← CPU results

cpu/einsum

profile: zen5-cpuhardware: x86_64date: 2026-10-28 00:16:12Ztenferro: 8fec1ca70852unit: msthreads: [1, 4, 8]

Legacy report: ingestion provenance is unrecorded; no raw artifact is inferred.

Compute-bound cases (144)

Absolute medians on a log time axis. Bars are shorter for faster calls; the dashed line is the best reference for each case.

casegroupvs PyTorchvs JAXnot measuredflags
bin_batched_matmul_b32_m128_n128_k128Threads: 1 / opt_flopseager 0.90x trace 1.69xeager 1.01x trace 1.88x--
bin_batched_matmul_b32_m128_n128_k128Threads: 1 / opt_sizeeager 0.90x trace 1.69xeager 1.01x trace 1.88x--
bin_batched_matmul_b32_m128_n128_k128Threads: 4 / opt_flopseager 3.29x trace 7.08xeager 1.77x trace 3.81x--
bin_batched_matmul_b32_m128_n128_k128Threads: 4 / opt_sizeeager 3.29x trace 7.08xeager 1.77x trace 3.81x--
bin_batched_matmul_b32_m128_n128_k128Threads: 8 / opt_flopseager 3.76x trace 6.07xeager 1.95x trace 3.14x--
bin_batched_matmul_b32_m128_n128_k128Threads: 8 / opt_sizeeager 3.76x trace 6.07xeager 1.95x trace 3.14x--
bin_batched_matmul_b32_m64_n64_k64Threads: 1 / opt_flopseager 0.92x trace 1.15xeager 0.90x trace 1.13x--
bin_batched_matmul_b32_m64_n64_k64Threads: 1 / opt_sizeeager 0.92x trace 1.15xeager 0.90x trace 1.13x--
bin_batched_matmul_b32_m64_n64_k64Threads: 4 / opt_flopseager 8.03x trace 17.32xeager 1.76x trace 3.79x--
bin_batched_matmul_b32_m64_n64_k64Threads: 4 / opt_sizeeager 8.03x trace 17.32xeager 1.76x trace 3.79x--
bin_batched_matmul_b32_m64_n64_k64Threads: 8 / opt_flopseager 13.65x trace 23.71xeager 1.43x trace 2.49x--
bin_batched_matmul_b32_m64_n64_k64Threads: 8 / opt_sizeeager 13.65x trace 23.71xeager 1.43x trace 2.49x--
bin_batched_outer_product_compact_j16_k16_o64_t64Threads: 1 / opt_flopseager 1.62x trace 1.81xeager 0.59x trace 0.66x--
bin_batched_outer_product_compact_j16_k16_o64_t64Threads: 1 / opt_sizeeager 1.62x trace 1.81xeager 0.59x trace 0.66x--
bin_batched_outer_product_compact_j16_k16_o64_t64Threads: 4 / opt_flopseager 19.03x trace 25.55xeager 3.47x trace 4.66x--
bin_batched_outer_product_compact_j16_k16_o64_t64Threads: 4 / opt_sizeeager 19.03x trace 25.55xeager 3.47x trace 4.66x--
bin_batched_outer_product_compact_j16_k16_o64_t64Threads: 8 / opt_flopseager 26.37x trace 39.67xeager 1.81x trace 2.72x--
bin_batched_outer_product_compact_j16_k16_o64_t64Threads: 8 / opt_sizeeager 26.37x trace 39.67xeager 1.81x trace 2.72x--
bin_batched_outer_product_noncompact_j16_k16_o64_t64Threads: 1 / opt_flopseager 1.62x trace 6.55xeager 0.61x trace 2.47x--
bin_batched_outer_product_noncompact_j16_k16_o64_t64Threads: 1 / opt_sizeeager 1.62x trace 6.55xeager 0.61x trace 2.47x--
bin_batched_outer_product_noncompact_j16_k16_o64_t64Threads: 4 / opt_flopseager 20.07x trace 11.81xeager 2.57x trace 1.51x--
bin_batched_outer_product_noncompact_j16_k16_o64_t64Threads: 4 / opt_sizeeager 20.07x trace 11.81xeager 2.57x trace 1.51x--
bin_batched_outer_product_noncompact_j16_k16_o64_t64Threads: 8 / opt_flopseager 13.50x trace 27.97xeager 0.85x trace 1.77x--
bin_batched_outer_product_noncompact_j16_k16_o64_t64Threads: 8 / opt_sizeeager 13.50x trace 27.97xeager 0.85x trace 1.77x--
bin_elementwise_mul_2048x2048Threads: 1 / opt_flopseager 0.24x trace 0.25xeager 0.16x trace 0.17x--
bin_elementwise_mul_2048x2048Threads: 1 / opt_sizeeager 0.24x trace 0.25xeager 0.16x trace 0.17x--
bin_elementwise_mul_2048x2048Threads: 4 / opt_flopseager 0.47x trace 0.61xeager 0.29x trace 0.37x--
bin_elementwise_mul_2048x2048Threads: 4 / opt_sizeeager 0.47x trace 0.61xeager 0.29x trace 0.37x--
bin_elementwise_mul_2048x2048Threads: 8 / opt_flopseager 0.75x trace 0.73xeager 0.40x trace 0.39x--
bin_elementwise_mul_2048x2048Threads: 8 / opt_sizeeager 0.75x trace 0.73xeager 0.40x trace 0.39x--
bin_elementwise_mul_512x512Threads: 1 / opt_flopseager 0.96x trace 0.98xeager 0.40x trace 0.41x--
bin_elementwise_mul_512x512Threads: 1 / opt_sizeeager 0.96x trace 0.98xeager 0.40x trace 0.41x--
bin_elementwise_mul_512x512Threads: 4 / opt_flopseager 24.47x trace 36.95xeager 2.09x trace 3.16x--
bin_elementwise_mul_512x512Threads: 4 / opt_sizeeager 24.47x trace 36.95xeager 2.09x trace 3.16x--
bin_elementwise_mul_512x512Threads: 8 / opt_flopseager 51.67x trace 116.57xeager 4.31x trace 9.71x--
bin_elementwise_mul_512x512Threads: 8 / opt_sizeeager 51.67x trace 116.57xeager 4.31x trace 9.71x--
bin_matmul_1024Threads: 1 / opt_flopseager 0.86x trace 1.04xeager 0.97x trace 1.17x--
bin_matmul_1024Threads: 1 / opt_sizeeager 0.86x trace 1.04xeager 0.97x trace 1.17x--
bin_matmul_1024Threads: 4 / opt_flopseager 1.10x trace 1.21xeager 1.85x trace 2.03x--
bin_matmul_1024Threads: 4 / opt_sizeeager 1.10x trace 1.21xeager 1.85x trace 2.03x--
bin_matmul_1024Threads: 8 / opt_flopseager 1.11x trace 1.46xeager 1.58x trace 2.08x--
bin_matmul_1024Threads: 8 / opt_sizeeager 1.11x trace 1.46xeager 1.58x trace 2.08x--
bin_matmul_256Threads: 1 / opt_flopseager 0.66x trace 0.82xeager 0.87x trace 1.08x--
bin_matmul_256Threads: 1 / opt_sizeeager 0.66x trace 0.82xeager 0.87x trace 1.08x--
bin_matmul_256Threads: 4 / opt_flopseager 0.40x trace 2.56xeager 0.61x trace 3.93x--
bin_matmul_256Threads: 4 / opt_sizeeager 0.40x trace 2.56xeager 0.61x trace 3.93x--
bin_matmul_256Threads: 8 / opt_flopseager 0.40x trace 6.22xeager 0.51x trace 8.02x--
bin_matmul_256Threads: 8 / opt_sizeeager 0.40x trace 6.22xeager 0.51x trace 8.02x--
bin_omeinsum_batched_matmul_8x8_batch_4Threads: 1 / opt_flopseager 0.27x trace 0.45xeager 0.05x trace 0.08x--
bin_omeinsum_batched_matmul_8x8_batch_4Threads: 1 / opt_sizeeager 0.27x trace 0.45xeager 0.05x trace 0.08x--
bin_omeinsum_batched_matmul_8x8_batch_4Threads: 4 / opt_flopseager 0.25x trace 0.42xeager 0.05x trace 0.08x--
bin_omeinsum_batched_matmul_8x8_batch_4Threads: 4 / opt_sizeeager 0.25x trace 0.42xeager 0.05x trace 0.08x--
bin_omeinsum_batched_matmul_8x8_batch_4Threads: 8 / opt_flopseager 0.28x trace 0.44xeager 0.05x trace 0.08x--
bin_omeinsum_batched_matmul_8x8_batch_4Threads: 8 / opt_sizeeager 0.28x trace 0.44xeager 0.05x trace 0.08x--
bin_omeinsum_high_d_12x12_contract_4_batch_4Threads: 1 / opt_flopseager 0.30x trace 0.33xeager 0.18x trace 0.21x--
bin_omeinsum_high_d_12x12_contract_4_batch_4Threads: 1 / opt_sizeeager 0.30x trace 0.33xeager 0.18x trace 0.21x--
bin_omeinsum_high_d_12x12_contract_4_batch_4Threads: 4 / opt_flopseager 0.31x trace 0.33xeager 0.16x trace 0.17x--
bin_omeinsum_high_d_12x12_contract_4_batch_4Threads: 4 / opt_sizeeager 0.31x trace 0.33xeager 0.16x trace 0.17x--
bin_omeinsum_high_d_12x12_contract_4_batch_4Threads: 8 / opt_flopseager 0.31x trace 0.33xeager 0.18x trace 0.20x--
bin_omeinsum_high_d_12x12_contract_4_batch_4Threads: 8 / opt_sizeeager 0.31x trace 0.33xeager 0.18x trace 0.20x--
bin_omeinsum_matmul_10x10Threads: 1 / opt_flopseager 0.45x trace 0.91xeager 0.11x trace 0.22x--
bin_omeinsum_matmul_10x10Threads: 1 / opt_sizeeager 0.45x trace 0.91xeager 0.11x trace 0.22x--
bin_omeinsum_matmul_10x10Threads: 4 / opt_flopseager 0.55x trace 0.91xeager 0.15x trace 0.26x--
bin_omeinsum_matmul_10x10Threads: 4 / opt_sizeeager 0.55x trace 0.91xeager 0.15x trace 0.26x--
bin_omeinsum_matmul_10x10Threads: 8 / opt_flopseager 0.53x trace 2.71xeager 0.14x trace 0.73x--
bin_omeinsum_matmul_10x10Threads: 8 / opt_sizeeager 0.53x trace 2.71xeager 0.14x trace 0.73x--
bin_outer_product_4096Threads: 1 / opt_flopseager 1.03x trace 1.03xeager 1.00x trace 1.00x--
bin_outer_product_4096Threads: 1 / opt_sizeeager 1.03x trace 1.03xeager 1.00x trace 1.00x--
bin_outer_product_4096Threads: 4 / opt_flopseager 1.52x trace 1.48xeager 0.88x trace 0.86x--
bin_outer_product_4096Threads: 4 / opt_sizeeager 1.52x trace 1.48xeager 0.88x trace 0.86x--
bin_outer_product_4096Threads: 8 / opt_flopseager 1.21x trace 1.26xeager 0.72x trace 0.75x--
bin_outer_product_4096Threads: 8 / opt_sizeeager 1.21x trace 1.26xeager 0.72x trace 0.75x--
bin_permuted_r4_abcd_dbef_acef_d32Threads: 1 / opt_flopseager 0.89x trace 1.11xeager 0.89x trace 1.11x--
bin_permuted_r4_abcd_dbef_acef_d32Threads: 1 / opt_sizeeager 0.89x trace 1.11xeager 0.89x trace 1.11x--
bin_permuted_r4_abcd_dbef_acef_d32Threads: 4 / opt_flopseager 1.43x trace 1.38xeager 1.50x trace 1.44x--
bin_permuted_r4_abcd_dbef_acef_d32Threads: 4 / opt_sizeeager 1.43x trace 1.38xeager 1.50x trace 1.44x--
bin_permuted_r4_abcd_dbef_acef_d32Threads: 8 / opt_flopseager 1.38x trace 2.08xeager 1.53x trace 2.32x--
bin_permuted_r4_abcd_dbef_acef_d32Threads: 8 / opt_sizeeager 1.38x trace 2.08xeager 1.53x trace 2.32x--
gm_queen5_5_3.wcspThreads: 1 / opt_flopseager 1.07x trace 1.28xeager 1.37x trace 1.64x--
gm_queen5_5_3.wcspThreads: 1 / opt_sizeeager 1.08x trace 1.01xeager 1.49x trace 1.39x--
gm_queen5_5_3.wcspThreads: 4 / opt_flopseager 1.73x trace 2.65xeager 1.37x trace 2.10x--
gm_queen5_5_3.wcspThreads: 4 / opt_sizeeager 1.64x trace 1.94xeager 1.48x trace 1.75x--
gm_queen5_5_3.wcspThreads: 8 / opt_flopseager 1.78x trace 3.22xeager 1.42x trace 2.56x--
gm_queen5_5_3.wcspThreads: 8 / opt_sizeeager 1.84x trace 2.40xeager 1.63x trace 2.13x--
lm_batch_likelihood_brackets_4_4dThreads: 1 / opt_flopseager 1.13x trace 0.77xeager 1.28x trace 0.87x--
lm_batch_likelihood_brackets_4_4dThreads: 1 / opt_sizeeager 1.12x trace 0.75xeager 1.36x trace 0.91x--
lm_batch_likelihood_brackets_4_4dThreads: 4 / opt_flopseager 3.54x trace 2.75xeager 1.89x trace 1.47x--
lm_batch_likelihood_brackets_4_4dThreads: 4 / opt_sizeeager 4.65x trace 3.51xeager 2.42x trace 1.82x--
lm_batch_likelihood_brackets_4_4dThreads: 8 / opt_flopseager 6.27x trace 3.37xeager 2.95x trace 1.58x--
lm_batch_likelihood_brackets_4_4dThreads: 8 / opt_sizeeager 7.24x trace 4.53xeager 3.29x trace 2.06x--
lm_batch_likelihood_sentence_3_12dThreads: 1 / opt_flopseager 1.31x trace 0.70xeager 1.45x trace 0.78x--
lm_batch_likelihood_sentence_3_12dThreads: 1 / opt_sizeeager 1.97x trace 1.07xeager 1.28x trace 0.70x--
lm_batch_likelihood_sentence_3_12dThreads: 4 / opt_flopseager 3.60x trace 1.62xeager 3.06x trace 1.38x--
lm_batch_likelihood_sentence_3_12dThreads: 4 / opt_sizeeager 4.05x trace 1.02xeager 4.34x trace 1.10x--
lm_batch_likelihood_sentence_3_12dThreads: 8 / opt_flopseager 4.51x trace 1.18xeager 5.55x trace 1.45x--
lm_batch_likelihood_sentence_3_12dThreads: 8 / opt_sizeeager 9.60x trace 1.60xeager 5.61x trace 0.93x--
lm_batch_likelihood_sentence_4_4dThreads: 1 / opt_flopseager 1.20x trace 0.75xeager 1.43x trace 0.89x--
lm_batch_likelihood_sentence_4_4dThreads: 1 / opt_sizeeager 1.31x trace 0.75xeager 1.70x trace 0.98x--
lm_batch_likelihood_sentence_4_4dThreads: 4 / opt_flopseager 6.19x trace 3.67xeager 3.47x trace 2.06x--
lm_batch_likelihood_sentence_4_4dThreads: 4 / opt_sizeeager 4.63x trace 2.08xeager 2.83x trace 1.27x--
lm_batch_likelihood_sentence_4_4dThreads: 8 / opt_flopseager 5.87x trace 3.24xeager 2.95x trace 1.63x--
lm_batch_likelihood_sentence_4_4dThreads: 8 / opt_sizeeager 6.25x trace 4.46xeager 3.19x trace 2.28x--
nary_matmul_chain_64Threads: 1 / opt_flopseager 0.74x trace 0.90xeager 0.32x trace 0.38x--
nary_matmul_chain_64Threads: 1 / opt_sizeeager 0.74x trace 0.90xeager 0.32x trace 0.38x--
nary_matmul_chain_64Threads: 4 / opt_flopseager 0.64x trace 2.52xeager 0.31x trace 1.24x--
nary_matmul_chain_64Threads: 4 / opt_sizeeager 0.64x trace 2.52xeager 0.31x trace 1.24x--
nary_matmul_chain_64Threads: 8 / opt_flopseager 2.48x trace 3.00xeager 0.91x trace 1.09x--
nary_matmul_chain_64Threads: 8 / opt_sizeeager 2.48x trace 3.00xeager 0.91x trace 1.09x--
str_matrix_chain_multiplication_100Threads: 1 / opt_flopseager 0.95x trace 0.91xeager 0.81x trace 0.78x--
str_matrix_chain_multiplication_100Threads: 1 / opt_sizeeager 0.95x trace 0.93xeager 0.81x trace 0.79x--
str_matrix_chain_multiplication_100Threads: 4 / opt_flopseager 0.94x trace 1.81xeager 0.64x trace 1.23x--
str_matrix_chain_multiplication_100Threads: 4 / opt_sizeeager 0.93x trace 0.89xeager 0.64x trace 0.61x--
str_matrix_chain_multiplication_100Threads: 8 / opt_flopseager 2.10x trace 1.99xeager 1.52x trace 1.44x--
str_matrix_chain_multiplication_100Threads: 8 / opt_sizeeager 2.10x trace 2.10xeager 1.53x trace 1.53x--
str_mps_varying_inner_product_200Threads: 1 / opt_flopseager 1.02x trace 0.86xeager 0.72x trace 0.60x--
str_mps_varying_inner_product_200Threads: 1 / opt_sizeeager 1.05x trace 1.04xeager 0.71x trace 0.70x--
str_mps_varying_inner_product_200Threads: 4 / opt_flopseager 1.11x trace 1.56xeager 0.87x trace 1.23x--
str_mps_varying_inner_product_200Threads: 4 / opt_sizeeager 1.20x trace 1.25xeager 0.99x trace 1.04x--
str_mps_varying_inner_product_200Threads: 8 / opt_flopseager 1.37x trace 1.58xeager 0.94x trace 1.08x--
str_mps_varying_inner_product_200Threads: 8 / opt_sizeeager 1.80x trace 1.57xeager 1.25x trace 1.09x--
str_nw_mera_closed_120Threads: 1 / opt_flopseager 0.93x trace 0.90xeager 0.88x trace 0.85x--
str_nw_mera_closed_120Threads: 1 / opt_sizeeager 1.00x trace 1.02xeager 0.95x trace 0.96x--
str_nw_mera_closed_120Threads: 4 / opt_flopseager 1.14x trace 1.08xeager 1.23x trace 1.16x--
str_nw_mera_closed_120Threads: 4 / opt_sizeeager 1.26x trace 1.31xeager 1.34x trace 1.38x--
str_nw_mera_closed_120Threads: 8 / opt_flopseager 1.27x trace 1.17xeager 1.38x trace 1.27x--
str_nw_mera_closed_120Threads: 8 / opt_sizeeager 1.37x trace 1.46xeager 1.54x trace 1.65x--
str_nw_mera_open_26Threads: 1 / opt_flopseager 1.24x trace 1.16xeager 0.88x trace 0.82x--
str_nw_mera_open_26Threads: 1 / opt_sizeeager 1.19x trace 1.12xeager 0.86x trace 0.81x--
str_nw_mera_open_26Threads: 4 / opt_flopseager 1.98x trace 1.76xeager 1.40x trace 1.25x--
str_nw_mera_open_26Threads: 4 / opt_sizeeager 1.87x trace 1.68xeager 1.36x trace 1.22x--
str_nw_mera_open_26Threads: 8 / opt_flopseager 1.95x trace 2.02xeager 1.56x trace 1.61x--
str_nw_mera_open_26Threads: 8 / opt_sizeeager 1.92x trace 2.03xeager 1.49x trace 1.57x--
tensornetwork_permutation_focus_step409_316Threads: 1 / opt_flopseager 0.73x trace 0.56xeager 0.82x trace 0.62x--
tensornetwork_permutation_focus_step409_316Threads: 1 / opt_sizeeager 0.73x trace 0.56xeager 0.82x trace 0.62x--
tensornetwork_permutation_focus_step409_316Threads: 4 / opt_flopseager 1.62x trace 1.44xeager 1.08x trace 0.96x--
tensornetwork_permutation_focus_step409_316Threads: 4 / opt_sizeeager 1.62x trace 1.44xeager 1.08x trace 0.96x--
tensornetwork_permutation_focus_step409_316Threads: 8 / opt_flopseager 2.14x trace 1.82xeager 0.99x trace 0.84x--
tensornetwork_permutation_focus_step409_316Threads: 8 / opt_sizeeager 2.14x trace 1.82xeager 0.99x trace 0.84x--
tensornetwork_permutation_light_415Threads: 1 / opt_flopseager 0.59x trace 0.57xeager 0.63x trace 0.60x--
tensornetwork_permutation_light_415Threads: 1 / opt_sizeeager 0.59x trace 0.57xeager 0.63x trace 0.60x--
tensornetwork_permutation_light_415Threads: 4 / opt_flopseager 1.44x trace 1.43xeager 0.91x trace 0.91x--
tensornetwork_permutation_light_415Threads: 4 / opt_sizeeager 1.44x trace 1.43xeager 0.91x trace 0.91x--
tensornetwork_permutation_light_415Threads: 8 / opt_flopseager 1.82x trace 1.73xeager 0.80x trace 0.76x--
tensornetwork_permutation_light_415Threads: 8 / opt_sizeeager 1.82x trace 1.73xeager 0.80x trace 0.76x--

Charts below are ordered best first (smallest ratio).

bin_elementwise_mul_2048x2048 Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (6.445 ms)pytorch-cpu6.445 msomeinsum_path7.190 msjax-cpu9.317 mstenferro-eager1.527 ms · 0.24xtenferro-trace1.607 ms · 0.25x

bin_elementwise_mul_2048x2048 Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (6.445 ms)pytorch-cpu6.445 msomeinsum_path7.190 msjax-cpu9.317 mstenferro-eager1.527 ms · 0.24xtenferro-trace1.607 ms · 0.25x

bin_omeinsum_batched_matmul_8x8_batch_4 Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.012 ms)omeinsum_path0.006 mspytorch-cpu0.012 msjax-cpu0.064 mstenferro-eager0.003 ms · 0.25xtenferro-trace0.005 ms · 0.42x

bin_omeinsum_batched_matmul_8x8_batch_4 Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.012 ms)omeinsum_path0.006 mspytorch-cpu0.012 msjax-cpu0.064 mstenferro-eager0.003 ms · 0.25xtenferro-trace0.005 ms · 0.42x

bin_omeinsum_batched_matmul_8x8_batch_4 Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.011 ms)omeinsum_path0.007 mspytorch-cpu0.011 msjax-cpu0.064 mstenferro-eager0.003 ms · 0.27xtenferro-trace0.005 ms · 0.45x

bin_omeinsum_batched_matmul_8x8_batch_4 Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.011 ms)omeinsum_path0.007 mspytorch-cpu0.011 msjax-cpu0.064 mstenferro-eager0.003 ms · 0.27xtenferro-trace0.005 ms · 0.45x

bin_omeinsum_batched_matmul_8x8_batch_4 Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.018 ms)omeinsum_path0.01 mspytorch-cpu0.018 msjax-cpu0.103 mstenferro-eager0.005 ms · 0.28xtenferro-trace0.008 ms · 0.44x

bin_omeinsum_batched_matmul_8x8_batch_4 Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.018 ms)omeinsum_path0.01 mspytorch-cpu0.018 msjax-cpu0.103 mstenferro-eager0.005 ms · 0.28xtenferro-trace0.008 ms · 0.44x

bin_omeinsum_high_d_12x12_contract_4_batch_4 Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.054 ms)omeinsum_path0.037 mspytorch-cpu0.054 msjax-cpu0.087 mstenferro-eager0.016 ms · 0.30xtenferro-trace0.018 ms · 0.33x

bin_omeinsum_high_d_12x12_contract_4_batch_4 Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.054 ms)omeinsum_path0.037 mspytorch-cpu0.054 msjax-cpu0.087 mstenferro-eager0.016 ms · 0.30xtenferro-trace0.018 ms · 0.33x

bin_omeinsum_high_d_12x12_contract_4_batch_4 Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.088 ms)omeinsum_path0.038 mspytorch-cpu0.088 msjax-cpu0.147 mstenferro-eager0.027 ms · 0.31xtenferro-trace0.029 ms · 0.33x

bin_omeinsum_high_d_12x12_contract_4_batch_4 Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.088 ms)omeinsum_path0.038 mspytorch-cpu0.088 msjax-cpu0.147 mstenferro-eager0.027 ms · 0.31xtenferro-trace0.029 ms · 0.33x

bin_omeinsum_high_d_12x12_contract_4_batch_4 Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.055 ms)omeinsum_path0.043 mspytorch-cpu0.055 msjax-cpu0.105 mstenferro-eager0.017 ms · 0.31xtenferro-trace0.018 ms · 0.33x

bin_omeinsum_high_d_12x12_contract_4_batch_4 Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.055 ms)omeinsum_path0.043 mspytorch-cpu0.055 msjax-cpu0.105 mstenferro-eager0.017 ms · 0.31xtenferro-trace0.018 ms · 0.33x

bin_matmul_256 Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.281 ms)jax-cpu0.218 mspytorch-cpu0.281 msomeinsum_path0.412 mstenferro-eager0.111 ms · 0.40xtenferro-trace1.749 ms · 6.22x

bin_matmul_256 Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.281 ms)jax-cpu0.218 mspytorch-cpu0.281 msomeinsum_path0.412 mstenferro-eager0.111 ms · 0.40xtenferro-trace1.749 ms · 6.22x

bin_matmul_256 Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.351 ms)jax-cpu0.229 msomeinsum_path0.338 mspytorch-cpu0.351 mstenferro-eager0.14 ms · 0.40xtenferro-trace0.899 ms · 2.56x

bin_matmul_256 Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.351 ms)jax-cpu0.229 msomeinsum_path0.338 mspytorch-cpu0.351 mstenferro-eager0.14 ms · 0.40xtenferro-trace0.899 ms · 2.56x

bin_omeinsum_matmul_10x10 Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.011 ms)pytorch-cpu0.011 msomeinsum_path0.012 msjax-cpu0.045 mstenferro-eager0.005 ms · 0.45xtenferro-trace0.01 ms · 0.91x

bin_omeinsum_matmul_10x10 Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.011 ms)pytorch-cpu0.011 msomeinsum_path0.012 msjax-cpu0.045 mstenferro-eager0.005 ms · 0.45xtenferro-trace0.01 ms · 0.91x

bin_elementwise_mul_2048x2048 Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (3.150 ms)pytorch-cpu3.150 msjax-cpu5.202 msomeinsum_path7.306 mstenferro-eager1.486 ms · 0.47xtenferro-trace1.929 ms · 0.61x

bin_elementwise_mul_2048x2048 Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (3.150 ms)pytorch-cpu3.150 msjax-cpu5.202 msomeinsum_path7.306 mstenferro-eager1.486 ms · 0.47xtenferro-trace1.929 ms · 0.61x

bin_omeinsum_matmul_10x10 Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.017 ms)pytorch-cpu0.017 msomeinsum_path0.018 msjax-cpu0.063 mstenferro-eager0.009 ms · 0.53xtenferro-trace0.046 ms · 2.71x

bin_omeinsum_matmul_10x10 Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.017 ms)pytorch-cpu0.017 msomeinsum_path0.018 msjax-cpu0.063 mstenferro-eager0.009 ms · 0.53xtenferro-trace0.046 ms · 2.71x

bin_omeinsum_matmul_10x10 Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.011 ms)pytorch-cpu0.011 msomeinsum_path0.012 msjax-cpu0.039 mstenferro-eager0.006 ms · 0.55xtenferro-trace0.01 ms · 0.91x

bin_omeinsum_matmul_10x10 Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.011 ms)pytorch-cpu0.011 msomeinsum_path0.012 msjax-cpu0.039 mstenferro-eager0.006 ms · 0.55xtenferro-trace0.01 ms · 0.91x

tensornetwork_permutation_focus_step409_316 Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (466.793 ms)omeinsum_path387.125 msjax-cpu417.033 mspytorch-cpu466.793 mstenferro-eager342.096 ms · 0.73xtenferro-trace259.165 ms · 0.56x

tensornetwork_permutation_focus_step409_316 Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (466.793 ms)omeinsum_path387.125 msjax-cpu417.033 mspytorch-cpu466.793 mstenferro-eager342.096 ms · 0.73xtenferro-trace259.165 ms · 0.56x

tensornetwork_permutation_light_415 Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (450.252 ms)omeinsum_path390.496 msjax-cpu423.649 mspytorch-cpu450.252 mstenferro-eager266.211 ms · 0.59xtenferro-trace256.016 ms · 0.57x

tensornetwork_permutation_light_415 Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (450.252 ms)omeinsum_path390.496 msjax-cpu423.649 mspytorch-cpu450.252 mstenferro-eager266.211 ms · 0.59xtenferro-trace256.016 ms · 0.57x

nary_matmul_chain_64 Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.025 ms)pytorch-cpu0.025 msomeinsum_path0.047 msjax-cpu0.051 mstenferro-eager0.016 ms · 0.64xtenferro-trace0.063 ms · 2.52x

nary_matmul_chain_64 Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.025 ms)pytorch-cpu0.025 msomeinsum_path0.047 msjax-cpu0.051 mstenferro-eager0.016 ms · 0.64xtenferro-trace0.063 ms · 2.52x

bin_matmul_256 Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.695 ms)jax-cpu0.53 mspytorch-cpu0.695 msomeinsum_path0.703 mstenferro-eager0.46 ms · 0.66xtenferro-trace0.573 ms · 0.82x

bin_matmul_256 Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.695 ms)jax-cpu0.53 mspytorch-cpu0.695 msomeinsum_path0.703 mstenferro-eager0.46 ms · 0.66xtenferro-trace0.573 ms · 0.82x

lm_batch_likelihood_sentence_3_12d Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (56.789 ms)jax-cpu51.156 mspytorch-cpu56.789 msomeinsum_path84.874 mstenferro-eager74.201 ms · 1.31xtenferro-trace39.877 ms · 0.70x

bin_elementwise_mul_2048x2048 Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (2.892 ms)pytorch-cpu2.892 msjax-cpu5.322 msomeinsum_path9.144 mstenferro-eager2.155 ms · 0.75xtenferro-trace2.097 ms · 0.73x

bin_elementwise_mul_2048x2048 Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (2.892 ms)pytorch-cpu2.892 msjax-cpu5.322 msomeinsum_path9.144 mstenferro-eager2.155 ms · 0.75xtenferro-trace2.097 ms · 0.73x

nary_matmul_chain_64 Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.031 ms)pytorch-cpu0.031 msomeinsum_path0.054 msjax-cpu0.073 mstenferro-eager0.023 ms · 0.74xtenferro-trace0.028 ms · 0.90x

nary_matmul_chain_64 Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.031 ms)pytorch-cpu0.031 msomeinsum_path0.054 msjax-cpu0.073 mstenferro-eager0.023 ms · 0.74xtenferro-trace0.028 ms · 0.90x

lm_batch_likelihood_sentence_4_4d Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (17.606 ms)jax-cpu14.778 mspytorch-cpu17.606 msomeinsum_path24.167 mstenferro-eager21.176 ms · 1.20xtenferro-trace13.188 ms · 0.75x

lm_batch_likelihood_brackets_4_4d Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (16.768 ms)jax-cpu13.805 mspytorch-cpu16.768 msomeinsum_path28.937 mstenferro-eager18.744 ms · 1.12xtenferro-trace12.578 ms · 0.75x

lm_batch_likelihood_sentence_4_4d Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (18.121 ms)jax-cpu13.925 mspytorch-cpu18.121 msomeinsum_path27.787 mstenferro-eager23.709 ms · 1.31xtenferro-trace13.679 ms · 0.75x

lm_batch_likelihood_brackets_4_4d Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (16.241 ms)jax-cpu14.358 mspytorch-cpu16.241 msomeinsum_path17.717 mstenferro-eager18.377 ms · 1.13xtenferro-trace12.527 ms · 0.77x

bin_matmul_1024 Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (34.676 ms)jax-cpu30.646 mspytorch-cpu34.676 msomeinsum_path49.426 mstenferro-eager29.722 ms · 0.86xtenferro-trace35.925 ms · 1.04x

bin_matmul_1024 Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (34.676 ms)jax-cpu30.646 mspytorch-cpu34.676 msomeinsum_path49.426 mstenferro-eager29.722 ms · 0.86xtenferro-trace35.925 ms · 1.04x

str_mps_varying_inner_product_200 Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (12.347 ms)pytorch-cpu12.347 msjax-cpu17.596 msomeinsum_path18.326 mstenferro-eager12.654 ms · 1.02xtenferro-trace10.612 ms · 0.86x

str_matrix_chain_multiplication_100 Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (3.608 ms)pytorch-cpu3.608 msjax-cpu5.224 msomeinsum_path7.341 mstenferro-eager3.363 ms · 0.93xtenferro-trace3.203 ms · 0.89x

bin_permuted_r4_abcd_dbef_acef_d32 Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (34.726 ms)pytorch-cpu34.726 msjax-cpu34.763 msomeinsum_path43.122 mstenferro-eager30.948 ms · 0.89xtenferro-trace38.449 ms · 1.11x

bin_permuted_r4_abcd_dbef_acef_d32 Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (34.726 ms)pytorch-cpu34.726 msjax-cpu34.763 msomeinsum_path43.122 mstenferro-eager30.948 ms · 0.89xtenferro-trace38.449 ms · 1.11x

str_nw_mera_closed_120 Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (1056.5 ms)pytorch-cpu1056.5 msjax-cpu1116.9 msomeinsum_path1189.8 mstenferro-eager987.014 ms · 0.93xtenferro-trace948.940 ms · 0.90x

bin_batched_matmul_b32_m128_n128_k128 Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (2.396 ms)jax-cpu2.154 mspytorch-cpu2.396 msomeinsum_path6.443 mstenferro-eager2.166 ms · 0.90xtenferro-trace4.047 ms · 1.69x

bin_batched_matmul_b32_m128_n128_k128 Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (2.396 ms)jax-cpu2.154 mspytorch-cpu2.396 msomeinsum_path6.443 mstenferro-eager2.166 ms · 0.90xtenferro-trace4.047 ms · 1.69x

str_matrix_chain_multiplication_100 Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (8.824 ms)pytorch-cpu8.824 msjax-cpu10.304 msomeinsum_path11.130 mstenferro-eager8.348 ms · 0.95xtenferro-trace8.009 ms · 0.91x

bin_batched_matmul_b32_m64_n64_k64 Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.331 ms)pytorch-cpu0.331 msjax-cpu0.338 msomeinsum_path0.457 mstenferro-eager0.305 ms · 0.92xtenferro-trace0.382 ms · 1.15x

bin_batched_matmul_b32_m64_n64_k64 Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.331 ms)pytorch-cpu0.331 msjax-cpu0.338 msomeinsum_path0.457 mstenferro-eager0.305 ms · 0.92xtenferro-trace0.382 ms · 1.15x

str_matrix_chain_multiplication_100 Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (8.828 ms)pytorch-cpu8.828 msjax-cpu10.342 msomeinsum_path11.006 mstenferro-eager8.348 ms · 0.95xtenferro-trace8.219 ms · 0.93x

str_matrix_chain_multiplication_100 Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (3.609 ms)pytorch-cpu3.609 msjax-cpu5.322 msomeinsum_path7.381 mstenferro-eager3.382 ms · 0.94xtenferro-trace6.540 ms · 1.81x

bin_elementwise_mul_512x512 Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.05 ms)pytorch-cpu0.05 msomeinsum_path0.087 msjax-cpu0.119 mstenferro-eager0.048 ms · 0.96xtenferro-trace0.049 ms · 0.98x

bin_elementwise_mul_512x512 Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.05 ms)pytorch-cpu0.05 msomeinsum_path0.087 msjax-cpu0.119 mstenferro-eager0.048 ms · 0.96xtenferro-trace0.049 ms · 0.98x

str_nw_mera_closed_120 Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (874.883 ms)pytorch-cpu874.883 msjax-cpu926.983 msomeinsum_path1073.0 mstenferro-eager877.532 ms · 1.00xtenferro-trace888.520 ms · 1.02x

gm_queen5_5_3.wcsp Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (1465.2 ms)jax-cpu1067.1 mspytorch-cpu1465.2 msomeinsum_path2009.9 mstenferro-eager1584.8 ms · 1.08xtenferro-trace1485.4 ms · 1.01x

lm_batch_likelihood_sentence_3_12d Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (27.052 ms)jax-cpu25.227 mspytorch-cpu27.052 msomeinsum_path66.763 mstenferro-eager109.431 ms · 4.05xtenferro-trace27.720 ms · 1.02x

bin_outer_product_4096 Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (20.847 ms)pytorch-cpu20.847 msjax-cpu21.456 msomeinsum_path154.639 mstenferro-eager21.532 ms · 1.03xtenferro-trace21.457 ms · 1.03x

bin_outer_product_4096 Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (20.847 ms)pytorch-cpu20.847 msjax-cpu21.456 msomeinsum_path154.639 mstenferro-eager21.532 ms · 1.03xtenferro-trace21.457 ms · 1.03x

str_mps_varying_inner_product_200 Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (11.428 ms)pytorch-cpu11.428 msomeinsum_path15.188 msjax-cpu16.857 mstenferro-eager12.047 ms · 1.05xtenferro-trace11.847 ms · 1.04x

gm_queen5_5_3.wcsp Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (3523.0 ms)jax-cpu2736.1 mspytorch-cpu3523.0 msomeinsum_path5008.0 mstenferro-eager3761.5 ms · 1.07xtenferro-trace4492.3 ms · 1.28x

lm_batch_likelihood_sentence_3_12d Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (41.633 ms)pytorch-cpu41.633 msjax-cpu64.109 msomeinsum_path91.663 mstenferro-eager81.928 ms · 1.97xtenferro-trace44.663 ms · 1.07x

str_nw_mera_closed_120 Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (386.730 ms)jax-cpu357.625 mspytorch-cpu386.730 msomeinsum_path547.987 mstenferro-eager440.974 ms · 1.14xtenferro-trace415.837 ms · 1.08x

bin_matmul_1024 Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (14.721 ms)jax-cpu8.802 mspytorch-cpu14.721 msomeinsum_path23.128 mstenferro-eager16.251 ms · 1.10xtenferro-trace17.878 ms · 1.21x

bin_matmul_1024 Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (14.721 ms)jax-cpu8.802 mspytorch-cpu14.721 msomeinsum_path23.128 mstenferro-eager16.251 ms · 1.10xtenferro-trace17.878 ms · 1.21x

str_mps_varying_inner_product_200 Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (8.135 ms)pytorch-cpu8.135 msjax-cpu10.361 msomeinsum_path18.246 mstenferro-eager9.058 ms · 1.11xtenferro-trace12.694 ms · 1.56x

bin_matmul_1024 Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (12.585 ms)jax-cpu8.858 mspytorch-cpu12.585 msomeinsum_path30.558 mstenferro-eager14.013 ms · 1.11xtenferro-trace18.409 ms · 1.46x

bin_matmul_1024 Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (12.585 ms)jax-cpu8.858 mspytorch-cpu12.585 msomeinsum_path30.558 mstenferro-eager14.013 ms · 1.11xtenferro-trace18.409 ms · 1.46x

str_nw_mera_open_26 Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (671.829 ms)pytorch-cpu671.829 msjax-cpu932.501 msomeinsum_path1070.4 mstenferro-eager798.729 ms · 1.19xtenferro-trace751.918 ms · 1.12x

str_nw_mera_open_26 Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (654.409 ms)pytorch-cpu654.409 msjax-cpu922.037 msomeinsum_path1093.9 mstenferro-eager808.681 ms · 1.24xtenferro-trace756.553 ms · 1.16x

str_nw_mera_closed_120 Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (357.773 ms)jax-cpu328.203 mspytorch-cpu357.773 msomeinsum_path565.853 mstenferro-eager454.306 ms · 1.27xtenferro-trace418.191 ms · 1.17x

lm_batch_likelihood_sentence_3_12d Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (27.876 ms)jax-cpu22.639 mspytorch-cpu27.876 msomeinsum_path74.549 mstenferro-eager125.616 ms · 4.51xtenferro-trace32.823 ms · 1.18x

str_mps_varying_inner_product_200 Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (8.178 ms)pytorch-cpu8.178 msjax-cpu9.884 msomeinsum_path15.921 mstenferro-eager9.826 ms · 1.20xtenferro-trace10.260 ms · 1.25x

bin_outer_product_4096 Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (7.454 ms)pytorch-cpu7.454 msjax-cpu12.498 msomeinsum_path197.616 mstenferro-eager9.046 ms · 1.21xtenferro-trace9.392 ms · 1.26x

bin_outer_product_4096 Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (7.454 ms)pytorch-cpu7.454 msjax-cpu12.498 msomeinsum_path197.616 mstenferro-eager9.046 ms · 1.21xtenferro-trace9.392 ms · 1.26x

str_nw_mera_closed_120 Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (293.342 ms)jax-cpu277.434 mspytorch-cpu293.342 msomeinsum_path461.747 mstenferro-eager370.964 ms · 1.26xtenferro-trace383.280 ms · 1.31x

str_nw_mera_closed_120 Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (262.954 ms)jax-cpu233.187 mspytorch-cpu262.954 msomeinsum_path468.168 mstenferro-eager359.499 ms · 1.37xtenferro-trace384.184 ms · 1.46x

str_mps_varying_inner_product_200 Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (9.944 ms)pytorch-cpu9.944 msjax-cpu14.521 msomeinsum_path22.127 mstenferro-eager13.648 ms · 1.37xtenferro-trace15.722 ms · 1.58x

bin_permuted_r4_abcd_dbef_acef_d32 Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (11.051 ms)jax-cpu9.936 mspytorch-cpu11.051 msomeinsum_path23.495 mstenferro-eager15.224 ms · 1.38xtenferro-trace23.002 ms · 2.08x

bin_permuted_r4_abcd_dbef_acef_d32 Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (11.051 ms)jax-cpu9.936 mspytorch-cpu11.051 msomeinsum_path23.495 mstenferro-eager15.224 ms · 1.38xtenferro-trace23.002 ms · 2.08x

bin_permuted_r4_abcd_dbef_acef_d32 Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (12.732 ms)jax-cpu12.206 mspytorch-cpu12.732 msomeinsum_path19.839 mstenferro-eager18.258 ms · 1.43xtenferro-trace17.571 ms · 1.38x

bin_permuted_r4_abcd_dbef_acef_d32 Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (12.732 ms)jax-cpu12.206 mspytorch-cpu12.732 msomeinsum_path19.839 mstenferro-eager18.258 ms · 1.43xtenferro-trace17.571 ms · 1.38x

tensornetwork_permutation_light_415 Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (144.041 ms)pytorch-cpu144.041 msjax-cpu227.021 msomeinsum_path265.425 mstenferro-eager207.306 ms · 1.44xtenferro-trace206.231 ms · 1.43x

tensornetwork_permutation_light_415 Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (144.041 ms)pytorch-cpu144.041 msjax-cpu227.021 msomeinsum_path265.425 mstenferro-eager207.306 ms · 1.44xtenferro-trace206.231 ms · 1.43x

tensornetwork_permutation_focus_step409_316 Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (147.208 ms)pytorch-cpu147.208 msjax-cpu220.242 msomeinsum_path273.073 mstenferro-eager238.555 ms · 1.62xtenferro-trace212.276 ms · 1.44x

tensornetwork_permutation_focus_step409_316 Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (147.208 ms)pytorch-cpu147.208 msjax-cpu220.242 msomeinsum_path273.073 mstenferro-eager238.555 ms · 1.62xtenferro-trace212.276 ms · 1.44x

bin_outer_product_4096 Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (7.507 ms)pytorch-cpu7.507 msjax-cpu12.945 msomeinsum_path140.577 mstenferro-eager11.441 ms · 1.52xtenferro-trace11.105 ms · 1.48x

bin_outer_product_4096 Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (7.507 ms)pytorch-cpu7.507 msjax-cpu12.945 msomeinsum_path140.577 mstenferro-eager11.441 ms · 1.52xtenferro-trace11.105 ms · 1.48x

str_mps_varying_inner_product_200 Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (9.840 ms)pytorch-cpu9.840 msjax-cpu14.167 msomeinsum_path19.372 mstenferro-eager17.712 ms · 1.80xtenferro-trace15.417 ms · 1.57x

lm_batch_likelihood_sentence_3_12d Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (15.592 ms)pytorch-cpu15.592 msjax-cpu26.671 msomeinsum_path72.385 mstenferro-eager149.707 ms · 9.60xtenferro-trace24.891 ms · 1.60x

lm_batch_likelihood_sentence_3_12d Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (16.965 ms)pytorch-cpu16.965 msjax-cpu19.949 msomeinsum_path58.235 mstenferro-eager61.126 ms · 3.60xtenferro-trace27.434 ms · 1.62x

bin_batched_outer_product_compact_j16_k16_o64_t64 Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.074 ms)pytorch-cpu0.074 msjax-cpu0.203 msomeinsum_path0.691 mstenferro-eager0.12 ms · 1.62xtenferro-trace0.134 ms · 1.81x

bin_batched_outer_product_compact_j16_k16_o64_t64 Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.074 ms)pytorch-cpu0.074 msjax-cpu0.203 msomeinsum_path0.691 mstenferro-eager0.12 ms · 1.62xtenferro-trace0.134 ms · 1.81x

bin_batched_outer_product_noncompact_j16_k16_o64_t64 Threads: 1 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.074 ms)pytorch-cpu0.074 msjax-cpu0.196 msomeinsum_path3.342 mstenferro-eager0.12 ms · 1.62xtenferro-trace0.485 ms · 6.55x

bin_batched_outer_product_noncompact_j16_k16_o64_t64 Threads: 1 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.074 ms)pytorch-cpu0.074 msjax-cpu0.196 msomeinsum_path3.342 mstenferro-eager0.12 ms · 1.62xtenferro-trace0.485 ms · 6.55x

gm_queen5_5_3.wcsp Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (578.710 ms)pytorch-cpu578.710 msjax-cpu640.613 msomeinsum_path1658.2 mstenferro-eager948.081 ms · 1.64xtenferro-trace1121.6 ms · 1.94x

str_nw_mera_open_26 Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (245.024 ms)pytorch-cpu245.024 msjax-cpu337.496 msomeinsum_path631.238 mstenferro-eager458.514 ms · 1.87xtenferro-trace411.038 ms · 1.68x

tensornetwork_permutation_light_415 Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (126.197 ms)pytorch-cpu126.197 msjax-cpu286.035 msomeinsum_path326.374 mstenferro-eager229.150 ms · 1.82xtenferro-trace218.170 ms · 1.73x

tensornetwork_permutation_light_415 Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (126.197 ms)pytorch-cpu126.197 msjax-cpu286.035 msomeinsum_path326.374 mstenferro-eager229.150 ms · 1.82xtenferro-trace218.170 ms · 1.73x

gm_queen5_5_3.wcsp Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (1375.7 ms)pytorch-cpu1375.7 msjax-cpu1736.2 msomeinsum_path4581.8 mstenferro-eager2379.5 ms · 1.73xtenferro-trace3647.3 ms · 2.65x

str_nw_mera_open_26 Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (236.201 ms)pytorch-cpu236.201 msjax-cpu332.863 msomeinsum_path642.859 mstenferro-eager466.773 ms · 1.98xtenferro-trace414.597 ms · 1.76x

gm_queen5_5_3.wcsp Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (1243.9 ms)pytorch-cpu1243.9 msjax-cpu1563.4 msomeinsum_path5626.6 mstenferro-eager2214.6 ms · 1.78xtenferro-trace4009.8 ms · 3.22x

tensornetwork_permutation_focus_step409_316 Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (128.458 ms)pytorch-cpu128.458 msjax-cpu277.987 msomeinsum_path334.694 mstenferro-eager274.506 ms · 2.14xtenferro-trace233.983 ms · 1.82x

tensornetwork_permutation_focus_step409_316 Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (128.458 ms)pytorch-cpu128.458 msjax-cpu277.987 msomeinsum_path334.694 mstenferro-eager274.506 ms · 2.14xtenferro-trace233.983 ms · 1.82x

gm_queen5_5_3.wcsp Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (510.419 ms)pytorch-cpu510.419 msjax-cpu575.919 msomeinsum_path1988.5 mstenferro-eager939.443 ms · 1.84xtenferro-trace1227.4 ms · 2.40x

str_nw_mera_open_26 Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (224.839 ms)pytorch-cpu224.839 msjax-cpu290.107 msomeinsum_path706.364 mstenferro-eager432.567 ms · 1.92xtenferro-trace455.975 ms · 2.03x

str_nw_mera_open_26 Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (222.787 ms)pytorch-cpu222.787 msjax-cpu278.737 msomeinsum_path705.791 mstenferro-eager434.313 ms · 1.95xtenferro-trace449.497 ms · 2.02x

str_matrix_chain_multiplication_100 Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (3.861 ms)pytorch-cpu3.861 msjax-cpu5.332 msomeinsum_path8.560 mstenferro-eager8.091 ms · 2.10xtenferro-trace7.690 ms · 1.99x

lm_batch_likelihood_sentence_4_4d Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (6.202 ms)pytorch-cpu6.202 msjax-cpu10.139 msomeinsum_path28.104 mstenferro-eager28.737 ms · 4.63xtenferro-trace12.901 ms · 2.08x

str_matrix_chain_multiplication_100 Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (3.851 ms)pytorch-cpu3.851 msjax-cpu5.284 msomeinsum_path8.224 mstenferro-eager8.069 ms · 2.10xtenferro-trace8.085 ms · 2.10x

nary_matmul_chain_64 Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.031 ms)pytorch-cpu0.031 msomeinsum_path0.08 msjax-cpu0.085 mstenferro-eager0.077 ms · 2.48xtenferro-trace0.093 ms · 3.00x

nary_matmul_chain_64 Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.031 ms)pytorch-cpu0.031 msomeinsum_path0.08 msjax-cpu0.085 mstenferro-eager0.077 ms · 2.48xtenferro-trace0.093 ms · 3.00x

lm_batch_likelihood_brackets_4_4d Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (5.370 ms)pytorch-cpu5.370 msjax-cpu10.050 msomeinsum_path15.527 mstenferro-eager19.000 ms · 3.54xtenferro-trace14.756 ms · 2.75x

lm_batch_likelihood_sentence_4_4d Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (6.656 ms)pytorch-cpu6.656 msjax-cpu13.259 msomeinsum_path44.301 mstenferro-eager39.085 ms · 5.87xtenferro-trace21.564 ms · 3.24x

bin_batched_matmul_b32_m128_n128_k128 Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.6 ms)pytorch-cpu0.6 msjax-cpu1.115 msomeinsum_path4.824 mstenferro-eager1.974 ms · 3.29xtenferro-trace4.246 ms · 7.08x

bin_batched_matmul_b32_m128_n128_k128 Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.6 ms)pytorch-cpu0.6 msjax-cpu1.115 msomeinsum_path4.824 mstenferro-eager1.974 ms · 3.29xtenferro-trace4.246 ms · 7.08x

lm_batch_likelihood_brackets_4_4d Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (6.353 ms)pytorch-cpu6.353 msjax-cpu13.519 msomeinsum_path50.697 mstenferro-eager39.825 ms · 6.27xtenferro-trace21.416 ms · 3.37x

lm_batch_likelihood_brackets_4_4d Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (5.397 ms)pytorch-cpu5.397 msjax-cpu10.382 msomeinsum_path30.606 mstenferro-eager25.079 ms · 4.65xtenferro-trace18.926 ms · 3.51x

lm_batch_likelihood_sentence_4_4d Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (5.624 ms)pytorch-cpu5.624 msjax-cpu10.030 msomeinsum_path35.908 mstenferro-eager34.835 ms · 6.19xtenferro-trace20.623 ms · 3.67x

bin_batched_matmul_b32_m128_n128_k128 Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.551 ms)pytorch-cpu0.551 msjax-cpu1.064 msomeinsum_path9.428 mstenferro-eager2.072 ms · 3.76xtenferro-trace3.345 ms · 6.07x

bin_batched_matmul_b32_m128_n128_k128 Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.551 ms)pytorch-cpu0.551 msjax-cpu1.064 msomeinsum_path9.428 mstenferro-eager2.072 ms · 3.76xtenferro-trace3.345 ms · 6.07x

lm_batch_likelihood_sentence_4_4d Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (6.729 ms)pytorch-cpu6.729 msjax-cpu13.175 msomeinsum_path35.087 mstenferro-eager42.034 ms · 6.25xtenferro-trace30.016 ms · 4.46x

lm_batch_likelihood_brackets_4_4d Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (5.971 ms)pytorch-cpu5.971 msjax-cpu13.143 msomeinsum_path35.960 mstenferro-eager43.212 ms · 7.24xtenferro-trace27.041 ms · 4.53x

bin_batched_matmul_b32_m64_n64_k64 Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.095 ms)pytorch-cpu0.095 msomeinsum_path0.422 msjax-cpu0.434 mstenferro-eager0.763 ms · 8.03xtenferro-trace1.645 ms · 17.32x

bin_batched_matmul_b32_m64_n64_k64 Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.095 ms)pytorch-cpu0.095 msomeinsum_path0.422 msjax-cpu0.434 mstenferro-eager0.763 ms · 8.03xtenferro-trace1.645 ms · 17.32x

bin_batched_outer_product_noncompact_j16_k16_o64_t64 Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.027 ms)pytorch-cpu0.027 msjax-cpu0.211 msomeinsum_path4.034 mstenferro-eager0.542 ms · 20.07xtenferro-trace0.319 ms · 11.81x

bin_batched_outer_product_noncompact_j16_k16_o64_t64 Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.027 ms)pytorch-cpu0.027 msjax-cpu0.211 msomeinsum_path4.034 mstenferro-eager0.542 ms · 20.07xtenferro-trace0.319 ms · 11.81x

bin_batched_outer_product_noncompact_j16_k16_o64_t64 Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.03 ms)pytorch-cpu0.03 msjax-cpu0.474 msomeinsum_path4.301 mstenferro-eager0.405 ms · 13.50xtenferro-trace0.839 ms · 27.97x

bin_batched_outer_product_noncompact_j16_k16_o64_t64 Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.03 ms)pytorch-cpu0.03 msjax-cpu0.474 msomeinsum_path4.301 mstenferro-eager0.405 ms · 13.50xtenferro-trace0.839 ms · 27.97x

bin_batched_matmul_b32_m64_n64_k64 Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.08 ms)pytorch-cpu0.08 msjax-cpu0.763 msomeinsum_path0.768 mstenferro-eager1.092 ms · 13.65xtenferro-trace1.897 ms · 23.71x

bin_batched_matmul_b32_m64_n64_k64 Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.08 ms)pytorch-cpu0.08 msjax-cpu0.763 msomeinsum_path0.768 mstenferro-eager1.092 ms · 13.65xtenferro-trace1.897 ms · 23.71x

bin_batched_outer_product_compact_j16_k16_o64_t64 Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.029 ms)pytorch-cpu0.029 msjax-cpu0.159 msomeinsum_path1.423 mstenferro-eager0.552 ms · 19.03xtenferro-trace0.741 ms · 25.55x

bin_batched_outer_product_compact_j16_k16_o64_t64 Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.029 ms)pytorch-cpu0.029 msjax-cpu0.159 msomeinsum_path1.423 mstenferro-eager0.552 ms · 19.03xtenferro-trace0.741 ms · 25.55x

bin_elementwise_mul_512x512 Threads: 4 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.019 ms)pytorch-cpu0.019 msomeinsum_path0.077 msjax-cpu0.222 mstenferro-eager0.465 ms · 24.47xtenferro-trace0.702 ms · 36.95x

bin_elementwise_mul_512x512 Threads: 4 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.019 ms)pytorch-cpu0.019 msomeinsum_path0.077 msjax-cpu0.222 mstenferro-eager0.465 ms · 24.47xtenferro-trace0.702 ms · 36.95x

bin_batched_outer_product_compact_j16_k16_o64_t64 Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.03 ms)pytorch-cpu0.03 msjax-cpu0.438 msomeinsum_path1.553 mstenferro-eager0.791 ms · 26.37xtenferro-trace1.190 ms · 39.67x

bin_batched_outer_product_compact_j16_k16_o64_t64 Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.03 ms)pytorch-cpu0.03 msjax-cpu0.438 msomeinsum_path1.553 mstenferro-eager0.791 ms · 26.37xtenferro-trace1.190 ms · 39.67x

bin_elementwise_mul_512x512 Threads: 8 / opt_flops · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.021 ms)pytorch-cpu0.021 msomeinsum_path0.116 msjax-cpu0.252 mstenferro-eager1.085 ms · 51.67xtenferro-trace2.448 ms · 116.57x

bin_elementwise_mul_512x512 Threads: 8 / opt_size · n=7

bar = median time (log scale) · shorter (left) is faster · dashed line = PyTorch (0.021 ms)pytorch-cpu0.021 msomeinsum_path0.116 msjax-cpu0.252 mstenferro-eager1.085 ms · 51.67xtenferro-trace2.448 ms · 116.57x

Original report

CPU Einsum Benchmark Results

Generated from the explicit runs below; each section retains its own provenance and thread settings.

Threads: 1

Source report: data/results/zen5-cpu/cpu/einsum/20261027_231715/report.md.

Latest run: ./scripts/run_all.sh 1.

This file is generated from one suite run under data/results/zen5-cpu/cpu/einsum/20261027_231715.

CPU Information

Thread Environment

Tenferro CPU BLAS Backend

Python Backend Providers

Julia / OMEinsum.jl Backend

Logs:

Strategy: opt_flops

Median ± IQR (ms). JULIA_NUM_THREADS=1, OMP_NUM_THREADS=1, RAYON_NUM_THREADS=1.

Instance tenferro-rs trace mode (ms) tenferro-rs eager mode (ms) PyTorch Python (ms) JAX Python (XLA CPU dot) (ms) OMEinsum.jl OpenBLAS (ms)
bin_batched_matmul_b32_m128_n128_k128 4.047 ± 0.081 2.166 ± 0.085 2.396 ± 0.034 2.154 ± 0.028 6.443 ± 5.395
bin_batched_matmul_b32_m64_n64_k64 0.382 ± 0.006 0.305 ± 0.005 0.331 ± 0.003 0.338 ± 0.010 0.457 ± 0.736
bin_batched_outer_product_compact_j16_k16_o64_t64 0.134 ± 0.014 0.120 ± 0.007 0.074 ± 0.001 0.203 ± 0.016 0.691 ± 1.113
bin_batched_outer_product_noncompact_j16_k16_o64_t64 0.485 ± 0.005 0.120 ± 0.003 0.074 ± 0.001 0.196 ± 0.015 3.342 ± 1.811
bin_elementwise_mul_2048x2048 1.607 ± 0.040 1.527 ± 0.026 6.445 ± 0.145 9.317 ± 1.050 7.190 ± 2.489
bin_elementwise_mul_512x512 0.049 ± 0.006 0.048 ± 0.008 0.050 ± 0.001 0.119 ± 0.009 0.087 ± 0.244
bin_matmul_1024 35.925 ± 0.179 29.722 ± 0.126 34.676 ± 0.106 30.646 ± 0.095 49.426 ± 16.034
bin_matmul_256 0.573 ± 0.021 0.460 ± 0.008 0.695 ± 0.014 0.530 ± 0.004 0.703 ± 0.006
bin_omeinsum_batched_matmul_8x8_batch_4 0.005 ± 0.000 0.003 ± 0.000 0.011 ± 0.000 0.064 ± 0.005 0.007 ± 0.001
bin_omeinsum_high_d_12x12_contract_4_batch_4 0.018 ± 0.000 0.016 ± 0.000 0.054 ± 0.000 0.087 ± 0.009 0.037 ± 0.009
bin_omeinsum_matmul_10x10 0.010 ± 0.000 0.005 ± 0.000 0.011 ± 0.000 0.045 ± 0.006 0.012 ± 0.001
bin_outer_product_4096 21.457 ± 0.654 21.532 ± 0.576 20.847 ± 0.912 21.456 ± 3.914 154.639 ± 30.207
bin_permuted_r4_abcd_dbef_acef_d32 38.449 ± 0.417 30.948 ± 0.136 34.726 ± 0.068 34.763 ± 0.194 43.122 ± 11.754
gm_queen5_5_3.wcsp 4492.272 ± 25.246 3761.453 ± 6.039 3522.951 ± 63.128 2736.080 ± 15.119 5007.983 ± 65.698
lm_batch_likelihood_brackets_4_4d 12.527 ± 0.057 18.377 ± 0.966 16.241 ± 0.156 14.358 ± 0.313 17.717 ± 7.574
lm_batch_likelihood_sentence_3_12d 39.877 ± 0.232 74.201 ± 0.350 56.789 ± 0.457 51.156 ± 0.137 84.874 ± 38.273
lm_batch_likelihood_sentence_4_4d 13.188 ± 0.176 21.176 ± 0.333 17.606 ± 0.291 14.778 ± 0.136 24.167 ± 23.864
nary_matmul_chain_64 0.028 ± 0.000 0.023 ± 0.000 0.031 ± 0.000 0.073 ± 0.006 0.054 ± 0.007
str_matrix_chain_multiplication_100 8.009 ± 0.058 8.348 ± 0.042 8.824 ± 0.045 10.304 ± 0.253 11.130 ± 0.941
str_mps_varying_inner_product_200 10.612 ± 0.107 12.654 ± 0.129 12.347 ± 0.031 17.596 ± 0.178 18.326 ± 29.974
str_nw_mera_closed_120 948.940 ± 2.771 987.014 ± 5.354 1056.507 ± 2.733 1116.942 ± 3.994 1189.776 ± 35.081
str_nw_mera_open_26 756.553 ± 2.166 808.681 ± 2.418 654.409 ± 1.261 922.037 ± 2.208 1093.886 ± 33.365
tensornetwork_permutation_focus_step409_316 259.165 ± 0.987 342.096 ± 1.296 466.793 ± 0.857 417.033 ± 1.571 387.125 ± 48.689
tensornetwork_permutation_light_415 256.016 ± 1.483 266.211 ± 1.241 450.252 ± 1.470 423.649 ± 1.472 390.496 ± 107.997
Strategy: opt_size

Median ± IQR (ms). JULIA_NUM_THREADS=1, OMP_NUM_THREADS=1, RAYON_NUM_THREADS=1.

Instance tenferro-rs trace mode (ms) tenferro-rs eager mode (ms) PyTorch Python (ms) JAX Python (XLA CPU dot) (ms) OMEinsum.jl OpenBLAS (ms)
bin_batched_matmul_b32_m128_n128_k128 4.047 ± 0.081 2.166 ± 0.085 2.396 ± 0.034 2.154 ± 0.028 6.443 ± 5.395
bin_batched_matmul_b32_m64_n64_k64 0.382 ± 0.006 0.305 ± 0.005 0.331 ± 0.003 0.338 ± 0.010 0.457 ± 0.736
bin_batched_outer_product_compact_j16_k16_o64_t64 0.134 ± 0.014 0.120 ± 0.007 0.074 ± 0.001 0.203 ± 0.016 0.691 ± 1.113
bin_batched_outer_product_noncompact_j16_k16_o64_t64 0.485 ± 0.005 0.120 ± 0.003 0.074 ± 0.001 0.196 ± 0.015 3.342 ± 1.811
bin_elementwise_mul_2048x2048 1.607 ± 0.040 1.527 ± 0.026 6.445 ± 0.145 9.317 ± 1.050 7.190 ± 2.489
bin_elementwise_mul_512x512 0.049 ± 0.006 0.048 ± 0.008 0.050 ± 0.001 0.119 ± 0.009 0.087 ± 0.244
bin_matmul_1024 35.925 ± 0.179 29.722 ± 0.126 34.676 ± 0.106 30.646 ± 0.095 49.426 ± 16.034
bin_matmul_256 0.573 ± 0.021 0.460 ± 0.008 0.695 ± 0.014 0.530 ± 0.004 0.703 ± 0.006
bin_omeinsum_batched_matmul_8x8_batch_4 0.005 ± 0.000 0.003 ± 0.000 0.011 ± 0.000 0.064 ± 0.005 0.007 ± 0.001
bin_omeinsum_high_d_12x12_contract_4_batch_4 0.018 ± 0.000 0.016 ± 0.000 0.054 ± 0.000 0.087 ± 0.009 0.037 ± 0.009
bin_omeinsum_matmul_10x10 0.010 ± 0.000 0.005 ± 0.000 0.011 ± 0.000 0.045 ± 0.006 0.012 ± 0.001
bin_outer_product_4096 21.457 ± 0.654 21.532 ± 0.576 20.847 ± 0.912 21.456 ± 3.914 154.639 ± 30.207
bin_permuted_r4_abcd_dbef_acef_d32 38.449 ± 0.417 30.948 ± 0.136 34.726 ± 0.068 34.763 ± 0.194 43.122 ± 11.754
gm_queen5_5_3.wcsp 1485.442 ± 5.491 1584.827 ± 2.815 1465.193 ± 6.489 1067.073 ± 1.826 2009.878 ± 15.527
lm_batch_likelihood_brackets_4_4d 12.578 ± 0.092 18.744 ± 0.322 16.768 ± 0.129 13.805 ± 0.157 28.937 ± 38.364
lm_batch_likelihood_sentence_3_12d 44.663 ± 0.446 81.928 ± 0.763 41.633 ± 0.205 64.109 ± 0.289 91.663 ± 24.404
lm_batch_likelihood_sentence_4_4d 13.679 ± 0.157 23.709 ± 0.758 18.121 ± 0.049 13.925 ± 0.160 27.787 ± 32.991
nary_matmul_chain_64 0.028 ± 0.000 0.023 ± 0.000 0.031 ± 0.000 0.073 ± 0.006 0.054 ± 0.007
str_matrix_chain_multiplication_100 8.219 ± 0.075 8.348 ± 0.031 8.828 ± 0.036 10.342 ± 0.173 11.006 ± 0.930
str_mps_varying_inner_product_200 11.847 ± 0.052 12.047 ± 0.076 11.428 ± 0.041 16.857 ± 0.307 15.188 ± 9.499
str_nw_mera_closed_120 888.520 ± 1.115 877.532 ± 4.800 874.883 ± 2.927 926.983 ± 1.513 1073.027 ± 21.064
str_nw_mera_open_26 751.918 ± 2.763 798.729 ± 1.986 671.829 ± 2.533 932.501 ± 1.794 1070.435 ± 32.258
tensornetwork_permutation_focus_step409_316 259.165 ± 0.987 342.096 ± 1.296 466.793 ± 0.857 417.033 ± 1.571 387.125 ± 48.689
tensornetwork_permutation_light_415 256.016 ± 1.483 266.211 ± 1.241 450.252 ± 1.470 423.649 ± 1.472 390.496 ± 107.997

Threads: 4

Source report: data/results/zen5-cpu/cpu/einsum/20261027_235042/report.md.

Latest run: ./scripts/run_all.sh 4.

This file is generated from one suite run under data/results/zen5-cpu/cpu/einsum/20261027_235042.

CPU Information

Thread Environment

Tenferro CPU BLAS Backend

Python Backend Providers

Julia / OMEinsum.jl Backend

Logs:

Strategy: opt_flops

Median ± IQR (ms). JULIA_NUM_THREADS=4, OMP_NUM_THREADS=4, RAYON_NUM_THREADS=4.

Instance tenferro-rs trace mode (ms) tenferro-rs eager mode (ms) PyTorch Python (ms) JAX Python (XLA CPU dot) (ms) OMEinsum.jl OpenBLAS (ms)
bin_batched_matmul_b32_m128_n128_k128 4.246 ± 1.510 1.974 ± 0.025 0.600 ± 0.007 1.115 ± 0.274 4.824 ± 3.754
bin_batched_matmul_b32_m64_n64_k64 1.645 ± 0.309 0.763 ± 0.005 0.095 ± 0.001 0.434 ± 0.360 0.422 ± 0.700
bin_batched_outer_product_compact_j16_k16_o64_t64 0.741 ± 0.009 0.552 ± 0.005 0.029 ± 0.000 0.159 ± 0.096 1.423 ± 0.276
bin_batched_outer_product_noncompact_j16_k16_o64_t64 0.319 ± 0.005 0.542 ± 0.003 0.027 ± 0.001 0.211 ± 0.153 4.034 ± 3.592
bin_elementwise_mul_2048x2048 1.929 ± 0.022 1.486 ± 0.484 3.150 ± 0.219 5.202 ± 0.613 7.306 ± 0.793
bin_elementwise_mul_512x512 0.702 ± 0.004 0.465 ± 0.003 0.019 ± 0.001 0.222 ± 0.099 0.077 ± 0.244
bin_matmul_1024 17.878 ± 0.589 16.251 ± 6.691 14.721 ± 0.170 8.802 ± 0.121 23.128 ± 12.476
bin_matmul_256 0.899 ± 0.004 0.140 ± 0.001 0.351 ± 0.007 0.229 ± 0.176 0.338 ± 0.021
bin_omeinsum_batched_matmul_8x8_batch_4 0.005 ± 0.000 0.003 ± 0.000 0.012 ± 0.001 0.064 ± 0.006 0.006 ± 0.001
bin_omeinsum_high_d_12x12_contract_4_batch_4 0.018 ± 0.000 0.017 ± 0.000 0.055 ± 0.001 0.105 ± 0.041 0.043 ± 0.012
bin_omeinsum_matmul_10x10 0.010 ± 0.000 0.006 ± 0.000 0.011 ± 0.001 0.039 ± 0.092 0.012 ± 0.001
bin_outer_product_4096 11.105 ± 3.773 11.441 ± 3.865 7.507 ± 0.442 12.945 ± 0.167 140.577 ± 4.374
bin_permuted_r4_abcd_dbef_acef_d32 17.571 ± 0.485 18.258 ± 7.728 12.732 ± 0.182 12.206 ± 0.389 19.839 ± 9.967
gm_queen5_5_3.wcsp 3647.344 ± 65.549 2379.539 ± 69.696 1375.702 ± 6.216 1736.151 ± 32.565 4581.752 ± 21.314
lm_batch_likelihood_brackets_4_4d 14.756 ± 0.047 19.000 ± 0.660 5.370 ± 0.050 10.050 ± 0.335 15.527 ± 5.423
lm_batch_likelihood_sentence_3_12d 27.434 ± 3.537 61.126 ± 48.272 16.965 ± 0.068 19.949 ± 1.095 58.235 ± 22.569
lm_batch_likelihood_sentence_4_4d 20.623 ± 5.836 34.835 ± 0.854 5.624 ± 0.065 10.030 ± 0.402 35.908 ± 19.399
nary_matmul_chain_64 0.063 ± 0.001 0.016 ± 0.000 0.025 ± 0.001 0.051 ± 0.008 0.047 ± 0.032
str_matrix_chain_multiplication_100 6.540 ± 0.020 3.382 ± 0.011 3.609 ± 0.021 5.322 ± 0.408 7.381 ± 0.617
str_mps_varying_inner_product_200 12.694 ± 0.101 9.058 ± 0.106 8.135 ± 0.055 10.361 ± 0.245 18.246 ± 3.301
str_nw_mera_closed_120 415.837 ± 3.099 440.974 ± 11.282 386.730 ± 1.625 357.625 ± 6.872 547.987 ± 15.014
str_nw_mera_open_26 414.597 ± 4.057 466.773 ± 102.083 236.201 ± 1.339 332.863 ± 5.670 642.859 ± 13.094
tensornetwork_permutation_focus_step409_316 212.276 ± 2.742 238.555 ± 20.734 147.208 ± 0.417 220.242 ± 3.315 273.073 ± 26.585
tensornetwork_permutation_light_415 206.231 ± 27.970 207.306 ± 22.942 144.041 ± 0.368 227.021 ± 7.230 265.425 ± 15.746
Strategy: opt_size

Median ± IQR (ms). JULIA_NUM_THREADS=4, OMP_NUM_THREADS=4, RAYON_NUM_THREADS=4.

Instance tenferro-rs trace mode (ms) tenferro-rs eager mode (ms) PyTorch Python (ms) JAX Python (XLA CPU dot) (ms) OMEinsum.jl OpenBLAS (ms)
bin_batched_matmul_b32_m128_n128_k128 4.246 ± 1.510 1.974 ± 0.025 0.600 ± 0.007 1.115 ± 0.274 4.824 ± 3.754
bin_batched_matmul_b32_m64_n64_k64 1.645 ± 0.309 0.763 ± 0.005 0.095 ± 0.001 0.434 ± 0.360 0.422 ± 0.700
bin_batched_outer_product_compact_j16_k16_o64_t64 0.741 ± 0.009 0.552 ± 0.005 0.029 ± 0.000 0.159 ± 0.096 1.423 ± 0.276
bin_batched_outer_product_noncompact_j16_k16_o64_t64 0.319 ± 0.005 0.542 ± 0.003 0.027 ± 0.001 0.211 ± 0.153 4.034 ± 3.592
bin_elementwise_mul_2048x2048 1.929 ± 0.022 1.486 ± 0.484 3.150 ± 0.219 5.202 ± 0.613 7.306 ± 0.793
bin_elementwise_mul_512x512 0.702 ± 0.004 0.465 ± 0.003 0.019 ± 0.001 0.222 ± 0.099 0.077 ± 0.244
bin_matmul_1024 17.878 ± 0.589 16.251 ± 6.691 14.721 ± 0.170 8.802 ± 0.121 23.128 ± 12.476
bin_matmul_256 0.899 ± 0.004 0.140 ± 0.001 0.351 ± 0.007 0.229 ± 0.176 0.338 ± 0.021
bin_omeinsum_batched_matmul_8x8_batch_4 0.005 ± 0.000 0.003 ± 0.000 0.012 ± 0.001 0.064 ± 0.006 0.006 ± 0.001
bin_omeinsum_high_d_12x12_contract_4_batch_4 0.018 ± 0.000 0.017 ± 0.000 0.055 ± 0.001 0.105 ± 0.041 0.043 ± 0.012
bin_omeinsum_matmul_10x10 0.010 ± 0.000 0.006 ± 0.000 0.011 ± 0.001 0.039 ± 0.092 0.012 ± 0.001
bin_outer_product_4096 11.105 ± 3.773 11.441 ± 3.865 7.507 ± 0.442 12.945 ± 0.167 140.577 ± 4.374
bin_permuted_r4_abcd_dbef_acef_d32 17.571 ± 0.485 18.258 ± 7.728 12.732 ± 0.182 12.206 ± 0.389 19.839 ± 9.967
gm_queen5_5_3.wcsp 1121.566 ± 7.438 948.081 ± 153.408 578.710 ± 1.033 640.613 ± 8.390 1658.170 ± 15.339
lm_batch_likelihood_brackets_4_4d 18.926 ± 0.093 25.079 ± 1.216 5.397 ± 0.069 10.382 ± 0.433 30.606 ± 18.535
lm_batch_likelihood_sentence_3_12d 27.720 ± 0.321 109.431 ± 44.509 27.052 ± 0.432 25.227 ± 1.251 66.763 ± 22.698
lm_batch_likelihood_sentence_4_4d 12.901 ± 3.499 28.737 ± 3.395 6.202 ± 0.116 10.139 ± 0.423 28.104 ± 19.357
nary_matmul_chain_64 0.063 ± 0.001 0.016 ± 0.000 0.025 ± 0.001 0.051 ± 0.008 0.047 ± 0.032
str_matrix_chain_multiplication_100 3.203 ± 0.013 3.363 ± 0.025 3.608 ± 0.043 5.224 ± 0.118 7.341 ± 0.650
str_mps_varying_inner_product_200 10.260 ± 0.044 9.826 ± 0.050 8.178 ± 0.021 9.884 ± 0.109 15.921 ± 5.393
str_nw_mera_closed_120 383.280 ± 2.502 370.964 ± 6.333 293.342 ± 1.175 277.434 ± 2.290 461.747 ± 27.942
str_nw_mera_open_26 411.038 ± 3.382 458.514 ± 99.271 245.024 ± 1.744 337.496 ± 6.789 631.238 ± 23.747
tensornetwork_permutation_focus_step409_316 212.276 ± 2.742 238.555 ± 20.734 147.208 ± 0.417 220.242 ± 3.315 273.073 ± 26.585
tensornetwork_permutation_light_415 206.231 ± 27.970 207.306 ± 22.942 144.041 ± 0.368 227.021 ± 7.230 265.425 ± 15.746

Threads: 8

Source report: data/results/zen5-cpu/cpu/einsum/20261028_001612/report.md.

Latest run: ./scripts/run_all.sh 8.

This file is generated from one suite run under data/results/zen5-cpu/cpu/einsum/20261028_001612.

CPU Information

Thread Environment

Tenferro CPU BLAS Backend

Python Backend Providers

Julia / OMEinsum.jl Backend

Logs:

Strategy: opt_flops

Median ± IQR (ms). JULIA_NUM_THREADS=8, OMP_NUM_THREADS=8, RAYON_NUM_THREADS=8.

Instance tenferro-rs trace mode (ms) tenferro-rs eager mode (ms) PyTorch Python (ms) JAX Python (XLA CPU dot) (ms) OMEinsum.jl OpenBLAS (ms)
bin_batched_matmul_b32_m128_n128_k128 3.345 ± 0.968 2.072 ± 0.023 0.551 ± 0.015 1.064 ± 0.087 9.428 ± 10.707
bin_batched_matmul_b32_m64_n64_k64 1.897 ± 0.350 1.092 ± 0.006 0.080 ± 0.002 0.763 ± 0.219 0.768 ± 0.817
bin_batched_outer_product_compact_j16_k16_o64_t64 1.190 ± 0.416 0.791 ± 0.007 0.030 ± 0.001 0.438 ± 0.247 1.553 ± 1.404
bin_batched_outer_product_noncompact_j16_k16_o64_t64 0.839 ± 0.008 0.405 ± 0.006 0.030 ± 0.005 0.474 ± 0.268 4.301 ± 1.597
bin_elementwise_mul_2048x2048 2.097 ± 0.037 2.155 ± 0.340 2.892 ± 0.277 5.322 ± 0.198 9.144 ± 0.522
bin_elementwise_mul_512x512 2.448 ± 0.140 1.085 ± 0.397 0.021 ± 0.001 0.252 ± 0.201 0.116 ± 0.399
bin_matmul_1024 18.409 ± 3.397 14.013 ± 1.839 12.585 ± 0.165 8.858 ± 0.394 30.558 ± 15.645
bin_matmul_256 1.749 ± 0.300 0.111 ± 0.002 0.281 ± 0.006 0.218 ± 0.056 0.412 ± 0.006
bin_omeinsum_batched_matmul_8x8_batch_4 0.008 ± 0.000 0.005 ± 0.000 0.018 ± 0.001 0.103 ± 0.006 0.010 ± 0.001
bin_omeinsum_high_d_12x12_contract_4_batch_4 0.029 ± 0.001 0.027 ± 0.001 0.088 ± 0.002 0.147 ± 0.074 0.038 ± 0.026
bin_omeinsum_matmul_10x10 0.046 ± 0.001 0.009 ± 0.000 0.017 ± 0.000 0.063 ± 0.029 0.018 ± 0.002
bin_outer_product_4096 9.392 ± 2.800 9.046 ± 0.626 7.454 ± 0.282 12.498 ± 0.468 197.616 ± 6.924
bin_permuted_r4_abcd_dbef_acef_d32 23.002 ± 0.384 15.224 ± 5.549 11.051 ± 0.247 9.936 ± 0.297 23.495 ± 6.107
gm_queen5_5_3.wcsp 4009.796 ± 39.209 2214.568 ± 49.833 1243.864 ± 12.744 1563.443 ± 27.959 5626.624 ± 34.861
lm_batch_likelihood_brackets_4_4d 21.416 ± 2.632 39.825 ± 8.985 6.353 ± 0.124 13.519 ± 0.457 50.697 ± 25.946
lm_batch_likelihood_sentence_3_12d 32.823 ± 0.397 125.616 ± 12.520 27.876 ± 0.155 22.639 ± 3.793 74.549 ± 27.278
lm_batch_likelihood_sentence_4_4d 21.564 ± 0.223 39.085 ± 3.573 6.656 ± 0.040 13.259 ± 0.177 44.301 ± 22.738
nary_matmul_chain_64 0.093 ± 0.001 0.077 ± 0.001 0.031 ± 0.001 0.085 ± 0.010 0.080 ± 0.026
str_matrix_chain_multiplication_100 7.690 ± 0.055 8.091 ± 0.048 3.861 ± 0.020 5.332 ± 0.152 8.560 ± 1.118
str_mps_varying_inner_product_200 15.722 ± 6.264 13.648 ± 7.724 9.944 ± 0.035 14.521 ± 0.123 22.127 ± 2.937
str_nw_mera_closed_120 418.191 ± 4.944 454.306 ± 16.615 357.773 ± 2.418 328.203 ± 16.327 565.853 ± 13.959
str_nw_mera_open_26 449.497 ± 2.409 434.313 ± 27.184 222.787 ± 1.270 278.737 ± 3.608 705.791 ± 16.419
tensornetwork_permutation_focus_step409_316 233.983 ± 15.852 274.506 ± 12.049 128.458 ± 0.885 277.987 ± 3.846 334.694 ± 38.522
tensornetwork_permutation_light_415 218.170 ± 2.015 229.150 ± 2.323 126.197 ± 0.933 286.035 ± 6.460 326.374 ± 34.140
Strategy: opt_size

Median ± IQR (ms). JULIA_NUM_THREADS=8, OMP_NUM_THREADS=8, RAYON_NUM_THREADS=8.

Instance tenferro-rs trace mode (ms) tenferro-rs eager mode (ms) PyTorch Python (ms) JAX Python (XLA CPU dot) (ms) OMEinsum.jl OpenBLAS (ms)
bin_batched_matmul_b32_m128_n128_k128 3.345 ± 0.968 2.072 ± 0.023 0.551 ± 0.015 1.064 ± 0.087 9.428 ± 10.707
bin_batched_matmul_b32_m64_n64_k64 1.897 ± 0.350 1.092 ± 0.006 0.080 ± 0.002 0.763 ± 0.219 0.768 ± 0.817
bin_batched_outer_product_compact_j16_k16_o64_t64 1.190 ± 0.416 0.791 ± 0.007 0.030 ± 0.001 0.438 ± 0.247 1.553 ± 1.404
bin_batched_outer_product_noncompact_j16_k16_o64_t64 0.839 ± 0.008 0.405 ± 0.006 0.030 ± 0.005 0.474 ± 0.268 4.301 ± 1.597
bin_elementwise_mul_2048x2048 2.097 ± 0.037 2.155 ± 0.340 2.892 ± 0.277 5.322 ± 0.198 9.144 ± 0.522
bin_elementwise_mul_512x512 2.448 ± 0.140 1.085 ± 0.397 0.021 ± 0.001 0.252 ± 0.201 0.116 ± 0.399
bin_matmul_1024 18.409 ± 3.397 14.013 ± 1.839 12.585 ± 0.165 8.858 ± 0.394 30.558 ± 15.645
bin_matmul_256 1.749 ± 0.300 0.111 ± 0.002 0.281 ± 0.006 0.218 ± 0.056 0.412 ± 0.006
bin_omeinsum_batched_matmul_8x8_batch_4 0.008 ± 0.000 0.005 ± 0.000 0.018 ± 0.001 0.103 ± 0.006 0.010 ± 0.001
bin_omeinsum_high_d_12x12_contract_4_batch_4 0.029 ± 0.001 0.027 ± 0.001 0.088 ± 0.002 0.147 ± 0.074 0.038 ± 0.026
bin_omeinsum_matmul_10x10 0.046 ± 0.001 0.009 ± 0.000 0.017 ± 0.000 0.063 ± 0.029 0.018 ± 0.002
bin_outer_product_4096 9.392 ± 2.800 9.046 ± 0.626 7.454 ± 0.282 12.498 ± 0.468 197.616 ± 6.924
bin_permuted_r4_abcd_dbef_acef_d32 23.002 ± 0.384 15.224 ± 5.549 11.051 ± 0.247 9.936 ± 0.297 23.495 ± 6.107
gm_queen5_5_3.wcsp 1227.358 ± 7.209 939.443 ± 178.223 510.419 ± 3.960 575.919 ± 11.914 1988.501 ± 81.016
lm_batch_likelihood_brackets_4_4d 27.041 ± 0.326 43.212 ± 0.660 5.971 ± 0.214 13.143 ± 0.296 35.960 ± 19.421
lm_batch_likelihood_sentence_3_12d 24.891 ± 1.440 149.707 ± 61.922 15.592 ± 0.142 26.671 ± 0.376 72.385 ± 25.219
lm_batch_likelihood_sentence_4_4d 30.016 ± 0.117 42.034 ± 12.687 6.729 ± 0.059 13.175 ± 0.373 35.087 ± 22.288
nary_matmul_chain_64 0.093 ± 0.001 0.077 ± 0.001 0.031 ± 0.001 0.085 ± 0.010 0.080 ± 0.026
str_matrix_chain_multiplication_100 8.085 ± 0.118 8.069 ± 0.224 3.851 ± 0.018 5.284 ± 0.161 8.224 ± 0.755
str_mps_varying_inner_product_200 15.417 ± 5.640 17.712 ± 5.773 9.840 ± 0.021 14.167 ± 0.066 19.372 ± 5.262
str_nw_mera_closed_120 384.184 ± 2.145 359.499 ± 6.435 262.954 ± 1.490 233.187 ± 8.547 468.168 ± 36.532
str_nw_mera_open_26 455.975 ± 6.457 432.567 ± 40.124 224.839 ± 2.295 290.107 ± 7.740 706.364 ± 46.942
tensornetwork_permutation_focus_step409_316 233.983 ± 15.852 274.506 ± 12.049 128.458 ± 0.885 277.987 ± 3.846 334.694 ± 38.522
tensornetwork_permutation_light_415 218.170 ± 2.015 229.150 ± 2.323 126.197 ± 0.933 286.035 ± 6.460 326.374 ± 34.140