Files
kernbench2/docs/sweeps/short_context_decode_sweep.csv
ywkang 711a9a257f gqa short-context: exclude KV deploy from wall; correct HBM peak; batch scaffold
Metric fixes (test harnesses, deploy artifact + wrong constant):
- wall = max(t_end) - min(t_start): exclude the one-time KV-cache deploy
  from the measured decode/prefill step (it was 92-99% of wall, mapping-
  invariant, and masked the real per-mapping separation).
- PEAK_PE_HBM_BPS 128 -> 256 B/ns (8 channels/PE x 32 GB/s); the old value
  was half the modeled per-PE HBM BW, so hbm_bw_util read >1.0 once wall
  was corrected. All six short-context sweep CSVs regenerated/repatched.

Result: the four KV mappings now separate along a {64,32,16,8}-active-PE
ladder (decode 8-kv/1-kv = 4.8x at 8K, 7.4x at 64K), not "modest/tied" as
before; decode is bandwidth-bound at 46-76% of the 256 GB/s per-PE ceiling.

Report (S5.2 rewrite):
- Replace the tied-wall / 8x-per-PE-util claims (both deploy artifacts)
  with the corrected separation and a density trade-off (per-CUBE KV
  footprint, Fig 16 top panel switched to a wall-invariant metric).
- Add a projected latency-vs-batched-throughput analysis (marked
  projected, not measured): dense mappings win throughput, 1-kv wins
  latency; converges at long context.
- Regenerate Fig 15/16/17/18; fix plot script hardcoded ROOT path.

Batch experiment (Part 2, cube_base):
- Add backward-compatible cube_base=0 scalar to the decode kernel so a
  batched user placed at DPPolicy.cube_start addresses 0-based shards.
  Default preserves single-user behavior (32 decode tests pass unchanged).
- New batch harness (skipped): concurrent B>=2 launches hit a sim routing
  issue (sip0.cube0.pe8); single-user path verified. Concurrency fix next.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-22 16:25:04 -07:00

1.6 KiB

1modekv_per_cubeCS_kvwall_usn_pegemm_utilmath_counthbm_bw_utilhbm_read_mbhbm_write_kbipcq_kbkv_cache_per_cube_mb
2A11881922.2471645000005376640.116655456242704286560.45768816.06258.057.752.0
3A118163843.6068965000013704640.1453570957746481213600.56904932.06258.057.754.0
4A118327686.326360500001814640.165747114790573427680.64816164.06258.057.758.0
5A1186553611.765288499959512640.1782490926598234255840.696668128.06258.057.7516.0
6A22481923.3555127500005475320.156246761392832366560.61182916.031258.024.754.0
7A224163846.074976750001311320.1726057634705611413600.67506432.031258.024.758.0
8A2243276811.513904750002315320.1821408154345100527680.71192264.031258.024.7516.0
9A2246553622.391760749934242320.1873146130328314255840.731921128.031258.024.7532.0
10A44281925.586135500000557160.18771044848442726560.73431816.0156258.08.258.0
11A4421638411.025063500001211160.1902167729010847413600.74357832.0156258.08.2516.0
12A4423276821.902919500003218160.1914952022720434227680.74830264.0156258.08.2532.0
13A4426553643.65863149988279160.1921408828404956355840.750688128.0156258.08.2564.0
14B81819210.81609100000094780.19389185982251686560.7581316.00781258.00.016.0
15B811638421.69394700000231480.193339828847147613600.75560232.00781258.00.032.0
16B813276843.44965900000638580.1930649904525503527680.75434464.00781258.00.064.0
17B816553686.9610829997691380.1929278640657596555840.753716128.00781258.00.0128.0