Spring RTS Engine

i kinda feel like my arms are all floppy

like if i dont move them aorund they are gonna fall off

TradeMark wrote:I dont think betalord can make such benchmark program. Hes just lobby coder.

work out first 500 prime numbers with an algorythm,

Var = starttime(microsecconds)

loop (workout prime numbers)

System Spec = (TimeNow(Micro) - starttime) Multipled by some factor

gives a number rating the speed of your system.

That only measures raw processing power, other factors like memory and gpu speed/vram also determine ingame performance.

its about 800 times more usefull than the cpu speed though.

and it would require 0 effort to convert the lobby to do that.

that doesn't take cache into account

yet again, its still better than the current system

im not saying its perfect but it works better than having a cpu value (especialy the stupidly high AMD ones)

Tobi wrote:TBH the CPU thing should just be removed.

actually it could help to determine sync error reasons while resync is still in development... but that's all.

Why dont we just show the name of the cpu instead?

My core 2 duo at 2.4Ghz per core is registering as 2.4Ghz in the lobby, and the lobby rounds up and down so my 2.66Ghz pentium D is 2.7Ghz in the lobby, when really they should show as 4.8Ghz and 5.2Ghz respectively at the very least.

I put my vote in for some kind of benchmark. Though not only a cpu benchmark... CPU, GPU, Ram, hard drive, and network connection.

CPU speed in fact, tells us absolutley nothing, almost like reyling on the BBC for detailed news. Mhz not amounting to any useful information is nothing new, anyone that has been around computers for awhile should understand what I mean when I mention Cyrix or Win-Bond.

If your buying a truck, or car, and you were concerned with how fast you can go in it/how much you can tow, would you ask the salesman "So, how many RPM's does this engine spin?", and then pick a car or truck with the fastest spinning engine? Hopefully not. Everyone here should have a basic understanding of what Mhz does, and does not mean.

It comes down to, the goals of the people working on this game. There are quite a few things that need work, if anyone really cares to have this game more "popular", some sort of benchmark scale should be considered. Either fix it, or get rid of it, theres no sense in having a broken system.

AF wrote:Why dont we just show the name of the cpu instead?

My core 2 duo at 2.4Ghz per core is registering as 2.4Ghz in the lobby, and the lobby rounds up and down so my 2.66Ghz pentium D is 2.7Ghz in the lobby, when really they should show as 4.8Ghz and 5.2Ghz respectively at the very least.

AF, Mhz isnt additive, ESPECIALLY with software that only uses one core.

But you can change the settings to use both. I know that my computer uses both cores with Spring, though there is a slight bias to the "first" of the two.

There are at least 3 thread in spring.

- main thread
- sound
- pathfinding

And depending on what you use

- AI

More so for the last one (OTAI and NTai mainly)

You also have AMD reverse hyperthreading.

But I agree, a basic rating system for hardware would be very nice. If you can all agree on a basic system and its implementation then I'd put it in AFLobby or cookiebot or whatever you wanted.

I think that we should run some small benchmark or just check system and determine one number "power" by taking other factors into account (CPU, GPU, RAM)

Some extra exacutable or parameter for spring that would run benchmark/test and report back one number would be best.

No, there is no pathfinding thread.

There is:

- main thread
- server thread, if hosting
- sound thread (depending on sound system)

AIs also run in the main thread but they can make their own threads.

imbaczek, need not really a benchmark for it, but rather a reporting of the results of CPUID instruction.

Hmm, how about if Spring made a record of average technical details for each game:

-CPU load
-Physical memory % usage
-Frames per second
-Latency

then sent the information to a central server (or simply stored it on the local machine). A trailing-average formula could be created to combine those factors into essentially a benchmark score. All without adding much of anything new... I'd anticipate that since at least #1, 3, 4 are already displayed and thus are already in the Spring engine, it wouldn't be much more difficult/resource-intensive to track them over the course of the game than it is to track things like metal produced etc.

Felix the Cat wrote:Hmm, how about if Spring made a record of average technical details for each game:

-CPU load
-Physical memory % usage
-Frames per second
-Latency

then sent the information to a central server (or simply stored it on the local machine). A trailing-average formula could be created to combine those factors into essentially a benchmark score. All without adding much of anything new... I'd anticipate that since at least #1, 3, 4 are already displayed and thus are already in the Spring engine, it wouldn't be much more difficult/resource-intensive to track them over the course of the game than it is to track things like metal produced etc.

That sounds like its really over-complicating it. There are already quite a few open-source benchmarking programs, it shouldnt be too hard. Have something simple (but not half-assed and useless, like calculating PI) run the first time the game is installed (or when a user wants to run it), save the results to a file, and rank it on a 1 to 10 scale, then have spring display that number next to the user, rather than displaying the CPUs clock-frequency. Network latency and throughput is another thing that probably should be measured, though that might be too much trouble.

Wait.

Having Spring record things that it already displays for the user is overcomplicated... compared to implementing and calibrating a completely new benchmarking routine?

You phail at logics.

Having Spring

1) Store lots of data for each game
2) Send the data to a central server
3) Combine all the data with a complex formula into a single benchmark number for each user

is more complex than having Spring

1) Run a small benchmark application every once in a while

I think we should all send in 5 bucks and hire a private detective who finds out the exact specs of everyone who plays spring, and also takes embarrassing pictures and display them next to your name instead of your cpu.

hehe :)

I like the idea with spring reporting performance of last game..

It could be something as simple as average cpu usage or average FPS (average FPS sounds best to me and also very simple to implement)

It could simply dump result into log on game end, and lobby could read it/average it and use it as a base for some performance metric.

It would be best, because it takes spring settings into account too..