You are incorrect about using latin languages. The first computers took their input in the form of punched cards. in fact, punched cards were in use for programmable weaving machines long before computers were invented. I think its safe to say that your computer pioneers would invent a language to communicate with their computers, not build computers and inputs to suit the language. lets look at our own history of development for ideas:
First, if their computer is binary based (likely) then they'll invent a language of ones and zeros, or holes and not holes, or black and white, or ups and downs, or tones and silence. Perhaps several of these. its the easiest/only way to work with primitive machines.
The next stage is that the computer language no longer matches one-to-one with the program in memory. Instead, a sufficiently advanced program on the computer (a compiler), takes a more complicated (easier for humans to read and write) language and translates it into the binary language. Our history took Assembly, a set of names for "slots" or storage spaces (registers) in the computer, and short names for commands which work on them. Each instruction in the language and the arguments to the instruction are literally translated to ones and zeros, which directly work in the cpu.
Dealing with literal addresses in memory is pretty inflexible, so the concept of names and labels was introduced, with a tool called a linker. It assigns addresses and keeps track of them, meaning you can refer toeyour variables and numbers by name, not location.
You can see by this point that a few critical linguistic concepts are in play... verbs, describing actions on data, which is named. A further class of instruction involves decisions... if the result of something is zero, go to a different memory address(area of program) and start executing the instructions there, otherwise continue. This, more or less, is the full set of instructions you need to solve pretty much anything (aka being Turing complete). you will need your language and computer to be turing complete if you want it to be a proper, flexible computer, but further refinements to the commands and language are basically for convenience, ease of use, speed of execution, not really essential to solve problems.
Depending on how your language works, and how advanced the computers are, some kind of symbolic references for the instructions and names would be developed. perhaps its ok with one key for each instruction (typically you would have some tens of instructions in a simple computer), and some way to compose variable names (combinations of phonetic sounds?), using numbers as a worst case would be prevalent, much like the latin character set got itself firmly embedded in modern computers. I find it hard to believe their primitive computers would be required to support thousands of characters just because their language demands it, they'd likely be required to develop a reduced character set representation of their language (probably phonetic) as a part of their computer development.
It's been said that had we discovered binary mathematics and boolean logic sooner our computer development would have occured much sooner too. If your culture can not find a cut-down wayeto represent commands and variables it'll likely prevent them from developing computers.
Only much, much later would computers be developed enough to handle the natural language. look as what we have today, you can (in a limited sense) say what you want and have the computer speak the response back to you. That took 50 years of breakneck speed development to achieve. Natural language handling is an extremely advanced computer science field, the fundamental computer design is long, long established before the input matures.