present in the RAM at all times.And as you can see, character stats are not even considered here.
30+ GB is required for only a single recipe and for only specific character stats.- 698 Craftsmanship, 678 Control – 30+ GB of data
- 699 Craftsmanship, 678 Control – 30+ GB of data (which is similar, but not the same)
- 700 Craftsmanship, 678 Control – 30+ GB of data (which is also similar, but not the same)
- Character level must be also considered.
- Recipe level must be also considered.
And so on.
So if we were to implement a database for Crafting Solver in the form as it is now, it would probably take more storage space than an entire humanity has available 🙂
Still, we are very greatful for your interest in the development process.
Please keep posting ideas and provide feedback!Thank you for asking again 🙂
We have just recently posted updates regarding the development progress. Please refer to these threads:
While we can not give an exact ETA, we are working very hard to deliver the new features as fast as possible.
Thank you for your patience and understanding!
You must be logged in to reply to this topic.