Skip to content


Repository files navigation

MIT licensed


Tiramisu is a compiler for expressing fast and portable data parallel computations. It provides a simple C++ API for expressing algorithms (Tiramisu expressions) and how these algorithms should be optimized by the compiler. Tiramisu can be used in areas such as linear and tensor algebra, deep learning, image processing, stencil computations and machine learning.

The Tiramisu compiler is based on the polyhedral model thus it can express a large set of loop optimizations and data layout transformations. Currently it targets (1) multicore X86 CPUs, (2) Nvidia GPUs, (3) Xilinx FPGAs (Vivado HLS) and (4) distributed machines (using MPI). It is designed to enable easy integration of code generators for new architectures.


The following is an example of a Tiramisu program specified using the C++ API.

// C++ code with a Tiramisu expression.
#include "tiramisu/tiramisu.h"
using namespace tiramisu;

void generate_code()
    // Specify the name of the function that you want to create.

    // Declare two iterator variables (i and j) such that 0<=i<100 and 0<=j<100.
    var i("i", 0, 100), j("j", 0, 100);

    // Declare a Tiramisu expression (algorithm) that is equivalent to the following C code
    // for (i=0; i<100; i++)
    //   for (j=0; j<100; j++)
    //     C(i,j) = 0;
    computation C({i,j}, 0);
    // Specify optimizations
    C.vectorize(j, 4);
    buffer b_C("b_C", {100, 100}, p_int32, a_output);

    // Generate code
    C.codegen({&b_C}, "generated_code.o");

Building Tiramisu from Sources

This section provides a description of how to build Tiramisu. The installation instructions below have been tested on Linux Ubuntu (18.04) and MacOS (13.0.1) but should work on other Linux and MacOS versions.


  1. CMake: version 3.22 or greater.

  2. Autoconf and libtool.

  3. Ninja.

  1. OpenMPI and OpenSSh: if you want to generate and run distributed code (MPI).

  2. CUDA Toolkit: if you want to generate and run CUDA code.

  3. Python 3.8 or higher if you want to use the python bindings. (Along with Pybind 2.10.2, Cython, and Numpy).

Build Methods

There are 3 ways to build Tiramisu:

  1. From spack, which will build everything from source for you.
  2. From source, but using system package managers for dependencies.
  3. Purely from source with our install script.

The last two only differ only in how they setup the dependenies.

Method 1: Build from spack

Install spack and then run:

spack install tiramisu

Method 2: Build from source but install dependencies using system package managers

There are two steps:

  1. Install the dependencies (either using Homebrew or using Apt).
  2. Use Cmake to build Tiramisu.
Install the dependencies
Install the dependencies using Homebrew

If you are on MacOS and using Homebrew, you can run the following commands to setup the dependencies:

brew install cmake
brew install llvm@14
brew install halide
brew install isl
brew link halide
brew link isl

If any of these ask you to update your path, do so. For example, using the following command, you can find the isl include and library directories:

brew info isl
Install the dependencies using Apt

If you are on Ubuntu/Debian, you can use apt to setup the dependencies:

chmod +x
sudo ./ 14 all
sudo apt-get install liblld-14-dev llvm-14-runtime
sudo apt-get install libllvm14 llvm-14-dev
sudo apt-get install llvm14-*
sudo apt-get install halide
sudo apt-get install libisl-dev

Using the following command, you can find the isl include and library directories:

dpkg -L libisl-dev
Building Tiramisu with cmake
  1. Get Tiramisu
git clone
cd tiramisu
mkdir build
  1. Setup the configure.cmake. In particular, choose if you want to use a GPU or MPI setup. Choose if you want to use the python bindings. Choose if you want to us the auto scheduler. You may need to add other options to support these.

  2. Configure:


If you want to install, add CMAKE_INSTALL_PREFIX. If you are installing the python bindings, add Tiramisu_INSTALL_PYTHONDIR to tell Tiramisu where to place a python package. You will need add these install locations to the relevant path variables such as PYTHONPATH and LD_LIBRARY_PATH.

  1. Build:
cmake --build build

You can also install if you want via cmake --install.

Method 3: Build from source, but install dependencies using our script

There are two steps:

  1. Install the dependencies using our script.
  2. Use Cmake to build Tiramisu.
Building Dependencies via Script
  1. Get Tiramisu
git clone
cd tiramisu
  1. Get and install Tiramisu submodules (ISL, LLVM and Halide). This step may take between few minutes to few hours (downloading and compiling LLVM is time consuming).
./utils/scripts/ <TIRAMISU_ROOT_DIR>
- Note: Make sure `<TIRAMISU_ROOT_DIR>` is absolute path!
  1. Optional: configure the tiramisu build by editing configure.cmake. Needed only if you want to generate MPI or GPU code, run the BLAS benchmarks, or if you want to build the autoscheduler module. A description of what each variable is and how it should be set is provided in comments in configure.cmake.

    • To use the GPU backend, set USE_GPU to TRUE. If the CUDA library is not found automatically while building Tiramisu, the user will be prompt to provide the path to the CUDA library.
    • To use the distributed backend, set USE_MPI to TRUE. If the MPI library is not found automatically, set the following variables: MPI_INCLUDE_DIR, MPI_LIB_DIR, and MPI_LIB_FLAGS.
    • To build the autoscheduler module, set USE_AUTO_SCHEDULER to TRUE.
  2. Add Halide's cmake to the CMAKE_PREFIX_PATH:

  1. Build the main Tiramisu library
mkdir build
cd build
cmake ..
cmake --build .
  1. If you want to build the autoscheduler module, set USE_AUTO_SCHEDULER to TRUE in configure.cmake, and after building Tiramisu :
make tiramisu_auto_scheduler

Old Tiramisu on a Virtual Machine

Users can use the Tiramisu virtual machine disk image. The image is created using virtual box (5.2.12) and has Tiramisu already pre-compiled and ready for use. It was compiled using the same instructions in this README file.

Once you download the image, unzip it and use virtual box to open the file 'TiramisuVM.vbox'.

Once the virtual machine has started, open a terminal, then go to the Tiramisu directory

cd /home/b/tiramisu/

If asked for a username/password


Getting Started

Run Tests

To run all the tests, assuming you are in the build/ directory

make test



To run only one test (test_01 for example)

ctest -R 01

This will compile and run the code generator and then the wrapper.

To view the output of a test pass the --verbose option to ctest.