Posit-LLVM

A pass for rewriting arithmetic operations between doubles either as operations between floats or 32-bit posits. Arithmetic operations are rewritten inline using LLVM. It's for LLVM 3.8.

Build:

$ cd cs6120-project3
$ mkdir build
$ cd build
$ cmake ..
$ make
$ cd ..
$ make install

Run:

To compile and run the C program `"src".c` with doubles replaced with floats for arithmetic operations, run:

$ make float_run SRC="src"

To compile and run `"src".c` with arithmetic operations on doubles replaced with the associated 32-bit posit arithmetic operations, run:

$ make posit_run SRC="src"

Output LLVM:

To write the LLVM code resulting from the posit LLVM pass, run:

$ make llvm_posit SRC=`src`

Similarly, to write the float LLVM pass, run:

$ make llvm_float SRC=`src`

Run:

$ clang -Xclang -load -Xclang build/posit/libPositPass.* foo.c