the initial code has been adapted

Thanks, worked

the input matrix should be const-qualified in C:

const int *const *matrix
// or better yet, to emphasize that it is an array:
const int *const matrix[];

the input is a matrix: int **, i.e. an array of 8-byte pointers to 4-byte int. [rdi] gives you the (address of the) first row of the matrix, not the first element of the first row

I dont get where array pointer in NASM.
; <-- RAX sumofmin(ro [qword [dword RDI]] nums, RSI rows, RDX cols) -->
So [rdi] should be first item in array?
I even did some function to print values in STDERR and everything worked fine but for [rdi].
Code i did and even checked on my machine works and would here too i believe if i could get this array.

Lua translation!

Fixed.

No random tests in Swift

Insufficient info, closing

Go Translation!

What language are you using? What bug have you experienced?

There's an bug in the initial solution.

For some reason, the random tests in the C# translation almost never fail, and you can return pretty much anything. I was able to submit a solution that has hard-coded return values for the three visible test cases, and just returns 0 for all other tests:

https://www.codewars.com/kata/reviews/5e5904021cff490001cdb237/groups/64ca1a6ca9a9ec0001fc2fb8

I'm still confused about arrays in C and assembly, I had some trouble with this.
In C it behaved like a standard 2D array, but in NASM I had to dereference rdi to get a pointer to the array. (I expected rdi to point to the first element in the array).
Is this what usually happens?

You're removing those numbers from the input, so, when the control function calculates its sum, it's using the other numbers: 136 + 128 + 83 = 347 Anyways, I've changed the tests so the expected value is computed first and your code messing with the input isn't a problem anymore, but be aware of that.