Skip to content

Use --find-links for MLIR-AIE and LLVM-AIE wheel sources#107

Open
x4v13r64 wants to merge 1 commit into
amd:develfrom
x4v13r64:fix/pin-package-indexes
Open

Use --find-links for MLIR-AIE and LLVM-AIE wheel sources#107
x4v13r64 wants to merge 1 commit into
amd:develfrom
x4v13r64:fix/pin-package-indexes

Conversation

@x4v13r64

Copy link
Copy Markdown

The GitHub Release URLs for mlir_aie and llvm-aie are currently added as --extra-index-url entries alongside PyPI. This means pip queries all indexes for all packages, which can cause resolution issues when the same package name exists on multiple indexes at different versions. Switching to --find-links adds the wheel files as candidates without creating a competing package index.

@github-actions

github-actions Bot commented May 20, 2026

Copy link
Copy Markdown
Contributor

CI Test Results

a678be7 (2026_07_09_22_51_46)

IRON - CI Summary

Examples

iron/applications/llama_3.2_1b
Test Krackan Status Krackan Phoenix Status Phoenix
test_llama_3_2_1b[llama_3.2_1b_prompt_1024_tokens_1] - - -
test_llama_3_2_1b[llama_3.2_1b_prompt_1024_tokens_40] - - -
test_llama_3_2_1b[llama_3.2_1b_prompt_13_tokens_1] - - -
test_llama_3_2_1b[llama_3.2_1b_prompt_13_tokens_40] - - -

Small

iron/operators/axpy
Test Krackan Status Krackan Latency (mean) Phoenix Status Phoenix Latency (mean)
test_axpy[input_length_2048-num_aie_columns_1-tile_size_2048-scalar_factor_3.0] 160.62 331.14
test_axpy[input_length_2048-num_aie_columns_2-tile_size_1024-scalar_factor_3.0] 185.76 443.18
test_axpy[input_length_2048-num_aie_columns_4-tile_size_512-scalar_factor_3.0] 190.18 467.34
test_axpy[input_length_2048-num_aie_columns_8-tile_size_256-scalar_factor_3.0] 222.02 - -
iron/operators/dequant
Test Krackan Status Krackan Latency (mean) Phoenix Status Phoenix Latency (mean)
test_dequant[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-group_size_32] 157.10 359.10
test_dequant[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-group_size_32] 158.52 775.30
test_dequant[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-group_size_32] 154.56 440.12
test_dequant[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-group_size_32] 168.00 468.46
test_dequant[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-group_size_32] 185.04 487.04
test_dequant[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-group_size_32] 163.76 726.38
test_dequant[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256-group_size_32] 191.56 - -
test_dequant[input_length_2048-num_aie_columns_8-num_channels_2-tile_size_128-group_size_32] 192.74 - -
iron/operators/elementwise_add
Test Krackan Status Krackan Latency (mean) Phoenix Status Phoenix Latency (mean)
test_elementwise_add[input_length_2048-num_aie_columns_1-tile_size_2048] 195.40 317.22
test_elementwise_add[input_length_2048-num_aie_columns_2-tile_size_1024] 191.10 431.86
test_elementwise_add[input_length_2048-num_aie_columns_4-tile_size_512] 207.18 333.66
test_elementwise_add[input_length_2048-num_aie_columns_8-tile_size_256] 199.42 - -
iron/operators/elementwise_mul
Test Krackan Status Krackan Latency (mean) Phoenix Status Phoenix Latency (mean)
test_elementwise_mul[input_length_2048-num_aie_columns_1-tile_size_2048] 177.74 428.70
test_elementwise_mul[input_length_2048-num_aie_columns_2-tile_size_1024] 160.20 526.02
test_elementwise_mul[input_length_2048-num_aie_columns_4-tile_size_512] 187.08 584.00
test_elementwise_mul[input_length_2048-num_aie_columns_8-tile_size_256] 242.84 - -
iron/operators/gelu
Test Krackan Status Krackan Latency (mean) Phoenix Status Phoenix Latency (mean)
test_gelu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048] 175.18 288.32
test_gelu[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024] 169.78 397.76
test_gelu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024] 187.14 370.66
test_gelu[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512] 175.76 306.66
test_gelu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512] 178.94 343.14
test_gelu[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256] 180.96 324.96
test_gelu[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256] 189.18 - -
test_gelu[input_length_2048-num_aie_columns_8-num_channels_2-tile_size_128] 215.72 - -
iron/operators/gemm
Test Krackan Status Krackan Latency (mean) Phoenix Status Phoenix Latency (mean)
test_gemm[M_1792-K_896-N_1152-num_aie_columns_8-b_col_maj_False-c_col_maj_True-m_64-k_32-n_48-trace_size_0-partition_N_1] 2379.14 - -
test_gemm[M_192-K_384-N_64-num_aie_columns_4-b_col_maj_False-c_col_maj_False-m_48-k_96-n_16-trace_size_0-partition_N_1] 252.24 504.82
test_gemm[M_192-K_384-N_64-num_aie_columns_4-b_col_maj_True-c_col_maj_True-m_48-k_96-n_16-trace_size_0-partition_N_1] 302.12 853.84
test_gemm[M_2048-K_2048-N_2048-num_aie_columns_1-b_col_maj_False-c_col_maj_False-m_64-k_64-n_64-trace_size_0-partition_N_1] 48560.14 82114.24
test_gemm[M_2048-K_2048-N_2048-num_aie_columns_2-b_col_maj_True-c_col_maj_False-m_64-k_64-n_64-trace_size_0-partition_N_1] 28677.34 25361.74
test_gemm[M_2048-K_2048-N_2048-num_aie_columns_8-b_col_maj_True-c_col_maj_True-m_64-k_64-n_64-trace_size_0-partition_N_1] 7880.12 - -
test_gemm[M_384-K_1536-N_1792-num_aie_columns_4-b_col_maj_True-c_col_maj_False-m_32-k_48-n_64-trace_size_0-partition_N_1] 2427.32 3617.20
test_gemm[M_64-K_512-N_256-num_aie_columns_4-b_col_maj_True-c_col_maj_False-m_16-k_64-n_64-trace_size_0-partition_N_4] 3841.52 6411.82
test_gemm[M_896-K_1792-N_640-num_aie_columns_8-b_col_maj_False-c_col_maj_True-m_32-k_64-n_80-trace_size_0-partition_N_1] 1447.74 - -
iron/operators/gemv
Test Krackan Status Krackan Latency (mean) Phoenix Status Phoenix Latency (mean)
test_gemv[M_128-K_128-num_aie_columns_1-tile_size_input_32-tile_size_output_128] 0.22 0.11
test_gemv[M_2048-K_8192-num_aie_columns_1-tile_size_input_1-tile_size_output_2048] 12.83 3.73
test_gemv[M_2048-K_8192-num_aie_columns_2-tile_size_input_1-tile_size_output_1024] 24.64 6.13
test_gemv[M_2048-K_8192-num_aie_columns_4-tile_size_input_1-tile_size_output_512] 38.36 9.87
test_gemv[M_2048-K_8192-num_aie_columns_8-tile_size_input_1-tile_size_output_256] 39.84 - -
test_gemv[M_8192-K_2048-num_aie_columns_1-tile_size_input_4-tile_size_output_1024] 12.93 3.58
test_gemv[M_8192-K_2048-num_aie_columns_2-tile_size_input_4-tile_size_output_1024] 24.53 5.89
test_gemv[M_8192-K_2048-num_aie_columns_4-tile_size_input_4-tile_size_output_1024] 41.30 11.25
test_gemv[M_8192-K_2048-num_aie_columns_8-tile_size_input_4-tile_size_output_1024] 43.25 - -
iron/operators/layer_norm
Test Krackan Status Krackan Latency (mean) Phoenix Status Phoenix Latency (mean)
test_layer_norm[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048] 143.94 295.42
test_layer_norm[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024] 192.28 440.66
test_layer_norm[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024] 180.82 358.10
test_layer_norm[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512] 170.56 526.34
test_layer_norm[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512] 190.62 679.00
test_layer_norm[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256] 189.68 481.54
test_layer_norm[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256] 184.60 - -
test_layer_norm[input_length_2048-num_aie_columns_8-num_channels_2-tile_size_128] 257.66 - -
iron/operators/leaky_relu
Test Krackan Status Krackan Latency (mean) Phoenix Status Phoenix Latency (mean)
test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-alpha_0.01] 146.62 422.60
test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-alpha_0.1] 157.28 312.92
test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-alpha_0.25] 166.16 277.28
test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-alpha_0.01] 141.80 341.34
test_leaky_relu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-alpha_0.01] 151.68 405.08
test_leaky_relu[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-alpha_0.01] 173.72 405.62
test_leaky_relu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-alpha_0.01] 162.02 432.74
test_leaky_relu[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-alpha_0.01] 174.38 543.68
test_leaky_relu[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256-alpha_0.01] 164.88 - -
test_leaky_relu[input_length_2048-num_aie_columns_8-num_channels_2-tile_size_128-alpha_0.01] 225.20 - -
iron/operators/mem_copy
Test Krackan Status Krackan Latency (mean) Phoenix Status Phoenix Latency (mean)
test_mem_copy[input_length_2048-num_cores_1-num_channels_1-bypass_False-tile_size_2048] 157.10 440.22
test_mem_copy[input_length_2048-num_cores_16-num_channels_2-bypass_False-tile_size_128] 212.18 - -
test_mem_copy[input_length_2048-num_cores_2-num_channels_1-bypass_False-tile_size_1024] 172.12 403.76
test_mem_copy[input_length_2048-num_cores_2-num_channels_2-bypass_False-tile_size_1024] 159.38 488.00
test_mem_copy[input_length_2048-num_cores_4-num_channels_1-bypass_False-tile_size_512] 189.30 602.68
test_mem_copy[input_length_2048-num_cores_4-num_channels_2-bypass_False-tile_size_512] 195.90 363.72
test_mem_copy[input_length_2048-num_cores_8-num_channels_1-bypass_False-tile_size_256] 213.58 - -
test_mem_copy[input_length_2048-num_cores_8-num_channels_2-bypass_False-tile_size_256] 178.20 488.80
iron/operators/relu
Test Krackan Status Krackan Latency (mean) Phoenix Status Phoenix Latency (mean)
test_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048] 169.48 285.18
test_relu[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024] 155.12 240.32
test_relu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024] 162.20 267.02
test_relu[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512] 201.66 338.10
test_relu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512] 173.72 350.14
test_relu[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256] 180.08 377.18
test_relu[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256] 196.38 - -
test_relu[input_length_2048-num_aie_columns_8-num_channels_2-tile_size_128] 223.76 - -
iron/operators/rms_norm
Test Krackan Status Krackan Latency (mean) Phoenix Status Phoenix Latency (mean)
test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-weighted_False] 196.30 413.28
test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-weighted_True] 164.60 304.50
test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-weighted_False] 162.64 360.94
test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-weighted_True] 147.40 417.88
test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-weighted_False] 196.30 399.40
test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-weighted_True] 163.82 393.70
test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-weighted_False] 173.10 385.72
test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-weighted_True] 161.28 369.60
test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-weighted_False] 164.80 604.96
test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-weighted_True] 176.12 498.86
test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-weighted_False] 181.82 377.50
test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-weighted_True] 179.82 - -
test_rms_norm[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256-weighted_False] 155.32 - -
test_rms_norm[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256-weighted_True] 190.18 - -
test_rms_norm[input_length_2048-num_aie_columns_8-num_channels_2-tile_size_128-weighted_False] 228.42 - -
iron/operators/rope
Test Krackan Status Krackan Latency (mean) Phoenix Status Phoenix Latency (mean)
test_rope[rows_32-cols_512-angle_rows_32-aie_columns_1-method_type_0] 203.56 431.12
test_rope[rows_32-cols_512-angle_rows_32-aie_columns_2-method_type_0] 178.62 403.56
test_rope[rows_32-cols_512-angle_rows_32-aie_columns_4-method_type_0] 199.80 1033.98
test_rope[rows_32-cols_512-angle_rows_32-aie_columns_8-method_type_0] 205.18 - -
test_rope[rows_32-cols_512-angle_rows_8-aie_columns_1-method_type_0] 189.50 354.14
test_rope[rows_32-cols_512-angle_rows_8-aie_columns_2-method_type_0] 193.44 348.28
test_rope[rows_32-cols_512-angle_rows_8-aie_columns_4-method_type_0] 178.64 361.62
test_rope[rows_32-cols_512-angle_rows_8-aie_columns_8-method_type_0] 177.78 - -
iron/operators/sigmoid
Test Krackan Status Krackan Latency (mean) Phoenix Status Phoenix Latency (mean)
test_sigmoid[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048] 210.96 367.60
test_sigmoid[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024] 174.26 424.42
test_sigmoid[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024] 188.84 534.94
test_sigmoid[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512] 190.88 360.54
test_sigmoid[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512] 175.86 412.92
test_sigmoid[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256] 234.70 430.38
test_sigmoid[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256] 205.66 - -
test_sigmoid[input_length_2048-num_aie_columns_8-num_channels_2-tile_size_128] 228.06 - -
iron/operators/silu
Test Krackan Status Krackan Latency (mean) Phoenix Status Phoenix Latency (mean)
test_silu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048] 182.50 434.06
test_silu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024] 156.94 358.88
test_silu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512] 162.80 530.90
test_silu[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256] 168.82 - -
iron/operators/softmax
Test Krackan Status Krackan Latency (mean) Phoenix Status Phoenix Latency (mean)
test_softmax[input_length_32768-num_aie_columns_2-num_channels_2-tile_size_1024] 163.92 693.14
test_softmax[input_length_32768-num_aie_columns_2-num_channels_2-tile_size_2048] 159.72 714.56
test_softmax[input_length_32768-num_aie_columns_2-num_channels_2-tile_size_512] 166.96 450.48
iron/operators/swiglu_decode
Test Krackan Status Krackan Latency (mean) Phoenix Status Phoenix Latency (mean)
test_swiglu_decode[embedding_dim_1024-hidden_dim_3584] 950.34 12270.66
test_swiglu_decode[embedding_dim_2048-hidden_dim_2048] 1014.12 13252.32
iron/operators/swiglu_prefill
Test Krackan Status Krackan Latency (mean) Phoenix Status Phoenix Latency (mean)
test_swiglu_prefill[seq_len_256-embedding_dim_2048-hidden_dim_2048-prio_accuracy_False] 2232.07 21981.84
iron/operators/tanh
Test Krackan Status Krackan Latency (mean) Phoenix Status Phoenix Latency (mean)
test_tanh[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048] 184.54 396.60
test_tanh[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024] 180.42 442.22
test_tanh[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024] 188.02 712.68
test_tanh[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512] 193.98 437.16
test_tanh[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512] 190.20 388.46
test_tanh[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256] 207.16 424.24
test_tanh[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256] 196.88 - -
test_tanh[input_length_2048-num_aie_columns_8-num_channels_2-tile_size_128] 239.32 - -
iron/operators/transpose
Test Krackan Status Krackan Latency (mean) Phoenix Status Phoenix Latency (mean)
test_transpose[M_2048-N_64-aie_columns_1-channels_1-m_64-n_64-s_8-num_batches_1] 209.94 498.28
test_transpose[M_2048-N_64-aie_columns_1-channels_1-m_64-n_64-s_8-num_batches_2] 216.92 1794.46
test_transpose[M_2048-N_64-aie_columns_1-channels_2-m_64-n_64-s_8-num_batches_1] 198.24 572.38
Krackan - Small

IRON

Tested on 2026_07_09_22_51_46 at commit a678be7.

iron/operators/axpy
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_axpy[input_length_2048-num_aie_columns_1-tile_size_2048-scalar_factor_3.0]✅ 5/5160.620.08n/a
test_axpy[input_length_2048-num_aie_columns_2-tile_size_1024-scalar_factor_3.0]✅ 5/5185.760.07n/a
test_axpy[input_length_2048-num_aie_columns_4-tile_size_512-scalar_factor_3.0]✅ 5/5190.180.07n/a
test_axpy[input_length_2048-num_aie_columns_8-tile_size_256-scalar_factor_3.0]✅ 5/5222.020.06n/a
iron/operators/dequant
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_dequant[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-group_size_32]✅ 5/5157.100.03n/a
test_dequant[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-group_size_32]✅ 5/5158.520.03n/a
test_dequant[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-group_size_32]✅ 5/5154.560.03n/a
test_dequant[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-group_size_32]✅ 5/5168.000.03n/a
test_dequant[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-group_size_32]✅ 5/5185.040.03n/a
test_dequant[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-group_size_32]✅ 5/5163.760.03n/a
test_dequant[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256-group_size_32]✅ 5/5191.560.03n/a
test_dequant[input_length_2048-num_aie_columns_8-num_channels_2-tile_size_128-group_size_32]✅ 5/5192.740.03n/a
iron/operators/elementwise_add
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_elementwise_add[input_length_2048-num_aie_columns_1-tile_size_2048]✅ 5/5195.400.07n/a
test_elementwise_add[input_length_2048-num_aie_columns_2-tile_size_1024]✅ 5/5191.100.07n/a
test_elementwise_add[input_length_2048-num_aie_columns_4-tile_size_512]✅ 5/5207.180.06n/a
test_elementwise_add[input_length_2048-num_aie_columns_8-tile_size_256]✅ 5/5199.420.06n/a
iron/operators/elementwise_mul
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_elementwise_mul[input_length_2048-num_aie_columns_1-tile_size_2048]✅ 5/5177.740.07n/a
test_elementwise_mul[input_length_2048-num_aie_columns_2-tile_size_1024]✅ 5/5160.200.08n/a
test_elementwise_mul[input_length_2048-num_aie_columns_4-tile_size_512]✅ 5/5187.080.07n/a
test_elementwise_mul[input_length_2048-num_aie_columns_8-tile_size_256]✅ 5/5242.840.05n/a
iron/operators/gelu
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_gelu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048]✅ 5/5175.180.05n/a
test_gelu[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024]✅ 5/5169.780.05n/a
test_gelu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024]✅ 5/5187.140.05n/a
test_gelu[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512]✅ 5/5175.760.05n/a
test_gelu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512]✅ 5/5178.940.05n/a
test_gelu[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256]✅ 5/5180.960.05n/a
test_gelu[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256]✅ 5/5189.180.04n/a
test_gelu[input_length_2048-num_aie_columns_8-num_channels_2-tile_size_128]✅ 5/5215.720.04n/a
iron/operators/gemm
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_gemm[M_1792-K_896-N_1152-num_aie_columns_8-b_col_maj_False-c_col_maj_True-m_64-k_32-n_48-trace_size_0-partition_N_1]✅ 5/52379.143.981566.11
test_gemm[M_192-K_384-N_64-num_aie_columns_4-b_col_maj_False-c_col_maj_False-m_48-k_96-n_16-trace_size_0-partition_N_1]✅ 5/5252.240.9038.49
test_gemm[M_192-K_384-N_64-num_aie_columns_4-b_col_maj_True-c_col_maj_True-m_48-k_96-n_16-trace_size_0-partition_N_1]✅ 5/5302.120.7732.97
test_gemm[M_2048-K_2048-N_2048-num_aie_columns_1-b_col_maj_False-c_col_maj_False-m_64-k_64-n_64-trace_size_0-partition_N_1]✅ 5/548560.140.52353.79
test_gemm[M_2048-K_2048-N_2048-num_aie_columns_2-b_col_maj_True-c_col_maj_False-m_64-k_64-n_64-trace_size_0-partition_N_1]✅ 5/528677.340.88599.08
test_gemm[M_2048-K_2048-N_2048-num_aie_columns_8-b_col_maj_True-c_col_maj_True-m_64-k_64-n_64-trace_size_0-partition_N_1]✅ 5/57880.123.202181.32
test_gemm[M_384-K_1536-N_1792-num_aie_columns_4-b_col_maj_True-c_col_maj_False-m_32-k_48-n_64-trace_size_0-partition_N_1]✅ 5/52427.323.36881.04
test_gemm[M_64-K_512-N_256-num_aie_columns_4-b_col_maj_True-c_col_maj_False-m_16-k_64-n_64-trace_size_0-partition_N_4]✅ 5/53841.520.3418.42
test_gemm[M_896-K_1792-N_640-num_aie_columns_8-b_col_maj_False-c_col_maj_True-m_32-k_64-n_80-trace_size_0-partition_N_1]✅ 5/51447.744.761470.22
iron/operators/gemv
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_gemv[M_128-K_128-num_aie_columns_1-tile_size_input_32-tile_size_output_128]✅ 5/5n/a0.220.22
test_gemv[M_2048-K_8192-num_aie_columns_1-tile_size_input_1-tile_size_output_2048]✅ 5/5n/a12.8312.82
test_gemv[M_2048-K_8192-num_aie_columns_2-tile_size_input_1-tile_size_output_1024]✅ 5/5n/a24.6424.62
test_gemv[M_2048-K_8192-num_aie_columns_4-tile_size_input_1-tile_size_output_512]✅ 5/5n/a38.3638.34
test_gemv[M_2048-K_8192-num_aie_columns_8-tile_size_input_1-tile_size_output_256]✅ 5/5n/a39.8439.82
test_gemv[M_8192-K_2048-num_aie_columns_1-tile_size_input_4-tile_size_output_1024]✅ 5/5n/a12.9312.92
test_gemv[M_8192-K_2048-num_aie_columns_2-tile_size_input_4-tile_size_output_1024]✅ 5/5n/a24.5324.51
test_gemv[M_8192-K_2048-num_aie_columns_4-tile_size_input_4-tile_size_output_1024]✅ 5/5n/a41.3041.27
test_gemv[M_8192-K_2048-num_aie_columns_8-tile_size_input_4-tile_size_output_1024]✅ 5/5n/a43.2543.23
iron/operators/layer_norm
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_layer_norm[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048]✅ 5/5143.940.06n/a
test_layer_norm[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024]✅ 5/5192.280.05n/a
test_layer_norm[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024]✅ 5/5180.820.05n/a
test_layer_norm[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512]✅ 5/5170.560.05n/a
test_layer_norm[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512]✅ 5/5190.620.04n/a
test_layer_norm[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256]✅ 5/5189.680.04n/a
test_layer_norm[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256]✅ 5/5184.600.05n/a
test_layer_norm[input_length_2048-num_aie_columns_8-num_channels_2-tile_size_128]✅ 5/5257.660.03n/a
iron/operators/leaky_relu
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-alpha_0.01]✅ 5/5146.620.06n/a
test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-alpha_0.1]✅ 5/5157.280.06n/a
test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-alpha_0.25]✅ 5/5166.160.05n/a
test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-alpha_0.01]✅ 5/5141.800.06n/a
test_leaky_relu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-alpha_0.01]✅ 5/5151.680.06n/a
test_leaky_relu[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-alpha_0.01]✅ 5/5173.720.05n/a
test_leaky_relu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-alpha_0.01]✅ 5/5162.020.05n/a
test_leaky_relu[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-alpha_0.01]✅ 5/5174.380.05n/a
test_leaky_relu[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256-alpha_0.01]✅ 5/5164.880.05n/a
test_leaky_relu[input_length_2048-num_aie_columns_8-num_channels_2-tile_size_128-alpha_0.01]✅ 5/5225.200.04n/a
iron/operators/mem_copy
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_mem_copy[input_length_2048-num_cores_1-num_channels_1-bypass_False-tile_size_2048]✅ 5/5157.100.05n/a
test_mem_copy[input_length_2048-num_cores_16-num_channels_2-bypass_False-tile_size_128]✅ 5/5212.180.04n/a
test_mem_copy[input_length_2048-num_cores_2-num_channels_1-bypass_False-tile_size_1024]✅ 5/5172.120.05n/a
test_mem_copy[input_length_2048-num_cores_2-num_channels_2-bypass_False-tile_size_1024]✅ 5/5159.380.05n/a
test_mem_copy[input_length_2048-num_cores_4-num_channels_1-bypass_False-tile_size_512]✅ 5/5189.300.04n/a
test_mem_copy[input_length_2048-num_cores_4-num_channels_2-bypass_False-tile_size_512]✅ 5/5195.900.04n/a
test_mem_copy[input_length_2048-num_cores_8-num_channels_1-bypass_False-tile_size_256]✅ 5/5213.580.04n/a
test_mem_copy[input_length_2048-num_cores_8-num_channels_2-bypass_False-tile_size_256]✅ 5/5178.200.05n/a
iron/operators/relu
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048]✅ 5/5169.480.05n/a
test_relu[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024]✅ 5/5155.120.05n/a
test_relu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024]✅ 5/5162.200.05n/a
test_relu[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512]✅ 5/5201.660.04n/a
test_relu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512]✅ 5/5173.720.05n/a
test_relu[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256]✅ 5/5180.080.05n/a
test_relu[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256]✅ 5/5196.380.04n/a
test_relu[input_length_2048-num_aie_columns_8-num_channels_2-tile_size_128]✅ 5/5223.760.04n/a
iron/operators/rms_norm
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-weighted_False]✅ 5/5196.300.05n/a
test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-weighted_True]✅ 5/5164.600.08n/a
test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-weighted_False]✅ 5/5162.640.05n/a
test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-weighted_True]✅ 5/5147.400.07n/a
test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-weighted_False]✅ 5/5196.300.04n/a
test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-weighted_True]✅ 5/5163.820.07n/a
test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-weighted_False]✅ 5/5173.100.05n/a
test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-weighted_True]✅ 5/5161.280.06n/a
test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-weighted_False]✅ 5/5164.800.05n/a
test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-weighted_True]✅ 5/5176.120.05n/a
test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-weighted_False]✅ 5/5181.820.05n/a
test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-weighted_True]✅ 5/5179.820.05n/a
test_rms_norm[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256-weighted_False]✅ 5/5155.320.05n/a
test_rms_norm[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256-weighted_True]✅ 5/5190.180.05n/a
test_rms_norm[input_length_2048-num_aie_columns_8-num_channels_2-tile_size_128-weighted_False]✅ 5/5228.420.04n/a
iron/operators/rope
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_rope[rows_32-cols_512-angle_rows_32-aie_columns_1-method_type_0]✅ 5/5203.560.53n/a
test_rope[rows_32-cols_512-angle_rows_32-aie_columns_2-method_type_0]✅ 5/5178.620.56n/a
test_rope[rows_32-cols_512-angle_rows_32-aie_columns_4-method_type_0]✅ 5/5199.800.52n/a
test_rope[rows_32-cols_512-angle_rows_32-aie_columns_8-method_type_0]✅ 5/5205.180.49n/a
test_rope[rows_32-cols_512-angle_rows_8-aie_columns_1-method_type_0]✅ 5/5189.500.40n/a
test_rope[rows_32-cols_512-angle_rows_8-aie_columns_2-method_type_0]✅ 5/5193.440.40n/a
test_rope[rows_32-cols_512-angle_rows_8-aie_columns_4-method_type_0]✅ 5/5178.640.42n/a
test_rope[rows_32-cols_512-angle_rows_8-aie_columns_8-method_type_0]✅ 5/5177.780.42n/a
iron/operators/sigmoid
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_sigmoid[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048]✅ 5/5210.960.04n/a
test_sigmoid[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024]✅ 5/5174.260.05n/a
test_sigmoid[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024]✅ 5/5188.840.04n/a
test_sigmoid[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512]✅ 5/5190.880.04n/a
test_sigmoid[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512]✅ 5/5175.860.05n/a
test_sigmoid[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256]✅ 5/5234.700.04n/a
test_sigmoid[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256]✅ 5/5205.660.04n/a
test_sigmoid[input_length_2048-num_aie_columns_8-num_channels_2-tile_size_128]✅ 5/5228.060.04n/a
iron/operators/silu
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_silu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048]✅ 5/5182.500.05n/a
test_silu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024]✅ 5/5156.940.05n/a
test_silu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512]✅ 5/5162.800.05n/a
test_silu[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256]✅ 5/5168.820.05n/a
iron/operators/softmax
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_softmax[input_length_32768-num_aie_columns_2-num_channels_2-tile_size_1024]✅ 5/5163.920.85n/a
test_softmax[input_length_32768-num_aie_columns_2-num_channels_2-tile_size_2048]✅ 5/5159.720.83n/a
test_softmax[input_length_32768-num_aie_columns_2-num_channels_2-tile_size_512]✅ 5/5166.960.81n/a
iron/operators/swiglu_decode
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_swiglu_decode[embedding_dim_1024-hidden_dim_3584]✅ 5/5950.340.00n/a
test_swiglu_decode[embedding_dim_2048-hidden_dim_2048]✅ 5/51014.120.01n/a
iron/operators/swiglu_prefill
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_swiglu_prefill[seq_len_256-embedding_dim_2048-hidden_dim_2048-prio_accuracy_False]✅ 5/52232.070.94n/a
iron/operators/tanh
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_tanh[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048]✅ 5/5184.540.05n/a
test_tanh[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024]✅ 5/5180.420.05n/a
test_tanh[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024]✅ 5/5188.020.04n/a
test_tanh[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512]✅ 5/5193.980.04n/a
test_tanh[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512]✅ 5/5190.200.04n/a
test_tanh[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256]✅ 5/5207.160.04n/a
test_tanh[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256]✅ 5/5196.880.04n/a
test_tanh[input_length_2048-num_aie_columns_8-num_channels_2-tile_size_128]✅ 5/5239.320.03n/a
iron/operators/transpose
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_transpose[M_2048-N_64-aie_columns_1-channels_1-m_64-n_64-s_8-num_batches_1]✅ 5/5209.942.55n/a
test_transpose[M_2048-N_64-aie_columns_1-channels_1-m_64-n_64-s_8-num_batches_2]✅ 5/5216.924.87n/a
test_transpose[M_2048-N_64-aie_columns_1-channels_2-m_64-n_64-s_8-num_batches_1]✅ 5/5198.242.75n/a

Trends:

IRON Trends

iron/operators/axpy

test_axpy[input_length_2048-num_aie_columns_1-tile_size_2048-scalar_factor_3.0]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.09 (+4.80%)0.08 (+11.02%)0.08 (+16.47%)0.06 (-3.52%)0.01 (+13.12%)200.20 (+3.68%)160.62 (-9.54%)158.00 (-14.13%)133.50 (-4.57%)24.94 (+16.19%)
c9bc036 — 2026-07-09 21:31:180.09 (n/a)0.07 (n/a)0.07 (n/a)0.06 (n/a)0.01 (n/a)193.10 (n/a)177.56 (n/a)184.00 (n/a)139.90 (n/a)21.47 (n/a)

test_axpy[input_length_2048-num_aie_columns_2-tile_size_1024-scalar_factor_3.0]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.07 (-19.73%)0.07 (-11.20%)0.07 (-7.33%)0.06 (-16.17%)0.01 (-27.55%)217.80 (+19.28%)185.76 (+12.34%)179.70 (+7.93%)167.50 (+24.54%)20.76 (+8.06%)
c9bc036 — 2026-07-09 21:31:180.09 (n/a)0.08 (n/a)0.07 (n/a)0.07 (n/a)0.01 (n/a)182.60 (n/a)165.36 (n/a)166.50 (n/a)134.50 (n/a)19.21 (n/a)

test_axpy[input_length_2048-num_aie_columns_4-tile_size_512-scalar_factor_3.0]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.08 (+9.41%)0.07 (-1.34%)0.07 (-1.54%)0.05 (-11.91%)0.01 (+71.98%)247.60 (+13.47%)190.18 (+2.92%)180.30 (+1.58%)157.50 (-8.59%)34.30 (+81.08%)
c9bc036 — 2026-07-09 21:31:180.07 (n/a)0.07 (n/a)0.07 (n/a)0.06 (n/a)0.01 (n/a)218.20 (n/a)184.78 (n/a)177.50 (n/a)172.30 (n/a)18.94 (n/a)

test_axpy[input_length_2048-num_aie_columns_8-tile_size_256-scalar_factor_3.0]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.08 (+8.17%)0.06 (+8.01%)0.06 (+2.51%)0.04 (-1.47%)0.02 (+33.67%)344.10 (+1.50%)222.02 (-4.86%)215.20 (-2.45%)159.70 (-7.53%)74.23 (+18.68%)
c9bc036 — 2026-07-09 21:31:180.07 (n/a)0.06 (n/a)0.06 (n/a)0.04 (n/a)0.01 (n/a)339.00 (n/a)233.36 (n/a)220.60 (n/a)172.70 (n/a)62.55 (n/a)
iron/operators/dequant

test_dequant[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-group_size_32]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.04 (+6.26%)0.03 (+8.32%)0.03 (+8.31%)0.03 (+7.91%)0.01 (+0.14%)181.50 (-7.35%)157.10 (-7.92%)163.10 (-7.64%)122.20 (-5.86%)22.10 (-13.10%)
c9bc036 — 2026-07-09 21:31:180.04 (n/a)0.03 (n/a)0.03 (n/a)0.03 (n/a)0.01 (n/a)195.90 (n/a)170.62 (n/a)176.60 (n/a)129.80 (n/a)25.43 (n/a)

test_dequant[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-group_size_32]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.04 (+2.99%)0.03 (+13.67%)0.03 (+15.54%)0.02 (-0.15%)0.01 (+13.15%)214.40 (+0.14%)158.52 (-11.42%)155.50 (-13.47%)126.00 (-2.93%)36.06 (+8.68%)
c9bc036 — 2026-07-09 21:31:180.04 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.01 (n/a)214.10 (n/a)178.96 (n/a)179.70 (n/a)129.80 (n/a)33.18 (n/a)

test_dequant[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-group_size_32]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.04 (+4.85%)0.03 (+21.80%)0.03 (+20.64%)0.03 (+88.26%)0.00 (-42.83%)180.40 (-46.89%)154.56 (-24.27%)152.20 (-17.10%)128.40 (-4.68%)22.25 (-72.17%)
c9bc036 — 2026-07-09 21:31:180.04 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.01 (n/a)339.70 (n/a)204.10 (n/a)183.60 (n/a)134.70 (n/a)79.97 (n/a)

test_dequant[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-group_size_32]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.04 (-9.52%)0.03 (+3.79%)0.03 (+13.37%)0.03 (+8.10%)0.00 (-22.87%)207.50 (-7.49%)168.00 (-4.69%)157.00 (-11.75%)147.50 (+10.49%)25.72 (-21.69%)
c9bc036 — 2026-07-09 21:31:180.04 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.01 (n/a)224.30 (n/a)176.26 (n/a)177.90 (n/a)133.50 (n/a)32.84 (n/a)

test_dequant[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-group_size_32]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.04 (+1.48%)0.03 (-12.43%)0.03 (-18.77%)0.02 (-17.27%)0.01 (+9.60%)242.90 (+20.85%)185.04 (+15.78%)171.70 (+23.08%)127.40 (-1.47%)45.32 (+29.57%)
c9bc036 — 2026-07-09 21:31:180.04 (n/a)0.03 (n/a)0.04 (n/a)0.03 (n/a)0.01 (n/a)201.00 (n/a)159.82 (n/a)139.50 (n/a)129.30 (n/a)34.97 (n/a)

test_dequant[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-group_size_32]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.04 (+24.78%)0.03 (+25.33%)0.03 (+22.24%)0.03 (+19.75%)0.01 (+30.46%)200.60 (-16.49%)163.76 (-19.98%)168.10 (-18.20%)125.90 (-19.86%)27.64 (-12.46%)
c9bc036 — 2026-07-09 21:31:180.03 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.00 (n/a)240.20 (n/a)204.66 (n/a)205.50 (n/a)157.10 (n/a)31.58 (n/a)

test_dequant[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256-group_size_32]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.03 (-6.63%)0.03 (-4.84%)0.03 (-4.20%)0.02 (+0.67%)0.00 (+0.91%)221.30 (-0.67%)191.56 (+5.24%)187.10 (+4.35%)160.30 (+7.08%)28.53 (+8.51%)
c9bc036 — 2026-07-09 21:31:180.04 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.00 (n/a)222.80 (n/a)182.02 (n/a)179.30 (n/a)149.70 (n/a)26.30 (n/a)

test_dequant[input_length_2048-num_aie_columns_8-num_channels_2-tile_size_128-group_size_32]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.03 (+14.22%)0.03 (+12.66%)0.03 (+18.61%)0.02 (+5.49%)0.00 (+56.63%)226.40 (-5.23%)192.74 (-10.59%)180.70 (-15.72%)164.40 (-12.46%)26.58 (+31.37%)
c9bc036 — 2026-07-09 21:31:180.03 (n/a)0.02 (n/a)0.02 (n/a)0.02 (n/a)0.00 (n/a)238.90 (n/a)215.56 (n/a)214.40 (n/a)187.80 (n/a)20.23 (n/a)
iron/operators/elementwise_add

test_elementwise_add[input_length_2048-num_aie_columns_1-tile_size_2048]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.09 (n/a)0.07 (n/a)0.07 (n/a)0.04 (n/a)0.02 (n/a)274.20 (n/a)195.40 (n/a)174.50 (n/a)137.40 (n/a)52.42 (n/a)

test_elementwise_add[input_length_2048-num_aie_columns_2-tile_size_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.08 (n/a)0.07 (n/a)0.06 (n/a)0.06 (n/a)0.01 (n/a)216.70 (n/a)191.10 (n/a)193.70 (n/a)155.80 (n/a)23.31 (n/a)

test_elementwise_add[input_length_2048-num_aie_columns_4-tile_size_512]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.06 (n/a)0.06 (n/a)0.06 (n/a)0.06 (n/a)0.00 (n/a)222.10 (n/a)207.18 (n/a)200.50 (n/a)197.40 (n/a)11.07 (n/a)

test_elementwise_add[input_length_2048-num_aie_columns_8-tile_size_256]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.07 (n/a)0.06 (n/a)0.06 (n/a)0.06 (n/a)0.01 (n/a)219.30 (n/a)199.42 (n/a)198.80 (n/a)171.50 (n/a)20.59 (n/a)
iron/operators/elementwise_mul

test_elementwise_mul[input_length_2048-num_aie_columns_1-tile_size_2048]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.10 (n/a)0.07 (n/a)0.06 (n/a)0.06 (n/a)0.02 (n/a)221.00 (n/a)177.74 (n/a)192.10 (n/a)121.10 (n/a)40.16 (n/a)

test_elementwise_mul[input_length_2048-num_aie_columns_2-tile_size_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.10 (n/a)0.08 (n/a)0.07 (n/a)0.07 (n/a)0.01 (n/a)175.10 (n/a)160.20 (n/a)169.60 (n/a)117.50 (n/a)24.20 (n/a)

test_elementwise_mul[input_length_2048-num_aie_columns_4-tile_size_512]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.07 (n/a)0.07 (n/a)0.06 (n/a)0.06 (n/a)0.01 (n/a)206.70 (n/a)187.08 (n/a)191.90 (n/a)165.80 (n/a)19.78 (n/a)

test_elementwise_mul[input_length_2048-num_aie_columns_8-tile_size_256]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.08 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.02 (n/a)337.90 (n/a)242.84 (n/a)244.80 (n/a)150.20 (n/a)68.81 (n/a)
iron/operators/gelu

test_gelu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)221.70 (n/a)175.18 (n/a)155.50 (n/a)142.50 (n/a)36.95 (n/a)

test_gelu[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)228.10 (n/a)169.78 (n/a)165.80 (n/a)130.40 (n/a)35.98 (n/a)

test_gelu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.06 (n/a)0.05 (n/a)0.05 (n/a)0.03 (n/a)0.01 (n/a)272.10 (n/a)187.14 (n/a)175.80 (n/a)140.20 (n/a)50.07 (n/a)

test_gelu[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)233.00 (n/a)175.76 (n/a)157.50 (n/a)138.40 (n/a)37.93 (n/a)

test_gelu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.07 (n/a)0.05 (n/a)0.04 (n/a)0.04 (n/a)0.01 (n/a)221.60 (n/a)178.94 (n/a)191.90 (n/a)123.90 (n/a)37.72 (n/a)

test_gelu[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.07 (n/a)0.05 (n/a)0.05 (n/a)0.03 (n/a)0.01 (n/a)238.30 (n/a)180.96 (n/a)172.20 (n/a)123.20 (n/a)45.97 (n/a)

test_gelu[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.05 (n/a)0.04 (n/a)0.04 (n/a)0.03 (n/a)0.01 (n/a)238.20 (n/a)189.18 (n/a)185.70 (n/a)158.10 (n/a)32.05 (n/a)

test_gelu[input_length_2048-num_aie_columns_8-num_channels_2-tile_size_128]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.04 (n/a)0.04 (n/a)0.04 (n/a)0.03 (n/a)0.01 (n/a)261.80 (n/a)215.72 (n/a)217.20 (n/a)184.70 (n/a)31.07 (n/a)
iron/operators/gemm

test_gemm[M_1792-K_896-N_1152-num_aie_columns_8-b_col_maj_False-c_col_maj_True-m_64-k_32-n_48-trace_size_0-partition_N_1]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:46:254.48 (-0.26%)3.98 (-4.33%)4.00 (-2.59%)3.54 (-8.79%)0.38 (+68.80%)2655.40 (+9.64%)2379.14 (+5.04%)2352.80 (+2.66%)2098.30 (+0.27%)224.50 (+87.04%)1763.05 (-0.26%)1566.11 (-4.33%)1572.34 (-2.59%)1393.16 (-8.79%)148.46 (+68.80%)
c9bc036 — 2026-07-09 21:31:184.49 (n/a)4.16 (n/a)4.10 (n/a)3.88 (n/a)0.22 (n/a)2421.90 (n/a)2264.96 (n/a)2291.90 (n/a)2092.70 (n/a)120.03 (n/a)1767.72 (n/a)1637.02 (n/a)1614.12 (n/a)1527.45 (n/a)87.95 (n/a)

test_gemm[M_192-K_384-N_64-num_aie_columns_4-b_col_maj_False-c_col_maj_False-m_48-k_96-n_16-trace_size_0-partition_N_1]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:46:251.03 (-8.81%)0.90 (+10.87%)0.98 (+45.90%)0.66 (+9.59%)0.16 (-38.94%)334.40 (-8.76%)252.24 (-13.92%)226.00 (-31.47%)213.80 (+9.70%)50.72 (-39.55%)44.14 (-8.81%)38.49 (+10.87%)41.75 (+45.90%)28.22 (+9.59%)6.72 (-38.94%)
c9bc036 — 2026-07-09 21:31:181.13 (n/a)0.81 (n/a)0.67 (n/a)0.60 (n/a)0.26 (n/a)366.50 (n/a)293.04 (n/a)329.80 (n/a)194.90 (n/a)83.90 (n/a)48.41 (n/a)34.71 (n/a)28.62 (n/a)25.75 (n/a)11.00 (n/a)

test_gemm[M_192-K_384-N_64-num_aie_columns_4-b_col_maj_True-c_col_maj_True-m_48-k_96-n_16-trace_size_0-partition_N_1]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:46:251.04 (-8.08%)0.77 (-13.62%)0.65 (-28.96%)0.59 (+2.44%)0.21 (-5.06%)377.60 (-2.38%)302.12 (+15.40%)340.90 (+40.75%)213.30 (+8.77%)74.28 (-2.18%)44.23 (-8.08%)32.97 (-13.62%)27.68 (-28.96%)24.99 (+2.44%)8.83 (-5.06%)
c9bc036 — 2026-07-09 21:31:181.13 (n/a)0.89 (n/a)0.91 (n/a)0.57 (n/a)0.22 (n/a)386.80 (n/a)261.80 (n/a)242.20 (n/a)196.10 (n/a)75.94 (n/a)48.12 (n/a)38.16 (n/a)38.97 (n/a)24.40 (n/a)9.30 (n/a)

test_gemm[M_2048-K_2048-N_2048-num_aie_columns_1-b_col_maj_False-c_col_maj_False-m_64-k_64-n_64-trace_size_0-partition_N_1]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:46:250.52 (+0.87%)0.52 (+0.47%)0.52 (+0.29%)0.52 (+0.47%)0.00 (+75.62%)48664.50 (-0.47%)48560.14 (-0.47%)48615.70 (-0.29%)48254.70 (-0.86%)172.16 (+73.10%)356.02 (+0.87%)353.79 (+0.47%)353.38 (+0.29%)353.03 (+0.47%)1.26 (+75.63%)
c9bc036 — 2026-07-09 21:31:180.52 (n/a)0.52 (n/a)0.52 (n/a)0.51 (n/a)0.00 (n/a)48892.60 (n/a)48787.46 (n/a)48757.60 (n/a)48673.40 (n/a)99.46 (n/a)352.96 (n/a)352.14 (n/a)352.35 (n/a)351.38 (n/a)0.72 (n/a)

test_gemm[M_2048-K_2048-N_2048-num_aie_columns_2-b_col_maj_True-c_col_maj_False-m_64-k_64-n_64-trace_size_0-partition_N_1]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:46:250.88 (+0.36%)0.88 (+0.08%)0.88 (+0.11%)0.87 (-0.05%)0.00 (+68.45%)28834.80 (+0.05%)28677.34 (-0.08%)28634.10 (-0.11%)28527.30 (-0.36%)126.37 (+67.96%)602.22 (+0.36%)599.08 (+0.08%)599.98 (+0.11%)595.80 (-0.05%)2.64 (+68.45%)
c9bc036 — 2026-07-09 21:31:180.88 (n/a)0.88 (n/a)0.88 (n/a)0.87 (n/a)0.00 (n/a)28819.90 (n/a)28699.84 (n/a)28666.20 (n/a)28631.40 (n/a)75.24 (n/a)600.04 (n/a)598.61 (n/a)599.31 (n/a)596.11 (n/a)1.57 (n/a)

test_gemm[M_2048-K_2048-N_2048-num_aie_columns_8-b_col_maj_True-c_col_maj_True-m_64-k_64-n_64-trace_size_0-partition_N_1]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:46:253.34 (+1.61%)3.20 (-0.53%)3.16 (-1.03%)3.15 (-1.11%)0.08 (+85.48%)8000.10 (+1.12%)7880.12 (+0.57%)7973.10 (+1.04%)7526.40 (-1.58%)200.75 (+84.47%)2282.60 (+1.61%)2181.32 (-0.53%)2154.72 (-1.03%)2147.46 (-1.11%)57.40 (+85.48%)
c9bc036 — 2026-07-09 21:31:183.29 (n/a)3.21 (n/a)3.19 (n/a)3.18 (n/a)0.05 (n/a)7911.60 (n/a)7835.22 (n/a)7890.80 (n/a)7647.30 (n/a)108.82 (n/a)2246.52 (n/a)2192.99 (n/a)2177.21 (n/a)2171.47 (n/a)30.95 (n/a)

test_gemm[M_384-K_1536-N_1792-num_aie_columns_4-b_col_maj_True-c_col_maj_False-m_32-k_48-n_64-trace_size_0-partition_N_1]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:46:253.79 (-8.19%)3.36 (-3.30%)3.48 (-0.00%)2.90 (-1.90%)0.40 (-18.43%)2783.70 (+1.93%)2427.32 (+2.99%)2318.50 (+0.00%)2129.10 (+8.92%)295.45 (-9.66%)992.86 (-8.19%)881.04 (-3.30%)911.75 (-0.00%)759.40 (-1.90%)104.40 (-18.43%)
c9bc036 — 2026-07-09 21:31:184.12 (n/a)3.47 (n/a)3.48 (n/a)2.95 (n/a)0.49 (n/a)2730.90 (n/a)2356.86 (n/a)2318.50 (n/a)1954.70 (n/a)327.04 (n/a)1081.47 (n/a)911.08 (n/a)911.77 (n/a)774.07 (n/a)127.99 (n/a)

test_gemm[M_64-K_512-N_256-num_aie_columns_4-b_col_maj_True-c_col_maj_False-m_16-k_64-n_64-trace_size_0-partition_N_4]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:46:250.51 (+49.68%)0.34 (+10.82%)0.30 (-4.85%)0.26 (-0.60%)0.10 (+244.75%)4724.70 (+0.60%)3841.52 (-5.54%)4170.60 (+5.10%)2428.60 (-33.19%)869.36 (+118.21%)27.63 (+49.68%)18.42 (+10.82%)16.09 (-4.85%)14.20 (-0.60%)5.33 (+244.75%)
c9bc036 — 2026-07-09 21:31:180.34 (n/a)0.31 (n/a)0.31 (n/a)0.27 (n/a)0.03 (n/a)4696.40 (n/a)4066.64 (n/a)3968.20 (n/a)3635.10 (n/a)398.40 (n/a)18.46 (n/a)16.62 (n/a)16.91 (n/a)14.29 (n/a)1.55 (n/a)

test_gemm[M_896-K_1792-N_640-num_aie_columns_8-b_col_maj_False-c_col_maj_True-m_32-k_64-n_80-trace_size_0-partition_N_1]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:46:256.40 (+33.46%)4.76 (+12.79%)4.70 (+2.75%)3.57 (+3.50%)1.03 (+61.49%)1863.00 (-3.38%)1447.74 (-9.95%)1414.00 (-2.68%)1039.80 (-25.08%)294.54 (+14.57%)1976.47 (+33.46%)1470.22 (+12.79%)1453.52 (+2.75%)1103.19 (+3.50%)318.19 (+61.49%)
c9bc036 — 2026-07-09 21:31:184.79 (n/a)4.22 (n/a)4.58 (n/a)3.45 (n/a)0.64 (n/a)1928.20 (n/a)1607.70 (n/a)1452.90 (n/a)1387.80 (n/a)257.09 (n/a)1480.95 (n/a)1303.54 (n/a)1414.57 (n/a)1065.85 (n/a)197.03 (n/a)
iron/operators/gemv

test_gemv[M_128-K_128-num_aie_columns_1-tile_size_input_32-tile_size_output_128]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:46:250.26 (+1.01%)0.22 (+2.94%)0.22 (-3.45%)0.18 (+27.63%)0.04 (-11.96%)0.26 (+1.01%)0.22 (+2.94%)0.22 (-3.45%)0.18 (+27.63%)0.04 (-11.96%)
c9bc036 — 2026-07-09 21:31:180.26 (n/a)0.22 (n/a)0.23 (n/a)0.14 (n/a)0.04 (n/a)0.26 (n/a)0.21 (n/a)0.22 (n/a)0.14 (n/a)0.04 (n/a)

test_gemv[M_2048-K_8192-num_aie_columns_1-tile_size_input_1-tile_size_output_2048]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:46:2513.42 (-0.25%)12.83 (+2.05%)13.06 (+0.19%)11.52 (+5.62%)0.76 (-26.14%)13.41 (-0.25%)12.82 (+2.05%)13.05 (+0.19%)11.51 (+5.62%)0.76 (-26.14%)
c9bc036 — 2026-07-09 21:31:1813.45 (n/a)12.57 (n/a)13.03 (n/a)10.90 (n/a)1.03 (n/a)13.44 (n/a)12.56 (n/a)13.03 (n/a)10.90 (n/a)1.03 (n/a)

test_gemv[M_2048-K_8192-num_aie_columns_2-tile_size_input_1-tile_size_output_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:46:2525.07 (+1.11%)24.64 (+4.87%)24.72 (+2.49%)24.11 (+18.00%)0.43 (-75.27%)25.06 (+1.11%)24.62 (+4.88%)24.70 (+2.49%)24.09 (+18.00%)0.43 (-75.27%)
c9bc036 — 2026-07-09 21:31:1824.80 (n/a)23.49 (n/a)24.12 (n/a)20.43 (n/a)1.74 (n/a)24.78 (n/a)23.48 (n/a)24.10 (n/a)20.42 (n/a)1.74 (n/a)

test_gemv[M_2048-K_8192-num_aie_columns_4-tile_size_input_1-tile_size_output_512]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:46:2544.16 (+8.06%)38.36 (-2.60%)40.01 (+1.58%)33.36 (-11.90%)4.72 (+342.28%)44.14 (+8.06%)38.34 (-2.60%)39.98 (+1.58%)33.34 (-11.90%)4.72 (+342.28%)
c9bc036 — 2026-07-09 21:31:1840.87 (n/a)39.39 (n/a)39.38 (n/a)37.87 (n/a)1.07 (n/a)40.84 (n/a)39.36 (n/a)39.36 (n/a)37.84 (n/a)1.07 (n/a)

test_gemv[M_2048-K_8192-num_aie_columns_8-tile_size_input_1-tile_size_output_256]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:46:2545.35 (+0.94%)39.84 (-7.38%)42.26 (-0.60%)25.34 (-39.57%)8.23 (+610.94%)45.32 (+0.94%)39.82 (-7.38%)42.24 (-0.60%)25.32 (-39.57%)8.23 (+610.94%)
c9bc036 — 2026-07-09 21:31:1844.93 (n/a)43.01 (n/a)42.52 (n/a)41.93 (n/a)1.16 (n/a)44.90 (n/a)42.99 (n/a)42.49 (n/a)41.91 (n/a)1.16 (n/a)

test_gemv[M_8192-K_2048-num_aie_columns_1-tile_size_input_4-tile_size_output_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:46:2513.53 (+1.54%)12.93 (+2.00%)13.18 (-0.25%)12.18 (+8.77%)0.55 (-39.05%)13.52 (+1.54%)12.92 (+2.00%)13.17 (-0.25%)12.17 (+8.77%)0.55 (-39.05%)
c9bc036 — 2026-07-09 21:31:1813.33 (n/a)12.67 (n/a)13.21 (n/a)11.20 (n/a)0.90 (n/a)13.32 (n/a)12.67 (n/a)13.21 (n/a)11.19 (n/a)0.90 (n/a)

test_gemv[M_8192-K_2048-num_aie_columns_2-tile_size_input_4-tile_size_output_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:46:2525.10 (+0.59%)24.53 (+0.84%)24.45 (+1.17%)24.09 (+0.38%)0.40 (+6.82%)25.08 (+0.59%)24.51 (+0.84%)24.44 (+1.17%)24.07 (+0.38%)0.40 (+6.82%)
c9bc036 — 2026-07-09 21:31:1824.95 (n/a)24.33 (n/a)24.17 (n/a)23.99 (n/a)0.37 (n/a)24.93 (n/a)24.31 (n/a)24.15 (n/a)23.98 (n/a)0.37 (n/a)

test_gemv[M_8192-K_2048-num_aie_columns_4-tile_size_input_4-tile_size_output_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:46:2542.79 (+4.32%)41.30 (+4.02%)41.13 (+1.31%)39.54 (+5.98%)1.21 (-25.53%)42.76 (+4.32%)41.27 (+4.02%)41.11 (+1.31%)39.51 (+5.98%)1.21 (-25.53%)
c9bc036 — 2026-07-09 21:31:1841.01 (n/a)39.70 (n/a)40.60 (n/a)37.30 (n/a)1.63 (n/a)40.99 (n/a)39.68 (n/a)40.58 (n/a)37.28 (n/a)1.63 (n/a)

test_gemv[M_8192-K_2048-num_aie_columns_8-tile_size_input_4-tile_size_output_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:46:2545.18 (+1.12%)43.25 (+1.39%)42.60 (+0.71%)42.15 (+2.49%)1.24 (-14.84%)45.15 (+1.12%)43.23 (+1.39%)42.58 (+0.71%)42.13 (+2.49%)1.24 (-14.84%)
c9bc036 — 2026-07-09 21:31:1844.68 (n/a)42.66 (n/a)42.30 (n/a)41.13 (n/a)1.45 (n/a)44.65 (n/a)42.63 (n/a)42.28 (n/a)41.10 (n/a)1.45 (n/a)
iron/operators/layer_norm

test_layer_norm[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.08 (n/a)0.06 (n/a)0.06 (n/a)0.04 (n/a)0.01 (n/a)201.50 (n/a)143.94 (n/a)131.40 (n/a)102.60 (n/a)37.09 (n/a)

test_layer_norm[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.08 (n/a)0.05 (n/a)0.06 (n/a)0.02 (n/a)0.02 (n/a)341.70 (n/a)192.28 (n/a)146.90 (n/a)104.10 (n/a)97.25 (n/a)

test_layer_norm[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.06 (n/a)0.05 (n/a)0.05 (n/a)0.03 (n/a)0.01 (n/a)237.90 (n/a)180.82 (n/a)177.30 (n/a)139.50 (n/a)38.02 (n/a)

test_layer_norm[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)191.60 (n/a)170.56 (n/a)177.40 (n/a)143.80 (n/a)21.00 (n/a)

test_layer_norm[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.05 (n/a)0.04 (n/a)0.04 (n/a)0.04 (n/a)0.00 (n/a)211.80 (n/a)190.62 (n/a)184.80 (n/a)175.70 (n/a)13.92 (n/a)

test_layer_norm[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.05 (n/a)0.04 (n/a)0.04 (n/a)0.04 (n/a)0.00 (n/a)202.20 (n/a)189.68 (n/a)189.80 (n/a)178.00 (n/a)10.43 (n/a)

test_layer_norm[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.07 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)232.40 (n/a)184.60 (n/a)175.00 (n/a)114.40 (n/a)48.47 (n/a)

test_layer_norm[input_length_2048-num_aie_columns_8-num_channels_2-tile_size_128]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.04 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.01 (n/a)378.10 (n/a)257.66 (n/a)240.20 (n/a)205.60 (n/a)69.13 (n/a)
iron/operators/leaky_relu

test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-alpha_0.01]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.07 (n/a)0.06 (n/a)0.06 (n/a)0.05 (n/a)0.01 (n/a)165.60 (n/a)146.62 (n/a)148.30 (n/a)118.90 (n/a)20.10 (n/a)

test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-alpha_0.1]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.08 (n/a)0.06 (n/a)0.06 (n/a)0.04 (n/a)0.02 (n/a)217.90 (n/a)157.28 (n/a)146.50 (n/a)108.60 (n/a)49.52 (n/a)

test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-alpha_0.25]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)205.20 (n/a)166.16 (n/a)172.80 (n/a)133.70 (n/a)28.31 (n/a)

test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-alpha_0.01]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.07 (n/a)0.06 (n/a)0.06 (n/a)0.05 (n/a)0.01 (n/a)159.80 (n/a)141.80 (n/a)140.70 (n/a)123.50 (n/a)13.02 (n/a)

test_leaky_relu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-alpha_0.01]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.06 (n/a)0.06 (n/a)0.06 (n/a)0.04 (n/a)0.01 (n/a)197.30 (n/a)151.68 (n/a)139.60 (n/a)128.00 (n/a)27.90 (n/a)

test_leaky_relu[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-alpha_0.01]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.07 (n/a)0.05 (n/a)0.04 (n/a)0.04 (n/a)0.01 (n/a)210.80 (n/a)173.72 (n/a)188.40 (n/a)121.20 (n/a)35.17 (n/a)

test_leaky_relu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-alpha_0.01]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.07 (n/a)0.05 (n/a)0.06 (n/a)0.04 (n/a)0.01 (n/a)218.50 (n/a)162.02 (n/a)148.50 (n/a)123.90 (n/a)38.73 (n/a)

test_leaky_relu[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-alpha_0.01]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.05 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)220.90 (n/a)174.38 (n/a)170.40 (n/a)151.10 (n/a)27.81 (n/a)

test_leaky_relu[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256-alpha_0.01]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.07 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)230.70 (n/a)164.88 (n/a)158.20 (n/a)115.20 (n/a)41.61 (n/a)

test_leaky_relu[input_length_2048-num_aie_columns_8-num_channels_2-tile_size_128-alpha_0.01]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.04 (n/a)0.04 (n/a)0.04 (n/a)0.03 (n/a)0.01 (n/a)287.30 (n/a)225.20 (n/a)211.50 (n/a)183.00 (n/a)40.01 (n/a)
iron/operators/mem_copy

test_mem_copy[input_length_2048-num_cores_1-num_channels_1-bypass_False-tile_size_2048]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.07 (-2.01%)0.05 (-4.81%)0.05 (-5.35%)0.04 (-8.39%)0.01 (+2.41%)192.70 (+9.18%)157.10 (+5.41%)161.50 (+5.69%)121.80 (+2.10%)26.66 (+13.40%)
c9bc036 — 2026-07-09 21:31:180.07 (n/a)0.06 (n/a)0.05 (n/a)0.05 (n/a)0.01 (n/a)176.50 (n/a)149.04 (n/a)152.80 (n/a)119.30 (n/a)23.51 (n/a)

test_mem_copy[input_length_2048-num_cores_16-num_channels_2-bypass_False-tile_size_128]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.04 (-24.66%)0.04 (-8.18%)0.04 (-0.46%)0.03 (-9.18%)0.00 (-58.57%)242.80 (+10.11%)212.18 (+6.42%)213.20 (+0.47%)183.90 (+32.78%)21.25 (-38.44%)
c9bc036 — 2026-07-09 21:31:180.06 (n/a)0.04 (n/a)0.04 (n/a)0.04 (n/a)0.01 (n/a)220.50 (n/a)199.38 (n/a)212.20 (n/a)138.50 (n/a)34.53 (n/a)

test_mem_copy[input_length_2048-num_cores_2-num_channels_1-bypass_False-tile_size_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.07 (+13.88%)0.05 (-4.22%)0.05 (-16.24%)0.04 (+17.52%)0.01 (+15.53%)198.80 (-14.90%)172.12 (+4.19%)179.90 (+19.38%)119.90 (-12.16%)31.84 (-18.55%)
c9bc036 — 2026-07-09 21:31:180.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)233.60 (n/a)165.20 (n/a)150.70 (n/a)136.50 (n/a)39.09 (n/a)

test_mem_copy[input_length_2048-num_cores_2-num_channels_2-bypass_False-tile_size_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.06 (+14.61%)0.05 (+30.02%)0.05 (+32.30%)0.04 (+72.75%)0.01 (-32.79%)202.00 (-42.12%)159.38 (-27.22%)151.80 (-24.40%)140.10 (-12.76%)24.71 (-67.04%)
c9bc036 — 2026-07-09 21:31:180.05 (n/a)0.04 (n/a)0.04 (n/a)0.02 (n/a)0.01 (n/a)349.00 (n/a)219.00 (n/a)200.80 (n/a)160.60 (n/a)74.97 (n/a)

test_mem_copy[input_length_2048-num_cores_4-num_channels_1-bypass_False-tile_size_512]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.05 (-7.49%)0.04 (-10.50%)0.05 (-3.22%)0.04 (-19.62%)0.01 (+21.49%)233.70 (+24.44%)189.30 (+12.73%)174.90 (+3.31%)161.10 (+8.12%)29.09 (+66.36%)
c9bc036 — 2026-07-09 21:31:180.05 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)187.80 (n/a)167.92 (n/a)169.30 (n/a)149.00 (n/a)17.49 (n/a)

test_mem_copy[input_length_2048-num_cores_4-num_channels_2-bypass_False-tile_size_512]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.05 (-26.68%)0.04 (-23.45%)0.04 (-24.21%)0.04 (-22.24%)0.00 (-47.56%)220.80 (+28.60%)195.90 (+29.72%)190.50 (+32.02%)174.30 (+36.38%)17.41 (-10.18%)
c9bc036 — 2026-07-09 21:31:180.06 (n/a)0.05 (n/a)0.06 (n/a)0.05 (n/a)0.01 (n/a)171.70 (n/a)151.02 (n/a)144.30 (n/a)127.80 (n/a)19.38 (n/a)

test_mem_copy[input_length_2048-num_cores_8-num_channels_1-bypass_False-tile_size_256]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.04 (-22.85%)0.04 (-18.13%)0.04 (-22.04%)0.04 (-8.56%)0.00 (-46.72%)227.40 (+9.38%)213.58 (+21.41%)221.00 (+28.26%)197.00 (+29.61%)14.99 (-26.01%)
c9bc036 — 2026-07-09 21:31:180.05 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)207.90 (n/a)175.92 (n/a)172.30 (n/a)152.00 (n/a)20.26 (n/a)

test_mem_copy[input_length_2048-num_cores_8-num_channels_2-bypass_False-tile_size_256]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.05 (-1.00%)0.05 (-3.87%)0.05 (+2.95%)0.04 (-14.52%)0.01 (+42.94%)227.80 (+17.00%)178.20 (+5.63%)155.90 (-2.87%)152.60 (+1.06%)33.80 (+66.67%)
c9bc036 — 2026-07-09 21:31:180.05 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)194.70 (n/a)168.70 (n/a)160.50 (n/a)151.00 (n/a)20.28 (n/a)
iron/operators/mha

test_mha[seq_len_16384-dim_64-num_heads_1-num_pipelines_8-num_kv_heads_0]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
573678d — 2026-06-29 18:46:460.21 (+0.02%)0.20 (+0.01%)0.20 (-0.07%)0.20 (+0.07%)0.00 (-18.47%)40961.20 (-0.07%)40924.92 (-0.01%)40941.50 (+0.07%)40859.60 (-0.02%)41.66 (-18.49%)
9c70ba8 — 2026-06-29 16:51:590.21 (n/a)0.20 (n/a)0.21 (n/a)0.20 (n/a)0.00 (n/a)40988.20 (n/a)40927.74 (n/a)40913.30 (n/a)40866.90 (n/a)51.11 (n/a)
iron/operators/rms_norm

test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-weighted_False]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.07 (+10.01%)0.05 (-6.22%)0.05 (+4.62%)0.02 (-46.79%)0.02 (+107.03%)385.40 (+87.91%)196.30 (+23.23%)149.70 (-4.41%)118.00 (-9.09%)108.52 (+276.00%)
c9bc036 — 2026-07-09 21:31:180.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)205.10 (n/a)159.30 (n/a)156.60 (n/a)129.80 (n/a)28.86 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-weighted_True]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.09 (+9.41%)0.08 (+7.50%)0.08 (+10.95%)0.06 (+5.37%)0.01 (+59.62%)196.40 (-5.12%)164.60 (-5.80%)156.90 (-9.88%)138.40 (-8.59%)28.19 (+36.78%)
c9bc036 — 2026-07-09 21:31:180.08 (n/a)0.07 (n/a)0.07 (n/a)0.06 (n/a)0.01 (n/a)207.00 (n/a)174.74 (n/a)174.10 (n/a)151.40 (n/a)20.61 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-weighted_False]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.06 (+11.25%)0.05 (+6.57%)0.05 (+16.75%)0.04 (-1.60%)0.01 (+50.42%)196.20 (+1.66%)162.64 (-4.90%)152.20 (-14.30%)128.70 (-10.06%)28.54 (+41.43%)
c9bc036 — 2026-07-09 21:31:180.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)193.00 (n/a)171.02 (n/a)177.60 (n/a)143.10 (n/a)20.18 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-weighted_True]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.08 (+19.23%)0.07 (+12.73%)0.07 (+7.09%)0.06 (+30.86%)0.01 (-8.03%)167.40 (-23.60%)147.40 (-12.21%)149.50 (-6.62%)121.10 (-16.14%)16.63 (-44.05%)
c9bc036 — 2026-07-09 21:31:180.07 (n/a)0.06 (n/a)0.06 (n/a)0.05 (n/a)0.01 (n/a)219.10 (n/a)167.90 (n/a)160.10 (n/a)144.40 (n/a)29.72 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-weighted_False]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.05 (-10.48%)0.04 (-20.10%)0.04 (-16.65%)0.03 (-26.35%)0.01 (+14.60%)244.80 (+35.77%)196.30 (+26.78%)190.50 (+19.96%)150.50 (+11.65%)35.00 (+78.11%)
c9bc036 — 2026-07-09 21:31:180.06 (n/a)0.05 (n/a)0.05 (n/a)0.05 (n/a)0.01 (n/a)180.30 (n/a)154.84 (n/a)158.80 (n/a)134.80 (n/a)19.65 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-weighted_True]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.08 (+22.91%)0.07 (+24.31%)0.07 (+25.27%)0.05 (+26.26%)0.01 (+19.27%)215.30 (-20.82%)163.82 (-19.85%)152.30 (-20.18%)122.30 (-18.63%)38.17 (-23.03%)
c9bc036 — 2026-07-09 21:31:180.07 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)271.90 (n/a)204.38 (n/a)190.80 (n/a)150.30 (n/a)49.60 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-weighted_False]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.06 (+38.17%)0.05 (+15.53%)0.05 (+9.08%)0.04 (+6.70%)0.01 (+169.95%)206.90 (-6.30%)173.10 (-11.15%)171.90 (-8.32%)129.10 (-27.63%)34.08 (+89.17%)
c9bc036 — 2026-07-09 21:31:180.05 (n/a)0.04 (n/a)0.04 (n/a)0.04 (n/a)0.00 (n/a)220.80 (n/a)194.82 (n/a)187.50 (n/a)178.40 (n/a)18.02 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-weighted_True]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.07 (+16.73%)0.06 (+10.21%)0.06 (+12.35%)0.04 (+6.20%)0.01 (+35.81%)209.40 (-5.85%)161.28 (-8.30%)151.50 (-10.99%)125.50 (-14.33%)33.17 (+10.25%)
c9bc036 — 2026-07-09 21:31:180.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)222.40 (n/a)175.88 (n/a)170.20 (n/a)146.50 (n/a)30.09 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-weighted_False]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.06 (+31.97%)0.05 (+13.60%)0.05 (+15.36%)0.04 (-10.33%)0.01 (+345.97%)214.40 (+11.55%)164.80 (-9.51%)159.50 (-13.32%)128.60 (-24.22%)32.40 (+283.06%)
c9bc036 — 2026-07-09 21:31:180.05 (n/a)0.05 (n/a)0.04 (n/a)0.04 (n/a)0.00 (n/a)192.20 (n/a)182.12 (n/a)184.00 (n/a)169.70 (n/a)8.46 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-weighted_True]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.07 (-5.22%)0.05 (-3.21%)0.06 (+3.62%)0.04 (-2.40%)0.01 (+10.14%)216.90 (+2.46%)176.12 (+4.25%)164.00 (-3.47%)138.00 (+5.50%)36.35 (+23.14%)
c9bc036 — 2026-07-09 21:31:180.07 (n/a)0.06 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)211.70 (n/a)168.94 (n/a)169.90 (n/a)130.80 (n/a)29.52 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-weighted_False]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.05 (-6.58%)0.05 (+1.96%)0.04 (-1.88%)0.04 (+13.94%)0.01 (-31.13%)203.50 (-12.25%)181.82 (-3.75%)193.20 (+1.90%)153.20 (+7.06%)24.43 (-35.99%)
c9bc036 — 2026-07-09 21:31:180.06 (n/a)0.04 (n/a)0.04 (n/a)0.04 (n/a)0.01 (n/a)231.90 (n/a)188.90 (n/a)189.60 (n/a)143.10 (n/a)38.17 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-weighted_True]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.07 (+35.07%)0.05 (+22.33%)0.04 (+17.40%)0.04 (+27.90%)0.01 (+39.26%)208.80 (-21.83%)179.82 (-17.89%)194.30 (-14.82%)127.40 (-25.93%)34.41 (-17.40%)
c9bc036 — 2026-07-09 21:31:180.05 (n/a)0.04 (n/a)0.04 (n/a)0.03 (n/a)0.01 (n/a)267.10 (n/a)219.00 (n/a)228.10 (n/a)172.00 (n/a)41.66 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256-weighted_False]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.06 (+23.70%)0.05 (+21.55%)0.05 (+27.48%)0.04 (+20.40%)0.01 (+14.12%)189.60 (-16.95%)155.32 (-17.87%)151.90 (-21.54%)128.50 (-19.13%)22.26 (-21.30%)
c9bc036 — 2026-07-09 21:31:180.05 (n/a)0.04 (n/a)0.04 (n/a)0.04 (n/a)0.01 (n/a)228.30 (n/a)189.12 (n/a)193.60 (n/a)158.90 (n/a)28.28 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_8-num_channels_1-tile_size_256-weighted_True]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.06 (+10.39%)0.05 (+7.32%)0.04 (+9.92%)0.04 (+6.76%)0.01 (+8.51%)212.10 (-6.36%)190.18 (-6.84%)196.00 (-9.01%)157.50 (-9.38%)21.31 (-8.98%)
c9bc036 — 2026-07-09 21:31:180.05 (n/a)0.04 (n/a)0.04 (n/a)0.04 (n/a)0.01 (n/a)226.50 (n/a)204.14 (n/a)215.40 (n/a)173.80 (n/a)23.42 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_8-num_channels_2-tile_size_128-weighted_False]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.05 (-13.06%)0.04 (-6.72%)0.04 (-4.88%)0.03 (-12.93%)0.01 (-11.58%)302.90 (+14.87%)228.42 (+7.41%)217.30 (+5.18%)176.90 (+15.02%)49.34 (+19.22%)
c9bc036 — 2026-07-09 21:31:180.05 (n/a)0.04 (n/a)0.04 (n/a)0.03 (n/a)0.01 (n/a)263.70 (n/a)212.66 (n/a)206.60 (n/a)153.80 (n/a)41.38 (n/a)
iron/operators/rope

test_rope[rows_32-cols_512-angle_rows_32-aie_columns_1-method_type_0]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.75 (+6.39%)0.53 (-5.10%)0.51 (-4.11%)0.32 (-23.91%)0.17 (+68.54%)303.90 (+31.44%)203.56 (+12.27%)192.20 (+4.29%)131.30 (-6.01%)69.51 (+107.04%)
c9bc036 — 2026-07-09 21:31:180.70 (n/a)0.56 (n/a)0.53 (n/a)0.43 (n/a)0.10 (n/a)231.20 (n/a)181.32 (n/a)184.30 (n/a)139.70 (n/a)33.57 (n/a)

test_rope[rows_32-cols_512-angle_rows_32-aie_columns_2-method_type_0]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.71 (-0.95%)0.56 (-11.68%)0.56 (-9.97%)0.46 (-23.60%)0.10 (+121.91%)212.70 (+30.89%)178.62 (+15.64%)176.10 (+11.03%)139.10 (+0.94%)31.05 (+197.18%)
c9bc036 — 2026-07-09 21:31:180.71 (n/a)0.64 (n/a)0.62 (n/a)0.60 (n/a)0.05 (n/a)162.50 (n/a)154.46 (n/a)158.60 (n/a)137.80 (n/a)10.45 (n/a)

test_rope[rows_32-cols_512-angle_rows_32-aie_columns_4-method_type_0]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.68 (+6.77%)0.52 (-13.72%)0.53 (-13.68%)0.32 (-41.69%)0.13 (+305.85%)304.30 (+71.44%)199.80 (+23.18%)186.60 (+15.90%)143.60 (-6.33%)61.93 (+573.74%)
c9bc036 — 2026-07-09 21:31:180.64 (n/a)0.61 (n/a)0.61 (n/a)0.55 (n/a)0.03 (n/a)177.50 (n/a)162.20 (n/a)161.00 (n/a)153.30 (n/a)9.19 (n/a)

test_rope[rows_32-cols_512-angle_rows_32-aie_columns_8-method_type_0]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.62 (-8.03%)0.49 (-15.87%)0.47 (-15.62%)0.43 (-21.13%)0.08 (+39.08%)230.90 (+26.80%)205.18 (+20.11%)208.90 (+18.56%)159.10 (+8.75%)27.46 (+87.64%)
c9bc036 — 2026-07-09 21:31:180.67 (n/a)0.58 (n/a)0.56 (n/a)0.54 (n/a)0.05 (n/a)182.10 (n/a)170.82 (n/a)176.20 (n/a)146.30 (n/a)14.64 (n/a)

test_rope[rows_32-cols_512-angle_rows_8-aie_columns_1-method_type_0]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.51 (-3.51%)0.40 (-8.74%)0.39 (-6.43%)0.30 (-27.05%)0.08 (+58.70%)248.50 (+37.07%)189.50 (+12.21%)190.30 (+6.85%)145.20 (+3.64%)39.49 (+125.27%)
c9bc036 — 2026-07-09 21:31:180.53 (n/a)0.44 (n/a)0.41 (n/a)0.41 (n/a)0.05 (n/a)181.30 (n/a)168.88 (n/a)178.10 (n/a)140.10 (n/a)17.53 (n/a)

test_rope[rows_32-cols_512-angle_rows_8-aie_columns_2-method_type_0]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.56 (+9.51%)0.40 (-8.02%)0.37 (-15.55%)0.30 (-16.66%)0.11 (+96.45%)242.60 (+19.98%)193.44 (+13.31%)199.30 (+18.42%)132.10 (-8.71%)48.07 (+118.64%)
c9bc036 — 2026-07-09 21:31:180.51 (n/a)0.44 (n/a)0.44 (n/a)0.36 (n/a)0.06 (n/a)202.20 (n/a)170.72 (n/a)168.30 (n/a)144.70 (n/a)21.99 (n/a)

test_rope[rows_32-cols_512-angle_rows_8-aie_columns_4-method_type_0]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.53 (-9.98%)0.42 (+10.62%)0.40 (+8.07%)0.37 (+83.80%)0.07 (-53.82%)200.20 (-45.58%)178.64 (-18.62%)185.50 (-7.48%)138.10 (+11.10%)23.88 (-73.65%)
c9bc036 — 2026-07-09 21:31:180.59 (n/a)0.38 (n/a)0.37 (n/a)0.20 (n/a)0.14 (n/a)367.90 (n/a)219.50 (n/a)200.50 (n/a)124.30 (n/a)90.62 (n/a)

test_rope[rows_32-cols_512-angle_rows_8-aie_columns_8-method_type_0]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.49 (+16.20%)0.42 (+6.10%)0.40 (-0.86%)0.37 (+5.22%)0.06 (+93.00%)200.00 (-4.99%)177.78 (-4.89%)182.30 (+0.89%)149.80 (-13.96%)22.68 (+57.59%)
c9bc036 — 2026-07-09 21:31:180.42 (n/a)0.40 (n/a)0.41 (n/a)0.35 (n/a)0.03 (n/a)210.50 (n/a)186.92 (n/a)180.70 (n/a)174.10 (n/a)14.39 (n/a)
iron/operators/softmax

test_softmax[input_length_32768-num_aie_columns_2-num_channels_2-tile_size_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:251.13 (+12.33%)0.85 (+7.64%)0.86 (+14.00%)0.60 (-14.81%)0.23 (+79.76%)218.00 (+17.39%)163.92 (-3.17%)151.90 (-12.30%)115.90 (-10.98%)44.44 (+94.67%)
c9bc036 — 2026-07-09 21:31:181.01 (n/a)0.79 (n/a)0.76 (n/a)0.71 (n/a)0.13 (n/a)185.70 (n/a)169.28 (n/a)173.20 (n/a)130.20 (n/a)22.83 (n/a)

test_softmax[input_length_32768-num_aie_columns_2-num_channels_2-tile_size_2048]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.88 (-8.35%)0.83 (+19.05%)0.86 (+31.45%)0.72 (+27.15%)0.07 (-56.59%)182.20 (-21.36%)159.72 (-18.32%)151.60 (-23.93%)149.60 (+9.12%)13.92 (-61.90%)
c9bc036 — 2026-07-09 21:31:180.96 (n/a)0.69 (n/a)0.66 (n/a)0.57 (n/a)0.15 (n/a)231.70 (n/a)195.54 (n/a)199.30 (n/a)137.10 (n/a)36.54 (n/a)

test_softmax[input_length_32768-num_aie_columns_2-num_channels_2-tile_size_512]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:251.02 (+23.72%)0.81 (+22.14%)0.86 (+18.46%)0.62 (+75.84%)0.16 (-16.20%)211.70 (-43.14%)166.96 (-23.29%)152.50 (-15.61%)128.30 (-19.16%)34.33 (-61.52%)
c9bc036 — 2026-07-09 21:31:180.83 (n/a)0.66 (n/a)0.73 (n/a)0.35 (n/a)0.19 (n/a)372.30 (n/a)217.66 (n/a)180.70 (n/a)158.70 (n/a)89.21 (n/a)
iron/operators/swiglu_decode

test_swiglu_decode[embedding_dim_1024-hidden_dim_3584]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.00 (+2.33%)0.00 (+3.37%)0.00 (+2.38%)0.00 (+2.50%)0.00 (+7.42%)989.94 (-4.26%)950.34 (-3.82%)949.32 (-2.10%)926.69 (-3.48%)24.27 (-22.54%)
c9bc036 — 2026-07-09 21:31:180.00 (n/a)0.00 (n/a)0.00 (n/a)0.00 (n/a)0.00 (n/a)1034.03 (n/a)988.07 (n/a)969.71 (n/a)960.13 (n/a)31.34 (n/a)

test_swiglu_decode[embedding_dim_2048-hidden_dim_2048]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.01 (+0.00%)0.01 (+2.02%)0.01 (+3.75%)0.01 (+0.00%)0.00 (+13.59%)1097.24 (+0.37%)1014.12 (-1.85%)989.42 (-3.17%)985.66 (-0.53%)47.41 (+19.81%)
c9bc036 — 2026-07-09 21:31:180.01 (n/a)0.01 (n/a)0.01 (n/a)0.01 (n/a)0.00 (n/a)1093.23 (n/a)1033.26 (n/a)1021.84 (n/a)990.94 (n/a)39.57 (n/a)
iron/operators/swiglu_prefill

test_swiglu_prefill[seq_len_256-embedding_dim_2048-hidden_dim_2048-prio_accuracy_False]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:250.95 (-0.13%)0.94 (-0.75%)0.94 (-1.02%)0.92 (-2.11%)0.01 (+153.22%)2276.59 (+2.16%)2232.07 (+0.77%)2237.22 (+1.03%)2201.83 (+0.13%)31.29 (+157.14%)
c9bc036 — 2026-07-09 21:31:180.95 (n/a)0.95 (n/a)0.95 (n/a)0.94 (n/a)0.01 (n/a)2228.51 (n/a)2214.97 (n/a)2214.31 (n/a)2199.00 (n/a)12.17 (n/a)
iron/operators/transpose

test_transpose[M_2048-N_64-aie_columns_1-channels_1-m_64-n_64-s_8-num_batches_1]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:253.25 (+3.11%)2.55 (-6.69%)2.47 (-4.44%)2.16 (-5.05%)0.43 (+15.99%)243.20 (+5.28%)209.94 (+7.78%)211.90 (+4.64%)161.20 (-3.01%)32.18 (+19.97%)
c9bc036 — 2026-07-09 21:31:183.15 (n/a)2.73 (n/a)2.59 (n/a)2.27 (n/a)0.37 (n/a)231.00 (n/a)194.78 (n/a)202.50 (n/a)166.20 (n/a)26.82 (n/a)

test_transpose[M_2048-N_64-aie_columns_1-channels_1-m_64-n_64-s_8-num_batches_2]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:255.70 (+21.48%)4.87 (+7.98%)4.82 (+6.68%)4.42 (+4.64%)0.51 (+159.74%)237.30 (-4.43%)216.92 (-6.78%)217.50 (-6.29%)183.80 (-17.69%)21.18 (+104.37%)
c9bc036 — 2026-07-09 21:31:184.70 (n/a)4.51 (n/a)4.52 (n/a)4.22 (n/a)0.20 (n/a)248.30 (n/a)232.70 (n/a)232.10 (n/a)223.30 (n/a)10.36 (n/a)

test_transpose[M_2048-N_64-aie_columns_1-channels_1-m_64-n_64-s_8]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
4bb8427 — 2026-06-23 23:08:223.72 (+20.32%)3.12 (+16.02%)3.04 (+6.32%)2.73 (+32.06%)0.43 (+8.15%)192.20 (-24.30%)170.32 (-14.26%)172.50 (-5.94%)141.00 (-16.86%)22.57 (-32.31%)
4d4b803 — 2026-06-22 17:54:573.09 (n/a)2.69 (n/a)2.86 (n/a)2.07 (n/a)0.40 (n/a)253.90 (n/a)198.64 (n/a)183.40 (n/a)169.60 (n/a)33.34 (n/a)

test_transpose[M_2048-N_64-aie_columns_1-channels_2-m_64-n_64-s_8-num_batches_1]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:46:253.67 (+38.48%)2.75 (+12.46%)2.65 (+4.62%)2.08 (+1.64%)0.63 (+155.74%)252.30 (-1.64%)198.24 (-8.24%)197.50 (-4.40%)142.90 (-27.79%)43.41 (+80.46%)
c9bc036 — 2026-07-09 21:31:182.65 (n/a)2.45 (n/a)2.54 (n/a)2.04 (n/a)0.25 (n/a)256.50 (n/a)216.04 (n/a)206.60 (n/a)197.90 (n/a)24.06 (n/a)

test_transpose[M_2048-N_64-aie_columns_1-channels_2-m_64-n_64-s_8]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
4bb8427 — 2026-06-23 23:08:224.36 (+7.95%)3.26 (+13.04%)3.33 (+22.37%)2.38 (+18.97%)0.79 (-14.92%)220.70 (-15.96%)168.40 (-14.68%)157.40 (-18.28%)120.30 (-7.39%)40.78 (-34.38%)
4d4b803 — 2026-06-22 17:54:574.04 (n/a)2.89 (n/a)2.72 (n/a)2.00 (n/a)0.93 (n/a)262.60 (n/a)197.38 (n/a)192.60 (n/a)129.90 (n/a)62.14 (n/a)
Krackan - Examples

IRON

Tested on 2026_07_09_23_14_32 at commit a678be7.

iron/applications/llama_3.2_1b
TestChecksTTFT (mean)TPS (mean)
test_llama_3_2_1b[llama_3.2_1b_prompt_1024_tokens_1]✅ 5/52.13n/a
test_llama_3_2_1b[llama_3.2_1b_prompt_1024_tokens_40]✅ 5/52.167.40
test_llama_3_2_1b[llama_3.2_1b_prompt_13_tokens_1]✅ 5/52.09n/a
test_llama_3_2_1b[llama_3.2_1b_prompt_13_tokens_40]✅ 5/52.087.23

Trends:

IRON Trends

iron/applications/llama_3.2_1b

test_llama_3_2_1b[llama_3.2_1b_prompt_1024_tokens_1]

Commit/Date TTFT (max)TTFT (mean)TTFT (median)TTFT (min)TTFT (stddev)
a678be7 — 2026-07-09 23:09:292.14 (+0.19%)2.13 (+0.14%)2.12 (-0.05%)2.11 (+0.14%)0.01 (+4.37%)
17769b8 — 2026-07-09 22:38:402.14 (n/a)2.12 (n/a)2.12 (n/a)2.11 (n/a)0.01 (n/a)

test_llama_3_2_1b[llama_3.2_1b_prompt_1024_tokens_40]

Commit/Date TPS (max)TPS (mean)TPS (median)TPS (min)TPS (stddev)TTFT (max)TTFT (mean)TTFT (median)TTFT (min)TTFT (stddev)
a678be7 — 2026-07-09 23:09:297.43 (+0.04%)7.40 (+0.11%)7.41 (+0.43%)7.37 (-0.04%)0.03 (+1.40%)2.27 (-0.18%)2.16 (-0.01%)2.13 (+0.14%)2.11 (-0.14%)0.07 (-3.03%)
17769b8 — 2026-07-09 22:38:407.43 (n/a)7.39 (n/a)7.38 (n/a)7.37 (n/a)0.03 (n/a)2.28 (n/a)2.16 (n/a)2.13 (n/a)2.12 (n/a)0.07 (n/a)

test_llama_3_2_1b[llama_3.2_1b_prompt_13_tokens_1]

Commit/Date TTFT (max)TTFT (mean)TTFT (median)TTFT (min)TTFT (stddev)
a678be7 — 2026-07-09 23:09:292.12 (+0.33%)2.09 (+0.13%)2.08 (-0.34%)2.07 (+0.63%)0.02 (-21.83%)
17769b8 — 2026-07-09 22:38:402.11 (n/a)2.08 (n/a)2.09 (n/a)2.06 (n/a)0.02 (n/a)

test_llama_3_2_1b[llama_3.2_1b_prompt_13_tokens_40]

Commit/Date TPS (max)TPS (mean)TPS (median)TPS (min)TPS (stddev)TTFT (max)TTFT (mean)TTFT (median)TTFT (min)TTFT (stddev)
a678be7 — 2026-07-09 23:09:297.28 (+0.39%)7.23 (-0.09%)7.22 (-0.19%)7.20 (-0.35%)0.03 (+238.58%)2.09 (-0.71%)2.08 (-0.56%)2.08 (-0.95%)2.07 (+0.15%)0.01 (-51.67%)
17769b8 — 2026-07-09 22:38:407.25 (n/a)7.24 (n/a)7.24 (n/a)7.22 (n/a)0.01 (n/a)2.10 (n/a)2.09 (n/a)2.10 (n/a)2.07 (n/a)0.02 (n/a)
Phoenix - Small

IRON

Tested on 2026_07_09_23_00_42 at commit a678be7.

iron/operators/axpy
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_axpy[input_length_2048-num_aie_columns_1-tile_size_2048-scalar_factor_3.0]✅ 5/5331.140.04n/a
test_axpy[input_length_2048-num_aie_columns_2-tile_size_1024-scalar_factor_3.0]✅ 5/5443.180.03n/a
test_axpy[input_length_2048-num_aie_columns_4-tile_size_512-scalar_factor_3.0]✅ 5/5467.340.03n/a
iron/operators/dequant
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_dequant[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-group_size_32]✅ 5/5359.100.02n/a
test_dequant[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-group_size_32]✅ 5/5775.300.01n/a
test_dequant[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-group_size_32]✅ 5/5440.120.01n/a
test_dequant[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-group_size_32]✅ 5/5468.460.01n/a
test_dequant[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-group_size_32]✅ 5/5487.040.01n/a
test_dequant[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-group_size_32]✅ 5/5726.380.01n/a
iron/operators/elementwise_add
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_elementwise_add[input_length_2048-num_aie_columns_1-tile_size_2048]✅ 5/5317.220.04n/a
test_elementwise_add[input_length_2048-num_aie_columns_2-tile_size_1024]✅ 5/5431.860.03n/a
test_elementwise_add[input_length_2048-num_aie_columns_4-tile_size_512]✅ 5/5333.660.04n/a
iron/operators/elementwise_mul
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_elementwise_mul[input_length_2048-num_aie_columns_1-tile_size_2048]✅ 5/5428.700.03n/a
test_elementwise_mul[input_length_2048-num_aie_columns_2-tile_size_1024]✅ 5/5526.020.03n/a
test_elementwise_mul[input_length_2048-num_aie_columns_4-tile_size_512]✅ 5/5584.000.02n/a
iron/operators/gelu
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_gelu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048]✅ 5/5288.320.03n/a
test_gelu[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024]✅ 5/5397.760.02n/a
test_gelu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024]✅ 5/5370.660.02n/a
test_gelu[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512]✅ 5/5306.660.03n/a
test_gelu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512]✅ 5/5343.140.03n/a
test_gelu[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256]✅ 5/5324.960.03n/a
iron/operators/gemm
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_gemm[M_192-K_384-N_64-num_aie_columns_4-b_col_maj_False-c_col_maj_False-m_48-k_96-n_16-trace_size_0-partition_N_1]✅ 5/5504.820.4418.95
test_gemm[M_192-K_384-N_64-num_aie_columns_4-b_col_maj_True-c_col_maj_True-m_48-k_96-n_16-trace_size_0-partition_N_1]✅ 5/5853.840.2711.70
test_gemm[M_2048-K_2048-N_2048-num_aie_columns_1-b_col_maj_False-c_col_maj_False-m_64-k_64-n_64-trace_size_0-partition_N_1]✅ 5/582114.240.31209.23
test_gemm[M_2048-K_2048-N_2048-num_aie_columns_2-b_col_maj_True-c_col_maj_False-m_64-k_64-n_64-trace_size_0-partition_N_1]✅ 5/525361.740.99677.62
test_gemm[M_384-K_1536-N_1792-num_aie_columns_4-b_col_maj_True-c_col_maj_False-m_32-k_48-n_64-trace_size_0-partition_N_1]✅ 5/53617.202.42635.20
test_gemm[M_64-K_512-N_256-num_aie_columns_4-b_col_maj_True-c_col_maj_False-m_16-k_64-n_64-trace_size_0-partition_N_4]✅ 5/56411.820.2111.23
iron/operators/gemv
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_gemv[M_128-K_128-num_aie_columns_1-tile_size_input_32-tile_size_output_128]✅ 5/5n/a0.110.11
test_gemv[M_2048-K_8192-num_aie_columns_1-tile_size_input_1-tile_size_output_2048]✅ 5/5n/a3.733.72
test_gemv[M_2048-K_8192-num_aie_columns_2-tile_size_input_1-tile_size_output_1024]✅ 5/5n/a6.136.13
test_gemv[M_2048-K_8192-num_aie_columns_4-tile_size_input_1-tile_size_output_512]✅ 5/5n/a9.879.87
test_gemv[M_8192-K_2048-num_aie_columns_1-tile_size_input_4-tile_size_output_1024]✅ 5/5n/a3.583.57
test_gemv[M_8192-K_2048-num_aie_columns_2-tile_size_input_4-tile_size_output_1024]✅ 5/5n/a5.895.89
test_gemv[M_8192-K_2048-num_aie_columns_4-tile_size_input_4-tile_size_output_1024]✅ 5/5n/a11.2511.24
iron/operators/layer_norm
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_layer_norm[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048]✅ 5/5295.420.03n/a
test_layer_norm[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024]✅ 5/5440.660.02n/a
test_layer_norm[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024]✅ 5/5358.100.03n/a
test_layer_norm[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512]✅ 5/5526.340.02n/a
test_layer_norm[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512]✅ 5/5679.000.02n/a
test_layer_norm[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256]✅ 5/5481.540.02n/a
iron/operators/leaky_relu
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-alpha_0.01]✅ 5/5422.600.02n/a
test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-alpha_0.1]✅ 5/5312.920.03n/a
test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-alpha_0.25]✅ 5/5277.280.03n/a
test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-alpha_0.01]✅ 5/5341.340.03n/a
test_leaky_relu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-alpha_0.01]✅ 5/5405.080.02n/a
test_leaky_relu[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-alpha_0.01]✅ 5/5405.620.02n/a
test_leaky_relu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-alpha_0.01]✅ 5/5432.740.02n/a
test_leaky_relu[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-alpha_0.01]✅ 5/5543.680.02n/a
iron/operators/mem_copy
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_mem_copy[input_length_2048-num_cores_1-num_channels_1-bypass_False-tile_size_2048]✅ 5/5440.220.02n/a
test_mem_copy[input_length_2048-num_cores_2-num_channels_1-bypass_False-tile_size_1024]✅ 5/5403.760.02n/a
test_mem_copy[input_length_2048-num_cores_2-num_channels_2-bypass_False-tile_size_1024]✅ 5/5488.000.02n/a
test_mem_copy[input_length_2048-num_cores_4-num_channels_1-bypass_False-tile_size_512]✅ 5/5602.680.02n/a
test_mem_copy[input_length_2048-num_cores_4-num_channels_2-bypass_False-tile_size_512]✅ 5/5363.720.02n/a
test_mem_copy[input_length_2048-num_cores_8-num_channels_2-bypass_False-tile_size_256]✅ 5/5488.800.02n/a
iron/operators/relu
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048]✅ 5/5285.180.03n/a
test_relu[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024]✅ 5/5240.320.03n/a
test_relu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024]✅ 5/5267.020.03n/a
test_relu[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512]✅ 5/5338.100.03n/a
test_relu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512]✅ 5/5350.140.03n/a
test_relu[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256]✅ 5/5377.180.02n/a
iron/operators/rms_norm
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-weighted_False]✅ 5/5413.280.02n/a
test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-weighted_True]✅ 5/5304.500.04n/a
test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-weighted_False]✅ 5/5360.940.03n/a
test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-weighted_True]✅ 5/5417.880.03n/a
test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-weighted_False]✅ 5/5399.400.02n/a
test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-weighted_True]✅ 5/5393.700.03n/a
test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-weighted_False]✅ 5/5385.720.02n/a
test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-weighted_True]✅ 5/5369.600.03n/a
test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-weighted_False]✅ 5/5604.960.02n/a
test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-weighted_True]✅ 5/5498.860.02n/a
test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-weighted_False]✅ 5/5377.500.02n/a
iron/operators/rope
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_rope[rows_32-cols_512-angle_rows_32-aie_columns_1-method_type_0]✅ 5/5431.120.26n/a
test_rope[rows_32-cols_512-angle_rows_32-aie_columns_2-method_type_0]✅ 5/5403.560.29n/a
test_rope[rows_32-cols_512-angle_rows_32-aie_columns_4-method_type_0]✅ 5/51033.980.16n/a
test_rope[rows_32-cols_512-angle_rows_8-aie_columns_1-method_type_0]✅ 5/5354.140.23n/a
test_rope[rows_32-cols_512-angle_rows_8-aie_columns_2-method_type_0]✅ 5/5348.280.23n/a
test_rope[rows_32-cols_512-angle_rows_8-aie_columns_4-method_type_0]✅ 5/5361.620.23n/a
iron/operators/sigmoid
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_sigmoid[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048]✅ 5/5367.600.02n/a
test_sigmoid[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024]✅ 5/5424.420.02n/a
test_sigmoid[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024]✅ 5/5534.940.02n/a
test_sigmoid[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512]✅ 5/5360.540.02n/a
test_sigmoid[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512]✅ 5/5412.920.02n/a
test_sigmoid[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256]✅ 5/5430.380.02n/a
iron/operators/silu
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_silu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048]✅ 5/5434.060.02n/a
test_silu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024]✅ 5/5358.880.03n/a
test_silu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512]✅ 5/5530.900.02n/a
iron/operators/softmax
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_softmax[input_length_32768-num_aie_columns_2-num_channels_2-tile_size_1024]✅ 5/5693.140.31n/a
test_softmax[input_length_32768-num_aie_columns_2-num_channels_2-tile_size_2048]✅ 5/5714.560.31n/a
test_softmax[input_length_32768-num_aie_columns_2-num_channels_2-tile_size_512]✅ 5/5450.480.32n/a
iron/operators/swiglu_decode
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_swiglu_decode[embedding_dim_1024-hidden_dim_3584]✅ 5/512270.660.00n/a
test_swiglu_decode[embedding_dim_2048-hidden_dim_2048]✅ 5/513252.320.00n/a
iron/operators/swiglu_prefill
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_swiglu_prefill[seq_len_256-embedding_dim_2048-hidden_dim_2048-prio_accuracy_False]✅ 5/521981.840.10n/a
iron/operators/tanh
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_tanh[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048]✅ 5/5396.600.02n/a
test_tanh[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024]✅ 5/5442.220.02n/a
test_tanh[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024]✅ 5/5712.680.02n/a
test_tanh[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512]✅ 5/5437.160.02n/a
test_tanh[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512]✅ 5/5388.460.02n/a
test_tanh[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256]✅ 5/5424.240.02n/a
iron/operators/transpose
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_transpose[M_2048-N_64-aie_columns_1-channels_1-m_64-n_64-s_8-num_batches_1]✅ 5/5498.281.09n/a
test_transpose[M_2048-N_64-aie_columns_1-channels_1-m_64-n_64-s_8-num_batches_2]✅ 5/51794.461.13n/a
test_transpose[M_2048-N_64-aie_columns_1-channels_2-m_64-n_64-s_8-num_batches_1]✅ 5/5572.380.95n/a

Trends:

IRON Trends

iron/operators/axpy

test_axpy[input_length_2048-num_aie_columns_1-tile_size_2048-scalar_factor_3.0]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.05 (+12.16%)0.04 (+25.22%)0.04 (+30.19%)0.02 (+6.17%)0.01 (+4.93%)595.40 (-5.82%)331.14 (-20.09%)274.90 (-23.19%)236.90 (-10.84%)148.62 (-5.06%)
17769b8 — 2026-07-09 22:38:300.05 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.01 (n/a)632.20 (n/a)414.38 (n/a)357.90 (n/a)265.70 (n/a)156.54 (n/a)

test_axpy[input_length_2048-num_aie_columns_2-tile_size_1024-scalar_factor_3.0]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.06 (+12.82%)0.03 (-10.64%)0.03 (-21.52%)0.01 (-28.25%)0.02 (+19.12%)874.30 (+39.38%)443.18 (+22.53%)356.20 (+27.40%)207.30 (-11.37%)260.57 (+56.15%)
17769b8 — 2026-07-09 22:38:300.05 (n/a)0.04 (n/a)0.04 (n/a)0.02 (n/a)0.01 (n/a)627.30 (n/a)361.70 (n/a)279.60 (n/a)233.90 (n/a)166.87 (n/a)

test_axpy[input_length_2048-num_aie_columns_4-tile_size_512-scalar_factor_3.0]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.05 (+119.86%)0.03 (+48.13%)0.02 (+19.25%)0.02 (+5.62%)0.01 (+594.59%)657.80 (-5.31%)467.34 (-22.32%)504.00 (-16.14%)244.80 (-54.52%)182.70 (+202.67%)
17769b8 — 2026-07-09 22:38:300.02 (n/a)0.02 (n/a)0.02 (n/a)0.02 (n/a)0.00 (n/a)694.70 (n/a)601.60 (n/a)601.00 (n/a)538.20 (n/a)60.36 (n/a)
iron/operators/dequant

test_dequant[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-group_size_32]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.02 (-19.37%)0.02 (-13.01%)0.02 (-21.53%)0.01 (+8.41%)0.00 (-45.95%)457.10 (-7.77%)359.10 (+6.00%)319.30 (+27.41%)272.60 (+24.02%)83.80 (-39.02%)
17769b8 — 2026-07-09 22:38:300.02 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)495.60 (n/a)338.76 (n/a)250.60 (n/a)219.80 (n/a)137.42 (n/a)

test_dequant[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-group_size_32]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.02 (+6.21%)0.01 (-8.69%)0.02 (+12.16%)0.00 (-78.73%)0.01 (+105.57%)2471.10 (+370.15%)775.30 (+112.13%)309.80 (-10.85%)233.80 (-5.84%)957.92 (+847.86%)
17769b8 — 2026-07-09 22:38:300.02 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.00 (n/a)525.60 (n/a)365.48 (n/a)347.50 (n/a)248.30 (n/a)101.06 (n/a)

test_dequant[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-group_size_32]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.02 (-31.39%)0.01 (-38.78%)0.01 (-40.04%)0.01 (-20.94%)0.00 (-38.58%)606.30 (+26.50%)440.12 (+57.47%)408.90 (+66.83%)264.00 (+45.70%)129.64 (+8.77%)
17769b8 — 2026-07-09 22:38:300.03 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)479.30 (n/a)279.50 (n/a)245.10 (n/a)181.20 (n/a)119.19 (n/a)

test_dequant[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-group_size_32]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.02 (+57.94%)0.01 (+9.61%)0.01 (-3.16%)0.01 (+0.32%)0.00 (+345.09%)527.40 (-0.32%)468.46 (-3.01%)514.40 (+3.27%)271.50 (-36.70%)110.25 (+177.70%)
17769b8 — 2026-07-09 22:38:300.01 (n/a)0.01 (n/a)0.01 (n/a)0.01 (n/a)0.00 (n/a)529.10 (n/a)483.00 (n/a)498.10 (n/a)428.90 (n/a)39.70 (n/a)

test_dequant[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-group_size_32]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.02 (-32.82%)0.01 (-32.09%)0.01 (-39.83%)0.01 (-42.25%)0.01 (-31.53%)816.50 (+73.17%)487.04 (+49.08%)453.60 (+66.21%)265.80 (+48.82%)213.56 (+64.47%)
17769b8 — 2026-07-09 22:38:300.03 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)471.50 (n/a)326.70 (n/a)272.90 (n/a)178.60 (n/a)129.85 (n/a)

test_dequant[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-group_size_32]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.02 (-14.14%)0.01 (-17.35%)0.01 (-18.47%)0.00 (-55.81%)0.01 (+0.49%)1568.80 (+126.28%)726.38 (+43.09%)610.90 (+22.65%)297.50 (+16.44%)490.22 (+198.50%)
17769b8 — 2026-07-09 22:38:300.02 (n/a)0.01 (n/a)0.01 (n/a)0.01 (n/a)0.01 (n/a)693.30 (n/a)507.64 (n/a)498.10 (n/a)255.50 (n/a)164.23 (n/a)
iron/operators/elementwise_add

test_elementwise_add[input_length_2048-num_aie_columns_1-tile_size_2048]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.05 (n/a)0.04 (n/a)0.05 (n/a)0.03 (n/a)0.01 (n/a)467.00 (n/a)317.22 (n/a)255.00 (n/a)240.80 (n/a)101.72 (n/a)

test_elementwise_add[input_length_2048-num_aie_columns_2-tile_size_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.04 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.01 (n/a)673.70 (n/a)431.86 (n/a)366.60 (n/a)306.10 (n/a)145.14 (n/a)

test_elementwise_add[input_length_2048-num_aie_columns_4-tile_size_512]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.05 (n/a)0.04 (n/a)0.04 (n/a)0.03 (n/a)0.01 (n/a)439.80 (n/a)333.66 (n/a)280.80 (n/a)248.30 (n/a)89.16 (n/a)
iron/operators/elementwise_mul

test_elementwise_mul[input_length_2048-num_aie_columns_1-tile_size_2048]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.05 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.01 (n/a)593.10 (n/a)428.70 (n/a)470.70 (n/a)261.60 (n/a)136.21 (n/a)

test_elementwise_mul[input_length_2048-num_aie_columns_2-tile_size_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.06 (n/a)0.03 (n/a)0.02 (n/a)0.02 (n/a)0.02 (n/a)810.60 (n/a)526.02 (n/a)543.50 (n/a)220.40 (n/a)210.14 (n/a)

test_elementwise_mul[input_length_2048-num_aie_columns_4-tile_size_512]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.02 (n/a)0.02 (n/a)0.02 (n/a)0.02 (n/a)0.00 (n/a)664.70 (n/a)584.00 (n/a)577.80 (n/a)503.90 (n/a)63.83 (n/a)
iron/operators/gelu

test_gelu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.03 (n/a)0.03 (n/a)0.03 (n/a)0.03 (n/a)0.00 (n/a)326.50 (n/a)288.32 (n/a)284.00 (n/a)246.60 (n/a)29.16 (n/a)

test_gelu[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.03 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)620.90 (n/a)397.76 (n/a)370.70 (n/a)277.00 (n/a)139.87 (n/a)

test_gelu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.03 (n/a)0.02 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)523.50 (n/a)370.66 (n/a)366.10 (n/a)247.90 (n/a)109.42 (n/a)

test_gelu[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.03 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.01 (n/a)497.20 (n/a)306.66 (n/a)254.50 (n/a)238.10 (n/a)109.00 (n/a)

test_gelu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.03 (n/a)0.03 (n/a)0.03 (n/a)0.01 (n/a)0.01 (n/a)552.10 (n/a)343.14 (n/a)284.40 (n/a)243.70 (n/a)124.87 (n/a)

test_gelu[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.03 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.00 (n/a)371.80 (n/a)324.96 (n/a)308.30 (n/a)296.40 (n/a)34.47 (n/a)
iron/operators/gemm

test_gemm[M_192-K_384-N_64-num_aie_columns_4-b_col_maj_False-c_col_maj_False-m_48-k_96-n_16-trace_size_0-partition_N_1]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:57:120.54 (-0.37%)0.44 (+8.06%)0.41 (+7.42%)0.40 (+18.73%)0.06 (-27.74%)555.80 (-15.78%)504.82 (-8.97%)533.30 (-6.91%)409.00 (+0.37%)62.54 (-38.37%)23.07 (-0.37%)18.95 (+8.06%)17.69 (+7.42%)16.98 (+18.73%)2.58 (-27.74%)
17769b8 — 2026-07-09 22:38:300.54 (n/a)0.41 (n/a)0.39 (n/a)0.34 (n/a)0.08 (n/a)659.90 (n/a)554.54 (n/a)572.90 (n/a)407.50 (n/a)101.48 (n/a)23.16 (n/a)17.54 (n/a)16.47 (n/a)14.30 (n/a)3.57 (n/a)

test_gemm[M_192-K_384-N_64-num_aie_columns_4-b_col_maj_True-c_col_maj_True-m_48-k_96-n_16-trace_size_0-partition_N_1]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:57:120.34 (-46.49%)0.27 (-37.21%)0.30 (-27.51%)0.18 (-42.93%)0.07 (-47.40%)1206.70 (+75.21%)853.84 (+57.89%)733.80 (+37.96%)656.30 (+86.87%)240.17 (+62.85%)14.38 (-46.49%)11.70 (-37.21%)12.86 (-27.51%)7.82 (-42.93%)2.91 (-47.40%)
17769b8 — 2026-07-09 22:38:300.63 (n/a)0.44 (n/a)0.42 (n/a)0.32 (n/a)0.13 (n/a)688.70 (n/a)540.78 (n/a)531.90 (n/a)351.20 (n/a)147.48 (n/a)26.87 (n/a)18.63 (n/a)17.74 (n/a)13.70 (n/a)5.53 (n/a)

test_gemm[M_2048-K_2048-N_2048-num_aie_columns_1-b_col_maj_False-c_col_maj_False-m_64-k_64-n_64-trace_size_0-partition_N_1]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:57:120.31 (-1.24%)0.31 (-0.01%)0.31 (-0.43%)0.30 (+1.52%)0.00 (-61.96%)83345.80 (-1.50%)82114.24 (-0.03%)81918.90 (+0.43%)81269.70 (+1.25%)762.17 (-61.95%)211.39 (-1.24%)209.23 (-0.01%)209.72 (-0.43%)206.13 (+1.52%)1.93 (-61.96%)
17769b8 — 2026-07-09 22:38:300.31 (n/a)0.31 (n/a)0.31 (n/a)0.30 (n/a)0.01 (n/a)84611.30 (n/a)82135.30 (n/a)81566.10 (n/a)80266.00 (n/a)2002.97 (n/a)214.04 (n/a)209.26 (n/a)210.62 (n/a)203.04 (n/a)5.07 (n/a)

test_gemm[M_2048-K_2048-N_2048-num_aie_columns_2-b_col_maj_True-c_col_maj_False-m_64-k_64-n_64-trace_size_0-partition_N_1]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:57:121.01 (-1.91%)0.99 (-2.44%)1.00 (-2.10%)0.97 (-3.22%)0.02 (+71.27%)26040.70 (+3.32%)25361.74 (+2.52%)25211.40 (+2.14%)24895.10 (+1.95%)518.92 (+79.66%)690.09 (-1.91%)677.62 (-2.44%)681.43 (-2.10%)659.73 (-3.22%)13.78 (+71.27%)
17769b8 — 2026-07-09 22:38:301.03 (n/a)1.02 (n/a)1.02 (n/a)1.00 (n/a)0.01 (n/a)25203.40 (n/a)24737.36 (n/a)24682.70 (n/a)24419.10 (n/a)288.84 (n/a)703.54 (n/a)694.57 (n/a)696.03 (n/a)681.65 (n/a)8.05 (n/a)

test_gemm[M_384-K_1536-N_1792-num_aie_columns_4-b_col_maj_True-c_col_maj_False-m_32-k_48-n_64-trace_size_0-partition_N_1]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:57:123.88 (+2.52%)2.42 (-1.38%)1.99 (+9.64%)1.80 (+14.56%)0.88 (-15.35%)4482.10 (-12.71%)3617.20 (-3.37%)4052.20 (-8.79%)2076.40 (-2.46%)1017.73 (-26.14%)1018.05 (+2.52%)635.20 (-1.38%)521.68 (+9.64%)471.64 (+14.56%)229.83 (-15.35%)
17769b8 — 2026-07-09 22:38:303.79 (n/a)2.46 (n/a)1.81 (n/a)1.57 (n/a)1.04 (n/a)5134.70 (n/a)3743.24 (n/a)4442.90 (n/a)2128.80 (n/a)1377.98 (n/a)992.99 (n/a)644.07 (n/a)475.80 (n/a)411.70 (n/a)271.52 (n/a)

test_gemm[M_64-K_512-N_256-num_aie_columns_4-b_col_maj_True-c_col_maj_False-m_16-k_64-n_64-trace_size_0-partition_N_4]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:57:120.28 (+30.37%)0.21 (+3.12%)0.22 (+8.19%)0.13 (-29.27%)0.06 (+387.70%)9551.90 (+41.39%)6411.82 (+3.73%)5707.30 (-7.57%)4463.30 (-23.30%)2010.25 (+439.06%)15.04 (+30.37%)11.23 (+3.12%)11.76 (+8.19%)7.03 (-29.27%)3.11 (+387.70%)
17769b8 — 2026-07-09 22:38:300.21 (n/a)0.20 (n/a)0.20 (n/a)0.18 (n/a)0.01 (n/a)6755.60 (n/a)6181.38 (n/a)6174.70 (n/a)5818.90 (n/a)372.92 (n/a)11.53 (n/a)10.89 (n/a)10.87 (n/a)9.93 (n/a)0.64 (n/a)
iron/operators/gemv

test_gemv[M_128-K_128-num_aie_columns_1-tile_size_input_32-tile_size_output_128]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:57:120.14 (+19.20%)0.11 (+52.39%)0.13 (+101.80%)0.06 (+9.42%)0.03 (+22.23%)0.14 (+19.20%)0.11 (+52.39%)0.13 (+101.80%)0.06 (+9.42%)0.03 (+22.23%)
17769b8 — 2026-07-09 22:38:300.12 (n/a)0.07 (n/a)0.06 (n/a)0.06 (n/a)0.03 (n/a)0.12 (n/a)0.07 (n/a)0.06 (n/a)0.06 (n/a)0.03 (n/a)

test_gemv[M_2048-K_8192-num_aie_columns_1-tile_size_input_1-tile_size_output_2048]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:57:123.95 (+0.37%)3.73 (-1.50%)3.64 (-6.49%)3.55 (+6.90%)0.19 (-27.97%)3.95 (+0.37%)3.72 (-1.50%)3.64 (-6.49%)3.55 (+6.90%)0.19 (-27.97%)
17769b8 — 2026-07-09 22:38:303.94 (n/a)3.78 (n/a)3.89 (n/a)3.32 (n/a)0.26 (n/a)3.94 (n/a)3.78 (n/a)3.89 (n/a)3.32 (n/a)0.26 (n/a)

test_gemv[M_2048-K_8192-num_aie_columns_2-tile_size_input_1-tile_size_output_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:57:127.56 (+0.06%)6.13 (-10.06%)5.72 (-19.07%)4.86 (-14.03%)1.07 (+47.76%)7.55 (+0.06%)6.13 (-10.06%)5.71 (-19.07%)4.86 (-14.03%)1.07 (+47.76%)
17769b8 — 2026-07-09 22:38:307.55 (n/a)6.82 (n/a)7.06 (n/a)5.65 (n/a)0.73 (n/a)7.55 (n/a)6.82 (n/a)7.06 (n/a)5.65 (n/a)0.73 (n/a)

test_gemv[M_2048-K_8192-num_aie_columns_4-tile_size_input_1-tile_size_output_512]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:57:1213.49 (-2.30%)9.87 (+3.03%)8.21 (-5.31%)6.35 (-18.77%)3.33 (+34.79%)13.48 (-2.30%)9.87 (+3.03%)8.21 (-5.31%)6.34 (-18.77%)3.33 (+34.79%)
17769b8 — 2026-07-09 22:38:3013.81 (n/a)9.58 (n/a)8.67 (n/a)7.82 (n/a)2.47 (n/a)13.80 (n/a)9.58 (n/a)8.67 (n/a)7.81 (n/a)2.47 (n/a)

test_gemv[M_8192-K_2048-num_aie_columns_1-tile_size_input_4-tile_size_output_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:57:123.82 (-3.70%)3.58 (-5.34%)3.58 (-6.32%)3.38 (-1.94%)0.17 (-16.44%)3.82 (-3.70%)3.57 (-5.34%)3.58 (-6.32%)3.38 (-1.94%)0.17 (-16.44%)
17769b8 — 2026-07-09 22:38:303.97 (n/a)3.78 (n/a)3.83 (n/a)3.45 (n/a)0.21 (n/a)3.97 (n/a)3.77 (n/a)3.82 (n/a)3.45 (n/a)0.21 (n/a)

test_gemv[M_8192-K_2048-num_aie_columns_2-tile_size_input_4-tile_size_output_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:57:127.25 (+1.71%)5.89 (-2.89%)5.66 (-7.16%)4.41 (-1.98%)1.06 (-2.66%)7.25 (+1.71%)5.89 (-2.89%)5.66 (-7.16%)4.41 (-1.98%)1.06 (-2.66%)
17769b8 — 2026-07-09 22:38:307.13 (n/a)6.06 (n/a)6.10 (n/a)4.50 (n/a)1.09 (n/a)7.13 (n/a)6.06 (n/a)6.09 (n/a)4.50 (n/a)1.09 (n/a)

test_gemv[M_8192-K_2048-num_aie_columns_4-tile_size_input_4-tile_size_output_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
a678be7 — 2026-07-09 22:57:1213.84 (-0.29%)11.25 (+3.56%)12.33 (+31.36%)8.12 (-0.78%)2.89 (+8.08%)13.83 (-0.29%)11.24 (+3.56%)12.32 (+31.36%)8.11 (-0.78%)2.89 (+8.08%)
17769b8 — 2026-07-09 22:38:3013.88 (n/a)10.86 (n/a)9.39 (n/a)8.18 (n/a)2.68 (n/a)13.88 (n/a)10.86 (n/a)9.38 (n/a)8.18 (n/a)2.68 (n/a)
iron/operators/layer_norm

test_layer_norm[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.04 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.01 (n/a)434.50 (n/a)295.42 (n/a)278.30 (n/a)227.40 (n/a)80.92 (n/a)

test_layer_norm[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.04 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)576.10 (n/a)440.66 (n/a)537.80 (n/a)231.80 (n/a)163.81 (n/a)

test_layer_norm[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.04 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.01 (n/a)523.00 (n/a)358.10 (n/a)291.90 (n/a)214.30 (n/a)138.52 (n/a)

test_layer_norm[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.03 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)706.40 (n/a)526.34 (n/a)524.30 (n/a)291.00 (n/a)163.11 (n/a)

test_layer_norm[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.04 (n/a)0.02 (n/a)0.02 (n/a)0.00 (n/a)0.01 (n/a)2032.20 (n/a)679.00 (n/a)344.60 (n/a)222.80 (n/a)765.65 (n/a)

test_layer_norm[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.04 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)612.50 (n/a)481.54 (n/a)522.80 (n/a)230.60 (n/a)147.52 (n/a)
iron/operators/leaky_relu

test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-alpha_0.01]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.03 (-5.95%)0.02 (-20.28%)0.02 (-34.22%)0.01 (-27.50%)0.01 (+31.49%)628.30 (+37.94%)422.60 (+31.87%)449.50 (+52.01%)269.90 (+6.30%)141.46 (+79.82%)
17769b8 — 2026-07-09 22:38:300.03 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.01 (n/a)455.50 (n/a)320.46 (n/a)295.70 (n/a)253.90 (n/a)78.67 (n/a)

test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-alpha_0.1]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.03 (-0.77%)0.03 (+5.03%)0.03 (+12.82%)0.02 (-5.41%)0.01 (+7.49%)515.50 (+5.72%)312.92 (-3.29%)260.90 (-11.38%)246.70 (+0.78%)114.28 (+16.55%)
17769b8 — 2026-07-09 22:38:300.03 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.01 (n/a)487.60 (n/a)323.56 (n/a)294.40 (n/a)244.80 (n/a)98.05 (n/a)

test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-alpha_0.25]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.04 (+12.52%)0.03 (+35.26%)0.03 (+74.41%)0.02 (+63.25%)0.01 (-37.63%)390.60 (-38.74%)277.28 (-36.38%)275.90 (-42.66%)204.80 (-11.11%)71.46 (-63.57%)
17769b8 — 2026-07-09 22:38:300.04 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)637.60 (n/a)435.86 (n/a)481.20 (n/a)230.40 (n/a)196.18 (n/a)

test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-alpha_0.01]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.03 (+33.60%)0.03 (+56.63%)0.03 (+106.96%)0.01 (-7.11%)0.01 (+70.18%)638.00 (+7.66%)341.34 (-31.11%)264.20 (-51.67%)234.70 (-25.14%)168.49 (+45.02%)
17769b8 — 2026-07-09 22:38:300.03 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)0.01 (n/a)592.60 (n/a)495.52 (n/a)546.70 (n/a)313.50 (n/a)116.18 (n/a)

test_leaky_relu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-alpha_0.01]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.03 (+14.13%)0.02 (+18.15%)0.02 (+38.87%)0.01 (+15.67%)0.01 (+3.29%)567.40 (-13.55%)405.08 (-17.32%)396.20 (-27.98%)236.60 (-12.37%)119.50 (-24.28%)
17769b8 — 2026-07-09 22:38:300.03 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)0.01 (n/a)656.30 (n/a)489.94 (n/a)550.10 (n/a)270.00 (n/a)157.82 (n/a)

test_leaky_relu[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-alpha_0.01]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.03 (-7.03%)0.02 (-13.95%)0.02 (-43.13%)0.02 (+19.38%)0.01 (-26.09%)503.00 (-16.24%)405.62 (+7.84%)486.40 (+75.85%)253.50 (+7.55%)123.07 (-30.04%)
17769b8 — 2026-07-09 22:38:300.03 (n/a)0.03 (n/a)0.03 (n/a)0.01 (n/a)0.01 (n/a)600.50 (n/a)376.12 (n/a)276.60 (n/a)235.70 (n/a)175.91 (n/a)

test_leaky_relu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-alpha_0.01]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.03 (+3.58%)0.02 (-11.08%)0.02 (-18.02%)0.02 (+25.04%)0.01 (-11.07%)502.60 (-20.02%)432.74 (+8.55%)471.70 (+21.98%)255.70 (-3.47%)102.56 (-30.53%)
17769b8 — 2026-07-09 22:38:300.03 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)628.40 (n/a)398.64 (n/a)386.70 (n/a)264.90 (n/a)147.64 (n/a)

test_leaky_relu[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-alpha_0.01]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.02 (-38.26%)0.02 (-27.10%)0.02 (-29.28%)0.01 (-4.55%)0.00 (-63.96%)694.30 (+4.77%)543.68 (+23.29%)541.70 (+41.40%)420.50 (+61.98%)108.27 (-40.75%)
17769b8 — 2026-07-09 22:38:300.03 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)662.70 (n/a)440.96 (n/a)383.10 (n/a)259.60 (n/a)182.72 (n/a)
iron/operators/mem_copy

test_mem_copy[input_length_2048-num_cores_1-num_channels_1-bypass_False-tile_size_2048]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.03 (-16.90%)0.02 (-19.34%)0.02 (-22.93%)0.01 (-20.91%)0.01 (-23.77%)669.70 (+26.43%)440.22 (+22.28%)436.80 (+29.77%)272.60 (+20.35%)155.80 (+16.34%)
17769b8 — 2026-07-09 22:38:300.04 (n/a)0.03 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)529.70 (n/a)360.02 (n/a)336.60 (n/a)226.50 (n/a)133.92 (n/a)

test_mem_copy[input_length_2048-num_cores_2-num_channels_1-bypass_False-tile_size_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.03 (-8.65%)0.02 (-1.04%)0.02 (+6.54%)0.01 (-6.34%)0.01 (+1.10%)629.70 (+6.78%)403.76 (+2.45%)374.80 (-6.14%)268.70 (+9.45%)152.97 (+13.47%)
17769b8 — 2026-07-09 22:38:300.03 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)589.70 (n/a)394.12 (n/a)399.30 (n/a)245.50 (n/a)134.81 (n/a)

test_mem_copy[input_length_2048-num_cores_2-num_channels_2-bypass_False-tile_size_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.02 (-20.59%)0.02 (-19.76%)0.02 (-28.83%)0.01 (-12.60%)0.00 (-6.23%)561.80 (+14.42%)488.00 (+25.24%)533.70 (+40.52%)396.80 (+25.93%)83.40 (+30.47%)
17769b8 — 2026-07-09 22:38:300.03 (n/a)0.02 (n/a)0.02 (n/a)0.02 (n/a)0.00 (n/a)491.00 (n/a)389.66 (n/a)379.80 (n/a)315.10 (n/a)63.92 (n/a)

test_mem_copy[input_length_2048-num_cores_4-num_channels_1-bypass_False-tile_size_512]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.04 (+21.13%)0.02 (+7.04%)0.02 (-2.72%)0.01 (-14.31%)0.01 (+33.42%)1303.40 (+16.70%)602.68 (+4.93%)538.80 (+2.80%)201.10 (-17.45%)415.62 (+27.55%)
17769b8 — 2026-07-09 22:38:300.03 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)1116.90 (n/a)574.36 (n/a)524.10 (n/a)243.60 (n/a)325.84 (n/a)

test_mem_copy[input_length_2048-num_cores_4-num_channels_2-bypass_False-tile_size_512]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.03 (+19.96%)0.02 (+10.94%)0.02 (-7.44%)0.02 (+25.89%)0.01 (+31.17%)476.20 (-20.55%)363.72 (-8.97%)373.80 (+8.03%)239.50 (-16.67%)109.51 (-13.29%)
17769b8 — 2026-07-09 22:38:300.03 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)599.40 (n/a)399.54 (n/a)346.00 (n/a)287.40 (n/a)126.29 (n/a)

test_mem_copy[input_length_2048-num_cores_8-num_channels_2-bypass_False-tile_size_256]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.03 (+1.15%)0.02 (+4.69%)0.02 (-14.09%)0.01 (+254.26%)0.01 (-26.08%)563.10 (-71.77%)488.80 (-33.79%)544.80 (+16.41%)267.80 (-1.14%)124.77 (-82.42%)
17769b8 — 2026-07-09 22:38:300.03 (n/a)0.02 (n/a)0.02 (n/a)0.00 (n/a)0.01 (n/a)1995.00 (n/a)738.24 (n/a)468.00 (n/a)270.90 (n/a)709.78 (n/a)
iron/operators/rms_norm

test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-weighted_False]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.03 (-4.58%)0.02 (-12.71%)0.02 (-34.29%)0.01 (-19.64%)0.01 (+3.73%)611.80 (+24.43%)413.28 (+17.08%)443.60 (+52.18%)253.60 (+4.84%)149.60 (+22.92%)
17769b8 — 2026-07-09 22:38:300.03 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.01 (n/a)491.70 (n/a)353.00 (n/a)291.50 (n/a)241.90 (n/a)121.70 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-weighted_True]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.05 (+3.50%)0.04 (+11.08%)0.04 (-7.65%)0.03 (+23.54%)0.01 (-23.74%)454.90 (-19.06%)304.50 (-15.28%)292.30 (+8.30%)233.30 (-3.36%)88.84 (-38.63%)
17769b8 — 2026-07-09 22:38:300.05 (n/a)0.04 (n/a)0.05 (n/a)0.02 (n/a)0.01 (n/a)562.00 (n/a)359.42 (n/a)269.90 (n/a)241.40 (n/a)144.78 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-weighted_False]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.05 (+38.18%)0.03 (+14.64%)0.03 (-0.65%)0.01 (+2.52%)0.01 (+49.63%)565.00 (-2.45%)360.94 (-6.09%)296.40 (+0.65%)169.00 (-27.65%)181.77 (+9.01%)
17769b8 — 2026-07-09 22:38:300.04 (n/a)0.02 (n/a)0.03 (n/a)0.01 (n/a)0.01 (n/a)579.20 (n/a)384.34 (n/a)294.50 (n/a)233.60 (n/a)166.75 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-weighted_True]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.04 (+4.88%)0.03 (-1.53%)0.02 (-1.05%)0.02 (+1.58%)0.01 (-13.57%)554.20 (-1.56%)417.88 (-1.12%)423.10 (+1.08%)271.40 (-4.64%)103.33 (-21.51%)
17769b8 — 2026-07-09 22:38:300.04 (n/a)0.03 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)563.00 (n/a)422.60 (n/a)418.60 (n/a)284.60 (n/a)131.65 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-weighted_False]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.04 (+13.38%)0.02 (+27.99%)0.02 (+25.91%)0.01 (+86.63%)0.01 (+8.84%)562.10 (-46.42%)399.40 (-27.28%)384.80 (-20.58%)232.90 (-11.78%)151.53 (-48.99%)
17769b8 — 2026-07-09 22:38:300.03 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)1049.10 (n/a)549.26 (n/a)484.50 (n/a)264.00 (n/a)297.02 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-weighted_True]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.04 (+10.46%)0.03 (+15.04%)0.03 (+6.24%)0.02 (+226.66%)0.01 (-18.82%)616.80 (-69.39%)393.70 (-42.41%)310.70 (-5.88%)246.80 (-9.46%)159.40 (-78.71%)
17769b8 — 2026-07-09 22:38:300.04 (n/a)0.03 (n/a)0.03 (n/a)0.01 (n/a)0.01 (n/a)2014.80 (n/a)683.68 (n/a)330.10 (n/a)272.60 (n/a)748.79 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-weighted_False]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.04 (-12.21%)0.02 (-2.97%)0.02 (-4.19%)0.02 (+8.91%)0.01 (-17.39%)526.90 (-8.17%)385.72 (-0.40%)443.30 (+4.38%)229.50 (+13.90%)128.00 (-12.99%)
17769b8 — 2026-07-09 22:38:300.04 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)573.80 (n/a)387.28 (n/a)424.70 (n/a)201.50 (n/a)147.11 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-weighted_True]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.04 (+71.84%)0.03 (+34.20%)0.03 (+46.13%)0.02 (-15.58%)0.01 (+393.66%)587.20 (+18.46%)369.60 (-15.97%)297.20 (-31.55%)219.50 (-41.81%)152.74 (+252.04%)
17769b8 — 2026-07-09 22:38:300.02 (n/a)0.02 (n/a)0.02 (n/a)0.02 (n/a)0.00 (n/a)495.70 (n/a)439.86 (n/a)434.20 (n/a)377.20 (n/a)43.39 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-weighted_False]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.03 (+56.48%)0.02 (+14.27%)0.01 (-6.08%)0.01 (-7.39%)0.01 (+104.50%)1108.50 (+7.98%)604.96 (-1.49%)563.70 (+6.48%)262.80 (-36.11%)311.58 (+30.42%)
17769b8 — 2026-07-09 22:38:300.02 (n/a)0.01 (n/a)0.02 (n/a)0.01 (n/a)0.00 (n/a)1026.60 (n/a)614.14 (n/a)529.40 (n/a)411.30 (n/a)238.90 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-weighted_True]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.03 (-27.56%)0.02 (-28.79%)0.02 (-40.63%)0.01 (-10.39%)0.01 (-39.61%)642.60 (+11.60%)498.86 (+35.18%)513.00 (+68.47%)363.80 (+38.06%)124.40 (-6.99%)
17769b8 — 2026-07-09 22:38:300.03 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.01 (n/a)575.80 (n/a)369.04 (n/a)304.50 (n/a)263.50 (n/a)133.75 (n/a)

test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-weighted_False]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.03 (+22.91%)0.02 (+44.86%)0.02 (+33.16%)0.02 (+139.97%)0.01 (-18.38%)453.50 (-58.32%)377.50 (-40.27%)426.40 (-24.90%)271.60 (-18.63%)91.16 (-71.14%)
17769b8 — 2026-07-09 22:38:300.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)0.01 (n/a)1088.10 (n/a)632.02 (n/a)567.80 (n/a)333.80 (n/a)315.84 (n/a)
iron/operators/rope

test_rope[rows_32-cols_512-angle_rows_32-aie_columns_1-method_type_0]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.39 (+3.71%)0.26 (+6.13%)0.24 (+19.04%)0.15 (+18.96%)0.09 (-18.08%)656.30 (-15.93%)431.12 (-12.98%)404.90 (-16.00%)252.30 (-3.59%)159.89 (-30.72%)
17769b8 — 2026-07-09 22:38:300.38 (n/a)0.24 (n/a)0.20 (n/a)0.13 (n/a)0.12 (n/a)780.70 (n/a)495.44 (n/a)482.00 (n/a)261.70 (n/a)230.78 (n/a)

test_rope[rows_32-cols_512-angle_rows_32-aie_columns_2-method_type_0]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.47 (+62.92%)0.29 (+24.04%)0.21 (-10.45%)0.17 (-10.44%)0.14 (+257.80%)571.60 (+11.66%)403.56 (-6.42%)462.30 (+11.67%)208.30 (-38.61%)167.08 (+141.43%)
17769b8 — 2026-07-09 22:38:300.29 (n/a)0.23 (n/a)0.24 (n/a)0.19 (n/a)0.04 (n/a)511.90 (n/a)431.24 (n/a)414.00 (n/a)339.30 (n/a)69.21 (n/a)

test_rope[rows_32-cols_512-angle_rows_32-aie_columns_4-method_type_0]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.28 (-26.50%)0.16 (-44.82%)0.17 (-50.98%)0.05 (-69.03%)0.10 (-4.27%)1947.20 (+222.87%)1033.98 (+159.55%)594.70 (+104.01%)353.00 (+36.08%)801.75 (+356.60%)
17769b8 — 2026-07-09 22:38:300.38 (n/a)0.29 (n/a)0.34 (n/a)0.16 (n/a)0.11 (n/a)603.10 (n/a)398.38 (n/a)291.50 (n/a)259.40 (n/a)175.59 (n/a)

test_rope[rows_32-cols_512-angle_rows_8-aie_columns_1-method_type_0]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.31 (+12.07%)0.23 (+15.78%)0.27 (+63.97%)0.14 (-3.12%)0.07 (+21.47%)535.40 (+3.22%)354.14 (-11.28%)276.40 (-39.01%)240.70 (-10.79%)129.45 (+16.01%)
17769b8 — 2026-07-09 22:38:300.27 (n/a)0.20 (n/a)0.16 (n/a)0.14 (n/a)0.06 (n/a)518.70 (n/a)399.16 (n/a)453.20 (n/a)269.80 (n/a)111.58 (n/a)

test_rope[rows_32-cols_512-angle_rows_8-aie_columns_2-method_type_0]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.29 (+18.94%)0.23 (+30.51%)0.24 (+57.82%)0.15 (+35.09%)0.07 (+20.86%)480.30 (-25.97%)348.28 (-23.69%)308.40 (-36.65%)251.80 (-15.93%)106.44 (-23.11%)
17769b8 — 2026-07-09 22:38:300.25 (n/a)0.17 (n/a)0.15 (n/a)0.11 (n/a)0.05 (n/a)648.80 (n/a)456.40 (n/a)486.80 (n/a)299.50 (n/a)138.44 (n/a)

test_rope[rows_32-cols_512-angle_rows_8-aie_columns_4-method_type_0]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.32 (+11.73%)0.23 (+3.88%)0.25 (-12.61%)0.14 (+44.85%)0.09 (-2.55%)535.80 (-30.96%)361.62 (-9.67%)299.80 (+14.43%)227.60 (-10.50%)147.45 (-34.86%)
17769b8 — 2026-07-09 22:38:300.29 (n/a)0.22 (n/a)0.28 (n/a)0.10 (n/a)0.09 (n/a)776.10 (n/a)400.32 (n/a)262.00 (n/a)254.30 (n/a)226.37 (n/a)
iron/operators/softmax

test_softmax[input_length_32768-num_aie_columns_2-num_channels_2-tile_size_1024]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.46 (+75.94%)0.31 (+36.51%)0.34 (+52.69%)0.07 (-68.21%)0.16 (+562.78%)1999.20 (+214.59%)693.14 (+20.63%)384.80 (-34.51%)285.00 (-43.16%)735.21 (+1147.10%)
17769b8 — 2026-07-09 22:38:300.26 (n/a)0.23 (n/a)0.22 (n/a)0.21 (n/a)0.02 (n/a)635.50 (n/a)574.58 (n/a)587.60 (n/a)501.40 (n/a)58.95 (n/a)

test_softmax[input_length_32768-num_aie_columns_2-num_channels_2-tile_size_2048]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.46 (+15.74%)0.31 (+17.81%)0.38 (+43.04%)0.06 (-60.70%)0.16 (+79.92%)2105.40 (+154.49%)714.56 (+32.26%)343.50 (-30.10%)282.80 (-13.60%)782.54 (+323.86%)
17769b8 — 2026-07-09 22:38:300.40 (n/a)0.27 (n/a)0.27 (n/a)0.16 (n/a)0.09 (n/a)827.30 (n/a)540.28 (n/a)491.40 (n/a)327.30 (n/a)184.62 (n/a)

test_softmax[input_length_32768-num_aie_columns_2-num_channels_2-tile_size_512]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.46 (+6.19%)0.32 (-1.77%)0.26 (-32.28%)0.20 (+62.35%)0.12 (-5.35%)658.50 (-38.41%)450.48 (-8.79%)498.00 (+47.69%)283.40 (-5.85%)159.05 (-51.16%)
17769b8 — 2026-07-09 22:38:300.44 (n/a)0.33 (n/a)0.39 (n/a)0.12 (n/a)0.13 (n/a)1069.10 (n/a)493.90 (n/a)337.20 (n/a)301.00 (n/a)325.66 (n/a)
iron/operators/swiglu_decode

test_swiglu_decode[embedding_dim_1024-hidden_dim_3584]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.00 (+16.67%)0.00 (+42.86%)0.00 (+50.00%)0.00 (+0.00%)0.00 (+31.10%)18960.21 (-12.22%)12270.66 (-22.40%)12705.55 (-25.69%)5798.21 (-13.41%)6019.29 (+9.61%)
17769b8 — 2026-07-09 22:38:300.00 (n/a)0.00 (n/a)0.00 (n/a)0.00 (n/a)0.00 (n/a)21598.80 (n/a)15813.49 (n/a)17097.23 (n/a)6695.96 (n/a)5491.52 (n/a)

test_swiglu_decode[embedding_dim_2048-hidden_dim_2048]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.00 (+0.00%)0.00 (-9.76%)0.00 (-16.67%)0.00 (+0.00%)0.00 (-2.82%)19120.20 (-10.91%)13252.32 (+8.07%)16294.34 (+22.58%)6125.42 (-2.45%)5724.31 (-7.78%)
17769b8 — 2026-07-09 22:38:300.00 (n/a)0.00 (n/a)0.00 (n/a)0.00 (n/a)0.00 (n/a)21462.12 (n/a)12263.19 (n/a)13292.95 (n/a)6279.46 (n/a)6206.94 (n/a)
iron/operators/swiglu_prefill

test_swiglu_prefill[seq_len_256-embedding_dim_2048-hidden_dim_2048-prio_accuracy_False]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:120.14 (+6.12%)0.10 (+9.17%)0.10 (+17.60%)0.07 (-3.82%)0.03 (+17.50%)28725.95 (+3.94%)21981.84 (-6.96%)20776.09 (-15.03%)14929.67 (-5.78%)5723.87 (+19.13%)
17769b8 — 2026-07-09 22:38:300.13 (n/a)0.09 (n/a)0.09 (n/a)0.08 (n/a)0.02 (n/a)27636.97 (n/a)23625.81 (n/a)24451.70 (n/a)15844.81 (n/a)4804.73 (n/a)
iron/operators/transpose

test_transpose[M_2048-N_64-aie_columns_1-channels_1-m_64-n_64-s_8-num_batches_1]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:121.39 (-21.90%)1.09 (-25.61%)0.99 (-34.80%)0.82 (-16.94%)0.23 (-20.49%)635.60 (+20.40%)498.28 (+34.00%)527.10 (+53.36%)377.40 (+28.06%)104.84 (+14.99%)
17769b8 — 2026-07-09 22:38:301.78 (n/a)1.47 (n/a)1.53 (n/a)0.99 (n/a)0.29 (n/a)527.90 (n/a)371.84 (n/a)343.70 (n/a)294.70 (n/a)91.18 (n/a)

test_transpose[M_2048-N_64-aie_columns_1-channels_1-m_64-n_64-s_8-num_batches_2]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:122.38 (-7.95%)1.13 (-46.75%)1.11 (-43.17%)0.29 (-83.60%)0.88 (+155.89%)3605.00 (+509.67%)1794.46 (+255.59%)943.10 (+75.95%)440.60 (+8.66%)1529.80 (+1863.16%)
17769b8 — 2026-07-09 22:38:302.59 (n/a)2.12 (n/a)1.96 (n/a)1.77 (n/a)0.34 (n/a)591.30 (n/a)504.64 (n/a)536.00 (n/a)405.50 (n/a)77.93 (n/a)

test_transpose[M_2048-N_64-aie_columns_1-channels_1-m_64-n_64-s_8]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
4bb8427 — 2026-06-25 20:01:371.58 (-2.67%)1.21 (+5.65%)1.29 (+41.89%)0.66 (-20.05%)0.38 (-2.53%)792.90 (+25.08%)479.12 (-3.80%)406.80 (-29.52%)331.10 (+2.76%)189.18 (+26.62%)
4bb8427 — 2026-06-23 22:46:491.63 (n/a)1.14 (n/a)0.91 (n/a)0.83 (n/a)0.38 (n/a)633.90 (n/a)498.02 (n/a)577.20 (n/a)322.20 (n/a)149.40 (n/a)

test_transpose[M_2048-N_64-aie_columns_1-channels_2-m_64-n_64-s_8-num_batches_1]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
a678be7 — 2026-07-09 22:57:121.22 (-38.99%)0.95 (-29.01%)0.86 (-31.70%)0.75 (-18.58%)0.20 (-52.57%)695.50 (+22.81%)572.38 (+35.21%)612.90 (+46.42%)431.30 (+63.93%)115.79 (-6.31%)
17769b8 — 2026-07-09 22:38:301.99 (n/a)1.34 (n/a)1.25 (n/a)0.93 (n/a)0.43 (n/a)566.30 (n/a)423.32 (n/a)418.60 (n/a)263.10 (n/a)123.59 (n/a)

test_transpose[M_2048-N_64-aie_columns_1-channels_2-m_64-n_64-s_8]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
4bb8427 — 2026-06-25 20:01:371.75 (-15.44%)1.32 (-0.50%)1.36 (+22.47%)0.77 (-19.20%)0.36 (-21.75%)678.10 (+23.76%)430.02 (-0.09%)385.10 (-18.34%)300.10 (+18.29%)146.96 (+19.46%)
4bb8427 — 2026-06-23 22:46:492.07 (n/a)1.32 (n/a)1.11 (n/a)0.96 (n/a)0.46 (n/a)547.90 (n/a)430.40 (n/a)471.60 (n/a)253.70 (n/a)123.02 (n/a)
Phoenix - Examples

IRON

Tested on 2026_07_09_23_12_10 at commit a678be7.

Trends:

IRON Trends

@andrej andrej force-pushed the fix/pin-package-indexes branch from f8cacd9 to d5ab18a Compare July 9, 2026 22:38
Replace --extra-index-url with --find-links for the GitHub Release
wheel sources. This avoids adding them as full package indexes that
pip queries for all packages. The CPU-only PyTorch index is kept as
an --extra-index-url since it serves a standard PEP 503 index.
@andrej andrej force-pushed the fix/pin-package-indexes branch from d5ab18a to c3464f5 Compare July 9, 2026 22:38
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant