• Joined on 2026-04-13
mukesh pushed to master at ywkang/kernbench2 2026-07-30 22:58:05 +00:00
fc4747668e paper(§5): add roofline, capacity planning, parallelism selection subsections
bfe3e0e8d1 analytical-viz: parallelism rules — Section 11 TP vs CP shared-budget
55653cbb3e analytical-viz: new 'Parallelism rules' tab
Compare 3 commits »
mukesh pushed to master at ywkang/kernbench2 2026-07-30 15:37:45 +00:00
5ce7a5b30b analytical-viz: capacity planning — symbol legend + plain-English formulas
bc5e704572 analytical-viz: Physical Layout — TTFT & TPOT full-model latency section
c39bbb16aa analytical-viz: capacity planning tab — add public-example B/replica table
77abc95d78 analytical-viz: capacity planning tab — add rules of thumb + deployment templates
f54822d247 analytical-viz: move capacity planning to its own tab, table-first layout
Compare 36 commits »
mukesh pushed to master at ywkang/kernbench2 2026-07-28 22:58:02 +00:00
c2b42999d8 analytical-viz: reorder tabs — Physical layout back to first
9afeb6bc16 analytical-viz: 3 sidebar apply-buttons (Attn / FFN / Attn+FFN)
bccc90db82 analytical-viz: Physical Layout scope filter for stage tables
1a6d52c58a analytical-viz: force-reload our modules on Streamlit rerun
09721040ca analytical-viz: 3 sweep buttons at top of Physical Layout tab
Compare 12 commits »
mukesh pushed to master at ywkang/kernbench2 2026-07-28 18:35:45 +00:00
09e9331026 topology: set flit_bytes=4096 (was implicit 256 default)
mukesh pushed to master at ywkang/kernbench2 2026-07-28 06:40:21 +00:00
f9d0077472 router: memoize Dijkstra path lookups
mukesh pushed to master at ywkang/kernbench2 2026-07-28 05:06:48 +00:00
65f358fdfa sim: report correct kernel latency under enable_data=False
mukesh pushed to master at ywkang/kernbench2 2026-07-27 05:12:56 +00:00
9fdde44922 analytical-viz: interactive Streamlit tool for SIP transformer analysis
mukesh pushed to master at ywkang/kernbench2 2026-07-22 16:54:48 +00:00
d0db9ff40e gqa-decode: topology-agnostic reduce + multi-model attention bench
mukesh pushed to master at ywkang/kernbench2 2026-07-14 04:50:59 +00:00
23a2bc3fb3 gqa-decode-composite: measure primitive hand-tiled (16×16×16) latency
mukesh pushed to master at ywkang/kernbench2 2026-07-13 17:18:39 +00:00
eec61838b5 sweep+plot: add primitive hand-tiled (16×16×16) variant to Case-6 decode composite study
mukesh pushed to master at ywkang/kernbench2 2026-06-18 22:38:25 +00:00
bb668a3ac3 paper(gqa): derive (m,l,O) AR per-PE cost from kernel topology
mukesh pushed to master at ywkang/kernbench2 2026-06-18 22:01:13 +00:00
0662c76fa7 paper(ipcq): de-emphasize flow-diagram outer frame so inner flow stands out
mukesh pushed to master at ywkang/kernbench2 2026-06-18 21:56:12 +00:00
37dc48fd27 paper(ipcq): add latency-model-style flow + stacked architecture diagrams
mukesh pushed to master at ywkang/kernbench2 2026-06-18 21:06:36 +00:00
7c45047e93 paper(ipcq): swap control-data overlay for latency PNG in figures/
mukesh pushed to master at ywkang/kernbench2 2026-06-18 21:02:46 +00:00
73e0b315fe paper(gqa): table cleanup — drop measured-comm column, narrower fig, wrapped title
mukesh pushed to master at ywkang/kernbench2 2026-06-18 20:27:56 +00:00
1552028c25 paper(ipcq): add IPCQ vs HW alternatives comparative study
mukesh pushed to master at ywkang/kernbench2 2026-06-18 18:51:33 +00:00
56c51b184a paper(gqa): refresh 4 decode 6-case PNGs in figures/ with actual 6-case content
mukesh pushed to master at ywkang/kernbench2 2026-06-18 18:49:16 +00:00
8102ddbe30 paper(gqa): rename long-ctx artifacts to gqa_long_ctx_6cases_* and add 4 decode 6-case charts
mukesh pushed to master at ywkang/kernbench2 2026-06-18 18:27:17 +00:00
84bb418e1e gqa(decode): add d_head-TP kernels + measurement runner + figure generators
dd3337f2e4 paper(gqa): add 6-case KV-sharding figures (analytical + measured)
Compare 2 commits »
mukesh pushed to master at ywkang/kernbench2 2026-06-17 16:39:46 +00:00
92b9221533 paper(gqa): drop committed long-ctx 4-cases derived chart PNGs