ywkang/kernbench2
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

gqa(decode-4cases): Case 3 — Cube-Repl × PE-SP (intra-CUBE AR only; 8× memory) (5C.C)

Browse Source 
Adds the third 4-cases panel: KV replicated per cube (8× memory
waste), PEs SP on S_kv within each cube, intra-CUBE 8-way reduce on
(m, ℓ, O), and no inter-CUBE comm (every cube ends with full answer;
designated writer = cube 0). Reuses the row-chain + col-bridge
intra-CUBE pattern that anchors Case 4 (21 ipcq_copy per cube × 8
cubes = 168 total). 12 tests pass (4 Case 4 + 4 Case 2 + 4 Case 3).

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This commit is contained in:
Mukesh Garg
2026-06-15 15:18:41 -07:00
parent 5672c8f3ef
commit ae942f6959
3 changed files with 302 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/kernbench/benches/milestone_gqa_decode_4cases.py
+48 -1
View File
@@ -32,6 +32,9 @@ import json
import os
from pathlib import Path
from kernbench.benches._gqa_attention_decode_cube_repl_pe_sp import (
gqa_attention_decode_cube_repl_pe_sp_kernel,
)
from kernbench.benches._gqa_attention_decode_cube_repl_pe_tp import (
gqa_attention_decode_cube_repl_pe_tp_kernel,
)
@@ -58,7 +61,8 @@ _SWEEP_JSON = _OUTPUT_DIR / "sweep.json"
_PANELS = (
"single_kv_group_decode_gqa_cube_sp_pe_sp", # Case 4 ★ optimal
"single_kv_group_decode_gqa_cube_repl_pe_tp", # Case 2 (no comm; 8× memory)
# Cases 1, 3 to be added by 5C.A/C
"single_kv_group_decode_gqa_cube_repl_pe_sp", # Case 3 (intra-CUBE AR only; 8× memory)
# Case 1 to be added by 5C.A
)
# Each entry: (kind, panel-specific params).
@@ -82,6 +86,14 @@ _PANEL_DISPATCH: dict[str, tuple[str, dict]] = {
"T_q": 1, "S_kv": 131_072,
"d_head": 128, "h_q": 8, "h_kv": 1,
}),
"single_kv_group_decode_gqa_cube_repl_pe_sp": ("decode_cube_repl_pe_sp", {
# Case 3: K, V replicated per cube (8× memory); PEs SP on S_kv
# within each cube. Intra-CUBE 8-way reduce; no inter-CUBE comm
# (every cube ends with full answer; designated writer = cube 0).
"C": 8, "P": 8,
"T_q": 1, "S_kv": 131_072,
"d_head": 128, "h_q": 8, "h_kv": 1,
}),
}
@@ -119,6 +131,39 @@ def _run_decode_panel_cube_repl_pe_tp(
)
def _run_decode_panel_cube_repl_pe_sp(
ctx, *, panel: str, C: int, P: int,
T_q: int, S_kv: int,
d_head: int, h_q: int, h_kv: int,
) -> None:
"""Case 3 runner: K, V replicated per cube; PEs SP on S_kv within cube.
DPPolicy models the cluster-wide 8× memory waste — every cube
holds full K, V in its HBM region, then splits the S_kv axis
row_wise across its 8 PEs. The kernel does an intra-CUBE 8-way
reduce on (m, ℓ, O); only cube 0's PE 0 writes the output.
"""
configure_sfr_intercube_multisip(ctx.engine, ctx.spec, _ccl_cfg())
dp_full = DPPolicy(cube="replicate", pe="replicate",
num_cubes=C, num_pes=P)
dp_kv = DPPolicy(cube="replicate", pe="row_wise",
num_cubes=C, num_pes=P)
q = ctx.zeros((T_q, h_q * d_head),
dtype="f16", dp=dp_full, name=f"{panel}_q")
k = ctx.zeros((S_kv, h_kv * d_head),
dtype="f16", dp=dp_kv, name=f"{panel}_k")
v = ctx.zeros((S_kv, h_kv * d_head),
dtype="f16", dp=dp_kv, name=f"{panel}_v")
o = ctx.empty((T_q, h_q * d_head),
dtype="f16", dp=dp_full, name=f"{panel}_o")
ctx.launch(
panel, gqa_attention_decode_cube_repl_pe_sp_kernel,
q, k, v, o,
T_q, S_kv, h_q, h_kv, d_head, C, P,
_auto_dim_remap=False,
)
def _make_bench_fn(panel: str):
kind, params = _PANEL_DISPATCH[panel]
@@ -127,6 +172,8 @@ def _make_bench_fn(panel: str):
_run_decode_panel(ctx, panel=panel, **params)
elif kind == "decode_cube_repl_pe_tp":
_run_decode_panel_cube_repl_pe_tp(ctx, panel=panel, **params)
elif kind == "decode_cube_repl_pe_sp":
_run_decode_panel_cube_repl_pe_sp(ctx, panel=panel, **params)
else:
raise RuntimeError(
f"milestone-gqa-decode-4cases panel {panel!r} has "