FGPU_Compiler

In this repository you can find the LLVM compiler for the FGPU inside a docker. With this you will be able to compile your openCL kernel into FGPU instructions. (for simulating and/or implementing)

Instructions

Install docker in your system.

    git clone this repository

    cd FGPU_Compiler/compiler

    docker build -t fgpu_compiler .

Alternatively you can pull an already built docker image (this haven't been tested enough times to make it the default option).

    docker pull munozher/fgpu_compiler

With the image already built, you can start it:

    cd ../kernels

    docker run -v $(pwd):/llvm/kernels -it fgpu_compiler 

NOTE: if you pulled the built image, the name of the image in the last command has to change from fgpu_compiler to munozher/fgpu_compiler

This will create a directory /llvm/kernels inside the container and mount the current directory over there, and then run the container into interactive mode (i.e., you get a shell inside the container).

    cd to llvm/kernels

    ./compile.sh or ./compile_and_log.sh 

NOTE: compile.sh outputs just the instructions for the FGPU while compile_and_log.sh will also provide an .objdump to see how the instructions were coded.

Later, its important to keep track of the docker instances because after created they will remain on your system forever unless deleted

You can see all the instances of docker with this line:

    docker container ls -a -s

Every time you do the docker run a new instance of docker will be created so you will need to delete the ones you already created to save space on your hardrive

    docker container rm *CONTID*

Replace CONTID with the ID from the list obtained in the previous line.

After compiling you can exit the docker instance with Ctrl + D.

The outputs can be found in kernels/compiler_outputs with the same name of your openCL kernel.