Cheat Engine

crtzrms · Newbie cheater Reputation: 0 Joined: 26 Apr 2020 Posts: 13

I've been using the first version of ultimap to write some really cool hacks this week but i moved on to a game that becomes pretty much unplayable using it, so i decided to give ultimap 2 another try. Turns out, performance while capturing events is MUCH better, but filtering process is obnoxiously slower, so i was looking for ways to improve performance, but i also have some concerns:

1- Does ultimap 2 uses any file as buffer? If so, where is this file? My concern is that my OS is on a SSD and if CE uses TEMP or any other system folder, it might kill my SSD real soon and that would suck.

2- I have 16GB RAM, but so far i've been using 16384KB as buffer per CPU; should i crank this up to hit my memory limits? I'm running on an I7 so i think it could be reasonable to have a buffer up to 1.3GB/core (total 10GB buffer) but since i don't know where CE is storing this information, i've been hesitant to mess with this configuration for now.

3- So far i've been using usermode. I'm not entirely sure if kernelmode would improve performance.

4- "Do not trigger interrupts when full" option seems to crash cheat engine and does not prevent BSOD sometimes; CE drivers are properly loaded tho, so i'm not entirely sure why this is happening.

5- I've been using "Process data while recording" option, and so far it seems to let me play the game with minor performance impact, but then again, since i don't know if CE will try to save a file somewhere on the SSD if i change this option, i didn't even try yet; If it uses my RAM only, it should provide a nice boost in itself.

6- Adding ranges doesn't seem to help all that much with performance (and usually gives me a BSOD), should i be more restrictive about the ranges i put?

7- Filters (Even when there are only 30.000 points being covered) take a whole eternity to process for some reason, there could be any specific configuration i'm missing?

As a final note, i've been using "Pause target while processing" because it sounds optimal, but leaving it off doesn't seem to have any major impact on performance. Another important note is that im working on a game running under an emulator (PCSX2).

Thanks in advance.

Dark Byte · Posted: Mon May 04, 2020 1:46 am Post subject:

1: If you use the default options the radiobutton "Process data later" will be set, and the folder will be set to d:\lotsofspace

So if you didn't change it, the results will be in d:\lotsofspace

2: The buffersize will only affect the speed inside the game itself as there will be less pauses while saving to disk, but the processing will still take a while

3: It won't. By not checking kernelmode you are not watching kernelmode code from being executed, which wouldn't contain anything useful anyhow

4: That only has an effect when using ranges. If empty not sure what will happen.

5: Just RAM

6: Did you let it load DBVM or did you load DBVM first? If not, you will BSOD yes. Also, what Mainboard do you use and are you using 7.1?
But since you're mentioning an emulator, ranges won't help you as the emulator likely JIT's the code

7: The filters have to parse the profiling stream and then go through it with the disassembler. In that way it is a bit slow yes

Also, what BSOD are you getting. A memorydump may help

crtzrms · Newbie cheater Reputation: 0 Joined: 26 Apr 2020 Posts: 13

Dark Byte · Posted: Thu May 07, 2020 2:59 am Post subject:

2:
It's not the process, the cpu will generate a log entry for every jump and and writes it to memory. Once that memory is full it has to be written to disk. And during that time the process will be suspended

6:
Freeze is DBVM likely not supported on your system
PROFILER_CONFIGURATION_ILLEGAL is using ranges but DBVM not loaded

but yeah, not useful for your situation right now

---

You're not wrong, it's just not that fast. Not sure if the affinity has an effect. Every CPU generates it's own log, but because it's a continuous stream only one thread can process the cpu log generated. So theoretically having more cores log should be faster(more threads processing the data), BUT windows sometimes assigns some processes to a specific cpu core itself, so most of the code might be executed on one core anyhow generating only a log for one cpu

crtzrms · Newbie cheater Reputation: 0 Joined: 26 Apr 2020 Posts: 13

Dark Byte · Posted: Thu May 07, 2020 2:27 pm Post subject:

In processdata yes, but the heavy lifting is mostly done in the intel process trace decoder library (itp dll)

crtzrms · Newbie cheater Reputation: 0 Joined: 26 Apr 2020 Posts: 13

Do you think its a pointless effort to try to improve the code? Is the dll key to it all?

Whenever i have time ill put lazarus here and try to see if i can work something out, if possible.