Everything in place to start optimizing.
The Cloud Console, the DOCC compiler, benchmarking on real hardware and your Performance Database are ready when you sign in. Connect your code and see your first results today.
Four parts, one loop.
Support for the languages you love.
The docc compiler meets your code where it is. Decorate a NumPy-style kernel, route a PyTorch module through torch.compile, or compile a C/C++ translation unit like you would with clang. Same compiler, same IR, no rewrites.
1from docc.compiler import native23@docc.native(target="sequential")4def kernel(TSTEPS, A, B):5 for t in range(1, TSTEPS):6 B[1:-1, 1:-1] = 0.2 * (7 A[1:-1, 1:-1] + A[1:-1, :-2] + A[1:-1, 2:] + A[2:, 1:-1] + A[:-2, 1:-1]8 )9 A[1:-1, 1:-1] = 0.2 * (10 B[1:-1, 1:-1] + B[1:-1, :-2] + B[1:-1, 2:] + B[2:, 1:-1] + B[:-2, 1:-1]11 )
Measure performance on every commit.
Every push carries your workloads straight from GitHub into the performance dashboard of our Cloud Console. We run them on real hardware, surface the hotspots, and place each part of your code on the roofline of the hardware, so you see where time goes and what limits it.
daisytuner/modelsmain · 6d6dbef
Performance dashboard2 workloads · todayLivepytorch_models2 workloadstorch—78.0 ms+3.0%torch-cuda—17.9 ms−7.5%docc-cuda93 regions11.2 ms1.60×+0.0%docc-omp45 regions41.5 ms1.88×−1.8%torch-cuda—36.4 ms−3.4%docc-cuda266 regions21.7 ms1.68×+0.6%Performance optimizations, reviewed like code.
Our compiler, docc, represents every code as a dataflow IR that makes data movement, parallelism and computation explicit. In the web IDE of our Cloud Console you compose optimization recipes yourself. The compiler checks each recipe for legality, and you're ready to open the result as a pull request.
[segformer] fused gemm and div
MergedServerCUDAA live database, queried by docc.
Your performance database stores every optimization recipe, and docc matches them against your code automatically during compilation. Purpose-built performance embeddings find similar regions and the ideal optimization for yours. A RAG method for compilers.
segformer.fused_gemm_divpull request · mergedCUDAinterchange ×2 · CUDA(32) · nested(8, 4)polyhedral · proven2.14×9%3 · transfer-tunableThree steps to your first results.
- 01Sign inOpen the Cloud Console. Runners, the web IDE and your database are already set up.Open the Cloud Console →
- 02Connect your codeLink a GitHub repository, or install docc locally with one command.Install docc →
- 03Push and optimizeEvery push is compiled, benchmarked on real hardware and ready to optimize in the browser.Read the docs →