Skip to content

ci: guard the PR comment against GitHub's 65536-character limit - #183

Open
atassis wants to merge 2 commits into
amd:develfrom
atassis:ci-pr-comment-size-limit
Open

ci: guard the PR comment against GitHub's 65536-character limit#183
atassis wants to merge 2 commits into
amd:develfrom
atassis:ci-pr-comment-size-limit

Conversation

@atassis

@atassis atassis commented Sep 1, 2026

Copy link
Copy Markdown
Contributor

pr-comment.yml posts the aggregate body in one call. Once all four suites have reported for a head SHA it can exceed what GitHub accepts: on run 33550798970 the body was 313,521 characters and the API returned 422 Validation Failed: body is too long (maximum is 65536 characters), so the PR received no results comment. Earlier runs on the same SHA posted fine because fewer suites had finished.

The 65536 in that message is not the enforced ceiling. Every aggregate comment on #150-#183 runs between 122,494 and 142,053 characters and all of them posted, so a guard comparing against a constant would strip the details off comments GitHub accepts today. Post the full body instead, and shed a section only when GitHub actually rejects one.

Sections go in size order. The trend tables are the bulk of an oversized body, so they go first: on run 33550798970 that takes the body from 313,521 to 83,961 characters and keeps every per-suite pass/fail table. After that the per-suite details, then the summary halved until it fits. The marker stays in the body on every rung, so the update branch keeps finding the existing comment instead of creating a new one each run.

Verified by replaying that run's four result artifacts through the script with a stubbed API rejecting above a given size: at 262144 it lands on the trends-dropped rung, at 65536 on the summary, and with nothing over the limit it makes a single call with the body unchanged.

Added

  • Fallback ladder in the Comment on PR step, with a core.warning when the comment had to be shortened.

Changed

  • Body assembled from a header constant so each rung can be rebuilt from it.
  • Create and update go through one post() helper.

Removed

PR Merge Checklist

  1. The PR is rebased on the latest devel commit and pointing to devel.
  2. Your PR has been reviewed and approved.
  3. All checks are passing.

The aggregate body is posted unchecked. Once all four suites have reported
for a head SHA it runs past the cap: on run 33550798970 it was over 265000
characters and the API returned 422 Validation Failed, so the PR received
no results comment at all.

Shed the per-suite details first, then truncate the summary, keeping the
marker so the next run still updates the same comment.

Claude-Session: https://claude.ai/code/session_014y7hzKhehcyoe74h2pRmfW
@andrej

andrej commented Sep 1, 2026

Copy link
Copy Markdown
Collaborator

Any good way to test this? Maybe we could make a lot of tests fail on purpose in this PR so we can see it in action (and we can fix them before we merge for real)?

@github-actions

github-actions Bot commented Sep 1, 2026

Copy link
Copy Markdown
Contributor

CI Test Results

7e49fd8 (2026_09_01_22_41_43)

IRON - CI Summary

Small

iron/operators/axpy
Test Krackan Status Krackan Phoenix Status Phoenix Latency (mean)
test_axpy[input_length_2048-num_aie_columns_1-tile_size_2048-scalar_factor_3.0] - - 275.80
test_axpy[input_length_2048-num_aie_columns_2-tile_size_1024-scalar_factor_3.0] - - 414.00
test_axpy[input_length_2048-num_aie_columns_4-tile_size_512-scalar_factor_3.0] - - 398.66
iron/operators/dequant
Test Krackan Status Krackan Phoenix Status Phoenix Latency (mean)
test_dequant[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-group_size_32] - - 352.06
test_dequant[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-group_size_32] - - 454.98
test_dequant[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-group_size_32] - - 353.44
test_dequant[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-group_size_32] - - 389.98
test_dequant[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-group_size_32] - - 462.86
test_dequant[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-group_size_32] - - 503.72
iron/operators/elementwise_add
Test Krackan Status Krackan Phoenix Status Phoenix Latency (mean)
test_elementwise_add[input_length_2048-num_aie_columns_1-tile_size_2048] - - 381.56
test_elementwise_add[input_length_2048-num_aie_columns_2-tile_size_1024] - - 324.84
test_elementwise_add[input_length_2048-num_aie_columns_4-tile_size_512] - - 514.56
iron/operators/elementwise_mul
Test Krackan Status Krackan Phoenix Status Phoenix Latency (mean)
test_elementwise_mul[input_length_2048-num_aie_columns_1-tile_size_2048] - - 286.90
test_elementwise_mul[input_length_2048-num_aie_columns_2-tile_size_1024] - - 315.54
test_elementwise_mul[input_length_2048-num_aie_columns_4-tile_size_512] - - 387.72
iron/operators/gelu
Test Krackan Status Krackan Phoenix Status Phoenix Latency (mean)
test_gelu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048] - - 323.04
test_gelu[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024] - - 706.06
test_gelu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024] - - 473.90
test_gelu[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512] - - 472.34
test_gelu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512] - - 1039.56
test_gelu[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256] - - 844.12
iron/operators/gemm
Test Krackan Status Krackan Phoenix Status Phoenix Latency (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] - - 626.36
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] - - 482.08
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] - - 82496.72
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] - - 24774.46
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] - - 4036.90
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] - - 7006.76
iron/operators/gemv
Test Krackan Status Krackan Phoenix Status Phoenix Latency (mean)
test_gemv[M_128-K_128-num_aie_columns_1-tile_size_input_32-tile_size_output_128] - - 0.09
test_gemv[M_2048-K_8192-num_aie_columns_1-tile_size_input_1-tile_size_output_2048] - - 3.71
test_gemv[M_2048-K_8192-num_aie_columns_2-tile_size_input_1-tile_size_output_1024] - - 6.70
test_gemv[M_2048-K_8192-num_aie_columns_4-tile_size_input_1-tile_size_output_512] - - 9.69
test_gemv[M_8192-K_2048-num_aie_columns_1-tile_size_input_4-tile_size_output_1024] - - 3.56
test_gemv[M_8192-K_2048-num_aie_columns_2-tile_size_input_4-tile_size_output_1024] - - 6.75
test_gemv[M_8192-K_2048-num_aie_columns_4-tile_size_input_4-tile_size_output_1024] - - 9.17
test_gemv_batched[M_1024-K_1024-num_aie_columns_1-tile_size_input_1-tile_size_output_64-num_batches_2] - - 1.92
test_gemv_batched[M_1026-K_64-num_aie_columns_1-tile_size_input_1-tile_size_output_2-num_batches_2] - - 0.37
test_gemv_batched[M_256-K_128-num_aie_columns_1-tile_size_input_1-tile_size_output_256-num_batches_4] - - 0.46
test_gemv_batched[M_64-K_1536-num_aie_columns_1-tile_size_input_1-tile_size_output_64-num_batches_8] - - 1.82
test_gemv_gelu[M_128-K_128-num_aie_columns_1-tile_size_input_32-tile_size_output_128] - - -
test_gemv_gelu[M_2048-K_8192-num_aie_columns_1-tile_size_input_1-tile_size_output_2048] - - -
test_gemv_gelu[M_8192-K_2048-num_aie_columns_1-tile_size_input_4-tile_size_output_1024] - - -
iron/operators/layer_norm
Test Krackan Status Krackan Phoenix Status Phoenix Latency (mean)
test_layer_norm[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048] - - 367.04
test_layer_norm[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024] - - 377.06
test_layer_norm[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024] - - 343.72
test_layer_norm[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512] - - 434.52
test_layer_norm[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512] - - 529.18
test_layer_norm[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256] - - 478.52
iron/operators/leaky_relu
Test Krackan Status Krackan Phoenix Status Phoenix Latency (mean)
test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-alpha_0.01] - - 367.64
test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-alpha_0.1] - - 382.46
test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-alpha_0.25] - - 834.94
test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-alpha_0.01] - - 406.94
test_leaky_relu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-alpha_0.01] - - 354.28
test_leaky_relu[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-alpha_0.01] - - 421.14
test_leaky_relu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-alpha_0.01] - - 848.96
test_leaky_relu[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-alpha_0.01] - - 644.66
iron/operators/mem_copy
Test Krackan Status Krackan Phoenix Status Phoenix Latency (mean)
test_mem_copy[input_length_2048-num_cores_1-num_channels_1-bypass_False-tile_size_2048] - - 283.88
test_mem_copy[input_length_2048-num_cores_2-num_channels_1-bypass_False-tile_size_1024] - - 310.02
test_mem_copy[input_length_2048-num_cores_2-num_channels_2-bypass_False-tile_size_1024] - - 289.90
test_mem_copy[input_length_2048-num_cores_4-num_channels_1-bypass_False-tile_size_512] - - 369.08
test_mem_copy[input_length_2048-num_cores_4-num_channels_2-bypass_False-tile_size_512] - - 393.40
test_mem_copy[input_length_2048-num_cores_8-num_channels_2-bypass_False-tile_size_256] - - 498.28
iron/operators/relu
Test Krackan Status Krackan Phoenix Status Phoenix Latency (mean)
test_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048] - - 261.90
test_relu[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024] - - 396.02
test_relu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024] - - 349.72
test_relu[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512] - - 383.92
test_relu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512] - - 657.88
test_relu[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256] - - 442.12
iron/operators/repeat
Test Krackan Status Krackan Phoenix Status Phoenix Latency (mean)
test_cols_without_a_legal_split_is_rejected[cols_1031-why_prime > 1023: the only divisors are 1 and cols, neither legal] - - -
test_cols_without_a_legal_split_is_rejected[cols_2062-why_2 x 1031: the only word-aligned chunk leaves a 1031-wide chunk count] - - -
test_cols_without_a_legal_split_is_rejected[cols_513-why_odd: every divisor is odd, so no chunk is a whole 32-bit word] - - -
test_repeat[rows_4-cols_1024-repeat_2-transfer_size_None] - - 366.20
test_repeat[rows_8-cols_512-repeat_4-transfer_size_64] - - 328.70
test_repeat[rows_8-cols_64-repeat_4-transfer_size_None] - - 241.82
iron/operators/rms_norm
Test Krackan Status Krackan Phoenix Status Phoenix Latency (mean)
test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-weighted_False] - - 310.66
test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-weighted_True] - - 413.84
test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-weighted_False] - - 397.32
test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-weighted_True] - - 365.90
test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-weighted_False] - - 414.02
test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-weighted_True] - - 627.98
test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-weighted_False] - - 366.06
test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-weighted_True] - - 345.32
test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-weighted_False] - - 347.38
test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-weighted_True] - - 441.64
test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-weighted_False] - - 443.26
iron/operators/rope
Test Krackan Status Krackan Phoenix Status Phoenix Latency (mean)
test_rope[rows_32-cols_512-angle_rows_32-aie_columns_1-method_type_0] - - 425.82
test_rope[rows_32-cols_512-angle_rows_32-aie_columns_2-method_type_0] - - 415.84
test_rope[rows_32-cols_512-angle_rows_32-aie_columns_4-method_type_0] - - 495.38
test_rope[rows_32-cols_512-angle_rows_8-aie_columns_1-method_type_0] - - 497.30
test_rope[rows_32-cols_512-angle_rows_8-aie_columns_2-method_type_0] - - 392.72
test_rope[rows_32-cols_512-angle_rows_8-aie_columns_4-method_type_0] - - 437.34
iron/operators/sigmoid
Test Krackan Status Krackan Phoenix Status Phoenix Latency (mean)
test_sigmoid[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048] - - 368.06
test_sigmoid[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024] - - 320.84
test_sigmoid[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024] - - 374.02
test_sigmoid[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512] - - 526.92
test_sigmoid[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512] - - 553.14
test_sigmoid[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256] - - 465.62
iron/operators/silu
Test Krackan Status Krackan Phoenix Status Phoenix Latency (mean)
test_silu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048] - - 468.66
test_silu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024] - - 503.32
test_silu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512] - - 353.00
iron/operators/softmax
Test Krackan Status Krackan Phoenix Status Phoenix Latency (mean)
test_softmax[input_length_32768-num_aie_columns_2-num_channels_2-tile_size_1024] - - 489.26
test_softmax[input_length_32768-num_aie_columns_2-num_channels_2-tile_size_2048] - - 658.28
test_softmax[input_length_32768-num_aie_columns_2-num_channels_2-tile_size_512] - - 387.70
iron/operators/strided_copy
Test Krackan Status Krackan Phoenix Status Phoenix Latency (mean)
test_strided_copy[chunked_transfer] - - 354.56
test_strided_copy[contiguous] - - 288.80
test_strided_copy[four_channels] - - 425.00
test_strided_copy[kv_slot0] - - 374.90
test_strided_copy[kv_slot5] - - 282.40
test_strided_copy[kv_slot5_four_channels] - - 435.44
test_strided_copy[kv_slot5_two_channels] - - 312.34
test_strided_copy[kv_slot_last] - - 380.76
test_strided_copy[two_channels] - - 475.14
test_strided_copy[two_channels_chunked] - - 338.86
test_transfer_size_not_dividing_per_channel_share_is_rejected[iter0] - - -
test_transfer_size_not_dividing_per_channel_share_is_rejected[iter1] - - -
test_transfer_size_not_dividing_per_channel_share_is_rejected[iter2] - - -
test_transfer_size_not_dividing_per_channel_share_is_rejected[iter3] - - -
test_transfer_size_not_dividing_per_channel_share_is_rejected[iter4] - - -
iron/operators/swiglu_decode
Test Krackan Status Krackan Phoenix Status Phoenix Latency (mean)
test_swiglu_decode[embedding_dim_1024-hidden_dim_3584] - - 15497.29
test_swiglu_decode[embedding_dim_2048-hidden_dim_2048] - - 12571.46
iron/operators/swiglu_prefill
Test Krackan Status Krackan Phoenix Status Phoenix Latency (mean)
test_swiglu_prefill[seq_len_256-embedding_dim_2048-hidden_dim_2048-prio_accuracy_False] - - 22580.84
iron/operators/tanh
Test Krackan Status Krackan Phoenix Status Phoenix Latency (mean)
test_tanh[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048] - - 351.44
test_tanh[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024] - - 431.04
test_tanh[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024] - - 354.90
test_tanh[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512] - - 705.30
test_tanh[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512] - - 459.22
test_tanh[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256] - - 515.32
iron/operators/transpose
Test Krackan Status Krackan 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] - - 378.92
test_transpose[M_2048-N_64-aie_columns_1-channels_1-m_64-n_64-s_8-num_batches_2] - - 1061.16
test_transpose[M_2048-N_64-aie_columns_1-channels_2-m_64-n_64-s_8-num_batches_1] - - 501.16
Phoenix - Small

IRON

Tested on 2026_09_01_22_41_43 at commit 7e49fd8.

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]✅ 5/5275.800.04n/a
test_axpy[input_length_2048-num_aie_columns_2-tile_size_1024-scalar_factor_3.0]✅ 5/5414.000.03n/a
test_axpy[input_length_2048-num_aie_columns_4-tile_size_512-scalar_factor_3.0]✅ 5/5398.660.03n/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]✅ 5/5352.060.02n/a
test_dequant[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-group_size_32]✅ 5/5454.980.01n/a
test_dequant[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-group_size_32]✅ 5/5353.440.02n/a
test_dequant[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-group_size_32]✅ 5/5389.980.01n/a
test_dequant[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-group_size_32]✅ 5/5462.860.01n/a
test_dequant[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-group_size_32]✅ 5/5503.720.01n/a
iron/operators/elementwise_add
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_elementwise_add[input_length_2048-num_aie_columns_1-tile_size_2048]✅ 5/5381.560.03n/a
test_elementwise_add[input_length_2048-num_aie_columns_2-tile_size_1024]✅ 5/5324.840.04n/a
test_elementwise_add[input_length_2048-num_aie_columns_4-tile_size_512]✅ 5/5514.560.03n/a
iron/operators/elementwise_mul
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_elementwise_mul[input_length_2048-num_aie_columns_1-tile_size_2048]✅ 5/5286.900.04n/a
test_elementwise_mul[input_length_2048-num_aie_columns_2-tile_size_1024]✅ 5/5315.540.04n/a
test_elementwise_mul[input_length_2048-num_aie_columns_4-tile_size_512]✅ 5/5387.720.03n/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]✅ 5/5323.040.03n/a
test_gelu[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024]✅ 5/5706.060.02n/a
test_gelu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024]✅ 5/5473.900.02n/a
test_gelu[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512]✅ 5/5472.340.02n/a
test_gelu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512]✅ 5/51039.560.02n/a
test_gelu[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256]✅ 5/5844.120.02n/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]✅ 5/5626.360.3615.42
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]✅ 5/5482.080.5021.23
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]✅ 5/582496.720.31208.28
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]✅ 5/524774.461.02693.62
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]✅ 5/54036.902.43637.23
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]✅ 5/57006.760.189.84
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/a0.090.09
test_gemv[M_2048-K_8192-num_aie_columns_1-tile_size_input_1-tile_size_output_2048]✅ 5/5n/a3.713.71
test_gemv[M_2048-K_8192-num_aie_columns_2-tile_size_input_1-tile_size_output_1024]✅ 5/5n/a6.706.69
test_gemv[M_2048-K_8192-num_aie_columns_4-tile_size_input_1-tile_size_output_512]✅ 5/5n/a9.699.69
test_gemv[M_8192-K_2048-num_aie_columns_1-tile_size_input_4-tile_size_output_1024]✅ 5/5n/a3.563.55
test_gemv[M_8192-K_2048-num_aie_columns_2-tile_size_input_4-tile_size_output_1024]✅ 5/5n/a6.756.74
test_gemv[M_8192-K_2048-num_aie_columns_4-tile_size_input_4-tile_size_output_1024]✅ 5/5n/a9.179.16
test_gemv_batched[M_1024-K_1024-num_aie_columns_1-tile_size_input_1-tile_size_output_64-num_batches_2]✅ 5/5n/a1.921.91
test_gemv_batched[M_1026-K_64-num_aie_columns_1-tile_size_input_1-tile_size_output_2-num_batches_2]✅ 5/5n/a0.370.36
test_gemv_batched[M_256-K_128-num_aie_columns_1-tile_size_input_1-tile_size_output_256-num_batches_4]✅ 5/5n/a0.460.45
test_gemv_batched[M_64-K_1536-num_aie_columns_1-tile_size_input_1-tile_size_output_64-num_batches_8]✅ 5/5n/a1.821.79
test_gemv_gelu[M_128-K_128-num_aie_columns_1-tile_size_input_32-tile_size_output_128]❌ 0/5n/an/an/a
test_gemv_gelu[M_2048-K_8192-num_aie_columns_1-tile_size_input_1-tile_size_output_2048]❌ 0/5n/an/an/a
test_gemv_gelu[M_8192-K_2048-num_aie_columns_1-tile_size_input_4-tile_size_output_1024]❌ 0/5n/an/an/a
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]✅ 5/5367.040.02n/a
test_layer_norm[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024]✅ 5/5377.060.02n/a
test_layer_norm[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024]✅ 5/5343.720.03n/a
test_layer_norm[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512]✅ 5/5434.520.02n/a
test_layer_norm[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512]✅ 5/5529.180.02n/a
test_layer_norm[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256]✅ 5/5478.520.02n/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]✅ 5/5367.640.02n/a
test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-alpha_0.1]✅ 5/5382.460.03n/a
test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-alpha_0.25]✅ 5/5834.940.02n/a
test_leaky_relu[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-alpha_0.01]✅ 5/5406.940.02n/a
test_leaky_relu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-alpha_0.01]✅ 5/5354.280.03n/a
test_leaky_relu[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-alpha_0.01]✅ 5/5421.140.02n/a
test_leaky_relu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-alpha_0.01]✅ 5/5848.960.02n/a
test_leaky_relu[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-alpha_0.01]✅ 5/5644.660.02n/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]✅ 5/5283.880.03n/a
test_mem_copy[input_length_2048-num_cores_2-num_channels_1-bypass_False-tile_size_1024]✅ 5/5310.020.03n/a
test_mem_copy[input_length_2048-num_cores_2-num_channels_2-bypass_False-tile_size_1024]✅ 5/5289.900.03n/a
test_mem_copy[input_length_2048-num_cores_4-num_channels_1-bypass_False-tile_size_512]✅ 5/5369.080.02n/a
test_mem_copy[input_length_2048-num_cores_4-num_channels_2-bypass_False-tile_size_512]✅ 5/5393.400.02n/a
test_mem_copy[input_length_2048-num_cores_8-num_channels_2-bypass_False-tile_size_256]✅ 5/5498.280.02n/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]✅ 5/5261.900.03n/a
test_relu[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024]✅ 5/5396.020.02n/a
test_relu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024]✅ 5/5349.720.03n/a
test_relu[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512]✅ 5/5383.920.02n/a
test_relu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512]✅ 5/5657.880.02n/a
test_relu[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256]✅ 5/5442.120.02n/a
iron/operators/repeat
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_cols_without_a_legal_split_is_rejected[cols_1031-why_prime > 1023: the only divisors are 1 and cols, neither legal]✅ 5/5n/an/an/a
test_cols_without_a_legal_split_is_rejected[cols_2062-why_2 x 1031: the only word-aligned chunk leaves a 1031-wide chunk count]✅ 5/5n/an/an/a
test_cols_without_a_legal_split_is_rejected[cols_513-why_odd: every divisor is odd, so no chunk is a whole 32-bit word]✅ 5/5n/an/an/a
test_repeat[rows_4-cols_1024-repeat_2-transfer_size_None]✅ 5/5366.200.07n/a
test_repeat[rows_8-cols_512-repeat_4-transfer_size_64]✅ 5/5328.700.14n/a
test_repeat[rows_8-cols_64-repeat_4-transfer_size_None]✅ 5/5241.820.02n/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]✅ 5/5310.660.03n/a
test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_1-tile_size_2048-weighted_True]✅ 5/5413.840.03n/a
test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-weighted_False]✅ 5/5397.320.02n/a
test_rms_norm[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024-weighted_True]✅ 5/5365.900.03n/a
test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-weighted_False]✅ 5/5414.020.02n/a
test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024-weighted_True]✅ 5/5627.980.03n/a
test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-weighted_False]✅ 5/5366.060.02n/a
test_rms_norm[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512-weighted_True]✅ 5/5345.320.03n/a
test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-weighted_False]✅ 5/5347.380.03n/a
test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512-weighted_True]✅ 5/5441.640.02n/a
test_rms_norm[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256-weighted_False]✅ 5/5443.260.02n/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]✅ 5/5425.820.25n/a
test_rope[rows_32-cols_512-angle_rows_32-aie_columns_2-method_type_0]✅ 5/5415.840.26n/a
test_rope[rows_32-cols_512-angle_rows_32-aie_columns_4-method_type_0]✅ 5/5495.380.23n/a
test_rope[rows_32-cols_512-angle_rows_8-aie_columns_1-method_type_0]✅ 5/5497.300.17n/a
test_rope[rows_32-cols_512-angle_rows_8-aie_columns_2-method_type_0]✅ 5/5392.720.21n/a
test_rope[rows_32-cols_512-angle_rows_8-aie_columns_4-method_type_0]✅ 5/5437.340.18n/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]✅ 5/5368.060.02n/a
test_sigmoid[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024]✅ 5/5320.840.03n/a
test_sigmoid[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024]✅ 5/5374.020.02n/a
test_sigmoid[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512]✅ 5/5526.920.02n/a
test_sigmoid[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512]✅ 5/5553.140.02n/a
test_sigmoid[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256]✅ 5/5465.620.02n/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]✅ 5/5468.660.02n/a
test_silu[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024]✅ 5/5503.320.02n/a
test_silu[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512]✅ 5/5353.000.03n/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]✅ 5/5489.260.28n/a
test_softmax[input_length_32768-num_aie_columns_2-num_channels_2-tile_size_2048]✅ 5/5658.280.22n/a
test_softmax[input_length_32768-num_aie_columns_2-num_channels_2-tile_size_512]✅ 5/5387.700.35n/a
iron/operators/strided_copy
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_strided_copy[chunked_transfer]✅ 5/5354.560.01n/a
test_strided_copy[contiguous]✅ 5/5288.800.02n/a
test_strided_copy[four_channels]✅ 5/5425.000.01n/a
test_strided_copy[kv_slot0]✅ 5/5374.900.39n/a
test_strided_copy[kv_slot5]✅ 5/5282.400.49n/a
test_strided_copy[kv_slot5_four_channels]✅ 5/5435.440.35n/a
test_strided_copy[kv_slot5_two_channels]✅ 5/5312.340.45n/a
test_strided_copy[kv_slot_last]✅ 5/5380.760.39n/a
test_strided_copy[two_channels]✅ 5/5475.140.01n/a
test_strided_copy[two_channels_chunked]✅ 5/5338.860.01n/a
test_transfer_size_not_dividing_per_channel_share_is_rejected[iter0]✅ 1/1n/an/an/a
test_transfer_size_not_dividing_per_channel_share_is_rejected[iter1]✅ 1/1n/an/an/a
test_transfer_size_not_dividing_per_channel_share_is_rejected[iter2]✅ 1/1n/an/an/a
test_transfer_size_not_dividing_per_channel_share_is_rejected[iter3]✅ 1/1n/an/an/a
test_transfer_size_not_dividing_per_channel_share_is_rejected[iter4]✅ 1/1n/an/an/a
iron/operators/swiglu_decode
TestChecksLatency (mean)Bandwidth (mean)Throughput (mean)
test_swiglu_decode[embedding_dim_1024-hidden_dim_3584]✅ 5/515497.290.00n/a
test_swiglu_decode[embedding_dim_2048-hidden_dim_2048]✅ 5/512571.460.00n/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]✅ 5/522580.840.10n/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]✅ 5/5351.440.02n/a
test_tanh[input_length_2048-num_aie_columns_1-num_channels_2-tile_size_1024]✅ 5/5431.040.02n/a
test_tanh[input_length_2048-num_aie_columns_2-num_channels_1-tile_size_1024]✅ 5/5354.900.03n/a
test_tanh[input_length_2048-num_aie_columns_2-num_channels_2-tile_size_512]✅ 5/5705.300.02n/a
test_tanh[input_length_2048-num_aie_columns_4-num_channels_1-tile_size_512]✅ 5/5459.220.02n/a
test_tanh[input_length_2048-num_aie_columns_4-num_channels_2-tile_size_256]✅ 5/5515.320.02n/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]✅ 5/5378.921.45n/a
test_transpose[M_2048-N_64-aie_columns_1-channels_1-m_64-n_64-s_8-num_batches_2]✅ 5/51061.161.86n/a
test_transpose[M_2048-N_64-aie_columns_1-channels_2-m_64-n_64-s_8-num_batches_1]✅ 5/5501.161.08n/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)
7e49fd8 — 2026-09-01 22:36:560.05 (+13.48%)0.04 (+25.33%)0.04 (+15.43%)0.04 (+82.39%)0.00 (-59.98%)294.60 (-45.17%)275.80 (-24.52%)276.30 (-13.36%)242.70 (-11.91%)20.88 (-80.69%)
a279467 — 2026-08-31 16:07:160.04 (n/a)0.04 (n/a)0.04 (n/a)0.02 (n/a)0.01 (n/a)537.30 (n/a)365.40 (n/a)318.90 (n/a)275.50 (n/a)108.14 (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)
7e49fd8 — 2026-09-01 22:36:560.06 (+31.25%)0.03 (+15.05%)0.02 (-12.52%)0.02 (+11.64%)0.02 (+66.88%)537.20 (-10.42%)414.00 (-6.97%)516.50 (+14.32%)221.70 (-23.81%)153.72 (+19.29%)
a279467 — 2026-08-31 16:07:160.04 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.01 (n/a)599.70 (n/a)445.02 (n/a)451.80 (n/a)291.00 (n/a)128.86 (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)
7e49fd8 — 2026-09-01 22:36:560.06 (-10.96%)0.03 (-16.17%)0.03 (-28.68%)0.02 (+7.49%)0.01 (-30.08%)574.60 (-6.98%)398.66 (+8.00%)430.50 (+40.23%)218.90 (+12.31%)134.93 (-28.34%)
a279467 — 2026-08-31 16:07:160.06 (n/a)0.04 (n/a)0.04 (n/a)0.02 (n/a)0.02 (n/a)617.70 (n/a)369.14 (n/a)307.00 (n/a)194.90 (n/a)188.30 (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)
7e49fd8 — 2026-09-01 22:36:560.02 (+8.01%)0.02 (+3.19%)0.02 (-9.41%)0.01 (+28.42%)0.00 (-16.86%)525.30 (-22.13%)352.06 (-9.23%)295.00 (+10.40%)243.50 (-7.41%)118.62 (-35.35%)
a279467 — 2026-08-31 16:07:160.02 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)674.60 (n/a)387.84 (n/a)267.20 (n/a)263.00 (n/a)183.50 (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)
7e49fd8 — 2026-09-01 22:36:560.02 (+1.11%)0.01 (-0.19%)0.01 (-12.65%)0.01 (-3.44%)0.00 (+14.32%)574.10 (+3.57%)454.98 (+2.16%)483.80 (+14.48%)287.40 (-1.07%)128.39 (+19.28%)
a279467 — 2026-08-31 16:07:160.02 (n/a)0.01 (n/a)0.01 (n/a)0.01 (n/a)0.00 (n/a)554.30 (n/a)445.34 (n/a)422.60 (n/a)290.50 (n/a)107.64 (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)
7e49fd8 — 2026-09-01 22:36:560.02 (-12.70%)0.02 (+3.52%)0.02 (-2.99%)0.01 (+410.01%)0.00 (-48.71%)484.80 (-80.39%)353.44 (-51.63%)293.80 (+3.09%)264.50 (+14.55%)103.44 (-89.40%)
a279467 — 2026-08-31 16:07:160.02 (n/a)0.02 (n/a)0.02 (n/a)0.00 (n/a)0.01 (n/a)2472.60 (n/a)730.70 (n/a)285.00 (n/a)230.90 (n/a)976.04 (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)
7e49fd8 — 2026-09-01 22:36:560.02 (-0.67%)0.01 (-3.01%)0.01 (+7.57%)0.01 (-7.43%)0.00 (-11.55%)564.50 (+8.02%)389.98 (+1.66%)350.00 (-7.04%)248.90 (+0.69%)134.43 (-1.67%)
a279467 — 2026-08-31 16:07:160.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)0.01 (n/a)522.60 (n/a)383.62 (n/a)376.50 (n/a)247.20 (n/a)136.71 (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)
7e49fd8 — 2026-09-01 22:36:560.02 (-5.91%)0.01 (-7.48%)0.01 (-10.18%)0.01 (+3.38%)0.01 (-5.91%)609.30 (-3.29%)462.86 (+7.27%)497.70 (+11.34%)223.90 (+6.32%)143.06 (-3.97%)
a279467 — 2026-08-31 16:07:160.02 (n/a)0.01 (n/a)0.01 (n/a)0.01 (n/a)0.01 (n/a)630.00 (n/a)431.48 (n/a)447.00 (n/a)210.60 (n/a)148.98 (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)
7e49fd8 — 2026-09-01 22:36:560.02 (-21.00%)0.01 (-17.51%)0.01 (-15.49%)0.01 (-17.91%)0.00 (-26.03%)595.50 (+21.80%)503.72 (+19.93%)554.50 (+18.33%)309.90 (+26.54%)114.13 (+13.41%)
a279467 — 2026-08-31 16:07:160.02 (n/a)0.01 (n/a)0.01 (n/a)0.01 (n/a)0.00 (n/a)488.90 (n/a)420.02 (n/a)468.60 (n/a)244.90 (n/a)100.64 (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)
7e49fd8 — 2026-09-01 22:36:560.05 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.01 (n/a)516.50 (n/a)381.56 (n/a)419.00 (n/a)241.50 (n/a)114.76 (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)
7e49fd8 — 2026-09-01 22:36:560.05 (n/a)0.04 (n/a)0.04 (n/a)0.03 (n/a)0.01 (n/a)425.40 (n/a)324.84 (n/a)288.30 (n/a)243.30 (n/a)79.28 (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)
7e49fd8 — 2026-09-01 22:36:560.04 (n/a)0.03 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)719.10 (n/a)514.56 (n/a)520.80 (n/a)279.30 (n/a)156.20 (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)
7e49fd8 — 2026-09-01 22:36:560.05 (n/a)0.04 (n/a)0.05 (n/a)0.03 (n/a)0.01 (n/a)370.50 (n/a)286.90 (n/a)272.20 (n/a)241.10 (n/a)49.36 (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)
7e49fd8 — 2026-09-01 22:36:560.05 (n/a)0.04 (n/a)0.04 (n/a)0.03 (n/a)0.01 (n/a)463.70 (n/a)315.54 (n/a)289.10 (n/a)232.00 (n/a)95.89 (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)
7e49fd8 — 2026-09-01 22:36:560.05 (n/a)0.03 (n/a)0.04 (n/a)0.02 (n/a)0.01 (n/a)603.70 (n/a)387.72 (n/a)325.30 (n/a)260.80 (n/a)141.59 (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)
7e49fd8 — 2026-09-01 22:36:560.04 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.01 (n/a)470.60 (n/a)323.04 (n/a)278.60 (n/a)196.70 (n/a)118.30 (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)
7e49fd8 — 2026-09-01 22:36:560.02 (n/a)0.02 (n/a)0.02 (n/a)0.00 (n/a)0.01 (n/a)1873.40 (n/a)706.06 (n/a)361.30 (n/a)344.70 (n/a)660.87 (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)
7e49fd8 — 2026-09-01 22:36:560.04 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)1017.30 (n/a)473.90 (n/a)332.60 (n/a)194.90 (n/a)326.82 (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)
7e49fd8 — 2026-09-01 22:36:560.02 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.00 (n/a)686.80 (n/a)472.34 (n/a)404.00 (n/a)375.20 (n/a)133.57 (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)
7e49fd8 — 2026-09-01 22:36:560.03 (n/a)0.02 (n/a)0.02 (n/a)0.00 (n/a)0.01 (n/a)2389.90 (n/a)1039.56 (n/a)369.50 (n/a)258.00 (n/a)1011.91 (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)
7e49fd8 — 2026-09-01 22:36:560.03 (n/a)0.02 (n/a)0.02 (n/a)0.00 (n/a)0.01 (n/a)2394.50 (n/a)844.12 (n/a)528.60 (n/a)235.90 (n/a)877.42 (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)
7e49fd8 — 2026-09-01 22:36:560.44 (-36.46%)0.36 (-8.22%)0.36 (+8.69%)0.29 (-6.36%)0.06 (-62.90%)776.00 (+6.78%)626.36 (+1.48%)621.30 (-8.00%)506.80 (+57.39%)107.79 (-35.48%)18.62 (-36.46%)15.42 (-8.22%)15.19 (+8.69%)12.16 (-6.36%)2.60 (-62.90%)
a279467 — 2026-08-31 16:07:160.69 (n/a)0.39 (n/a)0.33 (n/a)0.30 (n/a)0.16 (n/a)726.70 (n/a)617.20 (n/a)675.30 (n/a)322.00 (n/a)167.08 (n/a)29.31 (n/a)16.80 (n/a)13.97 (n/a)12.99 (n/a)7.01 (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)
7e49fd8 — 2026-09-01 22:36:560.70 (+31.88%)0.50 (+8.09%)0.41 (-11.26%)0.34 (-1.69%)0.16 (+112.70%)654.90 (+1.71%)482.08 (-2.14%)534.30 (+12.67%)315.70 (-24.17%)145.81 (+58.03%)29.89 (+31.88%)21.23 (+8.09%)17.66 (-11.26%)14.41 (-1.69%)6.90 (+112.70%)
a279467 — 2026-08-31 16:07:160.53 (n/a)0.46 (n/a)0.47 (n/a)0.34 (n/a)0.08 (n/a)643.90 (n/a)492.64 (n/a)474.20 (n/a)416.30 (n/a)92.27 (n/a)22.67 (n/a)19.64 (n/a)19.90 (n/a)14.66 (n/a)3.25 (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)
7e49fd8 — 2026-09-01 22:36:560.31 (-0.20%)0.31 (+0.69%)0.30 (+0.20%)0.30 (+2.22%)0.00 (-41.87%)83712.30 (-2.17%)82496.72 (-0.71%)82700.00 (-0.20%)81068.70 (+0.20%)1036.00 (-43.08%)211.92 (-0.20%)208.28 (+0.69%)207.74 (+0.20%)205.23 (+2.22%)2.62 (-41.87%)
a279467 — 2026-08-31 16:07:160.31 (n/a)0.30 (n/a)0.30 (n/a)0.29 (n/a)0.01 (n/a)85569.90 (n/a)83085.36 (n/a)82863.00 (n/a)80906.70 (n/a)1820.16 (n/a)212.34 (n/a)206.85 (n/a)207.33 (n/a)200.77 (n/a)4.51 (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)
7e49fd8 — 2026-09-01 22:36:561.03 (-0.91%)1.02 (-0.04%)1.02 (+0.42%)0.99 (-0.47%)0.02 (-3.94%)25499.70 (+0.47%)24774.46 (+0.04%)24589.50 (-0.41%)24398.20 (+0.92%)438.17 (-2.46%)704.15 (-0.91%)693.62 (-0.04%)698.67 (+0.42%)673.73 (-0.47%)12.08 (-3.94%)
a279467 — 2026-08-31 16:07:161.04 (n/a)1.02 (n/a)1.02 (n/a)0.99 (n/a)0.02 (n/a)25380.20 (n/a)24764.18 (n/a)24691.80 (n/a)24176.40 (n/a)449.22 (n/a)710.61 (n/a)693.92 (n/a)695.77 (n/a)676.90 (n/a)12.57 (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)
7e49fd8 — 2026-09-01 22:36:563.90 (+4.30%)2.43 (-2.43%)2.08 (-1.98%)1.13 (-14.74%)1.14 (+12.49%)7156.10 (+17.28%)4036.90 (+8.11%)3872.80 (+2.02%)2064.90 (-4.12%)2035.68 (+28.10%)1023.73 (+4.30%)637.23 (-2.43%)545.85 (-1.98%)295.40 (-14.74%)300.27 (+12.49%)
a279467 — 2026-08-31 16:07:163.74 (n/a)2.49 (n/a)2.12 (n/a)1.32 (n/a)1.02 (n/a)6101.60 (n/a)3734.00 (n/a)3796.00 (n/a)2153.60 (n/a)1589.10 (n/a)981.56 (n/a)653.07 (n/a)556.88 (n/a)346.46 (n/a)266.92 (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)
7e49fd8 — 2026-09-01 22:36:560.23 (+20.10%)0.18 (-0.41%)0.18 (-2.85%)0.14 (-17.78%)0.03 (+220.10%)8922.80 (+21.63%)7006.76 (+2.87%)6922.50 (+2.94%)5315.40 (-16.73%)1282.66 (+222.82%)12.63 (+20.10%)9.84 (-0.41%)9.69 (-2.85%)7.52 (-17.78%)1.82 (+220.10%)
a279467 — 2026-08-31 16:07:160.20 (n/a)0.18 (n/a)0.19 (n/a)0.17 (n/a)0.01 (n/a)7336.20 (n/a)6811.00 (n/a)6724.90 (n/a)6383.60 (n/a)397.33 (n/a)10.51 (n/a)9.88 (n/a)9.98 (n/a)9.15 (n/a)0.57 (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)
7e49fd8 — 2026-09-01 22:36:560.12 (n/a)0.09 (n/a)0.09 (n/a)0.07 (n/a)0.02 (n/a)0.12 (n/a)0.09 (n/a)0.08 (n/a)0.07 (n/a)0.02 (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)
7e49fd8 — 2026-09-01 22:36:563.93 (n/a)3.71 (n/a)3.75 (n/a)3.45 (n/a)0.18 (n/a)3.93 (n/a)3.71 (n/a)3.75 (n/a)3.45 (n/a)0.18 (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)
7e49fd8 — 2026-09-01 22:36:567.50 (-0.90%)6.70 (+5.42%)7.12 (+19.39%)5.78 (+1.64%)0.82 (+1.57%)7.50 (-0.90%)6.69 (+5.42%)7.12 (+19.39%)5.78 (+1.64%)0.82 (+1.57%)
a279467 — 2026-08-31 16:07:167.57 (n/a)6.35 (n/a)5.97 (n/a)5.69 (n/a)0.81 (n/a)7.57 (n/a)6.35 (n/a)5.96 (n/a)5.69 (n/a)0.81 (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)
7e49fd8 — 2026-09-01 22:36:5613.22 (+22.03%)9.69 (+8.84%)9.65 (+12.10%)7.45 (+0.45%)2.21 (+74.16%)13.21 (+22.03%)9.69 (+8.84%)9.64 (+12.10%)7.45 (+0.45%)2.21 (+74.16%)
a279467 — 2026-08-31 16:07:1610.84 (n/a)8.91 (n/a)8.61 (n/a)7.42 (n/a)1.27 (n/a)10.83 (n/a)8.90 (n/a)8.60 (n/a)7.42 (n/a)1.27 (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)
7e49fd8 — 2026-09-01 22:36:563.72 (n/a)3.56 (n/a)3.60 (n/a)3.38 (n/a)0.15 (n/a)3.72 (n/a)3.55 (n/a)3.60 (n/a)3.37 (n/a)0.15 (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)
7e49fd8 — 2026-09-01 22:36:567.36 (-4.38%)6.75 (+8.20%)7.11 (+18.36%)5.00 (+3.76%)0.99 (-8.87%)7.36 (-4.38%)6.74 (+8.20%)7.10 (+18.36%)4.99 (+3.76%)0.99 (-8.87%)
a279467 — 2026-08-31 16:07:167.70 (n/a)6.24 (n/a)6.00 (n/a)4.81 (n/a)1.09 (n/a)7.69 (n/a)6.23 (n/a)6.00 (n/a)4.81 (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)
7e49fd8 — 2026-09-01 22:36:5613.38 (-4.37%)9.17 (-8.55%)8.15 (-13.99%)7.72 (-2.50%)2.37 (+2.69%)13.37 (-4.37%)9.16 (-8.55%)8.14 (-13.99%)7.72 (-2.50%)2.37 (+2.69%)
a279467 — 2026-08-31 16:07:1613.99 (n/a)10.03 (n/a)9.47 (n/a)7.92 (n/a)2.31 (n/a)13.98 (n/a)10.02 (n/a)9.47 (n/a)7.92 (n/a)2.31 (n/a)

test_gemv_batched[M_1024-K_1024-num_aie_columns_1-tile_size_input_1-tile_size_output_64-num_batches_2]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
7e49fd8 — 2026-09-01 22:36:562.82 (-4.63%)1.92 (-3.45%)1.54 (-31.12%)1.00 (-4.52%)0.83 (-4.82%)2.82 (-4.63%)1.91 (-3.45%)1.54 (-31.12%)1.00 (-4.52%)0.83 (-4.82%)
a279467 — 2026-08-31 16:07:162.96 (n/a)1.99 (n/a)2.23 (n/a)1.05 (n/a)0.87 (n/a)2.95 (n/a)1.98 (n/a)2.23 (n/a)1.05 (n/a)0.87 (n/a)

test_gemv_batched[M_1026-K_64-num_aie_columns_1-tile_size_input_1-tile_size_output_2-num_batches_2]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
7e49fd8 — 2026-09-01 22:36:560.51 (-1.47%)0.37 (+18.21%)0.39 (+3.55%)0.14 (+24.47%)0.14 (-18.27%)0.51 (-1.47%)0.36 (+18.21%)0.39 (+3.55%)0.13 (+24.47%)0.14 (-18.27%)
a279467 — 2026-08-31 16:07:160.52 (n/a)0.31 (n/a)0.38 (n/a)0.11 (n/a)0.17 (n/a)0.51 (n/a)0.31 (n/a)0.37 (n/a)0.11 (n/a)0.17 (n/a)

test_gemv_batched[M_256-K_128-num_aie_columns_1-tile_size_input_1-tile_size_output_256-num_batches_4]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
7e49fd8 — 2026-09-01 22:36:560.62 (+0.27%)0.46 (+19.85%)0.42 (-7.00%)0.40 (+215.98%)0.09 (-60.39%)0.61 (+0.27%)0.45 (+19.85%)0.41 (-7.00%)0.40 (+215.98%)0.09 (-60.39%)
a279467 — 2026-08-31 16:07:160.62 (n/a)0.38 (n/a)0.45 (n/a)0.13 (n/a)0.23 (n/a)0.61 (n/a)0.38 (n/a)0.44 (n/a)0.13 (n/a)0.23 (n/a)

test_gemv_batched[M_64-K_1536-num_aie_columns_1-tile_size_input_1-tile_size_output_64-num_batches_8]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Throughput (max)Throughput (mean)Throughput (median)Throughput (min)Throughput (stddev)
7e49fd8 — 2026-09-01 22:36:562.75 (+14.90%)1.82 (-1.84%)1.79 (-23.19%)1.38 (+209.51%)0.56 (-33.06%)2.70 (+14.90%)1.79 (-1.84%)1.77 (-23.19%)1.36 (+209.51%)0.55 (-33.06%)
a279467 — 2026-08-31 16:07:162.39 (n/a)1.86 (n/a)2.34 (n/a)0.45 (n/a)0.83 (n/a)2.35 (n/a)1.83 (n/a)2.30 (n/a)0.44 (n/a)0.82 (n/a)

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

No metrics available.

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

No metrics available.

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

No metrics available.

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)
7e49fd8 — 2026-09-01 22:36:560.03 (n/a)0.02 (n/a)0.03 (n/a)0.01 (n/a)0.01 (n/a)546.10 (n/a)367.04 (n/a)290.60 (n/a)272.60 (n/a)123.99 (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)
7e49fd8 — 2026-09-01 22:36:560.03 (n/a)0.02 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)496.50 (n/a)377.06 (n/a)408.30 (n/a)240.20 (n/a)119.09 (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)
7e49fd8 — 2026-09-01 22:36:560.03 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.01 (n/a)543.30 (n/a)343.72 (n/a)298.00 (n/a)260.10 (n/a)115.02 (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)
7e49fd8 — 2026-09-01 22:36:560.03 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)614.70 (n/a)434.52 (n/a)374.00 (n/a)285.80 (n/a)155.99 (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)
7e49fd8 — 2026-09-01 22:36:560.03 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)1054.30 (n/a)529.18 (n/a)487.60 (n/a)242.70 (n/a)326.77 (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)
7e49fd8 — 2026-09-01 22:36:560.02 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.00 (n/a)633.60 (n/a)478.52 (n/a)446.90 (n/a)335.20 (n/a)114.22 (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)
7e49fd8 — 2026-09-01 22:36:560.03 (-0.26%)0.02 (+4.06%)0.03 (+32.37%)0.02 (-4.11%)0.01 (-6.02%)544.60 (+4.29%)367.64 (-4.25%)311.80 (-24.47%)248.20 (+0.24%)120.48 (+2.60%)
a279467 — 2026-08-31 16:07:160.03 (n/a)0.02 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)522.20 (n/a)383.94 (n/a)412.80 (n/a)247.60 (n/a)117.43 (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)
7e49fd8 — 2026-09-01 22:36:560.04 (+13.55%)0.03 (-3.01%)0.03 (+13.71%)0.01 (-25.86%)0.01 (+86.35%)638.90 (+34.87%)382.46 (+16.79%)269.80 (-12.06%)219.20 (-11.93%)184.22 (+115.29%)
a279467 — 2026-08-31 16:07:160.03 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.01 (n/a)473.70 (n/a)327.48 (n/a)306.80 (n/a)248.90 (n/a)85.57 (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)
7e49fd8 — 2026-09-01 22:36:560.03 (-27.29%)0.02 (-30.35%)0.02 (-12.81%)0.00 (-79.75%)0.01 (+3.27%)2484.30 (+393.80%)834.94 (+126.97%)436.50 (+14.69%)302.40 (+37.52%)927.34 (+689.49%)
a279467 — 2026-08-31 16:07:160.04 (n/a)0.02 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)503.10 (n/a)367.86 (n/a)380.60 (n/a)219.90 (n/a)117.46 (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)
7e49fd8 — 2026-09-01 22:36:560.03 (-29.55%)0.02 (-0.95%)0.02 (+29.05%)0.01 (+32.71%)0.01 (-52.49%)588.40 (-24.65%)406.94 (-16.01%)411.20 (-22.52%)252.70 (+41.97%)121.46 (-47.24%)
a279467 — 2026-08-31 16:07:160.05 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)780.90 (n/a)484.52 (n/a)530.70 (n/a)178.00 (n/a)230.19 (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)
7e49fd8 — 2026-09-01 22:36:560.04 (-17.68%)0.03 (-3.88%)0.02 (-20.95%)0.02 (+3.22%)0.01 (-24.28%)531.90 (-3.11%)354.28 (-2.50%)385.60 (+26.47%)223.70 (+21.51%)127.05 (-21.93%)
a279467 — 2026-08-31 16:07:160.04 (n/a)0.03 (n/a)0.03 (n/a)0.01 (n/a)0.01 (n/a)549.00 (n/a)363.36 (n/a)304.90 (n/a)184.10 (n/a)162.75 (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)
7e49fd8 — 2026-09-01 22:36:560.03 (+38.37%)0.02 (+18.74%)0.02 (+2.11%)0.02 (+1.62%)0.01 (+158.20%)532.70 (-1.59%)421.14 (-10.11%)466.20 (-2.06%)268.20 (-27.73%)126.04 (+89.34%)
a279467 — 2026-08-31 16:07:160.02 (n/a)0.02 (n/a)0.02 (n/a)0.02 (n/a)0.00 (n/a)541.30 (n/a)468.48 (n/a)476.00 (n/a)371.10 (n/a)66.57 (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)
7e49fd8 — 2026-09-01 22:36:560.03 (-13.65%)0.02 (-27.98%)0.02 (-9.12%)0.00 (-77.19%)0.01 (+1.00%)2445.60 (+338.44%)848.96 (+106.99%)483.40 (+10.01%)282.70 (+15.81%)898.76 (+498.26%)
a279467 — 2026-08-31 16:07:160.03 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)557.80 (n/a)410.14 (n/a)439.40 (n/a)244.10 (n/a)150.23 (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)
7e49fd8 — 2026-09-01 22:36:560.02 (-27.34%)0.02 (-23.10%)0.02 (-13.75%)0.01 (-54.90%)0.01 (-6.16%)1302.60 (+121.72%)644.66 (+45.93%)467.90 (+15.93%)381.20 (+37.62%)383.18 (+174.65%)
a279467 — 2026-08-31 16:07:160.03 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)587.50 (n/a)441.76 (n/a)403.60 (n/a)277.00 (n/a)139.52 (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)
7e49fd8 — 2026-09-01 22:36:560.03 (+6.80%)0.03 (+25.07%)0.03 (+14.30%)0.02 (+80.66%)0.00 (-50.22%)345.00 (-44.65%)283.88 (-27.70%)260.40 (-12.50%)247.90 (-6.38%)41.75 (-73.83%)
a279467 — 2026-08-31 16:07:160.03 (n/a)0.02 (n/a)0.03 (n/a)0.01 (n/a)0.01 (n/a)623.30 (n/a)392.66 (n/a)297.60 (n/a)264.80 (n/a)159.50 (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)
7e49fd8 — 2026-09-01 22:36:560.04 (+30.48%)0.03 (+55.27%)0.03 (+74.55%)0.02 (+92.02%)0.01 (+8.36%)398.20 (-47.93%)310.02 (-38.79%)292.10 (-42.71%)203.00 (-23.37%)84.63 (-52.26%)
a279467 — 2026-08-31 16:07:160.03 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)764.70 (n/a)506.50 (n/a)509.90 (n/a)264.90 (n/a)177.28 (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)
7e49fd8 — 2026-09-01 22:36:560.03 (-3.01%)0.03 (+21.62%)0.03 (+10.73%)0.02 (+78.43%)0.00 (-62.55%)341.20 (-43.95%)289.90 (-25.88%)278.30 (-9.67%)260.20 (+3.09%)32.99 (-78.61%)
a279467 — 2026-08-31 16:07:160.03 (n/a)0.02 (n/a)0.03 (n/a)0.01 (n/a)0.01 (n/a)608.70 (n/a)391.10 (n/a)308.10 (n/a)252.40 (n/a)154.22 (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)
7e49fd8 — 2026-09-01 22:36:560.03 (-4.10%)0.02 (+5.25%)0.02 (+38.31%)0.02 (+34.85%)0.01 (-36.36%)467.90 (-25.85%)369.08 (-14.66%)366.40 (-27.69%)241.70 (+4.27%)92.90 (-47.82%)
a279467 — 2026-08-31 16:07:160.04 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)631.00 (n/a)432.48 (n/a)506.70 (n/a)231.80 (n/a)178.04 (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)
7e49fd8 — 2026-09-01 22:36:560.03 (-10.58%)0.02 (-14.23%)0.02 (-35.88%)0.02 (+2.75%)0.01 (-29.39%)515.30 (-2.68%)393.40 (+9.36%)444.70 (+55.98%)252.20 (+11.84%)109.64 (-27.51%)
a279467 — 2026-08-31 16:07:160.04 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.01 (n/a)529.50 (n/a)359.72 (n/a)285.10 (n/a)225.50 (n/a)151.25 (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)
7e49fd8 — 2026-09-01 22:36:560.03 (-15.69%)0.02 (-8.21%)0.02 (+30.28%)0.01 (-11.47%)0.01 (-27.16%)767.30 (+12.95%)498.28 (+4.62%)434.80 (-23.23%)296.70 (+18.63%)190.11 (+1.06%)
a279467 — 2026-08-31 16:07:160.03 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)0.01 (n/a)679.30 (n/a)476.28 (n/a)566.40 (n/a)250.10 (n/a)188.11 (n/a)
iron/operators/repeat

test_cols_without_a_legal_split_is_rejected[cols_1031-why_prime > 1023: the only divisors are 1 and cols, neither legal]

No metrics available.

test_cols_without_a_legal_split_is_rejected[cols_2062-why_2 x 1031: the only word-aligned chunk leaves a 1031-wide chunk count]

No metrics available.

test_cols_without_a_legal_split_is_rejected[cols_513-why_odd: every divisor is odd, so no chunk is a whole 32-bit word]

No metrics available.

test_repeat[rows_4-cols_1024-repeat_2-transfer_size_None]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
7e49fd8 — 2026-09-01 22:36:560.11 (+15.03%)0.07 (-19.79%)0.06 (-32.96%)0.05 (-41.83%)0.02 (+1680.29%)470.70 (+71.91%)366.20 (+35.89%)400.70 (+49.18%)229.10 (-13.09%)111.91 (+2622.93%)
a279467 — 2026-08-31 16:07:160.09 (n/a)0.09 (n/a)0.09 (n/a)0.09 (n/a)0.00 (n/a)273.80 (n/a)269.48 (n/a)268.60 (n/a)263.60 (n/a)4.11 (n/a)

test_repeat[rows_8-cols_512-repeat_4-transfer_size_64]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
7e49fd8 — 2026-09-01 22:36:560.21 (+22.42%)0.14 (+16.04%)0.15 (+49.14%)0.08 (+3.40%)0.05 (+29.19%)484.40 (-3.29%)328.70 (-10.64%)271.60 (-32.95%)197.60 (-18.31%)129.66 (+13.00%)
a279467 — 2026-08-31 16:07:160.17 (n/a)0.12 (n/a)0.10 (n/a)0.08 (n/a)0.04 (n/a)500.90 (n/a)367.84 (n/a)405.10 (n/a)241.90 (n/a)114.74 (n/a)

test_repeat[rows_8-cols_64-repeat_4-transfer_size_None]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
7e49fd8 — 2026-09-01 22:36:560.02 (+23.68%)0.02 (+48.57%)0.02 (+42.28%)0.02 (+96.02%)0.00 (-59.72%)264.60 (-48.99%)241.82 (-36.24%)234.70 (-29.71%)225.10 (-19.15%)17.25 (-83.57%)
a279467 — 2026-08-31 16:07:160.02 (n/a)0.01 (n/a)0.02 (n/a)0.01 (n/a)0.00 (n/a)518.70 (n/a)379.24 (n/a)333.90 (n/a)278.40 (n/a)104.98 (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)
7e49fd8 — 2026-09-01 22:36:560.04 (+4.08%)0.03 (-4.04%)0.03 (-13.79%)0.02 (+21.49%)0.01 (-5.73%)454.30 (-17.70%)310.66 (+0.54%)293.10 (+15.99%)219.10 (-3.90%)95.36 (-30.25%)
a279467 — 2026-08-31 16:07:160.04 (n/a)0.03 (n/a)0.03 (n/a)0.01 (n/a)0.01 (n/a)552.00 (n/a)308.98 (n/a)252.70 (n/a)228.00 (n/a)136.71 (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)
7e49fd8 — 2026-09-01 22:36:560.05 (-15.40%)0.03 (-14.04%)0.02 (-22.81%)0.02 (-9.08%)0.01 (-6.95%)543.00 (+9.99%)413.84 (+18.43%)504.50 (+29.56%)242.40 (+18.24%)155.60 (+22.78%)
a279467 — 2026-08-31 16:07:160.06 (n/a)0.04 (n/a)0.03 (n/a)0.02 (n/a)0.02 (n/a)493.70 (n/a)349.44 (n/a)389.40 (n/a)205.00 (n/a)126.73 (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)
7e49fd8 — 2026-09-01 22:36:560.03 (-6.48%)0.02 (-12.99%)0.02 (-35.82%)0.01 (-4.71%)0.01 (-0.06%)554.10 (+4.94%)397.32 (+16.15%)457.70 (+55.84%)235.00 (+6.92%)145.46 (+8.45%)
a279467 — 2026-08-31 16:07:160.04 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.01 (n/a)528.00 (n/a)342.08 (n/a)293.70 (n/a)219.80 (n/a)134.12 (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)
7e49fd8 — 2026-09-01 22:36:560.06 (+39.53%)0.03 (+10.57%)0.03 (-29.59%)0.02 (+77.67%)0.02 (+12.24%)567.40 (-43.71%)365.90 (-21.27%)383.70 (+42.06%)168.50 (-28.33%)160.84 (-52.08%)
a279467 — 2026-08-31 16:07:160.04 (n/a)0.03 (n/a)0.04 (n/a)0.01 (n/a)0.02 (n/a)1008.00 (n/a)464.78 (n/a)270.10 (n/a)235.10 (n/a)335.65 (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)
7e49fd8 — 2026-09-01 22:36:560.04 (-2.10%)0.02 (-12.47%)0.02 (-34.67%)0.01 (+319.94%)0.01 (-38.02%)588.50 (-76.19%)414.02 (-42.99%)413.70 (+53.05%)222.00 (+2.12%)130.37 (-86.69%)
a279467 — 2026-08-31 16:07:160.04 (n/a)0.03 (n/a)0.03 (n/a)0.00 (n/a)0.01 (n/a)2471.30 (n/a)726.16 (n/a)270.30 (n/a)217.40 (n/a)979.47 (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)
7e49fd8 — 2026-09-01 22:36:560.06 (+50.57%)0.03 (+16.54%)0.04 (+73.15%)0.01 (-72.33%)0.02 (+122.36%)1916.40 (+261.45%)627.98 (+61.32%)254.20 (-42.24%)164.10 (-33.59%)735.71 (+496.68%)
a279467 — 2026-08-31 16:07:160.04 (n/a)0.03 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)530.20 (n/a)389.28 (n/a)440.10 (n/a)247.10 (n/a)123.30 (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)
7e49fd8 — 2026-09-01 22:36:560.03 (-6.61%)0.02 (+30.77%)0.03 (+59.12%)0.02 (+247.78%)0.01 (-34.17%)517.70 (-71.25%)366.06 (-46.85%)308.80 (-37.16%)266.70 (+7.11%)115.62 (-81.72%)
a279467 — 2026-08-31 16:07:160.03 (n/a)0.02 (n/a)0.02 (n/a)0.00 (n/a)0.01 (n/a)1800.50 (n/a)688.68 (n/a)491.40 (n/a)249.00 (n/a)632.45 (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)
7e49fd8 — 2026-09-01 22:36:560.04 (-3.21%)0.03 (+16.38%)0.03 (+17.14%)0.02 (+25.29%)0.01 (-22.34%)446.40 (-20.19%)345.32 (-17.50%)356.80 (-14.64%)241.20 (+3.34%)81.03 (-33.33%)
a279467 — 2026-08-31 16:07:160.04 (n/a)0.02 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)559.30 (n/a)418.58 (n/a)418.00 (n/a)233.40 (n/a)121.54 (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)
7e49fd8 — 2026-09-01 22:36:560.03 (+22.23%)0.03 (+45.99%)0.03 (+50.72%)0.02 (+88.55%)0.01 (+4.39%)470.60 (-46.97%)347.38 (-35.51%)318.20 (-33.65%)258.60 (-18.19%)95.72 (-56.29%)
a279467 — 2026-08-31 16:07:160.03 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)887.40 (n/a)538.62 (n/a)479.60 (n/a)316.10 (n/a)218.98 (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)
7e49fd8 — 2026-09-01 22:36:560.03 (-25.19%)0.02 (-9.87%)0.02 (+2.51%)0.01 (-15.58%)0.01 (-38.02%)625.90 (+18.45%)441.64 (+8.37%)413.90 (-2.45%)317.60 (+33.67%)114.38 (+7.89%)
a279467 — 2026-08-31 16:07:160.04 (n/a)0.02 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)528.40 (n/a)407.52 (n/a)424.30 (n/a)237.60 (n/a)106.01 (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)
7e49fd8 — 2026-09-01 22:36:560.03 (-20.65%)0.02 (-12.35%)0.02 (-33.33%)0.02 (+206.12%)0.01 (-57.35%)539.90 (-67.34%)443.26 (-27.60%)433.60 (+49.98%)289.50 (+26.03%)102.78 (-82.86%)
a279467 — 2026-08-31 16:07:160.04 (n/a)0.02 (n/a)0.03 (n/a)0.00 (n/a)0.01 (n/a)1652.90 (n/a)612.22 (n/a)289.10 (n/a)229.70 (n/a)599.70 (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)
7e49fd8 — 2026-09-01 22:36:560.37 (+6.03%)0.25 (+2.23%)0.27 (+41.43%)0.16 (-11.07%)0.08 (-3.96%)613.80 (+12.46%)425.82 (-2.50%)369.50 (-29.30%)264.00 (-5.68%)140.23 (+3.24%)
a279467 — 2026-08-31 16:07:160.35 (n/a)0.25 (n/a)0.19 (n/a)0.18 (n/a)0.09 (n/a)545.80 (n/a)436.72 (n/a)522.60 (n/a)279.90 (n/a)135.83 (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)
7e49fd8 — 2026-09-01 22:36:560.40 (+21.27%)0.26 (+60.25%)0.20 (+16.97%)0.18 (+338.94%)0.10 (-18.76%)551.80 (-77.22%)415.84 (-62.45%)487.00 (-14.50%)245.10 (-17.53%)134.21 (-85.80%)
a279467 — 2026-08-31 16:07:160.33 (n/a)0.16 (n/a)0.17 (n/a)0.04 (n/a)0.12 (n/a)2421.90 (n/a)1107.32 (n/a)569.60 (n/a)297.20 (n/a)945.06 (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)
7e49fd8 — 2026-09-01 22:36:560.34 (-7.95%)0.23 (-6.79%)0.18 (-24.42%)0.13 (-21.54%)0.10 (+18.69%)780.20 (+27.44%)495.38 (+14.35%)540.00 (+32.32%)285.30 (+8.64%)207.05 (+51.09%)
a279467 — 2026-08-31 16:07:160.37 (n/a)0.25 (n/a)0.24 (n/a)0.16 (n/a)0.08 (n/a)612.20 (n/a)433.20 (n/a)408.10 (n/a)262.60 (n/a)137.04 (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)
7e49fd8 — 2026-09-01 22:36:560.30 (-5.59%)0.17 (-25.42%)0.15 (-38.14%)0.11 (-29.44%)0.08 (+14.98%)666.00 (+41.73%)497.30 (+41.39%)503.20 (+61.64%)242.60 (+5.89%)164.31 (+56.97%)
a279467 — 2026-08-31 16:07:160.32 (n/a)0.23 (n/a)0.24 (n/a)0.16 (n/a)0.07 (n/a)469.90 (n/a)351.72 (n/a)311.30 (n/a)229.10 (n/a)104.67 (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)
7e49fd8 — 2026-09-01 22:36:560.29 (+3.39%)0.21 (+4.94%)0.19 (+17.69%)0.12 (-5.66%)0.07 (-9.71%)602.90 (+6.00%)392.72 (-6.63%)390.70 (-15.03%)250.50 (-3.28%)140.92 (-6.79%)
a279467 — 2026-08-31 16:07:160.28 (n/a)0.20 (n/a)0.16 (n/a)0.13 (n/a)0.08 (n/a)568.80 (n/a)420.60 (n/a)459.80 (n/a)259.00 (n/a)151.18 (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)
7e49fd8 — 2026-09-01 22:36:560.28 (-22.37%)0.18 (-5.68%)0.16 (+3.52%)0.14 (+12.95%)0.06 (-40.31%)523.20 (-11.47%)437.34 (-1.81%)462.40 (-3.41%)261.80 (+28.78%)104.20 (-28.33%)
a279467 — 2026-08-31 16:07:160.36 (n/a)0.19 (n/a)0.15 (n/a)0.12 (n/a)0.10 (n/a)591.00 (n/a)445.38 (n/a)478.70 (n/a)203.30 (n/a)145.38 (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)
7e49fd8 — 2026-09-01 22:36:560.40 (+0.85%)0.28 (+11.81%)0.26 (+13.96%)0.22 (+28.25%)0.08 (-12.64%)604.30 (-22.03%)489.26 (-13.05%)499.70 (-12.26%)325.10 (-0.85%)117.07 (-28.18%)
a279467 — 2026-08-31 16:07:160.40 (n/a)0.25 (n/a)0.23 (n/a)0.17 (n/a)0.09 (n/a)775.00 (n/a)562.68 (n/a)569.50 (n/a)327.90 (n/a)163.01 (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)
7e49fd8 — 2026-09-01 22:36:560.32 (+1.85%)0.22 (+2.92%)0.22 (+12.25%)0.13 (+26.73%)0.07 (-18.81%)1026.00 (-21.09%)658.28 (-9.45%)602.40 (-10.91%)412.70 (-1.81%)225.77 (-35.39%)
a279467 — 2026-08-31 16:07:160.31 (n/a)0.21 (n/a)0.19 (n/a)0.10 (n/a)0.08 (n/a)1300.20 (n/a)726.98 (n/a)676.20 (n/a)420.30 (n/a)349.42 (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)
7e49fd8 — 2026-09-01 22:36:560.45 (+19.68%)0.35 (+32.16%)0.38 (+38.60%)0.26 (+218.74%)0.08 (-28.99%)507.60 (-68.62%)387.70 (-41.83%)344.80 (-27.85%)288.20 (-16.44%)95.93 (-82.13%)
a279467 — 2026-08-31 16:07:160.38 (n/a)0.27 (n/a)0.27 (n/a)0.08 (n/a)0.12 (n/a)1617.80 (n/a)666.54 (n/a)477.90 (n/a)344.90 (n/a)536.92 (n/a)
iron/operators/strided_copy

test_strided_copy[chunked_transfer]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
7e49fd8 — 2026-09-01 22:36:560.01 (-19.53%)0.01 (-2.51%)0.01 (-0.21%)0.01 (+16.66%)0.00 (-35.98%)475.00 (-14.28%)354.56 (-2.76%)298.50 (+0.20%)292.20 (+24.29%)84.25 (-35.30%)
a279467 — 2026-08-31 16:07:160.02 (n/a)0.01 (n/a)0.01 (n/a)0.01 (n/a)0.00 (n/a)554.10 (n/a)364.62 (n/a)297.90 (n/a)235.10 (n/a)130.23 (n/a)

test_strided_copy[contiguous]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
7e49fd8 — 2026-09-01 22:36:560.02 (+2.08%)0.02 (+30.84%)0.02 (+6.61%)0.01 (+326.57%)0.00 (-43.13%)443.80 (-76.56%)288.80 (-54.82%)271.70 (-6.21%)217.20 (-2.03%)90.17 (-87.30%)
a279467 — 2026-08-31 16:07:160.02 (n/a)0.01 (n/a)0.01 (n/a)0.00 (n/a)0.01 (n/a)1893.30 (n/a)639.28 (n/a)289.70 (n/a)221.70 (n/a)710.18 (n/a)

test_strided_copy[four_channels]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
7e49fd8 — 2026-09-01 22:36:560.01 (-22.37%)0.01 (-18.67%)0.01 (-33.27%)0.01 (-25.86%)0.00 (-28.55%)733.20 (+34.88%)425.00 (+19.49%)417.60 (+49.89%)277.50 (+28.83%)186.04 (+16.42%)
a279467 — 2026-08-31 16:07:160.02 (n/a)0.01 (n/a)0.01 (n/a)0.01 (n/a)0.01 (n/a)543.60 (n/a)355.68 (n/a)278.60 (n/a)215.40 (n/a)159.81 (n/a)

test_strided_copy[kv_slot0]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
7e49fd8 — 2026-09-01 22:36:560.58 (+4.41%)0.39 (-16.29%)0.33 (-34.69%)0.25 (+13.07%)0.13 (-1.46%)518.10 (-11.56%)374.90 (+16.56%)404.10 (+53.13%)229.70 (-4.21%)117.98 (-20.59%)
a279467 — 2026-08-31 16:07:160.55 (n/a)0.46 (n/a)0.50 (n/a)0.23 (n/a)0.14 (n/a)585.80 (n/a)321.64 (n/a)263.90 (n/a)239.80 (n/a)148.57 (n/a)

test_strided_copy[kv_slot5]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
7e49fd8 — 2026-09-01 22:36:560.66 (+8.94%)0.49 (-7.28%)0.49 (-5.85%)0.35 (-23.84%)0.11 (+113.65%)379.20 (+31.30%)282.40 (+11.56%)271.00 (+6.23%)201.30 (-8.21%)64.54 (+158.11%)
a279467 — 2026-08-31 16:07:160.60 (n/a)0.53 (n/a)0.52 (n/a)0.46 (n/a)0.05 (n/a)288.80 (n/a)253.14 (n/a)255.10 (n/a)219.30 (n/a)25.00 (n/a)

test_strided_copy[kv_slot5_four_channels]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
7e49fd8 — 2026-09-01 22:36:560.54 (+27.45%)0.35 (+12.69%)0.28 (-4.19%)0.22 (-15.66%)0.15 (+131.33%)599.70 (+18.56%)435.44 (-0.24%)465.40 (+4.37%)243.20 (-21.55%)172.76 (+122.99%)
a279467 — 2026-08-31 16:07:160.43 (n/a)0.31 (n/a)0.30 (n/a)0.26 (n/a)0.07 (n/a)505.80 (n/a)436.48 (n/a)445.90 (n/a)310.00 (n/a)77.47 (n/a)

test_strided_copy[kv_slot5_two_channels]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
7e49fd8 — 2026-09-01 22:36:560.58 (+21.72%)0.45 (+34.59%)0.45 (+81.08%)0.27 (+12.96%)0.11 (-9.25%)480.80 (-11.49%)312.34 (-28.33%)291.10 (-44.77%)229.30 (-17.87%)97.59 (-30.16%)
a279467 — 2026-08-31 16:07:160.47 (n/a)0.33 (n/a)0.25 (n/a)0.24 (n/a)0.12 (n/a)543.20 (n/a)435.82 (n/a)527.10 (n/a)279.20 (n/a)139.73 (n/a)

test_strided_copy[kv_slot_last]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
7e49fd8 — 2026-09-01 22:36:560.58 (-5.83%)0.39 (-14.18%)0.41 (-2.55%)0.21 (-38.27%)0.14 (+25.99%)623.00 (+61.99%)380.76 (+24.90%)322.30 (+2.64%)227.20 (+6.17%)151.08 (+125.45%)
a279467 — 2026-08-31 16:07:160.62 (n/a)0.45 (n/a)0.42 (n/a)0.34 (n/a)0.11 (n/a)384.60 (n/a)304.84 (n/a)314.00 (n/a)214.00 (n/a)67.01 (n/a)

test_strided_copy[two_channels]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
7e49fd8 — 2026-09-01 22:36:560.02 (-12.92%)0.01 (-19.04%)0.01 (-35.64%)0.01 (+0.97%)0.00 (-20.70%)721.80 (-0.97%)475.14 (+15.38%)460.40 (+55.38%)231.50 (+14.83%)180.36 (-16.91%)
a279467 — 2026-08-31 16:07:160.02 (n/a)0.01 (n/a)0.01 (n/a)0.01 (n/a)0.01 (n/a)728.90 (n/a)411.80 (n/a)296.30 (n/a)201.60 (n/a)217.07 (n/a)

test_strided_copy[two_channels_chunked]

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
7e49fd8 — 2026-09-01 22:36:560.02 (+16.96%)0.01 (+7.30%)0.01 (-0.55%)0.01 (+15.88%)0.00 (+17.16%)453.30 (-13.71%)338.86 (-6.31%)294.40 (+0.55%)220.90 (-14.51%)107.11 (-7.76%)
a279467 — 2026-08-31 16:07:160.02 (n/a)0.01 (n/a)0.01 (n/a)0.01 (n/a)0.00 (n/a)525.30 (n/a)361.68 (n/a)292.80 (n/a)258.40 (n/a)116.12 (n/a)

test_transfer_size_not_dividing_per_channel_share_is_rejected[iter0]

No metrics available.

test_transfer_size_not_dividing_per_channel_share_is_rejected[iter1]

No metrics available.

test_transfer_size_not_dividing_per_channel_share_is_rejected[iter2]

No metrics available.

test_transfer_size_not_dividing_per_channel_share_is_rejected[iter3]

No metrics available.

test_transfer_size_not_dividing_per_channel_share_is_rejected[iter4]

No metrics available.

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)
7e49fd8 — 2026-09-01 22:36:560.00 (+66.67%)0.00 (+25.00%)0.00 (+0.00%)0.00 (+0.00%)0.00 (+158.20%)22662.29 (+9.02%)15497.29 (-10.66%)17632.59 (+0.42%)7525.71 (-48.88%)6036.78 (+137.67%)
a279467 — 2026-08-31 16:07:160.00 (n/a)0.00 (n/a)0.00 (n/a)0.00 (n/a)0.00 (n/a)20786.53 (n/a)17346.57 (n/a)17558.44 (n/a)14720.43 (n/a)2539.98 (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)
7e49fd8 — 2026-09-01 22:36:560.00 (-15.38%)0.00 (-2.56%)0.00 (+60.00%)0.00 (+0.00%)0.00 (-35.99%)20196.11 (+9.32%)12571.46 (-4.46%)10369.75 (-37.20%)7355.55 (+20.36%)5738.60 (-9.33%)
a279467 — 2026-08-31 16:07:160.00 (n/a)0.00 (n/a)0.00 (n/a)0.00 (n/a)0.00 (n/a)18473.88 (n/a)13158.86 (n/a)16512.32 (n/a)6111.54 (n/a)6329.32 (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)
7e49fd8 — 2026-09-01 22:36:560.13 (-17.14%)0.10 (-5.04%)0.09 (-2.85%)0.07 (+4.72%)0.02 (-28.37%)30500.83 (-4.49%)22580.84 (+1.77%)23701.37 (+3.00%)16368.93 (+20.70%)5758.23 (-18.58%)
a279467 — 2026-08-31 16:07:160.15 (n/a)0.10 (n/a)0.09 (n/a)0.07 (n/a)0.03 (n/a)31935.69 (n/a)22189.18 (n/a)23011.36 (n/a)13562.18 (n/a)7072.47 (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)
7e49fd8 — 2026-09-01 22:36:561.68 (-13.79%)1.45 (+29.13%)1.56 (+59.74%)0.92 (+268.20%)0.31 (-55.72%)571.60 (-72.84%)378.92 (-51.56%)336.10 (-37.40%)312.10 (+16.02%)108.63 (-85.69%)
a279467 — 2026-08-31 16:07:161.95 (n/a)1.13 (n/a)0.98 (n/a)0.25 (n/a)0.69 (n/a)2104.40 (n/a)782.18 (n/a)536.90 (n/a)269.00 (n/a)758.94 (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)
7e49fd8 — 2026-09-01 22:36:562.71 (-10.20%)1.86 (-2.86%)2.31 (-9.78%)0.31 (-4.99%)0.97 (-15.29%)3355.80 (+5.25%)1061.16 (+0.19%)454.50 (+10.83%)387.10 (+11.36%)1287.59 (+6.04%)
a279467 — 2026-08-31 16:07:163.02 (n/a)1.92 (n/a)2.56 (n/a)0.33 (n/a)1.14 (n/a)3188.30 (n/a)1059.16 (n/a)410.10 (n/a)347.60 (n/a)1214.20 (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)
7e49fd8 — 2026-09-01 22:36:561.40 (-14.86%)1.08 (+23.51%)0.97 (+55.73%)0.90 (+94.24%)0.22 (-54.34%)585.20 (-48.52%)501.16 (-31.87%)543.10 (-35.78%)375.10 (+17.48%)93.15 (-71.26%)
a279467 — 2026-08-31 16:07:161.64 (n/a)0.87 (n/a)0.62 (n/a)0.46 (n/a)0.48 (n/a)1136.80 (n/a)735.64 (n/a)845.70 (n/a)319.30 (n/a)324.12 (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_09_01_22_35_07 at commit 7e49fd8.

Trends:

IRON Trends

@atassis

atassis commented Sep 1, 2026

Copy link
Copy Markdown
Contributor Author

Testing it from this branch won't work: pr-comment.yml is workflow_run-triggered, so GitHub always runs devel's copy. This PR actually demonstrated that itself, its own comment runs (33567717223, 33568280474) hit the same 422.

Good news is we don't need broken tests. The failure reproduces on any SHA where all four suites have reported, a279467 does it every time.

Bad news is you shouldn't merge this yet. I checked the comments on #150-#183: they run 122k to 142k characters and all of them posted fine, so 65536 isn't the limit GitHub actually enforces. As written my guard would strip the details off every comment, not just the huge ones.

Let me redo it to shed only on a real 422, and drop trends.md first, that keeps the pass/fail tables. I'll ping you when it's up.

65536 is the number the error message names, not the ceiling GitHub enforces:
every aggregate comment on amd#150-amd#183 is 122k-142k characters and posted, so
the guard would have stripped the details off every comment. Try the full
body and drop a section only when the API rejects one, trends.md first.
@atassis

atassis commented Sep 1, 2026

Copy link
Copy Markdown
Contributor Author

Pushed as fe71df5. It now sheds only when the API actually rejects the body, trend tables first, which on the run that failed leaves an 84k comment with every pass/fail table intact. Body updated too.

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.

2 participants