tsne implements FIt-SNE algorithm for various GPU architectures (original CUDA source code is from here).

• CUDA: The original code was obtained from here
• SYCL: The CUDA code was migrated using Intel DPCT, and then the resulting code was modified to remove the dpct headers.
• HIP: Created from CUDA version using hipify-perl script.

For Intel GPU -
First, source icpx compiler. Then,

cd tsne/SYCL
mkdir build
cd build
CXX=icpx cmake -DGPU_AOT=pvc ..
make -sj

Note:

• To enable AOT compilation, please use the flag -DGPU_AOT=pvc for PVC.

For AMD GPU -
First source clang++ compiler. Then,

cd tsne/SYCL
mkdir build
cd build
CXX=clang++ cmake -DUSE_AMDHIP_BACKEND=gfx90a ..
make -sj

Note:

• We use the flag -DUSE_AMDHIP_BACKEND=gfx90a for MI250. Use the correct value for your GPU.

For NVIDIA GPU -
First source clang++ compiler. Then,

cd tsne/SYCL
mkdir build
cd build
CXX=clang++ cmake -DUSE_NVIDIA_BACKEND=YES -DUSE_SM=80 ..
make -sj

Note:

• We use the flag -DUSE_SM=80 for A100 or -DUSE_SM=90 for H100.

cd tsne/CUDA
mkdir build
cd build
CXX=nvcc cmake -DUSE_SM=80 ..
make -sj

Note:

• We use the flag -DUSE_SM=80 for A100 or -DUSE_SM=90 for H100.

cd tsne/HIP
mkdir build
cd build
CXX=hipcc cmake -DROCM_PATH=/opt/rocm ..
make -sj

After building, to run the workload, cd into the build folder. Then

./tsne

Output gives the total time for running the whole workload.