Arithmetic Expression Compiler

[FlatAssembler]

This is so far out of my wheelhouse that I couldn't comment intelligently on it.

Well, that's kind of how I feel when you talk about how Chrome is the only good browser because it is the only one that allows the users to publish voice-recordings to your website - I have no idea what you are talking about.

scripting old / small processors for use in robotics

Well, you know, the compiler algorithm I've made kind of breaks for ARM processors, it assumes the existence of a stack inside the processor (like the FPU stack in x86).
I mean, yeah, I suppose I could arrange the 15 general purpose registers (or even just a portion of them) in a stack and then "push" and "pop" the values using the macros (a decent assembly-language compiler, assuming such exists for ARM, would allow me to do that). The obvious problem with that is that it would lead to unreasonably large executables (but I don't think performance would be a problem, since ARM processors are known to move values between registers a lot faster than Intel or AMD processors).
Very few assembly-language compilers (I assume none of those available for ARM) would allow me to make statements such as:

Code:

mov [result],3.141592f

However, if I am using the Duktape framework, that actually wouldn't matter much because I could just invoke a C function from JavaScript that would do pointer aliasing to give me hexadecimal IEEE 754 code (and both x86 and ARM support little-endian). So, I could easily output something like:

Code:

mov r0,0x40490fd0 ;3.141592
mov [result],r0

The biggest problem is that ARM, being a RISC processor architecture, doesn't have anything like the x86 "fsin". That would lead to very different kinds of problems than what I had to solve when making an app that outputs x86 assembly (which has those instructions, but they often change the FPU stack in a counter-intuitive way). I could solve those problems by, for instance, making a complicated macro that would calculate trigonometric functions using factorials (assuming that can be done without corrupting the "processor stack", I am not sure if that's possible). Calculating logarithms and exponents will probably be just as problematic in ARM assembly as it is in x86 assembly.
Or, unless I am working on bare metal, I could simply temporarily "store" the stack somewhere in memory, request the kernel to do the complicated calculations for me, and then reload the stack into the processor registers. But, if I want to be consistent, I'd have to do that very often since, if I quickly look up online, the ARM assembly doesn't even have the equivalent of "fprem" (which I used quite often in my compiler).
Those are just the problems I can think of right now.

You would probably benefit from some formal education.

Why exactly? What do you think I am doing wrong that a formal education would help with?