cMeta automations to run programs in a portable way

[ Back to projects ]
[ Back ]

Every program artifact: a benchmark or application described once and compiled and run on any compute target (cpu, cuda, ...) by cx program run <name> <target>, the tools it needs set up on the way.

The commands:

Install the repository once (cx repo get cTuningLabs@cmeta-aops) and every row below is a command on your machine. The category itself - its commands and API - is here.

ArtifactComes fromWhat it is forTags
build-llama-cpp (cTuningLabs@cmeta-aops) Build llama.cpp from source for the selected compute target and run it on a GGUF model with a prompt dataset.
build-pytorch (cTuningLabs@cmeta-aops) Build PyTorch from source through the program pipeline (Python, compilers, CMake) for the selected target.
build-pytorchvision (cTuningLabs@cmeta-aops) Build torchvision from source against a locally built PyTorch.
build-torch-cpp (cTuningLabs@cmeta-aops) Build LibTorch, the PyTorch C++ distribution, from source with CMake and Ninja.
cbench-automotive-susan (cTuningLabs@cmeta-aops) cBench automotive susan image-processing benchmark (edges, corners, smoothing) on a PGM image dataset; runs on CPU and Android.
image-classification-onnx (cTuningLabs@cmeta-aops) Image classification with ONNX Runtime in Python on CPU, CUDA or XPU, with optional profiling.
image-classification-pytorch (cTuningLabs@cmeta-aops) Image classification with a pretrained PyTorch model in Python on the selected target.
lib-milepost (cTuningLabs@cmeta-aops) Build the MILEPOST codelet support library used by the MILEPOST benchmarks.
lib-polybench (cTuningLabs@cmeta-aops) Build the PolyBench support library used by the PolyBench kernels.
lib-xopenme (cTuningLabs@cmeta-aops) Build the xOpenME instrumentation library that records run-time statistics of benchmarks.
llama-cpp (cTuningLabs@cmeta-aops) Run a GGUF language model with a prebuilt or cached llama.cpp on a prompt dataset.
milepost-codelet-mibench-automotive-susan-e-src-susan-codelet-10-1 (cTuningLabs@cmeta-aops) A MILEPOST codelet extracted from MiBench susan edges (loop 10-1): a small kernel for compiler and hardware studies, on CPU and Android.
model-cnn-ylecun-mnist-pytorch (cTuningLabs@cmeta-aops) Train and run LeCun's CNN on MNIST with PyTorch on CPU or CUDA, with optional profiling.
polybench-cpu-gemm (cTuningLabs@cmeta-aops) PolyBench GEMM kernel (dense matrix multiplication) on CPU.
polybench-gemm-cpu-cuda (cTuningLabs@cmeta-aops) PolyBench GEMM kernel on CPU or CUDA with a configurable matrix size.
template-c-cpu (cTuningLabs@cmeta-aops) The base recipe every program inherits: select the target and compiler, compile the sources, run with datasets and models, collect outputs.
test-hello-c-cpu (cTuningLabs@cmeta-aops) Minimal hello-world C program that writes a statistics file; the smallest end-to-end check of the program pipeline.
test-nmm-c-cpu (cTuningLabs@cmeta-aops) Native matrix multiplication in C with OpenMP, OpenSSL and math libraries; configurable precision, sizes and repeats.
test-nmm-cpp-cpu (cTuningLabs@cmeta-aops) Native matrix multiplication in C++ with OpenMP, OpenSSL and math libraries; configurable precision, sizes and repeats.
test-nmm-go-cpu (cTuningLabs@cmeta-aops) Native matrix multiplication in Go; configurable precision, sizes and repeats.
test-nmm-java-cpu (cTuningLabs@cmeta-aops) Native matrix multiplication in Java, compiled with javac and run with java.
test-nmm-mojo-cpu (cTuningLabs@cmeta-aops) Native matrix multiplication in Mojo on CPU.
test-nmm-nvcc-cuda (cTuningLabs@cmeta-aops) Native matrix multiplication in CUDA compiled with nvcc; large default sizes for GPU timing.
test-nmm-objc-metal (cTuningLabs@cmeta-aops) Native matrix multiplication in Objective-C with Metal compute on Apple GPUs.
test-nmm-python-cpu (cTuningLabs@cmeta-aops) Native matrix multiplication in Python on CPU, with optional profiling.
test-nmm-rust-cpu (cTuningLabs@cmeta-aops) Native matrix multiplication in Rust on CPU.
test-nmm-swift-cpu (cTuningLabs@cmeta-aops) Native matrix multiplication in Swift on CPU.
test-nmm-torch-cpp (cTuningLabs@cmeta-aops) Matrix multiplication in C++ against LibTorch, the PyTorch C++ API.
test-pytorch-with-audio (cTuningLabs@cmeta-aops) Check PyTorch with torchaudio in Python on the selected target, with optional profiling.
test-pytorch-with-vision (cTuningLabs@cmeta-aops) Check PyTorch with torchvision in Python on the selected target (a C compiler is set up for extensions).
test-vllm (cTuningLabs@cmeta-aops) Start vLLM from Python on CPU or CUDA with a pinned or default version, with optional profiling.

Read live from the cMeta index of this server - the same data cx category list shows you locally.