amd-cpu5cf78c7ec0ad9516dd78bab546d5e7bd42fa31022cd0fd6445dcd9aad98990f76c17e33e773c4a7dGenerated from the explicit runs below; each section retains its own provenance and thread settings.
Source report: data/results/amd-cpu/cpu/einsum/20261008_044146/cpu_ops_report.md.
cpu/cpu_opsamd-cpu20261008_044146Latest run: ./scripts/run_all.sh 1.
This file is generated from one CPU ops run under data/results/amd-cpu/cpu/einsum/20261008_044146.
5cf78c7ec0ad9516dd78bab546d5e7bd42fa3102AMD Ryzen 9 9955HX 16-Core ProcessorAuthenticAMD3211621Linux-7.0.0-38-generic-x86_64-with-glibc2.3911FALSE11111111--xla_cpu_multi_thread_eigen=false --xla_cpu_experimental_ynn_fusion_type= intra_op_parallelism_threads=1system-mklblasmkl2026.0.1/opt/intel/oneapi/mkl/latest/opt/intel/oneapi/mkl/latest/lib/libmkl_rt.somkl, version 2.12.0+cpu, BLAS_INFO mkl, LAPACK_INFO mkl/workspaces/tenferro-benchmark/.venv/lib/python3.12/site-packages/torch/lib/libgomp.so.1JAX: dot backend xla_cpu, version 0.10.1, jaxlib 0.10.1, default backend cpu, LAPACK provider none_detected
CSV: data/results/amd-cpu/cpu/einsum/20261008_044146/cpu_ops_t1_20261008_044146.csv
data/results/amd-cpu/cpu/einsum/20261008_044146/cpu_ops_t1_20261008_044146.mdSmall rows use time/memory-bounded batches, normalized per operation. Tenferro eager/trace share a CPU execution scope outside timing. Eager primal rows reuse a borrowed session entered before sampling; eager backward rows without a borrowed-session API are unsupported. Raw JSONL records contain batch durations and operation counts.
Median ± IQR (ms). Missing backends are shown as -.
| suite | benchmark | dtype | threads | shape | tenferro-rs direct API (ms) | tenferro-rs eager mode (ms) | tenferro-rs trace mode (ms) | PyTorch Python (ms) | JAX Python (XLA CPU) (ms) | Julia (Base/LinearAlgebra) (ms) | Julia (Strided.jl) (ms) |
|---|---|---|---|---|---|---|---|---|---|---|---|
| batched | batched_eigh |
f64 | 1 | 16x16xbatch1024 (native batch layout) |
- | 20.298 ± 0.027 | 20.286 ± 0.029 | 21.490 ± 0.084 | 26.954 ± 0.323 | - | - |
| batched | batched_eigh |
f64 | 1 | 16x16xbatch16 (native batch layout) |
- | 0.305 ± 0.001 | 0.306 ± 0.000 | 0.324 ± 0.001 | 0.444 ± 0.014 | - | - |
| batched | batched_eigh |
f64 | 1 | 16x16xbatch256 (native batch layout) |
- | 5.069 ± 0.009 | 5.082 ± 0.014 | 5.422 ± 0.025 | 7.246 ± 0.150 | - | - |
| batched | batched_eigh |
f64 | 1 | 16x16xbatch64 (native batch layout) |
- | 1.240 ± 0.003 | 1.245 ± 0.002 | 1.332 ± 0.003 | 1.800 ± 0.068 | - | - |
| batched | batched_eigh |
f64 | 1 | 2x2xbatch1024 (native batch layout) |
- | 0.619 ± 0.002 | 0.621 ± 0.002 | 0.399 ± 0.001 | 0.365 ± 0.010 | - | - |
| batched | batched_eigh |
f64 | 1 | 2x2xbatch16 (native batch layout) |
- | 0.016 ± 0.000 | 0.018 ± 0.000 | 0.012 ± 0.000 | 0.013 ± 0.000 | - | - |
| batched | batched_eigh |
f64 | 1 | 2x2xbatch256 (native batch layout) |
- | 0.160 ± 0.000 | 0.162 ± 0.000 | 0.105 ± 0.001 | 0.094 ± 0.006 | - | - |
| batched | batched_eigh |
f64 | 1 | 2x2xbatch64 (native batch layout) |
- | 0.045 ± 0.000 | 0.047 ± 0.000 | 0.032 ± 0.000 | 0.026 ± 0.001 | - | - |
| batched | batched_eigh |
f64 | 1 | 4x4xbatch1024 (native batch layout) |
- | 2.239 ± 0.011 | 2.244 ± 0.003 | 2.079 ± 0.006 | 2.142 ± 0.076 | - | - |
| batched | batched_eigh |
f64 | 1 | 4x4xbatch16 (native batch layout) |
- | 0.041 ± 0.000 | 0.043 ± 0.000 | 0.039 ± 0.000 | 0.041 ± 0.000 | - | - |
| batched | batched_eigh |
f64 | 1 | 4x4xbatch256 (native batch layout) |
- | 0.563 ± 0.001 | 0.566 ± 0.001 | 0.524 ± 0.001 | 0.542 ± 0.022 | - | - |
| batched | batched_eigh |
f64 | 1 | 4x4xbatch64 (native batch layout) |
- | 0.145 ± 0.000 | 0.148 ± 0.000 | 0.136 ± 0.000 | 0.137 ± 0.007 | - | - |
| batched | batched_eigh |
f64 | 1 | 8x8xbatch1024 (native batch layout) |
- | 6.230 ± 0.009 | 6.235 ± 0.009 | 6.238 ± 0.022 | 7.372 ± 0.167 | - | - |
| batched | batched_eigh |
f64 | 1 | 8x8xbatch16 (native batch layout) |
- | 0.103 ± 0.000 | 0.104 ± 0.000 | 0.104 ± 0.000 | 0.119 ± 0.002 | - | - |
| batched | batched_eigh |
f64 | 1 | 8x8xbatch256 (native batch layout) |
- | 1.565 ± 0.002 | 1.565 ± 0.001 | 1.567 ± 0.004 | 1.864 ± 0.061 | - | - |
| batched | batched_eigh |
f64 | 1 | 8x8xbatch64 (native batch layout) |
- | 0.393 ± 0.001 | 0.395 ± 0.001 | 0.396 ± 0.001 | 0.460 ± 0.009 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 1 | 16x16xbatch1024 (native batch layout) |
- | 0.751 ± 0.004 | 0.894 ± 0.004 | 0.827 ± 0.010 | 0.921 ± 0.061 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 1 | 16x16xbatch16 (native batch layout) |
- | 0.017 ± 0.000 | 0.022 ± 0.000 | 0.024 ± 0.000 | 0.120 ± 0.001 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 1 | 16x16xbatch256 (native batch layout) |
- | 0.193 ± 0.002 | 0.234 ± 0.002 | 0.209 ± 0.002 | 0.174 ± 0.019 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 1 | 16x16xbatch64 (native batch layout) |
- | 0.053 ± 0.000 | 0.062 ± 0.000 | 0.059 ± 0.001 | 0.125 ± 0.001 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 1 | 2x2xbatch1024 (native batch layout) |
- | 0.100 ± 0.000 | 0.106 ± 0.000 | 0.034 ± 0.000 | 0.119 ± 0.001 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 1 | 2x2xbatch16 (native batch layout) |
- | 0.006 ± 0.000 | 0.011 ± 0.000 | 0.012 ± 0.000 | 0.116 ± 0.001 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 1 | 2x2xbatch256 (native batch layout) |
- | 0.029 ± 0.000 | 0.034 ± 0.000 | 0.017 ± 0.000 | 0.120 ± 0.001 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 1 | 2x2xbatch64 (native batch layout) |
- | 0.011 ± 0.000 | 0.016 ± 0.000 | 0.013 ± 0.000 | 0.118 ± 0.001 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 1 | 4x4xbatch1024 (native batch layout) |
- | 0.119 ± 0.000 | 0.130 ± 0.001 | 0.128 ± 0.000 | 0.125 ± 0.001 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 1 | 4x4xbatch16 (native batch layout) |
- | 0.007 ± 0.000 | 0.011 ± 0.000 | 0.013 ± 0.000 | 0.128 ± 0.001 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 1 | 4x4xbatch256 (native batch layout) |
- | 0.034 ± 0.000 | 0.039 ± 0.000 | 0.041 ± 0.000 | 0.119 ± 0.001 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 1 | 4x4xbatch64 (native batch layout) |
- | 0.012 ± 0.000 | 0.017 ± 0.000 | 0.019 ± 0.000 | 0.119 ± 0.001 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 1 | 8x8xbatch1024 (native batch layout) |
- | 0.205 ± 0.001 | 0.244 ± 0.002 | 0.191 ± 0.003 | 0.166 ± 0.005 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 1 | 8x8xbatch16 (native batch layout) |
- | 0.008 ± 0.000 | 0.013 ± 0.000 | 0.015 ± 0.000 | 0.119 ± 0.001 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 1 | 8x8xbatch256 (native batch layout) |
- | 0.055 ± 0.000 | 0.065 ± 0.000 | 0.055 ± 0.001 | 0.125 ± 0.001 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 1 | 8x8xbatch64 (native batch layout) |
- | 0.018 ± 0.000 | 0.023 ± 0.000 | 0.022 ± 0.000 | 0.121 ± 0.000 | - | - |
| batched | batched_qr |
f64 | 1 | 16x16xbatch1024 (native batch layout) |
- | 4.298 ± 0.011 | 4.281 ± 0.004 | 4.628 ± 0.027 | 7.303 ± 0.446 | - | - |
| batched | batched_qr |
f64 | 1 | 16x16xbatch16 (native batch layout) |
- | 0.073 ± 0.000 | 0.074 ± 0.000 | 0.076 ± 0.001 | 0.111 ± 0.004 | - | - |
| batched | batched_qr |
f64 | 1 | 16x16xbatch256 (native batch layout) |
- | 1.075 ± 0.002 | 1.080 ± 0.002 | 1.164 ± 0.009 | 1.742 ± 0.091 | - | - |
| batched | batched_qr |
f64 | 1 | 16x16xbatch64 (native batch layout) |
- | 0.271 ± 0.002 | 0.274 ± 0.001 | 0.292 ± 0.002 | 0.430 ± 0.014 | - | - |
| batched | batched_qr |
f64 | 1 | 2x2xbatch1024 (native batch layout) |
- | 0.204 ± 0.001 | 0.207 ± 0.000 | 0.204 ± 0.001 | 0.268 ± 0.008 | - | - |
| batched | batched_qr |
f64 | 1 | 2x2xbatch16 (native batch layout) |
- | 0.010 ± 0.000 | 0.012 ± 0.000 | 0.008 ± 0.000 | 0.012 ± 0.000 | - | - |
| batched | batched_qr |
f64 | 1 | 2x2xbatch256 (native batch layout) |
- | 0.056 ± 0.000 | 0.058 ± 0.000 | 0.055 ± 0.000 | 0.073 ± 0.005 | - | - |
| batched | batched_qr |
f64 | 1 | 2x2xbatch64 (native batch layout) |
- | 0.019 ± 0.000 | 0.021 ± 0.000 | 0.017 ± 0.000 | 0.024 ± 0.000 | - | - |
| batched | batched_qr |
f64 | 1 | 4x4xbatch1024 (native batch layout) |
- | 0.579 ± 0.002 | 0.582 ± 0.001 | 0.540 ± 0.002 | 0.777 ± 0.057 | - | - |
| batched | batched_qr |
f64 | 1 | 4x4xbatch16 (native batch layout) |
- | 0.015 ± 0.000 | 0.017 ± 0.000 | 0.013 ± 0.000 | 0.020 ± 0.000 | - | - |
| batched | batched_qr |
f64 | 1 | 4x4xbatch256 (native batch layout) |
- | 0.150 ± 0.000 | 0.152 ± 0.000 | 0.138 ± 0.001 | 0.202 ± 0.008 | - | - |
| batched | batched_qr |
f64 | 1 | 4x4xbatch64 (native batch layout) |
- | 0.042 ± 0.000 | 0.044 ± 0.000 | 0.038 ± 0.000 | 0.053 ± 0.002 | - | - |
| batched | batched_qr |
f64 | 1 | 8x8xbatch1024 (native batch layout) |
- | 1.378 ± 0.004 | 1.380 ± 0.004 | 1.367 ± 0.006 | 2.270 ± 0.057 | - | - |
| batched | batched_qr |
f64 | 1 | 8x8xbatch16 (native batch layout) |
- | 0.028 ± 0.000 | 0.030 ± 0.000 | 0.026 ± 0.000 | 0.041 ± 0.003 | - | - |
| batched | batched_qr |
f64 | 1 | 8x8xbatch256 (native batch layout) |
- | 0.351 ± 0.001 | 0.352 ± 0.001 | 0.347 ± 0.000 | 0.554 ± 0.014 | - | - |
| batched | batched_qr |
f64 | 1 | 8x8xbatch64 (native batch layout) |
- | 0.093 ± 0.000 | 0.094 ± 0.000 | 0.091 ± 0.000 | 0.149 ± 0.009 | - | - |
| batched | batched_solve |
f64 | 1 | 16x16xbatch1024 (native batch layout),rhs=1 |
- | 1.341 ± 0.003 | 1.400 ± 0.006 | 1.496 ± 0.006 | 2.610 ± 0.182 | - | - |
| batched | batched_solve |
f64 | 1 | 16x16xbatch16 (native batch layout),rhs=1 |
- | 0.030 ± 0.000 | 0.031 ± 0.000 | 0.038 ± 0.000 | 0.046 ± 0.002 | - | - |
| batched | batched_solve |
f64 | 1 | 16x16xbatch256 (native batch layout),rhs=1 |
- | 0.336 ± 0.001 | 0.341 ± 0.001 | 0.386 ± 0.003 | 0.614 ± 0.042 | - | - |
| batched | batched_solve |
f64 | 1 | 16x16xbatch64 (native batch layout),rhs=1 |
- | 0.091 ± 0.001 | 0.089 ± 0.000 | 0.107 ± 0.001 | 0.158 ± 0.010 | - | - |
| batched | batched_solve |
f64 | 1 | 2x2xbatch1024 (native batch layout),rhs=1 |
- | 0.086 ± 0.000 | 0.064 ± 0.000 | 0.083 ± 0.000 | 0.250 ± 0.007 | - | - |
| batched | batched_solve |
f64 | 1 | 2x2xbatch16 (native batch layout),rhs=1 |
- | 0.009 ± 0.000 | 0.011 ± 0.000 | 0.015 ± 0.000 | 0.012 ± 0.000 | - | - |
| batched | batched_solve |
f64 | 1 | 2x2xbatch256 (native batch layout),rhs=1 |
- | 0.028 ± 0.000 | 0.024 ± 0.000 | 0.033 ± 0.000 | 0.067 ± 0.002 | - | - |
| batched | batched_solve |
f64 | 1 | 2x2xbatch64 (native batch layout),rhs=1 |
- | 0.013 ± 0.000 | 0.014 ± 0.000 | 0.019 ± 0.000 | 0.021 ± 0.001 | - | - |
| batched | batched_solve |
f64 | 1 | 4x4xbatch1024 (native batch layout),rhs=1 |
- | 0.143 ± 0.000 | 0.107 ± 0.000 | 0.121 ± 0.002 | 0.392 ± 0.027 | - | - |
| batched | batched_solve |
f64 | 1 | 4x4xbatch16 (native batch layout),rhs=1 |
- | 0.010 ± 0.000 | 0.012 ± 0.000 | 0.017 ± 0.000 | 0.015 ± 0.000 | - | - |
| batched | batched_solve |
f64 | 1 | 4x4xbatch256 (native batch layout),rhs=1 |
- | 0.042 ± 0.000 | 0.034 ± 0.000 | 0.042 ± 0.000 | 0.103 ± 0.007 | - | - |
| batched | batched_solve |
f64 | 1 | 4x4xbatch64 (native batch layout),rhs=1 |
- | 0.017 ± 0.000 | 0.016 ± 0.000 | 0.022 ± 0.000 | 0.030 ± 0.001 | - | - |
| batched | batched_solve |
f64 | 1 | 8x8xbatch1024 (native batch layout),rhs=1 |
- | 0.447 ± 0.001 | 0.426 ± 0.001 | 0.431 ± 0.003 | 0.810 ± 0.037 | - | - |
| batched | batched_solve |
f64 | 1 | 8x8xbatch16 (native batch layout),rhs=1 |
- | 0.015 ± 0.000 | 0.016 ± 0.000 | 0.021 ± 0.000 | 0.019 ± 0.001 | - | - |
| batched | batched_solve |
f64 | 1 | 8x8xbatch256 (native batch layout),rhs=1 |
- | 0.116 ± 0.000 | 0.110 ± 0.000 | 0.119 ± 0.001 | 0.208 ± 0.016 | - | - |
| batched | batched_solve |
f64 | 1 | 8x8xbatch64 (native batch layout),rhs=1 |
- | 0.036 ± 0.000 | 0.036 ± 0.000 | 0.041 ± 0.000 | 0.058 ± 0.002 | - | - |
| batched | batched_svd |
f64 | 1 | 16x16xbatch1024 (native batch layout) |
- | 43.728 ± 0.040 | 43.712 ± 0.043 | 47.372 ± 0.106 | 47.504 ± 0.217 | - | - |
| batched | batched_svd |
f64 | 1 | 16x16xbatch16 (native batch layout) |
- | 0.671 ± 0.002 | 0.675 ± 0.001 | 0.727 ± 0.001 | 0.814 ± 0.012 | - | - |
| batched | batched_svd |
f64 | 1 | 16x16xbatch256 (native batch layout) |
- | 10.920 ± 0.023 | 10.932 ± 0.017 | 11.910 ± 0.038 | 12.250 ± 0.156 | - | - |
| batched | batched_svd |
f64 | 1 | 16x16xbatch64 (native batch layout) |
- | 2.715 ± 0.009 | 2.721 ± 0.003 | 2.949 ± 0.016 | 3.172 ± 0.107 | - | - |
| batched | batched_svd |
f64 | 1 | 2x2xbatch1024 (native batch layout) |
- | 1.361 ± 0.007 | 1.360 ± 0.004 | 1.113 ± 0.004 | 1.365 ± 0.053 | - | - |
| batched | batched_svd |
f64 | 1 | 2x2xbatch16 (native batch layout) |
- | 0.028 ± 0.000 | 0.031 ± 0.000 | 0.025 ± 0.000 | 0.030 ± 0.000 | - | - |
| batched | batched_svd |
f64 | 1 | 2x2xbatch256 (native batch layout) |
- | 0.346 ± 0.001 | 0.347 ± 0.001 | 0.285 ± 0.001 | 0.346 ± 0.010 | - | - |
| batched | batched_svd |
f64 | 1 | 2x2xbatch64 (native batch layout) |
- | 0.092 ± 0.000 | 0.094 ± 0.000 | 0.077 ± 0.000 | 0.089 ± 0.000 | - | - |
| batched | batched_svd |
f64 | 1 | 4x4xbatch1024 (native batch layout) |
- | 4.732 ± 0.008 | 4.740 ± 0.008 | 4.230 ± 0.011 | 4.213 ± 0.126 | - | - |
| batched | batched_svd |
f64 | 1 | 4x4xbatch16 (native batch layout) |
- | 0.080 ± 0.000 | 0.084 ± 0.000 | 0.073 ± 0.000 | 0.074 ± 0.000 | - | - |
| batched | batched_svd |
f64 | 1 | 4x4xbatch256 (native batch layout) |
- | 1.188 ± 0.002 | 1.192 ± 0.002 | 1.066 ± 0.001 | 1.054 ± 0.033 | - | - |
| batched | batched_svd |
f64 | 1 | 4x4xbatch64 (native batch layout) |
- | 0.302 ± 0.001 | 0.305 ± 0.000 | 0.271 ± 0.001 | 0.266 ± 0.008 | - | - |
| batched | batched_svd |
f64 | 1 | 8x8xbatch1024 (native batch layout) |
- | 15.250 ± 0.041 | 15.302 ± 0.010 | 14.465 ± 0.036 | 13.177 ± 0.178 | - | - |
| batched | batched_svd |
f64 | 1 | 8x8xbatch16 (native batch layout) |
- | 0.242 ± 0.000 | 0.246 ± 0.000 | 0.231 ± 0.000 | 0.224 ± 0.013 | - | - |
| batched | batched_svd |
f64 | 1 | 8x8xbatch256 (native batch layout) |
- | 3.817 ± 0.007 | 3.836 ± 0.007 | 3.624 ± 0.009 | 3.412 ± 0.094 | - | - |
| batched | batched_svd |
f64 | 1 | 8x8xbatch64 (native batch layout) |
- | 0.961 ± 0.002 | 0.960 ± 0.001 | 0.909 ± 0.002 | 0.882 ± 0.009 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 1 | 16x16xbatch1024 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 2.884 ± 0.026 | 7.845 ± 0.046 | 2.440 ± 0.252 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 1 | 16x16xbatch16 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.079 ± 0.000 | 0.209 ± 0.002 | 0.038 ± 0.003 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 1 | 16x16xbatch256 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.772 ± 0.004 | 2.033 ± 0.016 | 0.568 ± 0.035 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 1 | 16x16xbatch64 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.214 ± 0.001 | 0.577 ± 0.004 | 0.140 ± 0.005 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 1 | 2x2xbatch1024 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.325 ± 0.001 | 0.150 ± 0.001 | 0.023 ± 0.002 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 1 | 2x2xbatch16 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.037 ± 0.000 | 0.083 ± 0.000 | 0.010 ± 0.000 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 1 | 2x2xbatch256 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.107 ± 0.000 | 0.099 ± 0.000 | 0.017 ± 0.000 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 1 | 2x2xbatch64 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.052 ± 0.000 | 0.086 ± 0.001 | 0.014 ± 0.001 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 1 | 4x4xbatch1024 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.415 ± 0.001 | 0.442 ± 0.002 | 0.439 ± 0.020 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 1 | 4x4xbatch16 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.039 ± 0.000 | 0.088 ± 0.001 | 0.015 ± 0.001 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 1 | 4x4xbatch256 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.129 ± 0.000 | 0.173 ± 0.001 | 0.115 ± 0.003 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 1 | 4x4xbatch64 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.057 ± 0.000 | 0.104 ± 0.001 | 0.032 ± 0.001 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 1 | 8x8xbatch1024 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.861 ± 0.003 | 5.565 ± 0.029 | 0.795 ± 0.027 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 1 | 8x8xbatch16 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.046 ± 0.000 | 0.171 ± 0.001 | 0.017 ± 0.000 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 1 | 8x8xbatch256 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.238 ± 0.001 | 1.523 ± 0.014 | 0.207 ± 0.013 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 1 | 8x8xbatch64 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.083 ± 0.000 | 0.432 ± 0.001 | 0.055 ± 0.002 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 1 | 16x16xbatch1024 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 2.397 ± 0.006 | 3.029 ± 0.187 | 4.313 ± 0.365 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 1 | 16x16xbatch16 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.076 ± 0.000 | 0.107 ± 0.001 | 0.076 ± 0.002 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 1 | 16x16xbatch256 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.644 ± 0.002 | 0.821 ± 0.029 | 0.970 ± 0.033 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 1 | 16x16xbatch64 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.184 ± 0.001 | 0.251 ± 0.005 | 0.259 ± 0.005 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 1 | 2x2xbatch1024 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.254 ± 0.001 | 0.179 ± 0.001 | 0.531 ± 0.034 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 1 | 2x2xbatch16 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.044 ± 0.000 | 0.058 ± 0.000 | 0.021 ± 0.000 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 1 | 2x2xbatch256 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.094 ± 0.000 | 0.089 ± 0.000 | 0.137 ± 0.004 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 1 | 2x2xbatch64 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.054 ± 0.000 | 0.065 ± 0.000 | 0.042 ± 0.002 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 1 | 4x4xbatch1024 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.336 ± 0.001 | 0.287 ± 0.002 | 0.810 ± 0.023 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 1 | 4x4xbatch16 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.045 ± 0.000 | 0.062 ± 0.000 | 0.022 ± 0.001 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 1 | 4x4xbatch256 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.113 ± 0.000 | 0.116 ± 0.001 | 0.210 ± 0.008 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 1 | 4x4xbatch64 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.059 ± 0.000 | 0.073 ± 0.000 | 0.058 ± 0.002 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 1 | 8x8xbatch1024 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 1.089 ± 0.003 | 1.017 ± 0.011 | 1.506 ± 0.036 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 1 | 8x8xbatch16 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.056 ± 0.000 | 0.072 ± 0.000 | 0.032 ± 0.001 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 1 | 8x8xbatch256 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.301 ± 0.000 | 0.304 ± 0.003 | 0.376 ± 0.017 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 1 | 8x8xbatch64 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.104 ± 0.000 | 0.119 ± 0.000 | 0.103 ± 0.004 | - | - |
| large | eigh |
f64 | 1 | 1024x1024 |
- | 282.723 ± 0.915 | 282.184 ± 0.431 | 291.704 ± 0.378 | 198.438 ± 2.902 | - | - |
| large | eigh |
f64 | 1 | 128x128 |
- | 1.394 ± 0.012 | 1.399 ± 0.010 | 1.470 ± 0.005 | 1.464 ± 0.068 | - | - |
| large | eigh |
f64 | 1 | 256x256 |
- | 7.577 ± 0.018 | 7.561 ± 0.029 | 7.730 ± 0.020 | 6.760 ± 0.194 | - | - |
| large | eigh |
f64 | 1 | 512x512 |
- | 45.400 ± 0.105 | 45.351 ± 0.181 | 46.627 ± 0.124 | 34.732 ± 0.177 | - | - |
| large | eigh |
f64 | 1 | 64x64 |
- | 0.306 ± 0.000 | 0.308 ± 0.001 | 0.316 ± 0.001 | 0.374 ± 0.018 | - | - |
| large | grad_sum_eigh_jvp |
f64 | 1 | 1024x1024 |
- | - | 348.894 ± 0.537 | 493.351 ± 1.190 | 323.311 ± 0.574 | - | - |
| large | grad_sum_eigh_jvp |
f64 | 1 | 256x256 |
- | - | 8.925 ± 0.019 | 11.077 ± 0.037 | 8.384 ± 0.027 | - | - |
| large | grad_sum_eigh_jvp |
f64 | 1 | 512x512 |
- | - | 55.310 ± 0.214 | 74.532 ± 0.165 | 51.286 ± 0.247 | - | - |
| large | grad_sum_eigh_vjp |
f64 | 1 | 1024x1024 |
- | - | 350.559 ± 0.476 | 356.213 ± 0.625 | 320.925 ± 0.413 | - | - |
| large | grad_sum_eigh_vjp |
f64 | 1 | 256x256 |
- | - | 8.878 ± 0.023 | 8.978 ± 0.048 | 8.350 ± 0.064 | - | - |
| large | grad_sum_eigh_vjp |
f64 | 1 | 512x512 |
- | - | 55.276 ± 0.129 | 56.729 ± 0.082 | 51.124 ± 0.305 | - | - |
| large | grad_sum_lu_jvp |
f64 | 1 | 1024x1024 |
- | - | 282.367 ± 0.360 | 609.271 ± 2.501 | 234.359 ± 2.491 | - | - |
| large | grad_sum_lu_jvp |
f64 | 1 | 256x256 |
- | - | 5.342 ± 0.017 | 10.825 ± 0.076 | 4.599 ± 0.016 | - | - |
| large | grad_sum_lu_jvp |
f64 | 1 | 512x512 |
- | - | 38.496 ± 0.209 | 82.831 ± 0.187 | 32.653 ± 0.091 | - | - |
| large | grad_sum_lu_vjp |
f64 | 1 | 1024x1024 |
- | - | 276.074 ± 0.653 | 326.508 ± 5.939 | 235.357 ± 1.248 | - | - |
| large | grad_sum_lu_vjp |
f64 | 1 | 256x256 |
- | - | 5.465 ± 0.006 | 5.832 ± 0.023 | 4.504 ± 0.022 | - | - |
| large | grad_sum_lu_vjp |
f64 | 1 | 512x512 |
- | - | 38.204 ± 0.065 | 43.382 ± 0.117 | 31.662 ± 0.072 | - | - |
| large | grad_sum_matmul |
f64 | 1 | 1024x1024 |
- | 63.075 ± 0.306 | 59.155 ± 0.532 | 67.647 ± 0.602 | 61.157 ± 0.195 | - | - |
| large | grad_sum_matmul |
f64 | 1 | 128x128 |
- | 0.161 ± 0.001 | 0.155 ± 0.000 | 0.149 ± 0.003 | 0.157 ± 0.012 | - | - |
| large | grad_sum_matmul |
f64 | 1 | 256x256 |
- | 1.022 ± 0.004 | 1.002 ± 0.003 | 1.035 ± 0.007 | 1.073 ± 0.038 | - | - |
| large | grad_sum_matmul |
f64 | 1 | 512x512 |
- | 7.788 ± 0.035 | 7.788 ± 0.066 | 8.791 ± 0.034 | 8.054 ± 0.289 | - | - |
| large | grad_sum_matmul |
f64 | 1 | 64x64 |
- | 0.037 ± 0.000 | 0.032 ± 0.000 | 0.026 ± 0.000 | 0.029 ± 0.003 | - | - |
| large | grad_sum_matmul_backward |
f64 | 1 | 1024x1024 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 179.406 ± 0.802 | 201.239 ± 1.248 | 184.856 ± 1.855 | - | - |
| large | grad_sum_matmul_backward |
f64 | 1 | 128x128 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.490 ± 0.001 | 0.476 ± 0.002 | 0.429 ± 0.023 | - | - |
| large | grad_sum_matmul_backward |
f64 | 1 | 256x256 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 3.043 ± 0.007 | 3.059 ± 0.010 | 3.219 ± 0.103 | - | - |
| large | grad_sum_matmul_backward |
f64 | 1 | 512x512 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 23.303 ± 0.112 | 25.201 ± 0.071 | 23.433 ± 0.317 | - | - |
| large | grad_sum_matmul_backward |
f64 | 1 | 64x64 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.099 ± 0.001 | 0.098 ± 0.001 | 0.064 ± 0.002 | - | - |
| large | grad_sum_qr_jvp |
f64 | 1 | 1024x1024 |
- | - | 365.127 ± 0.494 | 418.926 ± 1.604 | 301.561 ± 1.265 | - | - |
| large | grad_sum_qr_jvp |
f64 | 1 | 256x256 |
- | - | 7.699 ± 0.038 | 8.272 ± 0.018 | 6.457 ± 0.017 | - | - |
| large | grad_sum_qr_jvp |
f64 | 1 | 512x512 |
- | - | 57.267 ± 0.091 | 66.265 ± 0.747 | 42.961 ± 0.350 | - | - |
| large | grad_sum_qr_vjp |
f64 | 1 | 1024x1024 |
- | - | 360.289 ± 0.408 | 407.364 ± 2.039 | 331.027 ± 0.619 | - | - |
| large | grad_sum_qr_vjp |
f64 | 1 | 256x256 |
- | - | 7.658 ± 0.014 | 8.287 ± 0.025 | 6.462 ± 0.041 | - | - |
| large | grad_sum_qr_vjp |
f64 | 1 | 512x512 |
- | - | 54.552 ± 0.101 | 68.759 ± 0.876 | 47.262 ± 0.587 | - | - |
| large | grad_sum_solve_backward |
f64 | 1 | 1024x1024,rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 28.162 ± 0.092 | 40.933 ± 0.322 | 22.921 ± 0.420 | - | - |
| large | grad_sum_solve_backward |
f64 | 1 | 128x128,rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.165 ± 0.001 | 0.204 ± 0.003 | 0.166 ± 0.007 | - | - |
| large | grad_sum_solve_backward |
f64 | 1 | 256x256,rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.741 ± 0.004 | 0.881 ± 0.008 | 0.701 ± 0.024 | - | - |
| large | grad_sum_solve_backward |
f64 | 1 | 512x512,rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 4.275 ± 0.015 | 5.399 ± 0.040 | 4.665 ± 0.160 | - | - |
| large | grad_sum_solve_backward |
f64 | 1 | 64x64,rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.071 ± 0.000 | 0.089 ± 0.001 | 0.051 ± 0.002 | - | - |
| large | grad_sum_solve_jvp |
f64 | 1 | 1024x1024,rhs=1 |
- | - | 27.206 ± 0.092 | 36.963 ± 0.129 | 22.241 ± 0.089 | - | - |
| large | grad_sum_solve_jvp |
f64 | 1 | 256x256,rhs=1 |
- | - | 0.720 ± 0.006 | 1.068 ± 0.017 | 0.610 ± 0.008 | - | - |
| large | grad_sum_solve_jvp |
f64 | 1 | 512x512,rhs=1 |
- | - | 4.274 ± 0.066 | 5.398 ± 0.025 | 4.051 ± 0.020 | - | - |
| large | grad_sum_solve_vjp |
f64 | 1 | 1024x1024,rhs=1 |
- | - | 26.804 ± 0.083 | 65.111 ± 0.487 | 22.017 ± 0.050 | - | - |
| large | grad_sum_solve_vjp |
f64 | 1 | 256x256,rhs=1 |
- | - | 0.705 ± 0.006 | 1.564 ± 0.010 | 0.623 ± 0.003 | - | - |
| large | grad_sum_solve_vjp |
f64 | 1 | 512x512,rhs=1 |
- | - | 4.216 ± 0.026 | 9.538 ± 0.026 | 4.087 ± 0.015 | - | - |
| large | grad_sum_svd_s_backward |
f64 | 1 | 1024x1024 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 681.870 ± 2.592 | 748.617 ± 2.088 | 611.029 ± 3.765 | - | - |
| large | grad_sum_svd_s_backward |
f64 | 1 | 128x128 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 2.977 ± 0.018 | 3.188 ± 0.016 | 2.925 ± 0.029 | - | - |
| large | grad_sum_svd_s_backward |
f64 | 1 | 256x256 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 16.134 ± 0.092 | 16.864 ± 0.039 | 14.794 ± 0.174 | - | - |
| large | grad_sum_svd_s_backward |
f64 | 1 | 512x512 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 108.161 ± 1.700 | 126.310 ± 0.256 | 100.161 ± 0.449 | - | - |
| large | grad_sum_svd_s_backward |
f64 | 1 | 64x64 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.642 ± 0.002 | 0.700 ± 0.010 | 0.605 ± 0.017 | - | - |
| large | grad_sum_svd_s_jvp |
f64 | 1 | 1024x1024 |
- | - | 676.216 ± 3.145 | 908.650 ± 4.960 | 607.557 ± 3.247 | - | - |
| large | grad_sum_svd_s_jvp |
f64 | 1 | 256x256 |
- | - | 16.134 ± 0.091 | 20.226 ± 0.068 | 14.537 ± 0.023 | - | - |
| large | grad_sum_svd_s_jvp |
f64 | 1 | 512x512 |
- | - | 107.400 ± 0.993 | 148.129 ± 0.664 | 99.509 ± 0.661 | - | - |
| large | grad_sum_svd_s_vjp |
f64 | 1 | 1024x1024 |
- | - | 661.978 ± 15.387 | 684.864 ± 6.642 | 607.308 ± 1.994 | - | - |
| large | grad_sum_svd_s_vjp |
f64 | 1 | 256x256 |
- | - | 16.094 ± 0.039 | 16.906 ± 0.064 | 14.569 ± 0.050 | - | - |
| large | grad_sum_svd_s_vjp |
f64 | 1 | 512x512 |
- | - | 106.149 ± 2.119 | 125.661 ± 0.489 | 99.615 ± 0.461 | - | - |
| large | matmul |
f64 | 1 | 1024x1024 |
- | 58.006 ± 0.103 | 58.975 ± 0.130 | 61.822 ± 0.287 | 60.039 ± 0.149 | - | - |
| large | matmul |
f64 | 1 | 128x128 |
- | 0.159 ± 0.001 | 0.151 ± 0.001 | 0.166 ± 0.006 | 0.143 ± 0.017 | - | - |
| large | matmul |
f64 | 1 | 256x256 |
- | 1.029 ± 0.004 | 0.998 ± 0.002 | 1.084 ± 0.012 | 1.120 ± 0.042 | - | - |
| large | matmul |
f64 | 1 | 512x512 |
- | 7.684 ± 0.044 | 7.631 ± 0.061 | 8.030 ± 0.022 | 7.916 ± 0.287 | - | - |
| large | matmul_rect |
f64 | 1 | 1024x256 * 256x1024 |
- | 14.489 ± 0.046 | 14.623 ± 0.057 | 15.236 ± 0.056 | 15.123 ± 0.152 | - | - |
| large | matmul_rect |
f64 | 1 | 256x1024 * 1024x256 |
- | 3.759 ± 0.008 | 3.879 ± 0.022 | 4.073 ± 0.028 | 4.233 ± 0.221 | - | - |
| large | qr |
f64 | 1 | 1024x1024 |
- | 130.893 ± 0.353 | 131.362 ± 1.052 | 159.844 ± 1.398 | 98.438 ± 1.121 | - | - |
| large | qr |
f64 | 1 | 128x128 |
- | 0.535 ± 0.001 | 0.537 ± 0.001 | 0.650 ± 0.003 | 0.609 ± 0.036 | - | - |
| large | qr |
f64 | 1 | 256x256 |
- | 3.139 ± 0.022 | 3.145 ± 0.012 | 3.834 ± 0.019 | 2.771 ± 0.121 | - | - |
| large | qr |
f64 | 1 | 512x512 |
- | 22.256 ± 0.113 | 22.569 ± 0.179 | 26.527 ± 0.070 | 17.895 ± 0.442 | - | - |
| large | qr |
f64 | 1 | 64x64 |
- | 0.096 ± 0.000 | 0.097 ± 0.000 | 0.105 ± 0.001 | 0.097 ± 0.005 | - | - |
| large | solve |
f64 | 1 | 1024x1024,rhs=1 |
- | 25.679 ± 0.037 | 25.851 ± 0.146 | 39.283 ± 0.336 | 22.535 ± 0.380 | - | - |
| large | solve |
f64 | 1 | 1024x1024,rhs=16 |
- | 27.069 ± 0.079 | 27.195 ± 0.059 | 41.276 ± 0.470 | 23.113 ± 0.640 | - | - |
| large | solve |
f64 | 1 | 1024x1024,rhs=64 |
- | 30.498 ± 0.112 | 30.523 ± 0.096 | 45.731 ± 0.374 | 25.874 ± 0.321 | - | - |
| large | solve |
f64 | 1 | 128x128,rhs=1 |
- | 0.109 ± 0.000 | 0.113 ± 0.001 | 0.138 ± 0.000 | 0.137 ± 0.007 | - | - |
| large | solve |
f64 | 1 | 128x128,rhs=16 |
- | 0.147 ± 0.003 | 0.153 ± 0.004 | 0.181 ± 0.002 | 0.394 ± 0.038 | - | - |
| large | solve |
f64 | 1 | 128x128,rhs=64 |
- | 0.247 ± 0.004 | 0.257 ± 0.003 | 0.289 ± 0.003 | 0.580 ± 0.296 | - | - |
| large | solve |
f64 | 1 | 256x256,rhs=1 |
- | 0.612 ± 0.002 | 0.630 ± 0.004 | 0.767 ± 0.004 | 0.636 ± 0.039 | - | - |
| large | solve |
f64 | 1 | 256x256,rhs=16 |
- | 0.728 ± 0.003 | 0.744 ± 0.001 | 0.887 ± 0.007 | 1.146 ± 0.275 | - | - |
| large | solve |
f64 | 1 | 256x256,rhs=64 |
- | 1.021 ± 0.009 | 1.040 ± 0.004 | 1.237 ± 0.011 | 1.398 ± 0.173 | - | - |
| large | solve |
f64 | 1 | 512x512,rhs=1 |
- | 3.932 ± 0.018 | 3.999 ± 0.033 | 5.073 ± 0.023 | 4.447 ± 0.390 | - | - |
| large | solve |
f64 | 1 | 512x512,rhs=16 |
- | 4.356 ± 0.009 | 4.410 ± 0.030 | 5.517 ± 0.034 | 4.606 ± 0.363 | - | - |
| large | solve |
f64 | 1 | 512x512,rhs=64 |
- | 5.352 ± 0.023 | 5.406 ± 0.008 | 6.674 ± 0.030 | 5.503 ± 0.231 | - | - |
| large | solve |
f64 | 1 | 64x64,rhs=1 |
- | 0.033 ± 0.000 | 0.036 ± 0.000 | 0.041 ± 0.000 | 0.040 ± 0.002 | - | - |
| large | solve |
f64 | 1 | 64x64,rhs=16 |
- | 0.044 ± 0.001 | 0.048 ± 0.000 | 0.054 ± 0.001 | 0.109 ± 0.013 | - | - |
| large | solve |
f64 | 1 | 64x64,rhs=64 |
- | 0.075 ± 0.000 | 0.079 ± 0.002 | 0.092 ± 0.002 | 0.127 ± 0.031 | - | - |
| large | svd |
f64 | 1 | 1024x1024 |
- | 561.278 ± 2.950 | 558.175 ± 4.959 | 620.126 ± 14.833 | 460.655 ± 0.877 | - | - |
| large | svd |
f64 | 1 | 128x128 |
- | 2.658 ± 0.013 | 2.677 ± 0.009 | 2.818 ± 0.009 | 2.718 ± 0.174 | - | - |
| large | svd |
f64 | 1 | 256x256 |
- | 14.186 ± 0.056 | 14.232 ± 0.054 | 15.282 ± 0.025 | 12.953 ± 0.137 | - | - |
| large | svd |
f64 | 1 | 512x512 |
- | 91.733 ± 0.686 | 91.706 ± 0.674 | 99.442 ± 0.293 | 82.792 ± 0.279 | - | - |
| large | svd |
f64 | 1 | 64x64 |
- | 0.576 ± 0.002 | 0.578 ± 0.003 | 0.599 ± 0.002 | 0.615 ± 0.031 | - | - |
| small | eigh |
f64 | 1 | 16x16 |
- | 0.025 ± 0.000 | 0.027 ± 0.000 | 0.026 ± 0.000 | 0.035 ± 0.000 | - | - |
| small | eigh |
f64 | 1 | 2x2 |
- | 0.006 ± 0.000 | 0.009 ± 0.000 | 0.006 ± 0.000 | 0.008 ± 0.000 | - | - |
| small | eigh |
f64 | 1 | 32x32 |
- | 0.078 ± 0.000 | 0.079 ± 0.000 | 0.084 ± 0.000 | 0.103 ± 0.005 | - | - |
| small | eigh |
f64 | 1 | 4x4 |
- | 0.008 ± 0.000 | 0.010 ± 0.000 | 0.008 ± 0.000 | 0.010 ± 0.000 | - | - |
| small | eigh |
f64 | 1 | 8x8 |
- | 0.012 ± 0.000 | 0.014 ± 0.000 | 0.012 ± 0.000 | 0.015 ± 0.000 | - | - |
| small | einsum_ij_jk_ik |
f64 | 1 | 16x16 |
- | 0.005 ± 0.000 | 0.011 ± 0.000 | 0.012 ± 0.000 | 0.127 ± 0.002 | - | - |
| small | einsum_ij_jk_ik |
f64 | 1 | 2x2 |
- | 0.003 ± 0.000 | 0.010 ± 0.000 | 0.011 ± 0.000 | 0.115 ± 0.001 | - | - |
| small | einsum_ij_jk_ik |
f64 | 1 | 32x32 |
- | 0.007 ± 0.000 | 0.014 ± 0.000 | 0.015 ± 0.000 | 0.131 ± 0.002 | - | - |
| small | einsum_ij_jk_ik |
f64 | 1 | 4x4 |
- | 0.003 ± 0.000 | 0.010 ± 0.000 | 0.011 ± 0.000 | 0.117 ± 0.002 | - | - |
| small | einsum_ij_jk_ik |
f64 | 1 | 8x8 |
- | 0.004 ± 0.000 | 0.010 ± 0.000 | 0.011 ± 0.000 | 0.117 ± 0.001 | - | - |
| small | grad_sum_eigh_jvp |
f64 | 1 | 16x16 |
- | - | 0.066 ± 0.000 | 0.124 ± 0.001 | 0.032 ± 0.000 | - | - |
| small | grad_sum_eigh_jvp |
f64 | 1 | 2x2 |
- | - | 0.046 ± 0.000 | 0.097 ± 0.002 | 0.009 ± 0.000 | - | - |
| small | grad_sum_eigh_jvp |
f64 | 1 | 32x32 |
- | - | 0.126 ± 0.001 | 0.197 ± 0.003 | 0.102 ± 0.000 | - | - |
| small | grad_sum_eigh_jvp |
f64 | 1 | 4x4 |
- | - | 0.047 ± 0.000 | 0.098 ± 0.001 | 0.011 ± 0.000 | - | - |
| small | grad_sum_eigh_jvp |
f64 | 1 | 8x8 |
- | - | 0.052 ± 0.000 | 0.105 ± 0.002 | 0.013 ± 0.000 | - | - |
| small | grad_sum_eigh_vjp |
f64 | 1 | 16x16 |
- | - | 0.068 ± 0.000 | 0.107 ± 0.001 | 0.033 ± 0.000 | - | - |
| small | grad_sum_eigh_vjp |
f64 | 1 | 2x2 |
- | - | 0.046 ± 0.000 | 0.083 ± 0.001 | 0.010 ± 0.000 | - | - |
| small | grad_sum_eigh_vjp |
f64 | 1 | 32x32 |
- | - | 0.125 ± 0.000 | 0.171 ± 0.004 | 0.103 ± 0.001 | - | - |
| small | grad_sum_eigh_vjp |
f64 | 1 | 4x4 |
- | - | 0.048 ± 0.000 | 0.086 ± 0.001 | 0.012 ± 0.000 | - | - |
| small | grad_sum_eigh_vjp |
f64 | 1 | 8x8 |
- | - | 0.053 ± 0.000 | 0.090 ± 0.001 | 0.013 ± 0.000 | - | - |
| small | grad_sum_lu_jvp |
f64 | 1 | 16x16 |
- | - | 0.055 ± 0.000 | 0.175 ± 0.004 | 0.015 ± 0.002 | - | - |
| small | grad_sum_lu_jvp |
f64 | 1 | 2x2 |
- | - | 0.049 ± 0.000 | 0.158 ± 0.003 | 0.011 ± 0.000 | - | - |
| small | grad_sum_lu_jvp |
f64 | 1 | 32x32 |
- | - | 0.079 ± 0.000 | 0.216 ± 0.004 | 0.030 ± 0.000 | - | - |
| small | grad_sum_lu_jvp |
f64 | 1 | 4x4 |
- | - | 0.049 ± 0.000 | 0.157 ± 0.003 | 0.011 ± 0.000 | - | - |
| small | grad_sum_lu_jvp |
f64 | 1 | 8x8 |
- | - | 0.051 ± 0.000 | 0.161 ± 0.004 | 0.017 ± 0.000 | - | - |
| small | grad_sum_lu_vjp |
f64 | 1 | 16x16 |
- | - | 0.056 ± 0.000 | 0.138 ± 0.001 | 0.014 ± 0.001 | - | - |
| small | grad_sum_lu_vjp |
f64 | 1 | 2x2 |
- | - | 0.049 ± 0.000 | 0.127 ± 0.001 | 0.011 ± 0.000 | - | - |
| small | grad_sum_lu_vjp |
f64 | 1 | 32x32 |
- | - | 0.083 ± 0.000 | 0.166 ± 0.004 | 0.031 ± 0.000 | - | - |
| small | grad_sum_lu_vjp |
f64 | 1 | 4x4 |
- | - | 0.050 ± 0.000 | 0.127 ± 0.001 | 0.012 ± 0.000 | - | - |
| small | grad_sum_lu_vjp |
f64 | 1 | 8x8 |
- | - | 0.051 ± 0.000 | 0.130 ± 0.001 | 0.013 ± 0.001 | - | - |
| small | grad_sum_matmul_backward |
f64 | 1 | 16x16 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.033 ± 0.000 | 0.036 ± 0.000 | 0.014 ± 0.000 | - | - |
| small | grad_sum_matmul_backward |
f64 | 1 | 2x2 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.030 ± 0.000 | 0.032 ± 0.000 | 0.010 ± 0.000 | - | - |
| small | grad_sum_matmul_backward |
f64 | 1 | 32x32 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.044 ± 0.000 | 0.046 ± 0.000 | 0.017 ± 0.001 | - | - |
| small | grad_sum_matmul_backward |
f64 | 1 | 4x4 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.030 ± 0.000 | 0.034 ± 0.000 | 0.010 ± 0.000 | - | - |
| small | grad_sum_matmul_backward |
f64 | 1 | 8x8 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.030 ± 0.000 | 0.033 ± 0.000 | 0.014 ± 0.001 | - | - |
| small | grad_sum_qr_jvp |
f64 | 1 | 16x16 |
- | - | 0.091 ± 0.001 | 0.124 ± 0.002 | 0.016 ± 0.000 | - | - |
| small | grad_sum_qr_jvp |
f64 | 1 | 2x2 |
- | - | 0.081 ± 0.000 | 0.112 ± 0.003 | 0.011 ± 0.001 | - | - |
| small | grad_sum_qr_jvp |
f64 | 1 | 32x32 |
- | - | 0.119 ± 0.001 | 0.156 ± 0.002 | 0.039 ± 0.001 | - | - |
| small | grad_sum_qr_jvp |
f64 | 1 | 4x4 |
- | - | 0.082 ± 0.000 | 0.110 ± 0.002 | 0.011 ± 0.000 | - | - |
| small | grad_sum_qr_jvp |
f64 | 1 | 8x8 |
- | - | 0.084 ± 0.001 | 0.114 ± 0.001 | 0.014 ± 0.000 | - | - |
| small | grad_sum_qr_vjp |
f64 | 1 | 16x16 |
- | - | 0.094 ± 0.000 | 0.131 ± 0.001 | 0.015 ± 0.000 | - | - |
| small | grad_sum_qr_vjp |
f64 | 1 | 2x2 |
- | - | 0.083 ± 0.000 | 0.118 ± 0.002 | 0.010 ± 0.000 | - | - |
| small | grad_sum_qr_vjp |
f64 | 1 | 32x32 |
- | - | 0.125 ± 0.000 | 0.161 ± 0.004 | 0.039 ± 0.000 | - | - |
| small | grad_sum_qr_vjp |
f64 | 1 | 4x4 |
- | - | 0.084 ± 0.000 | 0.117 ± 0.001 | 0.011 ± 0.000 | - | - |
| small | grad_sum_qr_vjp |
f64 | 1 | 8x8 |
- | - | 0.086 ± 0.000 | 0.119 ± 0.001 | 0.010 ± 0.001 | - | - |
| small | grad_sum_solve_backward |
f64 | 1 | 16x16,rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.040 ± 0.000 | 0.055 ± 0.000 | 0.017 ± 0.000 | - | - |
| small | grad_sum_solve_backward |
f64 | 1 | 2x2,rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.037 ± 0.000 | 0.051 ± 0.000 | 0.013 ± 0.000 | - | - |
| small | grad_sum_solve_backward |
f64 | 1 | 32x32,rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.048 ± 0.000 | 0.063 ± 0.001 | 0.019 ± 0.001 | - | - |
| small | grad_sum_solve_backward |
f64 | 1 | 4x4,rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.038 ± 0.000 | 0.052 ± 0.000 | 0.013 ± 0.000 | - | - |
| small | grad_sum_solve_backward |
f64 | 1 | 8x8,rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.039 ± 0.000 | 0.053 ± 0.000 | 0.014 ± 0.000 | - | - |
| small | grad_sum_solve_jvp |
f64 | 1 | 16x16,rhs=1 |
- | - | 0.030 ± 0.000 | 0.266 ± 0.002 | 0.014 ± 0.000 | - | - |
| small | grad_sum_solve_jvp |
f64 | 1 | 2x2,rhs=1 |
- | - | 0.028 ± 0.000 | 0.259 ± 0.002 | 0.011 ± 0.000 | - | - |
| small | grad_sum_solve_jvp |
f64 | 1 | 32x32,rhs=1 |
- | - | 0.038 ± 0.000 | 0.277 ± 0.002 | 0.015 ± 0.000 | - | - |
| small | grad_sum_solve_jvp |
f64 | 1 | 4x4,rhs=1 |
- | - | 0.029 ± 0.000 | 0.257 ± 0.002 | 0.011 ± 0.000 | - | - |
| small | grad_sum_solve_jvp |
f64 | 1 | 8x8,rhs=1 |
- | - | 0.029 ± 0.000 | 0.259 ± 0.001 | 0.012 ± 0.000 | - | - |
| small | grad_sum_solve_vjp |
f64 | 1 | 16x16,rhs=1 |
- | - | 0.030 ± 0.000 | 0.117 ± 0.001 | 0.016 ± 0.000 | - | - |
| small | grad_sum_solve_vjp |
f64 | 1 | 2x2,rhs=1 |
- | - | 0.028 ± 0.000 | 0.112 ± 0.001 | 0.012 ± 0.000 | - | - |
| small | grad_sum_solve_vjp |
f64 | 1 | 32x32,rhs=1 |
- | - | 0.037 ± 0.000 | 0.132 ± 0.002 | 0.017 ± 0.001 | - | - |
| small | grad_sum_solve_vjp |
f64 | 1 | 4x4,rhs=1 |
- | - | 0.028 ± 0.000 | 0.111 ± 0.001 | 0.012 ± 0.000 | - | - |
| small | grad_sum_solve_vjp |
f64 | 1 | 8x8,rhs=1 |
- | - | 0.029 ± 0.000 | 0.114 ± 0.001 | 0.013 ± 0.000 | - | - |
| small | grad_sum_svd_s_backward |
f64 | 1 | 16x16 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.069 ± 0.000 | 0.085 ± 0.000 | 0.057 ± 0.003 | - | - |
| small | grad_sum_svd_s_backward |
f64 | 1 | 2x2 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.026 ± 0.000 | 0.038 ± 0.000 | 0.010 ± 0.000 | - | - |
| small | grad_sum_svd_s_backward |
f64 | 1 | 32x32 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.173 ± 0.001 | 0.203 ± 0.001 | 0.173 ± 0.003 | - | - |
| small | grad_sum_svd_s_backward |
f64 | 1 | 4x4 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.029 ± 0.000 | 0.042 ± 0.000 | 0.013 ± 0.000 | - | - |
| small | grad_sum_svd_s_backward |
f64 | 1 | 8x8 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.040 ± 0.000 | 0.053 ± 0.000 | 0.020 ± 0.003 | - | - |
| small | grad_sum_svd_s_jvp |
f64 | 1 | 16x16 |
- | - | 0.070 ± 0.000 | 0.210 ± 0.005 | 0.053 ± 0.000 | - | - |
| small | grad_sum_svd_s_jvp |
f64 | 1 | 2x2 |
- | - | 0.027 ± 0.000 | 0.147 ± 0.001 | 0.011 ± 0.000 | - | - |
| small | grad_sum_svd_s_jvp |
f64 | 1 | 32x32 |
- | - | 0.174 ± 0.000 | 0.346 ± 0.009 | 0.158 ± 0.001 | - | - |
| small | grad_sum_svd_s_jvp |
f64 | 1 | 4x4 |
- | - | 0.030 ± 0.000 | 0.151 ± 0.001 | 0.013 ± 0.000 | - | - |
| small | grad_sum_svd_s_jvp |
f64 | 1 | 8x8 |
- | - | 0.041 ± 0.000 | 0.167 ± 0.005 | 0.018 ± 0.000 | - | - |
| small | grad_sum_svd_s_vjp |
f64 | 1 | 16x16 |
- | - | 0.069 ± 0.000 | 0.144 ± 0.001 | 0.054 ± 0.004 | - | - |
| small | grad_sum_svd_s_vjp |
f64 | 1 | 2x2 |
- | - | 0.026 ± 0.000 | 0.092 ± 0.001 | 0.011 ± 0.000 | - | - |
| small | grad_sum_svd_s_vjp |
f64 | 1 | 32x32 |
- | - | 0.173 ± 0.001 | 0.269 ± 0.005 | 0.160 ± 0.001 | - | - |
| small | grad_sum_svd_s_vjp |
f64 | 1 | 4x4 |
- | - | 0.030 ± 0.000 | 0.096 ± 0.001 | 0.014 ± 0.000 | - | - |
| small | grad_sum_svd_s_vjp |
f64 | 1 | 8x8 |
- | - | 0.041 ± 0.000 | 0.107 ± 0.001 | 0.018 ± 0.000 | - | - |
| small | matmul |
f64 | 1 | 16x16 |
- | 0.005 ± 0.000 | 0.009 ± 0.000 | 0.003 ± 0.000 | 0.008 ± 0.000 | - | - |
| small | matmul |
f64 | 1 | 2x2 |
- | 0.003 ± 0.000 | 0.009 ± 0.000 | 0.002 ± 0.000 | 0.006 ± 0.001 | - | - |
| small | matmul |
f64 | 1 | 32x32 |
- | 0.007 ± 0.000 | 0.012 ± 0.000 | 0.006 ± 0.000 | 0.008 ± 0.000 | - | - |
| small | matmul |
f64 | 1 | 4x4 |
- | 0.003 ± 0.000 | 0.008 ± 0.000 | 0.002 ± 0.000 | 0.006 ± 0.000 | - | - |
| small | matmul |
f64 | 1 | 64x64 |
- | 0.029 ± 0.002 | 0.029 ± 0.000 | 0.028 ± 0.001 | 0.029 ± 0.003 | - | - |
| small | matmul |
f64 | 1 | 8x8 |
- | 0.004 ± 0.000 | 0.009 ± 0.000 | 0.002 ± 0.000 | 0.008 ± 0.000 | - | - |
| small | qr |
f64 | 1 | 16x16 |
- | 0.010 ± 0.000 | 0.012 ± 0.000 | 0.009 ± 0.000 | 0.015 ± 0.000 | - | - |
| small | qr |
f64 | 1 | 2x2 |
- | 0.006 ± 0.000 | 0.008 ± 0.000 | 0.005 ± 0.000 | 0.009 ± 0.000 | - | - |
| small | qr |
f64 | 1 | 32x32 |
- | 0.022 ± 0.000 | 0.024 ± 0.000 | 0.025 ± 0.000 | 0.027 ± 0.001 | - | - |
| small | qr |
f64 | 1 | 4x4 |
- | 0.007 ± 0.000 | 0.008 ± 0.000 | 0.005 ± 0.000 | 0.009 ± 0.000 | - | - |
| small | qr |
f64 | 1 | 8x8 |
- | 0.007 ± 0.000 | 0.009 ± 0.000 | 0.006 ± 0.000 | 0.011 ± 0.000 | - | - |
| small | solve |
f64 | 1 | 16x16,rhs=1 |
- | 0.008 ± 0.000 | 0.011 ± 0.000 | 0.015 ± 0.000 | 0.010 ± 0.000 | - | - |
| small | solve |
f64 | 1 | 16x16,rhs=4 |
- | 0.009 ± 0.000 | 0.012 ± 0.000 | 0.016 ± 0.000 | 0.011 ± 0.000 | - | - |
| small | solve |
f64 | 1 | 2x2,rhs=1 |
- | 0.006 ± 0.000 | 0.010 ± 0.000 | 0.013 ± 0.000 | 0.007 ± 0.000 | - | - |
| small | solve |
f64 | 1 | 2x2,rhs=4 |
- | 0.006 ± 0.000 | 0.010 ± 0.000 | 0.013 ± 0.000 | 0.008 ± 0.000 | - | - |
| small | solve |
f64 | 1 | 32x32,rhs=1 |
- | 0.014 ± 0.000 | 0.018 ± 0.000 | 0.022 ± 0.000 | 0.012 ± 0.001 | - | - |
| small | solve |
f64 | 1 | 32x32,rhs=4 |
- | 0.015 ± 0.000 | 0.019 ± 0.000 | 0.023 ± 0.000 | 0.017 ± 0.003 | - | - |
| small | solve |
f64 | 1 | 4x4,rhs=1 |
- | 0.006 ± 0.000 | 0.010 ± 0.000 | 0.013 ± 0.000 | 0.008 ± 0.000 | - | - |
| small | solve |
f64 | 1 | 4x4,rhs=4 |
- | 0.006 ± 0.000 | 0.010 ± 0.000 | 0.013 ± 0.000 | 0.008 ± 0.000 | - | - |
| small | solve |
f64 | 1 | 8x8,rhs=1 |
- | 0.007 ± 0.000 | 0.010 ± 0.000 | 0.013 ± 0.000 | 0.008 ± 0.000 | - | - |
| small | solve |
f64 | 1 | 8x8,rhs=4 |
- | 0.007 ± 0.000 | 0.011 ± 0.000 | 0.014 ± 0.000 | 0.009 ± 0.000 | - | - |
| small | svd |
f64 | 1 | 16x16 |
- | 0.047 ± 0.000 | 0.050 ± 0.000 | 0.052 ± 0.000 | 0.057 ± 0.000 | - | - |
| small | svd |
f64 | 1 | 2x2 |
- | 0.008 ± 0.000 | 0.011 ± 0.000 | 0.008 ± 0.000 | 0.012 ± 0.000 | - | - |
| small | svd |
f64 | 1 | 32x32 |
- | 0.151 ± 0.000 | 0.153 ± 0.001 | 0.165 ± 0.001 | 0.171 ± 0.007 | - | - |
| small | svd |
f64 | 1 | 4x4 |
- | 0.011 ± 0.000 | 0.014 ± 0.000 | 0.011 ± 0.000 | 0.014 ± 0.000 | - | - |
| small | svd |
f64 | 1 | 8x8 |
- | 0.020 ± 0.000 | 0.023 ± 0.000 | 0.020 ± 0.000 | 0.022 ± 0.000 | - | - |
Rows with a successful cell more than 10x faster than the slowest successful cell are flagged for operation-specific review. This cross-backend spread check is a warning, not a correctness verdict.
batched/batched_matmul_ikb_kjb_ijb (f64, threads=1, shape=2x2xbatch16 (native batch layout)): tenferro-eager is 18.2x faster than the slowest successful cell (jax-cpu, 0.116 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.batched/batched_matmul_ikb_kjb_ijb (f64, threads=1, shape=2x2xbatch16 (native batch layout)): tenferro-trace is 10.6x faster than the slowest successful cell (jax-cpu, 0.116 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.batched/batched_matmul_ikb_kjb_ijb (f64, threads=1, shape=2x2xbatch64 (native batch layout)): tenferro-eager is 10.7x faster than the slowest successful cell (jax-cpu, 0.118 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.batched/batched_matmul_ikb_kjb_ijb (f64, threads=1, shape=4x4xbatch16 (native batch layout)): tenferro-eager is 18.6x faster than the slowest successful cell (jax-cpu, 0.128 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.batched/batched_matmul_ikb_kjb_ijb (f64, threads=1, shape=4x4xbatch16 (native batch layout)): tenferro-trace is 11.3x faster than the slowest successful cell (jax-cpu, 0.128 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.batched/batched_matmul_ikb_kjb_ijb (f64, threads=1, shape=8x8xbatch16 (native batch layout)): tenferro-eager is 14.3x faster than the slowest successful cell (jax-cpu, 0.119 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.batched/grad_sum_batched_matmul_backward (f64, threads=1, shape=2x2xbatch1024 (native batch layout)): jax-cpu is 14.0x faster than the slowest successful cell (tenferro-trace, 0.325 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=1, shape=16x16): pytorch-cpu is 10.3x faster than the slowest successful cell (jax-cpu, 0.127 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=1, shape=16x16): tenferro-eager is 25.5x faster than the slowest successful cell (jax-cpu, 0.127 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=1, shape=16x16): tenferro-trace is 11.5x faster than the slowest successful cell (jax-cpu, 0.127 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=1, shape=2x2): pytorch-cpu is 10.4x faster than the slowest successful cell (jax-cpu, 0.115 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=1, shape=2x2): tenferro-eager is 34.2x faster than the slowest successful cell (jax-cpu, 0.115 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=1, shape=2x2): tenferro-trace is 11.3x faster than the slowest successful cell (jax-cpu, 0.115 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=1, shape=32x32): tenferro-eager is 17.8x faster than the slowest successful cell (jax-cpu, 0.131 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=1, shape=4x4): pytorch-cpu is 10.5x faster than the slowest successful cell (jax-cpu, 0.117 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=1, shape=4x4): tenferro-eager is 34.6x faster than the slowest successful cell (jax-cpu, 0.117 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=1, shape=4x4): tenferro-trace is 11.5x faster than the slowest successful cell (jax-cpu, 0.117 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=1, shape=8x8): pytorch-cpu is 10.3x faster than the slowest successful cell (jax-cpu, 0.117 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=1, shape=8x8): tenferro-eager is 31.0x faster than the slowest successful cell (jax-cpu, 0.117 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=1, shape=8x8): tenferro-trace is 11.4x faster than the slowest successful cell (jax-cpu, 0.117 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_eigh_jvp (f64, threads=1, shape=2x2): jax-cpu is 10.2x faster than the slowest successful cell (pytorch-cpu, 0.097 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_lu_jvp (f64, threads=1, shape=16x16): jax-cpu is 11.3x faster than the slowest successful cell (pytorch-cpu, 0.175 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_lu_jvp (f64, threads=1, shape=2x2): jax-cpu is 14.5x faster than the slowest successful cell (pytorch-cpu, 0.158 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_lu_jvp (f64, threads=1, shape=4x4): jax-cpu is 14.0x faster than the slowest successful cell (pytorch-cpu, 0.157 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_lu_vjp (f64, threads=1, shape=2x2): jax-cpu is 11.7x faster than the slowest successful cell (pytorch-cpu, 0.127 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_lu_vjp (f64, threads=1, shape=4x4): jax-cpu is 10.6x faster than the slowest successful cell (pytorch-cpu, 0.127 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_qr_jvp (f64, threads=1, shape=2x2): jax-cpu is 10.6x faster than the slowest successful cell (pytorch-cpu, 0.112 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_qr_vjp (f64, threads=1, shape=2x2): jax-cpu is 11.6x faster than the slowest successful cell (pytorch-cpu, 0.118 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_qr_vjp (f64, threads=1, shape=4x4): jax-cpu is 10.2x faster than the slowest successful cell (pytorch-cpu, 0.117 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_qr_vjp (f64, threads=1, shape=8x8): jax-cpu is 11.5x faster than the slowest successful cell (pytorch-cpu, 0.119 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_solve_jvp (f64, threads=1, shape=16x16,rhs=1): jax-cpu is 19.0x faster than the slowest successful cell (pytorch-cpu, 0.266 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_solve_jvp (f64, threads=1, shape=2x2,rhs=1): jax-cpu is 24.3x faster than the slowest successful cell (pytorch-cpu, 0.259 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_solve_jvp (f64, threads=1, shape=32x32,rhs=1): jax-cpu is 18.5x faster than the slowest successful cell (pytorch-cpu, 0.277 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_solve_jvp (f64, threads=1, shape=4x4,rhs=1): jax-cpu is 23.5x faster than the slowest successful cell (pytorch-cpu, 0.257 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_solve_jvp (f64, threads=1, shape=8x8,rhs=1): jax-cpu is 22.3x faster than the slowest successful cell (pytorch-cpu, 0.259 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_svd_s_jvp (f64, threads=1, shape=2x2): jax-cpu is 13.7x faster than the slowest successful cell (pytorch-cpu, 0.147 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_svd_s_jvp (f64, threads=1, shape=4x4): jax-cpu is 11.4x faster than the slowest successful cell (pytorch-cpu, 0.151 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.This independent check derives minimum logical bytes from dtype and shape for materializing public API operations, plus conservative operation FLOPs for reductions, Frobenius norms, matrix products, triangular solves, and recognized dense linalg families. Decorated shapes (->, rhs=, and +) are parsed explicitly. Operations whose touched region cannot be inferred conservatively from the reported shape are left unmodeled. It assumes at most 100 GB/s and 50 GFLOP/s per requested CPU thread, capped at 1000 GB/s and 3200 GFLOP/s. A cell more than 10x faster than the resulting lower bound is flagged. These deliberately generous ceilings are a sanity screen, not a performance gate.
No cell exceeded the conservative physical bound by more than 10x.
Completed einsum measurements retained from the initial collection; CPU ops were recollected after the input-restoration fix. Component commits and original metadata are recorded in run.yaml and einsum_run.yaml. Exact commands: data/results/amd-cpu/cpu/refresh/20261008_044136/commands_remaining.sh.
Source report: data/results/amd-cpu/cpu/einsum/20261008_053424/cpu_ops_report.md.
cpu/cpu_opsamd-cpu20261008_053424Latest run: ./scripts/run_all.sh 4.
This file is generated from one CPU ops run under data/results/amd-cpu/cpu/einsum/20261008_053424.
5cf78c7ec0ad9516dd78bab546d5e7bd42fa3102AMD Ryzen 9 9955HX 16-Core ProcessorAuthenticAMD3211621Linux-7.0.0-38-generic-x86_64-with-glibc2.3944FALSE44444444--xla_cpu_multi_thread_eigen=true --xla_cpu_experimental_ynn_fusion_type= intra_op_parallelism_threads=4system-mklblasmkl2026.0.1/opt/intel/oneapi/mkl/latest/opt/intel/oneapi/mkl/latest/lib/libmkl_rt.somkl, version 2.12.0+cpu, BLAS_INFO mkl, LAPACK_INFO mkl/workspaces/tenferro-benchmark/.venv/lib/python3.12/site-packages/torch/lib/libgomp.so.1JAX: dot backend xla_cpu, version 0.10.1, jaxlib 0.10.1, default backend cpu, LAPACK provider none_detected
CSV: data/results/amd-cpu/cpu/einsum/20261008_053424/cpu_ops_t4_20261008_053424.csv
data/results/amd-cpu/cpu/einsum/20261008_053424/cpu_ops_t4_20261008_053424.mdSmall rows use time/memory-bounded batches, normalized per operation. Tenferro eager/trace share a CPU execution scope outside timing. Eager primal rows reuse a borrowed session entered before sampling; eager backward rows without a borrowed-session API are unsupported. Raw JSONL records contain batch durations and operation counts.
Median ± IQR (ms). Missing backends are shown as -.
| suite | benchmark | dtype | threads | shape | tenferro-rs direct API (ms) | tenferro-rs eager mode (ms) | tenferro-rs trace mode (ms) | PyTorch Python (ms) | JAX Python (XLA CPU) (ms) | Julia (Base/LinearAlgebra) (ms) | Julia (Strided.jl) (ms) |
|---|---|---|---|---|---|---|---|---|---|---|---|
| batched | batched_eigh |
f64 | 4 | 16x16xbatch1024 (native batch layout) |
- | 20.313 ± 0.022 | 20.315 ± 0.036 | 21.490 ± 0.079 | 14.120 ± 0.200 | - | - |
| batched | batched_eigh |
f64 | 4 | 16x16xbatch16 (native batch layout) |
- | 0.305 ± 0.001 | 0.306 ± 0.001 | 0.324 ± 0.001 | 0.311 ± 0.016 | - | - |
| batched | batched_eigh |
f64 | 4 | 16x16xbatch256 (native batch layout) |
- | 5.083 ± 0.014 | 5.072 ± 0.005 | 5.366 ± 0.015 | 3.230 ± 0.679 | - | - |
| batched | batched_eigh |
f64 | 4 | 16x16xbatch64 (native batch layout) |
- | 1.242 ± 0.002 | 1.244 ± 0.002 | 1.333 ± 0.009 | 0.653 ± 0.118 | - | - |
| batched | batched_eigh |
f64 | 4 | 2x2xbatch1024 (native batch layout) |
- | 0.606 ± 0.002 | 0.603 ± 0.002 | 0.403 ± 0.003 | 0.355 ± 0.014 | - | - |
| batched | batched_eigh |
f64 | 4 | 2x2xbatch16 (native batch layout) |
- | 0.016 ± 0.000 | 0.018 ± 0.000 | 0.013 ± 0.000 | 0.014 ± 0.000 | - | - |
| batched | batched_eigh |
f64 | 4 | 2x2xbatch256 (native batch layout) |
- | 0.157 ± 0.000 | 0.157 ± 0.000 | 0.105 ± 0.001 | 0.094 ± 0.003 | - | - |
| batched | batched_eigh |
f64 | 4 | 2x2xbatch64 (native batch layout) |
- | 0.044 ± 0.000 | 0.046 ± 0.000 | 0.031 ± 0.000 | 0.026 ± 0.001 | - | - |
| batched | batched_eigh |
f64 | 4 | 4x4xbatch1024 (native batch layout) |
- | 2.210 ± 0.005 | 2.207 ± 0.007 | 2.055 ± 0.004 | 1.520 ± 0.245 | - | - |
| batched | batched_eigh |
f64 | 4 | 4x4xbatch16 (native batch layout) |
- | 0.041 ± 0.000 | 0.042 ± 0.000 | 0.039 ± 0.000 | 0.041 ± 0.000 | - | - |
| batched | batched_eigh |
f64 | 4 | 4x4xbatch256 (native batch layout) |
- | 0.557 ± 0.001 | 0.558 ± 0.002 | 0.517 ± 0.002 | 0.545 ± 0.018 | - | - |
| batched | batched_eigh |
f64 | 4 | 4x4xbatch64 (native batch layout) |
- | 0.144 ± 0.000 | 0.145 ± 0.000 | 0.134 ± 0.000 | 0.139 ± 0.006 | - | - |
| batched | batched_eigh |
f64 | 4 | 8x8xbatch1024 (native batch layout) |
- | 6.230 ± 0.012 | 6.234 ± 0.020 | 6.230 ± 0.011 | 3.408 ± 0.391 | - | - |
| batched | batched_eigh |
f64 | 4 | 8x8xbatch16 (native batch layout) |
- | 0.102 ± 0.000 | 0.104 ± 0.000 | 0.103 ± 0.000 | 0.120 ± 0.004 | - | - |
| batched | batched_eigh |
f64 | 4 | 8x8xbatch256 (native batch layout) |
- | 1.564 ± 0.005 | 1.559 ± 0.005 | 1.571 ± 0.009 | 0.834 ± 0.161 | - | - |
| batched | batched_eigh |
f64 | 4 | 8x8xbatch64 (native batch layout) |
- | 0.394 ± 0.001 | 0.394 ± 0.000 | 0.397 ± 0.001 | 0.460 ± 0.019 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 4 | 16x16xbatch1024 (native batch layout) |
- | 0.238 ± 0.001 | 0.308 ± 0.009 | 0.436 ± 0.025 | 1.376 ± 0.293 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 4 | 16x16xbatch16 (native batch layout) |
- | 0.010 ± 0.000 | 0.017 ± 0.000 | 0.018 ± 0.001 | 0.118 ± 0.001 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 4 | 16x16xbatch256 (native batch layout) |
- | 0.063 ± 0.001 | 0.108 ± 0.001 | 0.114 ± 0.020 | 0.262 ± 0.052 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 4 | 16x16xbatch64 (native batch layout) |
- | 0.021 ± 0.000 | 0.037 ± 0.000 | 0.035 ± 0.005 | 0.127 ± 0.002 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 4 | 2x2xbatch1024 (native batch layout) |
- | 0.066 ± 0.001 | 0.075 ± 0.001 | 0.034 ± 0.000 | 0.123 ± 0.001 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 4 | 2x2xbatch16 (native batch layout) |
- | 0.007 ± 0.000 | 0.012 ± 0.000 | 0.012 ± 0.000 | 0.116 ± 0.002 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 4 | 2x2xbatch256 (native batch layout) |
- | 0.022 ± 0.000 | 0.027 ± 0.000 | 0.017 ± 0.000 | 0.119 ± 0.001 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 4 | 2x2xbatch64 (native batch layout) |
- | 0.010 ± 0.000 | 0.015 ± 0.000 | 0.013 ± 0.000 | 0.117 ± 0.001 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 4 | 4x4xbatch1024 (native batch layout) |
- | 0.071 ± 0.000 | 0.087 ± 0.000 | 0.063 ± 0.007 | 0.216 ± 0.008 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 4 | 4x4xbatch16 (native batch layout) |
- | 0.007 ± 0.000 | 0.012 ± 0.000 | 0.014 ± 0.000 | 0.119 ± 0.001 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 4 | 4x4xbatch256 (native batch layout) |
- | 0.023 ± 0.000 | 0.030 ± 0.000 | 0.041 ± 0.000 | 0.119 ± 0.001 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 4 | 4x4xbatch64 (native batch layout) |
- | 0.010 ± 0.000 | 0.016 ± 0.000 | 0.019 ± 0.000 | 0.118 ± 0.001 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 4 | 8x8xbatch1024 (native batch layout) |
- | 0.093 ± 0.001 | 0.137 ± 0.001 | 0.105 ± 0.010 | 0.417 ± 0.048 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 4 | 8x8xbatch16 (native batch layout) |
- | 0.007 ± 0.000 | 0.013 ± 0.000 | 0.014 ± 0.000 | 0.119 ± 0.001 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 4 | 8x8xbatch256 (native batch layout) |
- | 0.028 ± 0.000 | 0.044 ± 0.000 | 0.035 ± 0.006 | 0.127 ± 0.001 | - | - |
| batched | batched_matmul_ikb_kjb_ijb |
f64 | 4 | 8x8xbatch64 (native batch layout) |
- | 0.012 ± 0.000 | 0.019 ± 0.000 | 0.017 ± 0.000 | 0.119 ± 0.001 | - | - |
| batched | batched_qr |
f64 | 4 | 16x16xbatch1024 (native batch layout) |
- | 4.248 ± 0.017 | 4.247 ± 0.009 | 4.382 ± 0.033 | 3.485 ± 0.124 | - | - |
| batched | batched_qr |
f64 | 4 | 16x16xbatch16 (native batch layout) |
- | 0.072 ± 0.000 | 0.074 ± 0.001 | 0.075 ± 0.000 | 0.112 ± 0.001 | - | - |
| batched | batched_qr |
f64 | 4 | 16x16xbatch256 (native batch layout) |
- | 1.062 ± 0.005 | 1.067 ± 0.002 | 1.107 ± 0.006 | 0.913 ± 0.063 | - | - |
| batched | batched_qr |
f64 | 4 | 16x16xbatch64 (native batch layout) |
- | 0.270 ± 0.001 | 0.270 ± 0.001 | 0.296 ± 0.035 | 0.292 ± 0.005 | - | - |
| batched | batched_qr |
f64 | 4 | 2x2xbatch1024 (native batch layout) |
- | 0.197 ± 0.001 | 0.197 ± 0.001 | 0.167 ± 0.001 | 0.266 ± 0.009 | - | - |
| batched | batched_qr |
f64 | 4 | 2x2xbatch16 (native batch layout) |
- | 0.009 ± 0.000 | 0.012 ± 0.000 | 0.008 ± 0.000 | 0.012 ± 0.000 | - | - |
| batched | batched_qr |
f64 | 4 | 2x2xbatch256 (native batch layout) |
- | 0.054 ± 0.000 | 0.056 ± 0.000 | 0.046 ± 0.000 | 0.070 ± 0.004 | - | - |
| batched | batched_qr |
f64 | 4 | 2x2xbatch64 (native batch layout) |
- | 0.018 ± 0.000 | 0.021 ± 0.000 | 0.016 ± 0.000 | 0.024 ± 0.000 | - | - |
| batched | batched_qr |
f64 | 4 | 4x4xbatch1024 (native batch layout) |
- | 0.574 ± 0.002 | 0.600 ± 0.002 | 0.496 ± 0.003 | 0.823 ± 0.048 | - | - |
| batched | batched_qr |
f64 | 4 | 4x4xbatch16 (native batch layout) |
- | 0.015 ± 0.000 | 0.017 ± 0.000 | 0.014 ± 0.000 | 0.020 ± 0.000 | - | - |
| batched | batched_qr |
f64 | 4 | 4x4xbatch256 (native batch layout) |
- | 0.148 ± 0.000 | 0.150 ± 0.001 | 0.129 ± 0.001 | 0.204 ± 0.012 | - | - |
| batched | batched_qr |
f64 | 4 | 4x4xbatch64 (native batch layout) |
- | 0.042 ± 0.000 | 0.046 ± 0.000 | 0.037 ± 0.000 | 0.055 ± 0.002 | - | - |
| batched | batched_qr |
f64 | 4 | 8x8xbatch1024 (native batch layout) |
- | 1.376 ± 0.006 | 1.346 ± 0.002 | 1.287 ± 0.007 | 1.062 ± 0.104 | - | - |
| batched | batched_qr |
f64 | 4 | 8x8xbatch16 (native batch layout) |
- | 0.028 ± 0.000 | 0.029 ± 0.000 | 0.027 ± 0.000 | 0.042 ± 0.003 | - | - |
| batched | batched_qr |
f64 | 4 | 8x8xbatch256 (native batch layout) |
- | 0.340 ± 0.001 | 0.346 ± 0.003 | 0.333 ± 0.001 | 0.565 ± 0.017 | - | - |
| batched | batched_qr |
f64 | 4 | 8x8xbatch64 (native batch layout) |
- | 0.090 ± 0.000 | 0.091 ± 0.000 | 0.088 ± 0.000 | 0.152 ± 0.006 | - | - |
| batched | batched_solve |
f64 | 4 | 16x16xbatch1024 (native batch layout),rhs=1 |
- | 1.343 ± 0.005 | 1.398 ± 0.008 | 0.691 ± 0.006 | 1.835 ± 0.144 | - | - |
| batched | batched_solve |
f64 | 4 | 16x16xbatch16 (native batch layout),rhs=1 |
- | 0.030 ± 0.000 | 0.030 ± 0.000 | 0.029 ± 0.001 | 0.050 ± 0.003 | - | - |
| batched | batched_solve |
f64 | 4 | 16x16xbatch256 (native batch layout),rhs=1 |
- | 0.339 ± 0.000 | 0.345 ± 0.002 | 0.195 ± 0.008 | 0.525 ± 0.040 | - | - |
| batched | batched_solve |
f64 | 4 | 16x16xbatch64 (native batch layout),rhs=1 |
- | 0.089 ± 0.000 | 0.089 ± 0.000 | 0.072 ± 0.004 | 0.158 ± 0.014 | - | - |
| batched | batched_solve |
f64 | 4 | 2x2xbatch1024 (native batch layout),rhs=1 |
- | 0.086 ± 0.000 | 0.064 ± 0.000 | 0.070 ± 0.009 | 0.256 ± 0.007 | - | - |
| batched | batched_solve |
f64 | 4 | 2x2xbatch16 (native batch layout),rhs=1 |
- | 0.010 ± 0.000 | 0.011 ± 0.000 | 0.016 ± 0.000 | 0.012 ± 0.000 | - | - |
| batched | batched_solve |
f64 | 4 | 2x2xbatch256 (native batch layout),rhs=1 |
- | 0.028 ± 0.000 | 0.024 ± 0.000 | 0.029 ± 0.000 | 0.069 ± 0.002 | - | - |
| batched | batched_solve |
f64 | 4 | 2x2xbatch64 (native batch layout),rhs=1 |
- | 0.013 ± 0.000 | 0.014 ± 0.000 | 0.019 ± 0.000 | 0.022 ± 0.002 | - | - |
| batched | batched_solve |
f64 | 4 | 4x4xbatch1024 (native batch layout),rhs=1 |
- | 0.147 ± 0.000 | 0.105 ± 0.000 | 0.098 ± 0.003 | 0.389 ± 0.010 | - | - |
| batched | batched_solve |
f64 | 4 | 4x4xbatch16 (native batch layout),rhs=1 |
- | 0.010 ± 0.000 | 0.012 ± 0.000 | 0.017 ± 0.000 | 0.015 ± 0.000 | - | - |
| batched | batched_solve |
f64 | 4 | 4x4xbatch256 (native batch layout),rhs=1 |
- | 0.043 ± 0.000 | 0.034 ± 0.000 | 0.036 ± 0.000 | 0.101 ± 0.004 | - | - |
| batched | batched_solve |
f64 | 4 | 4x4xbatch64 (native batch layout),rhs=1 |
- | 0.017 ± 0.000 | 0.016 ± 0.000 | 0.021 ± 0.000 | 0.031 ± 0.001 | - | - |
| batched | batched_solve |
f64 | 4 | 8x8xbatch1024 (native batch layout),rhs=1 |
- | 0.466 ± 0.002 | 0.429 ± 0.001 | 0.266 ± 0.001 | 0.806 ± 0.039 | - | - |
| batched | batched_solve |
f64 | 4 | 8x8xbatch16 (native batch layout),rhs=1 |
- | 0.015 ± 0.000 | 0.016 ± 0.000 | 0.021 ± 0.000 | 0.021 ± 0.003 | - | - |
| batched | batched_solve |
f64 | 4 | 8x8xbatch256 (native batch layout),rhs=1 |
- | 0.118 ± 0.000 | 0.110 ± 0.000 | 0.090 ± 0.003 | 0.212 ± 0.012 | - | - |
| batched | batched_solve |
f64 | 4 | 8x8xbatch64 (native batch layout),rhs=1 |
- | 0.036 ± 0.000 | 0.035 ± 0.000 | 0.033 ± 0.000 | 0.059 ± 0.003 | - | - |
| batched | batched_svd |
f64 | 4 | 16x16xbatch1024 (native batch layout) |
- | 43.820 ± 0.031 | 43.840 ± 0.127 | 47.367 ± 0.130 | 23.766 ± 10.673 | - | - |
| batched | batched_svd |
f64 | 4 | 16x16xbatch16 (native batch layout) |
- | 0.675 ± 0.002 | 0.682 ± 0.003 | 0.727 ± 0.005 | 0.340 ± 0.052 | - | - |
| batched | batched_svd |
f64 | 4 | 16x16xbatch256 (native batch layout) |
- | 10.935 ± 0.008 | 11.014 ± 0.024 | 11.909 ± 0.023 | 5.096 ± 0.626 | - | - |
| batched | batched_svd |
f64 | 4 | 16x16xbatch64 (native batch layout) |
- | 2.731 ± 0.004 | 2.735 ± 0.004 | 2.969 ± 0.008 | 0.919 ± 0.158 | - | - |
| batched | batched_svd |
f64 | 4 | 2x2xbatch1024 (native batch layout) |
- | 1.307 ± 0.003 | 1.317 ± 0.003 | 1.052 ± 0.002 | 1.322 ± 0.032 | - | - |
| batched | batched_svd |
f64 | 4 | 2x2xbatch16 (native batch layout) |
- | 0.027 ± 0.000 | 0.030 ± 0.000 | 0.024 ± 0.000 | 0.029 ± 0.000 | - | - |
| batched | batched_svd |
f64 | 4 | 2x2xbatch256 (native batch layout) |
- | 0.332 ± 0.000 | 0.336 ± 0.001 | 0.269 ± 0.001 | 0.336 ± 0.017 | - | - |
| batched | batched_svd |
f64 | 4 | 2x2xbatch64 (native batch layout) |
- | 0.089 ± 0.000 | 0.092 ± 0.000 | 0.074 ± 0.000 | 0.087 ± 0.001 | - | - |
| batched | batched_svd |
f64 | 4 | 4x4xbatch1024 (native batch layout) |
- | 4.699 ± 0.011 | 4.722 ± 0.011 | 4.209 ± 0.014 | 1.735 ± 0.245 | - | - |
| batched | batched_svd |
f64 | 4 | 4x4xbatch16 (native batch layout) |
- | 0.080 ± 0.000 | 0.083 ± 0.000 | 0.073 ± 0.000 | 0.073 ± 0.000 | - | - |
| batched | batched_svd |
f64 | 4 | 4x4xbatch256 (native batch layout) |
- | 1.181 ± 0.001 | 1.189 ± 0.001 | 1.057 ± 0.002 | 1.065 ± 0.036 | - | - |
| batched | batched_svd |
f64 | 4 | 4x4xbatch64 (native batch layout) |
- | 0.300 ± 0.000 | 0.303 ± 0.000 | 0.269 ± 0.001 | 0.270 ± 0.006 | - | - |
| batched | batched_svd |
f64 | 4 | 8x8xbatch1024 (native batch layout) |
- | 15.292 ± 0.019 | 15.292 ± 0.055 | 14.464 ± 0.027 | 5.079 ± 1.443 | - | - |
| batched | batched_svd |
f64 | 4 | 8x8xbatch16 (native batch layout) |
- | 0.243 ± 0.001 | 0.248 ± 0.001 | 0.231 ± 0.001 | 0.213 ± 0.009 | - | - |
| batched | batched_svd |
f64 | 4 | 8x8xbatch256 (native batch layout) |
- | 3.819 ± 0.005 | 3.850 ± 0.008 | 3.636 ± 0.013 | 1.069 ± 0.098 | - | - |
| batched | batched_svd |
f64 | 4 | 8x8xbatch64 (native batch layout) |
- | 0.959 ± 0.003 | 0.961 ± 0.003 | 0.911 ± 0.003 | 0.587 ± 0.067 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 4 | 16x16xbatch1024 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 1.122 ± 0.013 | 7.643 ± 0.046 | 3.150 ± 0.553 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 4 | 16x16xbatch16 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.062 ± 0.000 | 0.204 ± 0.001 | 0.045 ± 0.003 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 4 | 16x16xbatch256 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.370 ± 0.002 | 1.941 ± 0.004 | 0.721 ± 0.086 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 4 | 16x16xbatch64 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.127 ± 0.000 | 0.549 ± 0.002 | 0.157 ± 0.010 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 4 | 2x2xbatch1024 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.231 ± 0.001 | 0.151 ± 0.000 | 0.024 ± 0.002 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 4 | 2x2xbatch16 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.040 ± 0.000 | 0.084 ± 0.000 | 0.009 ± 0.000 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 4 | 2x2xbatch256 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.087 ± 0.000 | 0.101 ± 0.001 | 0.017 ± 0.001 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 4 | 2x2xbatch64 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.050 ± 0.000 | 0.087 ± 0.000 | 0.012 ± 0.000 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 4 | 4x4xbatch1024 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.280 ± 0.004 | 0.215 ± 0.004 | 0.665 ± 0.058 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 4 | 4x4xbatch16 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.041 ± 0.000 | 0.088 ± 0.001 | 0.016 ± 0.002 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 4 | 4x4xbatch256 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.099 ± 0.000 | 0.174 ± 0.002 | 0.170 ± 0.007 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 4 | 4x4xbatch64 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.052 ± 0.001 | 0.105 ± 0.001 | 0.047 ± 0.002 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 4 | 8x8xbatch1024 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.476 ± 0.002 | 5.518 ± 0.034 | 1.353 ± 0.208 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 4 | 8x8xbatch16 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.044 ± 0.000 | 0.172 ± 0.001 | 0.024 ± 0.004 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 4 | 8x8xbatch256 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.153 ± 0.001 | 1.443 ± 0.010 | 0.270 ± 0.008 | - | - |
| batched | grad_sum_batched_matmul_backward |
f64 | 4 | 8x8xbatch64 (native batch layout) |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.067 ± 0.001 | 0.429 ± 0.001 | 0.073 ± 0.003 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 4 | 16x16xbatch1024 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 2.240 ± 0.021 | 1.341 ± 0.020 | 3.669 ± 0.434 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 4 | 16x16xbatch16 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.076 ± 0.000 | 0.099 ± 0.001 | 0.080 ± 0.005 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 4 | 16x16xbatch256 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.619 ± 0.002 | 0.441 ± 0.006 | 0.927 ± 0.162 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 4 | 16x16xbatch64 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.182 ± 0.000 | 0.211 ± 0.002 | 0.268 ± 0.017 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 4 | 2x2xbatch1024 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.223 ± 0.001 | 0.164 ± 0.001 | 0.534 ± 0.018 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 4 | 2x2xbatch16 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.044 ± 0.000 | 0.061 ± 0.001 | 0.021 ± 0.000 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 4 | 2x2xbatch256 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.087 ± 0.000 | 0.088 ± 0.002 | 0.136 ± 0.003 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 4 | 2x2xbatch64 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.053 ± 0.000 | 0.066 ± 0.000 | 0.043 ± 0.002 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 4 | 4x4xbatch1024 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.303 ± 0.001 | 0.258 ± 0.002 | 0.817 ± 0.038 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 4 | 4x4xbatch16 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.046 ± 0.000 | 0.063 ± 0.000 | 0.023 ± 0.003 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 4 | 4x4xbatch256 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.106 ± 0.001 | 0.110 ± 0.000 | 0.210 ± 0.007 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 4 | 4x4xbatch64 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.058 ± 0.000 | 0.072 ± 0.001 | 0.060 ± 0.002 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 4 | 8x8xbatch1024 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 1.045 ± 0.002 | 0.666 ± 0.010 | 1.695 ± 0.237 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 4 | 8x8xbatch16 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.057 ± 0.000 | 0.072 ± 0.001 | 0.036 ± 0.002 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 4 | 8x8xbatch256 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.292 ± 0.001 | 0.267 ± 0.002 | 0.384 ± 0.014 | - | - |
| batched | grad_sum_batched_solve_backward |
f64 | 4 | 8x8xbatch64 (native batch layout),rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.102 ± 0.000 | 0.113 ± 0.001 | 0.106 ± 0.005 | - | - |
| large | eigh |
f64 | 4 | 1024x1024 |
- | 90.456 ± 0.288 | 89.951 ± 0.274 | 93.152 ± 2.203 | 112.289 ± 6.261 | - | - |
| large | eigh |
f64 | 4 | 128x128 |
- | 0.801 ± 0.004 | 0.801 ± 0.010 | 0.879 ± 0.014 | 1.481 ± 0.068 | - | - |
| large | eigh |
f64 | 4 | 256x256 |
- | 3.670 ± 0.005 | 3.665 ± 0.035 | 3.859 ± 0.150 | 5.964 ± 0.499 | - | - |
| large | eigh |
f64 | 4 | 512x512 |
- | 16.318 ± 0.081 | 16.293 ± 0.107 | 17.235 ± 0.358 | 24.346 ± 0.306 | - | - |
| large | eigh |
f64 | 4 | 64x64 |
- | 0.217 ± 0.000 | 0.219 ± 0.001 | 0.238 ± 0.002 | 0.383 ± 0.008 | - | - |
| large | grad_sum_eigh_jvp |
f64 | 4 | 1024x1024 |
- | - | 110.595 ± 1.190 | 158.889 ± 0.821 | 142.499 ± 2.814 | - | - |
| large | grad_sum_eigh_jvp |
f64 | 4 | 256x256 |
- | - | 4.346 ± 0.042 | 4.765 ± 0.018 | 7.224 ± 0.444 | - | - |
| large | grad_sum_eigh_jvp |
f64 | 4 | 512x512 |
- | - | 19.898 ± 0.698 | 25.879 ± 0.099 | 30.225 ± 0.888 | - | - |
| large | grad_sum_eigh_vjp |
f64 | 4 | 1024x1024 |
- | - | 110.256 ± 0.489 | 112.010 ± 1.240 | 141.547 ± 1.556 | - | - |
| large | grad_sum_eigh_vjp |
f64 | 4 | 256x256 |
- | - | 4.343 ± 0.053 | 4.113 ± 0.017 | 6.640 ± 0.531 | - | - |
| large | grad_sum_eigh_vjp |
f64 | 4 | 512x512 |
- | - | 20.180 ± 0.389 | 19.726 ± 0.056 | 29.945 ± 1.244 | - | - |
| large | grad_sum_lu_jvp |
f64 | 4 | 1024x1024 |
- | - | 81.038 ± 0.600 | 180.950 ± 0.557 | 70.229 ± 3.164 | - | - |
| large | grad_sum_lu_jvp |
f64 | 4 | 256x256 |
- | - | 1.911 ± 0.025 | 4.312 ± 0.052 | 1.852 ± 0.325 | - | - |
| large | grad_sum_lu_jvp |
f64 | 4 | 512x512 |
- | - | 12.071 ± 1.972 | 30.860 ± 0.381 | 10.272 ± 1.694 | - | - |
| large | grad_sum_lu_vjp |
f64 | 4 | 1024x1024 |
- | - | 79.189 ± 0.214 | 92.926 ± 0.252 | 68.551 ± 1.687 | - | - |
| large | grad_sum_lu_vjp |
f64 | 4 | 256x256 |
- | - | 2.167 ± 0.024 | 2.125 ± 0.016 | 1.753 ± 0.087 | - | - |
| large | grad_sum_lu_vjp |
f64 | 4 | 512x512 |
- | - | 11.894 ± 0.095 | 14.338 ± 0.081 | 10.084 ± 0.540 | - | - |
| large | grad_sum_matmul |
f64 | 4 | 1024x1024 |
- | 16.928 ± 0.137 | 16.718 ± 0.361 | 21.150 ± 0.199 | 16.143 ± 0.184 | - | - |
| large | grad_sum_matmul |
f64 | 4 | 128x128 |
- | 0.075 ± 0.001 | 0.072 ± 0.000 | 0.076 ± 0.010 | 0.091 ± 0.013 | - | - |
| large | grad_sum_matmul |
f64 | 4 | 256x256 |
- | 0.364 ± 0.003 | 0.354 ± 0.008 | 0.367 ± 0.037 | 0.405 ± 0.029 | - | - |
| large | grad_sum_matmul |
f64 | 4 | 512x512 |
- | 2.594 ± 0.055 | 2.464 ± 0.057 | 4.255 ± 0.293 | 2.637 ± 0.221 | - | - |
| large | grad_sum_matmul |
f64 | 4 | 64x64 |
- | 0.030 ± 0.000 | 0.027 ± 0.000 | 0.022 ± 0.002 | 0.030 ± 0.003 | - | - |
| large | grad_sum_matmul_backward |
f64 | 4 | 1024x1024 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 52.029 ± 0.484 | 57.790 ± 0.072 | 48.312 ± 1.548 | - | - |
| large | grad_sum_matmul_backward |
f64 | 4 | 128x128 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.231 ± 0.007 | 0.216 ± 0.008 | 0.184 ± 0.023 | - | - |
| large | grad_sum_matmul_backward |
f64 | 4 | 256x256 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 1.091 ± 0.006 | 1.033 ± 0.031 | 1.057 ± 0.219 | - | - |
| large | grad_sum_matmul_backward |
f64 | 4 | 512x512 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 7.223 ± 0.064 | 9.424 ± 0.188 | 6.895 ± 0.331 | - | - |
| large | grad_sum_matmul_backward |
f64 | 4 | 64x64 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.080 ± 0.000 | 0.080 ± 0.002 | 0.065 ± 0.002 | - | - |
| large | grad_sum_qr_jvp |
f64 | 4 | 1024x1024 |
- | - | 113.904 ± 0.650 | 131.511 ± 0.602 | 112.847 ± 0.720 | - | - |
| large | grad_sum_qr_jvp |
f64 | 4 | 256x256 |
- | - | 3.322 ± 0.030 | 3.344 ± 0.594 | 3.794 ± 0.308 | - | - |
| large | grad_sum_qr_jvp |
f64 | 4 | 512x512 |
- | - | 22.380 ± 0.604 | 25.576 ± 0.157 | 19.457 ± 1.904 | - | - |
| large | grad_sum_qr_vjp |
f64 | 4 | 1024x1024 |
- | - | 111.495 ± 0.286 | 126.004 ± 0.331 | 114.857 ± 2.341 | - | - |
| large | grad_sum_qr_vjp |
f64 | 4 | 256x256 |
- | - | 3.191 ± 0.047 | 3.534 ± 0.046 | 3.846 ± 0.345 | - | - |
| large | grad_sum_qr_vjp |
f64 | 4 | 512x512 |
- | - | 20.052 ± 0.347 | 25.700 ± 0.527 | 20.776 ± 2.378 | - | - |
| large | grad_sum_solve_backward |
f64 | 4 | 1024x1024,rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 9.094 ± 0.106 | 11.418 ± 2.027 | 9.588 ± 0.765 | - | - |
| large | grad_sum_solve_backward |
f64 | 4 | 128x128,rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.168 ± 0.002 | 0.206 ± 0.001 | 0.155 ± 0.001 | - | - |
| large | grad_sum_solve_backward |
f64 | 4 | 256x256,rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.381 ± 0.015 | 0.459 ± 0.019 | 0.627 ± 0.075 | - | - |
| large | grad_sum_solve_backward |
f64 | 4 | 512x512,rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 1.524 ± 0.016 | 2.492 ± 0.190 | 2.535 ± 0.261 | - | - |
| large | grad_sum_solve_backward |
f64 | 4 | 64x64,rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.072 ± 0.000 | 0.089 ± 0.001 | 0.053 ± 0.002 | - | - |
| large | grad_sum_solve_jvp |
f64 | 4 | 1024x1024,rhs=1 |
- | - | 8.244 ± 1.917 | 11.528 ± 0.113 | 9.718 ± 0.888 | - | - |
| large | grad_sum_solve_jvp |
f64 | 4 | 256x256,rhs=1 |
- | - | 0.350 ± 0.002 | 0.624 ± 0.019 | 0.446 ± 0.013 | - | - |
| large | grad_sum_solve_jvp |
f64 | 4 | 512x512,rhs=1 |
- | - | 1.491 ± 0.079 | 3.028 ± 0.098 | 2.864 ± 0.761 | - | - |
| large | grad_sum_solve_vjp |
f64 | 4 | 1024x1024,rhs=1 |
- | - | 7.920 ± 0.100 | 18.763 ± 0.164 | 9.293 ± 0.231 | - | - |
| large | grad_sum_solve_vjp |
f64 | 4 | 256x256,rhs=1 |
- | - | 0.347 ± 0.009 | 0.701 ± 0.018 | 0.450 ± 0.020 | - | - |
| large | grad_sum_solve_vjp |
f64 | 4 | 512x512,rhs=1 |
- | - | 1.437 ± 0.006 | 3.944 ± 0.017 | 2.570 ± 0.497 | - | - |
| large | grad_sum_svd_s_backward |
f64 | 4 | 1024x1024 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 265.586 ± 4.853 | 306.521 ± 4.105 | 361.237 ± 6.286 | - | - |
| large | grad_sum_svd_s_backward |
f64 | 4 | 128x128 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 2.255 ± 0.047 | 2.591 ± 0.029 | 2.707 ± 0.078 | - | - |
| large | grad_sum_svd_s_backward |
f64 | 4 | 256x256 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 10.753 ± 0.041 | 12.119 ± 0.061 | 12.847 ± 0.799 | - | - |
| large | grad_sum_svd_s_backward |
f64 | 4 | 512x512 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 51.535 ± 0.476 | 69.670 ± 2.524 | 75.418 ± 1.222 | - | - |
| large | grad_sum_svd_s_backward |
f64 | 4 | 64x64 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.603 ± 0.004 | 0.659 ± 0.006 | 0.635 ± 0.065 | - | - |
| large | grad_sum_svd_s_jvp |
f64 | 4 | 1024x1024 |
- | - | 265.149 ± 2.585 | 344.704 ± 1.889 | 354.826 ± 6.893 | - | - |
| large | grad_sum_svd_s_jvp |
f64 | 4 | 256x256 |
- | - | 10.761 ± 0.090 | 12.931 ± 0.063 | 12.295 ± 0.613 | - | - |
| large | grad_sum_svd_s_jvp |
f64 | 4 | 512x512 |
- | - | 50.354 ± 1.528 | 68.240 ± 0.589 | 75.044 ± 1.311 | - | - |
| large | grad_sum_svd_s_vjp |
f64 | 4 | 1024x1024 |
- | - | 266.303 ± 5.420 | 280.895 ± 4.921 | 360.760 ± 4.876 | - | - |
| large | grad_sum_svd_s_vjp |
f64 | 4 | 256x256 |
- | - | 10.735 ± 0.024 | 12.043 ± 0.043 | 12.622 ± 1.063 | - | - |
| large | grad_sum_svd_s_vjp |
f64 | 4 | 512x512 |
- | - | 50.639 ± 0.336 | 68.901 ± 1.123 | 75.148 ± 0.982 | - | - |
| large | matmul |
f64 | 4 | 1024x1024 |
- | 14.772 ± 0.045 | 16.309 ± 0.058 | 17.586 ± 0.085 | 15.938 ± 0.087 | - | - |
| large | matmul |
f64 | 4 | 128x128 |
- | 0.062 ± 0.001 | 0.066 ± 0.000 | 0.081 ± 0.008 | 0.077 ± 0.014 | - | - |
| large | matmul |
f64 | 4 | 256x256 |
- | 0.319 ± 0.005 | 0.352 ± 0.002 | 0.386 ± 0.029 | 0.385 ± 0.057 | - | - |
| large | matmul |
f64 | 4 | 512x512 |
- | 2.175 ± 0.012 | 2.361 ± 0.045 | 3.003 ± 0.376 | 2.449 ± 0.267 | - | - |
| large | matmul_rect |
f64 | 4 | 1024x256 * 256x1024 |
- | 3.668 ± 0.007 | 3.885 ± 0.043 | 4.721 ± 0.356 | 4.407 ± 0.155 | - | - |
| large | matmul_rect |
f64 | 4 | 256x1024 * 1024x256 |
- | 1.073 ± 0.004 | 1.241 ± 0.020 | 1.562 ± 0.368 | 1.493 ± 0.235 | - | - |
| large | qr |
f64 | 4 | 1024x1024 |
- | 45.073 ± 0.465 | 44.836 ± 0.214 | 52.003 ± 0.354 | 59.527 ± 7.731 | - | - |
| large | qr |
f64 | 4 | 128x128 |
- | 0.420 ± 0.002 | 0.423 ± 0.006 | 0.412 ± 0.003 | 0.683 ± 0.075 | - | - |
| large | qr |
f64 | 4 | 256x256 |
- | 1.570 ± 0.007 | 1.580 ± 0.038 | 1.734 ± 0.027 | 3.460 ± 0.349 | - | - |
| large | qr |
f64 | 4 | 512x512 |
- | 8.984 ± 0.029 | 9.086 ± 0.018 | 10.307 ± 0.169 | 12.304 ± 0.660 | - | - |
| large | qr |
f64 | 4 | 64x64 |
- | 0.107 ± 0.000 | 0.108 ± 0.001 | 0.114 ± 0.002 | 0.099 ± 0.006 | - | - |
| large | solve |
f64 | 4 | 1024x1024,rhs=1 |
- | 6.948 ± 0.091 | 7.140 ± 0.025 | 9.780 ± 1.760 | 10.195 ± 0.803 | - | - |
| large | solve |
f64 | 4 | 1024x1024,rhs=16 |
- | 7.937 ± 0.055 | 8.122 ± 0.021 | 10.693 ± 1.943 | 9.494 ± 0.319 | - | - |
| large | solve |
f64 | 4 | 1024x1024,rhs=64 |
- | 8.703 ± 0.058 | 9.029 ± 0.100 | 12.180 ± 1.530 | 10.494 ± 1.066 | - | - |
| large | solve |
f64 | 4 | 128x128,rhs=1 |
- | 0.109 ± 0.000 | 0.115 ± 0.000 | 0.141 ± 0.001 | 0.138 ± 0.008 | - | - |
| large | solve |
f64 | 4 | 128x128,rhs=16 |
- | 0.132 ± 0.002 | 0.138 ± 0.001 | 0.164 ± 0.001 | 0.217 ± 0.044 | - | - |
| large | solve |
f64 | 4 | 128x128,rhs=64 |
- | 0.165 ± 0.004 | 0.174 ± 0.002 | 0.206 ± 0.002 | 0.369 ± 0.155 | - | - |
| large | solve |
f64 | 4 | 256x256,rhs=1 |
- | 0.250 ± 0.001 | 0.268 ± 0.002 | 0.353 ± 0.032 | 0.555 ± 0.137 | - | - |
| large | solve |
f64 | 4 | 256x256,rhs=16 |
- | 0.312 ± 0.004 | 0.332 ± 0.011 | 0.379 ± 0.070 | 0.805 ± 0.313 | - | - |
| large | solve |
f64 | 4 | 256x256,rhs=64 |
- | 0.409 ± 0.002 | 0.428 ± 0.007 | 0.486 ± 0.008 | 0.822 ± 0.232 | - | - |
| large | solve |
f64 | 4 | 512x512,rhs=1 |
- | 1.263 ± 0.004 | 1.327 ± 0.012 | 1.737 ± 0.199 | 2.228 ± 0.469 | - | - |
| large | solve |
f64 | 4 | 512x512,rhs=16 |
- | 1.461 ± 0.033 | 1.585 ± 0.005 | 1.822 ± 0.019 | 2.445 ± 0.209 | - | - |
| large | solve |
f64 | 4 | 512x512,rhs=64 |
- | 1.725 ± 0.014 | 1.853 ± 0.013 | 2.179 ± 0.044 | 2.884 ± 0.444 | - | - |
| large | solve |
f64 | 4 | 64x64,rhs=1 |
- | 0.034 ± 0.000 | 0.037 ± 0.000 | 0.041 ± 0.000 | 0.043 ± 0.007 | - | - |
| large | solve |
f64 | 4 | 64x64,rhs=16 |
- | 0.041 ± 0.000 | 0.045 ± 0.001 | 0.058 ± 0.001 | 0.064 ± 0.003 | - | - |
| large | solve |
f64 | 4 | 64x64,rhs=64 |
- | 0.053 ± 0.001 | 0.056 ± 0.001 | 0.067 ± 0.000 | 0.149 ± 0.059 | - | - |
| large | svd |
f64 | 4 | 1024x1024 |
- | 236.986 ± 2.403 | 224.274 ± 1.977 | 261.231 ± 6.794 | 306.170 ± 3.753 | - | - |
| large | svd |
f64 | 4 | 128x128 |
- | 2.135 ± 0.011 | 2.145 ± 0.008 | 2.293 ± 0.012 | 2.743 ± 0.151 | - | - |
| large | svd |
f64 | 4 | 256x256 |
- | 10.099 ± 0.036 | 10.072 ± 0.091 | 11.274 ± 0.063 | 12.331 ± 0.279 | - | - |
| large | svd |
f64 | 4 | 512x512 |
- | 46.421 ± 0.274 | 47.088 ± 0.485 | 50.423 ± 0.178 | 68.531 ± 0.241 | - | - |
| large | svd |
f64 | 4 | 64x64 |
- | 0.538 ± 0.002 | 0.544 ± 0.004 | 0.567 ± 0.005 | 0.604 ± 0.042 | - | - |
| small | eigh |
f64 | 4 | 16x16 |
- | 0.025 ± 0.000 | 0.027 ± 0.000 | 0.026 ± 0.000 | 0.036 ± 0.000 | - | - |
| small | eigh |
f64 | 4 | 2x2 |
- | 0.006 ± 0.000 | 0.009 ± 0.000 | 0.006 ± 0.000 | 0.008 ± 0.000 | - | - |
| small | eigh |
f64 | 4 | 32x32 |
- | 0.082 ± 0.000 | 0.083 ± 0.000 | 0.088 ± 0.000 | 0.106 ± 0.006 | - | - |
| small | eigh |
f64 | 4 | 4x4 |
- | 0.008 ± 0.000 | 0.010 ± 0.000 | 0.008 ± 0.000 | 0.010 ± 0.000 | - | - |
| small | eigh |
f64 | 4 | 8x8 |
- | 0.012 ± 0.000 | 0.014 ± 0.000 | 0.012 ± 0.000 | 0.015 ± 0.000 | - | - |
| small | einsum_ij_jk_ik |
f64 | 4 | 16x16 |
- | 0.005 ± 0.000 | 0.011 ± 0.000 | 0.012 ± 0.000 | 0.118 ± 0.001 | - | - |
| small | einsum_ij_jk_ik |
f64 | 4 | 2x2 |
- | 0.003 ± 0.000 | 0.010 ± 0.000 | 0.011 ± 0.000 | 0.115 ± 0.001 | - | - |
| small | einsum_ij_jk_ik |
f64 | 4 | 32x32 |
- | 0.008 ± 0.000 | 0.014 ± 0.000 | 0.016 ± 0.000 | 0.120 ± 0.002 | - | - |
| small | einsum_ij_jk_ik |
f64 | 4 | 4x4 |
- | 0.003 ± 0.000 | 0.010 ± 0.000 | 0.011 ± 0.000 | 0.118 ± 0.002 | - | - |
| small | einsum_ij_jk_ik |
f64 | 4 | 8x8 |
- | 0.004 ± 0.000 | 0.010 ± 0.000 | 0.012 ± 0.000 | 0.120 ± 0.001 | - | - |
| small | grad_sum_eigh_jvp |
f64 | 4 | 16x16 |
- | - | 0.066 ± 0.000 | 0.123 ± 0.002 | 0.034 ± 0.001 | - | - |
| small | grad_sum_eigh_jvp |
f64 | 4 | 2x2 |
- | - | 0.046 ± 0.000 | 0.095 ± 0.002 | 0.009 ± 0.000 | - | - |
| small | grad_sum_eigh_jvp |
f64 | 4 | 32x32 |
- | - | 0.130 ± 0.000 | 0.202 ± 0.003 | 0.103 ± 0.001 | - | - |
| small | grad_sum_eigh_jvp |
f64 | 4 | 4x4 |
- | - | 0.048 ± 0.000 | 0.097 ± 0.002 | 0.012 ± 0.000 | - | - |
| small | grad_sum_eigh_jvp |
f64 | 4 | 8x8 |
- | - | 0.052 ± 0.000 | 0.104 ± 0.002 | 0.013 ± 0.000 | - | - |
| small | grad_sum_eigh_vjp |
f64 | 4 | 16x16 |
- | - | 0.067 ± 0.000 | 0.106 ± 0.002 | 0.034 ± 0.001 | - | - |
| small | grad_sum_eigh_vjp |
f64 | 4 | 2x2 |
- | - | 0.047 ± 0.000 | 0.083 ± 0.002 | 0.010 ± 0.000 | - | - |
| small | grad_sum_eigh_vjp |
f64 | 4 | 32x32 |
- | - | 0.131 ± 0.000 | 0.176 ± 0.003 | 0.105 ± 0.001 | - | - |
| small | grad_sum_eigh_vjp |
f64 | 4 | 4x4 |
- | - | 0.048 ± 0.000 | 0.084 ± 0.001 | 0.012 ± 0.000 | - | - |
| small | grad_sum_eigh_vjp |
f64 | 4 | 8x8 |
- | - | 0.052 ± 0.000 | 0.088 ± 0.001 | 0.013 ± 0.001 | - | - |
| small | grad_sum_lu_jvp |
f64 | 4 | 16x16 |
- | - | 0.055 ± 0.000 | 0.179 ± 0.005 | 0.017 ± 0.001 | - | - |
| small | grad_sum_lu_jvp |
f64 | 4 | 2x2 |
- | - | 0.049 ± 0.000 | 0.163 ± 0.005 | 0.011 ± 0.000 | - | - |
| small | grad_sum_lu_jvp |
f64 | 4 | 32x32 |
- | - | 0.088 ± 0.001 | 0.220 ± 0.004 | 0.030 ± 0.005 | - | - |
| small | grad_sum_lu_jvp |
f64 | 4 | 4x4 |
- | - | 0.049 ± 0.000 | 0.162 ± 0.003 | 0.012 ± 0.000 | - | - |
| small | grad_sum_lu_jvp |
f64 | 4 | 8x8 |
- | - | 0.050 ± 0.000 | 0.163 ± 0.003 | 0.018 ± 0.004 | - | - |
| small | grad_sum_lu_vjp |
f64 | 4 | 16x16 |
- | - | 0.055 ± 0.000 | 0.140 ± 0.001 | 0.015 ± 0.001 | - | - |
| small | grad_sum_lu_vjp |
f64 | 4 | 2x2 |
- | - | 0.049 ± 0.000 | 0.127 ± 0.001 | 0.011 ± 0.000 | - | - |
| small | grad_sum_lu_vjp |
f64 | 4 | 32x32 |
- | - | 0.094 ± 0.000 | 0.169 ± 0.005 | 0.030 ± 0.005 | - | - |
| small | grad_sum_lu_vjp |
f64 | 4 | 4x4 |
- | - | 0.049 ± 0.000 | 0.129 ± 0.001 | 0.012 ± 0.000 | - | - |
| small | grad_sum_lu_vjp |
f64 | 4 | 8x8 |
- | - | 0.050 ± 0.000 | 0.130 ± 0.001 | 0.016 ± 0.000 | - | - |
| small | grad_sum_matmul_backward |
f64 | 4 | 16x16 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.033 ± 0.000 | 0.037 ± 0.000 | 0.015 ± 0.001 | - | - |
| small | grad_sum_matmul_backward |
f64 | 4 | 2x2 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.030 ± 0.000 | 0.033 ± 0.000 | 0.009 ± 0.000 | - | - |
| small | grad_sum_matmul_backward |
f64 | 4 | 32x32 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.046 ± 0.000 | 0.049 ± 0.000 | 0.016 ± 0.001 | - | - |
| small | grad_sum_matmul_backward |
f64 | 4 | 4x4 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.030 ± 0.000 | 0.034 ± 0.000 | 0.010 ± 0.000 | - | - |
| small | grad_sum_matmul_backward |
f64 | 4 | 8x8 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.030 ± 0.000 | 0.034 ± 0.000 | 0.012 ± 0.001 | - | - |
| small | grad_sum_qr_jvp |
f64 | 4 | 16x16 |
- | - | 0.090 ± 0.000 | 0.128 ± 0.002 | 0.015 ± 0.001 | - | - |
| small | grad_sum_qr_jvp |
f64 | 4 | 2x2 |
- | - | 0.081 ± 0.000 | 0.113 ± 0.002 | 0.011 ± 0.000 | - | - |
| small | grad_sum_qr_jvp |
f64 | 4 | 32x32 |
- | - | 0.121 ± 0.001 | 0.160 ± 0.005 | 0.041 ± 0.001 | - | - |
| small | grad_sum_qr_jvp |
f64 | 4 | 4x4 |
- | - | 0.082 ± 0.000 | 0.112 ± 0.002 | 0.012 ± 0.000 | - | - |
| small | grad_sum_qr_jvp |
f64 | 4 | 8x8 |
- | - | 0.084 ± 0.000 | 0.115 ± 0.002 | 0.016 ± 0.002 | - | - |
| small | grad_sum_qr_vjp |
f64 | 4 | 16x16 |
- | - | 0.093 ± 0.000 | 0.132 ± 0.001 | 0.015 ± 0.001 | - | - |
| small | grad_sum_qr_vjp |
f64 | 4 | 2x2 |
- | - | 0.082 ± 0.000 | 0.118 ± 0.001 | 0.010 ± 0.000 | - | - |
| small | grad_sum_qr_vjp |
f64 | 4 | 32x32 |
- | - | 0.123 ± 0.000 | 0.162 ± 0.004 | 0.041 ± 0.002 | - | - |
| small | grad_sum_qr_vjp |
f64 | 4 | 4x4 |
- | - | 0.084 ± 0.000 | 0.118 ± 0.001 | 0.012 ± 0.000 | - | - |
| small | grad_sum_qr_vjp |
f64 | 4 | 8x8 |
- | - | 0.085 ± 0.000 | 0.119 ± 0.002 | 0.014 ± 0.002 | - | - |
| small | grad_sum_solve_backward |
f64 | 4 | 16x16,rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.040 ± 0.000 | 0.056 ± 0.000 | 0.017 ± 0.001 | - | - |
| small | grad_sum_solve_backward |
f64 | 4 | 2x2,rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.037 ± 0.000 | 0.051 ± 0.000 | 0.012 ± 0.000 | - | - |
| small | grad_sum_solve_backward |
f64 | 4 | 32x32,rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.048 ± 0.000 | 0.064 ± 0.000 | 0.020 ± 0.002 | - | - |
| small | grad_sum_solve_backward |
f64 | 4 | 4x4,rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.037 ± 0.000 | 0.053 ± 0.000 | 0.013 ± 0.000 | - | - |
| small | grad_sum_solve_backward |
f64 | 4 | 8x8,rhs=1 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.039 ± 0.000 | 0.053 ± 0.000 | 0.014 ± 0.000 | - | - |
| small | grad_sum_solve_jvp |
f64 | 4 | 16x16,rhs=1 |
- | - | 0.030 ± 0.000 | 0.263 ± 0.002 | 0.014 ± 0.000 | - | - |
| small | grad_sum_solve_jvp |
f64 | 4 | 2x2,rhs=1 |
- | - | 0.029 ± 0.000 | 0.255 ± 0.001 | 0.011 ± 0.000 | - | - |
| small | grad_sum_solve_jvp |
f64 | 4 | 32x32,rhs=1 |
- | - | 0.038 ± 0.000 | 0.278 ± 0.003 | 0.017 ± 0.003 | - | - |
| small | grad_sum_solve_jvp |
f64 | 4 | 4x4,rhs=1 |
- | - | 0.028 ± 0.000 | 0.256 ± 0.001 | 0.011 ± 0.000 | - | - |
| small | grad_sum_solve_jvp |
f64 | 4 | 8x8,rhs=1 |
- | - | 0.029 ± 0.000 | 0.256 ± 0.002 | 0.012 ± 0.000 | - | - |
| small | grad_sum_solve_vjp |
f64 | 4 | 16x16,rhs=1 |
- | - | 0.030 ± 0.000 | 0.115 ± 0.002 | 0.016 ± 0.000 | - | - |
| small | grad_sum_solve_vjp |
f64 | 4 | 2x2,rhs=1 |
- | - | 0.028 ± 0.000 | 0.111 ± 0.001 | 0.012 ± 0.000 | - | - |
| small | grad_sum_solve_vjp |
f64 | 4 | 32x32,rhs=1 |
- | - | 0.037 ± 0.000 | 0.132 ± 0.001 | 0.017 ± 0.001 | - | - |
| small | grad_sum_solve_vjp |
f64 | 4 | 4x4,rhs=1 |
- | - | 0.028 ± 0.000 | 0.110 ± 0.001 | 0.012 ± 0.000 | - | - |
| small | grad_sum_solve_vjp |
f64 | 4 | 8x8,rhs=1 |
- | - | 0.028 ± 0.000 | 0.112 ± 0.001 | 0.013 ± 0.000 | - | - |
| small | grad_sum_svd_s_backward |
f64 | 4 | 16x16 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.068 ± 0.000 | 0.085 ± 0.000 | 0.060 ± 0.002 | - | - |
| small | grad_sum_svd_s_backward |
f64 | 4 | 2x2 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.026 ± 0.000 | 0.038 ± 0.000 | 0.010 ± 0.000 | - | - |
| small | grad_sum_svd_s_backward |
f64 | 4 | 32x32 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.182 ± 0.000 | 0.212 ± 0.002 | 0.178 ± 0.005 | - | - |
| small | grad_sum_svd_s_backward |
f64 | 4 | 4x4 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.029 ± 0.000 | 0.042 ± 0.000 | 0.014 ± 0.000 | - | - |
| small | grad_sum_svd_s_backward |
f64 | 4 | 8x8 |
- | unsupported: eager backward has no borrowed-session API; setup-inclusive timings disabled | 0.040 ± 0.000 | 0.053 ± 0.000 | 0.021 ± 0.002 | - | - |
| small | grad_sum_svd_s_jvp |
f64 | 4 | 16x16 |
- | - | 0.069 ± 0.000 | 0.206 ± 0.004 | 0.057 ± 0.000 | - | - |
| small | grad_sum_svd_s_jvp |
f64 | 4 | 2x2 |
- | - | 0.027 ± 0.000 | 0.144 ± 0.001 | 0.011 ± 0.000 | - | - |
| small | grad_sum_svd_s_jvp |
f64 | 4 | 32x32 |
- | - | 0.183 ± 0.001 | 0.354 ± 0.003 | 0.163 ± 0.002 | - | - |
| small | grad_sum_svd_s_jvp |
f64 | 4 | 4x4 |
- | - | 0.030 ± 0.000 | 0.149 ± 0.001 | 0.014 ± 0.000 | - | - |
| small | grad_sum_svd_s_jvp |
f64 | 4 | 8x8 |
- | - | 0.041 ± 0.000 | 0.165 ± 0.003 | 0.018 ± 0.001 | - | - |
| small | grad_sum_svd_s_vjp |
f64 | 4 | 16x16 |
- | - | 0.069 ± 0.000 | 0.141 ± 0.001 | 0.057 ± 0.000 | - | - |
| small | grad_sum_svd_s_vjp |
f64 | 4 | 2x2 |
- | - | 0.026 ± 0.000 | 0.090 ± 0.002 | 0.011 ± 0.000 | - | - |
| small | grad_sum_svd_s_vjp |
f64 | 4 | 32x32 |
- | - | 0.182 ± 0.000 | 0.272 ± 0.003 | 0.176 ± 0.003 | - | - |
| small | grad_sum_svd_s_vjp |
f64 | 4 | 4x4 |
- | - | 0.030 ± 0.000 | 0.094 ± 0.001 | 0.015 ± 0.000 | - | - |
| small | grad_sum_svd_s_vjp |
f64 | 4 | 8x8 |
- | - | 0.041 ± 0.000 | 0.104 ± 0.001 | 0.018 ± 0.002 | - | - |
| small | matmul |
f64 | 4 | 16x16 |
- | 0.005 ± 0.000 | 0.009 ± 0.000 | 0.003 ± 0.000 | 0.008 ± 0.001 | - | - |
| small | matmul |
f64 | 4 | 2x2 |
- | 0.003 ± 0.000 | 0.009 ± 0.000 | 0.002 ± 0.000 | 0.006 ± 0.000 | - | - |
| small | matmul |
f64 | 4 | 32x32 |
- | 0.008 ± 0.000 | 0.013 ± 0.000 | 0.007 ± 0.000 | 0.009 ± 0.000 | - | - |
| small | matmul |
f64 | 4 | 4x4 |
- | 0.003 ± 0.000 | 0.009 ± 0.000 | 0.002 ± 0.000 | 0.007 ± 0.000 | - | - |
| small | matmul |
f64 | 4 | 64x64 |
- | 0.027 ± 0.001 | 0.023 ± 0.000 | 0.022 ± 0.002 | 0.023 ± 0.002 | - | - |
| small | matmul |
f64 | 4 | 8x8 |
- | 0.004 ± 0.000 | 0.009 ± 0.000 | 0.002 ± 0.000 | 0.008 ± 0.002 | - | - |
| small | qr |
f64 | 4 | 16x16 |
- | 0.010 ± 0.000 | 0.012 ± 0.000 | 0.011 ± 0.000 | 0.015 ± 0.000 | - | - |
| small | qr |
f64 | 4 | 2x2 |
- | 0.006 ± 0.000 | 0.008 ± 0.000 | 0.005 ± 0.000 | 0.009 ± 0.000 | - | - |
| small | qr |
f64 | 4 | 32x32 |
- | 0.022 ± 0.000 | 0.024 ± 0.000 | 0.025 ± 0.000 | 0.029 ± 0.003 | - | - |
| small | qr |
f64 | 4 | 4x4 |
- | 0.007 ± 0.000 | 0.009 ± 0.000 | 0.006 ± 0.000 | 0.009 ± 0.000 | - | - |
| small | qr |
f64 | 4 | 8x8 |
- | 0.007 ± 0.000 | 0.009 ± 0.000 | 0.007 ± 0.000 | 0.011 ± 0.000 | - | - |
| small | solve |
f64 | 4 | 16x16,rhs=1 |
- | 0.008 ± 0.000 | 0.011 ± 0.000 | 0.015 ± 0.000 | 0.010 ± 0.000 | - | - |
| small | solve |
f64 | 4 | 16x16,rhs=4 |
- | 0.009 ± 0.000 | 0.012 ± 0.000 | 0.016 ± 0.000 | 0.011 ± 0.000 | - | - |
| small | solve |
f64 | 4 | 2x2,rhs=1 |
- | 0.006 ± 0.000 | 0.010 ± 0.000 | 0.013 ± 0.000 | 0.007 ± 0.000 | - | - |
| small | solve |
f64 | 4 | 2x2,rhs=4 |
- | 0.006 ± 0.000 | 0.010 ± 0.000 | 0.014 ± 0.000 | 0.008 ± 0.000 | - | - |
| small | solve |
f64 | 4 | 32x32,rhs=1 |
- | 0.014 ± 0.000 | 0.018 ± 0.000 | 0.022 ± 0.000 | 0.014 ± 0.001 | - | - |
| small | solve |
f64 | 4 | 32x32,rhs=4 |
- | 0.018 ± 0.000 | 0.020 ± 0.000 | 0.025 ± 0.000 | 0.017 ± 0.001 | - | - |
| small | solve |
f64 | 4 | 4x4,rhs=1 |
- | 0.006 ± 0.000 | 0.010 ± 0.000 | 0.013 ± 0.000 | 0.008 ± 0.000 | - | - |
| small | solve |
f64 | 4 | 4x4,rhs=4 |
- | 0.006 ± 0.000 | 0.010 ± 0.000 | 0.013 ± 0.000 | 0.008 ± 0.000 | - | - |
| small | solve |
f64 | 4 | 8x8,rhs=1 |
- | 0.006 ± 0.000 | 0.010 ± 0.000 | 0.013 ± 0.000 | 0.008 ± 0.000 | - | - |
| small | solve |
f64 | 4 | 8x8,rhs=4 |
- | 0.007 ± 0.000 | 0.011 ± 0.000 | 0.014 ± 0.000 | 0.009 ± 0.000 | - | - |
| small | svd |
f64 | 4 | 16x16 |
- | 0.048 ± 0.000 | 0.050 ± 0.000 | 0.052 ± 0.000 | 0.058 ± 0.000 | - | - |
| small | svd |
f64 | 4 | 2x2 |
- | 0.008 ± 0.000 | 0.011 ± 0.000 | 0.008 ± 0.000 | 0.012 ± 0.000 | - | - |
| small | svd |
f64 | 4 | 32x32 |
- | 0.157 ± 0.001 | 0.160 ± 0.000 | 0.172 ± 0.001 | 0.166 ± 0.010 | - | - |
| small | svd |
f64 | 4 | 4x4 |
- | 0.011 ± 0.000 | 0.014 ± 0.000 | 0.011 ± 0.000 | 0.014 ± 0.000 | - | - |
| small | svd |
f64 | 4 | 8x8 |
- | 0.020 ± 0.000 | 0.023 ± 0.000 | 0.020 ± 0.000 | 0.022 ± 0.000 | - | - |
Rows with a successful cell more than 10x faster than the slowest successful cell are flagged for operation-specific review. This cross-backend spread check is a warning, not a correctness verdict.
batched/batched_matmul_ikb_kjb_ijb (f64, threads=4, shape=16x16xbatch16 (native batch layout)): tenferro-eager is 12.1x faster than the slowest successful cell (jax-cpu, 0.118 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.batched/batched_matmul_ikb_kjb_ijb (f64, threads=4, shape=2x2xbatch16 (native batch layout)): tenferro-eager is 16.9x faster than the slowest successful cell (jax-cpu, 0.116 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.batched/batched_matmul_ikb_kjb_ijb (f64, threads=4, shape=2x2xbatch64 (native batch layout)): tenferro-eager is 11.7x faster than the slowest successful cell (jax-cpu, 0.117 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.batched/batched_matmul_ikb_kjb_ijb (f64, threads=4, shape=4x4xbatch16 (native batch layout)): tenferro-eager is 16.7x faster than the slowest successful cell (jax-cpu, 0.119 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.batched/batched_matmul_ikb_kjb_ijb (f64, threads=4, shape=4x4xbatch16 (native batch layout)): tenferro-trace is 10.1x faster than the slowest successful cell (jax-cpu, 0.119 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.batched/batched_matmul_ikb_kjb_ijb (f64, threads=4, shape=4x4xbatch64 (native batch layout)): tenferro-eager is 11.7x faster than the slowest successful cell (jax-cpu, 0.118 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.batched/batched_matmul_ikb_kjb_ijb (f64, threads=4, shape=8x8xbatch16 (native batch layout)): tenferro-eager is 15.9x faster than the slowest successful cell (jax-cpu, 0.119 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.batched/batched_matmul_ikb_kjb_ijb (f64, threads=4, shape=8x8xbatch64 (native batch layout)): tenferro-eager is 10.3x faster than the slowest successful cell (jax-cpu, 0.119 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.batched/grad_sum_batched_matmul_backward (f64, threads=4, shape=8x8xbatch1024 (native batch layout)): tenferro-trace is 11.6x faster than the slowest successful cell (pytorch-cpu, 5.518 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=4, shape=16x16): tenferro-eager is 23.9x faster than the slowest successful cell (jax-cpu, 0.118 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=4, shape=16x16): tenferro-trace is 10.7x faster than the slowest successful cell (jax-cpu, 0.118 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=4, shape=2x2): pytorch-cpu is 10.2x faster than the slowest successful cell (jax-cpu, 0.115 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=4, shape=2x2): tenferro-eager is 34.4x faster than the slowest successful cell (jax-cpu, 0.115 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=4, shape=2x2): tenferro-trace is 11.4x faster than the slowest successful cell (jax-cpu, 0.115 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=4, shape=32x32): tenferro-eager is 15.2x faster than the slowest successful cell (jax-cpu, 0.120 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=4, shape=4x4): pytorch-cpu is 10.5x faster than the slowest successful cell (jax-cpu, 0.118 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=4, shape=4x4): tenferro-eager is 35.0x faster than the slowest successful cell (jax-cpu, 0.118 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=4, shape=4x4): tenferro-trace is 11.4x faster than the slowest successful cell (jax-cpu, 0.118 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=4, shape=8x8): pytorch-cpu is 10.4x faster than the slowest successful cell (jax-cpu, 0.120 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=4, shape=8x8): tenferro-eager is 31.7x faster than the slowest successful cell (jax-cpu, 0.120 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/einsum_ij_jk_ik (f64, threads=4, shape=8x8): tenferro-trace is 11.7x faster than the slowest successful cell (jax-cpu, 0.120 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_eigh_jvp (f64, threads=4, shape=2x2): jax-cpu is 10.0x faster than the slowest successful cell (pytorch-cpu, 0.095 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_lu_jvp (f64, threads=4, shape=16x16): jax-cpu is 10.7x faster than the slowest successful cell (pytorch-cpu, 0.179 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_lu_jvp (f64, threads=4, shape=2x2): jax-cpu is 14.7x faster than the slowest successful cell (pytorch-cpu, 0.163 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_lu_jvp (f64, threads=4, shape=4x4): jax-cpu is 13.7x faster than the slowest successful cell (pytorch-cpu, 0.162 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_lu_vjp (f64, threads=4, shape=2x2): jax-cpu is 11.6x faster than the slowest successful cell (pytorch-cpu, 0.127 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_lu_vjp (f64, threads=4, shape=4x4): jax-cpu is 10.8x faster than the slowest successful cell (pytorch-cpu, 0.129 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_qr_jvp (f64, threads=4, shape=2x2): jax-cpu is 10.6x faster than the slowest successful cell (pytorch-cpu, 0.113 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_qr_vjp (f64, threads=4, shape=2x2): jax-cpu is 11.5x faster than the slowest successful cell (pytorch-cpu, 0.118 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_solve_jvp (f64, threads=4, shape=16x16,rhs=1): jax-cpu is 18.8x faster than the slowest successful cell (pytorch-cpu, 0.263 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_solve_jvp (f64, threads=4, shape=2x2,rhs=1): jax-cpu is 23.4x faster than the slowest successful cell (pytorch-cpu, 0.255 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_solve_jvp (f64, threads=4, shape=32x32,rhs=1): jax-cpu is 16.1x faster than the slowest successful cell (pytorch-cpu, 0.278 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_solve_jvp (f64, threads=4, shape=4x4,rhs=1): jax-cpu is 23.4x faster than the slowest successful cell (pytorch-cpu, 0.256 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_solve_jvp (f64, threads=4, shape=8x8,rhs=1): jax-cpu is 21.7x faster than the slowest successful cell (pytorch-cpu, 0.256 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_svd_s_jvp (f64, threads=4, shape=2x2): jax-cpu is 13.5x faster than the slowest successful cell (pytorch-cpu, 0.144 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.small/grad_sum_svd_s_jvp (f64, threads=4, shape=4x4): jax-cpu is 10.6x faster than the slowest successful cell (pytorch-cpu, 0.149 ms). Audit fixture semantics, synchronization, labels, and operation-specific bandwidth/FLOP bounds.This independent check derives minimum logical bytes from dtype and shape for materializing public API operations, plus conservative operation FLOPs for reductions, Frobenius norms, matrix products, triangular solves, and recognized dense linalg families. Decorated shapes (->, rhs=, and +) are parsed explicitly. Operations whose touched region cannot be inferred conservatively from the reported shape are left unmodeled. It assumes at most 100 GB/s and 50 GFLOP/s per requested CPU thread, capped at 1000 GB/s and 3200 GFLOP/s. A cell more than 10x faster than the resulting lower bound is flagged. These deliberately generous ceilings are a sanity screen, not a performance gate.
No cell exceeded the conservative physical bound by more than 10x.