GPU Benchmark Results

• Target profile: nvidia-gpu
• Suite: gpu/dense
• Suite file: benchmarks/gpu/dense.yaml
• Timestamp: 2026-10-08T12:59:42.398069+00:00
• tenferro-rs commit: b3f47296244ff7b7c55ac0a75f782cb0835418c1

GPU Information

• Device: cuda:0
• Name: NVIDIA L4
• UUID: GPU-a2742c63-0edb-1b0a-9cc7-5b8184ed07e0
• Memory: 23034 MiB
• Driver version: 580.126.20
• CUDA version: 13.0
• CUDA runtime: 12.8
• cuDNN version: 92000

CPU Information

• Model: AMD EPYC 9254 24-Core Processor
• Vendor: AuthenticAMD
• Logical CPUs: 48
• Sockets: 1
• Cores per socket: 24
• Threads per core: 2
• NUMA nodes: 1
• Python platform: Linux-6.8.0-106-generic-x86_64-with-glibc2.39

Median time is reported in milliseconds for ok records. Inputs are prepared on the GPU before timed runs; initial host-to-device transfer is outside the timed region. Timed runs include the host API call and backend-native device synchronization. tenferro-rs CUDA uses the explicit tenferro-rs synchronize API without downloading result tensors in the timed region. Dense and einsum inputs use the same deterministic benchmark generator in the Rust and Python runners; host-to-device layout conversion remains outside the timed region. tenferro-rs uses native column-major GPU tensors; PyTorch and vendor-wrapper columns use their native row-major framework tensors unless noted. The cuSOLVER column is torch.linalg with preferred_linalg_library=cusolver; for SVD it pins driver=gesvd as a QR-based cuSOLVER comparison. tenferro-rs CUDA SVD uses its backend default driver policy, currently gesvdj for matrices with both dimensions at most 1024 and gesvd otherwise. Non-ok cells show the structured backend status.

gpu/dense / batched_matmul / allocating output

gpu/dense / eigh / allocating output

gpu/dense / matmul / allocating output

gpu/dense / qr / allocating output

gpu/dense / solve / allocating output

gpu/dense / svd / allocating output

SVD note: SVD rows use synchronized timed regions and matched Rust/Python input generators. tenferro-rs CUDA uses its backend default driver policy, currently gesvdj for matrices with both dimensions at most 1024 and gesvd otherwise. The cuSOLVER column pins torch.linalg.svd driver=gesvd as a QR-based cuSOLVER comparison; PyTorch's default row may use a different SVD driver and row-major framework layout.