ladystarbreeze / mina-isa Goto Github PK

Writing LLVM and/or GCC backends enables people to write code targeting MINA in languages other than assembly.

Generally a helpful instruction to have as, unlike a branch, it doesn't have any penalties or prediction requirements as it doesn't modify the instruction pointer. Good for producing branch-free output, and directly maps to a common compiler operation (LLVM has a SELECT instruction in it's IR that exactly matches ISEL/FSEL semantics, GCC has conditional execution in it's RTL)
If/when vectors are introduced, a vector version of this should ideally use a bitmask and select on a per-value basis.

Integer Select
ISEL rD, rS1, rS2
If T == 1, rS1 is moved to rD, else rS2 is moved to rD.

Float Select
FSEL fD, fS1, fS2
if T == 1, fS1 is moved to fD, else fS2 is moved to rD.

The current MINA32 ISA overview is a pretty basic Markdown file. Adding a proper document in LaTeX and/or enhancing the current manual would make it much less painful to look at.

There is still no reference interpreter for the MINA32 base integer ISA. Adding one will be helpful for people interested in running MINA code on software.

There is no README yet. Adding one would give people a brief idea of what MINA is.

Are there any architectural guarantees regarding self-modifying code or is this implementation defined? (e.g. separate i-cache/d-cache, length of pipieline / guaranteed pipeline flushing, icache flush instructions, &c.)