Files
kernbench2/docs/sweeps/short_context_prefill_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
2A118819234.765078999497696640.0603235217739369652480.02968516.062564.0114688.02.0
3A1181638462.2566439945763640.06737118692698127108800.033024632.062564.0229376.04.0
4A11832768117.23649402130023640.07155287327558905221440.035006264.062564.0458752.08.0
5A11865536227.190164006345640.07384657726472504446720.0360931128.062564.0917504.016.0
6A224819234.98345650027832320.1198939275759681326240.058999316.062564.049152.04.0
7A2241638465.32612049105158320.1284112379081875654400.062945732.062564.098304.08.0
8A22432768126.01144852332584320.13314041062637966110720.065136964.062564.0196608.016.0
9A22465536247.3821043861434320.1356380732680083223360.0662942128.062564.0393216.032.0
10A442819246.571509501108665160.1801231716528187713120.088637916.062564.016384.08.0
11A4421638488.98265048367577160.1885447995627355727200.092422532.062564.032768.016.0
12A44232768173.195009542468160.1937378685946129555360.094783364.062564.065536.032.0
13A44265536341.9424492809754160.19625777419912674111680.0959226128.062564.0131072.064.0
14B81819268.5016219999948580.244917061963867266560.12052316.062564.00.016.0
15B8116384134.0376219999920780.2503359243421842713600.12271232.062564.00.032.0
16B8132768265.109621999998680.2531362818661586627680.12384364.062564.00.064.0
17B8165536527.253622000029380.254560087213432555840.124418128.062564.00.0128.0