Skip to content

Conversation

@andrej
Copy link
Collaborator

@andrej andrej commented Nov 13, 2025

Added

  • NPU-offloaded implementation of Llama 3.2 (1 billion parameters)
  • New CI test for llama

Changed

  • Moved example -> operators
  • Created examples directory to hold llama
  • Removed reliance on environment variables for MLIR AIE and Peano directories, since we no longer use an env setup script

PR Merge Checklist

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

@andrej andrej force-pushed the bring-in-the-llama branch 5 times, most recently from 99e576f to 62bf26b Compare November 13, 2025 15:29
@amd amd deleted a comment from github-actions bot Nov 13, 2025
@amd amd deleted a comment from github-actions bot Nov 13, 2025
@amd amd deleted a comment from github-actions bot Nov 13, 2025
@kurtis-b-1 kurtis-b-1 mentioned this pull request Nov 13, 2025
3 tasks
@andrej andrej force-pushed the bring-in-the-llama branch from ad36747 to 9da3796 Compare November 13, 2025 19:04
@github-actions
Copy link

📊 Test Results for Test Example Applications

3e0443f (2025_11_13_19_59_33)

IRONCLAD

Tested on 2025_11_13_19_59_33 at commit 3e0443f.

Test Checks TTFT (mean)TPS (mean)
llama_3.2_1b ✅ 5/5 24.63 0.83
📈 Trends (vs main branch) for Test Example Applications

3e0443f (2025_11_13_19_59_33)

IRONCLAD Trends

llama_3.2_1b

Commit/Date Num Tokens (max)Num Tokens (mean)Num Tokens (median)Num Tokens (min)Num Tokens (stddev)TPS (max)TPS (mean)TPS (median)TPS (min)TPS (stddev)TTFT (max)TTFT (mean)TTFT (median)TTFT (min)TTFT (stddev)Total (max)Total (mean)Total (median)Total (min)Total (stddev)
3e0443f — 2025-11-13 19:50:5440.00 (n/a)40.00 (n/a)40.00 (n/a)40.00 (n/a)0.00 (n/a)0.84 (n/a)0.83 (n/a)0.83 (n/a)0.81 (n/a)0.01 (n/a)24.90 (n/a)24.63 (n/a)24.83 (n/a)24.07 (n/a)0.32 (n/a)73.22 (n/a)71.94 (n/a)71.97 (n/a)70.77 (n/a)0.88 (n/a)

@amd amd deleted a comment from github-actions bot Nov 13, 2025
@amd amd deleted a comment from github-actions bot Nov 13, 2025
@amd amd deleted a comment from github-actions bot Nov 13, 2025
@amd amd deleted a comment from github-actions bot Nov 13, 2025
@github-actions
Copy link

📊 Test Results for Small Benchmark/Test Suite

3e0443f (2025_11_13_21_22_28)

IRONCLAD

Tested on 2025_11_13_21_22_28 at commit 3e0443f.

Test Checks Latency (mean)Bandwidth (mean)GFLOP/s (mean)
axpy_1_cols_2_channels_2048_tile_2048_3.0 ✅ 5/5 159.80 0.07 n/a
axpy_2_cols_2_channels_2048_tile_1024_3.0 ✅ 5/5 136.60 0.09 n/a
axpy_4_cols_2_channels_2048_tile_512_3.0 ✅ 5/5 160.00 0.07 n/a
axpy_8_cols_2_channels_2048_tile_256_3.0 ✅ 5/5 210.20 0.06 n/a
dequant_1_cols_1_channels_2048_tile_2048 ✅ 5/5 175.20 0.03 n/a
dequant_1_cols_2_channels_2048_tile_1024 ✅ 5/5 188.60 0.03 n/a
dequant_2_cols_1_channels_2048_tile_1024 ✅ 5/5 156.40 0.03 n/a
dequant_2_cols_2_channels_2048_tile_512 ✅ 5/5 187.20 0.03 n/a
dequant_4_cols_1_channels_2048_tile_512 ✅ 5/5 175.40 0.03 n/a
dequant_4_cols_2_channels_2048_tile_256 ✅ 5/5 214.40 0.02 n/a
dequant_8_cols_1_channels_2048_tile_256 ✅ 5/5 195.60 0.03 n/a
dequant_8_cols_2_channels_2048_tile_128 ✅ 5/5 238.40 0.02 n/a
eltwise_add_1_cols_2_channels_2048_tile_2048 ✅ 5/5 206.00 0.06 n/a
eltwise_add_2_cols_2_channels_2048_tile_1024 ✅ 5/5 193.40 0.06 n/a
eltwise_add_4_cols_2_channels_2048_tile_512 ✅ 5/5 175.60 0.07 n/a
eltwise_add_8_cols_2_channels_2048_tile_256 ✅ 5/5 198.20 0.06 n/a
eltwise_mul_1_cols_2_channels_2048_tile_2048 ✅ 5/5 194.20 0.06 n/a
eltwise_mul_2_cols_2_channels_2048_tile_1024 ✅ 5/5 198.00 0.06 n/a
eltwise_mul_4_cols_2_channels_2048_tile_512 ✅ 5/5 172.60 0.07 n/a
eltwise_mul_8_cols_2_channels_2048_tile_256 ✅ 5/5 209.40 0.06 n/a
gelu_1_cols_1_channels_2048_tile_2048 ✅ 5/5 185.80 0.05 n/a
gelu_1_cols_2_channels_2048_tile_1024 ✅ 5/5 196.20 0.04 n/a
gelu_2_cols_1_channels_2048_tile_1024 ✅ 5/5 185.40 0.04 n/a
gelu_2_cols_2_channels_2048_tile_512 ✅ 5/5 184.80 0.05 n/a
gelu_4_cols_1_channels_2048_tile_512 ✅ 5/5 235.80 0.04 n/a
gelu_4_cols_2_channels_2048_tile_256 ✅ 5/5 171.20 0.05 n/a
gelu_8_cols_1_channels_2048_tile_256 ✅ 5/5 141.00 0.06 n/a
gelu_8_cols_2_channels_2048_tile_128 ✅ 5/5 153.40 0.05 n/a
gemm_2048x2048x2048_64x64x64_8_cols_0_bcolmaj_0_ccolmaj_0 ✅ 5/5 6515.80 3.60 n/a
layer_norm_1_cols_1_channels_2048_tile_2048 ✅ 5/5 162.00 0.08 n/a
layer_norm_1_cols_2_channels_2048_tile_1024 ✅ 5/5 155.60 0.08 n/a
layer_norm_2_cols_1_channels_2048_tile_1024 ✅ 5/5 141.40 0.09 n/a
layer_norm_2_cols_2_channels_2048_tile_512 ✅ 5/5 137.80 0.09 n/a
layer_norm_4_cols_1_channels_2048_tile_512 ✅ 5/5 150.80 0.08 n/a
layer_norm_4_cols_2_channels_2048_tile_256 ✅ 5/5 141.80 0.09 n/a
layer_norm_8_cols_1_channels_2048_tile_256 ✅ 5/5 140.20 0.09 n/a
layer_norm_8_cols_2_channels_2048_tile_128 ✅ 5/5 189.60 0.07 n/a
matrix_vector_mul_128x128_32_1col ✅ 5/5 774.20 0.04 0.04
matrix_vector_mul_2048x8192_1_1col ✅ 5/5 3147.20 10.68 10.67
matrix_vector_mul_2048x8192_1_2col ✅ 5/5 1965.20 17.11 17.09
matrix_vector_mul_2048x8192_1_4col ✅ 5/5 1560.00 21.58 21.56
matrix_vector_mul_2048x8192_1_8col ✅ 5/5 2113.00 15.99 15.98
matrix_vector_mul_8192x2048_4_1col ✅ 5/5 3076.60 10.91 10.91
matrix_vector_mul_8192x2048_4_2col ✅ 5/5 1951.00 17.21 17.20
matrix_vector_mul_8192x2048_4_4col ✅ 5/5 1581.60 21.24 21.22
matrix_vector_mul_8192x2048_4_8col ✅ 5/5 1913.40 17.71 17.70
mem_copy_16_cores_2_chans_2048_tile_128_False ✅ 5/5 167.80 0.05 n/a
mem_copy_1_cores_1_chans_2048_tile_2048_False ✅ 5/5 190.20 0.04 n/a
mem_copy_2_cores_1_chans_2048_tile_1024_False ✅ 5/5 153.00 0.05 n/a
mem_copy_2_cores_2_chans_2048_tile_1024_False ✅ 5/5 162.00 0.05 n/a
mem_copy_4_cores_1_chans_2048_tile_512_False ✅ 5/5 145.00 0.05 n/a
mem_copy_4_cores_2_chans_2048_tile_512_False ✅ 5/5 175.40 0.05 n/a
mem_copy_8_cores_1_chans_2048_tile_256_False ✅ 5/5 176.00 0.04 n/a
mem_copy_8_cores_2_chans_2048_tile_256_False ✅ 5/5 174.00 0.04 n/a
mha ✅ 5/5 n/a 0.09 65.03
relu_1_cols_1_channels_2048_tile_2048 ✅ 5/5 134.40 0.06 n/a
relu_2_cols_1_channels_2048_tile_1024 ✅ 5/5 159.00 0.05 n/a
relu_4_cols_1_channels_2048_tile_512 ✅ 5/5 180.00 0.04 n/a
relu_8_cols_1_channels_2048_tile_256 ✅ 5/5 174.20 0.05 n/a
rms_norm_1_cols_1_channels_2048_tile_2048 ✅ 5/5 124.80 0.09 n/a
rms_norm_1_cols_2_channels_2048_tile_1024 ✅ 5/5 138.80 0.08 n/a
rms_norm_2_cols_1_channels_2048_tile_1024 ✅ 5/5 144.60 0.08 n/a
rms_norm_2_cols_2_channels_2048_tile_512 ✅ 5/5 161.80 0.07 n/a
rms_norm_4_cols_1_channels_2048_tile_512 ✅ 5/5 179.20 0.07 n/a
rms_norm_4_cols_2_channels_2048_tile_256 ✅ 5/5 156.60 0.07 n/a
rms_norm_8_cols_1_channels_2048_tile_256 ✅ 5/5 203.00 0.06 n/a
rms_norm_8_cols_2_channels_2048_tile_128 ✅ 5/5 233.80 0.05 n/a
rope_1_cols_2_channels_4096_tile_4096_0 ✅ 5/5 171.40 0.14 n/a
rope_2_cols_2_channels_4096_tile_2048_0 ✅ 5/5 152.20 0.15 n/a
rope_4_cols_2_channels_4096_tile_1024_0 ✅ 5/5 171.60 0.14 n/a
rope_8_cols_2_channels_4096_tile_512_0 ✅ 5/5 179.60 0.13 n/a
silu_1_cols_1_channels_2048_tile_2048 ✅ 5/5 146.00 0.05 n/a
silu_2_cols_1_channels_2048_tile_1024 ✅ 5/5 178.20 0.04 n/a
silu_4_cols_1_channels_2048_tile_512 ✅ 5/5 225.60 0.04 n/a
silu_8_cols_1_channels_2048_tile_256 ✅ 5/5 172.80 0.05 n/a
softmax_1_cols_2_channels_4096_tile_2048 ✅ 5/5 140.60 0.16 n/a
softmax_2_cols_2_channels_4096_tile_1024 ✅ 5/5 159.20 0.15 n/a
softmax_2_cols_2_channels_4096_tile_512 ✅ 5/5 159.40 0.16 n/a
swiglu ✅ 5/5 n/a n/a n/a
transpose_2048_M_64_N_1_cols_1_channels_64_m_64_n_8_s ✅ 5/5 177.20 2.86 n/a
transpose_2048_M_64_N_1_cols_2_channels_64_m_64_n_8_s ✅ 5/5 178.40 3.10 n/a
weighted_rms_norm_1_cols_2_channels_2048_weights_2048 ✅ 5/5 197.40 0.06 n/a
weighted_rms_norm_2_cols_2_channels_2048_weights_1024 ✅ 5/5 154.00 0.07 n/a
weighted_rms_norm_4_cols_2_channels_2048_weights_512 ✅ 5/5 170.20 0.07 n/a
weighted_rms_norm_8_cols_2_channels_2048_weights_256 ✅ 5/5 162.60 0.07 n/a
📈 Trends (vs main branch) for Small Benchmark/Test Suite

3e0443f (2025_11_13_21_22_28)

IRONCLAD Trends

axpy_1_cols_2_channels_2048_tile_2048_3.0

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.08 (-21.09%)0.07 (-14.48%)0.07 (-10.90%)0.06 (-9.50%)0.01 (-52.56%)179.00 (+10.49%)159.80 (+15.46%)156.00 (+12.23%)147.00 (+26.72%)12.42 (-33.54%)
0999a9a — 2025-11-13 19:24:080.10 (+16.38%)0.08 (+3.69%)0.08 (+1.44%)0.07 (-8.64%)0.01 (+372.63%)162.00 (+9.46%)138.40 (-1.84%)139.00 (-1.42%)116.00 (-14.07%)18.68 (+340.35%)
fbf1591 — 2025-11-13 03:39:220.08 (n/a)0.08 (n/a)0.08 (n/a)0.08 (n/a)0.00 (n/a)148.00 (n/a)141.00 (n/a)141.00 (n/a)135.00 (n/a)4.24 (n/a)

axpy_2_cols_2_channels_2048_tile_1024_3.0

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.11 (+8.65%)0.09 (+21.28%)0.08 (+10.87%)0.06 (+28.80%)0.02 (-10.18%)184.00 (-22.36%)136.60 (-20.77%)138.00 (-9.80%)104.00 (-7.96%)29.85 (-41.19%)
0999a9a — 2025-11-13 19:24:080.10 (-2.65%)0.07 (+4.47%)0.07 (+7.19%)0.05 (+18.14%)0.02 (+1.47%)237.00 (-15.36%)172.40 (-4.54%)153.00 (-6.71%)113.00 (+2.73%)50.76 (-8.77%)
fbf1591 — 2025-11-13 03:39:220.10 (n/a)0.07 (n/a)0.07 (n/a)0.04 (n/a)0.02 (n/a)280.00 (n/a)180.60 (n/a)164.00 (n/a)110.00 (n/a)55.64 (n/a)

axpy_4_cols_2_channels_2048_tile_512_3.0

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.09 (-25.58%)0.07 (-15.63%)0.08 (-7.33%)0.06 (-19.17%)0.01 (-34.64%)193.00 (+23.72%)160.00 (+17.65%)150.00 (+7.91%)129.00 (+34.38%)24.10 (+14.53%)
0999a9a — 2025-11-13 19:24:080.12 (+30.21%)0.09 (+13.66%)0.08 (+4.32%)0.07 (+35.90%)0.02 (+34.67%)156.00 (-26.42%)136.00 (-12.14%)139.00 (-4.14%)96.00 (-23.20%)21.04 (-29.55%)
fbf1591 — 2025-11-13 03:39:220.09 (n/a)0.08 (n/a)0.08 (n/a)0.05 (n/a)0.01 (n/a)212.00 (n/a)154.80 (n/a)145.00 (n/a)125.00 (n/a)29.87 (n/a)

axpy_8_cols_2_channels_2048_tile_256_3.0

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.07 (-30.49%)0.06 (-33.52%)0.06 (-33.68%)0.04 (-47.91%)0.01 (+2.82%)311.00 (+91.98%)210.20 (+55.24%)193.00 (+50.78%)164.00 (+43.86%)51.57 (+198.99%)
0999a9a — 2025-11-13 19:24:080.10 (+21.05%)0.09 (+32.73%)0.09 (+46.09%)0.07 (+33.95%)0.01 (-2.31%)162.00 (-25.35%)135.40 (-25.44%)128.00 (-31.55%)114.00 (-17.39%)17.25 (-38.93%)
fbf1591 — 2025-11-13 03:39:220.08 (n/a)0.06 (n/a)0.06 (n/a)0.05 (n/a)0.01 (n/a)217.00 (n/a)181.60 (n/a)187.00 (n/a)138.00 (n/a)28.24 (n/a)

dequant_1_cols_1_channels_2048_tile_2048

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.03 (-32.92%)0.03 (-23.60%)0.03 (-11.45%)0.02 (-21.57%)0.00 (-61.81%)204.00 (+27.50%)175.20 (+28.26%)166.00 (+12.93%)161.00 (+49.07%)16.19 (-26.93%)
0999a9a — 2025-11-13 19:24:080.05 (+58.33%)0.04 (+50.10%)0.03 (+30.61%)0.03 (+85.00%)0.01 (+48.14%)160.00 (-45.95%)136.60 (-34.07%)147.00 (-23.44%)108.00 (-36.84%)22.16 (-51.39%)
fbf1591 — 2025-11-13 03:39:220.03 (n/a)0.02 (n/a)0.03 (n/a)0.02 (n/a)0.00 (n/a)296.00 (n/a)207.20 (n/a)192.00 (n/a)171.00 (n/a)45.59 (n/a)

dequant_1_cols_2_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.03 (-24.52%)0.03 (-19.15%)0.03 (-10.50%)0.02 (-31.89%)0.00 (-15.78%)254.00 (+46.82%)188.60 (+24.74%)181.00 (+11.73%)155.00 (+32.48%)34.84 (+67.83%)
0999a9a — 2025-11-13 19:24:080.04 (-16.24%)0.03 (-9.53%)0.03 (-9.88%)0.03 (-2.31%)0.01 (-30.64%)173.00 (+2.37%)151.20 (+9.09%)162.00 (+10.96%)117.00 (+19.39%)20.76 (-12.76%)
fbf1591 — 2025-11-13 03:39:220.05 (n/a)0.04 (n/a)0.03 (n/a)0.03 (n/a)0.01 (n/a)169.00 (n/a)138.60 (n/a)146.00 (n/a)98.00 (n/a)23.80 (n/a)

dequant_2_cols_1_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.03 (-24.16%)0.03 (-21.79%)0.03 (-21.66%)0.03 (-15.85%)0.00 (-63.15%)164.00 (+18.84%)156.40 (+27.36%)157.00 (+27.64%)149.00 (+31.86%)5.12 (-42.21%)
0999a9a — 2025-11-13 19:24:080.04 (+10.62%)0.04 (+18.53%)0.04 (+20.33%)0.04 (+14.49%)0.00 (-3.03%)138.00 (-12.66%)122.80 (-15.78%)123.00 (-16.89%)113.00 (-9.60%)8.86 (-22.67%)
fbf1591 — 2025-11-13 03:39:220.04 (n/a)0.03 (n/a)0.03 (n/a)0.03 (n/a)0.00 (n/a)158.00 (n/a)145.80 (n/a)148.00 (n/a)125.00 (n/a)11.46 (n/a)

dequant_2_cols_2_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.03 (-16.45%)0.03 (-17.60%)0.03 (-13.87%)0.02 (-25.58%)0.00 (+12.92%)258.00 (+34.38%)187.20 (+23.65%)173.00 (+16.11%)152.00 (+19.69%)39.00 (+76.96%)
0999a9a — 2025-11-13 19:24:080.04 (+6.30%)0.03 (+17.59%)0.03 (+23.49%)0.03 (+8.85%)0.00 (-8.04%)192.00 (-8.13%)151.40 (-15.51%)149.00 (-19.02%)127.00 (-5.93%)22.04 (-19.09%)
fbf1591 — 2025-11-13 03:39:220.04 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.00 (n/a)209.00 (n/a)179.20 (n/a)184.00 (n/a)135.00 (n/a)27.24 (n/a)

dequant_4_cols_1_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.03 (-7.09%)0.03 (-12.47%)0.03 (-9.38%)0.02 (-37.45%)0.01 (+119.01%)267.00 (+59.88%)175.40 (+19.81%)160.00 (+10.34%)141.00 (+7.63%)46.70 (+286.19%)
0999a9a — 2025-11-13 19:24:080.04 (-16.79%)0.03 (-3.88%)0.03 (-8.97%)0.03 (+29.94%)0.00 (-64.56%)167.00 (-23.04%)146.40 (-0.95%)145.00 (+9.85%)131.00 (+20.18%)12.09 (-67.98%)
fbf1591 — 2025-11-13 03:39:220.04 (n/a)0.03 (n/a)0.04 (n/a)0.02 (n/a)0.01 (n/a)217.00 (n/a)147.80 (n/a)132.00 (n/a)109.00 (n/a)37.76 (n/a)

dequant_4_cols_2_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.02 (-25.63%)0.02 (-21.18%)0.02 (-13.37%)0.02 (-25.38%)0.00 (-26.77%)260.00 (+34.02%)214.40 (+26.86%)202.00 (+15.43%)199.00 (+34.46%)23.04 (+35.74%)
0999a9a — 2025-11-13 19:24:080.03 (-19.59%)0.03 (-7.11%)0.03 (-5.71%)0.03 (+0.00%)0.00 (-49.35%)194.00 (+0.00%)169.00 (+5.36%)175.00 (+6.06%)148.00 (+24.37%)16.97 (-38.04%)
fbf1591 — 2025-11-13 03:39:220.04 (n/a)0.03 (n/a)0.03 (n/a)0.03 (n/a)0.01 (n/a)194.00 (n/a)160.40 (n/a)165.00 (n/a)119.00 (n/a)27.39 (n/a)

dequant_8_cols_1_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.03 (-32.18%)0.03 (-25.26%)0.03 (-25.91%)0.02 (-9.01%)0.00 (-69.45%)222.00 (+9.90%)195.60 (+29.54%)193.00 (+34.97%)174.00 (+47.46%)15.34 (-49.88%)
0999a9a — 2025-11-13 19:24:080.04 (+24.58%)0.03 (+11.01%)0.03 (+12.59%)0.02 (-7.92%)0.01 (+137.56%)202.00 (+8.60%)151.00 (-7.13%)143.00 (-11.18%)118.00 (-19.73%)30.61 (+108.57%)
fbf1591 — 2025-11-13 03:39:220.03 (n/a)0.03 (n/a)0.03 (n/a)0.03 (n/a)0.00 (n/a)186.00 (n/a)162.60 (n/a)161.00 (n/a)147.00 (n/a)14.68 (n/a)

dequant_8_cols_2_channels_2048_tile_128

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.03 (-10.31%)0.02 (-3.64%)0.02 (-9.42%)0.01 (-4.13%)0.00 (-31.03%)339.00 (+4.31%)238.40 (+1.02%)223.00 (+10.40%)194.00 (+11.49%)51.55 (-16.60%)
0999a9a — 2025-11-13 19:24:080.03 (-6.32%)0.02 (+2.86%)0.02 (+8.91%)0.02 (+4.31%)0.01 (+1.62%)325.00 (-4.13%)236.00 (-2.32%)202.00 (-8.18%)174.00 (+6.75%)61.82 (+3.89%)
fbf1591 — 2025-11-13 03:39:220.03 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)339.00 (n/a)241.60 (n/a)220.00 (n/a)163.00 (n/a)59.50 (n/a)

eltwise_add_1_cols_2_channels_2048_tile_2048

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.07 (-31.06%)0.06 (-30.25%)0.06 (-22.22%)0.04 (-42.51%)0.01 (+1.88%)287.00 (+73.94%)206.00 (+47.99%)180.00 (+28.57%)161.00 (+45.05%)49.71 (+150.12%)
0999a9a — 2025-11-13 19:24:080.10 (+31.53%)0.08 (+10.02%)0.08 (+7.14%)0.07 (-6.06%)0.01 (+707.91%)165.00 (+6.45%)139.20 (-7.20%)140.00 (-6.67%)111.00 (-23.97%)19.87 (+555.21%)
fbf1591 — 2025-11-13 03:39:220.08 (n/a)0.08 (n/a)0.08 (n/a)0.07 (n/a)0.00 (n/a)155.00 (n/a)150.00 (n/a)150.00 (n/a)146.00 (n/a)3.03 (n/a)

eltwise_add_2_cols_2_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.08 (-16.67%)0.06 (-11.38%)0.07 (-7.60%)0.04 (-7.34%)0.01 (-22.17%)286.00 (+7.92%)193.40 (+11.41%)171.00 (+8.23%)150.00 (+20.00%)48.09 (+0.14%)
0999a9a — 2025-11-13 19:24:080.09 (-11.20%)0.07 (-26.87%)0.07 (-29.11%)0.04 (-46.79%)0.02 (+77.18%)265.00 (+87.94%)173.60 (+44.19%)158.00 (+41.07%)125.00 (+12.61%)48.02 (+301.50%)
fbf1591 — 2025-11-13 03:39:220.10 (n/a)0.10 (n/a)0.10 (n/a)0.08 (n/a)0.01 (n/a)141.00 (n/a)120.40 (n/a)112.00 (n/a)111.00 (n/a)11.96 (n/a)

eltwise_add_4_cols_2_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.08 (-2.13%)0.07 (-3.29%)0.07 (+3.03%)0.05 (-8.49%)0.01 (+11.77%)212.00 (+9.28%)175.60 (+4.03%)165.00 (-2.94%)141.00 (+2.17%)25.50 (+26.09%)
0999a9a — 2025-11-13 19:24:080.08 (-13.04%)0.07 (-17.24%)0.07 (-20.00%)0.06 (-12.37%)0.01 (-4.71%)194.00 (+14.12%)168.80 (+21.09%)170.00 (+25.00%)138.00 (+15.00%)20.22 (+22.30%)
fbf1591 — 2025-11-13 03:39:220.10 (n/a)0.08 (n/a)0.08 (n/a)0.07 (n/a)0.01 (n/a)170.00 (n/a)139.40 (n/a)136.00 (n/a)120.00 (n/a)16.54 (n/a)

eltwise_add_8_cols_2_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.07 (-3.43%)0.06 (-0.58%)0.06 (+11.05%)0.05 (-16.46%)0.01 (+65.10%)243.00 (+19.70%)198.20 (+1.75%)181.00 (-9.95%)175.00 (+3.55%)26.45 (+103.99%)
0999a9a — 2025-11-13 19:24:080.07 (-33.14%)0.06 (-30.15%)0.06 (-34.33%)0.06 (-19.70%)0.00 (-62.51%)203.00 (+24.54%)194.80 (+41.16%)201.00 (+52.27%)169.00 (+49.56%)12.97 (-31.96%)
fbf1591 — 2025-11-13 03:39:220.10 (n/a)0.08 (n/a)0.09 (n/a)0.07 (n/a)0.01 (n/a)163.00 (n/a)138.00 (n/a)132.00 (n/a)113.00 (n/a)19.06 (n/a)

eltwise_mul_1_cols_2_channels_2048_tile_2048

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.07 (-4.71%)0.06 (-5.04%)0.06 (-10.05%)0.05 (+11.97%)0.01 (-29.17%)234.00 (-10.69%)194.20 (+3.52%)189.00 (+11.18%)170.00 (+4.94%)23.79 (-36.41%)
0999a9a — 2025-11-13 19:24:080.07 (-23.46%)0.06 (-22.73%)0.07 (-15.88%)0.04 (-39.31%)0.01 (+34.51%)262.00 (+64.78%)187.60 (+32.49%)170.00 (+18.88%)162.00 (+30.65%)37.40 (+200.01%)
fbf1591 — 2025-11-13 03:39:220.09 (n/a)0.08 (n/a)0.08 (n/a)0.07 (n/a)0.01 (n/a)159.00 (n/a)141.60 (n/a)143.00 (n/a)124.00 (n/a)12.47 (n/a)

eltwise_mul_2_cols_2_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.08 (+8.16%)0.06 (-6.30%)0.06 (-2.20%)0.04 (-29.43%)0.01 (+165.41%)282.00 (+41.71%)198.00 (+11.74%)182.00 (+2.25%)147.00 (-7.55%)47.66 (+252.36%)
0999a9a — 2025-11-13 19:24:080.07 (-23.27%)0.06 (-20.20%)0.06 (-21.91%)0.06 (-16.58%)0.00 (-39.23%)199.00 (+19.88%)177.20 (+24.79%)178.00 (+28.06%)159.00 (+30.33%)13.53 (-5.57%)
fbf1591 — 2025-11-13 03:39:220.09 (n/a)0.08 (n/a)0.08 (n/a)0.07 (n/a)0.01 (n/a)166.00 (n/a)142.00 (n/a)139.00 (n/a)122.00 (n/a)14.32 (n/a)

eltwise_mul_4_cols_2_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.08 (-8.70%)0.07 (-12.62%)0.08 (-7.28%)0.04 (-39.23%)0.01 (+108.86%)260.00 (+64.56%)172.60 (+20.03%)151.00 (+7.86%)138.00 (+9.52%)45.54 (+274.09%)
0999a9a — 2025-11-13 19:24:080.09 (-10.32%)0.08 (-0.04%)0.08 (+2.86%)0.07 (+26.58%)0.01 (-53.20%)158.00 (-21.00%)143.80 (-2.84%)140.00 (-2.78%)126.00 (+11.50%)12.17 (-58.69%)
fbf1591 — 2025-11-13 03:39:220.10 (n/a)0.08 (n/a)0.08 (n/a)0.06 (n/a)0.01 (n/a)200.00 (n/a)148.00 (n/a)144.00 (n/a)113.00 (n/a)29.47 (n/a)

eltwise_mul_8_cols_2_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.06 (-20.99%)0.06 (-6.65%)0.06 (+4.26%)0.04 (-7.28%)0.01 (-38.69%)302.00 (+7.86%)209.40 (+3.97%)188.00 (-4.08%)181.00 (+26.57%)46.40 (-11.61%)
0999a9a — 2025-11-13 19:24:080.08 (-2.80%)0.06 (-12.49%)0.06 (-12.76%)0.04 (-23.57%)0.02 (+38.68%)280.00 (+30.84%)201.40 (+19.03%)196.00 (+14.62%)143.00 (+2.88%)52.49 (+87.02%)
fbf1591 — 2025-11-13 03:39:220.08 (n/a)0.07 (n/a)0.07 (n/a)0.05 (n/a)0.01 (n/a)214.00 (n/a)169.20 (n/a)171.00 (n/a)139.00 (n/a)28.07 (n/a)

gelu_1_cols_1_channels_2048_tile_2048

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.05 (-6.71%)0.05 (-10.78%)0.05 (-7.06%)0.03 (-32.23%)0.01 (+81.54%)273.00 (+47.57%)185.80 (+16.27%)170.00 (+7.59%)149.00 (+7.19%)44.61 (+198.77%)
0999a9a — 2025-11-13 19:24:080.06 (-8.63%)0.05 (+6.51%)0.05 (+14.56%)0.04 (+27.03%)0.00 (-57.98%)185.00 (-21.28%)159.80 (-10.22%)158.00 (-12.71%)139.00 (+9.45%)14.93 (-63.00%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.05 (n/a)0.05 (n/a)0.03 (n/a)0.01 (n/a)235.00 (n/a)178.00 (n/a)181.00 (n/a)127.00 (n/a)40.36 (n/a)

gelu_1_cols_2_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.04 (-29.89%)0.04 (-24.55%)0.04 (-18.92%)0.03 (-33.62%)0.00 (-6.14%)235.00 (+50.64%)196.20 (+33.11%)185.00 (+23.33%)184.00 (+42.64%)19.61 (+105.98%)
0999a9a — 2025-11-13 19:24:080.06 (-10.08%)0.06 (-4.24%)0.05 (-8.67%)0.05 (+26.92%)0.00 (-59.85%)156.00 (-21.21%)147.40 (+1.52%)150.00 (+9.49%)129.00 (+11.21%)9.52 (-66.37%)
fbf1591 — 2025-11-13 03:39:220.07 (n/a)0.06 (n/a)0.06 (n/a)0.04 (n/a)0.01 (n/a)198.00 (n/a)145.20 (n/a)137.00 (n/a)116.00 (n/a)28.31 (n/a)

gelu_2_cols_1_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.05 (-26.45%)0.04 (-27.61%)0.05 (-30.77%)0.04 (-28.64%)0.01 (-36.68%)220.00 (+40.13%)185.40 (+37.54%)182.00 (+44.44%)155.00 (+35.96%)20.90 (+17.74%)
0999a9a — 2025-11-13 19:24:080.07 (-4.39%)0.06 (+1.89%)0.07 (+8.73%)0.05 (+31.85%)0.01 (-34.72%)157.00 (-24.15%)134.80 (-4.94%)126.00 (-8.03%)114.00 (+4.59%)17.75 (-48.47%)
fbf1591 — 2025-11-13 03:39:220.08 (n/a)0.06 (n/a)0.06 (n/a)0.04 (n/a)0.01 (n/a)207.00 (n/a)141.80 (n/a)137.00 (n/a)109.00 (n/a)34.44 (n/a)

gelu_2_cols_2_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.06 (-19.08%)0.05 (-23.99%)0.05 (-32.02%)0.03 (-28.42%)0.01 (-13.30%)278.00 (+39.70%)184.80 (+33.53%)178.00 (+47.11%)131.00 (+23.58%)51.88 (+50.21%)
0999a9a — 2025-11-13 19:24:080.08 (+21.70%)0.06 (+6.40%)0.07 (+12.40%)0.04 (-15.58%)0.01 (+146.44%)199.00 (+18.45%)138.40 (-1.70%)121.00 (-11.03%)106.00 (-17.83%)34.54 (+139.17%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.06 (n/a)0.06 (n/a)0.05 (n/a)0.01 (n/a)168.00 (n/a)140.80 (n/a)136.00 (n/a)129.00 (n/a)14.44 (n/a)

gelu_4_cols_1_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.05 (-27.22%)0.04 (-34.00%)0.04 (-32.16%)0.02 (-53.89%)0.01 (+41.19%)360.00 (+116.87%)235.80 (+60.41%)227.00 (+47.40%)169.00 (+37.40%)67.32 (+329.21%)
0999a9a — 2025-11-13 19:24:080.07 (+24.39%)0.06 (+20.95%)0.05 (+4.55%)0.05 (+34.94%)0.01 (-4.62%)166.00 (-25.89%)147.00 (-18.15%)154.00 (-4.35%)123.00 (-19.61%)15.68 (-43.29%)
fbf1591 — 2025-11-13 03:39:220.05 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)224.00 (n/a)179.60 (n/a)161.00 (n/a)153.00 (n/a)27.66 (n/a)

gelu_4_cols_2_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.06 (+0.76%)0.05 (-7.95%)0.05 (-9.52%)0.04 (-10.15%)0.01 (+33.97%)197.00 (+11.30%)171.20 (+9.74%)168.00 (+10.53%)131.00 (-0.76%)23.39 (+46.28%)
0999a9a — 2025-11-13 19:24:080.06 (-15.15%)0.05 (-2.99%)0.05 (+6.58%)0.05 (+2.82%)0.01 (-43.65%)177.00 (-2.75%)156.00 (+1.30%)152.00 (-6.17%)132.00 (+17.86%)15.99 (-33.29%)
fbf1591 — 2025-11-13 03:39:220.07 (n/a)0.05 (n/a)0.05 (n/a)0.05 (n/a)0.01 (n/a)182.00 (n/a)154.00 (n/a)162.00 (n/a)112.00 (n/a)23.97 (n/a)

gelu_8_cols_1_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.07 (+11.76%)0.06 (+11.23%)0.06 (+7.14%)0.05 (+21.52%)0.01 (-18.38%)158.00 (-17.71%)141.00 (-11.10%)140.00 (-6.67%)119.00 (-10.53%)14.70 (-39.01%)
0999a9a — 2025-11-13 19:24:080.06 (+4.51%)0.05 (+10.75%)0.05 (+16.67%)0.04 (+18.23%)0.01 (+4.18%)192.00 (-15.42%)158.60 (-9.99%)150.00 (-14.29%)133.00 (-4.32%)24.10 (-16.91%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)227.00 (n/a)176.20 (n/a)175.00 (n/a)139.00 (n/a)29.00 (n/a)

gelu_8_cols_2_channels_2048_tile_128

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.06 (+30.34%)0.05 (+30.26%)0.05 (+29.03%)0.05 (+35.44%)0.00 (-5.97%)158.00 (-26.17%)153.40 (-23.30%)155.00 (-22.50%)145.00 (-23.28%)4.50 (-47.13%)
0999a9a — 2025-11-13 19:24:080.04 (-20.63%)0.04 (-16.87%)0.04 (-20.00%)0.04 (-6.54%)0.00 (-67.29%)214.00 (+7.00%)200.00 (+19.05%)200.00 (+25.00%)189.00 (+26.00%)8.51 (-55.43%)
fbf1591 — 2025-11-13 03:39:220.05 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)200.00 (n/a)168.00 (n/a)160.00 (n/a)150.00 (n/a)19.09 (n/a)

gemm_2048x2048x2048_64x64x64_8_cols_0_bcolmaj_0_ccolmaj_0

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:423.79 (+3.32%)3.60 (+1.52%)3.67 (+3.87%)3.42 (+0.95%)0.14 (+34.74%)6845.00 (-0.94%)6515.80 (-1.43%)6381.00 (-3.73%)6184.00 (-3.21%)261.03 (+31.02%)
0999a9a — 2025-11-13 19:24:083.67 (-0.59%)3.55 (-2.49%)3.54 (-3.59%)3.39 (-3.00%)0.11 (+46.06%)6910.00 (+3.09%)6610.40 (+2.61%)6628.00 (+3.72%)6389.00 (+0.60%)199.23 (+50.27%)
fbf1591 — 2025-11-13 03:39:223.69 (n/a)3.64 (n/a)3.67 (n/a)3.50 (n/a)0.07 (n/a)6703.00 (n/a)6442.40 (n/a)6390.00 (n/a)6351.00 (n/a)132.58 (n/a)

layer_norm_1_cols_1_channels_2048_tile_2048

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.09 (-7.19%)0.08 (+16.94%)0.07 (+25.88%)0.07 (+60.89%)0.01 (-59.63%)179.00 (-37.85%)162.00 (-19.88%)170.00 (-20.56%)139.00 (+7.75%)14.72 (-72.67%)
0999a9a — 2025-11-13 19:24:080.10 (-3.88%)0.07 (-12.46%)0.06 (-21.03%)0.04 (-31.25%)0.02 (+36.72%)288.00 (+45.45%)202.20 (+19.64%)214.00 (+26.63%)129.00 (+4.03%)53.87 (+105.25%)
fbf1591 — 2025-11-13 03:39:220.10 (n/a)0.07 (n/a)0.07 (n/a)0.06 (n/a)0.01 (n/a)198.00 (n/a)169.00 (n/a)169.00 (n/a)124.00 (n/a)26.24 (n/a)

layer_norm_1_cols_2_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.09 (-0.72%)0.08 (+13.82%)0.08 (+11.25%)0.07 (+42.05%)0.01 (-47.69%)176.00 (-29.60%)155.60 (-15.25%)160.00 (-10.11%)139.00 (+0.72%)14.54 (-63.62%)
0999a9a — 2025-11-13 19:24:080.09 (-5.07%)0.07 (-8.39%)0.07 (-6.18%)0.05 (-16.40%)0.01 (+21.90%)250.00 (+19.62%)183.60 (+11.41%)178.00 (+6.59%)138.00 (+5.34%)39.97 (+52.27%)
fbf1591 — 2025-11-13 03:39:220.09 (n/a)0.08 (n/a)0.07 (n/a)0.06 (n/a)0.01 (n/a)209.00 (n/a)164.80 (n/a)167.00 (n/a)131.00 (n/a)26.25 (n/a)

layer_norm_2_cols_1_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.11 (+0.89%)0.09 (+3.41%)0.08 (+1.99%)0.07 (+0.58%)0.01 (-0.84%)172.00 (-0.58%)141.40 (-3.42%)151.00 (-1.95%)112.00 (-0.88%)21.91 (-3.96%)
0999a9a — 2025-11-13 19:24:080.11 (+10.62%)0.09 (+16.05%)0.08 (+8.44%)0.07 (+26.59%)0.01 (+2.65%)173.00 (-21.00%)146.40 (-14.49%)154.00 (-7.78%)113.00 (-9.60%)22.82 (-25.99%)
fbf1591 — 2025-11-13 03:39:220.10 (n/a)0.07 (n/a)0.07 (n/a)0.06 (n/a)0.01 (n/a)219.00 (n/a)171.20 (n/a)167.00 (n/a)125.00 (n/a)30.83 (n/a)

layer_norm_2_cols_2_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.11 (+4.27%)0.09 (+16.45%)0.09 (+34.29%)0.08 (+34.00%)0.01 (-54.70%)150.00 (-25.37%)137.80 (-17.58%)140.00 (-25.53%)117.00 (-4.10%)11.03 (-68.26%)
0999a9a — 2025-11-13 19:24:080.10 (+21.31%)0.08 (+19.49%)0.07 (+2.66%)0.06 (+20.90%)0.02 (+64.00%)201.00 (-17.28%)167.20 (-14.43%)188.00 (-2.59%)122.00 (-17.57%)34.76 (+12.90%)
fbf1591 — 2025-11-13 03:39:220.08 (n/a)0.06 (n/a)0.06 (n/a)0.05 (n/a)0.01 (n/a)243.00 (n/a)195.40 (n/a)193.00 (n/a)148.00 (n/a)30.79 (n/a)

layer_norm_4_cols_1_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.11 (+3.48%)0.08 (-2.24%)0.08 (-4.03%)0.06 (-13.57%)0.01 (+42.94%)199.00 (+15.70%)150.80 (+4.14%)149.00 (+4.20%)115.00 (-3.36%)27.73 (+62.35%)
0999a9a — 2025-11-13 19:24:080.10 (+30.25%)0.09 (+45.91%)0.09 (+54.55%)0.07 (+60.47%)0.01 (-12.42%)172.00 (-37.68%)144.80 (-33.02%)143.00 (-35.29%)119.00 (-23.23%)17.08 (-57.59%)
fbf1591 — 2025-11-13 03:39:220.08 (n/a)0.06 (n/a)0.06 (n/a)0.04 (n/a)0.01 (n/a)276.00 (n/a)216.20 (n/a)221.00 (n/a)155.00 (n/a)40.28 (n/a)

layer_norm_4_cols_2_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.12 (-4.72%)0.09 (+1.49%)0.08 (+4.14%)0.07 (-0.57%)0.02 (-14.90%)176.00 (+0.57%)141.80 (-2.21%)145.00 (-3.97%)106.00 (+4.95%)22.52 (-8.11%)
0999a9a — 2025-11-13 19:24:080.12 (+50.49%)0.09 (+47.35%)0.08 (+51.66%)0.07 (+66.86%)0.02 (+21.43%)175.00 (-40.07%)145.00 (-33.79%)151.00 (-34.06%)101.00 (-33.55%)24.50 (-52.68%)
fbf1591 — 2025-11-13 03:39:220.08 (n/a)0.06 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)292.00 (n/a)219.00 (n/a)229.00 (n/a)152.00 (n/a)51.78 (n/a)

layer_norm_8_cols_1_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.11 (+46.55%)0.09 (+57.56%)0.09 (+33.33%)0.07 (+105.23%)0.01 (-20.47%)172.00 (-51.27%)140.20 (-40.54%)138.00 (-25.00%)116.00 (-31.76%)19.39 (-73.18%)
0999a9a — 2025-11-13 19:24:080.07 (-31.76%)0.06 (-27.11%)0.07 (-16.85%)0.03 (-25.21%)0.01 (-24.96%)353.00 (+33.71%)235.80 (+38.06%)184.00 (+20.26%)170.00 (+46.55%)72.33 (+41.07%)
fbf1591 — 2025-11-13 03:39:220.11 (n/a)0.08 (n/a)0.08 (n/a)0.05 (n/a)0.02 (n/a)264.00 (n/a)170.80 (n/a)153.00 (n/a)116.00 (n/a)51.27 (n/a)

layer_norm_8_cols_2_channels_2048_tile_128

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.08 (+6.92%)0.07 (+1.29%)0.06 (-7.35%)0.06 (-2.40%)0.01 (+71.92%)208.00 (+2.46%)189.60 (-0.52%)204.00 (+7.94%)159.00 (-6.47%)19.83 (+66.00%)
0999a9a — 2025-11-13 19:24:080.07 (+0.00%)0.06 (+2.85%)0.07 (+1.59%)0.06 (+23.15%)0.00 (-51.87%)203.00 (-18.80%)190.60 (-4.41%)189.00 (-1.56%)170.00 (+0.00%)11.94 (-60.19%)
fbf1591 — 2025-11-13 03:39:220.07 (n/a)0.06 (n/a)0.06 (n/a)0.05 (n/a)0.01 (n/a)250.00 (n/a)199.40 (n/a)192.00 (n/a)170.00 (n/a)30.00 (n/a)

matrix_vector_mul_128x128_32_1col

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)GFLOP/s (max)GFLOP/s (mean)GFLOP/s (median)GFLOP/s (min)GFLOP/s (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.05 (-4.83%)0.04 (-1.87%)0.04 (+2.83%)0.04 (-0.48%)0.00 (-37.77%)0.05 (-4.83%)0.04 (-1.87%)0.04 (+2.83%)0.04 (-0.48%)0.00 (-37.77%)831.00 (+0.48%)774.20 (+1.55%)777.00 (-2.75%)725.00 (+5.07%)38.45 (-33.65%)
0999a9a — 2025-11-13 19:24:080.05 (+4.35%)0.04 (+3.02%)0.04 (-0.25%)0.04 (+0.60%)0.00 (+62.83%)0.05 (+4.35%)0.04 (+3.02%)0.04 (-0.25%)0.04 (+0.60%)0.00 (+62.83%)827.00 (-0.60%)762.40 (-2.58%)799.00 (+0.25%)690.00 (-4.17%)57.94 (+53.99%)
fbf1591 — 2025-11-13 03:39:220.05 (n/a)0.04 (n/a)0.04 (n/a)0.04 (n/a)0.00 (n/a)0.05 (n/a)0.04 (n/a)0.04 (n/a)0.04 (n/a)0.00 (n/a)832.00 (n/a)782.60 (n/a)797.00 (n/a)720.00 (n/a)37.63 (n/a)

matrix_vector_mul_2048x8192_1_1col

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)GFLOP/s (max)GFLOP/s (mean)GFLOP/s (median)GFLOP/s (min)GFLOP/s (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:4211.20 (+1.10%)10.68 (-0.80%)10.64 (-1.77%)10.21 (-1.09%)0.35 (+29.57%)11.19 (+1.10%)10.67 (-0.80%)10.63 (-1.77%)10.20 (-1.09%)0.35 (+29.56%)3289.00 (+1.11%)3147.20 (+0.85%)3156.00 (+1.81%)2999.00 (-1.09%)102.49 (+29.42%)
0999a9a — 2025-11-13 19:24:0811.07 (-2.77%)10.77 (-0.84%)10.83 (+0.32%)10.32 (+0.46%)0.27 (-25.80%)11.07 (-2.77%)10.76 (-0.84%)10.82 (+0.32%)10.31 (+0.46%)0.27 (-25.79%)3253.00 (-0.46%)3120.80 (+0.79%)3100.00 (-0.32%)3032.00 (+2.85%)79.19 (-24.14%)
fbf1591 — 2025-11-13 03:39:2211.39 (n/a)10.86 (n/a)10.80 (n/a)10.27 (n/a)0.36 (n/a)11.38 (n/a)10.85 (n/a)10.79 (n/a)10.27 (n/a)0.36 (n/a)3268.00 (n/a)3096.20 (n/a)3110.00 (n/a)2948.00 (n/a)104.38 (n/a)

matrix_vector_mul_2048x8192_1_2col

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)GFLOP/s (max)GFLOP/s (mean)GFLOP/s (median)GFLOP/s (min)GFLOP/s (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:4218.03 (+0.91%)17.11 (-0.49%)16.87 (-1.96%)16.35 (-1.36%)0.60 (+39.05%)18.02 (+0.91%)17.09 (-0.49%)16.86 (-1.96%)16.34 (-1.36%)0.60 (+39.05%)2053.00 (+1.38%)1965.20 (+0.55%)1990.00 (+2.00%)1862.00 (-0.90%)67.79 (+39.43%)
0999a9a — 2025-11-13 19:24:0817.87 (+7.66%)17.19 (+6.72%)17.21 (+7.12%)16.58 (+6.07%)0.43 (+33.19%)17.86 (+7.66%)17.18 (+6.72%)17.20 (+7.12%)16.57 (+6.07%)0.43 (+33.18%)2025.00 (-5.73%)1954.40 (-6.27%)1951.00 (-6.65%)1879.00 (-7.12%)48.62 (+16.62%)
fbf1591 — 2025-11-13 03:39:2216.60 (n/a)16.11 (n/a)16.06 (n/a)15.63 (n/a)0.32 (n/a)16.59 (n/a)16.10 (n/a)16.05 (n/a)15.62 (n/a)0.32 (n/a)2148.00 (n/a)2085.20 (n/a)2090.00 (n/a)2023.00 (n/a)41.69 (n/a)

matrix_vector_mul_2048x8192_1_4col

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)GFLOP/s (max)GFLOP/s (mean)GFLOP/s (median)GFLOP/s (min)GFLOP/s (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:4222.86 (+5.45%)21.58 (+6.59%)21.94 (+6.67%)19.73 (+11.87%)1.07 (-23.63%)22.84 (+5.45%)21.56 (+6.59%)21.93 (+6.67%)19.71 (+11.87%)1.07 (-23.64%)1702.00 (-10.61%)1560.00 (-6.43%)1530.00 (-6.25%)1469.00 (-5.16%)80.47 (-35.66%)
0999a9a — 2025-11-13 19:24:0821.68 (+5.04%)20.24 (+6.30%)20.57 (+9.74%)17.63 (-4.10%)1.41 (+69.34%)21.66 (+5.04%)20.23 (+6.30%)20.56 (+9.74%)17.62 (-4.10%)1.41 (+69.35%)1904.00 (+4.27%)1667.20 (-5.61%)1632.00 (-8.88%)1549.00 (-4.79%)125.07 (+70.83%)
fbf1591 — 2025-11-13 03:39:2220.64 (n/a)19.04 (n/a)18.75 (n/a)18.39 (n/a)0.83 (n/a)20.62 (n/a)19.03 (n/a)18.73 (n/a)18.38 (n/a)0.83 (n/a)1826.00 (n/a)1766.20 (n/a)1791.00 (n/a)1627.00 (n/a)73.21 (n/a)

matrix_vector_mul_2048x8192_1_8col

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)GFLOP/s (max)GFLOP/s (mean)GFLOP/s (median)GFLOP/s (min)GFLOP/s (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:4218.45 (+4.67%)15.99 (-5.30%)15.85 (-7.46%)14.72 (-7.15%)1.31 (+112.58%)18.44 (+4.67%)15.98 (-5.30%)15.84 (-7.46%)14.71 (-7.15%)1.31 (+112.57%)2281.00 (+7.70%)2113.00 (+6.11%)2118.00 (+8.06%)1820.00 (-4.46%)159.87 (+115.34%)
0999a9a — 2025-11-13 19:24:0817.62 (-9.34%)16.88 (+0.64%)17.13 (+7.09%)15.85 (+17.61%)0.62 (-73.17%)17.61 (-9.34%)16.87 (+0.64%)17.12 (+7.09%)15.84 (+17.61%)0.62 (-73.17%)2118.00 (-14.97%)1991.40 (-2.39%)1960.00 (-6.62%)1905.00 (+10.31%)74.24 (-73.99%)
fbf1591 — 2025-11-13 03:39:2219.44 (n/a)16.78 (n/a)16.00 (n/a)13.48 (n/a)2.29 (n/a)19.43 (n/a)16.77 (n/a)15.99 (n/a)13.47 (n/a)2.29 (n/a)2491.00 (n/a)2040.20 (n/a)2099.00 (n/a)1727.00 (n/a)285.41 (n/a)

matrix_vector_mul_8192x2048_4_1col

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)GFLOP/s (max)GFLOP/s (mean)GFLOP/s (median)GFLOP/s (min)GFLOP/s (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:4211.05 (-1.25%)10.91 (-1.47%)10.91 (-1.33%)10.76 (-1.99%)0.10 (+37.72%)11.04 (-1.25%)10.91 (-1.47%)10.90 (-1.33%)10.76 (-1.99%)0.10 (+37.71%)3119.00 (+2.03%)3076.60 (+1.49%)3078.00 (+1.35%)3039.00 (+1.27%)27.73 (+42.29%)
0999a9a — 2025-11-13 19:24:0811.19 (-1.03%)11.08 (+3.60%)11.06 (+2.44%)10.98 (+13.09%)0.07 (-86.64%)11.18 (-1.03%)11.07 (+3.60%)11.05 (+2.44%)10.98 (+13.09%)0.07 (-86.64%)3057.00 (-11.57%)3031.40 (-3.72%)3037.00 (-2.38%)3001.00 (+1.04%)19.49 (-88.20%)
fbf1591 — 2025-11-13 03:39:2211.30 (n/a)10.69 (n/a)10.79 (n/a)9.71 (n/a)0.53 (n/a)11.30 (n/a)10.68 (n/a)10.79 (n/a)9.71 (n/a)0.53 (n/a)3457.00 (n/a)3148.60 (n/a)3111.00 (n/a)2970.00 (n/a)165.20 (n/a)

matrix_vector_mul_8192x2048_4_2col

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)GFLOP/s (max)GFLOP/s (mean)GFLOP/s (median)GFLOP/s (min)GFLOP/s (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:4217.71 (+2.16%)17.21 (+2.28%)17.17 (+0.56%)16.79 (+4.10%)0.30 (-38.04%)17.70 (+2.16%)17.20 (+2.28%)17.15 (+0.56%)16.78 (+4.10%)0.30 (-38.03%)2000.00 (-3.94%)1951.00 (-2.28%)1956.00 (-0.56%)1896.00 (-2.12%)33.36 (-41.81%)
0999a9a — 2025-11-13 19:24:0817.33 (+0.67%)16.83 (+5.22%)17.07 (+7.78%)16.13 (+11.53%)0.48 (-49.24%)17.32 (+0.67%)16.82 (+5.22%)17.06 (+7.78%)16.12 (+11.53%)0.48 (-49.24%)2082.00 (-10.34%)1996.60 (-5.22%)1967.00 (-7.22%)1937.00 (-0.67%)57.33 (-54.90%)
fbf1591 — 2025-11-13 03:39:2217.22 (n/a)15.99 (n/a)15.84 (n/a)14.46 (n/a)0.94 (n/a)17.21 (n/a)15.98 (n/a)15.83 (n/a)14.45 (n/a)0.94 (n/a)2322.00 (n/a)2106.60 (n/a)2120.00 (n/a)1950.00 (n/a)127.11 (n/a)

matrix_vector_mul_8192x2048_4_4col

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)GFLOP/s (max)GFLOP/s (mean)GFLOP/s (median)GFLOP/s (min)GFLOP/s (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:4222.10 (+8.16%)21.24 (+9.96%)21.05 (+8.53%)20.83 (+19.23%)0.45 (-57.74%)22.09 (+8.16%)21.22 (+9.96%)21.04 (+8.53%)20.82 (+19.23%)0.45 (-57.74%)1612.00 (-16.13%)1581.60 (-9.30%)1595.00 (-7.86%)1519.00 (-7.55%)32.64 (-67.39%)
0999a9a — 2025-11-13 19:24:0820.44 (+1.76%)19.31 (+0.88%)19.40 (+1.56%)17.47 (-4.73%)1.06 (+82.49%)20.42 (+1.77%)19.30 (+0.88%)19.38 (+1.56%)17.46 (-4.73%)1.06 (+82.49%)1922.00 (+4.97%)1743.80 (-0.65%)1731.00 (-1.54%)1643.00 (-1.73%)100.11 (+88.57%)
fbf1591 — 2025-11-13 03:39:2220.08 (n/a)19.15 (n/a)19.10 (n/a)18.34 (n/a)0.58 (n/a)20.07 (n/a)19.13 (n/a)19.09 (n/a)18.33 (n/a)0.58 (n/a)1831.00 (n/a)1755.20 (n/a)1758.00 (n/a)1672.00 (n/a)53.09 (n/a)

matrix_vector_mul_8192x2048_4_8col

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)GFLOP/s (max)GFLOP/s (mean)GFLOP/s (median)GFLOP/s (min)GFLOP/s (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:4219.54 (+2.74%)17.71 (+3.40%)18.33 (+11.68%)15.30 (-2.10%)1.65 (+21.23%)19.53 (+2.74%)17.70 (+3.40%)18.32 (+11.68%)15.29 (-2.10%)1.65 (+21.23%)2195.00 (+2.14%)1913.40 (-3.01%)1832.00 (-10.46%)1718.00 (-2.66%)185.23 (+21.14%)
0999a9a — 2025-11-13 19:24:0819.02 (+1.64%)17.12 (-4.33%)16.41 (-9.04%)15.62 (-3.86%)1.36 (+57.46%)19.01 (+1.64%)17.11 (-4.33%)16.40 (-9.04%)15.61 (-3.86%)1.36 (+57.46%)2149.00 (+4.02%)1972.80 (+4.91%)2046.00 (+9.94%)1765.00 (-1.62%)152.91 (+58.58%)
fbf1591 — 2025-11-13 03:39:2218.72 (n/a)17.90 (n/a)18.04 (n/a)16.25 (n/a)0.86 (n/a)18.70 (n/a)17.89 (n/a)18.03 (n/a)16.24 (n/a)0.86 (n/a)2066.00 (n/a)1880.40 (n/a)1861.00 (n/a)1794.00 (n/a)96.42 (n/a)

mem_copy_16_cores_2_chans_2048_tile_128_False

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.05 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.00 (n/a)193.00 (n/a)167.80 (n/a)162.00 (n/a)156.00 (n/a)13.67 (n/a)

mem_copy_1_cols_1_channels_2048_tile_2048

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
0999a9a — 2025-11-13 19:24:080.06 (+5.51%)0.05 (+5.10%)0.05 (-6.58%)0.04 (+18.29%)0.01 (-17.12%)175.00 (-15.46%)150.40 (-6.00%)152.00 (+7.04%)127.00 (-5.22%)18.28 (-33.91%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)207.00 (n/a)160.00 (n/a)142.00 (n/a)134.00 (n/a)27.66 (n/a)

mem_copy_1_cols_2_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
0999a9a — 2025-11-13 19:24:080.06 (-9.02%)0.05 (-8.21%)0.05 (+2.82%)0.03 (-34.51%)0.01 (+94.29%)226.00 (+52.70%)158.80 (+12.46%)142.00 (-2.74%)133.00 (+9.92%)34.49 (+235.98%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.05 (n/a)0.05 (n/a)0.05 (n/a)0.00 (n/a)148.00 (n/a)141.20 (n/a)146.00 (n/a)121.00 (n/a)10.26 (n/a)

mem_copy_1_cores_1_chans_2048_tile_2048_False

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.05 (n/a)0.04 (n/a)0.04 (n/a)0.03 (n/a)0.01 (n/a)257.00 (n/a)190.20 (n/a)180.00 (n/a)166.00 (n/a)34.07 (n/a)

mem_copy_2_cols_1_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
0999a9a — 2025-11-13 19:24:080.07 (+18.10%)0.06 (+4.90%)0.06 (+5.47%)0.05 (-7.36%)0.01 (+117.26%)163.00 (+7.95%)133.60 (-2.91%)128.00 (-5.19%)105.00 (-15.32%)20.58 (+98.52%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.06 (n/a)0.06 (n/a)0.05 (n/a)0.00 (n/a)151.00 (n/a)137.60 (n/a)135.00 (n/a)124.00 (n/a)10.37 (n/a)

mem_copy_2_cols_2_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
0999a9a — 2025-11-13 19:24:080.08 (+26.80%)0.06 (+22.42%)0.06 (+33.33%)0.05 (+43.98%)0.01 (+1.58%)166.00 (-30.54%)138.00 (-20.32%)138.00 (-25.00%)97.00 (-21.14%)23.37 (-44.22%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.05 (n/a)0.04 (n/a)0.03 (n/a)0.01 (n/a)239.00 (n/a)173.20 (n/a)184.00 (n/a)123.00 (n/a)41.89 (n/a)

mem_copy_2_cores_1_chans_2048_tile_1024_False

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)192.00 (n/a)153.00 (n/a)141.00 (n/a)131.00 (n/a)21.83 (n/a)

mem_copy_2_cores_2_chans_2048_tile_1024_False

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)209.00 (n/a)162.00 (n/a)154.00 (n/a)131.00 (n/a)26.85 (n/a)

mem_copy_4_cols_1_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
0999a9a — 2025-11-13 19:24:080.06 (-12.98%)0.05 (-13.35%)0.05 (-10.27%)0.04 (-19.21%)0.01 (+7.53%)203.00 (+23.78%)156.80 (+16.49%)146.00 (+11.45%)131.00 (+14.91%)25.69 (+52.83%)
fbf1591 — 2025-11-13 03:39:220.07 (n/a)0.06 (n/a)0.06 (n/a)0.05 (n/a)0.01 (n/a)164.00 (n/a)134.60 (n/a)131.00 (n/a)114.00 (n/a)16.81 (n/a)

mem_copy_4_cols_2_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
0999a9a — 2025-11-13 19:24:080.06 (+33.83%)0.05 (+30.33%)0.05 (+29.93%)0.04 (+27.00%)0.01 (+42.50%)200.00 (-21.26%)154.60 (-23.01%)147.00 (-23.04%)133.00 (-25.28%)23.48 (-14.85%)
fbf1591 — 2025-11-13 03:39:220.04 (n/a)0.04 (n/a)0.04 (n/a)0.03 (n/a)0.00 (n/a)254.00 (n/a)200.80 (n/a)191.00 (n/a)178.00 (n/a)27.58 (n/a)

mem_copy_4_cores_1_chans_2048_tile_512_False

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.00 (n/a)172.00 (n/a)145.00 (n/a)142.00 (n/a)130.00 (n/a)14.75 (n/a)

mem_copy_4_cores_2_chans_2048_tile_512_False

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.07 (n/a)0.05 (n/a)0.05 (n/a)0.03 (n/a)0.01 (n/a)293.00 (n/a)175.40 (n/a)164.00 (n/a)110.00 (n/a)63.30 (n/a)

mem_copy_8_cols_1_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
0999a9a — 2025-11-13 19:24:080.06 (+29.51%)0.06 (+23.73%)0.06 (+26.77%)0.04 (+1.00%)0.01 (+118.73%)200.00 (-0.99%)142.00 (-17.15%)127.00 (-21.12%)122.00 (-22.78%)29.37 (+72.65%)
fbf1591 — 2025-11-13 03:39:220.05 (n/a)0.04 (n/a)0.05 (n/a)0.04 (n/a)0.00 (n/a)202.00 (n/a)171.40 (n/a)161.00 (n/a)158.00 (n/a)17.01 (n/a)

mem_copy_8_cols_2_channels_2048_tile_128

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
0999a9a — 2025-11-13 19:24:080.05 (-1.86%)0.04 (+3.84%)0.04 (-0.95%)0.04 (-0.92%)0.01 (+18.78%)217.00 (+0.93%)193.20 (-3.11%)211.00 (+0.96%)161.00 (+1.90%)25.96 (+23.15%)
fbf1591 — 2025-11-13 03:39:220.05 (n/a)0.04 (n/a)0.04 (n/a)0.04 (n/a)0.00 (n/a)215.00 (n/a)199.40 (n/a)209.00 (n/a)158.00 (n/a)21.08 (n/a)

mem_copy_8_cores_1_chans_2048_tile_256_False

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.05 (n/a)0.04 (n/a)0.04 (n/a)0.04 (n/a)0.01 (n/a)211.00 (n/a)176.00 (n/a)173.00 (n/a)150.00 (n/a)23.71 (n/a)

mem_copy_8_cores_2_chans_2048_tile_256_False

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.05 (n/a)0.04 (n/a)0.04 (n/a)0.04 (n/a)0.00 (n/a)198.00 (n/a)174.00 (n/a)171.00 (n/a)149.00 (n/a)15.92 (n/a)

mha

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)GFLOP/s (max)GFLOP/s (mean)GFLOP/s (median)GFLOP/s (min)GFLOP/s (stddev)
3e0443f — 2025-11-13 21:21:420.10 (+15.33%)0.09 (+3.67%)0.09 (-0.09%)0.08 (+0.33%)0.01 (+137.78%)74.31 (+15.33%)65.03 (+3.67%)63.81 (-0.09%)58.80 (+0.33%)5.07 (+137.78%)
0999a9a — 2025-11-13 19:24:080.09 (-3.49%)0.09 (-2.49%)0.09 (+0.19%)0.08 (-6.78%)0.00 (+57.78%)64.43 (-3.49%)62.73 (-2.49%)63.87 (+0.19%)58.61 (-6.78%)2.13 (+57.77%)
fbf1591 — 2025-11-13 03:39:220.09 (n/a)0.09 (n/a)0.09 (n/a)0.09 (n/a)0.00 (n/a)66.76 (n/a)64.33 (n/a)63.75 (n/a)62.87 (n/a)1.35 (n/a)

relu_1_cols_1_channels_2048_tile_2048

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.06 (+18.70%)0.06 (+35.55%)0.06 (+37.10%)0.04 (+40.34%)0.01 (-0.42%)176.00 (-28.74%)134.40 (-27.19%)124.00 (-27.06%)123.00 (-15.75%)20.83 (-40.93%)
0999a9a — 2025-11-13 19:24:080.05 (-8.90%)0.04 (-12.92%)0.04 (-7.65%)0.03 (-17.00%)0.01 (-3.29%)247.00 (+20.49%)184.60 (+15.66%)170.00 (+8.28%)146.00 (+9.77%)35.26 (+32.92%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)205.00 (n/a)159.60 (n/a)157.00 (n/a)133.00 (n/a)26.53 (n/a)

relu_2_cols_1_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.05 (-7.75%)0.05 (-7.93%)0.05 (-6.88%)0.04 (-10.80%)0.00 (+4.77%)176.00 (+12.10%)159.00 (+8.76%)160.00 (+7.38%)142.00 (+8.40%)11.52 (+27.76%)
0999a9a — 2025-11-13 19:24:080.06 (-2.29%)0.05 (-3.00%)0.05 (-5.37%)0.05 (+3.82%)0.00 (-22.25%)157.00 (-3.68%)146.20 (+2.81%)149.00 (+5.67%)131.00 (+2.34%)9.02 (-24.35%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.05 (n/a)0.05 (n/a)0.05 (n/a)0.00 (n/a)163.00 (n/a)142.20 (n/a)141.00 (n/a)128.00 (n/a)11.92 (n/a)

relu_4_cols_1_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.05 (-15.34%)0.04 (-12.05%)0.05 (-12.50%)0.03 (-17.32%)0.01 (-15.78%)231.00 (+20.94%)180.00 (+13.78%)168.00 (+14.29%)163.00 (+18.12%)25.63 (+23.45%)
0999a9a — 2025-11-13 19:24:080.06 (-2.17%)0.05 (+19.50%)0.05 (+11.56%)0.04 (+60.73%)0.01 (-53.04%)191.00 (-37.79%)158.20 (-24.16%)147.00 (-10.37%)138.00 (+2.22%)20.76 (-71.43%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.04 (n/a)0.05 (n/a)0.02 (n/a)0.01 (n/a)307.00 (n/a)208.60 (n/a)164.00 (n/a)135.00 (n/a)72.66 (n/a)

relu_8_cols_1_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.05 (-7.38%)0.05 (-6.59%)0.05 (-11.88%)0.03 (-16.60%)0.01 (-8.20%)241.00 (+19.90%)174.20 (+7.40%)160.00 (+13.48%)149.00 (+7.97%)34.34 (+21.90%)
0999a9a — 2025-11-13 19:24:080.06 (-0.72%)0.05 (+0.25%)0.05 (+13.48%)0.04 (-15.42%)0.01 (+99.76%)201.00 (+18.24%)162.20 (+2.01%)141.00 (-11.88%)138.00 (+0.73%)28.17 (+137.04%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.00 (n/a)170.00 (n/a)159.00 (n/a)160.00 (n/a)137.00 (n/a)11.88 (n/a)

rms_norm_1_cols_1_channels_2048_tile_2048

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.11 (+10.68%)0.09 (+1.34%)0.09 (+1.61%)0.07 (-9.80%)0.01 (+64.53%)153.00 (+10.87%)124.80 (-0.32%)124.00 (-1.59%)103.00 (-9.65%)16.09 (+68.73%)
0999a9a — 2025-11-13 19:24:080.10 (+12.28%)0.09 (+14.91%)0.09 (+8.73%)0.08 (+26.09%)0.01 (-17.60%)138.00 (-20.69%)125.20 (-13.54%)126.00 (-8.03%)114.00 (-10.94%)9.54 (-42.73%)
fbf1591 — 2025-11-13 03:39:220.09 (n/a)0.08 (n/a)0.08 (n/a)0.07 (n/a)0.01 (n/a)174.00 (n/a)144.80 (n/a)137.00 (n/a)128.00 (n/a)16.65 (n/a)

rms_norm_1_cols_2_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.10 (+2.73%)0.08 (-3.87%)0.08 (-13.48%)0.07 (+26.74%)0.01 (-29.51%)172.00 (-21.10%)138.80 (+0.14%)141.00 (+15.57%)110.00 (-2.65%)20.87 (-47.74%)
0999a9a — 2025-11-13 19:24:080.10 (+13.27%)0.09 (+21.98%)0.09 (+26.23%)0.05 (+14.68%)0.02 (+21.77%)218.00 (-12.80%)138.60 (-17.40%)122.00 (-20.78%)113.00 (-11.72%)39.94 (-7.15%)
fbf1591 — 2025-11-13 03:39:220.09 (n/a)0.07 (n/a)0.07 (n/a)0.05 (n/a)0.01 (n/a)250.00 (n/a)167.80 (n/a)154.00 (n/a)128.00 (n/a)43.02 (n/a)

rms_norm_2_cols_1_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.10 (-15.32%)0.08 (-12.64%)0.08 (-19.58%)0.06 (+25.82%)0.01 (-44.50%)182.00 (-20.52%)144.60 (+7.11%)143.00 (+24.35%)111.00 (+18.09%)23.00 (-52.04%)
0999a9a — 2025-11-13 19:24:080.12 (+47.87%)0.09 (+28.65%)0.10 (+46.09%)0.05 (-25.76%)0.02 (+304.80%)229.00 (+34.71%)135.00 (-15.31%)115.00 (-31.55%)94.00 (-32.37%)47.96 (+296.61%)
fbf1591 — 2025-11-13 03:39:220.08 (n/a)0.07 (n/a)0.07 (n/a)0.07 (n/a)0.01 (n/a)170.00 (n/a)159.40 (n/a)168.00 (n/a)139.00 (n/a)12.09 (n/a)

rms_norm_2_cols_2_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.09 (-5.38%)0.07 (-19.54%)0.08 (-13.89%)0.05 (-45.06%)0.01 (+859.20%)233.00 (+82.03%)161.80 (+29.65%)144.00 (+16.13%)130.00 (+5.69%)37.08 (+1812.00%)
0999a9a — 2025-11-13 19:24:080.09 (-14.63%)0.09 (+21.34%)0.09 (+29.03%)0.09 (+88.28%)0.00 (-92.92%)128.00 (-46.89%)124.80 (-23.25%)124.00 (-22.50%)123.00 (+17.14%)1.94 (-95.64%)
fbf1591 — 2025-11-13 03:39:220.11 (n/a)0.08 (n/a)0.07 (n/a)0.05 (n/a)0.02 (n/a)241.00 (n/a)162.60 (n/a)160.00 (n/a)105.00 (n/a)44.50 (n/a)

rms_norm_4_cols_1_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.11 (+20.59%)0.07 (-23.22%)0.07 (-23.46%)0.04 (-52.47%)0.02 (+4167.78%)263.00 (+110.40%)179.20 (+44.75%)162.00 (+30.65%)102.00 (-17.07%)56.06 (+7391.61%)
0999a9a — 2025-11-13 19:24:080.09 (-0.81%)0.09 (+14.93%)0.09 (+16.13%)0.09 (+35.20%)0.00 (-94.54%)125.00 (-26.04%)123.80 (-14.38%)124.00 (-13.89%)123.00 (+0.82%)0.75 (-95.92%)
fbf1591 — 2025-11-13 03:39:220.09 (n/a)0.08 (n/a)0.08 (n/a)0.07 (n/a)0.01 (n/a)169.00 (n/a)144.60 (n/a)144.00 (n/a)122.00 (n/a)18.36 (n/a)

rms_norm_4_cols_2_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.08 (-5.84%)0.07 (-1.98%)0.08 (-4.67%)0.06 (+32.31%)0.01 (-44.36%)195.00 (-24.42%)156.60 (-2.97%)150.00 (+4.90%)137.00 (+6.20%)21.17 (-56.63%)
0999a9a — 2025-11-13 19:24:080.09 (+7.75%)0.08 (+6.49%)0.08 (+6.29%)0.04 (-1.94%)0.02 (+21.12%)258.00 (+1.98%)161.40 (-4.38%)143.00 (-5.92%)129.00 (-7.19%)48.81 (+14.45%)
fbf1591 — 2025-11-13 03:39:220.08 (n/a)0.07 (n/a)0.08 (n/a)0.05 (n/a)0.01 (n/a)253.00 (n/a)168.80 (n/a)152.00 (n/a)139.00 (n/a)42.65 (n/a)

rms_norm_8_cols_1_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.07 (-13.77%)0.06 (-19.44%)0.06 (-10.58%)0.05 (-28.22%)0.01 (+45.01%)241.00 (+39.31%)203.00 (+25.93%)189.00 (+11.83%)167.00 (+15.97%)29.18 (+139.75%)
0999a9a — 2025-11-13 19:24:080.08 (-3.47%)0.07 (-2.98%)0.07 (-10.06%)0.07 (-1.16%)0.01 (-2.33%)173.00 (+1.17%)161.20 (+3.07%)169.00 (+11.18%)144.00 (+3.60%)12.17 (+1.21%)
fbf1591 — 2025-11-13 03:39:220.08 (n/a)0.07 (n/a)0.08 (n/a)0.07 (n/a)0.01 (n/a)171.00 (n/a)156.40 (n/a)152.00 (n/a)139.00 (n/a)12.03 (n/a)

rms_norm_8_cols_2_channels_2048_tile_128

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.06 (-31.84%)0.05 (-33.62%)0.05 (-27.91%)0.04 (-46.49%)0.01 (+54.93%)299.00 (+86.88%)233.80 (+53.21%)215.00 (+38.71%)201.00 (+46.72%)35.11 (+338.10%)
0999a9a — 2025-11-13 19:24:080.08 (+11.68%)0.08 (+41.18%)0.07 (+32.90%)0.07 (+107.50%)0.00 (-70.36%)160.00 (-51.81%)152.60 (-34.17%)155.00 (-24.76%)137.00 (-10.46%)8.01 (-87.63%)
fbf1591 — 2025-11-13 03:39:220.07 (n/a)0.05 (n/a)0.06 (n/a)0.03 (n/a)0.01 (n/a)332.00 (n/a)231.80 (n/a)206.00 (n/a)153.00 (n/a)64.77 (n/a)

rope_1_cols_2_channels_4096_tile_4096_0

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.18 (-10.40%)0.14 (-13.55%)0.13 (-11.63%)0.11 (-21.60%)0.03 (+12.16%)213.00 (+27.54%)171.40 (+17.40%)172.00 (+13.16%)125.00 (+11.61%)30.02 (+63.66%)
0999a9a — 2025-11-13 19:24:080.20 (+24.11%)0.16 (+28.34%)0.15 (+15.13%)0.14 (+55.09%)0.02 (-16.27%)167.00 (-35.52%)146.00 (-24.66%)152.00 (-13.14%)112.00 (-19.42%)18.34 (-58.89%)
fbf1591 — 2025-11-13 03:39:220.16 (n/a)0.12 (n/a)0.13 (n/a)0.09 (n/a)0.03 (n/a)259.00 (n/a)193.80 (n/a)175.00 (n/a)139.00 (n/a)44.62 (n/a)

rope_2_cols_2_channels_4096_tile_2048_0

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.17 (+0.00%)0.15 (+1.02%)0.15 (+0.00%)0.12 (-3.23%)0.02 (+21.11%)186.00 (+3.33%)152.20 (-0.52%)152.00 (+0.00%)132.00 (+0.00%)19.08 (+23.80%)
0999a9a — 2025-11-13 19:24:080.17 (-12.88%)0.15 (-7.72%)0.15 (-1.97%)0.13 (-10.56%)0.01 (-32.96%)180.00 (+11.80%)153.00 (+7.59%)152.00 (+2.01%)132.00 (+14.78%)15.41 (-13.58%)
fbf1591 — 2025-11-13 03:39:220.20 (n/a)0.16 (n/a)0.15 (n/a)0.14 (n/a)0.02 (n/a)161.00 (n/a)142.20 (n/a)149.00 (n/a)115.00 (n/a)17.84 (n/a)

rope_4_cols_2_channels_4096_tile_1024_0

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.17 (+0.00%)0.14 (-4.50%)0.14 (-7.23%)0.10 (-9.87%)0.02 (-2.05%)223.00 (+10.95%)171.60 (+4.89%)166.00 (+7.79%)133.00 (+0.00%)31.59 (+8.59%)
0999a9a — 2025-11-13 19:24:080.17 (-7.52%)0.14 (-7.80%)0.15 (-11.04%)0.11 (+8.96%)0.02 (-10.02%)201.00 (-8.22%)163.60 (+7.49%)154.00 (+12.41%)133.00 (+8.13%)29.09 (-14.80%)
fbf1591 — 2025-11-13 03:39:220.19 (n/a)0.16 (n/a)0.17 (n/a)0.10 (n/a)0.03 (n/a)219.00 (n/a)152.20 (n/a)137.00 (n/a)123.00 (n/a)34.14 (n/a)

rope_8_cols_2_channels_4096_tile_512_0

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.16 (+8.90%)0.13 (+12.03%)0.13 (+12.71%)0.11 (+13.93%)0.02 (-2.36%)201.00 (-12.23%)179.60 (-11.09%)181.00 (-11.27%)146.00 (-8.18%)19.93 (-21.14%)
0999a9a — 2025-11-13 19:24:080.14 (+13.21%)0.12 (+2.72%)0.11 (+0.49%)0.10 (+3.49%)0.02 (+58.92%)229.00 (-3.38%)202.00 (-1.75%)204.00 (-0.49%)159.00 (-11.67%)25.27 (+34.93%)
fbf1591 — 2025-11-13 03:39:220.13 (n/a)0.11 (n/a)0.11 (n/a)0.10 (n/a)0.01 (n/a)237.00 (n/a)205.60 (n/a)205.00 (n/a)180.00 (n/a)18.73 (n/a)

silu_1_cols_1_channels_2048_tile_2048

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.06 (+11.38%)0.05 (+12.36%)0.05 (+9.21%)0.05 (+16.87%)0.01 (+2.92%)160.00 (-14.44%)146.00 (-11.19%)152.00 (-8.43%)123.00 (-10.22%)14.04 (-20.43%)
0999a9a — 2025-11-13 19:24:080.06 (+1.46%)0.05 (+0.63%)0.05 (-1.20%)0.04 (+11.76%)0.01 (-15.18%)187.00 (-10.53%)164.40 (-1.32%)166.00 (+1.22%)137.00 (-1.44%)17.65 (-26.23%)
fbf1591 — 2025-11-13 03:39:220.05 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)209.00 (n/a)166.60 (n/a)164.00 (n/a)139.00 (n/a)23.92 (n/a)

silu_2_cols_1_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.06 (-9.23%)0.04 (-11.70%)0.04 (-8.24%)0.03 (-14.96%)0.01 (+5.59%)234.00 (+17.59%)178.20 (+14.67%)170.00 (+8.97%)130.00 (+10.17%)35.86 (+37.60%)
0999a9a — 2025-11-13 19:24:080.06 (+15.25%)0.05 (+3.17%)0.05 (-5.13%)0.04 (+15.08%)0.01 (+3.79%)199.00 (-13.10%)155.40 (-3.84%)156.00 (+5.41%)118.00 (-13.24%)26.06 (-24.10%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.05 (n/a)0.05 (n/a)0.03 (n/a)0.01 (n/a)229.00 (n/a)161.60 (n/a)148.00 (n/a)136.00 (n/a)34.33 (n/a)

silu_4_cols_1_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.05 (-8.45%)0.04 (-23.74%)0.04 (-30.39%)0.02 (-46.74%)0.01 (+30.87%)368.00 (+87.76%)225.60 (+41.00%)204.00 (+43.66%)142.00 (+9.23%)78.76 (+165.75%)
0999a9a — 2025-11-13 19:24:080.06 (+4.62%)0.05 (-3.58%)0.05 (+0.70%)0.04 (-12.24%)0.01 (+100.17%)196.00 (+13.95%)160.00 (+6.38%)142.00 (-0.70%)130.00 (-4.41%)29.64 (+124.27%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.00 (n/a)172.00 (n/a)150.40 (n/a)143.00 (n/a)136.00 (n/a)13.22 (n/a)

silu_8_cols_1_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.05 (+6.90%)0.05 (+15.01%)0.05 (+31.79%)0.03 (-7.73%)0.01 (+56.42%)233.00 (+8.37%)172.80 (-11.29%)151.00 (-24.12%)145.00 (-6.45%)33.97 (+58.07%)
0999a9a — 2025-11-13 19:24:080.05 (-7.10%)0.04 (-9.14%)0.04 (-15.58%)0.04 (+26.51%)0.00 (-40.67%)215.00 (-20.96%)194.80 (+6.33%)199.00 (+18.45%)155.00 (+7.64%)21.49 (-52.75%)
fbf1591 — 2025-11-13 03:39:220.05 (n/a)0.04 (n/a)0.05 (n/a)0.03 (n/a)0.01 (n/a)272.00 (n/a)183.20 (n/a)168.00 (n/a)144.00 (n/a)45.48 (n/a)

softmax_1_cols_2_channels_4096_tile_2048

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.18 (+19.20%)0.16 (+29.13%)0.17 (+37.78%)0.14 (+23.46%)0.01 (+5.96%)162.00 (-19.00%)140.60 (-22.75%)135.00 (-27.42%)125.00 (-16.11%)13.25 (-26.73%)
0999a9a — 2025-11-13 19:24:080.15 (-20.13%)0.13 (-18.52%)0.12 (-20.43%)0.11 (-15.50%)0.01 (-28.63%)200.00 (+18.34%)182.00 (+22.31%)186.00 (+25.68%)149.00 (+25.21%)18.08 (+6.20%)
fbf1591 — 2025-11-13 03:39:220.19 (n/a)0.16 (n/a)0.15 (n/a)0.14 (n/a)0.02 (n/a)169.00 (n/a)148.80 (n/a)148.00 (n/a)119.00 (n/a)17.02 (n/a)

softmax_2_cols_2_channels_4096_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.16 (-12.95%)0.15 (+4.26%)0.16 (+17.48%)0.11 (+9.36%)0.02 (-26.85%)203.00 (-8.56%)159.20 (-5.58%)143.00 (-14.88%)139.00 (+14.88%)24.64 (-23.31%)
0999a9a — 2025-11-13 19:24:080.19 (+2.48%)0.14 (-11.63%)0.14 (-19.05%)0.10 (-23.42%)0.03 (+48.41%)222.00 (+30.59%)168.60 (+15.80%)168.00 (+23.53%)121.00 (-2.42%)32.13 (+84.60%)
fbf1591 — 2025-11-13 03:39:220.18 (n/a)0.16 (n/a)0.17 (n/a)0.13 (n/a)0.02 (n/a)170.00 (n/a)145.60 (n/a)136.00 (n/a)124.00 (n/a)17.41 (n/a)

softmax_2_cols_2_channels_4096_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.20 (+36.61%)0.16 (+22.76%)0.15 (+18.42%)0.09 (-2.36%)0.04 (+80.28%)254.00 (+2.42%)159.40 (-14.58%)152.00 (-15.56%)112.00 (-26.80%)50.17 (+42.20%)
0999a9a — 2025-11-13 19:24:080.15 (-1.96%)0.13 (-2.94%)0.13 (-5.55%)0.09 (+10.89%)0.02 (-13.74%)248.00 (-9.82%)186.60 (+1.30%)180.00 (+5.88%)153.00 (+2.00%)35.29 (-24.09%)
fbf1591 — 2025-11-13 03:39:220.15 (n/a)0.13 (n/a)0.13 (n/a)0.08 (n/a)0.03 (n/a)275.00 (n/a)184.20 (n/a)170.00 (n/a)150.00 (n/a)46.48 (n/a)

swiglu

No metrics available.

transpose_2048_M_64_N_1_cols_1_channels_64_m_64_n_8_s

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:423.59 (+0.00%)2.86 (-2.35%)2.65 (-5.98%)2.13 (-20.52%)0.56 (+67.07%)229.00 (+25.82%)177.20 (+5.23%)184.00 (+6.36%)136.00 (+0.00%)34.59 (+106.93%)
0999a9a — 2025-11-13 19:24:083.59 (+20.59%)2.93 (+26.88%)2.82 (+27.17%)2.68 (+52.75%)0.34 (-23.17%)182.00 (-34.53%)168.40 (-23.03%)173.00 (-21.36%)136.00 (-17.07%)16.72 (-58.74%)
fbf1591 — 2025-11-13 03:39:222.98 (n/a)2.31 (n/a)2.22 (n/a)1.76 (n/a)0.44 (n/a)278.00 (n/a)218.80 (n/a)220.00 (n/a)164.00 (n/a)40.51 (n/a)

transpose_2048_M_64_N_1_cols_2_channels_64_m_64_n_8_s

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:423.94 (+2.42%)3.10 (+24.54%)3.21 (+55.26%)1.48 (-3.32%)0.86 (-9.03%)331.00 (+3.44%)178.40 (-20.85%)152.00 (-35.59%)124.00 (-2.36%)77.12 (-1.95%)
0999a9a — 2025-11-13 19:24:083.84 (-6.30%)2.49 (-22.39%)2.07 (-30.93%)1.53 (-38.75%)0.95 (+70.33%)320.00 (+63.27%)225.40 (+43.93%)236.00 (+44.79%)127.00 (+6.72%)78.66 (+198.96%)
fbf1591 — 2025-11-13 03:39:224.10 (n/a)3.21 (n/a)3.00 (n/a)2.49 (n/a)0.56 (n/a)196.00 (n/a)156.60 (n/a)163.00 (n/a)119.00 (n/a)26.31 (n/a)

weighted_rms_norm_1_cols_2_channels_2048_weights_2048

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.07 (-24.03%)0.06 (-21.78%)0.07 (-10.29%)0.04 (-39.51%)0.01 (+6.05%)286.00 (+65.32%)197.40 (+31.42%)175.00 (+11.46%)154.00 (+31.62%)46.44 (+146.63%)
0999a9a — 2025-11-13 19:24:080.10 (-3.42%)0.08 (-10.60%)0.07 (-12.10%)0.07 (-13.87%)0.01 (+25.36%)173.00 (+16.11%)150.20 (+12.76%)157.00 (+13.77%)117.00 (+3.54%)18.83 (+48.05%)
fbf1591 — 2025-11-13 03:39:220.10 (n/a)0.09 (n/a)0.08 (n/a)0.08 (n/a)0.01 (n/a)149.00 (n/a)133.20 (n/a)138.00 (n/a)113.00 (n/a)12.72 (n/a)

weighted_rms_norm_2_cols_2_channels_2048_weights_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.09 (-17.42%)0.07 (-5.87%)0.07 (-7.45%)0.07 (+50.00%)0.01 (-67.20%)170.00 (-33.33%)154.00 (-2.28%)161.00 (+8.05%)132.00 (+21.10%)13.93 (-73.52%)
0999a9a — 2025-11-13 19:24:080.10 (+12.84%)0.08 (-5.42%)0.08 (-10.07%)0.04 (-41.96%)0.02 (+265.18%)255.00 (+72.30%)157.60 (+15.37%)149.00 (+11.19%)109.00 (-11.38%)52.60 (+447.71%)
fbf1591 — 2025-11-13 03:39:220.09 (n/a)0.08 (n/a)0.09 (n/a)0.08 (n/a)0.01 (n/a)148.00 (n/a)136.60 (n/a)134.00 (n/a)123.00 (n/a)9.60 (n/a)

weighted_rms_norm_4_cols_2_channels_2048_weights_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.08 (+14.60%)0.07 (+11.19%)0.06 (+10.67%)0.06 (+15.76%)0.01 (+12.12%)184.00 (-13.62%)170.20 (-10.14%)178.00 (-9.64%)137.00 (-12.74%)17.15 (-16.61%)
0999a9a — 2025-11-13 19:24:080.07 (-22.29%)0.06 (-29.29%)0.06 (-30.96%)0.05 (-34.27%)0.01 (+58.69%)213.00 (+52.14%)189.40 (+42.84%)197.00 (+44.85%)157.00 (+28.69%)20.57 (+210.64%)
fbf1591 — 2025-11-13 03:39:220.09 (n/a)0.09 (n/a)0.08 (n/a)0.08 (n/a)0.00 (n/a)140.00 (n/a)132.60 (n/a)136.00 (n/a)122.00 (n/a)6.62 (n/a)

weighted_rms_norm_8_cols_2_channels_2048_weights_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
3e0443f — 2025-11-13 21:21:420.08 (-7.86%)0.07 (-2.82%)0.07 (-4.27%)0.06 (+7.87%)0.01 (-33.84%)178.00 (-7.29%)162.60 (+1.88%)164.00 (+4.46%)140.00 (+8.53%)14.36 (-32.77%)
0999a9a — 2025-11-13 19:24:080.09 (+24.03%)0.07 (+20.14%)0.07 (+19.75%)0.06 (+28.65%)0.01 (+16.90%)192.00 (-22.27%)159.60 (-17.05%)157.00 (-16.49%)129.00 (-19.38%)21.36 (-28.23%)
fbf1591 — 2025-11-13 03:39:220.07 (n/a)0.06 (n/a)0.06 (n/a)0.05 (n/a)0.01 (n/a)247.00 (n/a)192.40 (n/a)188.00 (n/a)160.00 (n/a)29.76 (n/a)

@pvasireddy-amd
Copy link
Collaborator

pvasireddy-amd commented Nov 14, 2025

Just a suggestion: examples folder could be primitives or algorithms to avoid naming overlap with operators.

andrej and others added 5 commits November 14, 2025 13:29
Co-authored-by: pvasireddy-amd <pvasired@amd.com>
Co-authored-by: Victor Jung <33875047+Victor-Jung@users.noreply.github.com>
Co-authored-by: cubansil <CurtJohn.Bansil@amd.com>
Co-authored-by: André Rösti <an.roesti@gmail.com>
@andrej andrej force-pushed the bring-in-the-llama branch from 9da3796 to 28058d1 Compare November 14, 2025 20:46
@github-actions
Copy link

📊 Test Results for Test Example Applications

bef0e25 (2025_11_14_21_10_56)

IRONCLAD

Tested on 2025_11_14_21_10_56 at commit bef0e25.

Test Checks TTFT (mean)TPS (mean)
llama_3.2_1b ✅ 5/5 24.09 0.82
📈 Trends (vs main branch) for Test Example Applications

bef0e25 (2025_11_14_21_10_56)

IRONCLAD Trends

llama_3.2_1b

Commit/Date Num Tokens (max)Num Tokens (mean)Num Tokens (median)Num Tokens (min)Num Tokens (stddev)TPS (max)TPS (mean)TPS (median)TPS (min)TPS (stddev)TTFT (max)TTFT (mean)TTFT (median)TTFT (min)TTFT (stddev)Total (max)Total (mean)Total (median)Total (min)Total (stddev)
bef0e25 — 2025-11-14 21:02:1940.00 (n/a)40.00 (n/a)40.00 (n/a)40.00 (n/a)0.00 (n/a)0.83 (n/a)0.82 (n/a)0.83 (n/a)0.80 (n/a)0.01 (n/a)24.91 (n/a)24.09 (n/a)24.43 (n/a)23.14 (n/a)0.66 (n/a)73.07 (n/a)71.65 (n/a)71.86 (n/a)69.86 (n/a)1.24 (n/a)

@github-actions
Copy link

📊 Test Results for Test Example Applications

8c53610 (2025_11_14_22_53_33)

IRONCLAD

Tested on 2025_11_14_22_53_33 at commit 8c53610.

Test Checks TTFT (mean)TPS (mean)
llama_3.2_1b ✅ 5/5 24.06 0.83
📈 Trends (vs main branch) for Test Example Applications

8c53610 (2025_11_14_22_53_33)

IRONCLAD Trends

llama_3.2_1b

Commit/Date Num Tokens (max)Num Tokens (mean)Num Tokens (median)Num Tokens (min)Num Tokens (stddev)TPS (max)TPS (mean)TPS (median)TPS (min)TPS (stddev)TTFT (max)TTFT (mean)TTFT (median)TTFT (min)TTFT (stddev)Total (max)Total (mean)Total (median)Total (min)Total (stddev)
8c53610 — 2025-11-14 22:45:0040.00 (n/a)40.00 (n/a)40.00 (n/a)40.00 (n/a)0.00 (n/a)0.84 (n/a)0.83 (n/a)0.83 (n/a)0.82 (n/a)0.01 (n/a)24.71 (n/a)24.06 (n/a)24.09 (n/a)23.30 (n/a)0.50 (n/a)71.88 (n/a)71.02 (n/a)70.72 (n/a)70.68 (n/a)0.45 (n/a)

@andrej andrej force-pushed the bring-in-the-llama branch from 786a684 to d442395 Compare November 14, 2025 22:55
@github-actions
Copy link

📊 Test Results for Test Example Applications

82970e5 (2025_11_14_23_20_05)

IRONCLAD

Tested on 2025_11_14_23_20_05 at commit 82970e5.

Test Checks TTFT (mean)TPS (mean)
llama_3.2_1b ✅ 5/5 23.64 0.83
📈 Trends (vs main branch) for Test Example Applications

82970e5 (2025_11_14_23_20_05)

IRONCLAD Trends

llama_3.2_1b

Commit/Date Num Tokens (max)Num Tokens (mean)Num Tokens (median)Num Tokens (min)Num Tokens (stddev)TPS (max)TPS (mean)TPS (median)TPS (min)TPS (stddev)TTFT (max)TTFT (mean)TTFT (median)TTFT (min)TTFT (stddev)Total (max)Total (mean)Total (median)Total (min)Total (stddev)
82970e5 — 2025-11-14 23:11:3340.00 (n/a)40.00 (n/a)40.00 (n/a)40.00 (n/a)0.00 (n/a)0.85 (n/a)0.83 (n/a)0.83 (n/a)0.81 (n/a)0.01 (n/a)25.02 (n/a)23.64 (n/a)23.65 (n/a)22.81 (n/a)0.81 (n/a)71.99 (n/a)70.73 (n/a)70.85 (n/a)68.87 (n/a)1.03 (n/a)

@github-actions
Copy link

📊 Test Results for Test Example Applications

8404bd4 (2025_11_14_23_48_04)

IRONCLAD

Tested on 2025_11_14_23_48_04 at commit 8404bd4.

Test Checks TTFT (mean)TPS (mean)
llama_3.2_1b ✅ 5/5 23.63 0.84
📈 Trends (vs main branch) for Test Example Applications

8404bd4 (2025_11_14_23_48_04)

IRONCLAD Trends

llama_3.2_1b

Commit/Date Num Tokens (max)Num Tokens (mean)Num Tokens (median)Num Tokens (min)Num Tokens (stddev)TPS (max)TPS (mean)TPS (median)TPS (min)TPS (stddev)TTFT (max)TTFT (mean)TTFT (median)TTFT (min)TTFT (stddev)Total (max)Total (mean)Total (median)Total (min)Total (stddev)
8404bd4 — 2025-11-14 23:39:3540.00 (n/a)40.00 (n/a)40.00 (n/a)40.00 (n/a)0.00 (n/a)0.87 (n/a)0.84 (n/a)0.82 (n/a)0.82 (n/a)0.02 (n/a)24.59 (n/a)23.63 (n/a)23.71 (n/a)21.84 (n/a)1.01 (n/a)72.11 (n/a)70.24 (n/a)71.44 (n/a)66.72 (n/a)2.07 (n/a)

@github-actions
Copy link

📊 Test Results for Small Benchmark/Test Suite

8404bd4 (2025_11_15_00_04_44)

IRONCLAD

Tested on 2025_11_15_00_04_44 at commit 8404bd4.

Test Checks Latency (mean)Bandwidth (mean)GFLOP/s (mean)
axpy_1_cols_2_channels_2048_tile_2048_3.0 ✅ 5/5 167.40 0.07 n/a
axpy_2_cols_2_channels_2048_tile_1024_3.0 ✅ 5/5 158.00 0.08 n/a
axpy_4_cols_2_channels_2048_tile_512_3.0 ✅ 5/5 153.20 0.08 n/a
axpy_8_cols_2_channels_2048_tile_256_3.0 ✅ 5/5 171.20 0.07 n/a
dequant_1_cols_1_channels_2048_tile_2048 ✅ 5/5 148.40 0.03 n/a
dequant_1_cols_2_channels_2048_tile_1024 ✅ 5/5 170.40 0.03 n/a
dequant_2_cols_1_channels_2048_tile_1024 ✅ 5/5 217.40 0.02 n/a
dequant_2_cols_2_channels_2048_tile_512 ✅ 5/5 179.60 0.03 n/a
dequant_4_cols_1_channels_2048_tile_512 ✅ 5/5 139.00 0.04 n/a
dequant_4_cols_2_channels_2048_tile_256 ✅ 5/5 168.40 0.03 n/a
dequant_8_cols_1_channels_2048_tile_256 ✅ 5/5 164.80 0.03 n/a
dequant_8_cols_2_channels_2048_tile_128 ✅ 5/5 224.00 0.02 n/a
eltwise_add_1_cols_2_channels_2048_tile_2048 ✅ 5/5 196.60 0.06 n/a
eltwise_add_2_cols_2_channels_2048_tile_1024 ✅ 5/5 181.80 0.07 n/a
eltwise_add_4_cols_2_channels_2048_tile_512 ✅ 5/5 156.20 0.07 n/a
eltwise_add_8_cols_2_channels_2048_tile_256 ✅ 5/5 200.20 0.06 n/a
eltwise_mul_1_cols_2_channels_2048_tile_2048 ✅ 5/5 155.00 0.07 n/a
eltwise_mul_2_cols_2_channels_2048_tile_1024 ✅ 5/5 159.80 0.07 n/a
eltwise_mul_4_cols_2_channels_2048_tile_512 ✅ 5/5 156.00 0.07 n/a
eltwise_mul_8_cols_2_channels_2048_tile_256 ✅ 5/5 180.00 0.06 n/a
gelu_1_cols_1_channels_2048_tile_2048 ✅ 5/5 154.40 0.05 n/a
gelu_1_cols_2_channels_2048_tile_1024 ✅ 5/5 154.40 0.05 n/a
gelu_2_cols_1_channels_2048_tile_1024 ✅ 5/5 178.20 0.05 n/a
gelu_2_cols_2_channels_2048_tile_512 ✅ 5/5 170.20 0.05 n/a
gelu_4_cols_1_channels_2048_tile_512 ✅ 5/5 168.00 0.05 n/a
gelu_4_cols_2_channels_2048_tile_256 ✅ 5/5 175.20 0.05 n/a
gelu_8_cols_1_channels_2048_tile_256 ✅ 5/5 200.80 0.04 n/a
gelu_8_cols_2_channels_2048_tile_128 ✅ 5/5 200.20 0.04 n/a
gemm_2048x2048x2048_64x64x64_8_cols_0_bcolmaj_0_ccolmaj_0 ✅ 5/5 6449.20 3.64 n/a
layer_norm_1_cols_1_channels_2048_tile_2048 ✅ 5/5 185.20 0.07 n/a
layer_norm_1_cols_2_channels_2048_tile_1024 ✅ 5/5 159.80 0.08 n/a
layer_norm_2_cols_1_channels_2048_tile_1024 ✅ 5/5 149.40 0.08 n/a
layer_norm_2_cols_2_channels_2048_tile_512 ✅ 5/5 136.60 0.09 n/a
layer_norm_4_cols_1_channels_2048_tile_512 ✅ 5/5 151.80 0.08 n/a
layer_norm_4_cols_2_channels_2048_tile_256 ✅ 5/5 208.60 0.06 n/a
layer_norm_8_cols_1_channels_2048_tile_256 ✅ 5/5 139.20 0.09 n/a
layer_norm_8_cols_2_channels_2048_tile_128 ✅ 5/5 166.20 0.07 n/a
matrix_vector_mul_128x128_32_1col ✅ 5/5 756.20 0.04 0.04
matrix_vector_mul_2048x8192_1_1col ✅ 5/5 3228.60 10.47 10.46
matrix_vector_mul_2048x8192_1_2col ✅ 5/5 1985.80 16.92 16.91
matrix_vector_mul_2048x8192_1_4col ✅ 5/5 1625.60 20.76 20.75
matrix_vector_mul_2048x8192_1_8col ✅ 5/5 1927.40 17.70 17.69
matrix_vector_mul_8192x2048_4_1col ✅ 5/5 3061.20 10.97 10.96
matrix_vector_mul_8192x2048_4_2col ✅ 5/5 1919.80 17.49 17.48
matrix_vector_mul_8192x2048_4_4col ✅ 5/5 1664.40 20.31 20.29
matrix_vector_mul_8192x2048_4_8col ✅ 5/5 1946.20 17.41 17.40
mem_copy_16_cores_2_chans_2048_tile_128_False ✅ 5/5 157.60 0.05 n/a
mem_copy_1_cores_1_chans_2048_tile_2048_False ✅ 5/5 141.20 0.05 n/a
mem_copy_2_cores_1_chans_2048_tile_1024_False ✅ 5/5 175.00 0.05 n/a
mem_copy_2_cores_2_chans_2048_tile_1024_False ✅ 5/5 141.80 0.05 n/a
mem_copy_4_cores_1_chans_2048_tile_512_False ✅ 5/5 145.20 0.05 n/a
mem_copy_4_cores_2_chans_2048_tile_512_False ✅ 5/5 154.20 0.05 n/a
mem_copy_8_cores_1_chans_2048_tile_256_False ✅ 5/5 155.60 0.05 n/a
mem_copy_8_cores_2_chans_2048_tile_256_False ✅ 5/5 170.40 0.05 n/a
mha ✅ 5/5 n/a 0.09 65.80
relu_1_cols_1_channels_2048_tile_2048 ✅ 5/5 176.60 0.04 n/a
relu_2_cols_1_channels_2048_tile_1024 ✅ 5/5 168.80 0.05 n/a
relu_4_cols_1_channels_2048_tile_512 ✅ 5/5 168.60 0.05 n/a
relu_8_cols_1_channels_2048_tile_256 ✅ 5/5 178.00 0.04 n/a
rms_norm_1_cols_1_channels_2048_tile_2048 ✅ 5/5 148.00 0.08 n/a
rms_norm_1_cols_2_channels_2048_tile_1024 ✅ 5/5 149.20 0.08 n/a
rms_norm_2_cols_1_channels_2048_tile_1024 ✅ 5/5 142.40 0.08 n/a
rms_norm_2_cols_2_channels_2048_tile_512 ✅ 5/5 154.60 0.08 n/a
rms_norm_4_cols_1_channels_2048_tile_512 ✅ 5/5 145.40 0.08 n/a
rms_norm_4_cols_2_channels_2048_tile_256 ✅ 5/5 165.20 0.07 n/a
rms_norm_8_cols_1_channels_2048_tile_256 ✅ 5/5 145.60 0.08 n/a
rms_norm_8_cols_2_channels_2048_tile_128 ✅ 5/5 174.20 0.07 n/a
rope_1_cols_2_channels_4096_tile_4096_0 ✅ 5/5 151.40 0.16 n/a
rope_2_cols_2_channels_4096_tile_2048_0 ✅ 5/5 141.80 0.16 n/a
rope_4_cols_2_channels_4096_tile_1024_0 ✅ 5/5 144.00 0.16 n/a
rope_8_cols_2_channels_4096_tile_512_0 ✅ 5/5 177.00 0.14 n/a
silu_1_cols_1_channels_2048_tile_2048 ✅ 5/5 149.60 0.05 n/a
silu_2_cols_1_channels_2048_tile_1024 ✅ 5/5 147.60 0.06 n/a
silu_4_cols_1_channels_2048_tile_512 ✅ 5/5 169.60 0.05 n/a
silu_8_cols_1_channels_2048_tile_256 ✅ 5/5 202.20 0.04 n/a
softmax_1_cols_2_channels_4096_tile_2048 ✅ 5/5 137.80 0.17 n/a
softmax_2_cols_2_channels_4096_tile_1024 ✅ 5/5 142.40 0.16 n/a
softmax_2_cols_2_channels_4096_tile_512 ✅ 5/5 131.40 0.18 n/a
swiglu ✅ 5/5 n/a n/a n/a
transpose_2048_M_64_N_1_cols_1_channels_64_m_64_n_8_s ✅ 5/5 177.80 2.83 n/a
transpose_2048_M_64_N_1_cols_2_channels_64_m_64_n_8_s ✅ 5/5 168.20 2.94 n/a
weighted_rms_norm_1_cols_2_channels_2048_weights_2048 ✅ 5/5 167.60 0.07 n/a
weighted_rms_norm_2_cols_2_channels_2048_weights_1024 ✅ 5/5 128.60 0.09 n/a
weighted_rms_norm_4_cols_2_channels_2048_weights_512 ✅ 5/5 151.00 0.08 n/a
weighted_rms_norm_8_cols_2_channels_2048_weights_256 ✅ 5/5 179.00 0.07 n/a
📈 Trends (vs main branch) for Small Benchmark/Test Suite

8404bd4 (2025_11_15_00_04_44)

IRONCLAD Trends

axpy_1_cols_2_channels_2048_tile_2048_3.0

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.08 (-17.04%)0.07 (-10.21%)0.07 (-13.92%)0.05 (-5.56%)0.01 (-32.16%)234.00 (+5.88%)167.40 (+8.84%)158.00 (+16.18%)135.00 (+20.54%)34.87 (-11.48%)
003b49d — 2025-11-14 19:30:540.10 (+3.57%)0.08 (-6.28%)0.08 (+2.21%)0.05 (-26.70%)0.02 (+56.99%)221.00 (+36.42%)153.80 (+11.13%)136.00 (-2.16%)112.00 (-3.45%)39.40 (+110.88%)
0999a9a — 2025-11-13 19:24:080.10 (+16.38%)0.08 (+3.69%)0.08 (+1.44%)0.07 (-8.64%)0.01 (+372.63%)162.00 (+9.46%)138.40 (-1.84%)139.00 (-1.42%)116.00 (-14.07%)18.68 (+340.35%)
fbf1591 — 2025-11-13 03:39:220.08 (n/a)0.08 (n/a)0.08 (n/a)0.08 (n/a)0.00 (n/a)148.00 (n/a)141.00 (n/a)141.00 (n/a)135.00 (n/a)4.24 (n/a)

axpy_2_cols_2_channels_2048_tile_1024_3.0

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.10 (+30.00%)0.08 (+8.79%)0.07 (-6.98%)0.06 (+4.26%)0.02 (+89.36%)188.00 (-4.08%)158.00 (-6.06%)172.00 (+7.50%)110.00 (-23.08%)27.37 (+34.56%)
003b49d — 2025-11-14 19:30:540.08 (-20.98%)0.07 (-4.64%)0.07 (-4.37%)0.06 (+20.92%)0.01 (-60.42%)196.00 (-17.30%)168.20 (-2.44%)160.00 (+4.58%)143.00 (+26.55%)20.34 (-59.93%)
0999a9a — 2025-11-13 19:24:080.10 (-2.65%)0.07 (+4.47%)0.07 (+7.19%)0.05 (+18.14%)0.02 (+1.47%)237.00 (-15.36%)172.40 (-4.54%)153.00 (-6.71%)113.00 (+2.73%)50.76 (-8.77%)
fbf1591 — 2025-11-13 03:39:220.10 (n/a)0.07 (n/a)0.07 (n/a)0.04 (n/a)0.02 (n/a)280.00 (n/a)180.60 (n/a)164.00 (n/a)110.00 (n/a)55.64 (n/a)

axpy_4_cols_2_channels_2048_tile_512_3.0

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.08 (+9.35%)0.08 (+15.31%)0.08 (+15.49%)0.07 (+54.29%)0.01 (-38.21%)175.00 (-35.19%)153.20 (-16.28%)142.00 (-13.41%)139.00 (-8.55%)15.87 (-64.23%)
003b49d — 2025-11-14 19:30:540.08 (-36.84%)0.07 (-24.51%)0.07 (-15.24%)0.04 (-42.22%)0.01 (-27.05%)270.00 (+73.08%)183.00 (+34.56%)164.00 (+17.99%)152.00 (+58.33%)44.36 (+110.82%)
0999a9a — 2025-11-13 19:24:080.12 (+30.21%)0.09 (+13.66%)0.08 (+4.32%)0.07 (+35.90%)0.02 (+34.67%)156.00 (-26.42%)136.00 (-12.14%)139.00 (-4.14%)96.00 (-23.20%)21.04 (-29.55%)
fbf1591 — 2025-11-13 03:39:220.09 (n/a)0.08 (n/a)0.08 (n/a)0.05 (n/a)0.01 (n/a)212.00 (n/a)154.80 (n/a)145.00 (n/a)125.00 (n/a)29.87 (n/a)

axpy_8_cols_2_channels_2048_tile_256_3.0

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.07 (-9.55%)0.07 (+5.24%)0.07 (+13.95%)0.06 (+9.52%)0.00 (-58.98%)189.00 (-8.70%)171.20 (-6.65%)172.00 (-12.24%)157.00 (+10.56%)10.46 (-58.94%)
003b49d — 2025-11-14 19:30:540.08 (-19.72%)0.06 (-25.75%)0.06 (-34.69%)0.06 (-21.74%)0.01 (-7.63%)207.00 (+27.78%)183.40 (+35.45%)196.00 (+53.12%)142.00 (+24.56%)25.47 (+47.67%)
0999a9a — 2025-11-13 19:24:080.10 (+21.05%)0.09 (+32.73%)0.09 (+46.09%)0.07 (+33.95%)0.01 (-2.31%)162.00 (-25.35%)135.40 (-25.44%)128.00 (-31.55%)114.00 (-17.39%)17.25 (-38.93%)
fbf1591 — 2025-11-13 03:39:220.08 (n/a)0.06 (n/a)0.06 (n/a)0.05 (n/a)0.01 (n/a)217.00 (n/a)181.60 (n/a)187.00 (n/a)138.00 (n/a)28.24 (n/a)

dequant_1_cols_1_channels_2048_tile_2048

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.04 (+20.00%)0.03 (+19.04%)0.03 (+21.13%)0.03 (+31.43%)0.00 (+1.02%)175.00 (-23.91%)148.40 (-16.63%)142.00 (-17.44%)125.00 (-16.67%)17.11 (-37.60%)
003b49d — 2025-11-14 19:30:540.03 (-28.00%)0.03 (-23.83%)0.03 (-14.53%)0.02 (-30.43%)0.00 (-40.53%)230.00 (+43.75%)178.00 (+30.31%)172.00 (+17.01%)150.00 (+38.89%)27.42 (+23.72%)
0999a9a — 2025-11-13 19:24:080.05 (+58.33%)0.04 (+50.10%)0.03 (+30.61%)0.03 (+85.00%)0.01 (+48.14%)160.00 (-45.95%)136.60 (-34.07%)147.00 (-23.44%)108.00 (-36.84%)22.16 (-51.39%)
fbf1591 — 2025-11-13 03:39:220.03 (n/a)0.02 (n/a)0.03 (n/a)0.02 (n/a)0.00 (n/a)296.00 (n/a)207.20 (n/a)192.00 (n/a)171.00 (n/a)45.59 (n/a)

dequant_1_cols_2_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.04 (+8.40%)0.03 (+7.66%)0.03 (+2.53%)0.02 (+2.06%)0.01 (+8.77%)243.00 (-2.02%)170.40 (-6.78%)158.00 (-2.47%)131.00 (-7.75%)38.45 (+1.07%)
003b49d — 2025-11-14 19:30:540.03 (-17.61%)0.03 (-15.76%)0.03 (+0.00%)0.02 (-30.24%)0.01 (+3.67%)248.00 (+43.35%)182.80 (+20.90%)162.00 (+0.00%)142.00 (+21.37%)38.04 (+83.26%)
0999a9a — 2025-11-13 19:24:080.04 (-16.24%)0.03 (-9.53%)0.03 (-9.88%)0.03 (-2.31%)0.01 (-30.64%)173.00 (+2.37%)151.20 (+9.09%)162.00 (+10.96%)117.00 (+19.39%)20.76 (-12.76%)
fbf1591 — 2025-11-13 03:39:220.05 (n/a)0.04 (n/a)0.03 (n/a)0.03 (n/a)0.01 (n/a)169.00 (n/a)138.60 (n/a)146.00 (n/a)98.00 (n/a)23.80 (n/a)

dequant_2_cols_1_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.04 (+6.62%)0.02 (-25.20%)0.02 (-30.26%)0.02 (-35.20%)0.01 (+330.03%)250.00 (+54.32%)217.40 (+40.44%)228.00 (+43.40%)136.00 (-6.21%)41.74 (+503.98%)
003b49d — 2025-11-14 19:30:540.03 (-22.07%)0.03 (-20.91%)0.03 (-22.64%)0.03 (-14.81%)0.00 (-48.28%)162.00 (+17.39%)154.80 (+26.06%)159.00 (+29.27%)145.00 (+28.32%)6.91 (-22.03%)
0999a9a — 2025-11-13 19:24:080.04 (+10.62%)0.04 (+18.53%)0.04 (+20.33%)0.04 (+14.49%)0.00 (-3.03%)138.00 (-12.66%)122.80 (-15.78%)123.00 (-16.89%)113.00 (-9.60%)8.86 (-22.67%)
fbf1591 — 2025-11-13 03:39:220.04 (n/a)0.03 (n/a)0.03 (n/a)0.03 (n/a)0.00 (n/a)158.00 (n/a)145.80 (n/a)148.00 (n/a)125.00 (n/a)11.46 (n/a)

dequant_2_cols_2_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.04 (+14.63%)0.03 (+7.11%)0.03 (+19.15%)0.02 (-13.83%)0.01 (+83.68%)253.00 (+16.06%)179.60 (-0.22%)141.00 (-16.07%)123.00 (-12.77%)57.02 (+88.05%)
003b49d — 2025-11-14 19:30:540.03 (-9.93%)0.03 (-15.11%)0.03 (-11.31%)0.02 (-11.93%)0.00 (+8.00%)218.00 (+13.54%)180.00 (+18.89%)168.00 (+12.75%)141.00 (+11.02%)30.32 (+37.58%)
0999a9a — 2025-11-13 19:24:080.04 (+6.30%)0.03 (+17.59%)0.03 (+23.49%)0.03 (+8.85%)0.00 (-8.04%)192.00 (-8.13%)151.40 (-15.51%)149.00 (-19.02%)127.00 (-5.93%)22.04 (-19.09%)
fbf1591 — 2025-11-13 03:39:220.04 (n/a)0.03 (n/a)0.03 (n/a)0.02 (n/a)0.00 (n/a)209.00 (n/a)179.20 (n/a)184.00 (n/a)135.00 (n/a)27.24 (n/a)

dequant_4_cols_1_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.04 (+4.72%)0.04 (+20.34%)0.04 (+17.65%)0.03 (+60.00%)0.00 (-64.48%)155.00 (-37.50%)139.00 (-21.11%)136.00 (-15.00%)127.00 (-4.51%)9.49 (-78.28%)
003b49d — 2025-11-14 19:30:540.04 (-1.50%)0.03 (-12.66%)0.03 (-9.38%)0.02 (-32.66%)0.01 (+145.42%)248.00 (+48.50%)176.20 (+20.36%)160.00 (+10.34%)133.00 (+1.53%)43.68 (+261.18%)
0999a9a — 2025-11-13 19:24:080.04 (-16.79%)0.03 (-3.88%)0.03 (-8.97%)0.03 (+29.94%)0.00 (-64.56%)167.00 (-23.04%)146.40 (-0.95%)145.00 (+9.85%)131.00 (+20.18%)12.09 (-67.98%)
fbf1591 — 2025-11-13 03:39:220.04 (n/a)0.03 (n/a)0.04 (n/a)0.02 (n/a)0.01 (n/a)217.00 (n/a)147.80 (n/a)132.00 (n/a)109.00 (n/a)37.76 (n/a)

dequant_4_cols_2_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.04 (-6.98%)0.03 (+13.24%)0.03 (+18.71%)0.02 (+28.18%)0.01 (-22.06%)220.00 (-21.99%)168.40 (-15.72%)155.00 (-15.76%)129.00 (+7.50%)38.17 (-36.69%)
003b49d — 2025-11-14 19:30:540.04 (+23.33%)0.03 (-7.77%)0.03 (-4.89%)0.02 (-31.21%)0.01 (+188.72%)282.00 (+45.36%)199.80 (+18.22%)184.00 (+5.14%)120.00 (-18.92%)60.28 (+255.21%)
0999a9a — 2025-11-13 19:24:080.03 (-19.59%)0.03 (-7.11%)0.03 (-5.71%)0.03 (+0.00%)0.00 (-49.35%)194.00 (+0.00%)169.00 (+5.36%)175.00 (+6.06%)148.00 (+24.37%)16.97 (-38.04%)
fbf1591 — 2025-11-13 03:39:220.04 (n/a)0.03 (n/a)0.03 (n/a)0.03 (n/a)0.01 (n/a)194.00 (n/a)160.40 (n/a)165.00 (n/a)119.00 (n/a)27.39 (n/a)

dequant_8_cols_1_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.03 (-5.48%)0.03 (+2.20%)0.03 (+14.20%)0.02 (+0.00%)0.00 (-29.23%)205.00 (+0.00%)164.80 (-3.29%)162.00 (-12.43%)146.00 (+5.80%)21.27 (-21.36%)
003b49d — 2025-11-14 19:30:540.04 (-14.49%)0.03 (-12.43%)0.03 (-22.70%)0.02 (-1.46%)0.00 (-22.97%)205.00 (+1.49%)170.40 (+12.85%)185.00 (+29.37%)138.00 (+16.95%)27.05 (-11.63%)
0999a9a — 2025-11-13 19:24:080.04 (+24.58%)0.03 (+11.01%)0.03 (+12.59%)0.02 (-7.92%)0.01 (+137.56%)202.00 (+8.60%)151.00 (-7.13%)143.00 (-11.18%)118.00 (-19.73%)30.61 (+108.57%)
fbf1591 — 2025-11-13 03:39:220.03 (n/a)0.03 (n/a)0.03 (n/a)0.03 (n/a)0.00 (n/a)186.00 (n/a)162.60 (n/a)161.00 (n/a)147.00 (n/a)14.68 (n/a)

dequant_8_cols_2_channels_2048_tile_128

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.03 (+4.02%)0.02 (-9.02%)0.02 (-6.98%)0.02 (-19.63%)0.00 (+132.64%)270.00 (+24.42%)224.00 (+12.00%)215.00 (+7.50%)174.00 (-3.87%)32.64 (+179.04%)
003b49d — 2025-11-14 19:30:540.03 (-3.87%)0.02 (+11.01%)0.02 (+1.00%)0.02 (+49.77%)0.00 (-72.41%)217.00 (-33.23%)200.00 (-15.25%)200.00 (-0.99%)181.00 (+4.02%)11.70 (-81.08%)
0999a9a — 2025-11-13 19:24:080.03 (-6.32%)0.02 (+2.86%)0.02 (+8.91%)0.02 (+4.31%)0.01 (+1.62%)325.00 (-4.13%)236.00 (-2.32%)202.00 (-8.18%)174.00 (+6.75%)61.82 (+3.89%)
fbf1591 — 2025-11-13 03:39:220.03 (n/a)0.02 (n/a)0.02 (n/a)0.01 (n/a)0.01 (n/a)339.00 (n/a)241.60 (n/a)220.00 (n/a)163.00 (n/a)59.50 (n/a)

eltwise_add_1_cols_2_channels_2048_tile_2048

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.07 (-16.34%)0.06 (-10.34%)0.07 (-1.16%)0.04 (-14.84%)0.01 (-10.23%)283.00 (+17.43%)196.60 (+12.21%)172.00 (+1.18%)153.00 (+19.53%)46.37 (+26.79%)
003b49d — 2025-11-14 19:30:540.09 (-13.28%)0.07 (-19.00%)0.07 (-17.65%)0.05 (-31.54%)0.01 (+7.26%)241.00 (+46.06%)175.20 (+25.86%)170.00 (+21.43%)128.00 (+15.32%)36.57 (+84.01%)
0999a9a — 2025-11-13 19:24:080.10 (+31.53%)0.08 (+10.02%)0.08 (+7.14%)0.07 (-6.06%)0.01 (+707.91%)165.00 (+6.45%)139.20 (-7.20%)140.00 (-6.67%)111.00 (-23.97%)19.87 (+555.21%)
fbf1591 — 2025-11-13 03:39:220.08 (n/a)0.08 (n/a)0.08 (n/a)0.07 (n/a)0.00 (n/a)155.00 (n/a)150.00 (n/a)150.00 (n/a)146.00 (n/a)3.03 (n/a)

eltwise_add_2_cols_2_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.09 (-8.96%)0.07 (-22.02%)0.08 (-10.67%)0.05 (-42.86%)0.02 (+159.22%)252.00 (+75.00%)181.80 (+36.28%)150.00 (+11.94%)134.00 (+9.84%)48.40 (+402.86%)
003b49d — 2025-11-14 19:30:540.09 (+2.46%)0.09 (+22.92%)0.09 (+17.91%)0.08 (+84.03%)0.01 (-60.21%)144.00 (-45.66%)133.40 (-23.16%)134.00 (-15.19%)122.00 (-2.40%)9.62 (-79.96%)
0999a9a — 2025-11-13 19:24:080.09 (-11.20%)0.07 (-26.87%)0.07 (-29.11%)0.04 (-46.79%)0.02 (+77.18%)265.00 (+87.94%)173.60 (+44.19%)158.00 (+41.07%)125.00 (+12.61%)48.02 (+301.50%)
fbf1591 — 2025-11-13 03:39:220.10 (n/a)0.10 (n/a)0.10 (n/a)0.08 (n/a)0.01 (n/a)141.00 (n/a)120.40 (n/a)112.00 (n/a)111.00 (n/a)11.96 (n/a)

eltwise_add_4_cols_2_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.08 (+0.73%)0.07 (-6.80%)0.07 (-10.00%)0.07 (-12.50%)0.01 (+90.87%)176.00 (+14.29%)156.20 (+7.87%)160.00 (+11.11%)137.00 (-0.72%)13.09 (+116.50%)
003b49d — 2025-11-14 19:30:540.08 (+0.00%)0.08 (+15.03%)0.08 (+18.06%)0.07 (+25.97%)0.00 (-62.10%)154.00 (-20.62%)144.80 (-14.22%)144.00 (-15.29%)138.00 (+0.00%)6.05 (-70.10%)
0999a9a — 2025-11-13 19:24:080.08 (-13.04%)0.07 (-17.24%)0.07 (-20.00%)0.06 (-12.37%)0.01 (-4.71%)194.00 (+14.12%)168.80 (+21.09%)170.00 (+25.00%)138.00 (+15.00%)20.22 (+22.30%)
fbf1591 — 2025-11-13 03:39:220.10 (n/a)0.08 (n/a)0.08 (n/a)0.07 (n/a)0.01 (n/a)170.00 (n/a)139.40 (n/a)136.00 (n/a)120.00 (n/a)16.54 (n/a)

eltwise_add_8_cols_2_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.07 (-12.03%)0.06 (-9.38%)0.07 (-8.09%)0.04 (+17.60%)0.01 (-26.55%)267.00 (-14.97%)200.20 (+5.93%)173.00 (+8.81%)158.00 (+13.67%)42.97 (-32.74%)
003b49d — 2025-11-14 19:30:540.08 (+21.58%)0.07 (+11.50%)0.07 (+26.42%)0.04 (-35.35%)0.02 (+264.32%)314.00 (+54.68%)189.00 (-2.98%)159.00 (-20.90%)139.00 (-17.75%)63.88 (+392.60%)
0999a9a — 2025-11-13 19:24:080.07 (-33.14%)0.06 (-30.15%)0.06 (-34.33%)0.06 (-19.70%)0.00 (-62.51%)203.00 (+24.54%)194.80 (+41.16%)201.00 (+52.27%)169.00 (+49.56%)12.97 (-31.96%)
fbf1591 — 2025-11-13 03:39:220.10 (n/a)0.08 (n/a)0.09 (n/a)0.07 (n/a)0.01 (n/a)163.00 (n/a)138.00 (n/a)132.00 (n/a)113.00 (n/a)19.06 (n/a)

eltwise_mul_1_cols_2_channels_2048_tile_2048

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.09 (+3.73%)0.07 (+2.14%)0.07 (+1.24%)0.07 (+5.36%)0.01 (-12.43%)168.00 (-5.08%)155.00 (-2.39%)161.00 (-1.23%)134.00 (-3.60%)13.02 (-18.88%)
003b49d — 2025-11-14 19:30:540.08 (+16.55%)0.07 (+15.70%)0.07 (+4.29%)0.06 (+48.02%)0.01 (-22.91%)177.00 (-32.44%)158.80 (-15.35%)163.00 (-4.12%)139.00 (-14.20%)16.05 (-57.08%)
0999a9a — 2025-11-13 19:24:080.07 (-23.46%)0.06 (-22.73%)0.07 (-15.88%)0.04 (-39.31%)0.01 (+34.51%)262.00 (+64.78%)187.60 (+32.49%)170.00 (+18.88%)162.00 (+30.65%)37.40 (+200.01%)
fbf1591 — 2025-11-13 03:39:220.09 (n/a)0.08 (n/a)0.08 (n/a)0.07 (n/a)0.01 (n/a)159.00 (n/a)141.60 (n/a)143.00 (n/a)124.00 (n/a)12.47 (n/a)

eltwise_mul_2_cols_2_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.08 (-14.77%)0.07 (-0.73%)0.07 (-14.37%)0.07 (+44.91%)0.00 (-84.10%)167.00 (-30.99%)159.80 (-6.33%)160.00 (+16.79%)149.00 (+17.32%)6.31 (-87.00%)
003b49d — 2025-11-14 19:30:540.09 (+25.20%)0.07 (+11.24%)0.08 (+29.93%)0.05 (-17.77%)0.02 (+273.70%)242.00 (+21.61%)170.60 (-3.72%)137.00 (-23.03%)127.00 (-20.13%)48.49 (+258.50%)
0999a9a — 2025-11-13 19:24:080.07 (-23.27%)0.06 (-20.20%)0.06 (-21.91%)0.06 (-16.58%)0.00 (-39.23%)199.00 (+19.88%)177.20 (+24.79%)178.00 (+28.06%)159.00 (+30.33%)13.53 (-5.57%)
fbf1591 — 2025-11-13 03:39:220.09 (n/a)0.08 (n/a)0.08 (n/a)0.07 (n/a)0.01 (n/a)166.00 (n/a)142.00 (n/a)139.00 (n/a)122.00 (n/a)14.32 (n/a)

eltwise_mul_4_cols_2_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.09 (+0.00%)0.07 (+4.24%)0.08 (+8.05%)0.06 (-7.04%)0.01 (+21.40%)199.00 (+7.57%)156.00 (-3.35%)149.00 (-7.45%)133.00 (+0.00%)23.05 (+35.38%)
003b49d — 2025-11-14 19:30:540.09 (-5.26%)0.07 (-10.50%)0.07 (-13.04%)0.06 (-14.59%)0.01 (+16.82%)185.00 (+17.09%)161.40 (+12.24%)161.00 (+15.00%)133.00 (+5.56%)17.02 (+39.87%)
0999a9a — 2025-11-13 19:24:080.09 (-10.32%)0.08 (-0.04%)0.08 (+2.86%)0.07 (+26.58%)0.01 (-53.20%)158.00 (-21.00%)143.80 (-2.84%)140.00 (-2.78%)126.00 (+11.50%)12.17 (-58.69%)
fbf1591 — 2025-11-13 03:39:220.10 (n/a)0.08 (n/a)0.08 (n/a)0.06 (n/a)0.01 (n/a)200.00 (n/a)148.00 (n/a)144.00 (n/a)113.00 (n/a)29.47 (n/a)

eltwise_mul_8_cols_2_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.07 (-17.54%)0.06 (+1.17%)0.06 (+11.80%)0.06 (+4.71%)0.00 (-74.76%)191.00 (-4.50%)180.00 (-2.81%)178.00 (-10.55%)171.00 (+21.28%)6.69 (-70.55%)
003b49d — 2025-11-14 19:30:540.08 (+1.42%)0.06 (+3.51%)0.06 (-1.51%)0.06 (+40.00%)0.01 (-40.10%)200.00 (-28.57%)185.20 (-8.04%)199.00 (+1.53%)141.00 (-1.40%)22.73 (-56.70%)
0999a9a — 2025-11-13 19:24:080.08 (-2.80%)0.06 (-12.49%)0.06 (-12.76%)0.04 (-23.57%)0.02 (+38.68%)280.00 (+30.84%)201.40 (+19.03%)196.00 (+14.62%)143.00 (+2.88%)52.49 (+87.02%)
fbf1591 — 2025-11-13 03:39:220.08 (n/a)0.07 (n/a)0.07 (n/a)0.05 (n/a)0.01 (n/a)214.00 (n/a)169.20 (n/a)171.00 (n/a)139.00 (n/a)28.07 (n/a)

gelu_1_cols_1_channels_2048_tile_2048

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.06 (-8.09%)0.05 (-1.39%)0.05 (+1.32%)0.04 (-9.14%)0.01 (-8.06%)186.00 (+10.06%)154.40 (+1.45%)151.00 (-1.31%)136.00 (+8.80%)17.37 (+12.30%)
003b49d — 2025-11-14 19:30:540.07 (+11.20%)0.05 (+5.29%)0.05 (+3.27%)0.05 (+9.47%)0.01 (+27.90%)169.00 (-8.65%)152.20 (-4.76%)153.00 (-3.16%)125.00 (-10.07%)15.47 (+3.61%)
0999a9a — 2025-11-13 19:24:080.06 (-8.63%)0.05 (+6.51%)0.05 (+14.56%)0.04 (+27.03%)0.00 (-57.98%)185.00 (-21.28%)159.80 (-10.22%)158.00 (-12.71%)139.00 (+9.45%)14.93 (-63.00%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.05 (n/a)0.05 (n/a)0.03 (n/a)0.01 (n/a)235.00 (n/a)178.00 (n/a)181.00 (n/a)127.00 (n/a)40.36 (n/a)

gelu_1_cols_2_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.06 (-12.21%)0.05 (-0.13%)0.05 (-8.13%)0.05 (+28.82%)0.01 (-63.89%)170.00 (-22.37%)154.40 (-6.08%)160.00 (+8.84%)131.00 (+13.91%)13.48 (-69.75%)
003b49d — 2025-11-14 19:30:540.07 (+12.17%)0.05 (-4.06%)0.06 (+2.04%)0.04 (-28.77%)0.01 (+253.42%)219.00 (+40.38%)164.40 (+11.53%)147.00 (-2.00%)115.00 (-10.85%)44.59 (+368.31%)
0999a9a — 2025-11-13 19:24:080.06 (-10.08%)0.06 (-4.24%)0.05 (-8.67%)0.05 (+26.92%)0.00 (-59.85%)156.00 (-21.21%)147.40 (+1.52%)150.00 (+9.49%)129.00 (+11.21%)9.52 (-66.37%)
fbf1591 — 2025-11-13 03:39:220.07 (n/a)0.06 (n/a)0.06 (n/a)0.04 (n/a)0.01 (n/a)198.00 (n/a)145.20 (n/a)137.00 (n/a)116.00 (n/a)28.31 (n/a)

gelu_2_cols_1_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.06 (-24.66%)0.05 (-32.75%)0.05 (-36.81%)0.04 (-28.36%)0.01 (-17.17%)201.00 (+39.58%)178.20 (+49.00%)182.00 (+58.26%)146.00 (+32.73%)18.09 (+46.64%)
003b49d — 2025-11-14 19:30:540.07 (+3.64%)0.07 (+11.85%)0.07 (+9.57%)0.06 (+9.03%)0.01 (-21.10%)144.00 (-8.28%)119.60 (-11.28%)115.00 (-8.73%)110.00 (-3.51%)12.34 (-30.48%)
0999a9a — 2025-11-13 19:24:080.07 (-4.39%)0.06 (+1.89%)0.07 (+8.73%)0.05 (+31.85%)0.01 (-34.72%)157.00 (-24.15%)134.80 (-4.94%)126.00 (-8.03%)114.00 (+4.59%)17.75 (-48.47%)
fbf1591 — 2025-11-13 03:39:220.08 (n/a)0.06 (n/a)0.06 (n/a)0.04 (n/a)0.01 (n/a)207.00 (n/a)141.80 (n/a)137.00 (n/a)109.00 (n/a)34.44 (n/a)

gelu_2_cols_2_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.06 (-21.28%)0.05 (-15.38%)0.05 (-7.14%)0.03 (-30.42%)0.01 (-3.59%)240.00 (+43.71%)170.20 (+20.03%)154.00 (+7.69%)141.00 (+27.03%)36.70 (+76.06%)
003b49d — 2025-11-14 19:30:540.07 (-4.50%)0.06 (-5.43%)0.06 (-15.38%)0.05 (+19.16%)0.01 (-31.95%)167.00 (-16.08%)141.80 (+2.46%)143.00 (+18.18%)111.00 (+4.72%)20.85 (-39.65%)
0999a9a — 2025-11-13 19:24:080.08 (+21.70%)0.06 (+6.40%)0.07 (+12.40%)0.04 (-15.58%)0.01 (+146.44%)199.00 (+18.45%)138.40 (-1.70%)121.00 (-11.03%)106.00 (-17.83%)34.54 (+139.17%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.06 (n/a)0.06 (n/a)0.05 (n/a)0.01 (n/a)168.00 (n/a)140.80 (n/a)136.00 (n/a)129.00 (n/a)14.44 (n/a)

gelu_4_cols_1_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.06 (-8.09%)0.05 (-8.71%)0.06 (+4.90%)0.03 (-24.69%)0.01 (+36.44%)243.00 (+32.79%)168.00 (+12.75%)143.00 (-4.67%)136.00 (+8.80%)39.87 (+98.65%)
003b49d — 2025-11-14 19:30:540.07 (-1.60%)0.06 (-0.82%)0.05 (+2.67%)0.04 (-9.29%)0.01 (+13.57%)183.00 (+10.24%)149.00 (+1.36%)150.00 (-2.60%)125.00 (+1.63%)20.07 (+27.96%)
0999a9a — 2025-11-13 19:24:080.07 (+24.39%)0.06 (+20.95%)0.05 (+4.55%)0.05 (+34.94%)0.01 (-4.62%)166.00 (-25.89%)147.00 (-18.15%)154.00 (-4.35%)123.00 (-19.61%)15.68 (-43.29%)
fbf1591 — 2025-11-13 03:39:220.05 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)224.00 (n/a)179.60 (n/a)161.00 (n/a)153.00 (n/a)27.66 (n/a)

gelu_4_cols_2_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.05 (-15.53%)0.05 (-0.41%)0.05 (+1.18%)0.04 (+9.90%)0.00 (-50.18%)202.00 (-9.01%)175.20 (-1.46%)170.00 (-1.16%)161.00 (+18.38%)15.20 (-46.30%)
003b49d — 2025-11-14 19:30:540.06 (-2.94%)0.05 (-10.92%)0.05 (-11.63%)0.04 (-20.27%)0.01 (+38.43%)222.00 (+25.42%)177.80 (+13.97%)172.00 (+13.16%)136.00 (+3.03%)28.30 (+77.02%)
0999a9a — 2025-11-13 19:24:080.06 (-15.15%)0.05 (-2.99%)0.05 (+6.58%)0.05 (+2.82%)0.01 (-43.65%)177.00 (-2.75%)156.00 (+1.30%)152.00 (-6.17%)132.00 (+17.86%)15.99 (-33.29%)
fbf1591 — 2025-11-13 03:39:220.07 (n/a)0.05 (n/a)0.05 (n/a)0.05 (n/a)0.01 (n/a)182.00 (n/a)154.00 (n/a)162.00 (n/a)112.00 (n/a)23.97 (n/a)

gelu_8_cols_1_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.05 (-30.13%)0.04 (-35.29%)0.04 (-40.20%)0.03 (-37.82%)0.01 (-18.27%)275.00 (+60.82%)200.80 (+56.39%)199.00 (+67.23%)156.00 (+43.12%)40.58 (+85.26%)
003b49d — 2025-11-14 19:30:540.08 (+22.02%)0.07 (+23.76%)0.07 (+26.05%)0.05 (+12.28%)0.01 (+19.77%)171.00 (-10.94%)128.40 (-19.04%)119.00 (-20.67%)109.00 (-18.05%)21.91 (-9.09%)
0999a9a — 2025-11-13 19:24:080.06 (+4.51%)0.05 (+10.75%)0.05 (+16.67%)0.04 (+18.23%)0.01 (+4.18%)192.00 (-15.42%)158.60 (-9.99%)150.00 (-14.29%)133.00 (-4.32%)24.10 (-16.91%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)227.00 (n/a)176.20 (n/a)175.00 (n/a)139.00 (n/a)29.00 (n/a)

gelu_8_cols_2_channels_2048_tile_128

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.05 (-12.34%)0.04 (-16.38%)0.04 (-12.97%)0.03 (-22.18%)0.01 (+10.90%)266.00 (+28.50%)200.20 (+21.33%)185.00 (+14.91%)154.00 (+14.07%)37.83 (+62.19%)
003b49d — 2025-11-14 19:30:540.06 (+40.00%)0.05 (+23.27%)0.05 (+24.22%)0.04 (+3.38%)0.01 (+289.14%)207.00 (-3.27%)165.00 (-17.50%)161.00 (-19.50%)135.00 (-28.57%)23.32 (+174.11%)
0999a9a — 2025-11-13 19:24:080.04 (-20.63%)0.04 (-16.87%)0.04 (-20.00%)0.04 (-6.54%)0.00 (-67.29%)214.00 (+7.00%)200.00 (+19.05%)200.00 (+25.00%)189.00 (+26.00%)8.51 (-55.43%)
fbf1591 — 2025-11-13 03:39:220.05 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)200.00 (n/a)168.00 (n/a)160.00 (n/a)150.00 (n/a)19.09 (n/a)

gemm_2048x2048x2048_64x64x64_8_cols_0_bcolmaj_0_ccolmaj_0

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:573.69 (-3.45%)3.64 (-0.71%)3.68 (-0.28%)3.44 (+0.41%)0.10 (-25.16%)6805.00 (-0.41%)6449.20 (+0.66%)6363.00 (+0.28%)6347.00 (+3.57%)178.10 (-23.30%)
003b49d — 2025-11-14 19:30:543.82 (+4.26%)3.66 (+3.21%)3.69 (+4.46%)3.43 (+1.13%)0.13 (+21.15%)6833.00 (-1.11%)6407.20 (-3.07%)6345.00 (-4.27%)6128.00 (-4.09%)232.20 (+16.55%)
0999a9a — 2025-11-13 19:24:083.67 (-0.59%)3.55 (-2.49%)3.54 (-3.59%)3.39 (-3.00%)0.11 (+46.06%)6910.00 (+3.09%)6610.40 (+2.61%)6628.00 (+3.72%)6389.00 (+0.60%)199.23 (+50.27%)
fbf1591 — 2025-11-13 03:39:223.69 (n/a)3.64 (n/a)3.67 (n/a)3.50 (n/a)0.07 (n/a)6703.00 (n/a)6442.40 (n/a)6390.00 (n/a)6351.00 (n/a)132.58 (n/a)

layer_norm_1_cols_1_channels_2048_tile_2048

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.08 (-21.15%)0.07 (-24.11%)0.07 (-22.89%)0.05 (-36.54%)0.01 (+21.35%)260.00 (+57.58%)185.20 (+34.79%)166.00 (+29.69%)156.00 (+26.83%)38.19 (+148.90%)
003b49d — 2025-11-14 19:30:540.10 (+4.88%)0.09 (+38.25%)0.10 (+67.19%)0.07 (+74.55%)0.01 (-48.92%)165.00 (-42.71%)137.40 (-32.05%)128.00 (-40.19%)123.00 (-4.65%)15.34 (-71.52%)
0999a9a — 2025-11-13 19:24:080.10 (-3.88%)0.07 (-12.46%)0.06 (-21.03%)0.04 (-31.25%)0.02 (+36.72%)288.00 (+45.45%)202.20 (+19.64%)214.00 (+26.63%)129.00 (+4.03%)53.87 (+105.25%)
fbf1591 — 2025-11-13 03:39:220.10 (n/a)0.07 (n/a)0.07 (n/a)0.06 (n/a)0.01 (n/a)198.00 (n/a)169.00 (n/a)169.00 (n/a)124.00 (n/a)26.24 (n/a)

layer_norm_1_cols_2_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.09 (-17.52%)0.08 (-15.08%)0.08 (-12.10%)0.07 (-8.24%)0.01 (-44.09%)182.00 (+8.98%)159.80 (+16.30%)157.00 (+13.77%)137.00 (+21.24%)15.09 (-25.16%)
003b49d — 2025-11-14 19:30:540.11 (+22.12%)0.09 (+30.54%)0.09 (+28.99%)0.07 (+49.70%)0.01 (-6.97%)167.00 (-33.20%)137.40 (-25.16%)138.00 (-22.47%)113.00 (-18.12%)20.17 (-49.55%)
0999a9a — 2025-11-13 19:24:080.09 (-5.07%)0.07 (-8.39%)0.07 (-6.18%)0.05 (-16.40%)0.01 (+21.90%)250.00 (+19.62%)183.60 (+11.41%)178.00 (+6.59%)138.00 (+5.34%)39.97 (+52.27%)
fbf1591 — 2025-11-13 03:39:220.09 (n/a)0.08 (n/a)0.07 (n/a)0.06 (n/a)0.01 (n/a)209.00 (n/a)164.80 (n/a)167.00 (n/a)131.00 (n/a)26.25 (n/a)

layer_norm_2_cols_1_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.11 (+2.73%)0.08 (-9.74%)0.08 (-7.01%)0.07 (-13.79%)0.01 (+16.71%)174.00 (+16.00%)149.40 (+11.66%)157.00 (+7.53%)110.00 (-2.65%)21.32 (+28.12%)
003b49d — 2025-11-14 19:30:540.11 (+0.00%)0.09 (+8.35%)0.08 (+5.48%)0.08 (+15.33%)0.01 (-14.90%)150.00 (-13.29%)133.80 (-8.61%)146.00 (-5.19%)113.00 (+0.00%)16.64 (-27.06%)
0999a9a — 2025-11-13 19:24:080.11 (+10.62%)0.09 (+16.05%)0.08 (+8.44%)0.07 (+26.59%)0.01 (+2.65%)173.00 (-21.00%)146.40 (-14.49%)154.00 (-7.78%)113.00 (-9.60%)22.82 (-25.99%)
fbf1591 — 2025-11-13 03:39:220.10 (n/a)0.07 (n/a)0.07 (n/a)0.06 (n/a)0.01 (n/a)219.00 (n/a)171.20 (n/a)167.00 (n/a)125.00 (n/a)30.83 (n/a)

layer_norm_2_cols_2_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.10 (-9.02%)0.09 (-4.64%)0.09 (-7.58%)0.08 (+15.43%)0.01 (-44.79%)162.00 (-13.37%)136.60 (+2.40%)132.00 (+8.20%)122.00 (+9.91%)13.85 (-49.12%)
003b49d — 2025-11-14 19:30:540.11 (+9.91%)0.10 (+23.49%)0.10 (+54.10%)0.07 (+7.49%)0.02 (-12.01%)187.00 (-6.97%)133.40 (-20.22%)122.00 (-35.11%)111.00 (-9.02%)27.22 (-21.70%)
0999a9a — 2025-11-13 19:24:080.10 (+21.31%)0.08 (+19.49%)0.07 (+2.66%)0.06 (+20.90%)0.02 (+64.00%)201.00 (-17.28%)167.20 (-14.43%)188.00 (-2.59%)122.00 (-17.57%)34.76 (+12.90%)
fbf1591 — 2025-11-13 03:39:220.08 (n/a)0.06 (n/a)0.06 (n/a)0.05 (n/a)0.01 (n/a)243.00 (n/a)195.40 (n/a)193.00 (n/a)148.00 (n/a)30.79 (n/a)

layer_norm_4_cols_1_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.10 (-6.45%)0.08 (+8.60%)0.08 (+25.16%)0.07 (+17.74%)0.01 (-47.00%)186.00 (-15.07%)151.80 (-12.96%)155.00 (-20.10%)124.00 (+6.90%)21.39 (-52.48%)
003b49d — 2025-11-14 19:30:540.11 (+2.59%)0.08 (-11.68%)0.06 (-26.29%)0.06 (-21.46%)0.02 (+109.95%)219.00 (+27.33%)174.40 (+20.44%)194.00 (+35.66%)116.00 (-2.52%)45.00 (+163.45%)
0999a9a — 2025-11-13 19:24:080.10 (+30.25%)0.09 (+45.91%)0.09 (+54.55%)0.07 (+60.47%)0.01 (-12.42%)172.00 (-37.68%)144.80 (-33.02%)143.00 (-35.29%)119.00 (-23.23%)17.08 (-57.59%)
fbf1591 — 2025-11-13 03:39:220.08 (n/a)0.06 (n/a)0.06 (n/a)0.04 (n/a)0.01 (n/a)276.00 (n/a)216.20 (n/a)221.00 (n/a)155.00 (n/a)40.28 (n/a)

layer_norm_4_cols_2_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.09 (-12.23%)0.06 (-24.83%)0.06 (-30.85%)0.04 (-35.53%)0.02 (+55.63%)304.00 (+55.10%)208.60 (+42.88%)201.00 (+44.60%)139.00 (+13.93%)64.83 (+151.20%)
003b49d — 2025-11-14 19:30:540.10 (-17.21%)0.09 (-1.50%)0.09 (+8.63%)0.06 (-10.71%)0.01 (-28.46%)196.00 (+12.00%)146.00 (+0.69%)139.00 (-7.95%)122.00 (+20.79%)25.81 (+5.32%)
0999a9a — 2025-11-13 19:24:080.12 (+50.49%)0.09 (+47.35%)0.08 (+51.66%)0.07 (+66.86%)0.02 (+21.43%)175.00 (-40.07%)145.00 (-33.79%)151.00 (-34.06%)101.00 (-33.55%)24.50 (-52.68%)
fbf1591 — 2025-11-13 03:39:220.08 (n/a)0.06 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)292.00 (n/a)219.00 (n/a)229.00 (n/a)152.00 (n/a)51.78 (n/a)

layer_norm_8_cols_1_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.11 (+21.50%)0.09 (+37.43%)0.09 (+48.20%)0.07 (+27.65%)0.01 (-1.98%)170.00 (-21.66%)139.20 (-28.25%)139.00 (-32.52%)107.00 (-17.69%)20.73 (-36.02%)
003b49d — 2025-11-14 19:30:540.09 (+30.77%)0.07 (+16.03%)0.06 (-10.68%)0.06 (+62.67%)0.01 (-3.50%)217.00 (-38.53%)194.00 (-17.73%)206.00 (+11.96%)130.00 (-23.53%)32.40 (-55.20%)
0999a9a — 2025-11-13 19:24:080.07 (-31.76%)0.06 (-27.11%)0.07 (-16.85%)0.03 (-25.21%)0.01 (-24.96%)353.00 (+33.71%)235.80 (+38.06%)184.00 (+20.26%)170.00 (+46.55%)72.33 (+41.07%)
fbf1591 — 2025-11-13 03:39:220.11 (n/a)0.08 (n/a)0.08 (n/a)0.05 (n/a)0.02 (n/a)264.00 (n/a)170.80 (n/a)153.00 (n/a)116.00 (n/a)51.27 (n/a)

layer_norm_8_cols_2_channels_2048_tile_128

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.08 (+1.99%)0.07 (+0.26%)0.07 (-5.39%)0.07 (+20.67%)0.00 (-47.69%)179.00 (-17.13%)166.20 (-1.54%)167.00 (+5.70%)151.00 (-1.95%)9.95 (-58.07%)
003b49d — 2025-11-14 19:30:540.08 (+10.39%)0.07 (+14.33%)0.08 (+19.62%)0.06 (-6.02%)0.01 (+103.78%)216.00 (+6.40%)168.80 (-11.44%)158.00 (-16.40%)154.00 (-9.41%)23.73 (+98.66%)
0999a9a — 2025-11-13 19:24:080.07 (+0.00%)0.06 (+2.85%)0.07 (+1.59%)0.06 (+23.15%)0.00 (-51.87%)203.00 (-18.80%)190.60 (-4.41%)189.00 (-1.56%)170.00 (+0.00%)11.94 (-60.19%)
fbf1591 — 2025-11-13 03:39:220.07 (n/a)0.06 (n/a)0.06 (n/a)0.05 (n/a)0.01 (n/a)250.00 (n/a)199.40 (n/a)192.00 (n/a)170.00 (n/a)30.00 (n/a)

matrix_vector_mul_128x128_32_1col

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)GFLOP/s (max)GFLOP/s (mean)GFLOP/s (median)GFLOP/s (min)GFLOP/s (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.05 (+3.91%)0.04 (-4.57%)0.05 (-4.71%)0.03 (-15.57%)0.01 (+86.84%)0.05 (+3.91%)0.04 (-4.57%)0.05 (-4.71%)0.03 (-15.57%)0.01 (+86.84%)957.00 (+18.44%)756.20 (+6.30%)722.00 (+4.94%)639.00 (-3.77%)110.08 (+114.70%)
003b49d — 2025-11-14 19:30:540.05 (+3.92%)0.05 (+7.05%)0.05 (+16.13%)0.04 (+2.35%)0.00 (-8.09%)0.05 (+3.92%)0.05 (+7.05%)0.05 (+16.13%)0.04 (+2.35%)0.00 (-8.09%)808.00 (-2.30%)711.40 (-6.69%)688.00 (-13.89%)664.00 (-3.77%)51.27 (-11.52%)
0999a9a — 2025-11-13 19:24:080.05 (+4.35%)0.04 (+3.02%)0.04 (-0.25%)0.04 (+0.60%)0.00 (+62.83%)0.05 (+4.35%)0.04 (+3.02%)0.04 (-0.25%)0.04 (+0.60%)0.00 (+62.83%)827.00 (-0.60%)762.40 (-2.58%)799.00 (+0.25%)690.00 (-4.17%)57.94 (+53.99%)
fbf1591 — 2025-11-13 03:39:220.05 (n/a)0.04 (n/a)0.04 (n/a)0.04 (n/a)0.00 (n/a)0.05 (n/a)0.04 (n/a)0.04 (n/a)0.04 (n/a)0.00 (n/a)832.00 (n/a)782.60 (n/a)797.00 (n/a)720.00 (n/a)37.63 (n/a)

matrix_vector_mul_2048x8192_1_1col

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)GFLOP/s (max)GFLOP/s (mean)GFLOP/s (median)GFLOP/s (min)GFLOP/s (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:5711.20 (-1.10%)10.47 (-6.90%)11.06 (-1.91%)9.11 (-17.62%)0.84 (+762.18%)11.20 (-1.10%)10.46 (-6.90%)11.05 (-1.91%)9.11 (-17.62%)0.84 (+761.94%)3684.00 (+21.38%)3228.60 (+8.14%)3037.00 (+1.95%)2997.00 (+1.11%)273.75 (+947.14%)
003b49d — 2025-11-14 19:30:5411.33 (+2.29%)11.25 (+4.47%)11.27 (+4.06%)11.06 (+7.18%)0.10 (-63.80%)11.32 (+2.29%)11.24 (+4.47%)11.26 (+4.06%)11.06 (+7.18%)0.10 (-63.80%)3035.00 (-6.70%)2985.60 (-4.33%)2979.00 (-3.90%)2964.00 (-2.24%)26.14 (-66.99%)
0999a9a — 2025-11-13 19:24:0811.07 (-2.77%)10.77 (-0.84%)10.83 (+0.32%)10.32 (+0.46%)0.27 (-25.80%)11.07 (-2.77%)10.76 (-0.84%)10.82 (+0.32%)10.31 (+0.46%)0.27 (-25.79%)3253.00 (-0.46%)3120.80 (+0.79%)3100.00 (-0.32%)3032.00 (+2.85%)79.19 (-24.14%)
fbf1591 — 2025-11-13 03:39:2211.39 (n/a)10.86 (n/a)10.80 (n/a)10.27 (n/a)0.36 (n/a)11.38 (n/a)10.85 (n/a)10.79 (n/a)10.27 (n/a)0.36 (n/a)3268.00 (n/a)3096.20 (n/a)3110.00 (n/a)2948.00 (n/a)104.38 (n/a)

matrix_vector_mul_2048x8192_1_2col

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)GFLOP/s (max)GFLOP/s (mean)GFLOP/s (median)GFLOP/s (min)GFLOP/s (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:5717.51 (-2.66%)16.92 (-2.38%)16.65 (-6.25%)16.45 (+0.10%)0.45 (-34.59%)17.50 (-2.66%)16.91 (-2.38%)16.64 (-6.25%)16.44 (+0.10%)0.45 (-34.59%)2041.00 (-0.10%)1985.80 (+2.35%)2016.00 (+6.67%)1917.00 (+2.73%)52.40 (-33.11%)
003b49d — 2025-11-14 19:30:5417.99 (+0.70%)17.33 (+0.83%)17.76 (+3.23%)16.43 (-0.88%)0.69 (+60.54%)17.98 (+0.70%)17.32 (+0.83%)17.75 (+3.23%)16.42 (-0.88%)0.69 (+60.54%)2043.00 (+0.89%)1940.20 (-0.73%)1890.00 (-3.13%)1866.00 (-0.69%)78.34 (+61.13%)
0999a9a — 2025-11-13 19:24:0817.87 (+7.66%)17.19 (+6.72%)17.21 (+7.12%)16.58 (+6.07%)0.43 (+33.19%)17.86 (+7.66%)17.18 (+6.72%)17.20 (+7.12%)16.57 (+6.07%)0.43 (+33.18%)2025.00 (-5.73%)1954.40 (-6.27%)1951.00 (-6.65%)1879.00 (-7.12%)48.62 (+16.62%)
fbf1591 — 2025-11-13 03:39:2216.60 (n/a)16.11 (n/a)16.06 (n/a)15.63 (n/a)0.32 (n/a)16.59 (n/a)16.10 (n/a)16.05 (n/a)15.62 (n/a)0.32 (n/a)2148.00 (n/a)2085.20 (n/a)2090.00 (n/a)2023.00 (n/a)41.69 (n/a)

matrix_vector_mul_2048x8192_1_4col

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)GFLOP/s (max)GFLOP/s (mean)GFLOP/s (median)GFLOP/s (min)GFLOP/s (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:5722.50 (-0.80%)20.76 (-2.96%)20.45 (-4.63%)18.56 (-8.84%)1.49 (+65.81%)22.49 (-0.80%)20.75 (-2.96%)20.44 (-4.63%)18.55 (-8.84%)1.49 (+65.82%)1809.00 (+9.70%)1625.60 (+3.41%)1642.00 (+4.85%)1492.00 (+0.81%)117.99 (+79.96%)
003b49d — 2025-11-14 19:30:5422.69 (+4.66%)21.40 (+5.69%)21.44 (+4.21%)20.36 (+15.46%)0.90 (-36.26%)22.67 (+4.66%)21.38 (+5.69%)21.43 (+4.21%)20.35 (+15.46%)0.90 (-36.27%)1649.00 (-13.39%)1572.00 (-5.71%)1566.00 (-4.04%)1480.00 (-4.45%)65.57 (-47.58%)
0999a9a — 2025-11-13 19:24:0821.68 (+5.04%)20.24 (+6.30%)20.57 (+9.74%)17.63 (-4.10%)1.41 (+69.34%)21.66 (+5.04%)20.23 (+6.30%)20.56 (+9.74%)17.62 (-4.10%)1.41 (+69.35%)1904.00 (+4.27%)1667.20 (-5.61%)1632.00 (-8.88%)1549.00 (-4.79%)125.07 (+70.83%)
fbf1591 — 2025-11-13 03:39:2220.64 (n/a)19.04 (n/a)18.75 (n/a)18.39 (n/a)0.83 (n/a)20.62 (n/a)19.03 (n/a)18.73 (n/a)18.38 (n/a)0.83 (n/a)1826.00 (n/a)1766.20 (n/a)1791.00 (n/a)1627.00 (n/a)73.21 (n/a)

matrix_vector_mul_2048x8192_1_8col

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)GFLOP/s (max)GFLOP/s (mean)GFLOP/s (median)GFLOP/s (min)GFLOP/s (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:5720.04 (-0.12%)17.70 (-4.20%)18.36 (-2.08%)15.06 (-2.33%)2.21 (+34.70%)20.03 (-0.12%)17.69 (-4.20%)18.35 (-2.08%)15.05 (-2.33%)2.21 (+34.70%)2229.00 (+2.39%)1927.40 (+5.16%)1829.00 (+2.12%)1675.00 (+0.12%)248.25 (+37.59%)
003b49d — 2025-11-14 19:30:5420.07 (+13.87%)18.48 (+9.46%)18.75 (+9.44%)15.42 (-2.71%)1.64 (+166.63%)20.06 (+13.87%)18.47 (+9.46%)18.73 (+9.44%)15.41 (-2.71%)1.64 (+166.62%)2177.00 (+2.79%)1832.80 (-7.96%)1791.00 (-8.62%)1673.00 (-12.18%)180.43 (+143.03%)
0999a9a — 2025-11-13 19:24:0817.62 (-9.34%)16.88 (+0.64%)17.13 (+7.09%)15.85 (+17.61%)0.62 (-73.17%)17.61 (-9.34%)16.87 (+0.64%)17.12 (+7.09%)15.84 (+17.61%)0.62 (-73.17%)2118.00 (-14.97%)1991.40 (-2.39%)1960.00 (-6.62%)1905.00 (+10.31%)74.24 (-73.99%)
fbf1591 — 2025-11-13 03:39:2219.44 (n/a)16.78 (n/a)16.00 (n/a)13.48 (n/a)2.29 (n/a)19.43 (n/a)16.77 (n/a)15.99 (n/a)13.47 (n/a)2.29 (n/a)2491.00 (n/a)2040.20 (n/a)2099.00 (n/a)1727.00 (n/a)285.41 (n/a)

matrix_vector_mul_8192x2048_4_1col

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)GFLOP/s (max)GFLOP/s (mean)GFLOP/s (median)GFLOP/s (min)GFLOP/s (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:5711.16 (-1.99%)10.97 (+2.41%)11.01 (+1.08%)10.78 (+7.70%)0.15 (-73.26%)11.15 (-1.99%)10.96 (+2.41%)11.00 (+1.08%)10.77 (+7.70%)0.15 (-73.26%)3116.00 (-7.15%)3061.20 (-2.60%)3050.00 (-1.07%)3009.00 (+2.03%)41.72 (-74.80%)
003b49d — 2025-11-14 19:30:5411.39 (+1.76%)10.71 (-3.29%)10.89 (-1.49%)10.00 (-8.91%)0.56 (+682.28%)11.38 (+1.76%)10.71 (-3.29%)10.88 (-1.49%)10.00 (-8.91%)0.56 (+682.20%)3356.00 (+9.78%)3143.00 (+3.68%)3083.00 (+1.51%)2949.00 (-1.73%)165.56 (+749.51%)
0999a9a — 2025-11-13 19:24:0811.19 (-1.03%)11.08 (+3.60%)11.06 (+2.44%)10.98 (+13.09%)0.07 (-86.64%)11.18 (-1.03%)11.07 (+3.60%)11.05 (+2.44%)10.98 (+13.09%)0.07 (-86.64%)3057.00 (-11.57%)3031.40 (-3.72%)3037.00 (-2.38%)3001.00 (+1.04%)19.49 (-88.20%)
fbf1591 — 2025-11-13 03:39:2211.30 (n/a)10.69 (n/a)10.79 (n/a)9.71 (n/a)0.53 (n/a)11.30 (n/a)10.68 (n/a)10.79 (n/a)9.71 (n/a)0.53 (n/a)3457.00 (n/a)3148.60 (n/a)3111.00 (n/a)2970.00 (n/a)165.20 (n/a)

matrix_vector_mul_8192x2048_4_2col

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)GFLOP/s (max)GFLOP/s (mean)GFLOP/s (median)GFLOP/s (min)GFLOP/s (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:5717.81 (-0.11%)17.49 (+2.40%)17.43 (-1.51%)17.20 (+14.91%)0.23 (-79.05%)17.80 (-0.11%)17.48 (+2.40%)17.42 (-1.51%)17.19 (+14.91%)0.23 (-79.05%)1952.00 (-12.97%)1919.80 (-2.76%)1926.00 (+1.53%)1885.00 (+0.11%)24.87 (-81.85%)
003b49d — 2025-11-14 19:30:5417.83 (+2.87%)17.08 (+1.50%)17.70 (+3.69%)14.97 (-7.18%)1.08 (+127.09%)17.82 (+2.87%)17.07 (+1.50%)17.69 (+3.69%)14.96 (-7.18%)1.08 (+127.10%)2243.00 (+7.73%)1974.20 (-1.12%)1897.00 (-3.56%)1883.00 (-2.79%)137.03 (+139.02%)
0999a9a — 2025-11-13 19:24:0817.33 (+0.67%)16.83 (+5.22%)17.07 (+7.78%)16.13 (+11.53%)0.48 (-49.24%)17.32 (+0.67%)16.82 (+5.22%)17.06 (+7.78%)16.12 (+11.53%)0.48 (-49.24%)2082.00 (-10.34%)1996.60 (-5.22%)1967.00 (-7.22%)1937.00 (-0.67%)57.33 (-54.90%)
fbf1591 — 2025-11-13 03:39:2217.22 (n/a)15.99 (n/a)15.84 (n/a)14.46 (n/a)0.94 (n/a)17.21 (n/a)15.98 (n/a)15.83 (n/a)14.45 (n/a)0.94 (n/a)2322.00 (n/a)2106.60 (n/a)2120.00 (n/a)1950.00 (n/a)127.11 (n/a)

matrix_vector_mul_8192x2048_4_4col

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)GFLOP/s (max)GFLOP/s (mean)GFLOP/s (median)GFLOP/s (min)GFLOP/s (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:5722.69 (+4.05%)20.31 (-0.42%)19.32 (-5.64%)18.53 (-0.28%)1.68 (+29.34%)22.67 (+4.05%)20.29 (-0.42%)19.31 (-5.64%)18.52 (-0.28%)1.68 (+29.33%)1812.00 (+0.28%)1664.40 (+0.68%)1738.00 (+5.98%)1480.00 (-3.90%)133.61 (+25.34%)
003b49d — 2025-11-14 19:30:5421.80 (+6.69%)20.39 (+5.58%)20.47 (+5.55%)18.58 (+6.36%)1.30 (+22.12%)21.79 (+6.69%)20.38 (+5.58%)20.46 (+5.55%)18.57 (+6.36%)1.30 (+22.13%)1807.00 (-5.98%)1653.20 (-5.20%)1640.00 (-5.26%)1540.00 (-6.27%)106.60 (+6.49%)
0999a9a — 2025-11-13 19:24:0820.44 (+1.76%)19.31 (+0.88%)19.40 (+1.56%)17.47 (-4.73%)1.06 (+82.49%)20.42 (+1.77%)19.30 (+0.88%)19.38 (+1.56%)17.46 (-4.73%)1.06 (+82.49%)1922.00 (+4.97%)1743.80 (-0.65%)1731.00 (-1.54%)1643.00 (-1.73%)100.11 (+88.57%)
fbf1591 — 2025-11-13 03:39:2220.08 (n/a)19.15 (n/a)19.10 (n/a)18.34 (n/a)0.58 (n/a)20.07 (n/a)19.13 (n/a)19.09 (n/a)18.33 (n/a)0.58 (n/a)1831.00 (n/a)1755.20 (n/a)1758.00 (n/a)1672.00 (n/a)53.09 (n/a)

matrix_vector_mul_8192x2048_4_8col

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)GFLOP/s (max)GFLOP/s (mean)GFLOP/s (median)GFLOP/s (min)GFLOP/s (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:5719.70 (+2.35%)17.41 (-6.06%)17.96 (-3.80%)14.78 (-16.34%)1.64 (+213.84%)19.69 (+2.35%)17.40 (-6.06%)17.95 (-3.80%)14.78 (-16.34%)1.64 (+213.84%)2271.00 (+19.53%)1946.20 (+7.36%)1869.00 (+3.95%)1704.00 (-2.29%)191.36 (+269.03%)
003b49d — 2025-11-14 19:30:5419.25 (+1.20%)18.54 (+8.24%)18.67 (+13.79%)17.67 (+13.11%)0.52 (-61.54%)19.24 (+1.20%)18.52 (+8.24%)18.66 (+13.79%)17.66 (+13.10%)0.52 (-61.54%)1900.00 (-11.59%)1812.80 (-8.11%)1798.00 (-12.12%)1744.00 (-1.19%)51.86 (-66.09%)
0999a9a — 2025-11-13 19:24:0819.02 (+1.64%)17.12 (-4.33%)16.41 (-9.04%)15.62 (-3.86%)1.36 (+57.46%)19.01 (+1.64%)17.11 (-4.33%)16.40 (-9.04%)15.61 (-3.86%)1.36 (+57.46%)2149.00 (+4.02%)1972.80 (+4.91%)2046.00 (+9.94%)1765.00 (-1.62%)152.91 (+58.58%)
fbf1591 — 2025-11-13 03:39:2218.72 (n/a)17.90 (n/a)18.04 (n/a)16.25 (n/a)0.86 (n/a)18.70 (n/a)17.89 (n/a)18.03 (n/a)16.24 (n/a)0.86 (n/a)2066.00 (n/a)1880.40 (n/a)1861.00 (n/a)1794.00 (n/a)96.42 (n/a)

mem_copy_16_cores_2_chans_2048_tile_128_False

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)183.00 (n/a)157.60 (n/a)167.00 (n/a)125.00 (n/a)21.96 (n/a)

mem_copy_1_cols_1_channels_2048_tile_2048

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
003b49d — 2025-11-14 19:30:540.05 (-16.99%)0.05 (-12.47%)0.05 (-7.32%)0.04 (-11.17%)0.00 (-29.61%)197.00 (+12.57%)171.00 (+13.70%)164.00 (+7.89%)153.00 (+20.47%)17.49 (-4.32%)
0999a9a — 2025-11-13 19:24:080.06 (+5.51%)0.05 (+5.10%)0.05 (-6.58%)0.04 (+18.29%)0.01 (-17.12%)175.00 (-15.46%)150.40 (-6.00%)152.00 (+7.04%)127.00 (-5.22%)18.28 (-33.91%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)207.00 (n/a)160.00 (n/a)142.00 (n/a)134.00 (n/a)27.66 (n/a)

mem_copy_1_cols_2_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
003b49d — 2025-11-14 19:30:540.06 (+5.56%)0.05 (-1.59%)0.05 (-1.39%)0.03 (-14.07%)0.01 (+25.03%)263.00 (+16.37%)166.00 (+4.53%)144.00 (+1.41%)126.00 (-5.26%)49.38 (+43.17%)
0999a9a — 2025-11-13 19:24:080.06 (-9.02%)0.05 (-8.21%)0.05 (+2.82%)0.03 (-34.51%)0.01 (+94.29%)226.00 (+52.70%)158.80 (+12.46%)142.00 (-2.74%)133.00 (+9.92%)34.49 (+235.98%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.05 (n/a)0.05 (n/a)0.05 (n/a)0.00 (n/a)148.00 (n/a)141.20 (n/a)146.00 (n/a)121.00 (n/a)10.26 (n/a)

mem_copy_1_cores_1_chans_2048_tile_2048_False

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.06 (n/a)0.05 (n/a)0.06 (n/a)0.05 (n/a)0.01 (n/a)169.00 (n/a)141.20 (n/a)135.00 (n/a)128.00 (n/a)15.38 (n/a)

mem_copy_2_cols_1_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
003b49d — 2025-11-14 19:30:540.06 (-15.32%)0.05 (-17.09%)0.05 (-23.81%)0.04 (-15.54%)0.01 (-18.18%)193.00 (+18.40%)160.80 (+20.36%)168.00 (+31.25%)124.00 (+18.10%)22.92 (+11.39%)
0999a9a — 2025-11-13 19:24:080.07 (+18.10%)0.06 (+4.90%)0.06 (+5.47%)0.05 (-7.36%)0.01 (+117.26%)163.00 (+7.95%)133.60 (-2.91%)128.00 (-5.19%)105.00 (-15.32%)20.58 (+98.52%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.06 (n/a)0.06 (n/a)0.05 (n/a)0.00 (n/a)151.00 (n/a)137.60 (n/a)135.00 (n/a)124.00 (n/a)10.37 (n/a)

mem_copy_2_cols_2_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
003b49d — 2025-11-14 19:30:540.06 (-23.02%)0.05 (-6.98%)0.05 (-2.13%)0.04 (-15.31%)0.01 (-33.29%)196.00 (+18.07%)147.00 (+6.52%)141.00 (+2.17%)126.00 (+29.90%)25.27 (+8.13%)
0999a9a — 2025-11-13 19:24:080.08 (+26.80%)0.06 (+22.42%)0.06 (+33.33%)0.05 (+43.98%)0.01 (+1.58%)166.00 (-30.54%)138.00 (-20.32%)138.00 (-25.00%)97.00 (-21.14%)23.37 (-44.22%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.05 (n/a)0.04 (n/a)0.03 (n/a)0.01 (n/a)239.00 (n/a)173.20 (n/a)184.00 (n/a)123.00 (n/a)41.89 (n/a)

mem_copy_2_cores_1_chans_2048_tile_1024_False

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.06 (n/a)0.05 (n/a)0.05 (n/a)0.03 (n/a)0.01 (n/a)264.00 (n/a)175.00 (n/a)142.00 (n/a)123.00 (n/a)52.73 (n/a)

mem_copy_2_cores_2_chans_2048_tile_1024_False

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.06 (n/a)0.05 (n/a)0.05 (n/a)0.05 (n/a)0.01 (n/a)153.00 (n/a)141.80 (n/a)151.00 (n/a)123.00 (n/a)12.70 (n/a)

mem_copy_4_cols_1_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
003b49d — 2025-11-14 19:30:540.07 (+11.97%)0.05 (+3.09%)0.05 (+2.10%)0.03 (-11.35%)0.01 (+41.65%)229.00 (+12.81%)156.00 (-0.51%)143.00 (-2.05%)117.00 (-10.69%)38.21 (+48.71%)
0999a9a — 2025-11-13 19:24:080.06 (-12.98%)0.05 (-13.35%)0.05 (-10.27%)0.04 (-19.21%)0.01 (+7.53%)203.00 (+23.78%)156.80 (+16.49%)146.00 (+11.45%)131.00 (+14.91%)25.69 (+52.83%)
fbf1591 — 2025-11-13 03:39:220.07 (n/a)0.06 (n/a)0.06 (n/a)0.05 (n/a)0.01 (n/a)164.00 (n/a)134.60 (n/a)131.00 (n/a)114.00 (n/a)16.81 (n/a)

mem_copy_4_cols_2_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
003b49d — 2025-11-14 19:30:540.06 (+9.02%)0.05 (+5.59%)0.05 (+0.00%)0.04 (+3.63%)0.01 (+25.48%)193.00 (-3.50%)147.40 (-4.66%)147.00 (+0.00%)122.00 (-8.27%)25.24 (+7.48%)
0999a9a — 2025-11-13 19:24:080.06 (+33.83%)0.05 (+30.33%)0.05 (+29.93%)0.04 (+27.00%)0.01 (+42.50%)200.00 (-21.26%)154.60 (-23.01%)147.00 (-23.04%)133.00 (-25.28%)23.48 (-14.85%)
fbf1591 — 2025-11-13 03:39:220.04 (n/a)0.04 (n/a)0.04 (n/a)0.03 (n/a)0.00 (n/a)254.00 (n/a)200.80 (n/a)191.00 (n/a)178.00 (n/a)27.58 (n/a)

mem_copy_4_cores_1_chans_2048_tile_512_False

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.07 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)198.00 (n/a)145.20 (n/a)140.00 (n/a)117.00 (n/a)28.67 (n/a)

mem_copy_4_cores_2_chans_2048_tile_512_False

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)184.00 (n/a)154.20 (n/a)152.00 (n/a)134.00 (n/a)16.55 (n/a)

mem_copy_8_cols_1_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
003b49d — 2025-11-14 19:30:540.05 (-13.48%)0.05 (-17.79%)0.05 (-15.33%)0.03 (-25.09%)0.01 (+6.67%)267.00 (+33.50%)176.80 (+24.51%)150.00 (+18.11%)141.00 (+15.57%)47.41 (+61.41%)
0999a9a — 2025-11-13 19:24:080.06 (+29.51%)0.06 (+23.73%)0.06 (+26.77%)0.04 (+1.00%)0.01 (+118.73%)200.00 (-0.99%)142.00 (-17.15%)127.00 (-21.12%)122.00 (-22.78%)29.37 (+72.65%)
fbf1591 — 2025-11-13 03:39:220.05 (n/a)0.04 (n/a)0.05 (n/a)0.04 (n/a)0.00 (n/a)202.00 (n/a)171.40 (n/a)161.00 (n/a)158.00 (n/a)17.01 (n/a)

mem_copy_8_cols_2_channels_2048_tile_128

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
003b49d — 2025-11-14 19:30:540.05 (+13.38%)0.05 (+23.85%)0.05 (+37.91%)0.05 (+30.72%)0.00 (-56.85%)166.00 (-23.50%)153.40 (-20.60%)153.00 (-27.49%)142.00 (-11.80%)7.63 (-70.60%)
0999a9a — 2025-11-13 19:24:080.05 (-1.86%)0.04 (+3.84%)0.04 (-0.95%)0.04 (-0.92%)0.01 (+18.78%)217.00 (+0.93%)193.20 (-3.11%)211.00 (+0.96%)161.00 (+1.90%)25.96 (+23.15%)
fbf1591 — 2025-11-13 03:39:220.05 (n/a)0.04 (n/a)0.04 (n/a)0.04 (n/a)0.00 (n/a)215.00 (n/a)199.40 (n/a)209.00 (n/a)158.00 (n/a)21.08 (n/a)

mem_copy_8_cores_1_chans_2048_tile_256_False

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)202.00 (n/a)155.60 (n/a)148.00 (n/a)121.00 (n/a)29.76 (n/a)

mem_copy_8_cores_2_chans_2048_tile_256_False

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.05 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.00 (n/a)208.00 (n/a)170.40 (n/a)163.00 (n/a)156.00 (n/a)19.42 (n/a)

mha

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)GFLOP/s (max)GFLOP/s (mean)GFLOP/s (median)GFLOP/s (min)GFLOP/s (stddev)
8404bd4 — 2025-11-15 00:03:570.10 (+13.78%)0.09 (+4.69%)0.09 (+0.83%)0.09 (+4.72%)0.01 (+118.58%)74.38 (+13.78%)65.80 (+4.69%)63.68 (+0.83%)63.32 (+4.72%)4.30 (+118.58%)
003b49d — 2025-11-14 19:30:540.09 (+1.46%)0.09 (+0.19%)0.09 (-1.11%)0.08 (+3.18%)0.00 (-7.67%)65.37 (+1.46%)62.85 (+0.19%)63.16 (-1.11%)60.47 (+3.18%)1.97 (-7.67%)
0999a9a — 2025-11-13 19:24:080.09 (-3.49%)0.09 (-2.49%)0.09 (+0.19%)0.08 (-6.78%)0.00 (+57.78%)64.43 (-3.49%)62.73 (-2.49%)63.87 (+0.19%)58.61 (-6.78%)2.13 (+57.77%)
fbf1591 — 2025-11-13 03:39:220.09 (n/a)0.09 (n/a)0.09 (n/a)0.09 (n/a)0.00 (n/a)66.76 (n/a)64.33 (n/a)63.75 (n/a)62.87 (n/a)1.35 (n/a)

relu_1_cols_1_channels_2048_tile_2048

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.05 (-24.22%)0.04 (-10.91%)0.04 (-12.71%)0.04 (+16.58%)0.00 (-74.53%)187.00 (-14.22%)176.60 (+7.29%)181.00 (+14.56%)161.00 (+31.97%)10.69 (-70.81%)
003b49d — 2025-11-14 19:30:540.06 (+19.67%)0.05 (+13.97%)0.05 (+7.59%)0.03 (+13.30%)0.01 (+44.26%)218.00 (-11.74%)164.60 (-10.83%)158.00 (-7.06%)122.00 (-16.44%)36.61 (+3.83%)
0999a9a — 2025-11-13 19:24:080.05 (-8.90%)0.04 (-12.92%)0.04 (-7.65%)0.03 (-17.00%)0.01 (-3.29%)247.00 (+20.49%)184.60 (+15.66%)170.00 (+8.28%)146.00 (+9.77%)35.26 (+32.92%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)205.00 (n/a)159.60 (n/a)157.00 (n/a)133.00 (n/a)26.53 (n/a)

relu_2_cols_1_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.06 (+5.19%)0.05 (-1.21%)0.05 (-2.40%)0.04 (-16.28%)0.01 (+75.96%)215.00 (+19.44%)168.80 (+2.93%)167.00 (+2.45%)135.00 (-4.93%)26.52 (+103.41%)
003b49d — 2025-11-14 19:30:540.05 (-7.75%)0.05 (-10.61%)0.05 (-8.59%)0.04 (-12.78%)0.00 (+16.25%)180.00 (+14.65%)164.00 (+12.18%)163.00 (+9.40%)142.00 (+8.40%)13.04 (+44.55%)
0999a9a — 2025-11-13 19:24:080.06 (-2.29%)0.05 (-3.00%)0.05 (-5.37%)0.05 (+3.82%)0.00 (-22.25%)157.00 (-3.68%)146.20 (+2.81%)149.00 (+5.67%)131.00 (+2.34%)9.02 (-24.35%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.05 (n/a)0.05 (n/a)0.05 (n/a)0.00 (n/a)163.00 (n/a)142.20 (n/a)141.00 (n/a)128.00 (n/a)11.92 (n/a)

relu_4_cols_1_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.06 (-3.70%)0.05 (-2.67%)0.05 (-1.83%)0.04 (-11.21%)0.01 (+9.00%)214.00 (+12.63%)168.60 (+3.44%)164.00 (+1.86%)135.00 (+3.85%)26.55 (+29.69%)
003b49d — 2025-11-14 19:30:540.06 (+6.15%)0.05 (-2.89%)0.05 (-8.70%)0.04 (+0.53%)0.01 (+5.33%)190.00 (-0.52%)163.00 (+3.03%)161.00 (+9.52%)130.00 (-5.80%)20.47 (-1.37%)
0999a9a — 2025-11-13 19:24:080.06 (-2.17%)0.05 (+19.50%)0.05 (+11.56%)0.04 (+60.73%)0.01 (-53.04%)191.00 (-37.79%)158.20 (-24.16%)147.00 (-10.37%)138.00 (+2.22%)20.76 (-71.43%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.04 (n/a)0.05 (n/a)0.02 (n/a)0.01 (n/a)307.00 (n/a)208.60 (n/a)164.00 (n/a)135.00 (n/a)72.66 (n/a)

relu_8_cols_1_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.05 (-3.33%)0.04 (-5.59%)0.04 (-6.90%)0.03 (-19.40%)0.01 (+59.79%)232.00 (+24.06%)178.00 (+7.62%)174.00 (+7.41%)150.00 (+3.45%)28.71 (+109.67%)
003b49d — 2025-11-14 19:30:540.05 (-4.83%)0.05 (-4.03%)0.05 (-12.96%)0.04 (+7.49%)0.00 (-50.79%)187.00 (-6.97%)165.40 (+1.97%)162.00 (+14.89%)145.00 (+5.07%)13.69 (-51.39%)
0999a9a — 2025-11-13 19:24:080.06 (-0.72%)0.05 (+0.25%)0.05 (+13.48%)0.04 (-15.42%)0.01 (+99.76%)201.00 (+18.24%)162.20 (+2.01%)141.00 (-11.88%)138.00 (+0.73%)28.17 (+137.04%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.00 (n/a)170.00 (n/a)159.00 (n/a)160.00 (n/a)137.00 (n/a)11.88 (n/a)

rms_norm_1_cols_1_channels_2048_tile_2048

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.09 (-16.54%)0.08 (-8.93%)0.08 (-12.16%)0.07 (+0.60%)0.01 (-51.89%)166.00 (-0.60%)148.00 (+8.35%)148.00 (+13.85%)133.00 (+19.82%)10.75 (-42.84%)
003b49d — 2025-11-14 19:30:540.10 (+2.70%)0.09 (-7.17%)0.09 (-3.08%)0.07 (-17.37%)0.01 (+64.46%)167.00 (+21.01%)136.60 (+9.11%)130.00 (+3.17%)111.00 (-2.63%)18.81 (+97.23%)
0999a9a — 2025-11-13 19:24:080.10 (+12.28%)0.09 (+14.91%)0.09 (+8.73%)0.08 (+26.09%)0.01 (-17.60%)138.00 (-20.69%)125.20 (-13.54%)126.00 (-8.03%)114.00 (-10.94%)9.54 (-42.73%)
fbf1591 — 2025-11-13 03:39:220.09 (n/a)0.08 (n/a)0.08 (n/a)0.07 (n/a)0.01 (n/a)174.00 (n/a)144.80 (n/a)137.00 (n/a)128.00 (n/a)16.65 (n/a)

rms_norm_1_cols_2_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.09 (+4.88%)0.08 (-0.36%)0.08 (+0.00%)0.06 (-16.32%)0.01 (+95.38%)190.00 (+19.50%)149.20 (+2.05%)148.00 (+0.00%)123.00 (-4.65%)22.90 (+127.49%)
003b49d — 2025-11-14 19:30:540.09 (-12.40%)0.08 (-10.33%)0.08 (-17.57%)0.07 (+37.11%)0.01 (-68.45%)159.00 (-27.06%)146.20 (+5.48%)148.00 (+21.31%)129.00 (+14.16%)10.07 (-74.79%)
0999a9a — 2025-11-13 19:24:080.10 (+13.27%)0.09 (+21.98%)0.09 (+26.23%)0.05 (+14.68%)0.02 (+21.77%)218.00 (-12.80%)138.60 (-17.40%)122.00 (-20.78%)113.00 (-11.72%)39.94 (-7.15%)
fbf1591 — 2025-11-13 03:39:220.09 (n/a)0.07 (n/a)0.07 (n/a)0.05 (n/a)0.01 (n/a)250.00 (n/a)167.80 (n/a)154.00 (n/a)128.00 (n/a)43.02 (n/a)

rms_norm_2_cols_1_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.09 (+6.25%)0.08 (+6.97%)0.09 (+13.74%)0.07 (+2.96%)0.01 (+24.54%)169.00 (-2.87%)142.40 (-6.19%)131.00 (-12.08%)128.00 (-5.88%)16.04 (+13.52%)
003b49d — 2025-11-14 19:30:540.08 (-30.88%)0.08 (-18.19%)0.08 (-22.82%)0.07 (+31.61%)0.01 (-70.84%)174.00 (-24.02%)151.80 (+12.44%)149.00 (+29.57%)136.00 (+44.68%)14.13 (-70.53%)
0999a9a — 2025-11-13 19:24:080.12 (+47.87%)0.09 (+28.65%)0.10 (+46.09%)0.05 (-25.76%)0.02 (+304.80%)229.00 (+34.71%)135.00 (-15.31%)115.00 (-31.55%)94.00 (-32.37%)47.96 (+296.61%)
fbf1591 — 2025-11-13 03:39:220.08 (n/a)0.07 (n/a)0.07 (n/a)0.07 (n/a)0.01 (n/a)170.00 (n/a)159.40 (n/a)168.00 (n/a)139.00 (n/a)12.09 (n/a)

rms_norm_2_cols_2_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.09 (-3.08%)0.08 (+3.79%)0.07 (-7.69%)0.06 (+59.44%)0.01 (-48.81%)180.00 (-37.28%)154.60 (-10.22%)156.00 (+8.33%)130.00 (+3.17%)18.30 (-68.80%)
003b49d — 2025-11-14 19:30:540.09 (-2.38%)0.07 (-21.13%)0.08 (-13.89%)0.04 (-55.40%)0.02 (+1143.63%)287.00 (+124.22%)172.20 (+37.98%)144.00 (+16.13%)126.00 (+2.44%)58.67 (+2925.85%)
0999a9a — 2025-11-13 19:24:080.09 (-14.63%)0.09 (+21.34%)0.09 (+29.03%)0.09 (+88.28%)0.00 (-92.92%)128.00 (-46.89%)124.80 (-23.25%)124.00 (-22.50%)123.00 (+17.14%)1.94 (-95.64%)
fbf1591 — 2025-11-13 03:39:220.11 (n/a)0.08 (n/a)0.07 (n/a)0.05 (n/a)0.02 (n/a)241.00 (n/a)162.60 (n/a)160.00 (n/a)105.00 (n/a)44.50 (n/a)

rms_norm_4_cols_1_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.11 (+23.53%)0.08 (+8.76%)0.08 (+18.18%)0.06 (-5.32%)0.02 (+55.22%)188.00 (+5.62%)145.40 (-5.83%)143.00 (-15.38%)102.00 (-19.05%)30.58 (+35.45%)
003b49d — 2025-11-14 19:30:540.09 (-2.38%)0.08 (-17.97%)0.07 (-26.63%)0.06 (-29.78%)0.01 (+2002.65%)178.00 (+42.40%)154.40 (+24.72%)169.00 (+36.29%)126.00 (+2.44%)22.58 (+2917.33%)
0999a9a — 2025-11-13 19:24:080.09 (-0.81%)0.09 (+14.93%)0.09 (+16.13%)0.09 (+35.20%)0.00 (-94.54%)125.00 (-26.04%)123.80 (-14.38%)124.00 (-13.89%)123.00 (+0.82%)0.75 (-95.92%)
fbf1591 — 2025-11-13 03:39:220.09 (n/a)0.08 (n/a)0.08 (n/a)0.07 (n/a)0.01 (n/a)169.00 (n/a)144.60 (n/a)144.00 (n/a)122.00 (n/a)18.36 (n/a)

rms_norm_4_cols_2_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.08 (+17.65%)0.07 (+13.89%)0.07 (+16.87%)0.06 (+12.04%)0.01 (+28.22%)191.00 (-10.75%)165.20 (-11.94%)166.00 (-14.43%)136.00 (-15.00%)19.43 (-1.97%)
003b49d — 2025-11-14 19:30:540.07 (-19.37%)0.06 (-18.59%)0.06 (-26.29%)0.05 (+20.56%)0.01 (-59.00%)214.00 (-17.05%)187.60 (+16.23%)194.00 (+35.66%)160.00 (+24.03%)19.82 (-59.41%)
0999a9a — 2025-11-13 19:24:080.09 (+7.75%)0.08 (+6.49%)0.08 (+6.29%)0.04 (-1.94%)0.02 (+21.12%)258.00 (+1.98%)161.40 (-4.38%)143.00 (-5.92%)129.00 (-7.19%)48.81 (+14.45%)
fbf1591 — 2025-11-13 03:39:220.08 (n/a)0.07 (n/a)0.08 (n/a)0.05 (n/a)0.01 (n/a)253.00 (n/a)168.80 (n/a)152.00 (n/a)139.00 (n/a)42.65 (n/a)

rms_norm_8_cols_1_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.09 (-1.50%)0.08 (+6.65%)0.08 (+13.79%)0.07 (+8.81%)0.01 (-38.53%)159.00 (-8.09%)145.60 (-7.02%)145.00 (-12.12%)133.00 (+1.53%)9.83 (-43.38%)
003b49d — 2025-11-14 19:30:540.09 (+9.92%)0.07 (+3.67%)0.07 (+2.42%)0.07 (+0.00%)0.01 (+55.18%)173.00 (+0.00%)156.60 (-2.85%)165.00 (-2.37%)131.00 (-9.03%)17.36 (+42.64%)
0999a9a — 2025-11-13 19:24:080.08 (-3.47%)0.07 (-2.98%)0.07 (-10.06%)0.07 (-1.16%)0.01 (-2.33%)173.00 (+1.17%)161.20 (+3.07%)169.00 (+11.18%)144.00 (+3.60%)12.17 (+1.21%)
fbf1591 — 2025-11-13 03:39:220.08 (n/a)0.07 (n/a)0.08 (n/a)0.07 (n/a)0.01 (n/a)171.00 (n/a)156.40 (n/a)152.00 (n/a)139.00 (n/a)12.03 (n/a)

rms_norm_8_cols_2_channels_2048_tile_128

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.08 (+42.86%)0.07 (+45.77%)0.07 (+28.99%)0.06 (+83.50%)0.01 (-27.38%)200.00 (-45.50%)174.20 (-34.16%)169.00 (-22.48%)147.00 (-30.00%)18.08 (-71.58%)
003b49d — 2025-11-14 19:30:540.05 (-34.76%)0.05 (-39.42%)0.05 (-28.90%)0.03 (-56.40%)0.01 (+128.09%)367.00 (+129.38%)264.60 (+73.39%)218.00 (+40.65%)210.00 (+53.28%)63.64 (+693.95%)
0999a9a — 2025-11-13 19:24:080.08 (+11.68%)0.08 (+41.18%)0.07 (+32.90%)0.07 (+107.50%)0.00 (-70.36%)160.00 (-51.81%)152.60 (-34.17%)155.00 (-24.76%)137.00 (-10.46%)8.01 (-87.63%)
fbf1591 — 2025-11-13 03:39:220.07 (n/a)0.05 (n/a)0.06 (n/a)0.03 (n/a)0.01 (n/a)332.00 (n/a)231.80 (n/a)206.00 (n/a)153.00 (n/a)64.77 (n/a)

rope_1_cols_2_channels_4096_tile_4096_0

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.21 (+14.95%)0.16 (-9.76%)0.15 (-8.72%)0.11 (-31.86%)0.03 (+246.22%)204.00 (+46.76%)151.40 (+15.22%)149.00 (+9.56%)107.00 (-13.01%)30.94 (+345.41%)
003b49d — 2025-11-14 19:30:540.19 (-8.94%)0.17 (+9.40%)0.17 (+11.76%)0.16 (+20.14%)0.01 (-59.55%)139.00 (-16.77%)131.40 (-10.00%)136.00 (-10.53%)123.00 (+9.82%)6.95 (-62.13%)
0999a9a — 2025-11-13 19:24:080.20 (+24.11%)0.16 (+28.34%)0.15 (+15.13%)0.14 (+55.09%)0.02 (-16.27%)167.00 (-35.52%)146.00 (-24.66%)152.00 (-13.14%)112.00 (-19.42%)18.34 (-58.89%)
fbf1591 — 2025-11-13 03:39:220.16 (n/a)0.12 (n/a)0.13 (n/a)0.09 (n/a)0.03 (n/a)259.00 (n/a)193.80 (n/a)175.00 (n/a)139.00 (n/a)44.62 (n/a)

rope_2_cols_2_channels_4096_tile_2048_0

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.19 (+3.31%)0.16 (+7.87%)0.17 (+16.06%)0.13 (-5.29%)0.02 (+27.75%)170.00 (+5.59%)141.80 (-6.71%)137.00 (-13.84%)121.00 (-3.20%)17.99 (+31.79%)
003b49d — 2025-11-14 19:30:540.18 (+5.60%)0.15 (+0.61%)0.14 (-4.40%)0.14 (+11.80%)0.02 (+6.77%)161.00 (-10.56%)152.00 (-0.65%)159.00 (+4.61%)125.00 (-5.30%)13.65 (-11.43%)
0999a9a — 2025-11-13 19:24:080.17 (-12.88%)0.15 (-7.72%)0.15 (-1.97%)0.13 (-10.56%)0.01 (-32.96%)180.00 (+11.80%)153.00 (+7.59%)152.00 (+2.01%)132.00 (+14.78%)15.41 (-13.58%)
fbf1591 — 2025-11-13 03:39:220.20 (n/a)0.16 (n/a)0.15 (n/a)0.14 (n/a)0.02 (n/a)161.00 (n/a)142.20 (n/a)149.00 (n/a)115.00 (n/a)17.84 (n/a)

rope_4_cols_2_channels_4096_tile_1024_0

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.20 (+17.54%)0.16 (+8.90%)0.16 (+12.68%)0.14 (+4.85%)0.02 (+42.04%)165.00 (-4.62%)144.00 (-7.57%)142.00 (-11.25%)114.00 (-14.93%)18.03 (+14.57%)
003b49d — 2025-11-14 19:30:540.17 (-0.75%)0.15 (+2.89%)0.14 (-3.75%)0.13 (+16.18%)0.02 (-37.69%)173.00 (-13.93%)155.80 (-4.77%)160.00 (+3.90%)134.00 (+0.75%)15.74 (-45.89%)
0999a9a — 2025-11-13 19:24:080.17 (-7.52%)0.14 (-7.80%)0.15 (-11.04%)0.11 (+8.96%)0.02 (-10.02%)201.00 (-8.22%)163.60 (+7.49%)154.00 (+12.41%)133.00 (+8.13%)29.09 (-14.80%)
fbf1591 — 2025-11-13 03:39:220.19 (n/a)0.16 (n/a)0.17 (n/a)0.10 (n/a)0.03 (n/a)219.00 (n/a)152.20 (n/a)137.00 (n/a)123.00 (n/a)34.14 (n/a)

rope_8_cols_2_channels_4096_tile_512_0

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.18 (+11.29%)0.14 (+4.00%)0.15 (+15.79%)0.09 (-13.39%)0.03 (+73.25%)254.00 (+15.45%)177.00 (+0.68%)152.00 (-13.64%)124.00 (-10.14%)47.91 (+82.52%)
003b49d — 2025-11-14 19:30:540.17 (+15.22%)0.13 (+15.45%)0.13 (+15.91%)0.10 (+4.09%)0.02 (+24.38%)220.00 (-3.93%)175.80 (-12.97%)176.00 (-13.73%)138.00 (-13.21%)26.25 (+3.88%)
0999a9a — 2025-11-13 19:24:080.14 (+13.21%)0.12 (+2.72%)0.11 (+0.49%)0.10 (+3.49%)0.02 (+58.92%)229.00 (-3.38%)202.00 (-1.75%)204.00 (-0.49%)159.00 (-11.67%)25.27 (+34.93%)
fbf1591 — 2025-11-13 03:39:220.13 (n/a)0.11 (n/a)0.11 (n/a)0.10 (n/a)0.01 (n/a)237.00 (n/a)205.60 (n/a)205.00 (n/a)180.00 (n/a)18.73 (n/a)

silu_1_cols_1_channels_2048_tile_2048

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.07 (+8.85%)0.05 (-13.67%)0.05 (-14.58%)0.04 (-25.14%)0.01 (+367.73%)179.00 (+33.58%)149.60 (+19.30%)144.00 (+17.07%)113.00 (-8.13%)25.63 (+493.71%)
003b49d — 2025-11-14 19:30:540.06 (+11.38%)0.06 (+29.68%)0.06 (+34.96%)0.06 (+39.55%)0.00 (-62.09%)134.00 (-28.34%)125.40 (-23.72%)123.00 (-25.90%)123.00 (-10.22%)4.32 (-75.54%)
0999a9a — 2025-11-13 19:24:080.06 (+1.46%)0.05 (+0.63%)0.05 (-1.20%)0.04 (+11.76%)0.01 (-15.18%)187.00 (-10.53%)164.40 (-1.32%)166.00 (+1.22%)137.00 (-1.44%)17.65 (-26.23%)
fbf1591 — 2025-11-13 03:39:220.05 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.01 (n/a)209.00 (n/a)166.60 (n/a)164.00 (n/a)139.00 (n/a)23.92 (n/a)

silu_2_cols_1_channels_2048_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.07 (+4.63%)0.06 (-6.67%)0.06 (-8.70%)0.04 (-20.21%)0.01 (+71.90%)193.00 (+25.32%)147.60 (+12.16%)138.00 (+9.52%)108.00 (-4.42%)37.44 (+106.66%)
003b49d — 2025-11-14 19:30:540.07 (+4.42%)0.06 (+16.98%)0.06 (+23.81%)0.05 (+29.22%)0.01 (-6.38%)154.00 (-22.61%)131.60 (-15.32%)126.00 (-19.23%)113.00 (-4.24%)18.12 (-30.47%)
0999a9a — 2025-11-13 19:24:080.06 (+15.25%)0.05 (+3.17%)0.05 (-5.13%)0.04 (+15.08%)0.01 (+3.79%)199.00 (-13.10%)155.40 (-3.84%)156.00 (+5.41%)118.00 (-13.24%)26.06 (-24.10%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.05 (n/a)0.05 (n/a)0.03 (n/a)0.01 (n/a)229.00 (n/a)161.60 (n/a)148.00 (n/a)136.00 (n/a)34.33 (n/a)

silu_4_cols_1_channels_2048_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.06 (+10.22%)0.05 (+30.42%)0.05 (+53.15%)0.03 (+34.51%)0.01 (-9.38%)255.00 (-25.66%)169.60 (-26.07%)143.00 (-34.70%)137.00 (-9.27%)44.21 (-36.81%)
003b49d — 2025-11-14 19:30:540.05 (-13.91%)0.04 (-26.18%)0.03 (-35.16%)0.02 (-42.86%)0.01 (+22.03%)343.00 (+75.00%)229.40 (+43.38%)219.00 (+54.23%)151.00 (+16.15%)69.95 (+136.03%)
0999a9a — 2025-11-13 19:24:080.06 (+4.62%)0.05 (-3.58%)0.05 (+0.70%)0.04 (-12.24%)0.01 (+100.17%)196.00 (+13.95%)160.00 (+6.38%)142.00 (-0.70%)130.00 (-4.41%)29.64 (+124.27%)
fbf1591 — 2025-11-13 03:39:220.06 (n/a)0.05 (n/a)0.05 (n/a)0.04 (n/a)0.00 (n/a)172.00 (n/a)150.40 (n/a)143.00 (n/a)136.00 (n/a)13.22 (n/a)

silu_8_cols_1_channels_2048_tile_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.05 (+6.76%)0.04 (+11.09%)0.05 (+25.15%)0.03 (-3.52%)0.01 (+38.00%)284.00 (+3.65%)202.20 (-6.91%)163.00 (-20.10%)148.00 (-6.33%)58.05 (+30.37%)
003b49d — 2025-11-14 19:30:540.05 (-1.90%)0.04 (-7.67%)0.04 (-2.45%)0.03 (-21.53%)0.01 (+52.64%)274.00 (+27.44%)217.20 (+11.50%)204.00 (+2.51%)158.00 (+1.94%)44.53 (+107.23%)
0999a9a — 2025-11-13 19:24:080.05 (-7.10%)0.04 (-9.14%)0.04 (-15.58%)0.04 (+26.51%)0.00 (-40.67%)215.00 (-20.96%)194.80 (+6.33%)199.00 (+18.45%)155.00 (+7.64%)21.49 (-52.75%)
fbf1591 — 2025-11-13 03:39:220.05 (n/a)0.04 (n/a)0.05 (n/a)0.03 (n/a)0.01 (n/a)272.00 (n/a)183.20 (n/a)168.00 (n/a)144.00 (n/a)45.48 (n/a)

softmax_1_cols_2_channels_4096_tile_2048

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.20 (+18.75%)0.17 (+6.36%)0.18 (+12.40%)0.14 (-6.51%)0.02 (+150.19%)169.00 (+6.96%)137.80 (-4.31%)129.00 (-11.03%)112.00 (-15.79%)20.43 (+128.98%)
003b49d — 2025-11-14 19:30:540.17 (+12.03%)0.16 (+25.48%)0.16 (+28.28%)0.14 (+26.58%)0.01 (-30.51%)158.00 (-21.00%)144.00 (-20.88%)145.00 (-22.04%)133.00 (-10.74%)8.92 (-50.65%)
0999a9a — 2025-11-13 19:24:080.15 (-20.13%)0.13 (-18.52%)0.12 (-20.43%)0.11 (-15.50%)0.01 (-28.63%)200.00 (+18.34%)182.00 (+22.31%)186.00 (+25.68%)149.00 (+25.21%)18.08 (+6.20%)
fbf1591 — 2025-11-13 03:39:220.19 (n/a)0.16 (n/a)0.15 (n/a)0.14 (n/a)0.02 (n/a)169.00 (n/a)148.80 (n/a)148.00 (n/a)119.00 (n/a)17.02 (n/a)

softmax_2_cols_2_channels_4096_tile_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.19 (+2.44%)0.16 (+5.32%)0.17 (+14.07%)0.14 (+0.00%)0.02 (+19.18%)168.00 (+0.00%)142.40 (-4.69%)135.00 (-12.34%)123.00 (-2.38%)16.58 (+18.75%)
003b49d — 2025-11-14 19:30:540.18 (-3.97%)0.15 (+9.75%)0.15 (+9.09%)0.14 (+32.14%)0.02 (-44.61%)168.00 (-24.32%)149.40 (-11.39%)154.00 (-8.33%)126.00 (+4.13%)13.97 (-56.54%)
0999a9a — 2025-11-13 19:24:080.19 (+2.48%)0.14 (-11.63%)0.14 (-19.05%)0.10 (-23.42%)0.03 (+48.41%)222.00 (+30.59%)168.60 (+15.80%)168.00 (+23.53%)121.00 (-2.42%)32.13 (+84.60%)
fbf1591 — 2025-11-13 03:39:220.18 (n/a)0.16 (n/a)0.17 (n/a)0.13 (n/a)0.02 (n/a)170.00 (n/a)145.60 (n/a)136.00 (n/a)124.00 (n/a)17.41 (n/a)

softmax_2_cols_2_channels_4096_tile_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.20 (+2.59%)0.18 (+6.57%)0.19 (+5.93%)0.14 (+38.41%)0.02 (-30.68%)164.00 (-27.75%)131.40 (-9.75%)118.00 (-5.60%)116.00 (-2.52%)18.97 (-53.67%)
003b49d — 2025-11-14 19:30:540.19 (+28.57%)0.17 (+31.45%)0.18 (+44.00%)0.10 (+9.25%)0.03 (+53.94%)227.00 (-8.47%)145.60 (-21.97%)125.00 (-30.56%)119.00 (-22.22%)40.95 (+16.05%)
0999a9a — 2025-11-13 19:24:080.15 (-1.96%)0.13 (-2.94%)0.13 (-5.55%)0.09 (+10.89%)0.02 (-13.74%)248.00 (-9.82%)186.60 (+1.30%)180.00 (+5.88%)153.00 (+2.00%)35.29 (-24.09%)
fbf1591 — 2025-11-13 03:39:220.15 (n/a)0.13 (n/a)0.13 (n/a)0.08 (n/a)0.03 (n/a)275.00 (n/a)184.20 (n/a)170.00 (n/a)150.00 (n/a)46.48 (n/a)

swiglu

No metrics available.

transpose_2048_M_64_N_1_cols_1_channels_64_m_64_n_8_s

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:573.23 (-7.95%)2.83 (-4.91%)3.15 (+0.00%)2.08 (-14.47%)0.46 (+2.40%)235.00 (+16.92%)177.80 (+5.83%)155.00 (+0.00%)151.00 (+8.63%)32.68 (+24.58%)
003b49d — 2025-11-14 19:30:543.51 (-2.16%)2.98 (+1.47%)3.15 (+11.61%)2.43 (-9.45%)0.45 (+33.26%)201.00 (+10.44%)168.00 (-0.24%)155.00 (-10.40%)139.00 (+2.21%)26.23 (+56.91%)
0999a9a — 2025-11-13 19:24:083.59 (+20.59%)2.93 (+26.88%)2.82 (+27.17%)2.68 (+52.75%)0.34 (-23.17%)182.00 (-34.53%)168.40 (-23.03%)173.00 (-21.36%)136.00 (-17.07%)16.72 (-58.74%)
fbf1591 — 2025-11-13 03:39:222.98 (n/a)2.31 (n/a)2.22 (n/a)1.76 (n/a)0.44 (n/a)278.00 (n/a)218.80 (n/a)220.00 (n/a)164.00 (n/a)40.51 (n/a)

transpose_2048_M_64_N_1_cols_2_channels_64_m_64_n_8_s

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:573.37 (-1.38%)2.94 (+7.22%)3.00 (+12.27%)2.52 (+39.17%)0.31 (-46.07%)194.00 (-28.15%)168.20 (-10.34%)163.00 (-10.93%)145.00 (+1.40%)17.88 (-60.47%)
003b49d — 2025-11-14 19:30:543.41 (-11.19%)2.74 (+9.89%)2.67 (+28.96%)1.81 (+18.52%)0.57 (-39.76%)270.00 (-15.62%)187.60 (-16.77%)183.00 (-22.46%)143.00 (+12.60%)45.24 (-42.49%)
0999a9a — 2025-11-13 19:24:083.84 (-6.30%)2.49 (-22.39%)2.07 (-30.93%)1.53 (-38.75%)0.95 (+70.33%)320.00 (+63.27%)225.40 (+43.93%)236.00 (+44.79%)127.00 (+6.72%)78.66 (+198.96%)
fbf1591 — 2025-11-13 03:39:224.10 (n/a)3.21 (n/a)3.00 (n/a)2.49 (n/a)0.56 (n/a)196.00 (n/a)156.60 (n/a)163.00 (n/a)119.00 (n/a)26.31 (n/a)

weighted_rms_norm_1_cols_2_channels_2048_weights_2048

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.09 (+0.79%)0.07 (-3.64%)0.07 (+0.00%)0.05 (-25.32%)0.01 (+53.35%)233.00 (+33.91%)167.60 (+6.62%)158.00 (+0.00%)126.00 (-0.79%)35.33 (+114.32%)
003b49d — 2025-11-14 19:30:540.09 (-7.87%)0.07 (-4.96%)0.07 (-0.63%)0.07 (-0.57%)0.01 (-20.60%)174.00 (+0.58%)157.20 (+4.66%)158.00 (+0.64%)127.00 (+8.55%)16.49 (-12.45%)
0999a9a — 2025-11-13 19:24:080.10 (-3.42%)0.08 (-10.60%)0.07 (-12.10%)0.07 (-13.87%)0.01 (+25.36%)173.00 (+16.11%)150.20 (+12.76%)157.00 (+13.77%)117.00 (+3.54%)18.83 (+48.05%)
fbf1591 — 2025-11-13 03:39:220.10 (n/a)0.09 (n/a)0.08 (n/a)0.08 (n/a)0.01 (n/a)149.00 (n/a)133.20 (n/a)138.00 (n/a)113.00 (n/a)12.72 (n/a)

weighted_rms_norm_2_cols_2_channels_2048_weights_1024

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.11 (+16.98%)0.09 (+17.73%)0.10 (+33.33%)0.07 (+5.77%)0.01 (+78.97%)156.00 (-5.45%)128.60 (-13.58%)117.00 (-25.00%)106.00 (-14.52%)21.85 (+49.03%)
003b49d — 2025-11-14 19:30:540.09 (-12.10%)0.08 (-2.40%)0.07 (-4.49%)0.07 (+54.55%)0.01 (-62.01%)165.00 (-35.29%)148.80 (-5.58%)156.00 (+4.70%)124.00 (+13.76%)14.66 (-72.13%)
0999a9a — 2025-11-13 19:24:080.10 (+12.84%)0.08 (-5.42%)0.08 (-10.07%)0.04 (-41.96%)0.02 (+265.18%)255.00 (+72.30%)157.60 (+15.37%)149.00 (+11.19%)109.00 (-11.38%)52.60 (+447.71%)
fbf1591 — 2025-11-13 03:39:220.09 (n/a)0.08 (n/a)0.09 (n/a)0.08 (n/a)0.01 (n/a)148.00 (n/a)136.60 (n/a)134.00 (n/a)123.00 (n/a)9.60 (n/a)

weighted_rms_norm_4_cols_2_channels_2048_weights_512

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.10 (+15.04%)0.08 (+6.02%)0.07 (-3.80%)0.07 (+16.37%)0.01 (+22.80%)171.00 (-14.07%)151.00 (-5.51%)158.00 (+3.95%)113.00 (-13.08%)20.33 (-11.68%)
003b49d — 2025-11-14 19:30:540.09 (+20.77%)0.07 (+19.39%)0.08 (+29.61%)0.06 (+7.04%)0.01 (+42.62%)199.00 (-6.57%)159.80 (-15.63%)152.00 (-22.84%)130.00 (-17.20%)23.02 (+11.90%)
0999a9a — 2025-11-13 19:24:080.07 (-22.29%)0.06 (-29.29%)0.06 (-30.96%)0.05 (-34.27%)0.01 (+58.69%)213.00 (+52.14%)189.40 (+42.84%)197.00 (+44.85%)157.00 (+28.69%)20.57 (+210.64%)
fbf1591 — 2025-11-13 03:39:220.09 (n/a)0.09 (n/a)0.08 (n/a)0.08 (n/a)0.00 (n/a)140.00 (n/a)132.60 (n/a)136.00 (n/a)122.00 (n/a)6.62 (n/a)

weighted_rms_norm_8_cols_2_channels_2048_weights_256

Commit/Date Bandwidth (max)Bandwidth (mean)Bandwidth (median)Bandwidth (min)Bandwidth (stddev)Latency (max)Latency (mean)Latency (median)Latency (min)Latency (stddev)
8404bd4 — 2025-11-15 00:03:570.08 (+2.04%)0.07 (-5.61%)0.07 (+6.45%)0.05 (-21.55%)0.01 (+132.09%)232.00 (+27.47%)179.00 (+9.01%)155.00 (-6.06%)147.00 (-2.00%)34.11 (+190.21%)
003b49d — 2025-11-14 19:30:540.08 (-14.00%)0.07 (-4.06%)0.07 (-4.85%)0.06 (+5.49%)0.00 (-49.70%)182.00 (-5.21%)164.20 (+2.88%)165.00 (+5.10%)150.00 (+16.28%)11.75 (-44.97%)
0999a9a — 2025-11-13 19:24:080.09 (+24.03%)0.07 (+20.14%)0.07 (+19.75%)0.06 (+28.65%)0.01 (+16.90%)192.00 (-22.27%)159.60 (-17.05%)157.00 (-16.49%)129.00 (-19.38%)21.36 (-28.23%)
fbf1591 — 2025-11-13 03:39:220.07 (n/a)0.06 (n/a)0.06 (n/a)0.05 (n/a)0.01 (n/a)247.00 (n/a)192.40 (n/a)188.00 (n/a)160.00 (n/a)29.76 (n/a)

@andrej andrej merged commit bdca01f into devel Nov 15, 2025
4 checks passed
@andrej andrej deleted the bring-in-the-llama branch November 15, 2025 00:10
@kurtis-b-1 kurtis-b-1 mentioned this pull request Nov 17, 2025
3 tasks
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.

4 participants