Runs so Sloooow

[UP844]

I was planning to bother you all with another pair of AAR about CG002 - Prokhorovka.

I regret to give up this, but the delay between an order and its execution has become so long (70 seconds) I have been able to read an entire Osprey book during one half of a Movement Segment.

I was attempting to bring the Russian armored reinforcements (T-34s and SU-76s) to the front: I spent about half an hour to move a single tank [:(]. At such rate, moving 12 T-34s and 6 SU-76s will take approximately 9 hours.

I do not know if the LOS check is made for all the 630 hexes within the 14-hex LOS range, or only between the moving unit location and the enemy-occupied hexes. At night (with a 2-hex LOS range), the delay is negligible.

I do not think the issue resides with my PC: CPU use is 11-14% and memory use is 74.904 KB (17% of physical memory). I also tried installing Razer Game Booster, with no effect at all.

Some amount of delay has always been present in large scenario, but I feel it has dramatically increased since version 1.0.52.

I attach a saved game file: I would like to know how much delay you experience if you move the T-34 M41 in hex 28,32.

P.S. in this open terrain, the failure to fire at long range of AI infantry led to the almost complete destruction of the SS squads before the end of day 1. German infantry re-appeared on turn 16, when an engineer force (I suppose they are day 2 reinforcements) appeared on the southeast area of the map. Russian infantry losses were not light, but they were inflicted by tanks only - no German squad ever managed to come close enough to fire its inherent firepower.

[Paullus]

I'll take a look tomorrow. Just home from a party [;)]

[UP844]

Thanks, Paullus [&o]

[Paullus]

Well it does take some time on my computers as well, 35-45 seconds. Lowering the day visibility would help. Peter will start looking at optimizations on this logic starting ASAP.

[Peter Fisla]

Hi All,

The larger scenarios with maps that have a lot of elevations with units being too far apart will slow down the game engine because of Line of Sight and AI processing. UP844, I have done some testing with your saved game file and for me it takes about 30 seconds for the game engine to respond. The game does utilizes support for multi core CPUs, however the game engine wasn't written from group up to fully support multi-core CPUs. I'm hoping that soon to be released Visual Studio 2017 will further help me to better optimize my game engine however please keep in mind that putting too many assets on a large map will slow down the game play, depending how good your computer is. You need powerful desktop to run large scenarios. I will continue to update and better optimize the game engine for the future.

[Gerry4321]

In terms of LOS checking and the T-34 moving - the AI knows where all the German units are, even the "hidden" ones. So there should not be many hexes to check in the visibility range?

ORIGINAL: UP844

I was planning to bother you all with another pair of AAR about CG002 - Prokhorovka.

I regret to give up this, but the delay between an order and its execution has become so long (70 seconds) I have been able to read an entire Osprey book during one half of a Movement Segment.

I was attempting to bring the Russian armored reinforcements (T-34s and SU-76s) to the front: I spent about half an hour to move a single tank [:(]. At such rate, moving 12 T-34s and 6 SU-76s will take approximately 9 hours.

I do not know if the LOS check is made for all the 630 hexes within the 14-hex LOS range, or only between the moving unit location and the enemy-occupied hexes. At night (with a 2-hex LOS range), the delay is negligible.

I do not think the issue resides with my PC: CPU use is 11-14% and memory use is 74.904 KB (17% of physical memory). I also tried installing Razer Game Booster, with no effect at all.

Some amount of delay has always been present in large scenario, but I feel it has dramatically increased since version 1.0.52.

I attach a saved game file: I would like to know how much delay you experience if you move the T-34 M41 in hex 28,32.

P.S. in this open terrain, the failure to fire at long range of AI infantry led to the almost complete destruction of the SS squads before the end of day 1. German infantry re-appeared on turn 16, when an engineer force (I suppose they are day 2 reinforcements) appeared on the southeast area of the map. Russian infantry losses were not light, but they were inflicted by tanks only - no German squad ever managed to come close enough to fire its inherent firepower.

[Peter Fisla]

Every time your friendly unit moves, the game needs to check every enemy hex if it actually contains enemy units, to see if there is a LOS or not.

ORIGINAL: Gerry


In terms of LOS checking and the T-34 moving - the AI knows where all the German units are, even the "hidden" ones. So there should not be many hexes to check in the visibility range?

ORIGINAL: UP844

I was planning to bother you all with another pair of AAR about CG002 - Prokhorovka.

I regret to give up this, but the delay between an order and its execution has become so long (70 seconds) I have been able to read an entire Osprey book during one half of a Movement Segment.

I was attempting to bring the Russian armored reinforcements (T-34s and SU-76s) to the front: I spent about half an hour to move a single tank [:(]. At such rate, moving 12 T-34s and 6 SU-76s will take approximately 9 hours.

I do not know if the LOS check is made for all the 630 hexes within the 14-hex LOS range, or only between the moving unit location and the enemy-occupied hexes. At night (with a 2-hex LOS range), the delay is negligible.

I do not think the issue resides with my PC: CPU use is 11-14% and memory use is 74.904 KB (17% of physical memory). I also tried installing Razer Game Booster, with no effect at all.

Some amount of delay has always been present in large scenario, but I feel it has dramatically increased since version 1.0.52.

I attach a saved game file: I would like to know how much delay you experience if you move the T-34 M41 in hex 28,32.

P.S. in this open terrain, the failure to fire at long range of AI infantry led to the almost complete destruction of the SS squads before the end of day 1. German infantry re-appeared on turn 16, when an engineer force (I suppose they are day 2 reinforcements) appeared on the southeast area of the map. Russian infantry losses were not light, but they were inflicted by tanks only - no German squad ever managed to come close enough to fire its inherent firepower.

[Gerry4321]

Not sure I understand Peter. The AI keeps track of where units are so no other hexes need to be checked other than those hexes with enemy units, right?

[Rosseau]

What comes to mind is the ancient (and huge) HPS games moving along like lightning on almost any computer. They may have taken an easy way out LOS-wise, but there is an op-fire phase. I also sort of doubt their optimization was that great in 1999!

Not all games have big maps and lots of units, so maybe TotH is designed to be played out via smaller scenarios anyway. The real issue is with the DLC. After hundreds of user scenarios, I had to support it. But there are moral implications to selling a product that is mostly not going to work. That is for you folks to figure out.

[Peter Fisla]

ORIGINAL: Gerry

Not sure I understand Peter. The AI keeps track of where units are so no other hexes need to be checked other than those hexes with enemy units, right?

The LOS function needs to check enemy occupied hexes every time you move your unit, to see if your unit potentially moved into LOS of new enemy hex or your unit moved out of LOS of enemy hex (or both). The LOS function is also used for AI to evaluate possible human player targets because stuff like hindrance will affect the fire function.

[Missouri_Rebel]

Well I'm afraid that the performance issue seems to have found me as well.

I've spent a few nights carefully recreating Shilovo, a village on a hill overlooking the Don River. The problem is when I added the units and tried play testing it ran so slow as to make it unplayable. Maybe it's the hills? Maybe it's the 3 company's and a few more assets that ground things to a crawl but, honestly, it shouldnt have. I've seen way more units in the stock maps that ran better.

I am on a 6800k @ 3.4Ghz, installed on a 512 gb SSD, 32 gigs of 3000 RAM and a water cooled GTX 1080. Plenty of power to run any game I've come across at the highest settings. I even cut down on the map size to be 1X3 and got rid of the river section (but still has the hill) to no avail.

I'm pretty disappointed and think I'll be giving up on scenario creation as quickly as I picked it up.

EDIT

THE ISSUES SEEM TO BE MY OWN DOING BY NOT PROPERLY STRUCTURING A SCENARIO

[Missouri_Rebel]

Here is the unfinished scenario. the .txt extension to .scn .

[UP844]

ORIGINAL: Paullus

Well it does take some time on my computers as well, 35-45 seconds.

ORIGINAL: Peter Fisla

UP844, I have done some testing with your saved game file and for me it takes about 30 seconds for the game engine to respond.

Thanks for the tests: your replies confirm that - even with a PC more recent and powerful than mine - moving the T-34 closer to the front will then take a significant amount of time.

Assuming the tank is CE and able to use the road movement rate, it will take 10 to 15 minutes to move 20 hexes (spending some MPs to start, change elevation and changing its VCA). The Russian receives 18 AFVs as reinforcements on turns 13,14 and 15, which will then require 180 to 240 minutes to move.

ORIGINAL: Peter Fisla

Every time your friendly unit moves, the game needs to check every enemy hex if it actually contains enemy units, to see if there is a LOS or not.

The T-34 in the saved game file is approximately 50 hexes away from the nearest German unit (a broken SS squad in 3,81). Since the delay is much shorter in night turns, when visibility is 2 hexes, I suppose the LOS-checking routine does not perform the whole LOS-checking procedure if the range is greater than the visibility range. Is this assumption correct?

Some more remarks I hope could be helpful to solve the issue:

I have noticed, in CG001, that the delay does not significantly decrease even when there are very few enemy units on the map. By way of example, at the end of day 3, the only "enemy-occupied hexes" were a few hexes (about half a dozen) containing unpossessed German SW (which require a LOS check to decide if the SW counter must be displayed or not).

I also noticed that the delay does increase as the game progresses: I started another CG002 and moved a KV-1s on Turn 1: the delay is about 5-7 seconds (see attached save file), even though there are much more German units than in the saved game I attached above.

I launched the LOS tool to see if it could calculate the range between the T-34 and the nearest German unit: the message appearing in the LOS tool window reported that range was greater then the scenario visibility range, but it also stated "Check_LOS_FOW executed 2347 times" and "64.65 seconds".
Apparently, every LOS check took approximately 0.0275 seconds: what does, however, a single LOS check execution imply? Checking LOS between a unit and an enemy-occupied hex? In this case, checking LOS between the T-34 and all the enemy-occupied hexes (let's assume there are 30 of them) should not take more than 0.825 seconds. I'm quite puzzled [&:].

[CapnDarwin]

UP844, how big is the map in your test? Have you tested just one tank versus one tank starting out of visual range? I'm curious if the LOS routine is running the whole map instead of bailing out and saying "out of range".

[UP844]

The test was performed with the CG002 scenario: I have not measured the map, but I think it is approx. 100 hexes high and 30 hexes wide.

[UP844]

Just for the sake of it, I launched CG002 playing the German side and leaving the set-up to the computer.

The Germans (37 units) set up in 18 hexes. The Russians (20 units) set up in 6 hexes. I launched the LOS tool and the routine was executed 50 times in 2.5 seconds.

How comes the same routine is executed a staggering 2347 times at a later time (see above) with no approx. 35-40 Russian units in as many hexes and - I believe - no more than 20 German units?

[Hailstone]

I just upgraded my PC with an i7 8 core 7th generation processor at 4.7GHz 16M of DDR4 Ram and a video card with 8G of DDR4 memory and I still can't play the DLC. I can't complain because I knew this would happen before purchasing the game but I was hoping that supporting TotH development would address this and fix it. So far all I hear is crickets. Look, as the game is with the original scenarios it's fine. Just don't put out DLC's with ambitious scenarios that will make TotH look pretty bad.

[Paullus]

We are looking into this but it will take some time. Could you just give us some more information Hailstone of your new try with the DLC? What did you try and what did you experience?

[UP844]

Excuse me for intruding but I attached a save file game in post #40, stating that I had a 70-second delay before a unit completed its move. Both you and Peter tried the saved game, with a delays ranging from 30 to 45 seconds.

I have purchased Kursk as soon as it was released and I would gladly purchase it again tomorrow just to support "the Cause". I cannot avoid wondering, however, if the scenarios have been properly tested before release, given the results you obtained.

[dox44]

i don't see how anyone could have tested the campaign...its unplayable...well this side of eternity.