Home » SIRTECH CLASSICS » Jagged Alliance: Unfinished Business » Vanilla Modding » "Cool" patch...
| "Cool" patch...[message #102405]
|
Tue, 06 June 2006 00:07 
|
|
 mtx |
 |
Messages:10
Registered:December 2005
Location: Lisboa, Portugal |
|
|
Not exactly a source code project, but didn't know where to put it. The 1.13 forums don't deal with C code, and there isn't a public source code repository available...
I patched gameloop.cpp to avoid having ja2 keeping the CPU at 100%.
Problem: I usually play ja2 in a laptop, and hogging the CPU:
a) drains battery real quick
b) turns the laptop into a toaster. I mean HOT! :welder:
For desktop people: a cooler PC means less noise from the fans.
(btw, background tasks would get done faster if they could use the CPU, but that's not important)
Rationale: this game doesn't really need to have a huge framerate; 100Hz seems a good upper limit.
Solution: at each frame refresh, add a Sleep(SOMEMILLISECONDS) to pace the
refresh rate.
So:
--- gameloop.cpp.old 2006-06-02 19:24:15.000000000 +0100
+++ gameloop.cpp 2006-06-02 19:21:55.000000000 +0100
@@ -322,6 +322,8 @@
ReportMapscreenErrorLock();
}
#endif
+ //OK! 12% CPU:
+ Sleep(10);
//DebugMsg (TOPIC_JA2,DBG_LEVEL_3,"GameLoop done");
}
In my Asus (Centrino, 1.4GHz), there is no difference in gameplay, only in temperature and fan noise...
Tested with 1.13 (last November release).
If anyone tries it, please post here. Thank you.
Report message to a moderator
|
Private
|
|
|
|
|
|
|
|
| Re: "Cool" patch...[message #102408]
|
Wed, 07 June 2006 07:07 
|
|
| mtx |
|
Messages:10
Registered:December 2005
Location: Lisboa, Portugal |
|
|
Thanks for the suggestions.
Augustus:
I can try "V1.13 mod development". But I'm pleased to see that there are still interested readers here... And this patch works with the original sources (with some offset), not just 1.13.
I would really like to see it added under some form to 1.13, however.
defrog:
Your approach is undoubtedly more robust, but I don't know if it's worth the effort. I have a Pentium 200MMX and will try to install 1.13 to test it (it does play JA2, UB, UC well...)
However, I'm pretty confident that sleeping 10ms between screen updates - only done at the end of current computations - will not impact that much the playability, even in such an ancient CPU.
In my Asus, I can downscale CPU speed to 600MHz... And see no difference in the game.
But to feel more comfortable, let's replace the 10ms by 1. I never experimented with sleep(0), but will try; just yielding the current thread could just do the trick! Assuming windows would do it,
of course...
The whole point is to avoid having thousands of cycles per second doing mostly nothing in a busy loop...
The are better and cleverer ways of turning JA2 into a better windows citizen, no doubt. For example, turn gameloop() into an event-driven loop. This one has, however, simplicity on its side 
I would, of course, appreciate to have any feedback from testing in other setups.
PS: it was tested under Linux+Wine, and works perfectly as well.
Report message to a moderator
|
Private
|
|
|
|
| Re: "Cool" patch...[message #130588]
|
Fri, 23 February 2007 16:44
|
|
| mtx |
|
Messages:10
Registered:December 2005
Location: Lisboa, Portugal |
|
|
defrog:
I see your patch is merged in r725. Nice! Can't wait to
test it.
Meanwhile, I made some experiments with that windows Sleep()
call... It's trickier than I imagined, after reading some
docs. Examples (as I understood them):
- Sleep(0) isn't a noop! It essentialy does a thread yield.
Not much difference if you are not running other programs,
it seems.
- Sleep(1) is special! It allows window repainting and
window message processing. In my tests, using this in
the loop, although CPU usage remains high, it seems to
do away with most "motion chopping" and sound stutters
with slow CPUs (or using wine+linux...)
Your patch looks very good (and clever, I should add).
I'm eager to test this new option, especially under Wine/Linux (the most stringent test).
The obvious problem w/mine was that sometimes there are
calculations being done, even if the screen isn't changing
(e.g. enemy moves); in that case, it was just slowing
things down.
Sleep() should really only be invoked when nothing has
to done right now (e.g., user Paused the game), using
some global "workPending" flag or sth. similar... But
your ideia of using clock() is a very good approach to
that, and doesn't imply adding flag checking/toggling
everywhere. Yeah!
Report message to a moderator
|
Private
|
|
|
|
|
|
| Re: "Cool" patch...[message #130606]
|
Fri, 23 February 2007 22:07
|
|
| Sergeant_Kolja |
 |
Messages:42
Registered:July 2003
Location: Hannover, N-Germany |
|
|
defrogclock_t startTime = clock();
IMHO, GetTickCount() is a bit more efficient.
But Your idea to only free each remaining part of a 33fps slice is optimal!
MTX Sleep(0) isn't a noop! It essentialy does a thread yield.
Not much difference if you are not running other programs,
it seems.
Yes. Sleep(0) only 'stops' the current task of the current process (JA2.EXE), while all other tasks of the current process (JA2.EXE) inherit the freed remainder of the assigned time slice. Thus, all other processes (not JA2.EXE) are absolutely unchanged, if one of the remaining JA2 threads is also a time-eating one. Only, and only if, all the remaining JA2 threads are in so called waitable states (i.e. WaitForSingleObject() or more Sleep()), the OS gets the remainder of the time slice back and non JA2-processes or idle process can get this time.
MTXSleep(1) is special! It allows window repainting and window message processing. In my tests, using this in the loop, although CPU usage remains high, it seems to do away with most "motion chopping" and sound stutters with slow CPUs (or using wine+linux...)
Not special. It must depend on CPU power. Sleep(x) with x>0 says to the OS: please dont call this thread again before x ms are elapsed. IIRC, all other threads of the current process are running normal. Not inheriting the remainder of the slice.
I already experimented witch Sleep(1) right after the call of GameLoop(); It made my laptop slightly save energy, while Sleep(10) made the sound chopping (have very cheap AC'97 onboard sound in a dell inspiron 1300).
The defrog idea could be the best compromise.
But I'll try sometimes to further optimize it. The amount of Sleep'd time should also depend on BogoMips or CPU clock, because some people still using JA2 on 200 MHZ P5MMX and/or have CPU consuming sound drivers.
Report message to a moderator
|
|
|
|
|
|
|
|
|
| Re: "Cool" patch...[message #138866]
|
Tue, 15 May 2007 21:28
|
|
| mtx |
|
Messages:10
Registered:December 2005
Location: Lisboa, Portugal |
|
|
Sergeant_KoljaIs anybody still working on the gameloop sleep(1) issue?
I ask because I have an autocalibrating 30fps guaranteed patch in development, which looks fine here (CPU usage worst case 70%).
Not sure what you mean here; defrog's patch is in mailine 1.13 for quite some time, and handles this quite well, as far as I could test it.
Sergeant_Kolja
Would be hard to mix 2 patches together, so please contact me.
If you want to save a little more CPU cycles at the expense of a (unnoticeable?) delay in, e.g. "interface swooshes", you CAN mix the two...
That is, add a test after the calculations to always Sleep() at least 1ms (well, not exactly, check MSDN docs for details).
Personally, after defrog's work, I won't give it more thought. If you want to see a really "cool" (ahem, temperature wise ) JA2, check ja2-stracciatella - a port of vanilla JA2 to SDL. W/that, you can play JA2 in your laptop without worrying about batteries! 
Report message to a moderator
|
Private
|
|
|
|
Goto Forum:
Current Time: Tue Apr 16 06:56:34 GMT+3 2024
Total time taken to generate the page: 0.01641 seconds
|
Blog
Search
Help
Members
Register
Login
Home
