einsum-opt_size on zen5-cpu

What was measured

Hardware

Every timed cell matches a successful verification census, including operation variants. Each cell is median of process medians ± sample SD of those medians; W is the median within-process sample SD. All values are milliseconds. These are descriptive spreads, not confidence intervals. The between-process spread includes the rotating-order schedule. Each process retains 5 chronological calls; 2 complete process sets are recorded. Spread guide (computed from this run's own rows): per-engine between-start range median 0.65%, worst 6.54% among cells of 1 us or more (a sub-microsecond row is timer and scheduler jitter, not kernel time); 1 of 26 engine cells exceed 5% and carry !; median within-process relative SD 0.42%. A cell whose ±, ! or p is as large as the difference you are reading is not evidence of a regression. no row, so no ratio there is position-balanced). ! marks a cell where some engine's own per-start median range exceeds 5% of its median, and p a balanced cell whose estimated position-effect ratio exceeds 1.05. Flagged cells are retained, not discarded; the between-start variation can exceed the difference being compared. ! between-start range over 5%, p position effect over 5%, dispatch-dominated means the row is under 20 us so 4T/8T pay thread-region start-up). Per-start ratios and every chronological sample are in the run directory named at the top of this page (run*.csv). These describe the recorded revisions; they are not a regression claim.

cpueinsum, float64, 1T (CPU 4)

| case | opt_size (ms) | notes | |---|---| | bin_batched_matmul_b32_m64_n64_k64/n1/b1 | 0.55739 ± 0.00711208 (W 0.00208006) | | | bin_elementwise_mul_2048x2048/n1/b1 | 3.95889 ± 0.0176543 (W 0.0202886) | | | bin_matmul_256/n1/b1 | 1.07156 ± 0.00215314 (W 0.00112115) | | | bin_outer_product_4096/n1/b1 | 6.12403 ± 0.0293647 (W 0.00943961) | | | gm_queen5_5_3.wcsp/n1/b1 | 1610.03 ± 43.8141 (W 9.12662) | | | lm_batch_likelihood_brackets_4_4d/n1/b1 | 23.4018 ± 0.195935 (W 0.0259429) | | | lm_batch_likelihood_sentence_3_12d/n1/b1 | 86.2628 ± 0.232554 (W 0.350499) | | | lm_batch_likelihood_sentence_4_4d/n1/b1 | 27.3002 ± 0.112801 (W 0.0731404) | | | str_matrix_chain_multiplication_100/n1/b1 | 12.604 ± 0.0231641 (W 0.0243412) | | | str_nw_mera_closed_120/n1/b1 | 1688.97 ± 0.838807 (W 2.06144) | | | str_nw_mera_open_26/n1/b1 | 1178.71 ± 1.61889 (W 0.873193) | | | tensornetwork_permutation_focus_step409_316/n1/b1 | 242.53 ± 0.906205 (W 0.252631) | | | tensornetwork_permutation_light_415/n1/b1 | 237.961 ± 1.08615 (W 0.219394) | |

cpueinsum, float64, 4T (CPU 4-7)

| case | opt_size (ms) | notes | |---|---| | bin_batched_matmul_b32_m64_n64_k64/n1/b1 | 0.143487 ± 8.55599e-05 (W 0.00157028) | | | bin_elementwise_mul_2048x2048/n1/b1 | 2.13834 ± 0.00861468 (W 0.00983378) | | | bin_matmul_256/n1/b1 | 0.300314 ± 0.00139512 (W 0.00276267) | | | bin_outer_product_4096/n1/b1 | 4.20443 ± 0.0734139 (W 0.0261175) | | | gm_queen5_5_3.wcsp/n1/b1 | 529.766 ± 0.523006 (W 1.44494) | | | lm_batch_likelihood_brackets_4_4d/n1/b1 | 10.698 ± 0.103687 (W 0.0790943) | | | lm_batch_likelihood_sentence_3_12d/n1/b1 | 27.3832 ± 0.440866 (W 1.89911) | | | lm_batch_likelihood_sentence_4_4d/n1/b1 | 12.5424 ± 0.277788 (W 0.202493) | | | str_matrix_chain_multiplication_100/n1/b1 | 6.60194 ± 0.18446 (W 0.0791502) | | | str_nw_mera_closed_120/n1/b1 | 508.828 ± 1.82644 (W 5.02607) | | | str_nw_mera_open_26/n1/b1 | 323.745 ± 2.77233 (W 3.75721) | | | tensornetwork_permutation_focus_step409_316/n1/b1 | 88.3649 ± 0.781256 (W 1.24549) | | | tensornetwork_permutation_light_415/n1/b1 | 78.2631 ± 3.6189 (W 3.39598) | ! between-start range over 5% |

Repeated process sets describe observed repeatability under the recorded order schedule; they are not a paired baseline/candidate experiment and do not establish equal performance between engines.