Very good progress happened.

First of I've rewritten everything in VHDL, because the very first version of my processor was in VHDL and I feel much more confident with writing it than verilog.

As a second step I implemented a very basic BRAM module, tested most instructions (i just realized while typing this that I am still missing the shift instructions. Lazy me.) with a testbench and wired up the SoC.

Created a simple ROM too which is for the moment directly in the SoC module file and contains just 32 instructions max for the moment, BUT I got the very first program working looking like this:

load 0 into r0
load 1 into r1
load 16 into r2
load 0 into r3
load 5 into r5
add r0 = r0 + r1
set address (leds) 0x5 to r0
set lower than r3 = r1 < r2
branch when r3 != 0 --> 5 (to add instruction)
sub r0 = r0 - r1
set address (leds) 0x5 to r0
set lower than r3 = r0 < r1
branch when r3 = 0 --> 9 (sub instruction)
jump 0

This counts to 16 and back down to 0. And this works super nice! See embedded video:

Pretty happy about this. I think I will implement a very very simple assembler first, a nice ROM module next and then will see where it goes on. The updated VHDL code can be found in my gitlab repository