From a33ae4331d179d31b4800f24a7d99d328a6e080e Mon Sep 17 00:00:00 2001 From: "russell@unturf.com" Date: Fri, 5 Jun 2026 13:18:06 -0400 Subject: [PATCH] =?UTF-8?q?bend=20catalog=20Wave=203=20=E2=80=94=2015=20su?= =?UTF-8?q?rveyed=20forms=20(2026-06-05)?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Adds 15 new GPU primitive forms across domains not covered by Waves 1-2: ZK/SNARK provers, pairing crypto, tensor network contraction, sparse LA, CV primitives, numerical solvers, generic belief propagation, MD/CFD, convex opt, DSP beyond cuFFT, DB aggregations, graph theory beyond triangle/PageRank. Top hits by relevance to our missions: * cuda-mfcc-spectral (97x) — punters-cc, BT-DISC forensics, real-time CC pipeline. unsandbox HIGH. * cuda-batched-lp-simplex (95x over CPLEX) — Prime Mission workstation-to-queue balancing. unsandbox HIGH. * cuda-betweenness-centrality-weighted (30-150x) — workaholic-node detection on dependency DAG, MOAD-0001 model. undefect HIGH. * cuda-hungarian-lap (400M-var LAP ~13s) — workstation balancing. unsandbox HIGH. * cuda-msm-bls12-381 (27.86x) — Pippenger shares branchless modmul shape with our reversible secp256k1 inner loop. ECDSA HIGH. * cuda-tensornet-contract (tensor QR 100x vs Xeon 8480+) — alt to stabilizer/kickmix sim path; MPS/PEPS evaluates reversible secp256k1 circuits beyond Clifford. ECDSA HIGH. * cuda-loopy-bp-mrf (45x) — LBP substrate for FuzzingBrain probabilistic analysis. undefect HIGH. * cuda-ega-gpu-aggregation (6-29x) — defect-corpus aggregation at planetary scale; telemetry queue aggregation. undefect HIGH + unsandbox HIGH. Plus highlight numbers: FluidX3D LBM 100-200x vs ANSYS Fluent, CUDASW++4.0 5.71 TCUPS on H100 for sequence alignment, ICICLE-Snark the fastest Groth16 today. Wave 3 filter-outs noted: AMGX, GROMACS, NVOFA, junction-tree BP, batched L-BFGS — all below 10x or insufficiently general. CATALOG.md mirrors the table for in-repo reading; bend.html renders the public version with reference URLs. --- examples/cuda-fanout/CATALOG.md | 33 +++++++++++++++++++++++++++++++++ www/bend.html | 32 ++++++++++++++++++++++++++++++++ 2 files changed, 65 insertions(+) diff --git a/examples/cuda-fanout/CATALOG.md b/examples/cuda-fanout/CATALOG.md index 400fc06..e63507e 100644 --- a/examples/cuda-fanout/CATALOG.md +++ b/examples/cuda-fanout/CATALOG.md @@ -141,6 +141,39 @@ run on our fleet yet. - cuda-batched-mcts — 25-40x on Go-style rollouts; wrong shape for our search - cuda-faiss-ann — 5-12x; no embedding workload today +## Wave 3 — surveyed 2026-06-05 + +15 additional forms covering ZK/SNARK provers, pairing crypto, +tensor network contraction, sparse LA, CV primitives, numerical +solvers, generic BP, MD/CFD, convex opt, DSP beyond cuFFT, DB +aggregations, graph theory beyond triangle/PageRank. + +| form | speedup / throughput | hardware | relevance | +|-----------------------------------|------------------------------------------------|---------------------|-----------| +| cuda-fluidx3d-lbm | 100-200x vs ANSYS Fluent; 8799 MLUPS A100 | A100 | unsandbox MEDIUM (HPC reproducibility) | +| cuda-mfcc-spectral | ~97x CPU MFCC; STFT ~75x via cuSignal | GTX 580+ | unsandbox HIGH (punters-cc, BT-DISC forensics) | +| cuda-batched-lp-simplex | 95x over CPLEX; 5x over GLPK on 100K LPs | GTX 980-class | unsandbox HIGH (resource scheduling, Prime Mission) | +| cuda-betweenness-centrality-weighted | 30-150x warp-centric weighted BC | GTX+ | undefect HIGH (workaholic-node detection, MOAD-0001) | +| cuda-cudasift-orb-ransac | ~60x SIFT; 1.2ms on GTX 1060; ORB 11.3x | GTX 1060+ | unsandbox MEDIUM (visual evidence) | +| cuda-cudasw-gasal2 | CUDASW++4.0 16.2x; 5.71 TCUPS H100; GASAL2 750x | H100 | undefect MEDIUM (binary-diff patch-similarity) | +| cuda-loopy-bp-mrf | 45x over CPU LBP for stereo MRF | GTX 280+ | undefect HIGH (FuzzingBrain probabilistic analysis) | +| cuda-sgm-stereo | 42 fps @ 640x480, 128 disparities Tegra X1 | Tegra X1+ | unsandbox MEDIUM (edge node profile) | +| cuda-hungarian-lap | 400M-var LAP in ~13s; 10-50x class | NVIDIA | unsandbox HIGH (workstation-to-queue balancing); undefect MEDIUM | +| cuda-msm-bls12-381 | 27.86x over Pippenger RELIC AVX; 60% of Groth16 | A100 / 4090 | ECDSA HIGH (Pippenger shares branchless modmul shape) | +| cuda-pdwt-lifting | 15.9x over best CPU DWT | GTX / Tesla | unsandbox MEDIUM (audio-IPC analysis) | +| cuda-tensornet-contract | 8-20x vs CuPy; tensor QR ~100x vs Xeon 8480+ | A100 | ECDSA HIGH (alt to stabilizer/kickmix) | +| cuda-ega-gpu-aggregation | 6.45-29.12x multi-pass; group-by hash 19.4x | NVIDIA | undefect HIGH (corpus aggregation); unsandbox HIGH (telemetry) | +| cuda-bicgstab-ilu-spmv | SpTRSV 10.7x; ILU0 BiCGSTAB 3.2x; GMRES 6.9x | V100 / MI210 | ECDSA MEDIUM (sparse LA over GF(p)); undefect MEDIUM | +| cuda-icicle-snark-groth16 | Fastest Groth16 today; Mina 3x libsnark | RTX 4090 / A100 | ECDSA MEDIUM; undefect MEDIUM (zk-prover defect scanning) | + +### Wave 3 filtered out (below threshold) + +- AMGX algebraic multigrid: 2-5x — below 10x cutoff +- GROMACS GPU: 2-3x speedup vs CPU — below threshold +- NVIDIA Optical Flow SDK: 7-10x borderline, dedicated hardware unit +- Junction-tree BP per-message: 0.68-9.18x — below threshold +- Batched L-BFGS: 134x reported but single mesh-size case, not generalized + ## Build order — recommended 3-form runway 1. **A: `cuda-secp256k1-batched-mul`** — biggest immediate win. Drop VanitySearch's CUDA secp256k1 kernel as a BSHK worker; lumbda gets a 6.5 Gkeys/s point-mul oracle. Every candidate validation rides this form. Hardware already owned (3090 + 4090). diff --git a/www/bend.html b/www/bend.html index 6953f14..ffdbc8d 100644 --- a/www/bend.html +++ b/www/bend.html @@ -198,6 +198,38 @@ make gpu-worker LUMBDA=asm # smallest footprint +

Wave 3 — surveyed 2026-06-05

+

15 additional forms spanning ZK / SNARK provers, pairing crypto, tensor network contraction, sparse linear algebra, CV primitives, numerical solvers, generic belief propagation, MD/CFD kernels, convex optimization, DSP beyond cuFFT, DB aggregations, graph theory beyond triangle/PageRank. Sorted by reported speedup or absolute throughput descending.

+ + + + + + + + + + + + + + + + + + + + + + + + + + + +
formspeedup / throughputhardwarerelevanceref
cuda-fluidx3d-lbm100–200× vs ANSYS Fluent / OpenFOAM; 8,799 MLUPS single A100A100unsandbox MEDIUM (HPC reproducibility, OpenCL backend matches our fleet)FluidX3D
cuda-mfcc-spectral~97× CPU MFCC; STFT ~75× via cuSignal vs SciPyGTX 580 / RTX 30-seriesunsandbox HIGH — punters-cc, BT-DISC forensics, real-time CC pipelinecuSignal
cuda-batched-lp-simplex95× over CPLEX; 5× over GLPK on a batch of 100K LPsGTX 980-classunsandbox HIGH — resource scheduling, Prime Mission workstation balancingarXiv 1802.08557
cuda-betweenness-centrality-weighted30–150× warp-centric weighted BCGTX onwardsundefect HIGH — workaholic-node detection on dependency DAG, directly matches MOAD-0001 modelarXiv 1701.05975
cuda-cudasift-orb-ransac~60× SIFT CPU→GPU (11 fps 1920×1440); 1.2 ms on GTX 1060; ORB 11.3×GTX 1060+unsandbox MEDIUM (visual evidence pipeline for incident reports)CudaSift
cuda-cudasw-gasal2CUDASW++4.0 16.2× over v3.0; 134× over ADEPT; 5.71 TCUPS on H100; GASAL2 packing 750× vs NVBioH100 (TCUPS)undefect MEDIUM (binary-diff & patch-similarity at scale: SW reduces to opcode-sequence diff)CUDASW++4.0
cuda-loopy-bp-mrf45× over CPU LBP for stereo MRF inferenceGTX 280-class+undefect HIGH — LBP substrate for FuzzingBrain-style probabilistic program analysisarXiv 2509.22337
cuda-sgm-stereo42 fps at 640×480 with 128 disparities on Tegra X1; 46 fps on discrete GPUsTegra X1 / discreteunsandbox MEDIUM (embedded ARM+CUDA matches our edge node profile)arXiv 1610.04121
cuda-hungarian-lap10–50× class; 400 M-variable LAP in ~13 sNVIDIA GPUunsandbox HIGH — workstation-to-queue balancing per Prime Mission; defect-cluster ↔ patch-bundle assignment for undefectScienceDirect
cuda-msm-bls12-38127.86× over Pippenger (RELIC) AVX baseline; 60% of Groth16 prover time on single GPUA100 / RTX 4090ECDSA HIGH — Pippenger bucket sort + multi-G1 arithmetic shares branchless modmul shape with our reversible secp256k1 inner loopSimdMSM TCHES
cuda-pdwt-lifting15.9× over best optimized CPU DWT (lifting scheme)GTX / Teslaunsandbox MEDIUM (audio-IPC payload analysis, BT signal denoising)PDWT
cuda-tensornet-contract8–20× vs CuPy on contraction; tensor QR ~100× vs Xeon 8480+; tensor SVD ~10×A100ECDSA HIGH — alternative to stabilizer/kickmix sim path; MPS/PEPS evaluates reversible secp256k1 circuits beyond CliffordcuTensorNet
cuda-ega-gpu-aggregation6.45–29.12× over CPU multi-pass EGA; group-by hash 19.4×NVIDIA GPUundefect HIGH — defect-corpus aggregation at planetary scale; unsandbox HIGH — telemetry queue aggregationVLDB Top-k EGA
cuda-bicgstab-ilu-spmvSpTRSV 10.7×; ILU0 BiCGSTAB 3.2× vs cuSPARSE on MI210; GMRES(30) block-ISAI 1.4–6.9×V100 / MI210ECDSA MEDIUM (sparse LA over GF(p) underpins lattice / index-calc); undefect MEDIUM (spectral analysis on DAG)arXiv 2508.04917
cuda-icicle-snark-groth16ICICLE-Snark fastest Groth16 today; Mina GPU 3× over libsnark; NTT 91% of prover at large sizesRTX 4090 / A100ECDSA MEDIUM (zk + MSM stack shares finite-field discipline); undefect MEDIUM (zk-prover defect scanning)ICICLE-Snark
+

Wave 3 filter-outs: AMGX algebraic multigrid (2–5×), GROMACS GPU (2–3×), NVOFA optical flow (7–10× borderline, dedicated hardware unit), Junction-tree BP per-message (0.68–9.18×), batched L-BFGS (134× single-case, not generalized). All below 10× or insufficiently general; revisit when shape changes.

+

Skipped — revisit when shape changes