Tester in new build 604 doesn't work correct - page 11

 

hello, what i found about 610 very slows tests, after this very slow test finished, when you Show trades on chart, "open chart" related indicators are not shown on chart, oppsite to 509.

this is possibly related with indicators-variables defined not at the top of the block but inside the block (if blocks). or some continues create/init of variables on each tick?

(610 test is very slow but accurate, matches with 509, but takes many hours to finish.)


and this slowness will be valid for real executions also, under 610, i guess.

 

hi, possible cause for slowness: iMA(1000) and some other long period indicators run very slow, under 610 tester. under 509, these are comfortable. and working on real conditions..

if i reduce period like iMA(30), tester gets faster.. but again does not show used indicators on chart, when i press "open chart", only trades shown.. (509 shows indicators on chart correctly, if tester timeframe matches.)


MA5 = iMA(NULL, PERIOD_M5, 1000, 0, MODE_SMMA, PRICE_MEDIAN, 2);

 
cbalta:

hi, possible cause for slowness: iMA(1000) and some other long period indicators run very slow, under 610 tester. under 509, these are comfortable. and working on real conditions..

if i reduce period like iMA(30), tester gets faster.. but again does not show used indicators on chart, when i press "open chart", only trades shown.. (509 shows indicators on chart correctly, if tester timeframe matches.)


MA5 = iMA(NULL, PERIOD_M5, 1000, 0, MODE_SMMA, PRICE_MEDIAN, 2);



Have you reported either of these to the service desk for consideration?
 

I would start a fresh EA. Build it with the same indicators, one at a time . Start with a very basic EA structure, and only add one indicator at a time.

Test each time as you add to your EA, until you identify the exact functionality differences, between the new and old versions of mt4.

Starting a fully matured EA into a new platform update will be very difficult to isolate the individual functionality changes imo.

Starting with a skeletonized EA, will be much easier to sort out the changes, and find work arounds.

 
moneycode:

I would start a fresh EA. Build it with the same indicators, one at a time . Start with a very basic EA structure, and only add one indicator at a time.

Test each time as you add to your EA, until you identify the exact functionality differences, between the new and old versions of mt4.

Starting a fully matured EA into a new platform update will be very difficult to isolate the individual functionality changes imo.

Starting with a skeletonized EA, will be much easier to sort out the changes, and find work arounds.

sorry, i'm not a starter. i'm not developing new systems, i'm migrating proven code. what you will find is, some disease of new environment, when your builds finish.


meanwhile, it seems that, external icustom calls also very slow, for migrated code compared to 509.

 
Trevhib:

Have you reported either of these to the service desk for consideration?
thank you. added.
 
cbalta:

hello, what i found about 610 very slows tests, after this very slow test finished, when you Show trades on chart, "open chart" related indicators are not shown on chart, oppsite to 509.

this is possibly related with indicators-variables defined not at the top of the block but inside the block (if blocks). or some continues create/init of variables on each tick?

(610 test is very slow but accurate, matches with 509, but takes many hours to finish.)

and this slowness will be valid for real executions also, under 610, i guess.

I concur, I ran the test EA 600+ again on the same chart data as before 1 yr backtest. Results are very slow compared to previous.

New Speed Test: Test EA 1yr backtest on build 610 compared with previous test results.

MT4 build 509 MT4 build 606 MT4 build 610
Test1
Test2
Test3
13.510 seconds
13.588 seconds
13.603 seconds
12.075 seconds
11.778 seconds
12.043 seconds
26.317 seconds
25.943 seconds
27.815 seconds
Average13.567 seconds11.965 seconds26.692 seconds
Difference vs 50911.8% faster96.7% slower

OK here is an issue, I noticed terminal.exe was holding onto 750 MB RAM the whole time, even before I started the tests, So now I deleted the 500 MB .fxt file from tester. Terminal.exe immediately released most of that RAM. Now I run the 3 tests again on 610.

Terminal.exe only holding 68 MB RAM throughout these next tests. Newly created .fxt file is again 500 MB as expected for a 1yr backtest.

MT4 build 610 (after deleting old .fxt file)
Test1
Test2
Test3
11.622 seconds
11.934 seconds
11.481 seconds
Average11.678 seconds
Differnce vs 50913.9% faster

I believe that must mean there is a memory issue with build 610

 

My EA, under 509 release generates more than 2000 orders over year 2013.

After recompiling it, without ANY modification, under 610, it generates 3 orders and a list of : "tester: 0 symbol not found".

Any idea to fix will be appreciated.

 
FaceNord:

My EA, under 509 release generates more than 2000 orders over year 2013.

After recompiling it, without ANY modification, under 610, it generates 3 orders and a list of : "tester: 0 symbol not found".

Any idea to fix will be appreciated.


Welcome to mql4.com forum,

Is the symbol traded by your EA selected in Market Watch ?

 

So I went to the store accross the street and back I guess I was gone about 30 minutes now I look at task manager again, terminal.exe RAM usage climbed from 68 MB to 544 MB while I was gone, so whats up with that ?

Edit:

I deleted the .fxt file again this time it made no difference to RAM usage. I closed out the terminal but task manager still shows terminal.exe is running. So I repopened the terminal and reclosed it, this time it closed out of task manager too. I restart the terminal. It is now back down to using 56 MB RAM.

Reason: