FaaS Message Interface

Getting Started

C++

#include <fmi.h>
...
FMI::Communicator comm(peer_id, 3, config_path, comm_name);
FMI::Comm::Data<double> d;
FMI::Comm::Data<std::vector<int>> vec({0, 1, 2});
if (peer_id == 0) {
    d = 42.0;
}
FMI::Comm::Data<std::vector<int>> recv(1);
comm.bcast(d, 0);
comm.scatter(vec, recv, 0);
assert(d == 42.0);
assert(recv.get()[0] == peer_id);
FMI::Utils::Function<int> f([] (int a, int b) {return a + b;}, true, true);
FMI::Comm::Data<int> id = peer_id;
FMI::Comm::Data<int> id_sum;
comm.allreduce(id, id_sum, f);
assert(id_sum == 3);

Python

import fmi
...
comm = fmi.Communicator(peer_id, 3, config_path, comm_name)
if peer_id == 0:
    comm.bcast(42, 0, fmi.types(fmi.datatypes.int))
else:
    bc_res = comm.bcast(None, 0, fmi.types(fmi.datatypes.int))
    assert bc_res == 42
sc_res = comm.scatter([0, 1, 2], 0, fmi.types(fmi.datatypes.int_list, 3))
assert sc_res[0] == peer_id
id_sum = comm.allreduce(peer_id, fmi.func(fmi.op.sum),
                        fmi.types(fmi.datatypes.int))
assert id_sum == 3
id_custom = comm.allreduce(peer_id,
             fmi.func(fmi.op.custom, lambda a, b: 3 * a + b, False, False),
             fmi.types(fmi.datatypes.int))
assert id_custom == 5 # Left-to-right order because of non-assoc. / non-comm.

Installation / Compilation

C++

To facilitate compiling, there is a Docker image with all necessary dependencies:

docker pull ghcr.io/opencorech/fmi-build-cpp:latest

Python

AWS Lambda Layer

The simplest way to use the library are the AWS Lambda layers, available under the following ARNs:

• arn:aws:lambda:eu-central-1:386971375191:layer:fmi-python36:1
• arn:aws:lambda:eu-central-1:386971375191:layer:fmi-python37:1
• arn:aws:lambda:eu-central-1:386971375191:layer:fmi-python38:1
• arn:aws:lambda:eu-central-1:386971375191:layer:fmi-python39:1

If you are using CloudFormation, you simply have to reference the ARN with Layers:, a minimal example for Python 3.9 looks like this:

AWSTemplateFormatVersion: '2010-09-09'
Transform: 'AWS::Serverless-2016-10-31'
 
Resources:
  lambdaml:
    Type: AWS::Serverless::Function
    Properties:
      FunctionName: fmi-example
      Runtime: python3.9
      Layers:
        - arn:aws:lambda:eu-central-1:386971375191:layer:fmi-python39:1

When using the AWS Management Console, you can add it under Layers -> Add a layer -> Specify an ARN:

Compilation

To compile the library by yourself, there are preconfigured Docker images for all Python versions that contain the correct dependencies (FMI dependencies, Python headers, and the correct Boost.Python):

docker pull ghcr.io/opencorech/fmi-build-python39:latest