Fix bug with more load on process than it should be about xray-16 HOT 8 CLOSED

The original code was

if (p_cnt < (nWorkers * 64))
    nWorkers = 1;

Currently, with that fix, code looks like this:

if (p_cnt < (nWorkers * 20))
    nWorkers = 1;

I found in r8384 (leaked COP svn revision ~March 2012) this code:

if (p_cnt < nWorkers)
   nWorkers = p_cnt;

Which code is better and what does it do?

from xray-16.

@Xottab-DUTY
Hi. I don`t know but on my old computer with Core 2 Duo E7300 code "if (p_cnt < (nWorkers * 20))" provides 100% load and particles bugs(it`s look like rainbows instead anomalies) instead "if (p_cnt < (nWorkers * 64))"

Code "if (p_cnt < (nWorkers * 64))" i found in sources of release engine version of COP

Code "if (p_cnt < nWorkers) = p_cnt;" i did not see and don`t tried to use

from xray-16.

Now i noticed that this change is not in the dev and another branches of project. But why?

from xray-16.

@nitrocaster fixed this.. Maybe this commit was occasionally dropped.
So, we need to find, what solution is better

if (p_cnt < (nWorkers * 64))
    nWorkers = 1;

or

if (p_cnt < nWorkers)
   nWorkers = p_cnt;

from xray-16.

CPU usage is 90% on my 6-core with:

if (p_cnt < nWorkers)
nWorkers = p_cnt;

if (p_cnt < (nWorkers * 64))
nWorkers = 1;

Is how it is in Shadow of Chernobyl and Clear Sky source and does seem to run better then * 20. Only 20% CPU usage.

I downloaded CoP revision 8384. It seems to be much older then whatever Nitro used as the base for OpenXray.

from xray-16.

Oh, thank you!
As far, as I know, nitrocaster used 1.6.0.2 and r8384 should be newer... Looks like I catch some bit of a nonsense here. =)
But whatever, * 64 seems to be a better solution.

P.S. by the way, @ViHtarb says that (nWorkers * 20) gives graphical bugs. @revolucas what about graphics look with if (p_cnt < nWorkers) nWorkers = p_cnt; variant? Just interesting)

from xray-16.

I didn't notice any graphical differences, only lag when using if (p_cnt < nWorkers) nWorkers = p_cnt;

I don't know about r8384. It's strange as it has many slight differences, almost like it's sabotage to make it run as shitty as possible.

from xray-16.

In theory if (p_cnt < nWorkers) nWorkers = p_cnt; should increase performance because it creates more workers if there's many particles.. But workers themselves needed to be optimized to take effect of multiplying them... Well.. all the code needs optimizing I'd say))

from xray-16.