CHEF-FP examples repository

NOTE: Some of these scripts and commands might take upto several minutes to finish (depending on your machine).

This repository contains benchmarks for CHEF-FP comparing it against the state of the art ADAPT-FP tool.

Building a docker image for setting up CHEF-FP

To build the docker image run:

docker build -t "cheffp:Dockerfile" .

This installs CHEF-FP, ADAPT-FP along with their dependencies.

Using the docker image

Run the following command to open a bash terminal in a container:

docker run -it --rm cheffp:Dockerfile /bin/bash

Run the CHEF-FP vs ADAPT benchmarks in the docker image

To run the benchmarks, run the docker image and execute:

cd /code/chef-fp-examples
./runner.sh

The results of the benchmark will be in the /code/results.txt file.

To run the benchmarks with CHEF-FP's error reports, run:

cd /code/chef-fp-examples
./runChefResults.sh

Using CHEF-FP for mixed precision analysis

To follow this tutorial, you can either use the docker image or install clad and build the PrintModel on your system.

Lets take the following code as an example:

double f(double x) {
    double y = x * x;
    return y;
}

int main() {
    f(2.3);
}

Include the ErrorFunc.h header from the PrintModel/ directory. Then, mark the function you want to run error estimation on using clad::estimate_error and run the generated function using execute. The following example uses the error estimation model defined in the paper allowing us to print the mixed precision analysis results through clad::printErrorReport():

#include "clad/Differentiator/Differentiator.h"
#include "ErrorFunc.h"

double f(double x) {
    double y = x * x;
    return y;
}

int main() {
    // Generate the floating point error estimation code for 'f'
    auto df = clad::estimate_error(f);

    // Declare the necessary variables and call the generated code
    double dx = 0, final_error = 0;
    df.execute(2.3, &dx, final_error);

    // Print the mixed precision configuration
    clad::printErrorReport();

    // Optionally print clad's estimated error 
    // std::cout << final_error;
}

To compile use the following command:

clang++-13 \
        -Xclang -add-plugin -Xclang clad -Xclang -load -Xclang clad.so \
        -Xclang -plugin-arg-clad -Xclang -fcustom-estimation-model \
        -Xclang -plugin-arg-clad -Xclang /code/chef-fp-examples/PrintModel/libPrintModel.so \
        -lstdc++ -lm -std=c++11 -O3 \
        source.cpp

And run it:

./a.out

Create your own custom models

To create your own custom models you can refer /code/clad/demos/ErrorEstimation/CustomModel/README.md in the docker container or it's online version