MT5 - Cloud Agents stuck on Processing

New comment
 

Hello,

I am attempting to perform a two year back test on XAUUSD using my custom made EA. I am well aware that the more inputs that are used, the longer it will take to process / run.

I see the max combinations for my EA is 1.71228266496e+22

I see 256 agents are being used, but it is just stuck on processing.

If I leave it for a few hours, will it eventually start? I haven't been blocked from the network for such a large test no?

If I could get some clarification that would be great!


 

If you were wondering, there were also no errors when starting up the cloud network


 

I was originally seeing a max combination total of 1.71228266496e+22 for my EA.

I've now reduced my amount of inputs considerably but I'm still stuck on processing:



 
How long takes one pass on your pc?
 
Carl Schreiber #:
How long takes one pass on your pc?

I don't know, when I tried running the optimised back test with just my laptop, it got past the processing stage, to busy, then each core got stuck on 0.0.

I am using a HP pavilion plus 14, so the power is decent, but not enough to perform a back test of this size.

I was using the cloud network just fine on friday so I dont know what has suddenly happened 

 
SpoonsInL #: I don't know, when I tried running the optimised back test with just my laptop, it got past the processing stage, to busy, then each core got stuck on 0.0. I am using a HP pavilion plus 14, so the power is decent, but not enough to perform a back test of this size. I was using the cloud network just fine on friday so I dont know what has suddenly happened 

Imagen that it was able to do one pass in just one second (which is extremely optimistic and unlikely).

So if the entire optimisation takes 2080423437926400000 passes, that would take 65969794 millennia (x1000 years) on a single thread.

Do you really think that feasible, even with the aid of the MQL5 Cloud?

Could you even afford that amount of testing on the MQL5 cloud?

Be realistic. Reduce the number of passes. Be more astute about your parameter range choices.

 

So first do a profile of your code !! Editor => Debug => Profiling with Hist. Data and set in Editor Tools the start and end time of your test time.

Otherwise you pay a lot of money for CPUs spinning around!

 
Fernando Carreiro #:

Imagen that it was able to do one pass in just one second (which is extremely optimistic and unlikely).

So if the entire optimisation takes 2080423437926400000 passes, that would take 65969794 millennia (1000 years).

Do you really think that feasible, even with the aid of the MQL5 Cloud?

Could you even afford that much amount of testing on the MQL5 cloud?

Be realistic. Reduce the number of passes. Be more astute about your parameter range choices.

Okay point taken 😅
 
Carl Schreiber #:

So first do a profile of your code !! Editor => Debug => Profiling with Hist. Data and set in Editor Tools the start and end time of your test time.

Otherwise you pay a lot of money for CPUs spinning around!

I'll take a look at this
 

Okay I went into my code, changed some defaults for some of the Boolean values, so...

From:

2080423437926400000

To:

650132324352000

Now it makes sense why it wasn't starting, thanks for the blunt advice, I'll post an update here if the back test is actually successful @ Fernando Carreiro 👌

 
SpoonsInL #: Okay I went into my code, changed some defaults for some of the Boolean values, so...

From: 2080423437926400000 To: 650132324352000

Now it makes sense why it wasn't starting, thanks for the blunt advice, I'll post an update here if the back test is actually successful @ Fernando Carreiro 👌

So you have reduced it to a measly 20615 millennia.

So if you run it on 20615 threads, it will only take 1 millennium.

Yes, it is now totally feasible!!! 🙄

12
New comment
Reason: