Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →GPU-accelerate Python code using CuPy, Numba CUDA, Warp, cuDF, cuML, cuGraph, KvikIO, cuCIM, cuxfilter, cuVS, cuSpatial, and RAFT. Use whenever the user mentions GPU/CUDA/NVIDIA acceleration, or wants to speed up NumPy, pandas, scikit-learn, scikit-image, NetworkX, GeoPandas, or Faiss workloads. Covers physics simulation, differentiable rendering, mesh ray casting, particle systems (DEM/SPH/fluids), vector/similarity search, GPUDirect Storage file IO, interactive dashboards, geospatial analysis,
.claude/skills/mkurman-optimize-for-gpu/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -3% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 33% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -9% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-21 | ✗→✓ | ▲ Improved | -7% | 0% |
|-------------| | references/cupy.md | User has NumPy/SciPy code, or needs array operations on GPU | | references/numba.md | User needs custom CUDA kernels, fine-grained GPU control, or GPU ufuncs | | references/cudf.md | User has pandas code, or needs dataframe operations on GPU | | references/cuml.md | User has scikit-learn code, or needs ML training/inference/preprocessing on GPU | | references/cugraph.md | User has NetworkX code, or needs graph analytics on GPU | | references/warp.md | User needs GPU simulation, spatial computing, mesh/volume queries, differentiable programming, or robotics | | references/kvikio.md | User needs high-performance file IO to/from GPU, GPUDirect Storage, reading S3/HTTP to GPU, or Zarr on GPU | | references/cuxfilter.md | User wants GPU-accelerated interactive dashboards, cross-filtering, or EDA visualization | | references/cucim.md | User has scikit-image code, or needs image processing, digital pathology, or WSI reading on GPU | | references/cuvs.md | User needs vector search, nearest neighbors, similarity search, or RAG retrieval on GPU | | references/cuspatial.md | User has GeoPandas/shapely code, or needs spatial joins, distance calculations, or trajectory analysis on GPU | | references/raft.md | User needs sparse eigensolvers, device memory management, or multi-GPU primitives |
Read the specific reference before writing code — they contain detailed API patterns, optimization techniques, and pitfalls specific to each library.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 14,662 | 12,708 | -13% | 1 | 1 | 0% | 2,822 | 2,739 | -3% | 0 | 0 | — |
case-06 | fail→pass | 36,394 | 10,287 | -72% | 1 | 1 | 0% | 1,898 | 2,533 | +33% | 0 | 0 | — |
case-02 | pass→pass | 13,784 | 27,705 | +101% | 1 | 1 | 0% | 2,661 | 3,050 | +15% | 0 | 0 | — |
case-03 | pass→pass | 16,664 | 15,518 | -7% | 1 | 1 | 0% | 3,216 | 2,866 | -11% | 0 | 0 | — |
case-04 | pass→pass | 10,065 | 9,290 | -8% | 1 | 1 | 0% | 1,895 | 2,148 | +13% | 0 | 0 | — |
case-05 | pass→pass | 15,306 | 15,938 | +4% | 1 | 1 | 0% | 2,949 | 2,976 | +1% | 0 | 0 | — |
case-07 | pass→pass | 16,141 | 16,397 | +2% | 1 | 1 | 0% | 2,844 | 3,353 | +18% | 0 | 0 | — |
case-08 | pass→pass | 16,323 | 13,984 | -14% | 1 | 1 | 0% | 2,960 | 3,044 | +3% | 0 | 0 | — |
case-09 | pass→pass | 19,868 | 17,232 | -13% | 1 | 1 | 0% | 3,644 | 3,631 | -0% | 0 | 0 | — |
case-10 | pass→pass | 17,616 | 13,524 | -23% | 1 | 1 | 0% | 3,382 | 2,975 | -12% | 0 | 0 | — |
case-11 | fail→pass | 23,144 | 21,254 | -8% | 1 | 1 | 0% | 4,607 | 4,206 | -9% | 0 | 0 | — |
case-12 | pass→pass | 16,664 | 10,263 | -38% | 1 | 1 | 0% | 3,201 | 2,214 | -31% | 0 | 0 | — |
case-13 | pass→pass | 11,810 | 12,038 | +2% | 1 | 1 | 0% | 2,104 | 2,526 | +20% | 0 | 0 | — |
case-14 | pass→pass | 13,937 | 12,540 | -10% | 1 | 1 | 0% | 2,742 | 2,275 | -17% | 0 | 0 | — |
case-15 | pass→pass | 16,006 | 13,163 | -18% | 1 | 1 | 0% | 3,020 | 2,957 | -2% | 0 | 0 | — |
case-16 | pass→pass | 20,176 | 16,430 | -19% | 1 | 1 | 0% | 3,638 | 3,337 | -8% | 0 | 0 | — |
case-17 | fail→pass | 20,468 | 14,930 | -27% | 1 | 1 | 0% | 3,559 | 3,144 | -12% | 0 | 0 | — |
case-18 | pass→pass | 17,944 | 14,523 | -19% | 1 | 1 | 0% | 3,301 | 3,100 | -6% | 0 | 0 | — |
case-19 | pass→pass | 13,293 | 14,611 | +10% | 1 | 1 | 0% | 2,584 | 3,249 | +26% | 0 | 0 | — |
case-20 | fail→fail | 28,188 | 2,646 | -91% | 1 | 1 | 0% | 5,189 | 629 | -88% | 0 | 0 | — |
case-21 | fail→pass | 19,152 | 18,299 | -4% | 1 | 1 | 0% | 3,664 | 3,404 | -7% | 0 | 0 | — |
case-22 | pass→pass | 18,140 | 15,290 | -16% | 1 | 1 | 0% | 3,640 | 3,514 | -3% | 0 | 0 | — |
case-23 | pass→pass | 18,478 | 17,807 | -4% | 1 | 1 | 0% | 3,972 | 4,229 | +6% | 0 | 0 | — |
case-24 | pass→pass | 11,716 | 11,028 | -6% | 1 | 1 | 0% | 2,309 | 2,527 | +9% | 0 | 0 | — |
case-25 | pass→pass | 15,609 | 15,219 | -2% | 1 | 1 | 0% | 3,300 | 3,709 | +12% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 25 cases were attempted, and 24 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +20 percentage points is the difference between those two pass rates over the 24 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.