# MemVanta trained-model layer profile
prefill_ms=46169.3 decode_total_ms=8907.17 sampling_ms=0.364908
prefill_cycles_per_token=882053442.88 decode_cycles_per_token=1361352603.50
qkv_projection_ms=12651.28 attention_output_projection_ms=3887.31 ffn_gemm_ms=34086.64 output_head_ms=3225.07
prefill_qkv_projection_ms=11232.52 prefill_attention_output_projection_ms=3411.80 prefill_ffn_gemm_ms=30274.30 prefill_output_head_ms=375.66 prefill_kernel_ms=45294.28
decode_qkv_projection_ms=1418.76 decode_attention_output_projection_ms=475.50 decode_ffn_gemm_ms=3812.34 decode_output_head_ms=2849.41 decode_kernel_ms=8556.01
non_gemm_core_ms=1225.86 (attention softmax/KV + RMSNorm + RoPE + elementwise/residual)
layer,qkv_ms,o_proj_ms,ffn_ms,total_projection_ms,pct_kernel_ms,projection_ms_per_token
0,396.32,120.67,1065.56,1582.55,2.94,10.99
1,398.22,122.30,1082.30,1602.82,2.98,11.13
2,394.48,120.46,1059.48,1574.43,2.92,10.93
3,393.84,120.49,1076.49,1590.83,2.95,11.05
4,394.17,121.67,1056.73,1572.57,2.92,10.92
5,393.62,121.68,1063.85,1579.14,2.93,10.97
6,393.38,122.40,1068.01,1583.79,2.94,11.00
7,393.87,123.11,1072.00,1588.98,2.95,11.03
8,397.17,120.42,1070.28,1587.87,2.95,11.03
9,395.96,121.74,1059.27,1576.97,2.93,10.95
10,397.60,122.93,1046.99,1567.52,2.91,10.89
11,391.86,121.68,1066.33,1579.87,2.93,10.97
12,394.18,123.58,1059.46,1577.23,2.93,10.95
13,396.50,121.15,1063.60,1581.25,2.94,10.98
14,391.89,120.99,1068.42,1581.30,2.94,10.98
15,396.78,121.66,1079.29,1597.73,2.97,11.10
16,396.25,121.49,1078.37,1596.11,2.96,11.08
17,394.94,121.95,1054.64,1571.53,2.92,10.91
18,395.45,122.26,1055.78,1573.50,2.92,10.93
19,392.28,120.32,1060.39,1572.99,2.92,10.92
20,395.46,120.21,1074.03,1589.70,2.95,11.04
21,394.07,121.76,1052.54,1568.36,2.91,10.89
22,395.36,122.92,1070.46,1588.74,2.95,11.03
23,391.14,120.29,1057.50,1568.93,2.91,10.90
24,391.70,120.01,1051.80,1563.51,2.90,10.86
25,392.20,120.94,1057.25,1570.40,2.92,10.91
26,390.97,120.31,1052.55,1563.82,2.90,10.86
27,394.25,119.95,1054.65,1568.86,2.91,10.89
28,392.74,122.98,1073.44,1589.15,2.95,11.04
29,396.00,121.59,1080.14,1597.73,2.97,11.10
30,407.93,121.45,1089.07,1618.46,3.01,11.24
31,410.67,121.94,1065.97,1598.58,2.97,11.10
profile_csv=sevenb-ffn-kernel-ab/optimized-2.csv
