tenferro-benchmark

Benchmark suite for tenferro-rs.

The repository keeps the latest human-facing reports per target profile:

• macOS CPU: result/mac-cpu/cpu/einsum.md, result/mac-cpu/cpu/cpu_ops.md, result/mac-cpu/cpu/linalg_jvp_vjp.md
• Linux/AMD CPU: result/amd-cpu/cpu/einsum.md, result/amd-cpu/cpu/cpu_ops.md, result/amd-cpu/cpu/linalg_jvp_vjp.md
• NVIDIA GPU: result/nvidia-gpu/gpu/dense.md, result/nvidia-gpu/gpu/einsum.md, result/nvidia-gpu/gpu/sparse.md, result/nvidia-gpu/gpu/tensornetwork.md

Historical reports are not archived in extra files. Use git history when older results are needed.

The GPU tensor network benchmark uses problem data from extern/TensorNetworkBenchmarks/, based on the upstream TensorNetworkBenchmarks repository. See GPU tensor network suite.

Workflows

• macOS CPU workflow: native run, no Docker, Accelerate.
• Linux CPU devcontainer workflow: Docker/devcontainer, OpenBLAS for tenferro, detected PyTorch provider.
• NVIDIA GPU devcontainer workflow: CUDA devcontainer.
• Einsum suite and instance selection: source benchmark, selection rules, diagnostic cases, path strategies.
• GPU tensor network suite: TensorNetworkBenchmarks parity on CUDA.
• Result layout and metadata: target_profile, suite_id, run.yaml, latest reports.
• Architecture terminology: suite, runner, backend, strategy, target profile.
• PyTorch einsum dispatch notes: PyTorch source investigation notes.

Quick Smoke

macOS:

uv sync
./scripts/setup_extern_deps.sh
BENCHMARK_TARGET_PROFILE=mac-cpu \
BENCH_INSTANCE=bin_matmul_256 \
BENCH_RUNS=1 \
BENCH_WARMUPS=0 \
PUBLICATION_GATE_SUITE=small \
  ./scripts/run_all.sh 1

Linux devcontainer from the host:

devcontainer up --workspace-folder .
devcontainer exec --workspace-folder . bash -lc '\
  BENCHMARK_TARGET_PROFILE=amd-cpu \
  BENCH_INSTANCE=bin_matmul_256 \
  BENCH_RUNS=1 \
  BENCH_WARMUPS=0 \
  PUBLICATION_GATE_SUITE=small \
    ./scripts/run_all.sh 1'

GPU devcontainer from the host:

devcontainer up --workspace-folder . --config .devcontainer/cuda/devcontainer.json
devcontainer exec --workspace-folder . --config .devcontainer/cuda/devcontainer.json \
  bash -lc 'BENCHMARK_TARGET_PROFILE=nvidia-gpu ./scripts/run_gpu_suite.sh'

Comparison Backends

CPU reports compare:

• tenferro-trace
• tenferro-eager
• pytorch-cpu
• jax-cpu

GPU reports compare:

• tenferro-cuda-trace
• tenferro-cuda-eager
• pytorch-cuda
• jax-cuda
• vendor-specific CUDA backends where meaningful

C++ Torch/LibTorch runners are intentionally removed. PyTorch Python is the ATen comparison backend. The PyTorch CPU provider is detected at run time and recorded in run.yaml and generated reports; Linux does not source-build PyTorch to force OpenBLAS.

Development Checks

Run these after changing benchmark scripts or schemas:

uv run python scripts/validate_benchmark_suite.py benchmarks/cpu/einsum.yaml
uv run python scripts/validate_benchmark_suite.py benchmarks/gpu/dense.yaml benchmarks/gpu/einsum.yaml benchmarks/gpu/sparse.yaml benchmarks/gpu/tensornetwork.yaml
bash tests/test_suite_result_layout.sh
bash tests/test_run_all_docs_outputs.sh
bash tests/test_clean_extern_deps.sh
bash tests/test_setup_extern_tenferro_checkout.sh
cmake -S cpp -B build/cpp-plan-test
cmake --build build/cpp-plan-test --target einsum_plan_test
ctest --test-dir build/cpp-plan-test --output-on-failure

License

MIT