Cheat Engine

sgsgwv$6263 · Posted: Fri Nov 24, 2023 8:09 am Post subject: Something wrong with my code

Can anyone tell me whats wrong here:

ParkourPenguin · Posted: Fri Nov 24, 2023 12:55 pm Post subject:

The "Auto Assembler tutorials" forum is for tutorials, not help

Why are you using x87 instructions in 64-bit code? Use SSE instructions instead

You're not specifying data size when accessing memory. e.g. `fld [rbx+320]` could be accessing a float or a double. Use `dword ptr[...]` or `qword ptr[...]` for float or double respectively.

Those `fmul` instructions that square values could just be `fmul st(0),st(0)`

`fadd st(i)` is a little ambiguous as it could mean `fadd st(i),st(0)` or `fadd st(0),st(i)`. CE uses the latter.

You're not popping one of the values from the FPU stack. i.e. `fadd st(1)` should be just `faddp` (no arguments)

Why `fcomp` / `fstsw ax` / `sahf`? Do you know what the FPU status word and EFLAGS register are? `fcomip` is a thing that exists. The unordered version is probably more appropriate (i.e. `fucomip`)

`add esp,4` - no

You're modifying rax without backing it up

Don't mix SSE and x87 instructions. If the game is using SSE (i.e. `movss [...],xmm0`), you should too.

sgsgwv$6263 · Posted: Fri Nov 24, 2023 2:36 pm Post subject:

Okay. I can see that I am wrong on multiple fronts here. What I am doing is:

Calculating the magnitude of a vector whose x and y components are in rbx+320 and +324. And then comparing it with limit variable. If it is less than jump to speeder code label otherwise continue with original code label execution.

Can you help me write this in sse instructions?

I dont completely understand sse instructions because I dont know how they backup xmm registers.

Also,in comparison instructions of floats or double, its not as simple as integer types. I cant just

ParkourPenguin · Posted: Fri Nov 24, 2023 3:15 pm Post subject:

You can usually get away with not backing them up. Most XMM registers won't be in use at most injection points. You can look at the injection point to get a good idea. Include the comment at the bottom of the script that shows the code around the injection point.

If you need to back up xmm registers, use `movups` and put each register some place where there's 16 bytes available. The stack is the most convenient choice.

`ucomiss` / `ucomisd` sets EFLAGS- specifically ZF, PF, CF. Use equal, above, or below conditionals (e.g. jae, jne, jb, etc.)

sgsgwv$6263 · Posted: Fri Nov 24, 2023 4:04 pm Post subject:

Thanks a ton. That worked like a charm. You are a saviour.

Few more questions:
Why did you allocate 20 bytes in the stack when you said it requires 16 bytes per xmm register?

Why can I not use jg,jle,je etc after ucomiss but only those that you specified?

Can you share snippet to do the same compare using FPU syntax?

Thanks again in adavnce.

ParkourPenguin · Posted: Fri Nov 24, 2023 7:36 pm Post subject:

That's 0x20 bytes. 0x20 = 32 in decimal. 2 xmm registers, 16 bytes per register.

jg / jge / jl / jle check different flags in the EFLAGS register. ucomiss just sets those flags to 0 regardless of the result.

The x87 version is pretty much what you wrote but with the changes I previously mentioned.

sgsgwv$6263 · Posted: Sat Nov 25, 2023 10:42 pm Post subject:

Why does the game crash when I do:

ParkourPenguin · Posted: Sat Nov 25, 2023 10:51 pm Post subject:

That code is fine- it writes the float 0.01 to the address "myvar". Something else is wrong.

sgsgwv$6263 · Posted: Sun Nov 26, 2023 2:39 am Post subject:

In

ParkourPenguin · Posted: Sun Nov 26, 2023 12:25 pm Post subject:

sgsgwv$6263 · Posted: Wed Nov 29, 2023 9:06 am Post subject:

Thanks for the help so far.

Whats wrong here:

ParkourPenguin · Posted: Wed Nov 29, 2023 11:58 am Post subject:

sgsgwv$6263 · Posted: Wed Nov 29, 2023 12:24 pm Post subject:

Suprisingly min_h works fine. But if you intialize another such variable liks that the game crashes.