ywkang

0 Followers · 0 Following

• Joined on 2026-03-04

Repositories

1

Projects

1

Packages Public Activity Starred Repositories

ywkang pushed to master at ywkang/kernbench2 2026-06-11 05:56:27 +00:00

3d4a3d43c4 gqa(adr-0065): N2 — softmax_merge recipe MATH ops compute in data mode

ywkang pushed to master at ywkang/kernbench2 2026-06-11 05:46:40 +00:00

4089e18770 gqa(adr-0065): N1 — composite GEMM computes numerically in data mode

ywkang pushed to master at ywkang/kernbench2 2026-06-11 05:29:13 +00:00

e8d6c283d8 gqa(adr-0065): D8 — composite returns output handle + output-space DMA

ywkang pushed to master at ywkang/kernbench2 2026-06-11 05:15:15 +00:00

dd525bfcb7 paper: add /paper skill + 1H HW-SW codesign report (GEMM, All-Reduce, fused GQA)

ywkang pushed to master at ywkang/kernbench2 2026-06-11 04:04:46 +00:00

4a55ae5c0b gqa(adr-0065): P6 — close dispatch-ratio loop (ADR-0064 Test #9 direction)

ywkang pushed to master at ywkang/kernbench2 2026-06-11 04:02:29 +00:00

35453cc4fe gqa(adr-0065): P5(B) — decode opt2 two-composite kernel + dispatch measurement

ywkang pushed to master at ywkang/kernbench2 2026-06-11 03:36:31 +00:00

17fb94086e gqa(adr-0065): P4 — strict-FIFO RW hazard tracker

ywkang pushed to master at ywkang/kernbench2 2026-06-11 03:24:36 +00:00

55f025c4b1 gqa(adr-0065): P3 — flat-ops PE_SCHEDULER plan (position-scan + prologue stages)

ywkang pushed to master at ywkang/kernbench2 2026-06-11 02:48:30 +00:00

184f654295 gqa(adr-0065): P2 — softmax_merge recipe + TLContext prologue lowering

ywkang pushed to master at ywkang/kernbench2 2026-06-11 02:18:12 +00:00

47e2c78c66 gqa(adr-0064/0065): flat-ops CompositeCmd (P1) + structural dispatch cost (ADR-0064 Rev2); promote ADR-0064

ywkang pushed to master at ywkang/kernbench2 2026-06-10 23:34:36 +00:00

3647e2d8f5 gqa(adr): ADR-0064/0065 second-pass review — 1024B rationale, FIFO as ordering source, robust formula-based tests, D5/D6 auto-bind alignment

ywkang pushed to master at ywkang/kernbench2 2026-06-10 23:29:41 +00:00

a8d04750e6 gqa(adr): ADR-0065 review fixes — KERNEL phase/repetition split, auto-bind conflict + MATH-TCM invariants, FIFO conservatism note, logical_bytes per-op summation

ywkang pushed to master at ywkang/kernbench2 2026-06-10 23:22:24 +00:00

95cecccd8e gqa(adr): ADR-0064/0065 review fixes — composite size cap, type-aware logical_bytes, R recalibration

ywkang pushed to master at ywkang/kernbench2 2026-06-10 23:11:27 +00:00

a8c50238c6 gqa(adr): ADR-0064 Rev2 (structural dispatch cost) + ADR-0065/DDD-0065 (flat-ops composite + softmax_merge recipe)

ywkang pushed to master at ywkang/kernbench2 2026-06-05 07:52:40 +00:00

ab324c155c gqa(adr): review fixes — unify KV placement to contiguous, ring/reduce cost model, opt3 wording, rank def

ywkang pushed to master at ywkang/kernbench2 2026-06-05 06:09:14 +00:00

9103661098 gqa(adr): mark DDD-0060 as synced in ADR-0060 SB (EN+KO)

ywkang pushed to master at ywkang/kernbench2 2026-06-05 06:08:03 +00:00

c699e2ddfc gqa(ddd): rewrite DDD-0060 to match current ADR (two-kernel composite hybrid)

ywkang pushed to master at ywkang/kernbench2 2026-06-05 06:02:50 +00:00

973a452b77 gqa(adr): sync ADR-0060 KO mirror to current EN (two-kernel design)

ywkang pushed to master at ywkang/kernbench2 2026-06-05 05:50:00 +00:00

8de13efb9d gqa(adr): opt2 ex_composite decode needs tl.wait() before the reduction

ywkang pushed to master at ywkang/kernbench2 2026-06-05 05:40:13 +00:00

9c554e3f64 gqa(adr): fix prefill TL;DR pseudocode — K/V are TCM handles for the ring