admin/matmul-analysis
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix issues #23-#25 (+#26 标注): GM读写共享总线累加计时 / ASW-切M/N 共享块 GM首读1次+L2重复读(n-1)次、场景按单batch判定、分组预算不再除B / StreamK 按plan实际tile芯片口径评估 / 降核线性带宽假设文档标注

Browse Source
This commit is contained in:
admin
2026-09-04 11:43:23 +08:00
parent f4b23d9f05
commit 4843053ad3
10 changed files with 248 additions and 180 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
BMM/BMM_Theory/README.md
View File

@@ -49,10 +49,11 @@ python -m unittest discover -s tests -v
### 时延模型要点 (读结果前必看)
- **搬入 MTE2 分两段**: `t_mte2_gm` (GM→L1 直读, 按 GM 带宽 1.6TB/s; 无论是否随路驻留 L2 都不再累加 L2→L1) 与 `t_mte2_l2` (数据驻留 L2 后重复读命中, 按 L2 带宽 5.2TB/s);
- **Cube 计算** `t_mmad`: 单核算力 ≈15.2 TFLOPS, MergeBatch 的冗余计算计入;
- **Fixpipe 搬出** `t_fixpipe`: 数据量按 **C 矩阵 dtype** 计 (fp16/fp8 时随路转换、写出量减半); StreamK 中间部分和为防精度丢失按 4B (L0C dtype) 计;
- **总时延** `t_total = max(各级) + t_drain` (双缓冲稳态取最大 + 末级排空暴露);
- **GM 是读写共享总线 (1.6TB/s)**: MTE2 的 GM 读与 Fixpipe 直写 GM 并发时按 **(读+写)/1.6TB/s 累加**计入 MTE2 搬移链 (issue#23);
- **搬入 MTE2 分两段、同链相加**: `t_mte2_gm` (首读按 GM 带宽; 随路驻留 L2 不重复累加) + `t_mte2_l2` (共享块驻留 L2 后被其它核重复读, 按 L2 读口 5.2TB/s) + DMA 命令开销 (issue#24);
- **Cube 计算** `t_mmad`: 芯片算力 ≈486 TFLOPS (单核 ≈15.2 TFLOPS), MergeBatch 冗余计算计入;
- **Fixpipe 搬出**: 直写 GM 计入 GM 共享总线; 驻留 L2 走 5.2TB/s 写口 (独立计时); 数据量按 **C 矩阵 dtype** 计 (fp16/fp8 随路转换减半); StreamK 部分和按 4B (L0C dtype);
- **总时延** `t_total = max(MTE2搬移链, MMAD, Fixpipe-L2) + t_drain` (稳态取最大 + 末级排空暴露; REDUCE 串行追加);
- **瓶颈交换**: 搬移瓶颈可牺牲算力换搬移效率 (MergeBatch), 计算瓶颈可牺牲搬移换计算效率 (ASW_Basic 切 M/N).
## 目录结构

94
BMM/BMM_Theory/bmm_theory/branches/asw_basic.py
View File

@@ -80,21 +80,22 @@ class AswBasicBranch(Branch):
# Step 4: swizzle 窗口 W = max{d | d|C, d <= floor(sqrt(C))}
swizzle_w = self._swizzle_w()
# Step 5: L2 分组判断
s_in = b * (m * k + k * n) * dt
s_out = b * m * n * case.dtype_out_bytes
if s_in + s_out <= s.l2_bytes:
l2_scene = "A_全驻留"
# Step 5: L2 分组判断 (issue#24: 按单 batch 工作集判定, 同一时刻单 batch 激活)
a_b = m * k * dt
bb_b = k * n * dt
out_batch = m * n * case.dtype_out_bytes
if a_b + bb_b + out_batch <= s.l2_bytes:
l2_scene = "A_单batch全驻留(输入+输出)"
l2_out = "resident(输出驻留L2异步回写)"
r_in = 1.0
elif s_in <= s.l2_bytes:
l2_scene = "B_输入驻留输出直写GM"
elif a_b + bb_b <= s.l2_bytes:
l2_scene = "B_单batch输入驻留输出直写GM"
l2_out = "direct_gm(输出直写GM不占L2)"
r_in = 1.0
else:
l2_scene = "C_输入超L2分组执行"
l2_scene = "C_单batch输入超L2分组执行"
l2_out = "direct_gm(输出直写GM不占L2)"
r_in = self._l2_group_r_in(case, single_m, single_n)
r_in = self._l2_group_r_gm(case, single_m, single_n)
# Step 6: k_l1 (GM->L1 K 向粒度, 须 >= dValue 下限; K 小于下限时整 K 一次搬入不切)
dv_min_elems = max(s.dvalue_hw_min // dt, 1)
@@ -124,7 +125,7 @@ class AswBasicBranch(Branch):
tail_block_cnt=tail["r"], tail_wave_num=tail["n_wave"],
fixpipe_unitflag=True,
out_dtype_bytes=case.dtype_out_bytes,
note=f"L2场景{l2_scene}, r_in={r_in:.2f}; 尾轮: {tail['reason']}",
note=f"L2场景{l2_scene}, GM首读倍率={r_in:.2f}; 尾轮: {tail['reason']}",
)
# ------------------------------------------------------------------
@@ -189,12 +190,18 @@ class AswBasicBranch(Branch):
w = d
return w
def _l2_group_r_in(self, case, sm, sn) -> float:
"""场景 C: L2 分组的重复读倍率 r_in = (n_grp*M + m_grp*N)/(M+N)."""
def _l2_group_r_gm(self, case, sm, sn) -> float:
"""场景 C (单 batch 输入仍超 L2): 分组执行的 GM 重复读倍率.
每组 (m_grp x n_grp 个 tile) 输入工作集 <= L2; 组间共享块复用落空回 GM.
预算 D 按**单 batch** 计 (issue#24: 线性映射同一时刻只激活 1 个 batch,
不再除以总 batch 数).
r_gm = (ceil(n_cnt/n_grp)*M + ceil(m_cnt/m_grp)*N)/(M+N).
"""
s = self.spec
m, n, k, b = case.m, case.n, case.k, case.batch_c
m, n, k = case.m, case.n, case.k
dt = case.dtype_in_bytes
d = s.l2_bytes / (b * k * dt)
d = s.l2_bytes / (k * dt)
m_grp = max(1, int(d / (2 * sm)))
n_grp = max(1, int(d / (2 * sn)))
m_cnt = ceil_div(m, sm)
@@ -261,37 +268,64 @@ class AswBasicBranch(Branch):
# ------------------------------------------------------------------
def evaluate(self, case: BmmCase, plan: ImplPlan) -> HardwareTiming:
"""MTE2 两段块级模型 (issue#24, 用户澄清):
- 首读走 GM (按 GM 带宽, 不叠加 L2); 共享块 (A 行块被 n_cnt 个 tile 读、
B 列块被 m_cnt 个 tile 读) 驻留 L2 后其余 (n_cnt-1)/(m_cnt-1) 次读走
L2 读口 (5.2TB/s 独享) —— 场景 A/B (单 batch 工作集可驻留);
- 单 batch 输入超 L2 -> 场景 C 分组执行, GM 重复按组间落空计 (r_gm), 窗口内
复用未另计 (保守);
- 输出: 场景 A 驻留 L2 (5.2 写口) / 场景 B,C 直写 GM —— GM 读写共享总线
累加由 assemble 的 MTE2 链处理 (issue#23).
"""
s = self.spec
b = case.batch_c
m, n, k = case.m, case.n, case.k
dt = case.dtype_in_bytes
out_b = case.dtype_out_bytes
used = max(plan.used_core_num, 1)
flops = 2.0 * b * m * n * k
t_mmad = flops / (plan.used_core_num * s.q16)
t_mmad = flops / (used * s.q16)
in_bytes = b * (m * k + k * n) * dt
out_bytes = b * m * n * out_b
# r_in 从 note 里解析困难, 重新计算
s_in = in_bytes
s_out = out_bytes
if s_in + s_out <= s.l2_bytes or s_in <= s.l2_bytes:
r_in = 1.0
# ---- 字节量 (每 batch 口径) ----
a_b = m * k * dt # 每 batch A 字节
bb_b = k * n * dt # 每 batch B 字节
out_all = b * m * n * out_b # 输出总字节
m_cnt = max(plan.m_cnt, 1)
n_cnt = max(plan.n_cnt, 1)
# ---- 场景判定: 按单 batch 工作集 (同一时刻单 batch 激活) ----
if a_b + bb_b + m * n * out_b <= s.l2_bytes:
scene, to_l2, r_gm = "A_单batch全驻留(输入+输出)", True, 1.0
elif a_b + bb_b <= s.l2_bytes:
scene, to_l2, r_gm = "B_单batch输入驻留,输出直写GM", False, 1.0
else:
r_in = self._l2_group_r_in(case, plan.single_core_m, plan.single_core_n)
scene, to_l2, r_gm = "C_单batch输入超L2分组执行", False, \
self._l2_group_r_gm(case, plan.single_core_m, plan.single_core_n)
t_mte2 = r_in * in_bytes / (plan.used_core_num * s.bw_pc)
to_l2 = "resident" in plan.l2_policy_out
t_fix = out_bytes / (plan.used_core_num * (s.bw_l2_pc if to_l2 else s.bw_pc))
# ---- MTE2: GM 段 (首读, 每字节一次) + L2 段 (共享块重复读) ----
gm_read = r_gm * b * (a_b + bb_b)
if scene.startswith(("A_", "B_")):
# 驻留命中: A 行块被 n_cnt 个 tile 复用 -> 其余 (n_cnt-1) 次走 L2 读口
l2_read = b * ((n_cnt - 1) * a_b + (m_cnt - 1) * bb_b)
else:
l2_read = 0.0 # 场景 C: 组间复用落空(已计 GM), 窗口内复用保守不计
t_gm = gm_read / (used * s.bw_pc)
t_l2 = l2_read / (used * s.bw_l2_pc)
# ---- Fixpipe ----
t_fix = out_all / (used * (s.bw_l2_pc if to_l2 else s.bw_pc))
# drain: 尾轮暴露 (方案 B 已均匀重切, drain 小; A1b 尾轮凑满, drain 小; A0 尾轮 r 核空转)
t_drain = 0.0
if plan.tail_strategy == "A0" and plan.tail_block_cnt > 0:
t_drain = max(t_mmad, t_mte2, t_fix) # 尾轮空转一个整块
t_drain = max(t_mmad, t_gm + t_l2, t_fix) # 尾轮空转一个整块
return assemble_timing(
t_mte2_gm=t_mte2, t_mte2_l2=0.0, t_dma_cmd=0.0,
t_mte2_gm=t_gm, t_mte2_l2=t_l2, t_dma_cmd=0.0,
t_mmad=t_mmad, t_fixpipe=t_fix, t_reduce=0.0, t_drain=t_drain,
gm_read_bytes=r_in * in_bytes, l2_read_bytes=0.0, dma_cmd_count=0.0,
cube_flops=flops, fixpipe_bytes=out_bytes,
gm_read_bytes=gm_read, l2_read_bytes=l2_read, dma_cmd_count=0.0,
cube_flops=flops, fixpipe_bytes=out_all,
fixpipe_to_gm=(not to_l2),
)

63
BMM/BMM_Theory/bmm_theory/branches/stream_k.py
View File

@@ -129,40 +129,59 @@ class StreamKBranch(Branch):
# ------------------------------------------------------------------
def evaluate(self, case: BmmCase, plan: ImplPlan) -> HardwareTiming:
"""芯片口径评估, 按 plan 实际 tile 布局 (issue#25).
切 K 组 (tile = single_m x single_n) 由 grid_k 核协作: 组内 K 段零重复读
(GM 一次); 组间共享 (A 行块被 n_cnt 个 tile 用) 在单 batch 工作集可驻留时
走 L2 读口 (与 ASW 同规则). 全部 tile-组在 C 核上分波执行:
waves = ceil(b*m_cnt*n_cnt*grid_k / C).
- GM 读: 首读总量 = b*(m*k + k*n)*dt (组间共享命中 L2 时), 或按 r_gm 放大
(工作集超 L2, 保守同 ASW 场景 C 公式);
- MMAD: 芯片总量 2*b*m*n*k / (C核) —— 全核忙稳态;
- 归约 (部分和 4B 写 L2/AIV 读回求和/最终写回): 每 tile-组一次、组间串行
追加为 drain: t_drain = waves * eval_streamk_reduce(tile, grid_k, out).
"""
s = self.spec
b = case.batch_c
m, n, k = case.m, case.n, case.k
dt = case.dtype_in_bytes
out_b = case.dtype_out_bytes
grid_k = plan.grid_k
grid_k = max(plan.grid_k, 1)
m_cnt = max(plan.m_cnt, 1)
n_cnt = max(plan.n_cnt, 1)
tile_m = max(plan.single_core_m, 1)
tile_n = max(plan.single_core_n, 1)
tile = tile_m * tile_n # 每核实际 tile 元素数
# 单 tile (L0C 满载基本块)
tile_elems = s.l0c_elems # 65536
t_mmad_tile = 2.0 * k * s.l0c_bytes / 4 / s.q16 # 2K/Q16 * L0C/4B
t_mte2_tile = k * (2 * math.sqrt(tile_elems)) * dt / s.bw_pc # 近似方形 tile
# ---- MTE2: GM 段 + L2 段 (同 ASW 块级规则) ----
a_b = m * k * dt
bb_b = k * n * dt
if a_b + bb_b <= s.l2_bytes:
gm_read = b * (a_b + bb_b) # 组间共享命中 L2: GM 每字节一次
l2_read = b * ((n_cnt - 1) * a_b + (m_cnt - 1) * bb_b)
else:
gm_read = b * (a_b + bb_b) # 保守: 共享跨组回落 GM 未另计
l2_read = 0.0
t_gm = gm_read / s.bw_gm # 芯片口径 (全核并发)
t_l2 = l2_read / s.bw_l2
t_mte2 = t_gm + t_l2
# 切 K 后流水时延缩 grid_k 倍
t_mmad = t_mmad_tile / grid_k
t_mte2 = t_mte2_tile / grid_k
# ---- MMAD: 芯片总量 ----
flops = 2.0 * b * m * n * k
t_mmad = flops / (s.aic_num * s.q16)
# 归约: 部分和 4B 驻留 L2, AIV 归约 (含部分和写/读回/求和/最终按 C dtype 写回)
# 口径 (issue#11/#17): 归约整体为串行追加 (t_drain=t_reduce, reduce_serial=True),
# 部分和写出已计入 eval_streamk_reduce 的 t_write_partial —— 稳态 Fixpipe 不再
# 重复计账. 若再按 grid_k*tile*4B/单核带宽份额另计一次, 既重复计账又把整组
# 部分和串行压到单核写口, 高估 grid_k 倍 (streamk_demo 曾虚高到 55us/FIXPIPE;
# 第三轮曾回退该修复, issue#17 恢复).
t_reduce = eval_streamk_reduce(tile_elems, grid_k, out_b, s)
# ---- 归约: 每 tile-组一次, 组间分波串行追加 ----
waves = ceil_div(b * m_cnt * n_cnt * grid_k, s.aic_num)
t_reduce_group = eval_streamk_reduce(tile, grid_k, out_b, s)
t_reduce = waves * t_reduce_group
fix_bytes = 0.0
t_fix = 0.0
flops_pc = 2.0 * tile_elems * k / grid_k
gm_bytes = k * (2 * math.sqrt(tile_elems)) * dt / grid_k
return assemble_timing(
t_mte2_gm=t_mte2, t_mte2_l2=0.0, t_dma_cmd=0.0,
t_mte2_gm=t_gm, t_mte2_l2=t_l2, t_dma_cmd=0.0,
t_mmad=t_mmad, t_fixpipe=t_fix, t_reduce=t_reduce,
t_drain=t_reduce, # 归约串行追加 (reduce_serial 默认 True, 不进稳态 max)
gm_read_bytes=gm_bytes, l2_read_bytes=0.0, dma_cmd_count=0.0,
cube_flops=flops_pc, fixpipe_bytes=fix_bytes,
reduce_serial=True,
gm_read_bytes=gm_read, l2_read_bytes=l2_read, dma_cmd_count=0.0,
cube_flops=flops, fixpipe_bytes=fix_bytes,
reduce_serial=True, fixpipe_to_gm=False,
)

9
BMM/BMM_Theory/bmm_theory/evaluator.py
View File

@@ -66,11 +66,10 @@ class PlanEvaluator:
if not res.feasible:
tips.append("方案违反硬件约束, 需先修正: " + res.violations)
bn = t.bottleneck
if bn == "MTE2_GM":
tips.append("瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, "
"或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)")
elif bn == "MTE2_L2":
tips.append("瓶颈在 L2 重复读: 优化核间分配/swizzle 窗口压低活跃工作集")
if bn == "MTE2":
tips.append("瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile "
"提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/"
"分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争")
elif bn == "MMAD":
tips.append("瓶颈在 Cube 计算: 已接近理论算力上限, 检查是否有冗余计算 "
"(MergeBatch 交叉项) 可消除")

7
BMM/BMM_Theory/bmm_theory/hardware/ascend950pr.py
View File

@@ -39,11 +39,16 @@ class NpuSpec:
dvalue_hw_min: int = 256 # DMA 硬件突发下限 (Byte) —— 尾轮文档 §2.3
min_tile_size: int = 16 * 1024 # min_TileSize: 单块搬移最小量 16KB
min_datamount_per_core: int = 480 * 1024 # min_DatamountPerCore: 单核搬移总量下限 480KB
min_core_num_ratio: float = 0.8 # minCoreNum ≈ 0.8 * C
min_core_num_ratio: float = 0.8 # minCoreNum ≈ 0.8 * C (经验: 约 3/4 核并发才达 90%+ 带宽利用率)
# ---- DMA 固定开销 ----
t_cmd_ns: float = 50.0 # T_cmd: 单次 GM->L1 DMA 命令固定开销 (ns, 估计值, 需实测标定)
# ---- 带宽模型假设 (issue#26) ----
# GM 1.6TB/s 为读写共享总线 (读+写累加计时); L2 带宽读写各自独享 5.2TB/s;
# active_cores < C 时按"每核份额线性配平"为模型假设 (如 1 核也给 50GB/s),
# 真实低核数带宽利用率低于线性, 需 msProf 实测曲线标定后替换该假设.
# ---- Cube 计算粒度 ----
fractal: int = 16 # 16x16x16 基本块

25
BMM/BMM_Theory/bmm_theory/timing.py
View File

@@ -99,19 +99,31 @@ def assemble_timing(t_mte2_gm: float, t_mte2_l2: float, t_dma_cmd: float,
gm_read_bytes: float, l2_read_bytes: float,
dma_cmd_count: float, cube_flops: float,
fixpipe_bytes: float,
reduce_serial: bool = True) -> HardwareTiming:
reduce_serial: bool = True,
fixpipe_to_gm: bool = True) -> HardwareTiming:
"""汇总各级时延, 判定瓶颈.
带宽端口口径 (issue#23, 用户澄清 + KB):
- GM 1.6TB/s 为**读写共享总线**: MTE2 的 GM 读与 Fixpipe 直写 GM 并发时无法
拆分读写占用, 时延累加 (读+写)/1.6TB/s, 并入 MTE2 搬移链;
- L2 带宽读写各自独享 5.2TB/s: L2 重复读段(读口) 与 Fixpipe→L2 写(写口)
互不竞争; MTE2 引擎顺序服务 GM/L2 装载, 两段相加 (官方 T≈HBM/1.6+L2/5.2);
- Fixpipe→L2 (resident) 写出独立为 FIXPIPE 级, 不进 GM 总线.
归约计账约定 (issue#9): REDUCE 默认**串行追加** (reduce_serial=True, 归约不可
掩盖, 体现在 t_drain 中), 不进稳态 max(); 仅当调用方显式声明归约可流水掩盖
(reduce_serial=False) 时才进稳态 max(). 避免"既取最大又串行追加"的双倍计账.
"""
t_mte2 = t_mte2_gm + t_mte2_l2 + t_dma_cmd
fix_gm = t_fixpipe if fixpipe_to_gm else 0.0 # Fixpipe 直写 GM 的时延
fix_l2 = 0.0 if fixpipe_to_gm else t_fixpipe # Fixpipe→L2 (5.2TB/s 写口)
# MTE2 搬移链 = GM 总线(读写共享, 读+直写累加) + L2 重复读段 + DMA 命令开销
# (同一 MTE2 引擎顺序服务 GM/L2 两类装载; GM 写由 Fixpipe 并发发起, 共享 GM 总线)
mte2_chain = t_mte2_gm + t_dma_cmd + fix_gm + t_mte2_l2
stages = {
"MTE2_GM": t_mte2_gm + t_dma_cmd,
"MTE2_L2": t_mte2_l2,
"MTE2": mte2_chain,
"MMAD": t_mmad,
"FIXPIPE": t_fixpipe,
"FIXPIPE": fix_l2, # 仅 Fixpipe→L2 (5.2 写口, 与 L2 读口互不竞争)
}
if not reduce_serial:
stages["REDUCE"] = t_reduce
@@ -135,8 +147,7 @@ def assemble_timing(t_mte2_gm: float, t_mte2_l2: float, t_dma_cmd: float,
def bound_type_of(bottleneck: str) -> str:
return {
"MMAD": "计算Bound",
"MTE2_GM": "访存Bound(GM)",
"MTE2_L2": "访存Bound(L2)",
"FIXPIPE": "写出Bound",
"MTE2": "访存Bound(GM读写共享+L2重复读)",
"FIXPIPE": "写出Bound(L2写口)",
"REDUCE": "归约Bound",
}.get(bottleneck, "")

44
BMM/BMM_Theory/examples/plans.csv
View File

@@ -7,7 +7,7 @@ merge_iter_arbitrate,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮
iter_demo_form_b,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),4,1,64,64,256,256,2,b_双batch乒乓,64,64,256,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,双batch乒乓: 2*(MK+KN)*dtype=128KB <= L1
iter_demo_form_d,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),2,1,64,64,8192,1024,1,d_两侧都切K,64,64,256,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"两侧都切K: k_L1=1024, K段成对流水, batch边界天然无缝; dValueA=2048B/dValueB=128B"
streamk_demo,StreamK,Ascend950PR,batch_mat_mul_v3,32,1,1,1,32,"B/M/N切出4块, 每块32核切K归约 (归约组内核c负责K段[c*K/32,(c+1)*K/32))",1,1,128,128,320,256,1,K段标准分块流水,128,128,64,allocate(部分和驻留L2),"resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换)",0,8388608,grid_K=32路切K+归约,1,1,32,0,0,0,0,True,4,"P=1.00, grid_K=32, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终"
asw_demo_full,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,1024,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,52,52,1,52,52,2,139,True,2,"L2场景B_输入驻留输出直写GM, r_in=1.00; 尾轮: 周长型主导, rho=0.06<rho_dv=0.53, A1b被dValue卡死, 方案B反超"
asw_demo_full,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,1024,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,52,52,1,52,52,2,139,True,2,"L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: 周长型主导, rho=0.06<rho_dv=0.53, A1b被dValue卡死, 方案B反超"
asw_demo_reduce_core,ASW_Basic_降核,Ascend950PR,batch_mat_mul_v3,16,1,1,1,1,"降核: 只用16核, 每核一个L0C满载输出块, 其余核闲置",0,1,256,256,128,128,1,标准核内流水,256,256,64,allocate,direct_gm,0,0,不涉及(每核一块无尾轮),1,1,1,0,0,0,0,True,2,"P=16.00<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢)"
b4_m1_n128_k256,ASW_Basic_降核,Ascend950PR,batch_mat_mul_v3,1,1,1,1,1,"降核: 只用1核, 每核一个L0C满载输出块, 其余核闲置",0,1,1,128,256,256,1,标准核内流水,1,128,128,allocate,direct_gm,0,0,不涉及(每核一块无尾轮),1,1,1,0,0,0,0,True,2,"P=0.01<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢)"
b8_m2_n192_k128,ASW_Basic_降核,Ascend950PR,batch_mat_mul_v3,1,1,1,1,1,"降核: 只用1核, 每核一个L0C满载输出块, 其余核闲置",0,1,2,192,128,128,1,标准核内流水,2,192,80,allocate,direct_gm,0,0,不涉及(每核一块无尾轮),1,1,1,0,0,0,0,True,2,"P=0.05<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢)"
@@ -15,31 +15,31 @@ b16_m4_n256_k192,ASW_Basic_降核,Ascend950PR,batch_mat_mul_v3,1,1,1,1,1,"降核
b32_m8_n128_k256,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),1,1,8,128,256,256,1,a_单batch全驻留,8,128,128,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,单batch全驻留: (MK+KN)*dtype=68KB <= L1
b64_m16_n256_k512,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),2,1,16,256,512,240,1,c_一侧驻留+对侧切K,16,256,64,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"一侧驻留(A)+对侧切K: A驻留16KB, 预算L1/2, k_L1=240, dValueA=480B/dValueB=512B"
b128_m8_n192_k256,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),4,1,8,192,256,256,2,b_双batch乒乓,8,192,80,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,双batch乒乓: 2*(MK+KN)*dtype=200KB <= L1
b32_m16_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,1,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,47,True,2,"L2场景C_输入超L2分组执行, r_in=1.09; 尾轮: r=0 无尾轮"
b32_m16_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,1,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,47,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮"
b4_m1_n8192_k8192,StreamK,Ascend950PR,batch_mat_mul_v3,32,1,1,1,32,"B/M/N切出4块, 每块32核切K归约 (归约组内核c负责K段[c*K/32,(c+1)*K/32))",1,1,1,8192,256,256,1,K段标准分块流水,1,128,64,allocate(部分和驻留L2),"resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换)",0,4194304,grid_K=32路切K+归约,1,1,32,0,0,0,0,True,4,"P=0.50, grid_K=32, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终"
b16_m2_n4096_k7168,StreamK,Ascend950PR,batch_mat_mul_v3,32,1,1,1,16,"B/M/N切出16块, 每块16核切K归约 (归约组内核c负责K段[c*K/16,(c+1)*K/16))",1,1,2,4096,448,256,1,K段标准分块流水,2,128,64,allocate(部分和驻留L2),"resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换)",0,8388608,grid_K=16路切K+归约,1,1,16,0,0,0,0,True,4,"P=2.00, grid_K=16, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终"
b32_m64_n64_k7168,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),1,1,64,64,7168,1024,1,d_两侧都切K,64,64,256,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"两侧都切K: k_L1=1024, K段成对流水, batch边界天然无缝; dValueA=2048B/dValueB=128B"
b64_m1024_n1024_k7168,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),2,1,1024,1024,7168,64,1,d_两侧都切K,176,176,64,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"两侧都切K: k_L1=64, K段成对流水, batch边界天然无缝; dValueA=128B/dValueB=2048B"
b128_m2048_n2048_k1536,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,12,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,1536,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,576,True,2,"L2场景C_输入超L2分组执行, r_in=12.00; 尾轮: r=0 无尾轮"
b64_m1024_n8192_k2048,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,6,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,2048,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,564,True,2,"L2场景C_输入超L2分组执行, r_in=10.56; 尾轮: r=0 无尾轮"
b128_m2048_n2048_k1536,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,12,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,1536,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,576,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮"
b64_m1024_n8192_k2048,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,6,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,2048,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,564,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮"
b32_m1024_n1024_k512,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),1,1,1024,1024,512,64,1,d_两侧都切K,176,176,64,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"两侧都切K: k_L1=64, K段成对流水, batch边界天然无缝; dValueA=128B/dValueB=2048B"
b128_m4096_n4096_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,24,24,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,2304,True,2,"L2场景C_输入超L2分组执行, r_in=24.00; 尾轮: r=0 无尾轮"
b32_m8192_n4096_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,24,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,1128,True,2,"L2场景C_输入超L2分组执行, r_in=31.67; 尾轮: r=0 无尾轮"
b32_m2048_n2048_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,12,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,144,True,2,"L2场景C_输入超L2分组执行, r_in=12.00; 尾轮: r=0 无尾轮"
b64_m4096_n2048_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,24,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,576,True,2,"L2场景B_输入驻留输出直写GM, r_in=1.00; 尾轮: r=0 无尾轮"
b16_m1024_n1024_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,6,6,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,18,True,2,"L2场景C_输入超L2分组执行, r_in=6.00; 尾轮: r=0 无尾轮"
b4_m32768_n128_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,1,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,方案B,205,1,1,205,1,12,24,True,2,"L2场景A_全驻留, r_in=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=28.96us), 选方案B工程简洁 (一套tile)"
b8_m32768_n2048_k512,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,512,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,561,True,2,"L2场景C_输入超L2分组执行, r_in=1.24; 尾轮: r=0 无尾轮"
b4_m131072_n128_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,1,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,820,1,1,820,1,4,94,True,2,"L2场景C_输入超L2分组执行, r_in=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=115.39us), 选方案B工程简洁 (一套tile)"
b8_m131072_n1024_k256,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,6,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,256,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,820,7,1,820,7,16,1118,True,2,"L2场景C_输入超L2分组执行, r_in=1.06; 尾轮: 面积型主导, A1b与方案B严格打平(T=1384.63us), 选方案B工程简洁 (一套tile)"
b16_m32768_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,205,52,1,205,52,16,4395,True,2,"L2场景C_输入超L2分组执行, r_in=75.00; 尾轮: 周长型主导, rho=0.50<rho_dv=0.53, A1b被dValue卡死, 方案B反超"
b4_m32768_n128_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,1,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,205,1,1,205,1,12,24,True,2,"L2场景C_输入超L2分组执行, r_in=1.14; 尾轮: 周长型主导, rho=0.38<rho_dv=0.53, A1b被dValue卡死, 方案B反超"
b32_m131072_n8192_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,35015,True,2,"L2场景C_输入超L2分组执行, r_in=2.88; 尾轮: r=0 无尾轮"
b64_m32768_n8192_k1536,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,1536,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,17578,True,2,"L2场景C_输入超L2分组执行, r_in=75.00; 尾轮: r=0 无尾轮"
b8_m131072_n8192_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A1b,2,2,1,2,2,24,8754,True,2,"L2场景C_输入超L2分组执行, r_in=44.53; 尾轮: 周长型主导, rho=0.75>=rho_dv=0.53, A1b恒优 (尾轮tile缩√rho=0.87倍凑满核)"
b128_m4096_n4096_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,24,24,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,2304,True,2,"L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮"
b32_m8192_n4096_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,24,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,1128,True,2,"L2场景C_单batch输入超L2分组执行, GM首读倍率=1.33; 尾轮: r=0 无尾轮"
b32_m2048_n2048_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,12,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,144,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮"
b64_m4096_n2048_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,24,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,576,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮"
b16_m1024_n1024_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,6,6,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,18,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮"
b4_m32768_n128_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,1,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,方案B,205,1,1,205,1,12,24,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=28.96us), 选方案B工程简洁 (一套tile)"
b8_m32768_n2048_k512,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,512,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,561,True,2,"L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮"
b4_m131072_n128_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,1,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,方案B,820,1,1,820,1,4,94,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=115.39us), 选方案B工程简洁 (一套tile)"
b8_m131072_n1024_k256,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,6,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,256,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,820,7,1,820,7,16,1118,True,2,"L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=1384.63us), 选方案B工程简洁 (一套tile)"
b16_m32768_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,205,52,1,205,52,16,4395,True,2,"L2场景C_单batch输入超L2分组执行, GM首读倍率=3.20; 尾轮: 周长型主导, rho=0.50<rho_dv=0.53, A1b被dValue卡死, 方案B反超"
b4_m32768_n128_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,1,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,205,1,1,205,1,12,24,True,2,"L2场景C_单batch输入超L2分组执行, GM首读倍率=1.03; 尾轮: 周长型主导, rho=0.38<rho_dv=0.53, A1b被dValue卡死, 方案B反超"
b32_m131072_n8192_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,35015,True,2,"L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮"
b64_m32768_n8192_k1536,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,1536,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,17578,True,2,"L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮"
b8_m131072_n8192_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A1b,2,2,1,2,2,24,8754,True,2,"L2场景C_单batch输入超L2分组执行, GM首读倍率=4.76; 尾轮: 周长型主导, rho=0.75>=rho_dv=0.53, A1b恒优 (尾轮tile缩√rho=0.87倍凑满核)"
b8_m16_n7168_k1536,StreamK,Ascend950PR,batch_mat_mul_v3,32,1,1,2,2,"B/M/N切出16块, 每块2核切K归约 (归约组内核c负责K段[c*K/2,(c+1)*K/2))",1,1,16,3584,768,256,1,K段标准分块流水,16,128,64,allocate(部分和驻留L2),"resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换)",0,3670016,grid_K=2路切K+归约,1,1,2,0,0,0,0,True,4,"P=14.00, grid_K=2, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终"
b64_m1024_n7168_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,6,41,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,492,True,2,"L2场景C_输入超L2分组执行, r_in=10.38; 尾轮: r=0 无尾轮"
b32_m8192_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,2209,True,2,"L2场景C_输入超L2分组执行, r_in=47.00; 尾轮: r=0 无尾轮"
b8_m4096_n4096_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,24,24,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,144,True,2,"L2场景B_输入驻留输出直写GM, r_in=1.00; 尾轮: r=0 无尾轮"
b128_m8192_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,8836,True,2,"L2场景C_输入超L2分组执行, r_in=47.00; 尾轮: r=0 无尾轮"
b64_m1024_n7168_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,6,41,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,492,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮"
b32_m8192_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,2209,True,2,"L2场景C_单batch输入超L2分组执行, GM首读倍率=2.00; 尾轮: r=0 无尾轮"
b8_m4096_n4096_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,24,24,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,144,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮"
b128_m8192_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,8836,True,2,"L2场景C_单batch输入超L2分组执行, GM首读倍率=2.00; 尾轮: r=0 无尾轮"
special_k1_b64,特殊分支,Ascend950PR,batch_mat_mul_v3,64,1,1,1,1,AIV 核间按行均分 (无 Cube tile 概念),0,1,0,0,1,0,1,UB驻留(AIV) AIV单缓冲,0,0,0,allocate,direct_gm,0,0,不涉及(AIV逐元素),1,1,1,0,0,0,0,False,2,"K=1逐元素乘: 退化为 C=A⊙B 无累加深度, Cube 16x16x16 粒度浪费 15/16; 走 AIV 通路 GM->UB->Mul->GM, AIV单缓冲 (B<2*AIV 逐batch单缓冲串行)"
1 case_id branch npu op used_core_num split_b m_cnt n_cnt grid_k core_map b_core merge_b0 single_core_m single_core_n single_core_k k_l1 b_l1 l1_form base_m base_n base_k l2_policy_in l2_policy_out swizzle_w workspace_bytes tail_strategy tail_m_cnt tail_n_cnt tail_k_cnt tail_m_main tail_n_main tail_block_cnt tail_wave_num fixpipe_unitflag out_dtype_bytes note
7 iter_demo_form_b IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 4 1 64 64 256 256 2 b_双batch乒乓 64 64 256 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 双batch乒乓: 2*(MK+KN)*dtype=128KB <= L1
8 iter_demo_form_d IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 2 1 64 64 8192 1024 1 d_两侧都切K 64 64 256 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 两侧都切K: k_L1=1024, K段成对流水, batch边界天然无缝; dValueA=2048B/dValueB=128B
9 streamk_demo StreamK Ascend950PR batch_mat_mul_v3 32 1 1 1 32 B/M/N切出4块, 每块32核切K归约 (归约组内核c负责K段[c*K/32,(c+1)*K/32)) 1 1 128 128 320 256 1 K段标准分块流水 128 128 64 allocate(部分和驻留L2) resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换) 0 8388608 grid_K=32路切K+归约 1 1 32 0 0 0 0 True 4 P=1.00, grid_K=32, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终
10 asw_demo_full ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 47 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 1024 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 方案B 52 52 1 52 52 2 139 True 2 L2场景B_输入驻留输出直写GM, r_in=1.00; 尾轮: 周长型主导, rho=0.06<rho_dv=0.53, A1b被dValue卡死, 方案B反超 L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: 周长型主导, rho=0.06<rho_dv=0.53, A1b被dValue卡死, 方案B反超
11 asw_demo_reduce_core ASW_Basic_降核 Ascend950PR batch_mat_mul_v3 16 1 1 1 1 降核: 只用16核, 每核一个L0C满载输出块, 其余核闲置 0 1 256 256 128 128 1 标准核内流水 256 256 64 allocate direct_gm 0 0 不涉及(每核一块无尾轮) 1 1 1 0 0 0 0 True 2 P=16.00<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢)
12 b4_m1_n128_k256 ASW_Basic_降核 Ascend950PR batch_mat_mul_v3 1 1 1 1 1 降核: 只用1核, 每核一个L0C满载输出块, 其余核闲置 0 1 1 128 256 256 1 标准核内流水 1 128 128 allocate direct_gm 0 0 不涉及(每核一块无尾轮) 1 1 1 0 0 0 0 True 2 P=0.01<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢)
13 b8_m2_n192_k128 ASW_Basic_降核 Ascend950PR batch_mat_mul_v3 1 1 1 1 1 降核: 只用1核, 每核一个L0C满载输出块, 其余核闲置 0 1 2 192 128 128 1 标准核内流水 2 192 80 allocate direct_gm 0 0 不涉及(每核一块无尾轮) 1 1 1 0 0 0 0 True 2 P=0.05<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢)
15 b32_m8_n128_k256 IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 1 1 8 128 256 256 1 a_单batch全驻留 8 128 128 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 单batch全驻留: (MK+KN)*dtype=68KB <= L1
16 b64_m16_n256_k512 IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 2 1 16 256 512 240 1 c_一侧驻留+对侧切K 16 256 64 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 一侧驻留(A)+对侧切K: A驻留16KB, 预算L1/2, k_L1=240, dValueA=480B/dValueB=512B
17 b128_m8_n192_k256 IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 4 1 8 192 256 256 2 b_双batch乒乓 8 192 80 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 双batch乒乓: 2*(MK+KN)*dtype=200KB <= L1
18 b32_m16_n8192_k7168 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 1 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 7168 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 47 True 2 L2场景C_输入超L2分组执行, r_in=1.09; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮
19 b4_m1_n8192_k8192 StreamK Ascend950PR batch_mat_mul_v3 32 1 1 1 32 B/M/N切出4块, 每块32核切K归约 (归约组内核c负责K段[c*K/32,(c+1)*K/32)) 1 1 1 8192 256 256 1 K段标准分块流水 1 128 64 allocate(部分和驻留L2) resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换) 0 4194304 grid_K=32路切K+归约 1 1 32 0 0 0 0 True 4 P=0.50, grid_K=32, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终
20 b16_m2_n4096_k7168 StreamK Ascend950PR batch_mat_mul_v3 32 1 1 1 16 B/M/N切出16块, 每块16核切K归约 (归约组内核c负责K段[c*K/16,(c+1)*K/16)) 1 1 2 4096 448 256 1 K段标准分块流水 2 128 64 allocate(部分和驻留L2) resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换) 0 8388608 grid_K=16路切K+归约 1 1 16 0 0 0 0 True 4 P=2.00, grid_K=16, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终
21 b32_m64_n64_k7168 IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 1 1 64 64 7168 1024 1 d_两侧都切K 64 64 256 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 两侧都切K: k_L1=1024, K段成对流水, batch边界天然无缝; dValueA=2048B/dValueB=128B
22 b64_m1024_n1024_k7168 IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 2 1 1024 1024 7168 64 1 d_两侧都切K 176 176 64 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 两侧都切K: k_L1=64, K段成对流水, batch边界天然无缝; dValueA=128B/dValueB=2048B
23 b128_m2048_n2048_k1536 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 12 12 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 1536 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 576 True 2 L2场景C_输入超L2分组执行, r_in=12.00; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮
24 b64_m1024_n8192_k2048 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 6 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 2048 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 564 True 2 L2场景C_输入超L2分组执行, r_in=10.56; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮
25 b32_m1024_n1024_k512 IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 1 1 1024 1024 512 64 1 d_两侧都切K 176 176 64 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 两侧都切K: k_L1=64, K段成对流水, batch边界天然无缝; dValueA=128B/dValueB=2048B
26 b128_m4096_n4096_k8192 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 24 24 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 8192 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A0 1 1 1 0 0 0 2304 True 2 L2场景C_输入超L2分组执行, r_in=24.00; 尾轮: r=0 无尾轮 L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮
27 b32_m8192_n4096_k7168 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 47 24 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 7168 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A0 1 1 1 0 0 0 1128 True 2 L2场景C_输入超L2分组执行, r_in=31.67; 尾轮: r=0 无尾轮 L2场景C_单batch输入超L2分组执行, GM首读倍率=1.33; 尾轮: r=0 无尾轮
28 b32_m2048_n2048_k8192 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 12 12 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 8192 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 144 True 2 L2场景C_输入超L2分组执行, r_in=12.00; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮
29 b64_m4096_n2048_k128 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 24 12 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 128 128 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 576 True 2 L2场景B_输入驻留输出直写GM, r_in=1.00; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮
30 b16_m1024_n1024_k8192 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 6 6 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 8192 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 18 True 2 L2场景C_输入超L2分组执行, r_in=6.00; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮
31 b4_m32768_n128_k128 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 187 1 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 128 128 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) resident(输出驻留L2异步回写) 4 0 方案B 205 1 1 205 1 12 24 True 2 L2场景A_全驻留, r_in=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=28.96us), 选方案B工程简洁 (一套tile) L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=28.96us), 选方案B工程简洁 (一套tile)
32 b8_m32768_n2048_k512 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 187 12 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 512 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A0 1 1 1 0 0 0 561 True 2 L2场景C_输入超L2分组执行, r_in=1.24; 尾轮: r=0 无尾轮 L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮
33 b4_m131072_n128_k128 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 745 1 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 128 128 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 方案B 820 1 1 820 1 4 94 True 2 L2场景C_输入超L2分组执行, r_in=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=115.39us), 选方案B工程简洁 (一套tile) L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=115.39us), 选方案B工程简洁 (一套tile)
34 b8_m131072_n1024_k256 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 745 6 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 256 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 方案B 820 7 1 820 7 16 1118 True 2 L2场景C_输入超L2分组执行, r_in=1.06; 尾轮: 面积型主导, A1b与方案B严格打平(T=1384.63us), 选方案B工程简洁 (一套tile) L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=1384.63us), 选方案B工程简洁 (一套tile)
35 b16_m32768_n8192_k7168 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 187 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 7168 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 方案B 205 52 1 205 52 16 4395 True 2 L2场景C_输入超L2分组执行, r_in=75.00; 尾轮: 周长型主导, rho=0.50<rho_dv=0.53, A1b被dValue卡死, 方案B反超 L2场景C_单batch输入超L2分组执行, GM首读倍率=3.20; 尾轮: 周长型主导, rho=0.50<rho_dv=0.53, A1b被dValue卡死, 方案B反超
36 b4_m32768_n128_k8192 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 187 1 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 8192 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 方案B 205 1 1 205 1 12 24 True 2 L2场景C_输入超L2分组执行, r_in=1.14; 尾轮: 周长型主导, rho=0.38<rho_dv=0.53, A1b被dValue卡死, 方案B反超 L2场景C_单batch输入超L2分组执行, GM首读倍率=1.03; 尾轮: 周长型主导, rho=0.38<rho_dv=0.53, A1b被dValue卡死, 方案B反超
37 b32_m131072_n8192_k128 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 745 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 128 128 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A0 1 1 1 0 0 0 35015 True 2 L2场景C_输入超L2分组执行, r_in=2.88; 尾轮: r=0 无尾轮 L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮
38 b64_m32768_n8192_k1536 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 187 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 1536 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A0 1 1 1 0 0 0 17578 True 2 L2场景C_输入超L2分组执行, r_in=75.00; 尾轮: r=0 无尾轮 L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮
39 b8_m131072_n8192_k8192 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 745 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 8192 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A1b 2 2 1 2 2 24 8754 True 2 L2场景C_输入超L2分组执行, r_in=44.53; 尾轮: 周长型主导, rho=0.75>=rho_dv=0.53, A1b恒优 (尾轮tile缩√rho=0.87倍凑满核) L2场景C_单batch输入超L2分组执行, GM首读倍率=4.76; 尾轮: 周长型主导, rho=0.75>=rho_dv=0.53, A1b恒优 (尾轮tile缩√rho=0.87倍凑满核)
40 b8_m16_n7168_k1536 StreamK Ascend950PR batch_mat_mul_v3 32 1 1 2 2 B/M/N切出16块, 每块2核切K归约 (归约组内核c负责K段[c*K/2,(c+1)*K/2)) 1 1 16 3584 768 256 1 K段标准分块流水 16 128 64 allocate(部分和驻留L2) resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换) 0 3670016 grid_K=2路切K+归约 1 1 2 0 0 0 0 True 4 P=14.00, grid_K=2, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终
41 b64_m1024_n7168_k7168 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 6 41 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 7168 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 492 True 2 L2场景C_输入超L2分组执行, r_in=10.38; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮
42 b32_m8192_n8192_k7168 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 47 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 7168 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A0 1 1 1 0 0 0 2209 True 2 L2场景C_输入超L2分组执行, r_in=47.00; 尾轮: r=0 无尾轮 L2场景C_单batch输入超L2分组执行, GM首读倍率=2.00; 尾轮: r=0 无尾轮
43 b8_m4096_n4096_k128 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 24 24 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 128 128 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 144 True 2 L2场景B_输入驻留输出直写GM, r_in=1.00; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮
44 b128_m8192_n8192_k7168 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 47 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 7168 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A0 1 1 1 0 0 0 8836 True 2 L2场景C_输入超L2分组执行, r_in=47.00; 尾轮: r=0 无尾轮 L2场景C_单batch输入超L2分组执行, GM首读倍率=2.00; 尾轮: r=0 无尾轮
45 special_k1_b64 特殊分支 Ascend950PR batch_mat_mul_v3 64 1 1 1 1 AIV 核间按行均分 (无 Cube tile 概念) 0 1 0 0 1 0 1 UB驻留(AIV) AIV单缓冲 0 0 0 allocate direct_gm 0 0 不涉及(AIV逐元素) 1 1 1 0 0 0 0 False 2 K=1逐元素乘: 退化为 C=A⊙B 无累加深度, Cube 16x16x16 粒度浪费 15/16; 走 AIV 通路 GM->UB->Mul->GM, AIV单缓冲 (B<2*AIV 逐batch单缓冲串行)

84
BMM/BMM_Theory/examples/result_evaluate.csv
View File

@@ -1,45 +1,45 @@
case_id,batch_a,batch_b,m,n,k,dtype_a,dtype_b,dtype_c,trans_a,trans_b,has_bias,out_nd,deterministic_level,plan_case_id,plan_branch,plan_npu,plan_op,plan_used_core_num,plan_split_b,plan_m_cnt,plan_n_cnt,plan_grid_k,plan_core_map,plan_b_core,plan_merge_b0,plan_single_core_m,plan_single_core_n,plan_single_core_k,plan_k_l1,plan_b_l1,plan_l1_form,plan_base_m,plan_base_n,plan_base_k,plan_l2_policy_in,plan_l2_policy_out,plan_swizzle_w,plan_workspace_bytes,plan_tail_strategy,plan_tail_m_cnt,plan_tail_n_cnt,plan_tail_k_cnt,plan_tail_m_main,plan_tail_n_main,plan_tail_block_cnt,plan_tail_wave_num,plan_fixpipe_unitflag,plan_out_dtype_bytes,plan_note,gm_read_bytes,l2_read_bytes,t_mte2_gm,t_mte2_l2,t_mte2,dma_cmd_count,t_dma_cmd,cube_flops,t_mmad,fixpipe_bytes,t_fixpipe,t_reduce,t_steady,t_drain,t_total,bottleneck,feasible,violations,bound_type,advice
to_matmul_demo,1,1,2048,2048,2048,bf16,bf16,bf16,False,False,False,True,0,to_matmul_demo,转Matmul,Ascend950PR,batch_mat_mul_v3,32,1,0,0,1,"折叠为 Matmul [2048,2048]x[2048,2048], 复用 Matmul 切分体系",0,1,0,0,2048,0,1,,0,0,0,,,0,0,转Matmul后由 Matmul 体系决定,1,1,1,0,0,0,0,True,2,"BatchB=1免费折叠: 左矩阵 [1,2048,2048] 视图折叠为 [2048,2048], 零重排零 split",16777216,0.0,1.048576e-05,0.0,1.048576e-05,0.0,0.0,17179869184.0,3.534952506995885e-05,8388608,5.24288e-06,0.0,3.534952506995885e-05,0.0,3.534952506995885e-05,MMAD,True,,计算Bound,"瓶颈在 Cube 计算: 已接近理论算力上限, 检查是否有冗余计算 (MergeBatch 交叉项) 可消除"
special_k0_demo,128,128,256,256,0,bf16,bf16,bf16,False,False,False,True,0,special_k0_demo,特殊分支,Ascend950PR,batch_mat_mul_v3,64,1,1,1,1,AIV 核间按行均分 (无 Cube tile 概念),0,1,0,0,0,0,1,UB驻留(AIV),0,0,0,allocate,direct_gm,0,0,不涉及(AIV逐元素),1,1,1,0,0,0,0,False,2,"K=0纯写值: 无任何计算, C=bias 或 0, 纯 AIV 写值; 按行均分到 AIV 核",0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,16777216,1.048576e-05,0.0,1.048576e-05,0.0,1.048576e-05,FIXPIPE,True,,写出Bound,"瓶颈在 Fixpipe 写出: 检查输出 dtype (fp16/fp8 可减半写出量), 或评估输出驻留 L2 异步回写策略"
special_k1_demo,128,128,256,256,1,bf16,bf16,bf16,False,False,False,True,0,special_k1_demo,特殊分支,Ascend950PR,batch_mat_mul_v3,64,1,1,1,1,AIV 核间按行均分 (无 Cube tile 概念),0,1,0,0,1,0,1,UB驻留(AIV) UB乒乓,0,0,0,allocate,direct_gm,0,0,不涉及(AIV逐元素),1,1,1,0,0,0,0,False,2,"K=1逐元素乘: 退化为 C=A⊙B 无累加深度, Cube 16x16x16 粒度浪费 15/16; 走 AIV 通路 GM->UB->Mul->GM, UB乒乓 (B>=2*AIV 双batch乒乓流水)",131072,0.0,8.192e-08,0.0,8.192e-08,0.0,0.0,8388608.0,6.206060606060606e-07,16777216,1.048576e-05,0.0,1.048576e-05,0.0,1.048576e-05,FIXPIPE,True,,写出Bound,"瓶颈在 Fixpipe 写出: 检查输出 dtype (fp16/fp8 可减半写出量), 或评估输出驻留 L2 异步回写策略"
merge_demo_k_trunc,2048,2048,32,32,256,bf16,bf16,bf16,False,False,False,True,0,merge_demo_k_trunc,MergeBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B均分(核间零重复读零依赖),64,4,128,128,256,256,8,合并驻留,128,128,128,allocate(GM->L1随路驻留L2),direct_gm,0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"b0=4 (L0C上限5.7/算存比上限19.0/b_core=64); K截断; 合并后单次DMA搬入 A'[128,256]+B'[256,128]",2097152,0.0,4.194304e-05,0.0,4.2743039999999997e-05,16.0,8.000000000000001e-07,134217728.0,8.837381267489712e-06,131072,2.62144e-06,0.0,4.2743039999999997e-05,3.0192408230452674e-07,4.3044964082304525e-05,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
merge_iter_arbitrate,128,128,64,64,512,bf16,bf16,bf16,False,False,False,True,0,merge_iter_arbitrate,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),4,1,64,64,512,512,2,b_双batch乒乓,64,64,256,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,双batch乒乓: 2*(MK+KN)*dtype=256KB <= L1,524288,0.0,1.048576e-05,0.0,1.0685759999999999e-05,4,2.0000000000000002e-07,16777216.0,1.104672658436214e-06,32768,6.5536e-07,0.0,1.0685759999999999e-05,4.400081646090535e-07,1.1125768164609053e-05,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
iter_demo_form_b,128,128,64,64,256,bf16,bf16,bf16,False,False,False,True,0,iter_demo_form_b,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),4,1,64,64,256,256,2,b_双batch乒乓,64,64,256,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,双batch乒乓: 2*(MK+KN)*dtype=128KB <= L1,262144,0.0,5.24288e-06,0.0,5.4428799999999995e-06,4,2.0000000000000002e-07,8388608.0,5.52336329218107e-07,32768,6.5536e-07,0.0,5.4428799999999995e-06,3.0192408230452674e-07,5.744804082304526e-06,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
iter_demo_form_d,64,64,64,64,8192,bf16,bf16,bf16,False,False,False,True,0,iter_demo_form_d,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),2,1,64,64,8192,1024,1,d_两侧都切K,64,64,256,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"两侧都切K: k_L1=1024, K段成对流水, batch边界天然无缝; dValueA=2048B/dValueB=128B",4194304,0.0,8.388608e-05,0.0,8.468608e-05,16,8.000000000000001e-07,134217728.0,8.837381267489712e-06,16384,3.2768e-07,0.0,8.468608e-05,7.16176329218107e-07,8.540225632921811e-05,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
streamk_demo,4,4,128,128,10240,bf16,bf16,bf16,False,False,False,True,0,streamk_demo,StreamK,Ascend950PR,batch_mat_mul_v3,32,1,1,1,32,"B/M/N切出4块, 每块32核切K归约 (归约组内核c负责K段[c*K/32,(c+1)*K/32))",1,1,128,128,320,256,1,K段标准分块流水,128,128,64,allocate(部分和驻留L2),"resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换)",0,8388608,grid_K=32路切K+归约,1,1,32,0,0,0,0,True,4,"P=1.00, grid_K=32, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终",327680.0,0.0,6.5536e-06,0.0,6.5536e-06,0.0,0.0,41943040.0,2.761681646090535e-06,0.0,0.0,3.4067453613053613e-06,6.5536e-06,3.4067453613053613e-06,9.960345361305361e-06,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
asw_demo_full,2,2,8192,8192,1024,bf16,bf16,bf16,False,False,False,True,0,asw_demo_full,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,1024,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,52,52,1,52,52,2,139,True,2,"L2场景B_输入驻留输出直写GM, r_in=1.00; 尾轮: 周长型主导, rho=0.06<rho_dv=0.53, A1b被dValue卡死, 方案B反超",67108864.0,0.0,4.194304e-05,0.0,4.194304e-05,0.0,0.0,274877906944.0,0.0005655924011193416,268435456,0.00016777216,0.0,0.0005655924011193416,0.0,0.0005655924011193416,MMAD,True,,计算Bound,"瓶颈在 Cube 计算: 已接近理论算力上限, 检查是否有冗余计算 (MergeBatch 交叉项) 可消除"
asw_demo_reduce_core,16,16,256,256,128,bf16,bf16,bf16,False,False,False,True,0,asw_demo_reduce_core,ASW_Basic_降核,Ascend950PR,batch_mat_mul_v3,16,1,1,1,1,"降核: 只用16核, 每核一个L0C满载输出块, 其余核闲置",0,1,256,256,128,128,1,标准核内流水,256,256,64,allocate,direct_gm,0,0,不涉及(每核一块无尾轮),1,1,1,0,0,0,0,True,2,"P=16.00<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢)",2097152.0,0.0,2.62144e-06,0.0,2.62144e-06,0.0,0.0,268435456.0,1.104672658436214e-06,2097152,2.62144e-06,0.0,2.62144e-06,0.0,2.62144e-06,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b4_m1_n128_k256,4,4,1,128,256,bf16,bf16,bf16,False,False,False,True,0,b4_m1_n128_k256,ASW_Basic_降核,Ascend950PR,batch_mat_mul_v3,1,1,1,1,1,"降核: 只用1核, 每核一个L0C满载输出块, 其余核闲置",0,1,1,128,256,256,1,标准核内流水,1,128,128,allocate,direct_gm,0,0,不涉及(每核一块无尾轮),1,1,1,0,0,0,0,True,2,"P=0.01<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢)",264192.0,0.0,5.28384e-06,0.0,5.28384e-06,0.0,0.0,262144.0,1.7260510288065844e-08,1024,2.048e-08,0.0,5.28384e-06,0.0,5.28384e-06,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b8_m2_n192_k128,8,8,2,192,128,bf16,bf16,bf16,False,False,False,True,0,b8_m2_n192_k128,ASW_Basic_降核,Ascend950PR,batch_mat_mul_v3,1,1,1,1,1,"降核: 只用1核, 每核一个L0C满载输出块, 其余核闲置",0,1,2,192,128,128,1,标准核内流水,2,192,80,allocate,direct_gm,0,0,不涉及(每核一块无尾轮),1,1,1,0,0,0,0,True,2,"P=0.05<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢)",397312.0,0.0,7.94624e-06,0.0,7.94624e-06,0.0,0.0,786432.0,5.178153086419753e-08,6144,1.2288e-07,0.0,7.94624e-06,0.0,7.94624e-06,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b16_m4_n256_k192,16,16,4,256,192,bf16,bf16,bf16,False,False,False,True,0,b16_m4_n256_k192,ASW_Basic_降核,Ascend950PR,batch_mat_mul_v3,1,1,1,1,1,"降核: 只用1核, 每核一个L0C满载输出块, 其余核闲置",0,1,4,256,192,192,1,标准核内流水,4,256,64,allocate,direct_gm,0,0,不涉及(每核一块无尾轮),1,1,1,0,0,0,0,True,2,"P=0.25<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢)",1597440.0,0.0,3.19488e-05,0.0,3.19488e-05,0.0,0.0,6291456.0,4.1425224691358024e-07,32768,6.5536e-07,0.0,3.19488e-05,0.0,3.19488e-05,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b32_m8_n128_k256,32,32,8,128,256,bf16,bf16,bf16,False,False,False,True,0,b32_m8_n128_k256,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),1,1,8,128,256,256,1,a_单batch全驻留,8,128,128,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,单batch全驻留: (MK+KN)*dtype=68KB <= L1,69632,0.0,1.39264e-06,0.0,1.44264e-06,1,5.0000000000000004e-08,524288.0,3.452102057613169e-08,2048,4.096e-08,0.0,1.44264e-06,7.548102057613169e-08,1.5181210205761316e-06,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b64_m16_n256_k512,64,64,16,256,512,bf16,bf16,bf16,False,False,False,True,0,b64_m16_n256_k512,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),2,1,16,256,512,240,1,c_一侧驻留+对侧切K,16,256,64,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"一侧驻留(A)+对侧切K: A驻留16KB, 预算L1/2, k_L1=240, dValueA=480B/dValueB=512B",557056,0.0,1.56672e-05,0.0,1.5967200000000002e-05,6,3.0000000000000004e-07,8388608.0,5.52336329218107e-07,16384,3.2768e-07,0.0,1.5967200000000002e-05,2.932938271604938e-07,1.6260493827160496e-05,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b128_m8_n192_k256,128,128,8,192,256,bf16,bf16,bf16,False,False,False,True,0,b128_m8_n192_k256,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),4,1,8,192,256,256,2,b_双batch乒乓,8,192,80,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,双batch乒乓: 2*(MK+KN)*dtype=200KB <= L1,409600,0.0,8.192e-06,0.0,8.392e-06,4,2.0000000000000002e-07,3145728.0,2.0712612345679012e-07,12288,2.4576e-07,0.0,8.392e-06,1.1322153086419754e-07,8.505221530864198e-06,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b32_m16_n8192_k7168,32,32,16,8192,7168,bf16,bf16,bf16,False,False,False,True,0,b32_m16_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,1,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,47,True,2,"L2场景C_输入超L2分组执行, r_in=1.09; 尾轮: r=0 无尾轮",4103077888.0000005,0.0,0.0025644236800000005,0.0,0.0025644236800000005,0.0,0.0,60129542144.0,0.00012372333774485596,8388608,5.24288e-06,0.0,0.0025644236800000005,0.0,0.0025644236800000005,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b4_m1_n8192_k8192,4,4,1,8192,8192,bf16,bf16,bf16,False,False,False,True,0,b4_m1_n8192_k8192,StreamK,Ascend950PR,batch_mat_mul_v3,32,1,1,1,32,"B/M/N切出4块, 每块32核切K归约 (归约组内核c负责K段[c*K/32,(c+1)*K/32))",1,1,1,8192,256,256,1,K段标准分块流水,1,128,64,allocate(部分和驻留L2),"resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换)",0,4194304,grid_K=32路切K+归约,1,1,32,0,0,0,0,True,4,"P=0.50, grid_K=32, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终",262144.0,0.0,5.24288e-06,0.0,5.24288e-06,0.0,0.0,33554432.0,2.209345316872428e-06,0.0,0.0,3.4067453613053613e-06,5.24288e-06,3.4067453613053613e-06,8.64962536130536e-06,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b16_m2_n4096_k7168,16,16,2,4096,7168,bf16,bf16,bf16,False,False,False,True,0,b16_m2_n4096_k7168,StreamK,Ascend950PR,batch_mat_mul_v3,32,1,1,1,16,"B/M/N切出16块, 每块16核切K归约 (归约组内核c负责K段[c*K/16,(c+1)*K/16))",1,1,2,4096,448,256,1,K段标准分块流水,2,128,64,allocate(部分和驻留L2),"resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换)",0,8388608,grid_K=16路切K+归约,1,1,16,0,0,0,0,True,4,"P=2.00, grid_K=16, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终",458752.0,0.0,9.17504e-06,0.0,9.17504e-06,0.0,0.0,58720256.0,3.866354304526749e-06,0.0,0.0,1.7159757575757577e-06,9.17504e-06,1.7159757575757577e-06,1.0891015757575759e-05,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b32_m64_n64_k7168,32,32,64,64,7168,bf16,bf16,bf16,False,False,False,True,0,b32_m64_n64_k7168,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),1,1,64,64,7168,1024,1,d_两侧都切K,64,64,256,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"两侧都切K: k_L1=1024, K段成对流水, batch边界天然无缝; dValueA=2048B/dValueB=128B",1835008,0.0,3.6700159999999995e-05,0.0,3.705016e-05,7,3.5000000000000004e-07,58720256.0,3.866354304526749e-06,8192,1.6384e-07,0.0,3.705016e-05,7.16176329218107e-07,3.77663363292181e-05,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
special_k0_demo,128,128,256,256,0,bf16,bf16,bf16,False,False,False,True,0,special_k0_demo,特殊分支,Ascend950PR,batch_mat_mul_v3,64,1,1,1,1,AIV 核间按行均分 (无 Cube tile 概念),0,1,0,0,0,0,1,UB驻留(AIV),0,0,0,allocate,direct_gm,0,0,不涉及(AIV逐元素),1,1,1,0,0,0,0,False,2,"K=0纯写值: 无任何计算, C=bias 或 0, 纯 AIV 写值; 按行均分到 AIV 核",0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,16777216,1.048576e-05,0.0,1.048576e-05,0.0,1.048576e-05,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
special_k1_demo,128,128,256,256,1,bf16,bf16,bf16,False,False,False,True,0,special_k1_demo,特殊分支,Ascend950PR,batch_mat_mul_v3,64,1,1,1,1,AIV 核间按行均分 (无 Cube tile 概念),0,1,0,0,1,0,1,UB驻留(AIV) UB乒乓,0,0,0,allocate,direct_gm,0,0,不涉及(AIV逐元素),1,1,1,0,0,0,0,False,2,"K=1逐元素乘: 退化为 C=A⊙B 无累加深度, Cube 16x16x16 粒度浪费 15/16; 走 AIV 通路 GM->UB->Mul->GM, UB乒乓 (B>=2*AIV 双batch乒乓流水)",131072,0.0,8.192e-08,0.0,8.192e-08,0.0,0.0,8388608.0,6.206060606060606e-07,16777216,1.048576e-05,0.0,1.0567679999999999e-05,0.0,1.0567679999999999e-05,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
merge_demo_k_trunc,2048,2048,32,32,256,bf16,bf16,bf16,False,False,False,True,0,merge_demo_k_trunc,MergeBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B均分(核间零重复读零依赖),64,4,128,128,256,256,8,合并驻留,128,128,128,allocate(GM->L1随路驻留L2),direct_gm,0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"b0=4 (L0C上限5.7/算存比上限19.0/b_core=64); K截断; 合并后单次DMA搬入 A'[128,256]+B'[256,128]",2097152,0.0,4.194304e-05,0.0,4.2743039999999997e-05,16.0,8.000000000000001e-07,134217728.0,8.837381267489712e-06,131072,2.62144e-06,0.0,4.5364479999999994e-05,3.0192408230452674e-07,4.566640408230452e-05,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
merge_iter_arbitrate,128,128,64,64,512,bf16,bf16,bf16,False,False,False,True,0,merge_iter_arbitrate,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),4,1,64,64,512,512,2,b_双batch乒乓,64,64,256,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,双batch乒乓: 2*(MK+KN)*dtype=256KB <= L1,524288,0.0,1.048576e-05,0.0,1.0685759999999999e-05,4,2.0000000000000002e-07,16777216.0,1.104672658436214e-06,32768,6.5536e-07,0.0,1.1341119999999999e-05,4.400081646090535e-07,1.1781128164609052e-05,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
iter_demo_form_b,128,128,64,64,256,bf16,bf16,bf16,False,False,False,True,0,iter_demo_form_b,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),4,1,64,64,256,256,2,b_双batch乒乓,64,64,256,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,双batch乒乓: 2*(MK+KN)*dtype=128KB <= L1,262144,0.0,5.24288e-06,0.0,5.4428799999999995e-06,4,2.0000000000000002e-07,8388608.0,5.52336329218107e-07,32768,6.5536e-07,0.0,6.09824e-06,3.0192408230452674e-07,6.400164082304526e-06,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
iter_demo_form_d,64,64,64,64,8192,bf16,bf16,bf16,False,False,False,True,0,iter_demo_form_d,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),2,1,64,64,8192,1024,1,d_两侧都切K,64,64,256,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"两侧都切K: k_L1=1024, K段成对流水, batch边界天然无缝; dValueA=2048B/dValueB=128B",4194304,0.0,8.388608e-05,0.0,8.468608e-05,16,8.000000000000001e-07,134217728.0,8.837381267489712e-06,16384,3.2768e-07,0.0,8.501376e-05,7.16176329218107e-07,8.572993632921812e-05,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
streamk_demo,4,4,128,128,10240,bf16,bf16,bf16,False,False,False,True,0,streamk_demo,StreamK,Ascend950PR,batch_mat_mul_v3,32,1,1,1,32,"B/M/N切出4块, 每块32核切K归约 (归约组内核c负责K段[c*K/32,(c+1)*K/32))",1,1,128,128,320,256,1,K段标准分块流水,128,128,64,allocate(部分和驻留L2),"resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换)",0,8388608,grid_K=32路切K+归约,1,1,32,0,0,0,0,True,4,"P=1.00, grid_K=32, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终",20971520,0,1.31072e-05,0.0,1.31072e-05,0.0,0.0,1342177280.0,2.761681646090535e-06,0.0,0.0,3.4067453613053613e-06,1.31072e-05,3.4067453613053613e-06,1.651394536130536e-05,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
asw_demo_full,2,2,8192,8192,1024,bf16,bf16,bf16,False,False,False,True,0,asw_demo_full,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,1024,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,52,52,1,52,52,2,139,True,2,"L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: 周长型主导, rho=0.06<rho_dv=0.53, A1b被dValue卡死, 方案B反超",67108864.0,3087007744,4.194304e-05,0.0005936553353846154,0.0006355983753846153,0.0,0.0,274877906944.0,0.0005655924011193416,268435456,0.00016777216,0.0,0.0008033705353846154,0.0,0.0008033705353846154,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
asw_demo_reduce_core,16,16,256,256,128,bf16,bf16,bf16,False,False,False,True,0,asw_demo_reduce_core,ASW_Basic_降核,Ascend950PR,batch_mat_mul_v3,16,1,1,1,1,"降核: 只用16核, 每核一个L0C满载输出块, 其余核闲置",0,1,256,256,128,128,1,标准核内流水,256,256,64,allocate,direct_gm,0,0,不涉及(每核一块无尾轮),1,1,1,0,0,0,0,True,2,"P=16.00<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢)",2097152.0,0,2.62144e-06,0.0,2.62144e-06,0.0,0.0,268435456.0,1.104672658436214e-06,2097152,8.065969230769231e-07,0.0,2.62144e-06,0.0,2.62144e-06,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b4_m1_n128_k256,4,4,1,128,256,bf16,bf16,bf16,False,False,False,True,0,b4_m1_n128_k256,ASW_Basic_降核,Ascend950PR,batch_mat_mul_v3,1,1,1,1,1,"降核: 只用1核, 每核一个L0C满载输出块, 其余核闲置",0,1,1,128,256,256,1,标准核内流水,1,128,128,allocate,direct_gm,0,0,不涉及(每核一块无尾轮),1,1,1,0,0,0,0,True,2,"P=0.01<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢)",264192.0,0,5.28384e-06,0.0,5.28384e-06,0.0,0.0,262144.0,1.7260510288065844e-08,1024,6.3015384615384615e-09,0.0,5.28384e-06,0.0,5.28384e-06,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b8_m2_n192_k128,8,8,2,192,128,bf16,bf16,bf16,False,False,False,True,0,b8_m2_n192_k128,ASW_Basic_降核,Ascend950PR,batch_mat_mul_v3,1,1,1,1,1,"降核: 只用1核, 每核一个L0C满载输出块, 其余核闲置",0,1,2,192,128,128,1,标准核内流水,2,192,80,allocate,direct_gm,0,0,不涉及(每核一块无尾轮),1,1,1,0,0,0,0,True,2,"P=0.05<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢)",397312.0,0,7.94624e-06,0.0,7.94624e-06,0.0,0.0,786432.0,5.178153086419753e-08,6144,3.7809230769230766e-08,0.0,7.94624e-06,0.0,7.94624e-06,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b16_m4_n256_k192,16,16,4,256,192,bf16,bf16,bf16,False,False,False,True,0,b16_m4_n256_k192,ASW_Basic_降核,Ascend950PR,batch_mat_mul_v3,1,1,1,1,1,"降核: 只用1核, 每核一个L0C满载输出块, 其余核闲置",0,1,4,256,192,192,1,标准核内流水,4,256,64,allocate,direct_gm,0,0,不涉及(每核一块无尾轮),1,1,1,0,0,0,0,True,2,"P=0.25<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢)",1597440.0,0,3.19488e-05,0.0,3.19488e-05,0.0,0.0,6291456.0,4.1425224691358024e-07,32768,2.0164923076923077e-07,0.0,3.19488e-05,0.0,3.19488e-05,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b32_m8_n128_k256,32,32,8,128,256,bf16,bf16,bf16,False,False,False,True,0,b32_m8_n128_k256,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),1,1,8,128,256,256,1,a_单batch全驻留,8,128,128,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,单batch全驻留: (MK+KN)*dtype=68KB <= L1,69632,0.0,1.39264e-06,0.0,1.44264e-06,1,5.0000000000000004e-08,524288.0,3.452102057613169e-08,2048,4.096e-08,0.0,1.4836e-06,7.548102057613169e-08,1.5590810205761317e-06,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b64_m16_n256_k512,64,64,16,256,512,bf16,bf16,bf16,False,False,False,True,0,b64_m16_n256_k512,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),2,1,16,256,512,240,1,c_一侧驻留+对侧切K,16,256,64,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"一侧驻留(A)+对侧切K: A驻留16KB, 预算L1/2, k_L1=240, dValueA=480B/dValueB=512B",557056,0.0,1.56672e-05,0.0,1.5967200000000002e-05,6,3.0000000000000004e-07,8388608.0,5.52336329218107e-07,16384,3.2768e-07,0.0,1.629488e-05,2.932938271604938e-07,1.6588173827160495e-05,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b128_m8_n192_k256,128,128,8,192,256,bf16,bf16,bf16,False,False,False,True,0,b128_m8_n192_k256,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),4,1,8,192,256,256,2,b_双batch乒乓,8,192,80,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,双batch乒乓: 2*(MK+KN)*dtype=200KB <= L1,409600,0.0,8.192e-06,0.0,8.392e-06,4,2.0000000000000002e-07,3145728.0,2.0712612345679012e-07,12288,2.4576e-07,0.0,8.637760000000001e-06,1.1322153086419754e-07,8.7509815308642e-06,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b32_m16_n8192_k7168,32,32,16,8192,7168,bf16,bf16,bf16,False,False,False,True,0,b32_m16_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,1,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,47,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮",3765436416.0,337641472,0.00235339776,6.493105230769231e-05,0.0024183288123076924,0.0,0.0,60129542144.0,0.00012372333774485596,8388608,1.6131938461538462e-06,0.0,0.0024183288123076924,0.0,0.0024183288123076924,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b4_m1_n8192_k8192,4,4,1,8192,8192,bf16,bf16,bf16,False,False,False,True,0,b4_m1_n8192_k8192,StreamK,Ascend950PR,batch_mat_mul_v3,32,1,1,1,32,"B/M/N切出4块, 每块32核切K归约 (归约组内核c负责K段[c*K/32,(c+1)*K/32))",1,1,1,8192,256,256,1,K段标准分块流水,1,128,64,allocate(部分和驻留L2),"resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换)",0,4194304,grid_K=32路切K+归约,1,1,32,0,0,0,0,True,4,"P=0.50, grid_K=32, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终",536936448,0.0,0.00033558528,0.0,0.00033558528,0.0,0.0,536870912.0,1.104672658436214e-06,0.0,0.0,1.7033726806526807e-06,0.00033558528,1.7033726806526807e-06,0.00033728865268065273,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b16_m2_n4096_k7168,16,16,2,4096,7168,bf16,bf16,bf16,False,False,False,True,0,b16_m2_n4096_k7168,StreamK,Ascend950PR,batch_mat_mul_v3,32,1,1,1,16,"B/M/N切出16块, 每块16核切K归约 (归约组内核c负责K段[c*K/16,(c+1)*K/16))",1,1,2,4096,448,256,1,K段标准分块流水,2,128,64,allocate(部分和驻留L2),"resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换)",0,8388608,grid_K=16路切K+归约,1,1,16,0,0,0,0,True,4,"P=2.00, grid_K=16, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终",939982848,0,0.00058748928,0.0,0.00058748928,0.0,0.0,1879048192.0,3.866354304526749e-06,0.0,0.0,1.7159757575757577e-06,0.00058748928,1.7159757575757577e-06,0.0005892052557575758,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b32_m64_n64_k7168,32,32,64,64,7168,bf16,bf16,bf16,False,False,False,True,0,b32_m64_n64_k7168,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),1,1,64,64,7168,1024,1,d_两侧都切K,64,64,256,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"两侧都切K: k_L1=1024, K段成对流水, batch边界天然无缝; dValueA=2048B/dValueB=128B",1835008,0.0,3.6700159999999995e-05,0.0,3.705016e-05,7,3.5000000000000004e-07,58720256.0,3.866354304526749e-06,8192,1.6384e-07,0.0,3.7214e-05,7.16176329218107e-07,3.7930176329218104e-05,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b64_m1024_n1024_k7168,64,64,1024,1024,7168,bf16,bf16,bf16,False,False,False,True,0,b64_m1024_n1024_k7168,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),2,1,1024,1024,7168,64,1,d_两侧都切K,176,176,64,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"两侧都切K: k_L1=64, K段成对流水, batch边界天然无缝; dValueA=128B/dValueB=2048B",58720256,0.0,0.0011744051199999998,0.0,0.00118560512,224,1.1200000000000001e-05,30064771072.0,0.0019795734039176954,4194304,8.388608e-05,0.0,0.0019795734039176954,5.078042126748971e-05,0.0020303538251851853,MMAD,True,,计算Bound,"瓶颈在 Cube 计算: 已接近理论算力上限, 检查是否有冗余计算 (MergeBatch 交叉项) 可消除"
b128_m2048_n2048_k1536,128,128,2048,2048,1536,bf16,bf16,bf16,False,False,False,True,0,b128_m2048_n2048_k1536,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,12,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,1536,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,576,True,2,"L2场景C_输入超L2分组执行, r_in=12.00; 尾轮: r=0 无尾轮",19327352832.0,0.0,0.01207959552,0.0,0.01207959552,0.0,0.0,1649267441664.0,0.0033935544067160493,1073741824,0.00067108864,0.0,0.01207959552,0.0,0.01207959552,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b64_m1024_n8192_k2048,64,64,1024,8192,2048,bf16,bf16,bf16,False,False,False,True,0,b64_m1024_n8192_k2048,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,6,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,2048,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,564,True,2,"L2场景C_输入超L2分组执行, r_in=10.56; 尾轮: r=0 无尾轮",25501368320.0,0.0,0.0159383552,0.0,0.0159383552,0.0,0.0,2199023255552.0,0.004524739208954733,1073741824,0.00067108864,0.0,0.0159383552,0.0,0.0159383552,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b32_m1024_n1024_k512,32,32,1024,1024,512,bf16,bf16,bf16,False,False,False,True,0,b32_m1024_n1024_k512,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),1,1,1024,1024,512,64,1,d_两侧都切K,176,176,64,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"两侧都切K: k_L1=64, K段成对流水, batch边界天然无缝; dValueA=128B/dValueB=2048B",2097152,0.0,4.194304e-05,0.0,4.234304e-05,8,4.0000000000000003e-07,1073741824.0,7.06990501399177e-05,2097152,4.194304e-05,0.0,7.06990501399177e-05,5.078042126748971e-05,0.0001214794714074074,MMAD,True,,计算Bound,"瓶颈在 Cube 计算: 已接近理论算力上限, 检查是否有冗余计算 (MergeBatch 交叉项) 可消除"
b128_m4096_n4096_k8192,128,128,4096,4096,8192,bf16,bf16,bf16,False,False,False,True,0,b128_m4096_n4096_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,24,24,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,2304,True,2,"L2场景C_输入超L2分组执行, r_in=24.00; 尾轮: r=0 无尾轮",412316860416.0,0.0,0.25769803776,0.0,0.25769803776,0.0,0.0,35184372088832.0,0.07239582734327572,4294967296,0.00268435456,0.0,0.25769803776,0.0,0.25769803776,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b32_m8192_n4096_k7168,32,32,8192,4096,7168,bf16,bf16,bf16,False,False,False,True,0,b32_m8192_n4096_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,24,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,1128,True,2,"L2场景C_输入超L2分组执行, r_in=31.67; 尾轮: r=0 无尾轮",178509578240.0,0.0,0.1115684864,0.0,0.1115684864,0.0,0.0,15393162788864.0,0.031673174462683126,2147483648,0.00134217728,0.0,0.1115684864,0.0,0.1115684864,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b32_m2048_n2048_k8192,32,32,2048,2048,8192,bf16,bf16,bf16,False,False,False,True,0,b32_m2048_n2048_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,12,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,144,True,2,"L2场景C_输入超L2分组执行, r_in=12.00; 尾轮: r=0 无尾轮",25769803776.0,0.0,0.01610612736,0.0,0.01610612736,0.0,0.0,2199023255552.0,0.004524739208954733,268435456,0.00016777216,0.0,0.01610612736,0.0,0.01610612736,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b64_m4096_n2048_k128,64,64,4096,2048,128,bf16,bf16,bf16,False,False,False,True,0,b64_m4096_n2048_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,24,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,576,True,2,"L2场景B_输入驻留输出直写GM, r_in=1.00; 尾轮: r=0 无尾轮",100663296.0,0.0,6.291456e-05,0.0,6.291456e-05,0.0,0.0,137438953472.0,0.0002827962005596708,1073741824,0.00067108864,0.0,0.00067108864,0.0,0.00067108864,FIXPIPE,True,,写出Bound,"瓶颈在 Fixpipe 写出: 检查输出 dtype (fp16/fp8 可减半写出量), 或评估输出驻留 L2 异步回写策略"
b16_m1024_n1024_k8192,16,16,1024,1024,8192,bf16,bf16,bf16,False,False,False,True,0,b16_m1024_n1024_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,6,6,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,18,True,2,"L2场景C_输入超L2分组执行, r_in=6.00; 尾轮: r=0 无尾轮",3221225472.0,0.0,0.00201326592,0.0,0.00201326592,0.0,0.0,274877906944.0,0.0005655924011193416,33554432,2.097152e-05,0.0,0.00201326592,0.0,0.00201326592,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b4_m32768_n128_k128,4,4,32768,128,128,bf16,bf16,bf16,False,False,False,True,0,b4_m32768_n128_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,1,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,方案B,205,1,1,205,1,12,24,True,2,"L2场景A_全驻留, r_in=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=28.96us), 选方案B工程简洁 (一套tile)",33685504.0,0.0,2.105344e-05,0.0,2.105344e-05,0.0,0.0,4294967296.0,8.837381267489712e-06,33554432,6.452775384615385e-06,0.0,2.105344e-05,0.0,2.105344e-05,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b8_m32768_n2048_k512,8,8,32768,2048,512,bf16,bf16,bf16,False,False,False,True,0,b8_m32768_n2048_k512,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,512,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,561,True,2,"L2场景C_输入超L2分组执行, r_in=1.24; 尾轮: r=0 无尾轮",352321536.0,0.0,0.00022020096,0.0,0.00022020096,0.0,0.0,549755813888.0,0.0011311848022386832,1073741824,0.00067108864,0.0,0.0011311848022386832,0.0,0.0011311848022386832,MMAD,True,,计算Bound,"瓶颈在 Cube 计算: 已接近理论算力上限, 检查是否有冗余计算 (MergeBatch 交叉项) 可消除"
b4_m131072_n128_k128,4,4,131072,128,128,bf16,bf16,bf16,False,False,False,True,0,b4_m131072_n128_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,1,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,820,1,1,820,1,4,94,True,2,"L2场景C_输入超L2分组执行, r_in=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=115.39us), 选方案B工程简洁 (一套tile)",134610944.0,0.0,8.413184e-05,0.0,8.413184e-05,0.0,0.0,17179869184.0,3.534952506995885e-05,134217728,8.388608e-05,0.0,8.413184e-05,0.0,8.413184e-05,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b8_m131072_n1024_k256,8,8,131072,1024,256,bf16,bf16,bf16,False,False,False,True,0,b8_m131072_n1024_k256,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,6,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,256,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,820,7,1,820,7,16,1118,True,2,"L2场景C_输入超L2分组执行, r_in=1.06; 尾轮: 面积型主导, A1b与方案B严格打平(T=1384.63us), 选方案B工程简洁 (一套tile)",574619648.0,0.0,0.00035913728,0.0,0.00035913728,0.0,0.0,549755813888.0,0.0011311848022386832,2147483648,0.00134217728,0.0,0.00134217728,0.0,0.00134217728,FIXPIPE,True,,写出Bound,"瓶颈在 Fixpipe 写出: 检查输出 dtype (fp16/fp8 可减半写出量), 或评估输出驻留 L2 异步回写策略"
b16_m32768_n8192_k7168,16,16,32768,8192,7168,bf16,bf16,bf16,False,False,False,True,0,b16_m32768_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,205,52,1,205,52,16,4395,True,2,"L2场景C_输入超L2分组执行, r_in=75.00; 尾轮: 周长型主导, rho=0.50<rho_dv=0.53, A1b被dValue卡死, 方案B反超",704643072000.0,0.0,0.44040192,0.0,0.44040192,0.0,0.0,61572651155456.0,0.1266926978507325,8589934592,0.00536870912,0.0,0.44040192,0.0,0.44040192,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b4_m32768_n128_k8192,4,4,32768,128,8192,bf16,bf16,bf16,False,False,False,True,0,b4_m32768_n128_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,1,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,205,1,1,205,1,12,24,True,2,"L2场景C_输入超L2分组执行, r_in=1.14; 尾轮: 周长型主导, rho=0.38<rho_dv=0.53, A1b被dValue卡死, 方案B反超",2466250752.0,0.0,0.00154140672,0.0,0.00154140672,0.0,0.0,274877906944.0,0.0005655924011193416,33554432,2.097152e-05,0.0,0.00154140672,0.0,0.00154140672,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b32_m131072_n8192_k128,32,32,131072,8192,128,bf16,bf16,bf16,False,False,False,True,0,b32_m131072_n8192_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,35015,True,2,"L2场景C_输入超L2分组执行, r_in=2.88; 尾轮: r=0 无尾轮",3288334336.0,0.0,0.00205520896,0.0,0.00205520896,0.0,0.0,8796093022208.0,0.01809895683581893,68719476736,0.04294967296,0.0,0.04294967296,0.0,0.04294967296,FIXPIPE,True,,写出Bound,"瓶颈在 Fixpipe 写出: 检查输出 dtype (fp16/fp8 可减半写出量), 或评估输出驻留 L2 异步回写策略"
b64_m32768_n8192_k1536,64,64,32768,8192,1536,bf16,bf16,bf16,False,False,False,True,0,b64_m32768_n8192_k1536,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,1536,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,17578,True,2,"L2场景C_输入超L2分组执行, r_in=75.00; 尾轮: r=0 无尾轮",603979776000.0,0.0,0.37748736,0.0,0.37748736,0.0,0.0,52776558133248.0,0.10859374101491358,34359738368,0.02147483648,0.0,0.37748736,0.0,0.37748736,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b8_m131072_n8192_k8192,8,8,131072,8192,8192,bf16,bf16,bf16,False,False,False,True,0,b8_m131072_n8192_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A1b,2,2,1,2,2,24,8754,True,2,"L2场景C_输入超L2分组执行, r_in=44.53; 尾轮: 周长型主导, rho=0.75>=rho_dv=0.53, A1b恒优 (尾轮tile缩√rho=0.87倍凑满核)",812822560768.0,0.0,0.50801410048,0.0,0.50801410048,0.0,0.0,140737488355328.0,0.2895833093731029,17179869184,0.01073741824,0.0,0.50801410048,0.0,0.50801410048,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b8_m16_n7168_k1536,8,8,16,7168,1536,bf16,bf16,bf16,False,False,False,True,0,b8_m16_n7168_k1536,StreamK,Ascend950PR,batch_mat_mul_v3,32,1,1,2,2,"B/M/N切出16块, 每块2核切K归约 (归约组内核c负责K段[c*K/2,(c+1)*K/2))",1,1,16,3584,768,256,1,K段标准分块流水,16,128,64,allocate(部分和驻留L2),"resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换)",0,3670016,grid_K=2路切K+归约,1,1,2,0,0,0,0,True,4,"P=14.00, grid_K=2, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终",786432.0,0.0,1.572864e-05,0.0,1.572864e-05,0.0,0.0,100663296.0,6.628035950617284e-06,0.0,0.0,2.3655235431235433e-07,1.572864e-05,2.3655235431235433e-07,1.5965192354312353e-05,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b64_m1024_n7168_k7168,64,64,1024,7168,7168,bf16,bf16,bf16,False,False,False,True,0,b64_m1024_n7168_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,6,41,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,492,True,2,"L2场景C_输入超L2分组执行, r_in=10.38; 尾轮: r=0 无尾轮",77980499968.0,0.0,0.04873781248,0.0,0.04873781248,0.0,0.0,6734508720128.0,0.013857013827423869,939524096,0.00058720256,0.0,0.04873781248,0.0,0.04873781248,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b32_m8192_n8192_k7168,32,32,8192,8192,7168,bf16,bf16,bf16,False,False,False,True,0,b32_m8192_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,2209,True,2,"L2场景C_输入超L2分组执行, r_in=47.00; 尾轮: r=0 无尾轮",353261060096.0,0.0,0.22078816256,0.0,0.22078816256,0.0,0.0,30786325577728.0,0.06334634892536625,4294967296,0.00268435456,0.0,0.22078816256,0.0,0.22078816256,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
b8_m4096_n4096_k128,8,8,4096,4096,128,bf16,bf16,bf16,False,False,False,True,0,b8_m4096_n4096_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,24,24,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,144,True,2,"L2场景B_输入驻留输出直写GM, r_in=1.00; 尾轮: r=0 无尾轮",16777216.0,0.0,1.048576e-05,0.0,1.048576e-05,0.0,0.0,34359738368.0,7.06990501399177e-05,268435456,0.00016777216,0.0,0.00016777216,0.0,0.00016777216,FIXPIPE,True,,写出Bound,"瓶颈在 Fixpipe 写出: 检查输出 dtype (fp16/fp8 可减半写出量), 或评估输出驻留 L2 异步回写策略"
b128_m8192_n8192_k7168,128,128,8192,8192,7168,bf16,bf16,bf16,False,False,False,True,0,b128_m8192_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,8836,True,2,"L2场景C_输入超L2分组执行, r_in=47.00; 尾轮: r=0 无尾轮",1413044240384.0,0.0,0.88315265024,0.0,0.88315265024,0.0,0.0,123145302310912.0,0.253385395701465,17179869184,0.01073741824,0.0,0.88315265024,0.0,0.88315265024,MTE2_GM,True,,访存Bound(GM),"瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向)"
special_k1_b64,64,64,8192,512,1,bf16,bf16,bf16,False,False,False,True,0,special_k1_b64,特殊分支,Ascend950PR,batch_mat_mul_v3,64,1,1,1,1,AIV 核间按行均分 (无 Cube tile 概念),0,1,0,0,1,0,1,UB驻留(AIV) AIV单缓冲,0,0,0,allocate,direct_gm,0,0,不涉及(AIV逐元素),1,1,1,0,0,0,0,False,2,"K=1逐元素乘: 退化为 C=A⊙B 无累加深度, Cube 16x16x16 粒度浪费 15/16; 走 AIV 通路 GM->UB->Mul->GM, AIV单缓冲 (B<2*AIV 逐batch单缓冲串行)",1114112,0.0,6.9632e-07,0.0,6.9632e-07,0.0,0.0,268435456.0,1.985939393939394e-05,536870912,0.00033554432,0.0,0.00033554432,0.0,0.00033554432,FIXPIPE,True,,写出Bound,"瓶颈在 Fixpipe 写出: 检查输出 dtype (fp16/fp8 可减半写出量), 或评估输出驻留 L2 异步回写策略"
b128_m2048_n2048_k1536,128,128,2048,2048,1536,bf16,bf16,bf16,False,False,False,True,0,b128_m2048_n2048_k1536,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,12,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,1536,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,576,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮",1610612736.0,17716740096,0.00100663296,0.003407065403076923,0.004413698363076923,0.0,0.0,1649267441664.0,0.0033935544067160493,1073741824,0.0002064888123076923,0.0,0.004413698363076923,0.0,0.004413698363076923,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b64_m1024_n8192_k2048,64,64,1024,8192,2048,bf16,bf16,bf16,False,False,False,True,0,b64_m1024_n8192_k2048,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,6,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,2048,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,564,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮",2415919104.0,23085449216,0.00150994944,0.004439509464615385,0.005949458904615385,0.0,0.0,2199023255552.0,0.004524739208954733,1073741824,0.0002064888123076923,0.0,0.005949458904615385,0.0,0.005949458904615385,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b32_m1024_n1024_k512,32,32,1024,1024,512,bf16,bf16,bf16,False,False,False,True,0,b32_m1024_n1024_k512,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),1,1,1024,1024,512,64,1,d_两侧都切K,176,176,64,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"两侧都切K: k_L1=64, K段成对流水, batch边界天然无缝; dValueA=128B/dValueB=2048B",2097152,0.0,4.194304e-05,0.0,4.234304e-05,8,4.0000000000000003e-07,1073741824.0,7.06990501399177e-05,2097152,4.194304e-05,0.0,8.428607999999999e-05,5.078042126748971e-05,0.00013506650126748969,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b128_m4096_n4096_k8192,128,128,4096,4096,8192,bf16,bf16,bf16,False,False,False,True,0,b128_m4096_n4096_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,24,24,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,2304,True,2,"L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮",17179869184.0,395136991232,0.01073741824,0.07598788292923077,0.08672530116923077,0.0,0.0,35184372088832.0,0.07239582734327572,4294967296,0.00268435456,0.0,0.08940965572923076,0.0,0.08940965572923076,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b32_m8192_n4096_k7168,32,32,8192,4096,7168,bf16,bf16,bf16,False,False,False,True,0,b32_m8192_n4096_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,24,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,1128,True,2,"L2场景C_单batch输入超L2分组执行, GM首读倍率=1.33; 尾轮: r=0 无尾轮",7516192768.0,0.0,0.00469762048,0.0,0.00469762048,0.0,0.0,15393162788864.0,0.031673174462683126,2147483648,0.00134217728,0.0,0.031673174462683126,0.0,0.031673174462683126,MMAD,True,,计算Bound,"瓶颈在 Cube 计算: 已接近理论算力上限, 检查是否有冗余计算 (MergeBatch 交叉项) 可消除"
b32_m2048_n2048_k8192,32,32,2048,2048,8192,bf16,bf16,bf16,False,False,False,True,0,b32_m2048_n2048_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,12,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,144,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮",2147483648.0,23622320128,0.00134217728,0.004542753870769231,0.005884931150769231,0.0,0.0,2199023255552.0,0.004524739208954733,268435456,5.162220307692308e-05,0.0,0.005884931150769231,0.0,0.005884931150769231,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b64_m4096_n2048_k128,64,64,4096,2048,128,bf16,bf16,bf16,False,False,False,True,0,b64_m4096_n2048_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,24,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,576,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮",100663296.0,1509949440,6.291456e-05,0.0002903748923076923,0.0003532894523076923,0.0,0.0,137438953472.0,0.0002827962005596708,1073741824,0.0002064888123076923,0.0,0.0003532894523076923,0.0,0.0003532894523076923,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b16_m1024_n1024_k8192,16,16,1024,1024,8192,bf16,bf16,bf16,False,False,False,True,0,b16_m1024_n1024_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,6,6,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,18,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮",536870912.0,2684354560,0.00033554432,0.0005162220307692308,0.0008517663507692308,0.0,0.0,274877906944.0,0.0005655924011193416,33554432,6.452775384615385e-06,0.0,0.0008517663507692308,0.0,0.0008517663507692308,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b4_m32768_n128_k128,4,4,32768,128,128,bf16,bf16,bf16,False,False,False,True,0,b4_m32768_n128_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,1,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,方案B,205,1,1,205,1,12,24,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=28.96us), 选方案B工程简洁 (一套tile)",33685504.0,24379392,2.105344e-05,4.688344615384615e-06,2.5741784615384616e-05,0.0,0.0,4294967296.0,8.837381267489712e-06,33554432,6.452775384615385e-06,0.0,2.5741784615384616e-05,0.0,2.5741784615384616e-05,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b8_m32768_n2048_k512,8,8,32768,2048,512,bf16,bf16,bf16,False,False,False,True,0,b8_m32768_n2048_k512,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,512,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,561,True,2,"L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮",285212672.0,6073352192,0.00017825792,0.0011679523446153847,0.0013462102646153848,0.0,0.0,549755813888.0,0.0011311848022386832,1073741824,0.00067108864,0.0,0.0020172989046153846,0.0,0.0020172989046153846,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b4_m131072_n128_k128,4,4,131072,128,128,bf16,bf16,bf16,False,False,False,True,0,b4_m131072_n128_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,1,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,方案B,820,1,1,820,1,4,94,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=115.39us), 选方案B工程简洁 (一套tile)",134348800.0,97517568,8.3968e-05,1.875337846153846e-05,0.00010272137846153847,0.0,0.0,17179869184.0,3.534952506995885e-05,134217728,2.581110153846154e-05,0.0,0.00010272137846153847,0.0,0.00010272137846153847,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b8_m131072_n1024_k256,8,8,131072,1024,256,bf16,bf16,bf16,False,False,False,True,0,b8_m131072_n1024_k256,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,6,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,256,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,820,7,1,820,7,16,1118,True,2,"L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=1384.63us), 选方案B工程简洁 (一套tile)",541065216.0,5804916736,0.00033816576,0.0011163301415384615,0.0014544959015384616,0.0,0.0,549755813888.0,0.0011311848022386832,2147483648,0.00134217728,0.0,0.0027966731815384617,0.0,0.0027966731815384617,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b16_m32768_n8192_k7168,16,16,32768,8192,7168,bf16,bf16,bf16,False,False,False,True,0,b16_m32768_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,205,52,1,205,52,16,4395,True,2,"L2场景C_单batch输入超L2分组执行, GM首读倍率=3.20; 尾轮: 周长型主导, rho=0.50<rho_dv=0.53, A1b被dValue卡死, 方案B反超",30064771072.0,0.0,0.01879048192,0.0,0.01879048192,0.0,0.0,61572651155456.0,0.1266926978507325,8589934592,0.00536870912,0.0,0.1266926978507325,0.0,0.1266926978507325,MMAD,True,,计算Bound,"瓶颈在 Cube 计算: 已接近理论算力上限, 检查是否有冗余计算 (MergeBatch 交叉项) 可消除"
b4_m32768_n128_k8192,4,4,32768,128,8192,bf16,bf16,bf16,False,False,False,True,0,b4_m32768_n128_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,1,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,205,1,1,205,1,12,24,True,2,"L2场景C_单batch输入超L2分组执行, GM首读倍率=1.03; 尾轮: 周长型主导, rho=0.38<rho_dv=0.53, A1b被dValue卡死, 方案B反超",2222981120.0,0.0,0.0013893632,0.0,0.0013893632,0.0,0.0,274877906944.0,0.0005655924011193416,33554432,2.097152e-05,0.0,0.00141033472,0.0,0.00141033472,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b32_m131072_n8192_k128,32,32,131072,8192,128,bf16,bf16,bf16,False,False,False,True,0,b32_m131072_n8192_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,35015,True,2,"L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮",1140850688.0,99321118720,0.00071303168,0.019100215138461538,0.019813246818461538,0.0,0.0,8796093022208.0,0.01809895683581893,68719476736,0.04294967296,0.0,0.06276291977846153,0.0,0.06276291977846153,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b64_m32768_n8192_k1536,64,64,32768,8192,1536,bf16,bf16,bf16,False,False,False,True,0,b64_m32768_n8192_k1536,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,1536,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,17578,True,2,"L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮",8053063680.0,595926712320,0.0050331648,0.11460129083076923,0.11963445563076923,0.0,0.0,52776558133248.0,0.10859374101491358,34359738368,0.02147483648,0.0,0.14110929211076922,0.0,0.14110929211076922,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b8_m131072_n8192_k8192,8,8,131072,8192,8192,bf16,bf16,bf16,False,False,False,True,0,b8_m131072_n8192_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A1b,2,2,1,2,2,24,8754,True,2,"L2场景C_单batch输入超L2分组执行, GM首读倍率=4.76; 尾轮: 周长型主导, rho=0.75>=rho_dv=0.53, A1b恒优 (尾轮tile缩√rho=0.87倍凑满核)",86973087744.0,0.0,0.05435817984,0.0,0.05435817984,0.0,0.0,140737488355328.0,0.2895833093731029,17179869184,0.01073741824,0.0,0.2895833093731029,0.0,0.2895833093731029,MMAD,True,,计算Bound,"瓶颈在 Cube 计算: 已接近理论算力上限, 检查是否有冗余计算 (MergeBatch 交叉项) 可消除"
b8_m16_n7168_k1536,8,8,16,7168,1536,bf16,bf16,bf16,False,False,False,True,0,b8_m16_n7168_k1536,StreamK,Ascend950PR,batch_mat_mul_v3,32,1,1,2,2,"B/M/N切出16块, 每块2核切K归约 (归约组内核c负责K段[c*K/2,(c+1)*K/2))",1,1,16,3584,768,256,1,K段标准分块流水,16,128,64,allocate(部分和驻留L2),"resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换)",0,3670016,grid_K=2路切K+归约,1,1,2,0,0,0,0,True,4,"P=14.00, grid_K=2, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终",176553984,393216,0.00011034624,7.561846153846153e-08,0.00011042185846153846,0.0,0.0,2818572288.0,5.799531456790123e-06,0.0,0.0,2.0698331002331e-07,0.00011042185846153846,2.0698331002331e-07,0.00011062884177156177,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b64_m1024_n7168_k7168,64,64,1024,7168,7168,bf16,bf16,bf16,False,False,False,True,0,b64_m1024_n7168_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,6,41,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,492,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮",7516192768.0,70464307200,0.00469762048,0.013550828307692308,0.018248448787692308,0.0,0.0,6734508720128.0,0.013857013827423869,939524096,0.00018067771076923076,0.0,0.018248448787692308,0.0,0.018248448787692308,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b32_m8192_n8192_k7168,32,32,8192,8192,7168,bf16,bf16,bf16,False,False,False,True,0,b32_m8192_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,2209,True,2,"L2场景C_单batch输入超L2分组执行, GM首读倍率=2.00; 尾轮: r=0 无尾轮",15032385536.0,0.0,0.00939524096,0.0,0.00939524096,0.0,0.0,30786325577728.0,0.06334634892536625,4294967296,0.00268435456,0.0,0.06334634892536625,0.0,0.06334634892536625,MMAD,True,,计算Bound,"瓶颈在 Cube 计算: 已接近理论算力上限, 检查是否有冗余计算 (MergeBatch 交叉项) 可消除"
b8_m4096_n4096_k128,8,8,4096,4096,128,bf16,bf16,bf16,False,False,False,True,0,b8_m4096_n4096_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,24,24,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,144,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮",16777216.0,385875968,1.048576e-05,7.420691692307692e-05,8.469267692307693e-05,0.0,0.0,34359738368.0,7.06990501399177e-05,268435456,5.162220307692308e-05,0.0,8.469267692307693e-05,0.0,8.469267692307693e-05,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
b128_m8192_n8192_k7168,128,128,8192,8192,7168,bf16,bf16,bf16,False,False,False,True,0,b128_m8192_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,8836,True,2,"L2场景C_单batch输入超L2分组执行, GM首读倍率=2.00; 尾轮: r=0 无尾轮",60129542144.0,0.0,0.03758096384,0.0,0.03758096384,0.0,0.0,123145302310912.0,0.253385395701465,17179869184,0.01073741824,0.0,0.253385395701465,0.0,0.253385395701465,MMAD,True,,计算Bound,"瓶颈在 Cube 计算: 已接近理论算力上限, 检查是否有冗余计算 (MergeBatch 交叉项) 可消除"
special_k1_b64,64,64,8192,512,1,bf16,bf16,bf16,False,False,False,True,0,special_k1_b64,特殊分支,Ascend950PR,batch_mat_mul_v3,64,1,1,1,1,AIV 核间按行均分 (无 Cube tile 概念),0,1,0,0,1,0,1,UB驻留(AIV) AIV单缓冲,0,0,0,allocate,direct_gm,0,0,不涉及(AIV逐元素),1,1,1,0,0,0,0,False,2,"K=1逐元素乘: 退化为 C=A⊙B 无累加深度, Cube 16x16x16 粒度浪费 15/16; 走 AIV 通路 GM->UB->Mul->GM, AIV单缓冲 (B<2*AIV 逐batch单缓冲串行)",1114112,0.0,6.9632e-07,0.0,6.9632e-07,0.0,0.0,268435456.0,1.985939393939394e-05,536870912,0.00033554432,0.0,0.00033624063999999996,0.0,0.00033624063999999996,MTE2,True,,访存Bound(GM读写共享+L2重复读),"瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争"
1 case_id batch_a batch_b m n k dtype_a dtype_b dtype_c trans_a trans_b has_bias out_nd deterministic_level plan_case_id plan_branch plan_npu plan_op plan_used_core_num plan_split_b plan_m_cnt plan_n_cnt plan_grid_k plan_core_map plan_b_core plan_merge_b0 plan_single_core_m plan_single_core_n plan_single_core_k plan_k_l1 plan_b_l1 plan_l1_form plan_base_m plan_base_n plan_base_k plan_l2_policy_in plan_l2_policy_out plan_swizzle_w plan_workspace_bytes plan_tail_strategy plan_tail_m_cnt plan_tail_n_cnt plan_tail_k_cnt plan_tail_m_main plan_tail_n_main plan_tail_block_cnt plan_tail_wave_num plan_fixpipe_unitflag plan_out_dtype_bytes plan_note gm_read_bytes l2_read_bytes t_mte2_gm t_mte2_l2 t_mte2 dma_cmd_count t_dma_cmd cube_flops t_mmad fixpipe_bytes t_fixpipe t_reduce t_steady t_drain t_total bottleneck feasible violations bound_type advice
2 to_matmul_demo 1 1 2048 2048 2048 bf16 bf16 bf16 False False False True 0 to_matmul_demo 转Matmul Ascend950PR batch_mat_mul_v3 32 1 0 0 1 折叠为 Matmul [2048,2048]x[2048,2048], 复用 Matmul 切分体系 0 1 0 0 2048 0 1 0 0 0 0 0 转Matmul后由 Matmul 体系决定 1 1 1 0 0 0 0 True 2 BatchB=1免费折叠: 左矩阵 [1,2048,2048] 视图折叠为 [2048,2048], 零重排零 split 16777216 0.0 1.048576e-05 0.0 1.048576e-05 0.0 0.0 17179869184.0 3.534952506995885e-05 8388608 5.24288e-06 0.0 3.534952506995885e-05 0.0 3.534952506995885e-05 MMAD True 计算Bound 瓶颈在 Cube 计算: 已接近理论算力上限, 检查是否有冗余计算 (MergeBatch 交叉项) 可消除
3 special_k0_demo 128 128 256 256 0 bf16 bf16 bf16 False False False True 0 special_k0_demo 特殊分支 Ascend950PR batch_mat_mul_v3 64 1 1 1 1 AIV 核间按行均分 (无 Cube tile 概念) 0 1 0 0 0 0 1 UB驻留(AIV) 0 0 0 allocate direct_gm 0 0 不涉及(AIV逐元素) 1 1 1 0 0 0 0 False 2 K=0纯写值: 无任何计算, C=bias 或 0, 纯 AIV 写值; 按行均分到 AIV 核 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 16777216 1.048576e-05 0.0 1.048576e-05 0.0 1.048576e-05 FIXPIPE MTE2 True 写出Bound 访存Bound(GM读写共享+L2重复读) 瓶颈在 Fixpipe 写出: 检查输出 dtype (fp16/fp8 可减半写出量), 或评估输出驻留 L2 异步回写策略 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
4 special_k1_demo 128 128 256 256 1 bf16 bf16 bf16 False False False True 0 special_k1_demo 特殊分支 Ascend950PR batch_mat_mul_v3 64 1 1 1 1 AIV 核间按行均分 (无 Cube tile 概念) 0 1 0 0 1 0 1 UB驻留(AIV) UB乒乓 0 0 0 allocate direct_gm 0 0 不涉及(AIV逐元素) 1 1 1 0 0 0 0 False 2 K=1逐元素乘: 退化为 C=A⊙B 无累加深度, Cube 16x16x16 粒度浪费 15/16; 走 AIV 通路 GM->UB->Mul->GM, UB乒乓 (B>=2*AIV 双batch乒乓流水) 131072 0.0 8.192e-08 0.0 8.192e-08 0.0 0.0 8388608.0 6.206060606060606e-07 16777216 1.048576e-05 0.0 1.048576e-05 1.0567679999999999e-05 0.0 1.048576e-05 1.0567679999999999e-05 FIXPIPE MTE2 True 写出Bound 访存Bound(GM读写共享+L2重复读) 瓶颈在 Fixpipe 写出: 检查输出 dtype (fp16/fp8 可减半写出量), 或评估输出驻留 L2 异步回写策略 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
5 merge_demo_k_trunc 2048 2048 32 32 256 bf16 bf16 bf16 False False False True 0 merge_demo_k_trunc MergeBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B均分(核间零重复读零依赖) 64 4 128 128 256 256 8 合并驻留 128 128 128 allocate(GM->L1随路驻留L2) direct_gm 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 b0=4 (L0C上限5.7/算存比上限19.0/b_core=64); K截断; 合并后单次DMA搬入 A'[128,256]+B'[256,128] 2097152 0.0 4.194304e-05 0.0 4.2743039999999997e-05 16.0 8.000000000000001e-07 134217728.0 8.837381267489712e-06 131072 2.62144e-06 0.0 4.2743039999999997e-05 4.5364479999999994e-05 3.0192408230452674e-07 4.3044964082304525e-05 4.566640408230452e-05 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
6 merge_iter_arbitrate 128 128 64 64 512 bf16 bf16 bf16 False False False True 0 merge_iter_arbitrate IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 4 1 64 64 512 512 2 b_双batch乒乓 64 64 256 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 双batch乒乓: 2*(MK+KN)*dtype=256KB <= L1 524288 0.0 1.048576e-05 0.0 1.0685759999999999e-05 4 2.0000000000000002e-07 16777216.0 1.104672658436214e-06 32768 6.5536e-07 0.0 1.0685759999999999e-05 1.1341119999999999e-05 4.400081646090535e-07 1.1125768164609053e-05 1.1781128164609052e-05 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
7 iter_demo_form_b 128 128 64 64 256 bf16 bf16 bf16 False False False True 0 iter_demo_form_b IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 4 1 64 64 256 256 2 b_双batch乒乓 64 64 256 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 双batch乒乓: 2*(MK+KN)*dtype=128KB <= L1 262144 0.0 5.24288e-06 0.0 5.4428799999999995e-06 4 2.0000000000000002e-07 8388608.0 5.52336329218107e-07 32768 6.5536e-07 0.0 5.4428799999999995e-06 6.09824e-06 3.0192408230452674e-07 5.744804082304526e-06 6.400164082304526e-06 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
8 iter_demo_form_d 64 64 64 64 8192 bf16 bf16 bf16 False False False True 0 iter_demo_form_d IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 2 1 64 64 8192 1024 1 d_两侧都切K 64 64 256 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 两侧都切K: k_L1=1024, K段成对流水, batch边界天然无缝; dValueA=2048B/dValueB=128B 4194304 0.0 8.388608e-05 0.0 8.468608e-05 16 8.000000000000001e-07 134217728.0 8.837381267489712e-06 16384 3.2768e-07 0.0 8.468608e-05 8.501376e-05 7.16176329218107e-07 8.540225632921811e-05 8.572993632921812e-05 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
9 streamk_demo 4 4 128 128 10240 bf16 bf16 bf16 False False False True 0 streamk_demo StreamK Ascend950PR batch_mat_mul_v3 32 1 1 1 32 B/M/N切出4块, 每块32核切K归约 (归约组内核c负责K段[c*K/32,(c+1)*K/32)) 1 1 128 128 320 256 1 K段标准分块流水 128 128 64 allocate(部分和驻留L2) resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换) 0 8388608 grid_K=32路切K+归约 1 1 32 0 0 0 0 True 4 P=1.00, grid_K=32, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终 327680.0 20971520 0.0 0 6.5536e-06 1.31072e-05 0.0 6.5536e-06 1.31072e-05 0.0 0.0 41943040.0 1342177280.0 2.761681646090535e-06 0.0 0.0 3.4067453613053613e-06 6.5536e-06 1.31072e-05 3.4067453613053613e-06 9.960345361305361e-06 1.651394536130536e-05 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
10 asw_demo_full 2 2 8192 8192 1024 bf16 bf16 bf16 False False False True 0 asw_demo_full ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 47 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 1024 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 方案B 52 52 1 52 52 2 139 True 2 L2场景B_输入驻留输出直写GM, r_in=1.00; 尾轮: 周长型主导, rho=0.06<rho_dv=0.53, A1b被dValue卡死, 方案B反超 L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: 周长型主导, rho=0.06<rho_dv=0.53, A1b被dValue卡死, 方案B反超 67108864.0 0.0 3087007744 4.194304e-05 0.0 0.0005936553353846154 4.194304e-05 0.0006355983753846153 0.0 0.0 274877906944.0 0.0005655924011193416 268435456 0.00016777216 0.0 0.0005655924011193416 0.0008033705353846154 0.0 0.0005655924011193416 0.0008033705353846154 MMAD MTE2 True 计算Bound 访存Bound(GM读写共享+L2重复读) 瓶颈在 Cube 计算: 已接近理论算力上限, 检查是否有冗余计算 (MergeBatch 交叉项) 可消除 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
11 asw_demo_reduce_core 16 16 256 256 128 bf16 bf16 bf16 False False False True 0 asw_demo_reduce_core ASW_Basic_降核 Ascend950PR batch_mat_mul_v3 16 1 1 1 1 降核: 只用16核, 每核一个L0C满载输出块, 其余核闲置 0 1 256 256 128 128 1 标准核内流水 256 256 64 allocate direct_gm 0 0 不涉及(每核一块无尾轮) 1 1 1 0 0 0 0 True 2 P=16.00<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢) 2097152.0 0.0 0 2.62144e-06 0.0 2.62144e-06 0.0 0.0 268435456.0 1.104672658436214e-06 2097152 2.62144e-06 8.065969230769231e-07 0.0 2.62144e-06 0.0 2.62144e-06 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
12 b4_m1_n128_k256 4 4 1 128 256 bf16 bf16 bf16 False False False True 0 b4_m1_n128_k256 ASW_Basic_降核 Ascend950PR batch_mat_mul_v3 1 1 1 1 1 降核: 只用1核, 每核一个L0C满载输出块, 其余核闲置 0 1 1 128 256 256 1 标准核内流水 1 128 128 allocate direct_gm 0 0 不涉及(每核一块无尾轮) 1 1 1 0 0 0 0 True 2 P=0.01<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢) 264192.0 0.0 0 5.28384e-06 0.0 5.28384e-06 0.0 0.0 262144.0 1.7260510288065844e-08 1024 2.048e-08 6.3015384615384615e-09 0.0 5.28384e-06 0.0 5.28384e-06 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
13 b8_m2_n192_k128 8 8 2 192 128 bf16 bf16 bf16 False False False True 0 b8_m2_n192_k128 ASW_Basic_降核 Ascend950PR batch_mat_mul_v3 1 1 1 1 1 降核: 只用1核, 每核一个L0C满载输出块, 其余核闲置 0 1 2 192 128 128 1 标准核内流水 2 192 80 allocate direct_gm 0 0 不涉及(每核一块无尾轮) 1 1 1 0 0 0 0 True 2 P=0.05<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢) 397312.0 0.0 0 7.94624e-06 0.0 7.94624e-06 0.0 0.0 786432.0 5.178153086419753e-08 6144 1.2288e-07 3.7809230769230766e-08 0.0 7.94624e-06 0.0 7.94624e-06 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
14 b16_m4_n256_k192 16 16 4 256 192 bf16 bf16 bf16 False False False True 0 b16_m4_n256_k192 ASW_Basic_降核 Ascend950PR batch_mat_mul_v3 1 1 1 1 1 降核: 只用1核, 每核一个L0C满载输出块, 其余核闲置 0 1 4 256 192 192 1 标准核内流水 4 256 64 allocate direct_gm 0 0 不涉及(每核一块无尾轮) 1 1 1 0 0 0 0 True 2 P=0.25<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢) 1597440.0 0.0 0 3.19488e-05 0.0 3.19488e-05 0.0 0.0 6291456.0 4.1425224691358024e-07 32768 6.5536e-07 2.0164923076923077e-07 0.0 3.19488e-05 0.0 3.19488e-05 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
15 b32_m8_n128_k256 32 32 8 128 256 bf16 bf16 bf16 False False False True 0 b32_m8_n128_k256 IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 1 1 8 128 256 256 1 a_单batch全驻留 8 128 128 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 单batch全驻留: (MK+KN)*dtype=68KB <= L1 69632 0.0 1.39264e-06 0.0 1.44264e-06 1 5.0000000000000004e-08 524288.0 3.452102057613169e-08 2048 4.096e-08 0.0 1.44264e-06 1.4836e-06 7.548102057613169e-08 1.5181210205761316e-06 1.5590810205761317e-06 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
16 b64_m16_n256_k512 64 64 16 256 512 bf16 bf16 bf16 False False False True 0 b64_m16_n256_k512 IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 2 1 16 256 512 240 1 c_一侧驻留+对侧切K 16 256 64 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 一侧驻留(A)+对侧切K: A驻留16KB, 预算L1/2, k_L1=240, dValueA=480B/dValueB=512B 557056 0.0 1.56672e-05 0.0 1.5967200000000002e-05 6 3.0000000000000004e-07 8388608.0 5.52336329218107e-07 16384 3.2768e-07 0.0 1.5967200000000002e-05 1.629488e-05 2.932938271604938e-07 1.6260493827160496e-05 1.6588173827160495e-05 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
17 b128_m8_n192_k256 128 128 8 192 256 bf16 bf16 bf16 False False False True 0 b128_m8_n192_k256 IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 4 1 8 192 256 256 2 b_双batch乒乓 8 192 80 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 双batch乒乓: 2*(MK+KN)*dtype=200KB <= L1 409600 0.0 8.192e-06 0.0 8.392e-06 4 2.0000000000000002e-07 3145728.0 2.0712612345679012e-07 12288 2.4576e-07 0.0 8.392e-06 8.637760000000001e-06 1.1322153086419754e-07 8.505221530864198e-06 8.7509815308642e-06 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
18 b32_m16_n8192_k7168 32 32 16 8192 7168 bf16 bf16 bf16 False False False True 0 b32_m16_n8192_k7168 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 1 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 7168 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 47 True 2 L2场景C_输入超L2分组执行, r_in=1.09; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮 4103077888.0000005 3765436416.0 0.0 337641472 0.0025644236800000005 0.00235339776 0.0 6.493105230769231e-05 0.0025644236800000005 0.0024183288123076924 0.0 0.0 60129542144.0 0.00012372333774485596 8388608 5.24288e-06 1.6131938461538462e-06 0.0 0.0025644236800000005 0.0024183288123076924 0.0 0.0025644236800000005 0.0024183288123076924 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
19 b4_m1_n8192_k8192 4 4 1 8192 8192 bf16 bf16 bf16 False False False True 0 b4_m1_n8192_k8192 StreamK Ascend950PR batch_mat_mul_v3 32 1 1 1 32 B/M/N切出4块, 每块32核切K归约 (归约组内核c负责K段[c*K/32,(c+1)*K/32)) 1 1 1 8192 256 256 1 K段标准分块流水 1 128 64 allocate(部分和驻留L2) resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换) 0 4194304 grid_K=32路切K+归约 1 1 32 0 0 0 0 True 4 P=0.50, grid_K=32, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终 262144.0 536936448 0.0 5.24288e-06 0.00033558528 0.0 5.24288e-06 0.00033558528 0.0 0.0 33554432.0 536870912.0 2.209345316872428e-06 1.104672658436214e-06 0.0 0.0 3.4067453613053613e-06 1.7033726806526807e-06 5.24288e-06 0.00033558528 3.4067453613053613e-06 1.7033726806526807e-06 8.64962536130536e-06 0.00033728865268065273 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
20 b16_m2_n4096_k7168 16 16 2 4096 7168 bf16 bf16 bf16 False False False True 0 b16_m2_n4096_k7168 StreamK Ascend950PR batch_mat_mul_v3 32 1 1 1 16 B/M/N切出16块, 每块16核切K归约 (归约组内核c负责K段[c*K/16,(c+1)*K/16)) 1 1 2 4096 448 256 1 K段标准分块流水 2 128 64 allocate(部分和驻留L2) resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换) 0 8388608 grid_K=16路切K+归约 1 1 16 0 0 0 0 True 4 P=2.00, grid_K=16, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终 458752.0 939982848 0.0 0 9.17504e-06 0.00058748928 0.0 9.17504e-06 0.00058748928 0.0 0.0 58720256.0 1879048192.0 3.866354304526749e-06 0.0 0.0 1.7159757575757577e-06 9.17504e-06 0.00058748928 1.7159757575757577e-06 1.0891015757575759e-05 0.0005892052557575758 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
21 b32_m64_n64_k7168 32 32 64 64 7168 bf16 bf16 bf16 False False False True 0 b32_m64_n64_k7168 IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 1 1 64 64 7168 1024 1 d_两侧都切K 64 64 256 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 两侧都切K: k_L1=1024, K段成对流水, batch边界天然无缝; dValueA=2048B/dValueB=128B 1835008 0.0 3.6700159999999995e-05 0.0 3.705016e-05 7 3.5000000000000004e-07 58720256.0 3.866354304526749e-06 8192 1.6384e-07 0.0 3.705016e-05 3.7214e-05 7.16176329218107e-07 3.77663363292181e-05 3.7930176329218104e-05 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
22 b64_m1024_n1024_k7168 64 64 1024 1024 7168 bf16 bf16 bf16 False False False True 0 b64_m1024_n1024_k7168 IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 2 1 1024 1024 7168 64 1 d_两侧都切K 176 176 64 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 两侧都切K: k_L1=64, K段成对流水, batch边界天然无缝; dValueA=128B/dValueB=2048B 58720256 0.0 0.0011744051199999998 0.0 0.00118560512 224 1.1200000000000001e-05 30064771072.0 0.0019795734039176954 4194304 8.388608e-05 0.0 0.0019795734039176954 5.078042126748971e-05 0.0020303538251851853 MMAD True 计算Bound 瓶颈在 Cube 计算: 已接近理论算力上限, 检查是否有冗余计算 (MergeBatch 交叉项) 可消除
23 b128_m2048_n2048_k1536 128 128 2048 2048 1536 bf16 bf16 bf16 False False False True 0 b128_m2048_n2048_k1536 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 12 12 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 1536 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 576 True 2 L2场景C_输入超L2分组执行, r_in=12.00; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮 19327352832.0 1610612736.0 0.0 17716740096 0.01207959552 0.00100663296 0.0 0.003407065403076923 0.01207959552 0.004413698363076923 0.0 0.0 1649267441664.0 0.0033935544067160493 1073741824 0.00067108864 0.0002064888123076923 0.0 0.01207959552 0.004413698363076923 0.0 0.01207959552 0.004413698363076923 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
24 b64_m1024_n8192_k2048 64 64 1024 8192 2048 bf16 bf16 bf16 False False False True 0 b64_m1024_n8192_k2048 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 6 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 2048 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 564 True 2 L2场景C_输入超L2分组执行, r_in=10.56; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮 25501368320.0 2415919104.0 0.0 23085449216 0.0159383552 0.00150994944 0.0 0.004439509464615385 0.0159383552 0.005949458904615385 0.0 0.0 2199023255552.0 0.004524739208954733 1073741824 0.00067108864 0.0002064888123076923 0.0 0.0159383552 0.005949458904615385 0.0 0.0159383552 0.005949458904615385 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
25 b32_m1024_n1024_k512 32 32 1024 1024 512 bf16 bf16 bf16 False False False True 0 b32_m1024_n1024_k512 IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 1 1 1024 1024 512 64 1 d_两侧都切K 176 176 64 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 两侧都切K: k_L1=64, K段成对流水, batch边界天然无缝; dValueA=128B/dValueB=2048B 2097152 0.0 4.194304e-05 0.0 4.234304e-05 8 4.0000000000000003e-07 1073741824.0 7.06990501399177e-05 2097152 4.194304e-05 0.0 7.06990501399177e-05 8.428607999999999e-05 5.078042126748971e-05 0.0001214794714074074 0.00013506650126748969 MMAD MTE2 True 计算Bound 访存Bound(GM读写共享+L2重复读) 瓶颈在 Cube 计算: 已接近理论算力上限, 检查是否有冗余计算 (MergeBatch 交叉项) 可消除 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
26 b128_m4096_n4096_k8192 128 128 4096 4096 8192 bf16 bf16 bf16 False False False True 0 b128_m4096_n4096_k8192 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 24 24 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 8192 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A0 1 1 1 0 0 0 2304 True 2 L2场景C_输入超L2分组执行, r_in=24.00; 尾轮: r=0 无尾轮 L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮 412316860416.0 17179869184.0 0.0 395136991232 0.25769803776 0.01073741824 0.0 0.07598788292923077 0.25769803776 0.08672530116923077 0.0 0.0 35184372088832.0 0.07239582734327572 4294967296 0.00268435456 0.0 0.25769803776 0.08940965572923076 0.0 0.25769803776 0.08940965572923076 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
27 b32_m8192_n4096_k7168 32 32 8192 4096 7168 bf16 bf16 bf16 False False False True 0 b32_m8192_n4096_k7168 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 47 24 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 7168 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A0 1 1 1 0 0 0 1128 True 2 L2场景C_输入超L2分组执行, r_in=31.67; 尾轮: r=0 无尾轮 L2场景C_单batch输入超L2分组执行, GM首读倍率=1.33; 尾轮: r=0 无尾轮 178509578240.0 7516192768.0 0.0 0.1115684864 0.00469762048 0.0 0.1115684864 0.00469762048 0.0 0.0 15393162788864.0 0.031673174462683126 2147483648 0.00134217728 0.0 0.1115684864 0.031673174462683126 0.0 0.1115684864 0.031673174462683126 MTE2_GM MMAD True 访存Bound(GM) 计算Bound 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 Cube 计算: 已接近理论算力上限, 检查是否有冗余计算 (MergeBatch 交叉项) 可消除
28 b32_m2048_n2048_k8192 32 32 2048 2048 8192 bf16 bf16 bf16 False False False True 0 b32_m2048_n2048_k8192 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 12 12 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 8192 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 144 True 2 L2场景C_输入超L2分组执行, r_in=12.00; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮 25769803776.0 2147483648.0 0.0 23622320128 0.01610612736 0.00134217728 0.0 0.004542753870769231 0.01610612736 0.005884931150769231 0.0 0.0 2199023255552.0 0.004524739208954733 268435456 0.00016777216 5.162220307692308e-05 0.0 0.01610612736 0.005884931150769231 0.0 0.01610612736 0.005884931150769231 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
29 b64_m4096_n2048_k128 64 64 4096 2048 128 bf16 bf16 bf16 False False False True 0 b64_m4096_n2048_k128 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 24 12 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 128 128 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 576 True 2 L2场景B_输入驻留输出直写GM, r_in=1.00; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮 100663296.0 0.0 1509949440 6.291456e-05 0.0 0.0002903748923076923 6.291456e-05 0.0003532894523076923 0.0 0.0 137438953472.0 0.0002827962005596708 1073741824 0.00067108864 0.0002064888123076923 0.0 0.00067108864 0.0003532894523076923 0.0 0.00067108864 0.0003532894523076923 FIXPIPE MTE2 True 写出Bound 访存Bound(GM读写共享+L2重复读) 瓶颈在 Fixpipe 写出: 检查输出 dtype (fp16/fp8 可减半写出量), 或评估输出驻留 L2 异步回写策略 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
30 b16_m1024_n1024_k8192 16 16 1024 1024 8192 bf16 bf16 bf16 False False False True 0 b16_m1024_n1024_k8192 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 6 6 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 8192 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 18 True 2 L2场景C_输入超L2分组执行, r_in=6.00; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮 3221225472.0 536870912.0 0.0 2684354560 0.00201326592 0.00033554432 0.0 0.0005162220307692308 0.00201326592 0.0008517663507692308 0.0 0.0 274877906944.0 0.0005655924011193416 33554432 2.097152e-05 6.452775384615385e-06 0.0 0.00201326592 0.0008517663507692308 0.0 0.00201326592 0.0008517663507692308 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
31 b4_m32768_n128_k128 4 4 32768 128 128 bf16 bf16 bf16 False False False True 0 b4_m32768_n128_k128 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 187 1 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 128 128 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) resident(输出驻留L2异步回写) 4 0 方案B 205 1 1 205 1 12 24 True 2 L2场景A_全驻留, r_in=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=28.96us), 选方案B工程简洁 (一套tile) L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=28.96us), 选方案B工程简洁 (一套tile) 33685504.0 0.0 24379392 2.105344e-05 0.0 4.688344615384615e-06 2.105344e-05 2.5741784615384616e-05 0.0 0.0 4294967296.0 8.837381267489712e-06 33554432 6.452775384615385e-06 0.0 2.105344e-05 2.5741784615384616e-05 0.0 2.105344e-05 2.5741784615384616e-05 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
32 b8_m32768_n2048_k512 8 8 32768 2048 512 bf16 bf16 bf16 False False False True 0 b8_m32768_n2048_k512 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 187 12 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 512 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A0 1 1 1 0 0 0 561 True 2 L2场景C_输入超L2分组执行, r_in=1.24; 尾轮: r=0 无尾轮 L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮 352321536.0 285212672.0 0.0 6073352192 0.00022020096 0.00017825792 0.0 0.0011679523446153847 0.00022020096 0.0013462102646153848 0.0 0.0 549755813888.0 0.0011311848022386832 1073741824 0.00067108864 0.0 0.0011311848022386832 0.0020172989046153846 0.0 0.0011311848022386832 0.0020172989046153846 MMAD MTE2 True 计算Bound 访存Bound(GM读写共享+L2重复读) 瓶颈在 Cube 计算: 已接近理论算力上限, 检查是否有冗余计算 (MergeBatch 交叉项) 可消除 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
33 b4_m131072_n128_k128 4 4 131072 128 128 bf16 bf16 bf16 False False False True 0 b4_m131072_n128_k128 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 745 1 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 128 128 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 方案B 820 1 1 820 1 4 94 True 2 L2场景C_输入超L2分组执行, r_in=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=115.39us), 选方案B工程简洁 (一套tile) L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=115.39us), 选方案B工程简洁 (一套tile) 134610944.0 134348800.0 0.0 97517568 8.413184e-05 8.3968e-05 0.0 1.875337846153846e-05 8.413184e-05 0.00010272137846153847 0.0 0.0 17179869184.0 3.534952506995885e-05 134217728 8.388608e-05 2.581110153846154e-05 0.0 8.413184e-05 0.00010272137846153847 0.0 8.413184e-05 0.00010272137846153847 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
34 b8_m131072_n1024_k256 8 8 131072 1024 256 bf16 bf16 bf16 False False False True 0 b8_m131072_n1024_k256 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 745 6 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 256 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 方案B 820 7 1 820 7 16 1118 True 2 L2场景C_输入超L2分组执行, r_in=1.06; 尾轮: 面积型主导, A1b与方案B严格打平(T=1384.63us), 选方案B工程简洁 (一套tile) L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=1384.63us), 选方案B工程简洁 (一套tile) 574619648.0 541065216.0 0.0 5804916736 0.00035913728 0.00033816576 0.0 0.0011163301415384615 0.00035913728 0.0014544959015384616 0.0 0.0 549755813888.0 0.0011311848022386832 2147483648 0.00134217728 0.0 0.00134217728 0.0027966731815384617 0.0 0.00134217728 0.0027966731815384617 FIXPIPE MTE2 True 写出Bound 访存Bound(GM读写共享+L2重复读) 瓶颈在 Fixpipe 写出: 检查输出 dtype (fp16/fp8 可减半写出量), 或评估输出驻留 L2 异步回写策略 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
35 b16_m32768_n8192_k7168 16 16 32768 8192 7168 bf16 bf16 bf16 False False False True 0 b16_m32768_n8192_k7168 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 187 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 7168 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 方案B 205 52 1 205 52 16 4395 True 2 L2场景C_输入超L2分组执行, r_in=75.00; 尾轮: 周长型主导, rho=0.50<rho_dv=0.53, A1b被dValue卡死, 方案B反超 L2场景C_单batch输入超L2分组执行, GM首读倍率=3.20; 尾轮: 周长型主导, rho=0.50<rho_dv=0.53, A1b被dValue卡死, 方案B反超 704643072000.0 30064771072.0 0.0 0.44040192 0.01879048192 0.0 0.44040192 0.01879048192 0.0 0.0 61572651155456.0 0.1266926978507325 8589934592 0.00536870912 0.0 0.44040192 0.1266926978507325 0.0 0.44040192 0.1266926978507325 MTE2_GM MMAD True 访存Bound(GM) 计算Bound 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 Cube 计算: 已接近理论算力上限, 检查是否有冗余计算 (MergeBatch 交叉项) 可消除
36 b4_m32768_n128_k8192 4 4 32768 128 8192 bf16 bf16 bf16 False False False True 0 b4_m32768_n128_k8192 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 187 1 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 8192 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 方案B 205 1 1 205 1 12 24 True 2 L2场景C_输入超L2分组执行, r_in=1.14; 尾轮: 周长型主导, rho=0.38<rho_dv=0.53, A1b被dValue卡死, 方案B反超 L2场景C_单batch输入超L2分组执行, GM首读倍率=1.03; 尾轮: 周长型主导, rho=0.38<rho_dv=0.53, A1b被dValue卡死, 方案B反超 2466250752.0 2222981120.0 0.0 0.00154140672 0.0013893632 0.0 0.00154140672 0.0013893632 0.0 0.0 274877906944.0 0.0005655924011193416 33554432 2.097152e-05 0.0 0.00154140672 0.00141033472 0.0 0.00154140672 0.00141033472 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
37 b32_m131072_n8192_k128 32 32 131072 8192 128 bf16 bf16 bf16 False False False True 0 b32_m131072_n8192_k128 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 745 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 128 128 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A0 1 1 1 0 0 0 35015 True 2 L2场景C_输入超L2分组执行, r_in=2.88; 尾轮: r=0 无尾轮 L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮 3288334336.0 1140850688.0 0.0 99321118720 0.00205520896 0.00071303168 0.0 0.019100215138461538 0.00205520896 0.019813246818461538 0.0 0.0 8796093022208.0 0.01809895683581893 68719476736 0.04294967296 0.0 0.04294967296 0.06276291977846153 0.0 0.04294967296 0.06276291977846153 FIXPIPE MTE2 True 写出Bound 访存Bound(GM读写共享+L2重复读) 瓶颈在 Fixpipe 写出: 检查输出 dtype (fp16/fp8 可减半写出量), 或评估输出驻留 L2 异步回写策略 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
38 b64_m32768_n8192_k1536 64 64 32768 8192 1536 bf16 bf16 bf16 False False False True 0 b64_m32768_n8192_k1536 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 187 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 1536 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A0 1 1 1 0 0 0 17578 True 2 L2场景C_输入超L2分组执行, r_in=75.00; 尾轮: r=0 无尾轮 L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮 603979776000.0 8053063680.0 0.0 595926712320 0.37748736 0.0050331648 0.0 0.11460129083076923 0.37748736 0.11963445563076923 0.0 0.0 52776558133248.0 0.10859374101491358 34359738368 0.02147483648 0.0 0.37748736 0.14110929211076922 0.0 0.37748736 0.14110929211076922 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
39 b8_m131072_n8192_k8192 8 8 131072 8192 8192 bf16 bf16 bf16 False False False True 0 b8_m131072_n8192_k8192 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 745 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 8192 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A1b 2 2 1 2 2 24 8754 True 2 L2场景C_输入超L2分组执行, r_in=44.53; 尾轮: 周长型主导, rho=0.75>=rho_dv=0.53, A1b恒优 (尾轮tile缩√rho=0.87倍凑满核) L2场景C_单batch输入超L2分组执行, GM首读倍率=4.76; 尾轮: 周长型主导, rho=0.75>=rho_dv=0.53, A1b恒优 (尾轮tile缩√rho=0.87倍凑满核) 812822560768.0 86973087744.0 0.0 0.50801410048 0.05435817984 0.0 0.50801410048 0.05435817984 0.0 0.0 140737488355328.0 0.2895833093731029 17179869184 0.01073741824 0.0 0.50801410048 0.2895833093731029 0.0 0.50801410048 0.2895833093731029 MTE2_GM MMAD True 访存Bound(GM) 计算Bound 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 Cube 计算: 已接近理论算力上限, 检查是否有冗余计算 (MergeBatch 交叉项) 可消除
40 b8_m16_n7168_k1536 8 8 16 7168 1536 bf16 bf16 bf16 False False False True 0 b8_m16_n7168_k1536 StreamK Ascend950PR batch_mat_mul_v3 32 1 1 2 2 B/M/N切出16块, 每块2核切K归约 (归约组内核c负责K段[c*K/2,(c+1)*K/2)) 1 1 16 3584 768 256 1 K段标准分块流水 16 128 64 allocate(部分和驻留L2) resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换) 0 3670016 grid_K=2路切K+归约 1 1 2 0 0 0 0 True 4 P=14.00, grid_K=2, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终 786432.0 176553984 0.0 393216 1.572864e-05 0.00011034624 0.0 7.561846153846153e-08 1.572864e-05 0.00011042185846153846 0.0 0.0 100663296.0 2818572288.0 6.628035950617284e-06 5.799531456790123e-06 0.0 0.0 2.3655235431235433e-07 2.0698331002331e-07 1.572864e-05 0.00011042185846153846 2.3655235431235433e-07 2.0698331002331e-07 1.5965192354312353e-05 0.00011062884177156177 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
41 b64_m1024_n7168_k7168 64 64 1024 7168 7168 bf16 bf16 bf16 False False False True 0 b64_m1024_n7168_k7168 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 6 41 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 7168 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 492 True 2 L2场景C_输入超L2分组执行, r_in=10.38; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮 77980499968.0 7516192768.0 0.0 70464307200 0.04873781248 0.00469762048 0.0 0.013550828307692308 0.04873781248 0.018248448787692308 0.0 0.0 6734508720128.0 0.013857013827423869 939524096 0.00058720256 0.00018067771076923076 0.0 0.04873781248 0.018248448787692308 0.0 0.04873781248 0.018248448787692308 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
42 b32_m8192_n8192_k7168 32 32 8192 8192 7168 bf16 bf16 bf16 False False False True 0 b32_m8192_n8192_k7168 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 47 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 7168 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A0 1 1 1 0 0 0 2209 True 2 L2场景C_输入超L2分组执行, r_in=47.00; 尾轮: r=0 无尾轮 L2场景C_单batch输入超L2分组执行, GM首读倍率=2.00; 尾轮: r=0 无尾轮 353261060096.0 15032385536.0 0.0 0.22078816256 0.00939524096 0.0 0.22078816256 0.00939524096 0.0 0.0 30786325577728.0 0.06334634892536625 4294967296 0.00268435456 0.0 0.22078816256 0.06334634892536625 0.0 0.22078816256 0.06334634892536625 MTE2_GM MMAD True 访存Bound(GM) 计算Bound 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 Cube 计算: 已接近理论算力上限, 检查是否有冗余计算 (MergeBatch 交叉项) 可消除
43 b8_m4096_n4096_k128 8 8 4096 4096 128 bf16 bf16 bf16 False False False True 0 b8_m4096_n4096_k128 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 24 24 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 128 128 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 144 True 2 L2场景B_输入驻留输出直写GM, r_in=1.00; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮 16777216.0 0.0 385875968 1.048576e-05 0.0 7.420691692307692e-05 1.048576e-05 8.469267692307693e-05 0.0 0.0 34359738368.0 7.06990501399177e-05 268435456 0.00016777216 5.162220307692308e-05 0.0 0.00016777216 8.469267692307693e-05 0.0 0.00016777216 8.469267692307693e-05 FIXPIPE MTE2 True 写出Bound 访存Bound(GM读写共享+L2重复读) 瓶颈在 Fixpipe 写出: 检查输出 dtype (fp16/fp8 可减半写出量), 或评估输出驻留 L2 异步回写策略 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争
44 b128_m8192_n8192_k7168 128 128 8192 8192 7168 bf16 bf16 bf16 False False False True 0 b128_m8192_n8192_k7168 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 47 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 7168 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A0 1 1 1 0 0 0 8836 True 2 L2场景C_输入超L2分组执行, r_in=47.00; 尾轮: r=0 无尾轮 L2场景C_单batch输入超L2分组执行, GM首读倍率=2.00; 尾轮: r=0 无尾轮 1413044240384.0 60129542144.0 0.0 0.88315265024 0.03758096384 0.0 0.88315265024 0.03758096384 0.0 0.0 123145302310912.0 0.253385395701465 17179869184 0.01073741824 0.0 0.88315265024 0.253385395701465 0.0 0.88315265024 0.253385395701465 MTE2_GM MMAD True 访存Bound(GM) 计算Bound 瓶颈在 GM 搬入: 可考虑增大 tile 提升 dValue/单核搬移量, 或利用 L2 驻留吸收重复读 (MergeBatch/ASW swizzle 方向) 瓶颈在 Cube 计算: 已接近理论算力上限, 检查是否有冗余计算 (MergeBatch 交叉项) 可消除
45 special_k1_b64 64 64 8192 512 1 bf16 bf16 bf16 False False False True 0 special_k1_b64 特殊分支 Ascend950PR batch_mat_mul_v3 64 1 1 1 1 AIV 核间按行均分 (无 Cube tile 概念) 0 1 0 0 1 0 1 UB驻留(AIV) AIV单缓冲 0 0 0 allocate direct_gm 0 0 不涉及(AIV逐元素) 1 1 1 0 0 0 0 False 2 K=1逐元素乘: 退化为 C=A⊙B 无累加深度, Cube 16x16x16 粒度浪费 15/16; 走 AIV 通路 GM->UB->Mul->GM, AIV单缓冲 (B<2*AIV 逐batch单缓冲串行) 1114112 0.0 6.9632e-07 0.0 6.9632e-07 0.0 0.0 268435456.0 1.985939393939394e-05 536870912 0.00033554432 0.0 0.00033554432 0.00033624063999999996 0.0 0.00033554432 0.00033624063999999996 FIXPIPE MTE2 True 写出Bound 访存Bound(GM读写共享+L2重复读) 瓶颈在 Fixpipe 写出: 检查输出 dtype (fp16/fp8 可减半写出量), 或评估输出驻留 L2 异步回写策略 瓶颈在 MTE2 搬移链 (GM 读写共享总线 + L2 重复读): 可增大 tile 提升 dValue/单核搬移量、利用 L2 驻留吸收重复读 (ASW swizzle/分组方向), 或评估输出驻留 L2 以减少 GM 直写与读竞争

84
BMM/BMM_Theory/examples/result_recommend.csv
View File

@@ -1,45 +1,45 @@
case_id,batch_a,batch_b,m,n,k,dtype_a,dtype_b,dtype_c,trans_a,trans_b,has_bias,out_nd,deterministic_level,plan_case_id,plan_branch,plan_npu,plan_op,plan_used_core_num,plan_split_b,plan_m_cnt,plan_n_cnt,plan_grid_k,plan_core_map,plan_b_core,plan_merge_b0,plan_single_core_m,plan_single_core_n,plan_single_core_k,plan_k_l1,plan_b_l1,plan_l1_form,plan_base_m,plan_base_n,plan_base_k,plan_l2_policy_in,plan_l2_policy_out,plan_swizzle_w,plan_workspace_bytes,plan_tail_strategy,plan_tail_m_cnt,plan_tail_n_cnt,plan_tail_k_cnt,plan_tail_m_main,plan_tail_n_main,plan_tail_block_cnt,plan_tail_wave_num,plan_fixpipe_unitflag,plan_out_dtype_bytes,plan_note,gm_read_bytes,l2_read_bytes,t_mte2_gm,t_mte2_l2,t_mte2,dma_cmd_count,t_dma_cmd,cube_flops,t_mmad,fixpipe_bytes,t_fixpipe,t_reduce,t_steady,t_drain,t_total,bottleneck,feasible,violations,bound_type,advice
to_matmul_demo,1,1,2048,2048,2048,bf16,bf16,bf16,False,False,False,True,0,to_matmul_demo,转Matmul,Ascend950PR,batch_mat_mul_v3,32,1,0,0,1,"折叠为 Matmul [2048,2048]x[2048,2048], 复用 Matmul 切分体系",0,1,0,0,2048,0,1,,0,0,0,,,0,0,转Matmul后由 Matmul 体系决定,1,1,1,0,0,0,0,True,2,"BatchB=1免费折叠: 左矩阵 [1,2048,2048] 视图折叠为 [2048,2048], 零重排零 split",16777216,0.0,1.048576e-05,0.0,1.048576e-05,0.0,0.0,17179869184.0,3.534952506995885e-05,8388608,5.24288e-06,0.0,3.534952506995885e-05,0.0,3.534952506995885e-05,MMAD,True,,计算Bound,"BatchA=1或BatchB=1, 折叠转普通Matmul"
special_k0_demo,128,128,256,256,0,bf16,bf16,bf16,False,False,False,True,0,special_k0_demo,特殊分支,Ascend950PR,batch_mat_mul_v3,64,1,1,1,1,AIV 核间按行均分 (无 Cube tile 概念),0,1,0,0,0,0,1,UB驻留(AIV),0,0,0,allocate,direct_gm,0,0,不涉及(AIV逐元素),1,1,1,0,0,0,0,False,2,"K=0纯写值: 无任何计算, C=bias 或 0, 纯 AIV 写值; 按行均分到 AIV 核",0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,16777216,1.048576e-05,0.0,1.048576e-05,0.0,1.048576e-05,FIXPIPE,True,,写出Bound,K=0纯写值
special_k1_demo,128,128,256,256,1,bf16,bf16,bf16,False,False,False,True,0,special_k1_demo,特殊分支,Ascend950PR,batch_mat_mul_v3,64,1,1,1,1,AIV 核间按行均分 (无 Cube tile 概念),0,1,0,0,1,0,1,UB驻留(AIV) UB乒乓,0,0,0,allocate,direct_gm,0,0,不涉及(AIV逐元素),1,1,1,0,0,0,0,False,2,"K=1逐元素乘: 退化为 C=A⊙B 无累加深度, Cube 16x16x16 粒度浪费 15/16; 走 AIV 通路 GM->UB->Mul->GM, UB乒乓 (B>=2*AIV 双batch乒乓流水)",131072,0.0,8.192e-08,0.0,8.192e-08,0.0,0.0,8388608.0,6.206060606060606e-07,16777216,1.048576e-05,0.0,1.048576e-05,0.0,1.048576e-05,FIXPIPE,True,,写出Bound,"K=1逐元素乘, 走AIV向量通路"
merge_demo_k_trunc,2048,2048,32,32,256,bf16,bf16,bf16,False,False,False,True,0,merge_demo_k_trunc,MergeBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B均分(核间零重复读零依赖),64,4,128,128,256,256,8,合并驻留,128,128,128,allocate(GM->L1随路驻留L2),direct_gm,0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"b0=4 (L0C上限5.7/算存比上限19.0/b_core=64); K截断; 合并后单次DMA搬入 A'[128,256]+B'[256,128]",2097152,0.0,4.194304e-05,0.0,4.2743039999999997e-05,16.0,8.000000000000001e-07,134217728.0,8.837381267489712e-06,131072,2.62144e-06,0.0,4.2743039999999997e-05,3.0192408230452674e-07,4.3044964082304525e-05,MTE2_GM,True,,访存Bound(GM),"两分支均合法, 仲裁: [分界条件] MergeBatch最优=True (k_L1=K(截断); b_core=64 vs 阈值 b0*(T_comp+T_write)/T_cmd=6.0; drain惩罚=(b0-1)*(T_comp+T_write)=0.23us, 搬移节省=b_core*(1-1/b0)*T_cmd=2.40us); [时延模型] T_MergeBatch=43.04us vs T_IterBatch=45.22us -> MergeBatch更优; [裁决] MergeBatch"
merge_iter_arbitrate,128,128,64,64,512,bf16,bf16,bf16,False,False,False,True,0,merge_iter_arbitrate,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),4,1,64,64,512,512,2,b_双batch乒乓,64,64,256,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,双batch乒乓: 2*(MK+KN)*dtype=256KB <= L1,524288,0.0,1.048576e-05,0.0,1.0685759999999999e-05,4,2.0000000000000002e-07,16777216.0,1.104672658436214e-06,32768,6.5536e-07,0.0,1.0685759999999999e-05,4.400081646090535e-07,1.1125768164609053e-05,MTE2_GM,True,,访存Bound(GM),"两分支均合法, 仲裁: [分界条件] MergeBatch最优=False (k_L1=K(截断); b_core=4 vs 阈值 b0*(T_comp+T_write)/T_cmd=17.6; drain惩罚=(b0-1)*(T_comp+T_write)=0.44us, 搬移节省=b_core*(1-1/b0)*T_cmd=0.10us); [时延模型] T_MergeBatch=11.49us vs T_IterBatch=11.13us -> IterBatch更优; [裁决] IterBatch"
iter_demo_form_b,128,128,64,64,256,bf16,bf16,bf16,False,False,False,True,0,iter_demo_form_b,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),4,1,64,64,256,256,2,b_双batch乒乓,64,64,256,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,双batch乒乓: 2*(MK+KN)*dtype=128KB <= L1,262144,0.0,5.24288e-06,0.0,5.4428799999999995e-06,4,2.0000000000000002e-07,8388608.0,5.52336329218107e-07,32768,6.5536e-07,0.0,5.4428799999999995e-06,3.0192408230452674e-07,5.744804082304526e-06,MTE2_GM,True,,访存Bound(GM),仅 IterBatch 条件满足
iter_demo_form_d,64,64,64,64,8192,bf16,bf16,bf16,False,False,False,True,0,iter_demo_form_d,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),2,1,64,64,8192,1024,1,d_两侧都切K,64,64,256,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"两侧都切K: k_L1=1024, K段成对流水, batch边界天然无缝; dValueA=2048B/dValueB=128B",4194304,0.0,8.388608e-05,0.0,8.468608e-05,16,8.000000000000001e-07,134217728.0,8.837381267489712e-06,16384,3.2768e-07,0.0,8.468608e-05,7.16176329218107e-07,8.540225632921811e-05,MTE2_GM,True,,访存Bound(GM),仅 IterBatch 条件满足
streamk_demo,4,4,128,128,10240,bf16,bf16,bf16,False,False,False,True,0,streamk_demo,StreamK,Ascend950PR,batch_mat_mul_v3,32,1,1,1,32,"B/M/N切出4块, 每块32核切K归约 (归约组内核c负责K段[c*K/32,(c+1)*K/32))",1,1,128,128,320,256,1,K段标准分块流水,128,128,64,allocate(部分和驻留L2),"resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换)",0,8388608,grid_K=32路切K+归约,1,1,32,0,0,0,0,True,4,"P=1.00, grid_K=32, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终",327680.0,0.0,6.5536e-06,0.0,6.5536e-06,0.0,0.0,41943040.0,2.761681646090535e-06,0.0,0.0,3.4067453613053613e-06,6.5536e-06,3.4067453613053613e-06,9.960345361305361e-06,MTE2_GM,True,,访存Bound(GM),"P<=C/2, B/M/N并行度买不满, 切K (grid_K=32)"
asw_demo_full,2,2,8192,8192,1024,bf16,bf16,bf16,False,False,False,True,0,asw_demo_full,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,1024,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,52,52,1,52,52,2,139,True,2,"L2场景B_输入驻留输出直写GM, r_in=1.00; 尾轮: 周长型主导, rho=0.06<rho_dv=0.53, A1b被dValue卡死, 方案B反超",67108864.0,0.0,4.194304e-05,0.0,4.194304e-05,0.0,0.0,274877906944.0,0.0005655924011193416,268435456,0.00016777216,0.0,0.0005655924011193416,0.0,0.0005655924011193416,MMAD,True,,计算Bound,ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
asw_demo_reduce_core,16,16,256,256,128,bf16,bf16,bf16,False,False,False,True,0,asw_demo_reduce_core,ASW_Basic_降核,Ascend950PR,batch_mat_mul_v3,16,1,1,1,1,"降核: 只用16核, 每核一个L0C满载输出块, 其余核闲置",0,1,256,256,128,128,1,标准核内流水,256,256,64,allocate,direct_gm,0,0,不涉及(每核一块无尾轮),1,1,1,0,0,0,0,True,2,"P=16.00<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢)",2097152.0,0.0,2.62144e-06,0.0,2.62144e-06,0.0,0.0,268435456.0,1.104672658436214e-06,2097152,2.62144e-06,0.0,2.62144e-06,0.0,2.62144e-06,MTE2_GM,True,,访存Bound(GM),ASW_Basic兜底 (StreamK未过: 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2)
b4_m1_n128_k256,4,4,1,128,256,bf16,bf16,bf16,False,False,False,True,0,b4_m1_n128_k256,ASW_Basic_降核,Ascend950PR,batch_mat_mul_v3,1,1,1,1,1,"降核: 只用1核, 每核一个L0C满载输出块, 其余核闲置",0,1,1,128,256,256,1,标准核内流水,1,128,128,allocate,direct_gm,0,0,不涉及(每核一块无尾轮),1,1,1,0,0,0,0,True,2,"P=0.01<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢)",264192.0,0.0,5.28384e-06,0.0,5.28384e-06,0.0,0.0,262144.0,1.7260510288065844e-08,1024,2.048e-08,0.0,5.28384e-06,0.0,5.28384e-06,MTE2_GM,True,,访存Bound(GM),ASW_Basic兜底 (StreamK未过: 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受: K > grid_K^2/(grid_K-1)*theta_c)
b8_m2_n192_k128,8,8,2,192,128,bf16,bf16,bf16,False,False,False,True,0,b8_m2_n192_k128,ASW_Basic_降核,Ascend950PR,batch_mat_mul_v3,1,1,1,1,1,"降核: 只用1核, 每核一个L0C满载输出块, 其余核闲置",0,1,2,192,128,128,1,标准核内流水,2,192,80,allocate,direct_gm,0,0,不涉及(每核一块无尾轮),1,1,1,0,0,0,0,True,2,"P=0.05<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢)",397312.0,0.0,7.94624e-06,0.0,7.94624e-06,0.0,0.0,786432.0,5.178153086419753e-08,6144,1.2288e-07,0.0,7.94624e-06,0.0,7.94624e-06,MTE2_GM,True,,访存Bound(GM),ASW_Basic兜底 (StreamK未过: 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受: K > grid_K^2/(grid_K-1)*theta_c)
b16_m4_n256_k192,16,16,4,256,192,bf16,bf16,bf16,False,False,False,True,0,b16_m4_n256_k192,ASW_Basic_降核,Ascend950PR,batch_mat_mul_v3,1,1,1,1,1,"降核: 只用1核, 每核一个L0C满载输出块, 其余核闲置",0,1,4,256,192,192,1,标准核内流水,4,256,64,allocate,direct_gm,0,0,不涉及(每核一块无尾轮),1,1,1,0,0,0,0,True,2,"P=0.25<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢)",1597440.0,0.0,3.19488e-05,0.0,3.19488e-05,0.0,0.0,6291456.0,4.1425224691358024e-07,32768,6.5536e-07,0.0,3.19488e-05,0.0,3.19488e-05,MTE2_GM,True,,访存Bound(GM),ASW_Basic兜底 (StreamK未过: 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受: K > grid_K^2/(grid_K-1)*theta_c)
b32_m8_n128_k256,32,32,8,128,256,bf16,bf16,bf16,False,False,False,True,0,b32_m8_n128_k256,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),1,1,8,128,256,256,1,a_单batch全驻留,8,128,128,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,单batch全驻留: (MK+KN)*dtype=68KB <= L1,69632,0.0,1.39264e-06,0.0,1.44264e-06,1,5.0000000000000004e-08,524288.0,3.452102057613169e-08,2048,4.096e-08,0.0,1.44264e-06,7.548102057613169e-08,1.5181210205761316e-06,MTE2_GM,True,,访存Bound(GM),仅 IterBatch 条件满足
b64_m16_n256_k512,64,64,16,256,512,bf16,bf16,bf16,False,False,False,True,0,b64_m16_n256_k512,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),2,1,16,256,512,240,1,c_一侧驻留+对侧切K,16,256,64,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"一侧驻留(A)+对侧切K: A驻留16KB, 预算L1/2, k_L1=240, dValueA=480B/dValueB=512B",557056,0.0,1.56672e-05,0.0,1.5967200000000002e-05,6,3.0000000000000004e-07,8388608.0,5.52336329218107e-07,16384,3.2768e-07,0.0,1.5967200000000002e-05,2.932938271604938e-07,1.6260493827160496e-05,MTE2_GM,True,,访存Bound(GM),仅 IterBatch 条件满足
b128_m8_n192_k256,128,128,8,192,256,bf16,bf16,bf16,False,False,False,True,0,b128_m8_n192_k256,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),4,1,8,192,256,256,2,b_双batch乒乓,8,192,80,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,双batch乒乓: 2*(MK+KN)*dtype=200KB <= L1,409600,0.0,8.192e-06,0.0,8.392e-06,4,2.0000000000000002e-07,3145728.0,2.0712612345679012e-07,12288,2.4576e-07,0.0,8.392e-06,1.1322153086419754e-07,8.505221530864198e-06,MTE2_GM,True,,访存Bound(GM),仅 IterBatch 条件满足
b32_m16_n8192_k7168,32,32,16,8192,7168,bf16,bf16,bf16,False,False,False,True,0,b32_m16_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,1,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,47,True,2,"L2场景C_输入超L2分组执行, r_in=1.09; 尾轮: r=0 无尾轮",4103077888.0000005,0.0,0.0025644236800000005,0.0,0.0025644236800000005,0.0,0.0,60129542144.0,0.00012372333774485596,8388608,5.24288e-06,0.0,0.0025644236800000005,0.0,0.0025644236800000005,MTE2_GM,True,,访存Bound(GM),"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B"
b4_m1_n8192_k8192,4,4,1,8192,8192,bf16,bf16,bf16,False,False,False,True,0,b4_m1_n8192_k8192,StreamK,Ascend950PR,batch_mat_mul_v3,32,1,1,1,32,"B/M/N切出4块, 每块32核切K归约 (归约组内核c负责K段[c*K/32,(c+1)*K/32))",1,1,1,8192,256,256,1,K段标准分块流水,1,128,64,allocate(部分和驻留L2),"resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换)",0,4194304,grid_K=32路切K+归约,1,1,32,0,0,0,0,True,4,"P=0.50, grid_K=32, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终",262144.0,0.0,5.24288e-06,0.0,5.24288e-06,0.0,0.0,33554432.0,2.209345316872428e-06,0.0,0.0,3.4067453613053613e-06,5.24288e-06,3.4067453613053613e-06,8.64962536130536e-06,MTE2_GM,True,,访存Bound(GM),"P<=C/2, B/M/N并行度买不满, 切K (grid_K=32)"
b16_m2_n4096_k7168,16,16,2,4096,7168,bf16,bf16,bf16,False,False,False,True,0,b16_m2_n4096_k7168,StreamK,Ascend950PR,batch_mat_mul_v3,32,1,1,1,16,"B/M/N切出16块, 每块16核切K归约 (归约组内核c负责K段[c*K/16,(c+1)*K/16))",1,1,2,4096,448,256,1,K段标准分块流水,2,128,64,allocate(部分和驻留L2),"resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换)",0,8388608,grid_K=16路切K+归约,1,1,16,0,0,0,0,True,4,"P=2.00, grid_K=16, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终",458752.0,0.0,9.17504e-06,0.0,9.17504e-06,0.0,0.0,58720256.0,3.866354304526749e-06,0.0,0.0,1.7159757575757577e-06,9.17504e-06,1.7159757575757577e-06,1.0891015757575759e-05,MTE2_GM,True,,访存Bound(GM),"P<=C/2, B/M/N并行度买不满, 切K (grid_K=16)"
b32_m64_n64_k7168,32,32,64,64,7168,bf16,bf16,bf16,False,False,False,True,0,b32_m64_n64_k7168,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),1,1,64,64,7168,1024,1,d_两侧都切K,64,64,256,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"两侧都切K: k_L1=1024, K段成对流水, batch边界天然无缝; dValueA=2048B/dValueB=128B",1835008,0.0,3.6700159999999995e-05,0.0,3.705016e-05,7,3.5000000000000004e-07,58720256.0,3.866354304526749e-06,8192,1.6384e-07,0.0,3.705016e-05,7.16176329218107e-07,3.77663363292181e-05,MTE2_GM,True,,访存Bound(GM),仅 IterBatch 条件满足
special_k0_demo,128,128,256,256,0,bf16,bf16,bf16,False,False,False,True,0,special_k0_demo,特殊分支,Ascend950PR,batch_mat_mul_v3,64,1,1,1,1,AIV 核间按行均分 (无 Cube tile 概念),0,1,0,0,0,0,1,UB驻留(AIV),0,0,0,allocate,direct_gm,0,0,不涉及(AIV逐元素),1,1,1,0,0,0,0,False,2,"K=0纯写值: 无任何计算, C=bias 或 0, 纯 AIV 写值; 按行均分到 AIV 核",0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,16777216,1.048576e-05,0.0,1.048576e-05,0.0,1.048576e-05,MTE2,True,,访存Bound(GM读写共享+L2重复读),K=0纯写值
special_k1_demo,128,128,256,256,1,bf16,bf16,bf16,False,False,False,True,0,special_k1_demo,特殊分支,Ascend950PR,batch_mat_mul_v3,64,1,1,1,1,AIV 核间按行均分 (无 Cube tile 概念),0,1,0,0,1,0,1,UB驻留(AIV) UB乒乓,0,0,0,allocate,direct_gm,0,0,不涉及(AIV逐元素),1,1,1,0,0,0,0,False,2,"K=1逐元素乘: 退化为 C=A⊙B 无累加深度, Cube 16x16x16 粒度浪费 15/16; 走 AIV 通路 GM->UB->Mul->GM, UB乒乓 (B>=2*AIV 双batch乒乓流水)",131072,0.0,8.192e-08,0.0,8.192e-08,0.0,0.0,8388608.0,6.206060606060606e-07,16777216,1.048576e-05,0.0,1.0567679999999999e-05,0.0,1.0567679999999999e-05,MTE2,True,,访存Bound(GM读写共享+L2重复读),"K=1逐元素乘, 走AIV向量通路"
merge_demo_k_trunc,2048,2048,32,32,256,bf16,bf16,bf16,False,False,False,True,0,merge_demo_k_trunc,MergeBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B均分(核间零重复读零依赖),64,4,128,128,256,256,8,合并驻留,128,128,128,allocate(GM->L1随路驻留L2),direct_gm,0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"b0=4 (L0C上限5.7/算存比上限19.0/b_core=64); K截断; 合并后单次DMA搬入 A'[128,256]+B'[256,128]",2097152,0.0,4.194304e-05,0.0,4.2743039999999997e-05,16.0,8.000000000000001e-07,134217728.0,8.837381267489712e-06,131072,2.62144e-06,0.0,4.5364479999999994e-05,3.0192408230452674e-07,4.566640408230452e-05,MTE2,True,,访存Bound(GM读写共享+L2重复读),"两分支均合法, 仲裁: [分界条件] MergeBatch最优=True (k_L1=K(截断); b_core=64 vs 阈值 b0*(T_comp+T_write)/T_cmd=6.0; drain惩罚=(b0-1)*(T_comp+T_write)=0.23us, 搬移节省=b_core*(1-1/b0)*T_cmd=2.40us); [时延模型] T_MergeBatch=45.67us vs T_IterBatch=47.84us -> MergeBatch更优; [裁决] MergeBatch"
merge_iter_arbitrate,128,128,64,64,512,bf16,bf16,bf16,False,False,False,True,0,merge_iter_arbitrate,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),4,1,64,64,512,512,2,b_双batch乒乓,64,64,256,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,双batch乒乓: 2*(MK+KN)*dtype=256KB <= L1,524288,0.0,1.048576e-05,0.0,1.0685759999999999e-05,4,2.0000000000000002e-07,16777216.0,1.104672658436214e-06,32768,6.5536e-07,0.0,1.1341119999999999e-05,4.400081646090535e-07,1.1781128164609052e-05,MTE2,True,,访存Bound(GM读写共享+L2重复读),"两分支均合法, 仲裁: [分界条件] MergeBatch最优=False (k_L1=K(截断); b_core=4 vs 阈值 b0*(T_comp+T_write)/T_cmd=17.6; drain惩罚=(b0-1)*(T_comp+T_write)=0.44us, 搬移节省=b_core*(1-1/b0)*T_cmd=0.10us); [时延模型] T_MergeBatch=12.14us vs T_IterBatch=11.78us -> IterBatch更优; [裁决] IterBatch"
iter_demo_form_b,128,128,64,64,256,bf16,bf16,bf16,False,False,False,True,0,iter_demo_form_b,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),4,1,64,64,256,256,2,b_双batch乒乓,64,64,256,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,双batch乒乓: 2*(MK+KN)*dtype=128KB <= L1,262144,0.0,5.24288e-06,0.0,5.4428799999999995e-06,4,2.0000000000000002e-07,8388608.0,5.52336329218107e-07,32768,6.5536e-07,0.0,6.09824e-06,3.0192408230452674e-07,6.400164082304526e-06,MTE2,True,,访存Bound(GM读写共享+L2重复读),仅 IterBatch 条件满足
iter_demo_form_d,64,64,64,64,8192,bf16,bf16,bf16,False,False,False,True,0,iter_demo_form_d,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),2,1,64,64,8192,1024,1,d_两侧都切K,64,64,256,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"两侧都切K: k_L1=1024, K段成对流水, batch边界天然无缝; dValueA=2048B/dValueB=128B",4194304,0.0,8.388608e-05,0.0,8.468608e-05,16,8.000000000000001e-07,134217728.0,8.837381267489712e-06,16384,3.2768e-07,0.0,8.501376e-05,7.16176329218107e-07,8.572993632921812e-05,MTE2,True,,访存Bound(GM读写共享+L2重复读),仅 IterBatch 条件满足
streamk_demo,4,4,128,128,10240,bf16,bf16,bf16,False,False,False,True,0,streamk_demo,StreamK,Ascend950PR,batch_mat_mul_v3,32,1,1,1,32,"B/M/N切出4块, 每块32核切K归约 (归约组内核c负责K段[c*K/32,(c+1)*K/32))",1,1,128,128,320,256,1,K段标准分块流水,128,128,64,allocate(部分和驻留L2),"resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换)",0,8388608,grid_K=32路切K+归约,1,1,32,0,0,0,0,True,4,"P=1.00, grid_K=32, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终",20971520,0,1.31072e-05,0.0,1.31072e-05,0.0,0.0,1342177280.0,2.761681646090535e-06,0.0,0.0,3.4067453613053613e-06,1.31072e-05,3.4067453613053613e-06,1.651394536130536e-05,MTE2,True,,访存Bound(GM读写共享+L2重复读),"P<=C/2, B/M/N并行度买不满, 切K (grid_K=32)"
asw_demo_full,2,2,8192,8192,1024,bf16,bf16,bf16,False,False,False,True,0,asw_demo_full,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,1024,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,52,52,1,52,52,2,139,True,2,"L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: 周长型主导, rho=0.06<rho_dv=0.53, A1b被dValue卡死, 方案B反超",67108864.0,3087007744,4.194304e-05,0.0005936553353846154,0.0006355983753846153,0.0,0.0,274877906944.0,0.0005655924011193416,268435456,0.00016777216,0.0,0.0008033705353846154,0.0,0.0008033705353846154,MTE2,True,,访存Bound(GM读写共享+L2重复读),ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
asw_demo_reduce_core,16,16,256,256,128,bf16,bf16,bf16,False,False,False,True,0,asw_demo_reduce_core,ASW_Basic_降核,Ascend950PR,batch_mat_mul_v3,16,1,1,1,1,"降核: 只用16核, 每核一个L0C满载输出块, 其余核闲置",0,1,256,256,128,128,1,标准核内流水,256,256,64,allocate,direct_gm,0,0,不涉及(每核一块无尾轮),1,1,1,0,0,0,0,True,2,"P=16.00<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢)",2097152.0,0,2.62144e-06,0.0,2.62144e-06,0.0,0.0,268435456.0,1.104672658436214e-06,2097152,8.065969230769231e-07,0.0,2.62144e-06,0.0,2.62144e-06,MTE2,True,,访存Bound(GM读写共享+L2重复读),ASW_Basic兜底 (StreamK未过: 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2)
b4_m1_n128_k256,4,4,1,128,256,bf16,bf16,bf16,False,False,False,True,0,b4_m1_n128_k256,ASW_Basic_降核,Ascend950PR,batch_mat_mul_v3,1,1,1,1,1,"降核: 只用1核, 每核一个L0C满载输出块, 其余核闲置",0,1,1,128,256,256,1,标准核内流水,1,128,128,allocate,direct_gm,0,0,不涉及(每核一块无尾轮),1,1,1,0,0,0,0,True,2,"P=0.01<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢)",264192.0,0,5.28384e-06,0.0,5.28384e-06,0.0,0.0,262144.0,1.7260510288065844e-08,1024,6.3015384615384615e-09,0.0,5.28384e-06,0.0,5.28384e-06,MTE2,True,,访存Bound(GM读写共享+L2重复读),ASW_Basic兜底 (StreamK未过: 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受: K > grid_K^2/(grid_K-1)*theta_c)
b8_m2_n192_k128,8,8,2,192,128,bf16,bf16,bf16,False,False,False,True,0,b8_m2_n192_k128,ASW_Basic_降核,Ascend950PR,batch_mat_mul_v3,1,1,1,1,1,"降核: 只用1核, 每核一个L0C满载输出块, 其余核闲置",0,1,2,192,128,128,1,标准核内流水,2,192,80,allocate,direct_gm,0,0,不涉及(每核一块无尾轮),1,1,1,0,0,0,0,True,2,"P=0.05<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢)",397312.0,0,7.94624e-06,0.0,7.94624e-06,0.0,0.0,786432.0,5.178153086419753e-08,6144,3.7809230769230766e-08,0.0,7.94624e-06,0.0,7.94624e-06,MTE2,True,,访存Bound(GM读写共享+L2重复读),ASW_Basic兜底 (StreamK未过: 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受: K > grid_K^2/(grid_K-1)*theta_c)
b16_m4_n256_k192,16,16,4,256,192,bf16,bf16,bf16,False,False,False,True,0,b16_m4_n256_k192,ASW_Basic_降核,Ascend950PR,batch_mat_mul_v3,1,1,1,1,1,"降核: 只用1核, 每核一个L0C满载输出块, 其余核闲置",0,1,4,256,192,192,1,标准核内流水,4,256,64,allocate,direct_gm,0,0,不涉及(每核一块无尾轮),1,1,1,0,0,0,0,True,2,"P=0.25<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢)",1597440.0,0,3.19488e-05,0.0,3.19488e-05,0.0,0.0,6291456.0,4.1425224691358024e-07,32768,2.0164923076923077e-07,0.0,3.19488e-05,0.0,3.19488e-05,MTE2,True,,访存Bound(GM读写共享+L2重复读),ASW_Basic兜底 (StreamK未过: 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受: K > grid_K^2/(grid_K-1)*theta_c)
b32_m8_n128_k256,32,32,8,128,256,bf16,bf16,bf16,False,False,False,True,0,b32_m8_n128_k256,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),1,1,8,128,256,256,1,a_单batch全驻留,8,128,128,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,单batch全驻留: (MK+KN)*dtype=68KB <= L1,69632,0.0,1.39264e-06,0.0,1.44264e-06,1,5.0000000000000004e-08,524288.0,3.452102057613169e-08,2048,4.096e-08,0.0,1.4836e-06,7.548102057613169e-08,1.5590810205761317e-06,MTE2,True,,访存Bound(GM读写共享+L2重复读),仅 IterBatch 条件满足
b64_m16_n256_k512,64,64,16,256,512,bf16,bf16,bf16,False,False,False,True,0,b64_m16_n256_k512,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),2,1,16,256,512,240,1,c_一侧驻留+对侧切K,16,256,64,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"一侧驻留(A)+对侧切K: A驻留16KB, 预算L1/2, k_L1=240, dValueA=480B/dValueB=512B",557056,0.0,1.56672e-05,0.0,1.5967200000000002e-05,6,3.0000000000000004e-07,8388608.0,5.52336329218107e-07,16384,3.2768e-07,0.0,1.629488e-05,2.932938271604938e-07,1.6588173827160495e-05,MTE2,True,,访存Bound(GM读写共享+L2重复读),仅 IterBatch 条件满足
b128_m8_n192_k256,128,128,8,192,256,bf16,bf16,bf16,False,False,False,True,0,b128_m8_n192_k256,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),4,1,8,192,256,256,2,b_双batch乒乓,8,192,80,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,双batch乒乓: 2*(MK+KN)*dtype=200KB <= L1,409600,0.0,8.192e-06,0.0,8.392e-06,4,2.0000000000000002e-07,3145728.0,2.0712612345679012e-07,12288,2.4576e-07,0.0,8.637760000000001e-06,1.1322153086419754e-07,8.7509815308642e-06,MTE2,True,,访存Bound(GM读写共享+L2重复读),仅 IterBatch 条件满足
b32_m16_n8192_k7168,32,32,16,8192,7168,bf16,bf16,bf16,False,False,False,True,0,b32_m16_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,1,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,47,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮",3765436416.0,337641472,0.00235339776,6.493105230769231e-05,0.0024183288123076924,0.0,0.0,60129542144.0,0.00012372333774485596,8388608,1.6131938461538462e-06,0.0,0.0024183288123076924,0.0,0.0024183288123076924,MTE2,True,,访存Bound(GM读写共享+L2重复读),"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B"
b4_m1_n8192_k8192,4,4,1,8192,8192,bf16,bf16,bf16,False,False,False,True,0,b4_m1_n8192_k8192,StreamK,Ascend950PR,batch_mat_mul_v3,32,1,1,1,32,"B/M/N切出4块, 每块32核切K归约 (归约组内核c负责K段[c*K/32,(c+1)*K/32))",1,1,1,8192,256,256,1,K段标准分块流水,1,128,64,allocate(部分和驻留L2),"resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换)",0,4194304,grid_K=32路切K+归约,1,1,32,0,0,0,0,True,4,"P=0.50, grid_K=32, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终",536936448,0.0,0.00033558528,0.0,0.00033558528,0.0,0.0,536870912.0,1.104672658436214e-06,0.0,0.0,1.7033726806526807e-06,0.00033558528,1.7033726806526807e-06,0.00033728865268065273,MTE2,True,,访存Bound(GM读写共享+L2重复读),"P<=C/2, B/M/N并行度买不满, 切K (grid_K=32)"
b16_m2_n4096_k7168,16,16,2,4096,7168,bf16,bf16,bf16,False,False,False,True,0,b16_m2_n4096_k7168,StreamK,Ascend950PR,batch_mat_mul_v3,32,1,1,1,16,"B/M/N切出16块, 每块16核切K归约 (归约组内核c负责K段[c*K/16,(c+1)*K/16))",1,1,2,4096,448,256,1,K段标准分块流水,2,128,64,allocate(部分和驻留L2),"resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换)",0,8388608,grid_K=16路切K+归约,1,1,16,0,0,0,0,True,4,"P=2.00, grid_K=16, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终",939982848,0,0.00058748928,0.0,0.00058748928,0.0,0.0,1879048192.0,3.866354304526749e-06,0.0,0.0,1.7159757575757577e-06,0.00058748928,1.7159757575757577e-06,0.0005892052557575758,MTE2,True,,访存Bound(GM读写共享+L2重复读),"P<=C/2, B/M/N并行度买不满, 切K (grid_K=16)"
b32_m64_n64_k7168,32,32,64,64,7168,bf16,bf16,bf16,False,False,False,True,0,b32_m64_n64_k7168,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),1,1,64,64,7168,1024,1,d_两侧都切K,64,64,256,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"两侧都切K: k_L1=1024, K段成对流水, batch边界天然无缝; dValueA=2048B/dValueB=128B",1835008,0.0,3.6700159999999995e-05,0.0,3.705016e-05,7,3.5000000000000004e-07,58720256.0,3.866354304526749e-06,8192,1.6384e-07,0.0,3.7214e-05,7.16176329218107e-07,3.7930176329218104e-05,MTE2,True,,访存Bound(GM读写共享+L2重复读),仅 IterBatch 条件满足
b64_m1024_n1024_k7168,64,64,1024,1024,7168,bf16,bf16,bf16,False,False,False,True,0,b64_m1024_n1024_k7168,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),2,1,1024,1024,7168,64,1,d_两侧都切K,176,176,64,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"两侧都切K: k_L1=64, K段成对流水, batch边界天然无缝; dValueA=128B/dValueB=2048B",58720256,0.0,0.0011744051199999998,0.0,0.00118560512,224,1.1200000000000001e-05,30064771072.0,0.0019795734039176954,4194304,8.388608e-05,0.0,0.0019795734039176954,5.078042126748971e-05,0.0020303538251851853,MMAD,True,,计算Bound,仅 IterBatch 条件满足
b128_m2048_n2048_k1536,128,128,2048,2048,1536,bf16,bf16,bf16,False,False,False,True,0,b128_m2048_n2048_k1536,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,12,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,1536,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,576,True,2,"L2场景C_输入超L2分组执行, r_in=12.00; 尾轮: r=0 无尾轮",19327352832.0,0.0,0.01207959552,0.0,0.01207959552,0.0,0.0,1649267441664.0,0.0033935544067160493,1073741824,0.00067108864,0.0,0.01207959552,0.0,0.01207959552,MTE2_GM,True,,访存Bound(GM),"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0"
b64_m1024_n8192_k2048,64,64,1024,8192,2048,bf16,bf16,bf16,False,False,False,True,0,b64_m1024_n8192_k2048,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,6,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,2048,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,564,True,2,"L2场景C_输入超L2分组执行, r_in=10.56; 尾轮: r=0 无尾轮",25501368320.0,0.0,0.0159383552,0.0,0.0159383552,0.0,0.0,2199023255552.0,0.004524739208954733,1073741824,0.00067108864,0.0,0.0159383552,0.0,0.0159383552,MTE2_GM,True,,访存Bound(GM),"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0"
b32_m1024_n1024_k512,32,32,1024,1024,512,bf16,bf16,bf16,False,False,False,True,0,b32_m1024_n1024_k512,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),1,1,1024,1024,512,64,1,d_两侧都切K,176,176,64,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"两侧都切K: k_L1=64, K段成对流水, batch边界天然无缝; dValueA=128B/dValueB=2048B",2097152,0.0,4.194304e-05,0.0,4.234304e-05,8,4.0000000000000003e-07,1073741824.0,7.06990501399177e-05,2097152,4.194304e-05,0.0,7.06990501399177e-05,5.078042126748971e-05,0.0001214794714074074,MMAD,True,,计算Bound,仅 IterBatch 条件满足
b128_m4096_n4096_k8192,128,128,4096,4096,8192,bf16,bf16,bf16,False,False,False,True,0,b128_m4096_n4096_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,24,24,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,2304,True,2,"L2场景C_输入超L2分组执行, r_in=24.00; 尾轮: r=0 无尾轮",412316860416.0,0.0,0.25769803776,0.0,0.25769803776,0.0,0.0,35184372088832.0,0.07239582734327572,4294967296,0.00268435456,0.0,0.25769803776,0.0,0.25769803776,MTE2_GM,True,,访存Bound(GM),"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0"
b32_m8192_n4096_k7168,32,32,8192,4096,7168,bf16,bf16,bf16,False,False,False,True,0,b32_m8192_n4096_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,24,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,1128,True,2,"L2场景C_输入超L2分组执行, r_in=31.67; 尾轮: r=0 无尾轮",178509578240.0,0.0,0.1115684864,0.0,0.1115684864,0.0,0.0,15393162788864.0,0.031673174462683126,2147483648,0.00134217728,0.0,0.1115684864,0.0,0.1115684864,MTE2_GM,True,,访存Bound(GM),"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0"
b32_m2048_n2048_k8192,32,32,2048,2048,8192,bf16,bf16,bf16,False,False,False,True,0,b32_m2048_n2048_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,12,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,144,True,2,"L2场景C_输入超L2分组执行, r_in=12.00; 尾轮: r=0 无尾轮",25769803776.0,0.0,0.01610612736,0.0,0.01610612736,0.0,0.0,2199023255552.0,0.004524739208954733,268435456,0.00016777216,0.0,0.01610612736,0.0,0.01610612736,MTE2_GM,True,,访存Bound(GM),"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0"
b64_m4096_n2048_k128,64,64,4096,2048,128,bf16,bf16,bf16,False,False,False,True,0,b64_m4096_n2048_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,24,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,576,True,2,"L2场景B_输入驻留输出直写GM, r_in=1.00; 尾轮: r=0 无尾轮",100663296.0,0.0,6.291456e-05,0.0,6.291456e-05,0.0,0.0,137438953472.0,0.0002827962005596708,1073741824,0.00067108864,0.0,0.00067108864,0.0,0.00067108864,FIXPIPE,True,,写出Bound,"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0"
b16_m1024_n1024_k8192,16,16,1024,1024,8192,bf16,bf16,bf16,False,False,False,True,0,b16_m1024_n1024_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,6,6,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,18,True,2,"L2场景C_输入超L2分组执行, r_in=6.00; 尾轮: r=0 无尾轮",3221225472.0,0.0,0.00201326592,0.0,0.00201326592,0.0,0.0,274877906944.0,0.0005655924011193416,33554432,2.097152e-05,0.0,0.00201326592,0.0,0.00201326592,MTE2_GM,True,,访存Bound(GM),ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
b4_m32768_n128_k128,4,4,32768,128,128,bf16,bf16,bf16,False,False,False,True,0,b4_m32768_n128_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,1,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,方案B,205,1,1,205,1,12,24,True,2,"L2场景A_全驻留, r_in=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=28.96us), 选方案B工程简洁 (一套tile)",33685504.0,0.0,2.105344e-05,0.0,2.105344e-05,0.0,0.0,4294967296.0,8.837381267489712e-06,33554432,6.452775384615385e-06,0.0,2.105344e-05,0.0,2.105344e-05,MTE2_GM,True,,访存Bound(GM),ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
b8_m32768_n2048_k512,8,8,32768,2048,512,bf16,bf16,bf16,False,False,False,True,0,b8_m32768_n2048_k512,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,512,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,561,True,2,"L2场景C_输入超L2分组执行, r_in=1.24; 尾轮: r=0 无尾轮",352321536.0,0.0,0.00022020096,0.0,0.00022020096,0.0,0.0,549755813888.0,0.0011311848022386832,1073741824,0.00067108864,0.0,0.0011311848022386832,0.0,0.0011311848022386832,MMAD,True,,计算Bound,ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
b4_m131072_n128_k128,4,4,131072,128,128,bf16,bf16,bf16,False,False,False,True,0,b4_m131072_n128_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,1,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,820,1,1,820,1,4,94,True,2,"L2场景C_输入超L2分组执行, r_in=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=115.39us), 选方案B工程简洁 (一套tile)",134610944.0,0.0,8.413184e-05,0.0,8.413184e-05,0.0,0.0,17179869184.0,3.534952506995885e-05,134217728,8.388608e-05,0.0,8.413184e-05,0.0,8.413184e-05,MTE2_GM,True,,访存Bound(GM),ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
b8_m131072_n1024_k256,8,8,131072,1024,256,bf16,bf16,bf16,False,False,False,True,0,b8_m131072_n1024_k256,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,6,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,256,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,820,7,1,820,7,16,1118,True,2,"L2场景C_输入超L2分组执行, r_in=1.06; 尾轮: 面积型主导, A1b与方案B严格打平(T=1384.63us), 选方案B工程简洁 (一套tile)",574619648.0,0.0,0.00035913728,0.0,0.00035913728,0.0,0.0,549755813888.0,0.0011311848022386832,2147483648,0.00134217728,0.0,0.00134217728,0.0,0.00134217728,FIXPIPE,True,,写出Bound,ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
b16_m32768_n8192_k7168,16,16,32768,8192,7168,bf16,bf16,bf16,False,False,False,True,0,b16_m32768_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,205,52,1,205,52,16,4395,True,2,"L2场景C_输入超L2分组执行, r_in=75.00; 尾轮: 周长型主导, rho=0.50<rho_dv=0.53, A1b被dValue卡死, 方案B反超",704643072000.0,0.0,0.44040192,0.0,0.44040192,0.0,0.0,61572651155456.0,0.1266926978507325,8589934592,0.00536870912,0.0,0.44040192,0.0,0.44040192,MTE2_GM,True,,访存Bound(GM),ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
b4_m32768_n128_k8192,4,4,32768,128,8192,bf16,bf16,bf16,False,False,False,True,0,b4_m32768_n128_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,1,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,205,1,1,205,1,12,24,True,2,"L2场景C_输入超L2分组执行, r_in=1.14; 尾轮: 周长型主导, rho=0.38<rho_dv=0.53, A1b被dValue卡死, 方案B反超",2466250752.0,0.0,0.00154140672,0.0,0.00154140672,0.0,0.0,274877906944.0,0.0005655924011193416,33554432,2.097152e-05,0.0,0.00154140672,0.0,0.00154140672,MTE2_GM,True,,访存Bound(GM),ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
b32_m131072_n8192_k128,32,32,131072,8192,128,bf16,bf16,bf16,False,False,False,True,0,b32_m131072_n8192_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,35015,True,2,"L2场景C_输入超L2分组执行, r_in=2.88; 尾轮: r=0 无尾轮",3288334336.0,0.0,0.00205520896,0.0,0.00205520896,0.0,0.0,8796093022208.0,0.01809895683581893,68719476736,0.04294967296,0.0,0.04294967296,0.0,0.04294967296,FIXPIPE,True,,写出Bound,"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0"
b64_m32768_n8192_k1536,64,64,32768,8192,1536,bf16,bf16,bf16,False,False,False,True,0,b64_m32768_n8192_k1536,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,1536,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,17578,True,2,"L2场景C_输入超L2分组执行, r_in=75.00; 尾轮: r=0 无尾轮",603979776000.0,0.0,0.37748736,0.0,0.37748736,0.0,0.0,52776558133248.0,0.10859374101491358,34359738368,0.02147483648,0.0,0.37748736,0.0,0.37748736,MTE2_GM,True,,访存Bound(GM),"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0"
b8_m131072_n8192_k8192,8,8,131072,8192,8192,bf16,bf16,bf16,False,False,False,True,0,b8_m131072_n8192_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A1b,2,2,1,2,2,24,8754,True,2,"L2场景C_输入超L2分组执行, r_in=44.53; 尾轮: 周长型主导, rho=0.75>=rho_dv=0.53, A1b恒优 (尾轮tile缩√rho=0.87倍凑满核)",812822560768.0,0.0,0.50801410048,0.0,0.50801410048,0.0,0.0,140737488355328.0,0.2895833093731029,17179869184,0.01073741824,0.0,0.50801410048,0.0,0.50801410048,MTE2_GM,True,,访存Bound(GM),ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
b8_m16_n7168_k1536,8,8,16,7168,1536,bf16,bf16,bf16,False,False,False,True,0,b8_m16_n7168_k1536,StreamK,Ascend950PR,batch_mat_mul_v3,32,1,1,2,2,"B/M/N切出16块, 每块2核切K归约 (归约组内核c负责K段[c*K/2,(c+1)*K/2))",1,1,16,3584,768,256,1,K段标准分块流水,16,128,64,allocate(部分和驻留L2),"resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换)",0,3670016,grid_K=2路切K+归约,1,1,2,0,0,0,0,True,4,"P=14.00, grid_K=2, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终",786432.0,0.0,1.572864e-05,0.0,1.572864e-05,0.0,0.0,100663296.0,6.628035950617284e-06,0.0,0.0,2.3655235431235433e-07,1.572864e-05,2.3655235431235433e-07,1.5965192354312353e-05,MTE2_GM,True,,访存Bound(GM),"P<=C/2, B/M/N并行度买不满, 切K (grid_K=2)"
b64_m1024_n7168_k7168,64,64,1024,7168,7168,bf16,bf16,bf16,False,False,False,True,0,b64_m1024_n7168_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,6,41,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,492,True,2,"L2场景C_输入超L2分组执行, r_in=10.38; 尾轮: r=0 无尾轮",77980499968.0,0.0,0.04873781248,0.0,0.04873781248,0.0,0.0,6734508720128.0,0.013857013827423869,939524096,0.00058720256,0.0,0.04873781248,0.0,0.04873781248,MTE2_GM,True,,访存Bound(GM),"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0"
b32_m8192_n8192_k7168,32,32,8192,8192,7168,bf16,bf16,bf16,False,False,False,True,0,b32_m8192_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,2209,True,2,"L2场景C_输入超L2分组执行, r_in=47.00; 尾轮: r=0 无尾轮",353261060096.0,0.0,0.22078816256,0.0,0.22078816256,0.0,0.0,30786325577728.0,0.06334634892536625,4294967296,0.00268435456,0.0,0.22078816256,0.0,0.22078816256,MTE2_GM,True,,访存Bound(GM),"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0"
b8_m4096_n4096_k128,8,8,4096,4096,128,bf16,bf16,bf16,False,False,False,True,0,b8_m4096_n4096_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,24,24,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,144,True,2,"L2场景B_输入驻留输出直写GM, r_in=1.00; 尾轮: r=0 无尾轮",16777216.0,0.0,1.048576e-05,0.0,1.048576e-05,0.0,0.0,34359738368.0,7.06990501399177e-05,268435456,0.00016777216,0.0,0.00016777216,0.0,0.00016777216,FIXPIPE,True,,写出Bound,ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
b128_m8192_n8192_k7168,128,128,8192,8192,7168,bf16,bf16,bf16,False,False,False,True,0,b128_m8192_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,8836,True,2,"L2场景C_输入超L2分组执行, r_in=47.00; 尾轮: r=0 无尾轮",1413044240384.0,0.0,0.88315265024,0.0,0.88315265024,0.0,0.0,123145302310912.0,0.253385395701465,17179869184,0.01073741824,0.0,0.88315265024,0.0,0.88315265024,MTE2_GM,True,,访存Bound(GM),"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0"
special_k1_b64,64,64,8192,512,1,bf16,bf16,bf16,False,False,False,True,0,special_k1_b64,特殊分支,Ascend950PR,batch_mat_mul_v3,64,1,1,1,1,AIV 核间按行均分 (无 Cube tile 概念),0,1,0,0,1,0,1,UB驻留(AIV) AIV单缓冲,0,0,0,allocate,direct_gm,0,0,不涉及(AIV逐元素),1,1,1,0,0,0,0,False,2,"K=1逐元素乘: 退化为 C=A⊙B 无累加深度, Cube 16x16x16 粒度浪费 15/16; 走 AIV 通路 GM->UB->Mul->GM, AIV单缓冲 (B<2*AIV 逐batch单缓冲串行)",1114112,0.0,6.9632e-07,0.0,6.9632e-07,0.0,0.0,268435456.0,1.985939393939394e-05,536870912,0.00033554432,0.0,0.00033554432,0.0,0.00033554432,FIXPIPE,True,,写出Bound,"K=1逐元素乘, 走AIV向量通路"
b128_m2048_n2048_k1536,128,128,2048,2048,1536,bf16,bf16,bf16,False,False,False,True,0,b128_m2048_n2048_k1536,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,12,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,1536,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,576,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮",1610612736.0,17716740096,0.00100663296,0.003407065403076923,0.004413698363076923,0.0,0.0,1649267441664.0,0.0033935544067160493,1073741824,0.0002064888123076923,0.0,0.004413698363076923,0.0,0.004413698363076923,MTE2,True,,访存Bound(GM读写共享+L2重复读),"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0"
b64_m1024_n8192_k2048,64,64,1024,8192,2048,bf16,bf16,bf16,False,False,False,True,0,b64_m1024_n8192_k2048,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,6,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,2048,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,564,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮",2415919104.0,23085449216,0.00150994944,0.004439509464615385,0.005949458904615385,0.0,0.0,2199023255552.0,0.004524739208954733,1073741824,0.0002064888123076923,0.0,0.005949458904615385,0.0,0.005949458904615385,MTE2,True,,访存Bound(GM读写共享+L2重复读),"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0"
b32_m1024_n1024_k512,32,32,1024,1024,512,bf16,bf16,bf16,False,False,False,True,0,b32_m1024_n1024_k512,IterBatch,Ascend950PR,batch_mat_mul_v3,32,32,1,1,1,切B轮转分配(核间零重复读零依赖),1,1,1024,1024,512,64,1,d_两侧都切K,176,176,64,allocate(GM->L1随路驻留L2),"direct_gm(输出仅写一次,直写GM不占L2)",0,0,不涉及(核内不切M/N),1,1,1,0,0,0,0,True,2,"两侧都切K: k_L1=64, K段成对流水, batch边界天然无缝; dValueA=128B/dValueB=2048B",2097152,0.0,4.194304e-05,0.0,4.234304e-05,8,4.0000000000000003e-07,1073741824.0,7.06990501399177e-05,2097152,4.194304e-05,0.0,8.428607999999999e-05,5.078042126748971e-05,0.00013506650126748969,MTE2,True,,访存Bound(GM读写共享+L2重复读),仅 IterBatch 条件满足
b128_m4096_n4096_k8192,128,128,4096,4096,8192,bf16,bf16,bf16,False,False,False,True,0,b128_m4096_n4096_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,24,24,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,2304,True,2,"L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮",17179869184.0,395136991232,0.01073741824,0.07598788292923077,0.08672530116923077,0.0,0.0,35184372088832.0,0.07239582734327572,4294967296,0.00268435456,0.0,0.08940965572923076,0.0,0.08940965572923076,MTE2,True,,访存Bound(GM读写共享+L2重复读),"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0"
b32_m8192_n4096_k7168,32,32,8192,4096,7168,bf16,bf16,bf16,False,False,False,True,0,b32_m8192_n4096_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,24,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,1128,True,2,"L2场景C_单batch输入超L2分组执行, GM首读倍率=1.33; 尾轮: r=0 无尾轮",7516192768.0,0.0,0.00469762048,0.0,0.00469762048,0.0,0.0,15393162788864.0,0.031673174462683126,2147483648,0.00134217728,0.0,0.031673174462683126,0.0,0.031673174462683126,MMAD,True,,计算Bound,"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0"
b32_m2048_n2048_k8192,32,32,2048,2048,8192,bf16,bf16,bf16,False,False,False,True,0,b32_m2048_n2048_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,12,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,144,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮",2147483648.0,23622320128,0.00134217728,0.004542753870769231,0.005884931150769231,0.0,0.0,2199023255552.0,0.004524739208954733,268435456,5.162220307692308e-05,0.0,0.005884931150769231,0.0,0.005884931150769231,MTE2,True,,访存Bound(GM读写共享+L2重复读),"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0"
b64_m4096_n2048_k128,64,64,4096,2048,128,bf16,bf16,bf16,False,False,False,True,0,b64_m4096_n2048_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,24,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,576,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮",100663296.0,1509949440,6.291456e-05,0.0002903748923076923,0.0003532894523076923,0.0,0.0,137438953472.0,0.0002827962005596708,1073741824,0.0002064888123076923,0.0,0.0003532894523076923,0.0,0.0003532894523076923,MTE2,True,,访存Bound(GM读写共享+L2重复读),"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0"
b16_m1024_n1024_k8192,16,16,1024,1024,8192,bf16,bf16,bf16,False,False,False,True,0,b16_m1024_n1024_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,6,6,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,18,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮",536870912.0,2684354560,0.00033554432,0.0005162220307692308,0.0008517663507692308,0.0,0.0,274877906944.0,0.0005655924011193416,33554432,6.452775384615385e-06,0.0,0.0008517663507692308,0.0,0.0008517663507692308,MTE2,True,,访存Bound(GM读写共享+L2重复读),ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
b4_m32768_n128_k128,4,4,32768,128,128,bf16,bf16,bf16,False,False,False,True,0,b4_m32768_n128_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,1,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,方案B,205,1,1,205,1,12,24,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=28.96us), 选方案B工程简洁 (一套tile)",33685504.0,24379392,2.105344e-05,4.688344615384615e-06,2.5741784615384616e-05,0.0,0.0,4294967296.0,8.837381267489712e-06,33554432,6.452775384615385e-06,0.0,2.5741784615384616e-05,0.0,2.5741784615384616e-05,MTE2,True,,访存Bound(GM读写共享+L2重复读),ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
b8_m32768_n2048_k512,8,8,32768,2048,512,bf16,bf16,bf16,False,False,False,True,0,b8_m32768_n2048_k512,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,12,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,512,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,561,True,2,"L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮",285212672.0,6073352192,0.00017825792,0.0011679523446153847,0.0013462102646153848,0.0,0.0,549755813888.0,0.0011311848022386832,1073741824,0.00067108864,0.0,0.0020172989046153846,0.0,0.0020172989046153846,MTE2,True,,访存Bound(GM读写共享+L2重复读),ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
b4_m131072_n128_k128,4,4,131072,128,128,bf16,bf16,bf16,False,False,False,True,0,b4_m131072_n128_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,1,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,方案B,820,1,1,820,1,4,94,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=115.39us), 选方案B工程简洁 (一套tile)",134348800.0,97517568,8.3968e-05,1.875337846153846e-05,0.00010272137846153847,0.0,0.0,17179869184.0,3.534952506995885e-05,134217728,2.581110153846154e-05,0.0,0.00010272137846153847,0.0,0.00010272137846153847,MTE2,True,,访存Bound(GM读写共享+L2重复读),ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
b8_m131072_n1024_k256,8,8,131072,1024,256,bf16,bf16,bf16,False,False,False,True,0,b8_m131072_n1024_k256,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,6,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,256,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,820,7,1,820,7,16,1118,True,2,"L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=1384.63us), 选方案B工程简洁 (一套tile)",541065216.0,5804916736,0.00033816576,0.0011163301415384615,0.0014544959015384616,0.0,0.0,549755813888.0,0.0011311848022386832,2147483648,0.00134217728,0.0,0.0027966731815384617,0.0,0.0027966731815384617,MTE2,True,,访存Bound(GM读写共享+L2重复读),ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
b16_m32768_n8192_k7168,16,16,32768,8192,7168,bf16,bf16,bf16,False,False,False,True,0,b16_m32768_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,205,52,1,205,52,16,4395,True,2,"L2场景C_单batch输入超L2分组执行, GM首读倍率=3.20; 尾轮: 周长型主导, rho=0.50<rho_dv=0.53, A1b被dValue卡死, 方案B反超",30064771072.0,0.0,0.01879048192,0.0,0.01879048192,0.0,0.0,61572651155456.0,0.1266926978507325,8589934592,0.00536870912,0.0,0.1266926978507325,0.0,0.1266926978507325,MMAD,True,,计算Bound,ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
b4_m32768_n128_k8192,4,4,32768,128,8192,bf16,bf16,bf16,False,False,False,True,0,b4_m32768_n128_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,1,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,方案B,205,1,1,205,1,12,24,True,2,"L2场景C_单batch输入超L2分组执行, GM首读倍率=1.03; 尾轮: 周长型主导, rho=0.38<rho_dv=0.53, A1b被dValue卡死, 方案B反超",2222981120.0,0.0,0.0013893632,0.0,0.0013893632,0.0,0.0,274877906944.0,0.0005655924011193416,33554432,2.097152e-05,0.0,0.00141033472,0.0,0.00141033472,MTE2,True,,访存Bound(GM读写共享+L2重复读),ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
b32_m131072_n8192_k128,32,32,131072,8192,128,bf16,bf16,bf16,False,False,False,True,0,b32_m131072_n8192_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,35015,True,2,"L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮",1140850688.0,99321118720,0.00071303168,0.019100215138461538,0.019813246818461538,0.0,0.0,8796093022208.0,0.01809895683581893,68719476736,0.04294967296,0.0,0.06276291977846153,0.0,0.06276291977846153,MTE2,True,,访存Bound(GM读写共享+L2重复读),"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0"
b64_m32768_n8192_k1536,64,64,32768,8192,1536,bf16,bf16,bf16,False,False,False,True,0,b64_m32768_n8192_k1536,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,187,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,1536,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,17578,True,2,"L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮",8053063680.0,595926712320,0.0050331648,0.11460129083076923,0.11963445563076923,0.0,0.0,52776558133248.0,0.10859374101491358,34359738368,0.02147483648,0.0,0.14110929211076922,0.0,0.14110929211076922,MTE2,True,,访存Bound(GM读写共享+L2重复读),"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0"
b8_m131072_n8192_k8192,8,8,131072,8192,8192,bf16,bf16,bf16,False,False,False,True,0,b8_m131072_n8192_k8192,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,745,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,8192,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A1b,2,2,1,2,2,24,8754,True,2,"L2场景C_单batch输入超L2分组执行, GM首读倍率=4.76; 尾轮: 周长型主导, rho=0.75>=rho_dv=0.53, A1b恒优 (尾轮tile缩√rho=0.87倍凑满核)",86973087744.0,0.0,0.05435817984,0.0,0.05435817984,0.0,0.0,140737488355328.0,0.2895833093731029,17179869184,0.01073741824,0.0,0.2895833093731029,0.0,0.2895833093731029,MMAD,True,,计算Bound,ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
b8_m16_n7168_k1536,8,8,16,7168,1536,bf16,bf16,bf16,False,False,False,True,0,b8_m16_n7168_k1536,StreamK,Ascend950PR,batch_mat_mul_v3,32,1,1,2,2,"B/M/N切出16块, 每块2核切K归约 (归约组内核c负责K段[c*K/2,(c+1)*K/2))",1,1,16,3584,768,256,1,K段标准分块流水,16,128,64,allocate(部分和驻留L2),"resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换)",0,3670016,grid_K=2路切K+归约,1,1,2,0,0,0,0,True,4,"P=14.00, grid_K=2, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终",176553984,393216,0.00011034624,7.561846153846153e-08,0.00011042185846153846,0.0,0.0,2818572288.0,5.799531456790123e-06,0.0,0.0,2.0698331002331e-07,0.00011042185846153846,2.0698331002331e-07,0.00011062884177156177,MTE2,True,,访存Bound(GM读写共享+L2重复读),"P<=C/2, B/M/N并行度买不满, 切K (grid_K=2)"
b64_m1024_n7168_k7168,64,64,1024,7168,7168,bf16,bf16,bf16,False,False,False,True,0,b64_m1024_n7168_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,6,41,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,492,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮",7516192768.0,70464307200,0.00469762048,0.013550828307692308,0.018248448787692308,0.0,0.0,6734508720128.0,0.013857013827423869,939524096,0.00018067771076923076,0.0,0.018248448787692308,0.0,0.018248448787692308,MTE2,True,,访存Bound(GM读写共享+L2重复读),"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0"
b32_m8192_n8192_k7168,32,32,8192,8192,7168,bf16,bf16,bf16,False,False,False,True,0,b32_m8192_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,2209,True,2,"L2场景C_单batch输入超L2分组执行, GM首读倍率=2.00; 尾轮: r=0 无尾轮",15032385536.0,0.0,0.00939524096,0.0,0.00939524096,0.0,0.0,30786325577728.0,0.06334634892536625,4294967296,0.00268435456,0.0,0.06334634892536625,0.0,0.06334634892536625,MMAD,True,,计算Bound,"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0"
b8_m4096_n4096_k128,8,8,4096,4096,128,bf16,bf16,bf16,False,False,False,True,0,b8_m4096_n4096_k128,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,24,24,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,128,128,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),resident(输出驻留L2异步回写),4,0,A0,1,1,1,0,0,0,144,True,2,"L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮",16777216.0,385875968,1.048576e-05,7.420691692307692e-05,8.469267692307693e-05,0.0,0.0,34359738368.0,7.06990501399177e-05,268435456,5.162220307692308e-05,0.0,8.469267692307693e-05,0.0,8.469267692307693e-05,MTE2,True,,访存Bound(GM读写共享+L2重复读),ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
b128_m8192_n8192_k7168,128,128,8192,8192,7168,bf16,bf16,bf16,False,False,False,True,0,b128_m8192_n8192_k7168,ASW_Basic,Ascend950PR,batch_mat_mul_v3,32,1,47,47,1,B->M->N线性映射+ASW滑窗蛇形(W=4),0,1,176,176,7168,256,1,双缓冲驻留当前tile输入,176,176,80,allocate(输入驻留L2吸收重复读),direct_gm(输出直写GM不占L2),4,0,A0,1,1,1,0,0,0,8836,True,2,"L2场景C_单batch输入超L2分组执行, GM首读倍率=2.00; 尾轮: r=0 无尾轮",60129542144.0,0.0,0.03758096384,0.0,0.03758096384,0.0,0.0,123145302310912.0,0.253385395701465,17179869184,0.01073741824,0.0,0.253385395701465,0.0,0.253385395701465,MMAD,True,,计算Bound,"IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0"
special_k1_b64,64,64,8192,512,1,bf16,bf16,bf16,False,False,False,True,0,special_k1_b64,特殊分支,Ascend950PR,batch_mat_mul_v3,64,1,1,1,1,AIV 核间按行均分 (无 Cube tile 概念),0,1,0,0,1,0,1,UB驻留(AIV) AIV单缓冲,0,0,0,allocate,direct_gm,0,0,不涉及(AIV逐元素),1,1,1,0,0,0,0,False,2,"K=1逐元素乘: 退化为 C=A⊙B 无累加深度, Cube 16x16x16 粒度浪费 15/16; 走 AIV 通路 GM->UB->Mul->GM, AIV单缓冲 (B<2*AIV 逐batch单缓冲串行)",1114112,0.0,6.9632e-07,0.0,6.9632e-07,0.0,0.0,268435456.0,1.985939393939394e-05,536870912,0.00033554432,0.0,0.00033624063999999996,0.0,0.00033624063999999996,MTE2,True,,访存Bound(GM读写共享+L2重复读),"K=1逐元素乘, 走AIV向量通路"
1 case_id batch_a batch_b m n k dtype_a dtype_b dtype_c trans_a trans_b has_bias out_nd deterministic_level plan_case_id plan_branch plan_npu plan_op plan_used_core_num plan_split_b plan_m_cnt plan_n_cnt plan_grid_k plan_core_map plan_b_core plan_merge_b0 plan_single_core_m plan_single_core_n plan_single_core_k plan_k_l1 plan_b_l1 plan_l1_form plan_base_m plan_base_n plan_base_k plan_l2_policy_in plan_l2_policy_out plan_swizzle_w plan_workspace_bytes plan_tail_strategy plan_tail_m_cnt plan_tail_n_cnt plan_tail_k_cnt plan_tail_m_main plan_tail_n_main plan_tail_block_cnt plan_tail_wave_num plan_fixpipe_unitflag plan_out_dtype_bytes plan_note gm_read_bytes l2_read_bytes t_mte2_gm t_mte2_l2 t_mte2 dma_cmd_count t_dma_cmd cube_flops t_mmad fixpipe_bytes t_fixpipe t_reduce t_steady t_drain t_total bottleneck feasible violations bound_type advice
2 to_matmul_demo 1 1 2048 2048 2048 bf16 bf16 bf16 False False False True 0 to_matmul_demo 转Matmul Ascend950PR batch_mat_mul_v3 32 1 0 0 1 折叠为 Matmul [2048,2048]x[2048,2048], 复用 Matmul 切分体系 0 1 0 0 2048 0 1 0 0 0 0 0 转Matmul后由 Matmul 体系决定 1 1 1 0 0 0 0 True 2 BatchB=1免费折叠: 左矩阵 [1,2048,2048] 视图折叠为 [2048,2048], 零重排零 split 16777216 0.0 1.048576e-05 0.0 1.048576e-05 0.0 0.0 17179869184.0 3.534952506995885e-05 8388608 5.24288e-06 0.0 3.534952506995885e-05 0.0 3.534952506995885e-05 MMAD True 计算Bound BatchA=1或BatchB=1, 折叠转普通Matmul
3 special_k0_demo 128 128 256 256 0 bf16 bf16 bf16 False False False True 0 special_k0_demo 特殊分支 Ascend950PR batch_mat_mul_v3 64 1 1 1 1 AIV 核间按行均分 (无 Cube tile 概念) 0 1 0 0 0 0 1 UB驻留(AIV) 0 0 0 allocate direct_gm 0 0 不涉及(AIV逐元素) 1 1 1 0 0 0 0 False 2 K=0纯写值: 无任何计算, C=bias 或 0, 纯 AIV 写值; 按行均分到 AIV 核 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 16777216 1.048576e-05 0.0 1.048576e-05 0.0 1.048576e-05 FIXPIPE MTE2 True 写出Bound 访存Bound(GM读写共享+L2重复读) K=0纯写值
4 special_k1_demo 128 128 256 256 1 bf16 bf16 bf16 False False False True 0 special_k1_demo 特殊分支 Ascend950PR batch_mat_mul_v3 64 1 1 1 1 AIV 核间按行均分 (无 Cube tile 概念) 0 1 0 0 1 0 1 UB驻留(AIV) UB乒乓 0 0 0 allocate direct_gm 0 0 不涉及(AIV逐元素) 1 1 1 0 0 0 0 False 2 K=1逐元素乘: 退化为 C=A⊙B 无累加深度, Cube 16x16x16 粒度浪费 15/16; 走 AIV 通路 GM->UB->Mul->GM, UB乒乓 (B>=2*AIV 双batch乒乓流水) 131072 0.0 8.192e-08 0.0 8.192e-08 0.0 0.0 8388608.0 6.206060606060606e-07 16777216 1.048576e-05 0.0 1.048576e-05 1.0567679999999999e-05 0.0 1.048576e-05 1.0567679999999999e-05 FIXPIPE MTE2 True 写出Bound 访存Bound(GM读写共享+L2重复读) K=1逐元素乘, 走AIV向量通路
5 merge_demo_k_trunc 2048 2048 32 32 256 bf16 bf16 bf16 False False False True 0 merge_demo_k_trunc MergeBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B均分(核间零重复读零依赖) 64 4 128 128 256 256 8 合并驻留 128 128 128 allocate(GM->L1随路驻留L2) direct_gm 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 b0=4 (L0C上限5.7/算存比上限19.0/b_core=64); K截断; 合并后单次DMA搬入 A'[128,256]+B'[256,128] 2097152 0.0 4.194304e-05 0.0 4.2743039999999997e-05 16.0 8.000000000000001e-07 134217728.0 8.837381267489712e-06 131072 2.62144e-06 0.0 4.2743039999999997e-05 4.5364479999999994e-05 3.0192408230452674e-07 4.3044964082304525e-05 4.566640408230452e-05 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 两分支均合法, 仲裁: [分界条件] MergeBatch最优=True (k_L1=K(截断); b_core=64 vs 阈值 b0*(T_comp+T_write)/T_cmd=6.0; drain惩罚=(b0-1)*(T_comp+T_write)=0.23us, 搬移节省=b_core*(1-1/b0)*T_cmd=2.40us); [时延模型] T_MergeBatch=43.04us vs T_IterBatch=45.22us -> MergeBatch更优; [裁决] MergeBatch 两分支均合法, 仲裁: [分界条件] MergeBatch最优=True (k_L1=K(截断); b_core=64 vs 阈值 b0*(T_comp+T_write)/T_cmd=6.0; drain惩罚=(b0-1)*(T_comp+T_write)=0.23us, 搬移节省=b_core*(1-1/b0)*T_cmd=2.40us); [时延模型] T_MergeBatch=45.67us vs T_IterBatch=47.84us -> MergeBatch更优; [裁决] MergeBatch
6 merge_iter_arbitrate 128 128 64 64 512 bf16 bf16 bf16 False False False True 0 merge_iter_arbitrate IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 4 1 64 64 512 512 2 b_双batch乒乓 64 64 256 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 双batch乒乓: 2*(MK+KN)*dtype=256KB <= L1 524288 0.0 1.048576e-05 0.0 1.0685759999999999e-05 4 2.0000000000000002e-07 16777216.0 1.104672658436214e-06 32768 6.5536e-07 0.0 1.0685759999999999e-05 1.1341119999999999e-05 4.400081646090535e-07 1.1125768164609053e-05 1.1781128164609052e-05 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 两分支均合法, 仲裁: [分界条件] MergeBatch最优=False (k_L1=K(截断); b_core=4 vs 阈值 b0*(T_comp+T_write)/T_cmd=17.6; drain惩罚=(b0-1)*(T_comp+T_write)=0.44us, 搬移节省=b_core*(1-1/b0)*T_cmd=0.10us); [时延模型] T_MergeBatch=11.49us vs T_IterBatch=11.13us -> IterBatch更优; [裁决] IterBatch 两分支均合法, 仲裁: [分界条件] MergeBatch最优=False (k_L1=K(截断); b_core=4 vs 阈值 b0*(T_comp+T_write)/T_cmd=17.6; drain惩罚=(b0-1)*(T_comp+T_write)=0.44us, 搬移节省=b_core*(1-1/b0)*T_cmd=0.10us); [时延模型] T_MergeBatch=12.14us vs T_IterBatch=11.78us -> IterBatch更优; [裁决] IterBatch
7 iter_demo_form_b 128 128 64 64 256 bf16 bf16 bf16 False False False True 0 iter_demo_form_b IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 4 1 64 64 256 256 2 b_双batch乒乓 64 64 256 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 双batch乒乓: 2*(MK+KN)*dtype=128KB <= L1 262144 0.0 5.24288e-06 0.0 5.4428799999999995e-06 4 2.0000000000000002e-07 8388608.0 5.52336329218107e-07 32768 6.5536e-07 0.0 5.4428799999999995e-06 6.09824e-06 3.0192408230452674e-07 5.744804082304526e-06 6.400164082304526e-06 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 仅 IterBatch 条件满足
8 iter_demo_form_d 64 64 64 64 8192 bf16 bf16 bf16 False False False True 0 iter_demo_form_d IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 2 1 64 64 8192 1024 1 d_两侧都切K 64 64 256 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 两侧都切K: k_L1=1024, K段成对流水, batch边界天然无缝; dValueA=2048B/dValueB=128B 4194304 0.0 8.388608e-05 0.0 8.468608e-05 16 8.000000000000001e-07 134217728.0 8.837381267489712e-06 16384 3.2768e-07 0.0 8.468608e-05 8.501376e-05 7.16176329218107e-07 8.540225632921811e-05 8.572993632921812e-05 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 仅 IterBatch 条件满足
9 streamk_demo 4 4 128 128 10240 bf16 bf16 bf16 False False False True 0 streamk_demo StreamK Ascend950PR batch_mat_mul_v3 32 1 1 1 32 B/M/N切出4块, 每块32核切K归约 (归约组内核c负责K段[c*K/32,(c+1)*K/32)) 1 1 128 128 320 256 1 K段标准分块流水 128 128 64 allocate(部分和驻留L2) resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换) 0 8388608 grid_K=32路切K+归约 1 1 32 0 0 0 0 True 4 P=1.00, grid_K=32, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终 327680.0 20971520 0.0 0 6.5536e-06 1.31072e-05 0.0 6.5536e-06 1.31072e-05 0.0 0.0 41943040.0 1342177280.0 2.761681646090535e-06 0.0 0.0 3.4067453613053613e-06 6.5536e-06 1.31072e-05 3.4067453613053613e-06 9.960345361305361e-06 1.651394536130536e-05 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) P<=C/2, B/M/N并行度买不满, 切K (grid_K=32)
10 asw_demo_full 2 2 8192 8192 1024 bf16 bf16 bf16 False False False True 0 asw_demo_full ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 47 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 1024 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 方案B 52 52 1 52 52 2 139 True 2 L2场景B_输入驻留输出直写GM, r_in=1.00; 尾轮: 周长型主导, rho=0.06<rho_dv=0.53, A1b被dValue卡死, 方案B反超 L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: 周长型主导, rho=0.06<rho_dv=0.53, A1b被dValue卡死, 方案B反超 67108864.0 0.0 3087007744 4.194304e-05 0.0 0.0005936553353846154 4.194304e-05 0.0006355983753846153 0.0 0.0 274877906944.0 0.0005655924011193416 268435456 0.00016777216 0.0 0.0005655924011193416 0.0008033705353846154 0.0 0.0005655924011193416 0.0008033705353846154 MMAD MTE2 True 计算Bound 访存Bound(GM读写共享+L2重复读) ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
11 asw_demo_reduce_core 16 16 256 256 128 bf16 bf16 bf16 False False False True 0 asw_demo_reduce_core ASW_Basic_降核 Ascend950PR batch_mat_mul_v3 16 1 1 1 1 降核: 只用16核, 每核一个L0C满载输出块, 其余核闲置 0 1 256 256 128 128 1 标准核内流水 256 256 64 allocate direct_gm 0 0 不涉及(每核一块无尾轮) 1 1 1 0 0 0 0 True 2 P=16.00<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢) 2097152.0 0.0 0 2.62144e-06 0.0 2.62144e-06 0.0 0.0 268435456.0 1.104672658436214e-06 2097152 2.62144e-06 8.065969230769231e-07 0.0 2.62144e-06 0.0 2.62144e-06 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) ASW_Basic兜底 (StreamK未过: 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2)
12 b4_m1_n128_k256 4 4 1 128 256 bf16 bf16 bf16 False False False True 0 b4_m1_n128_k256 ASW_Basic_降核 Ascend950PR batch_mat_mul_v3 1 1 1 1 1 降核: 只用1核, 每核一个L0C满载输出块, 其余核闲置 0 1 1 128 256 256 1 标准核内流水 1 128 128 allocate direct_gm 0 0 不涉及(每核一块无尾轮) 1 1 1 0 0 0 0 True 2 P=0.01<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢) 264192.0 0.0 0 5.28384e-06 0.0 5.28384e-06 0.0 0.0 262144.0 1.7260510288065844e-08 1024 2.048e-08 6.3015384615384615e-09 0.0 5.28384e-06 0.0 5.28384e-06 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) ASW_Basic兜底 (StreamK未过: 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受: K > grid_K^2/(grid_K-1)*theta_c)
13 b8_m2_n192_k128 8 8 2 192 128 bf16 bf16 bf16 False False False True 0 b8_m2_n192_k128 ASW_Basic_降核 Ascend950PR batch_mat_mul_v3 1 1 1 1 1 降核: 只用1核, 每核一个L0C满载输出块, 其余核闲置 0 1 2 192 128 128 1 标准核内流水 2 192 80 allocate direct_gm 0 0 不涉及(每核一块无尾轮) 1 1 1 0 0 0 0 True 2 P=0.05<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢) 397312.0 0.0 0 7.94624e-06 0.0 7.94624e-06 0.0 0.0 786432.0 5.178153086419753e-08 6144 1.2288e-07 3.7809230769230766e-08 0.0 7.94624e-06 0.0 7.94624e-06 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) ASW_Basic兜底 (StreamK未过: 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受: K > grid_K^2/(grid_K-1)*theta_c)
14 b16_m4_n256_k192 16 16 4 256 192 bf16 bf16 bf16 False False False True 0 b16_m4_n256_k192 ASW_Basic_降核 Ascend950PR batch_mat_mul_v3 1 1 1 1 1 降核: 只用1核, 每核一个L0C满载输出块, 其余核闲置 0 1 4 256 192 192 1 标准核内流水 4 256 64 allocate direct_gm 0 0 不涉及(每核一块无尾轮) 1 1 1 0 0 0 0 True 2 P=0.25<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢) 1597440.0 0.0 0 3.19488e-05 0.0 3.19488e-05 0.0 0.0 6291456.0 4.1425224691358024e-07 32768 6.5536e-07 2.0164923076923077e-07 0.0 3.19488e-05 0.0 3.19488e-05 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) ASW_Basic兜底 (StreamK未过: 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受: K > grid_K^2/(grid_K-1)*theta_c)
15 b32_m8_n128_k256 32 32 8 128 256 bf16 bf16 bf16 False False False True 0 b32_m8_n128_k256 IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 1 1 8 128 256 256 1 a_单batch全驻留 8 128 128 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 单batch全驻留: (MK+KN)*dtype=68KB <= L1 69632 0.0 1.39264e-06 0.0 1.44264e-06 1 5.0000000000000004e-08 524288.0 3.452102057613169e-08 2048 4.096e-08 0.0 1.44264e-06 1.4836e-06 7.548102057613169e-08 1.5181210205761316e-06 1.5590810205761317e-06 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 仅 IterBatch 条件满足
16 b64_m16_n256_k512 64 64 16 256 512 bf16 bf16 bf16 False False False True 0 b64_m16_n256_k512 IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 2 1 16 256 512 240 1 c_一侧驻留+对侧切K 16 256 64 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 一侧驻留(A)+对侧切K: A驻留16KB, 预算L1/2, k_L1=240, dValueA=480B/dValueB=512B 557056 0.0 1.56672e-05 0.0 1.5967200000000002e-05 6 3.0000000000000004e-07 8388608.0 5.52336329218107e-07 16384 3.2768e-07 0.0 1.5967200000000002e-05 1.629488e-05 2.932938271604938e-07 1.6260493827160496e-05 1.6588173827160495e-05 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 仅 IterBatch 条件满足
17 b128_m8_n192_k256 128 128 8 192 256 bf16 bf16 bf16 False False False True 0 b128_m8_n192_k256 IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 4 1 8 192 256 256 2 b_双batch乒乓 8 192 80 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 双batch乒乓: 2*(MK+KN)*dtype=200KB <= L1 409600 0.0 8.192e-06 0.0 8.392e-06 4 2.0000000000000002e-07 3145728.0 2.0712612345679012e-07 12288 2.4576e-07 0.0 8.392e-06 8.637760000000001e-06 1.1322153086419754e-07 8.505221530864198e-06 8.7509815308642e-06 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 仅 IterBatch 条件满足
18 b32_m16_n8192_k7168 32 32 16 8192 7168 bf16 bf16 bf16 False False False True 0 b32_m16_n8192_k7168 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 1 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 7168 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 47 True 2 L2场景C_输入超L2分组执行, r_in=1.09; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮 4103077888.0000005 3765436416.0 0.0 337641472 0.0025644236800000005 0.00235339776 0.0 6.493105230769231e-05 0.0025644236800000005 0.0024183288123076924 0.0 0.0 60129542144.0 0.00012372333774485596 8388608 5.24288e-06 1.6131938461538462e-06 0.0 0.0025644236800000005 0.0024183288123076924 0.0 0.0025644236800000005 0.0024183288123076924 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B
19 b4_m1_n8192_k8192 4 4 1 8192 8192 bf16 bf16 bf16 False False False True 0 b4_m1_n8192_k8192 StreamK Ascend950PR batch_mat_mul_v3 32 1 1 1 32 B/M/N切出4块, 每块32核切K归约 (归约组内核c负责K段[c*K/32,(c+1)*K/32)) 1 1 1 8192 256 256 1 K段标准分块流水 1 128 64 allocate(部分和驻留L2) resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换) 0 4194304 grid_K=32路切K+归约 1 1 32 0 0 0 0 True 4 P=0.50, grid_K=32, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终 262144.0 536936448 0.0 5.24288e-06 0.00033558528 0.0 5.24288e-06 0.00033558528 0.0 0.0 33554432.0 536870912.0 2.209345316872428e-06 1.104672658436214e-06 0.0 0.0 3.4067453613053613e-06 1.7033726806526807e-06 5.24288e-06 0.00033558528 3.4067453613053613e-06 1.7033726806526807e-06 8.64962536130536e-06 0.00033728865268065273 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) P<=C/2, B/M/N并行度买不满, 切K (grid_K=32)
20 b16_m2_n4096_k7168 16 16 2 4096 7168 bf16 bf16 bf16 False False False True 0 b16_m2_n4096_k7168 StreamK Ascend950PR batch_mat_mul_v3 32 1 1 1 16 B/M/N切出16块, 每块16核切K归约 (归约组内核c负责K段[c*K/16,(c+1)*K/16)) 1 1 2 4096 448 256 1 K段标准分块流水 2 128 64 allocate(部分和驻留L2) resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换) 0 8388608 grid_K=16路切K+归约 1 1 16 0 0 0 0 True 4 P=2.00, grid_K=16, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终 458752.0 939982848 0.0 0 9.17504e-06 0.00058748928 0.0 9.17504e-06 0.00058748928 0.0 0.0 58720256.0 1879048192.0 3.866354304526749e-06 0.0 0.0 1.7159757575757577e-06 9.17504e-06 0.00058748928 1.7159757575757577e-06 1.0891015757575759e-05 0.0005892052557575758 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) P<=C/2, B/M/N并行度买不满, 切K (grid_K=16)
21 b32_m64_n64_k7168 32 32 64 64 7168 bf16 bf16 bf16 False False False True 0 b32_m64_n64_k7168 IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 1 1 64 64 7168 1024 1 d_两侧都切K 64 64 256 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 两侧都切K: k_L1=1024, K段成对流水, batch边界天然无缝; dValueA=2048B/dValueB=128B 1835008 0.0 3.6700159999999995e-05 0.0 3.705016e-05 7 3.5000000000000004e-07 58720256.0 3.866354304526749e-06 8192 1.6384e-07 0.0 3.705016e-05 3.7214e-05 7.16176329218107e-07 3.77663363292181e-05 3.7930176329218104e-05 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) 仅 IterBatch 条件满足
22 b64_m1024_n1024_k7168 64 64 1024 1024 7168 bf16 bf16 bf16 False False False True 0 b64_m1024_n1024_k7168 IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 2 1 1024 1024 7168 64 1 d_两侧都切K 176 176 64 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 两侧都切K: k_L1=64, K段成对流水, batch边界天然无缝; dValueA=128B/dValueB=2048B 58720256 0.0 0.0011744051199999998 0.0 0.00118560512 224 1.1200000000000001e-05 30064771072.0 0.0019795734039176954 4194304 8.388608e-05 0.0 0.0019795734039176954 5.078042126748971e-05 0.0020303538251851853 MMAD True 计算Bound 仅 IterBatch 条件满足
23 b128_m2048_n2048_k1536 128 128 2048 2048 1536 bf16 bf16 bf16 False False False True 0 b128_m2048_n2048_k1536 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 12 12 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 1536 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 576 True 2 L2场景C_输入超L2分组执行, r_in=12.00; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮 19327352832.0 1610612736.0 0.0 17716740096 0.01207959552 0.00100663296 0.0 0.003407065403076923 0.01207959552 0.004413698363076923 0.0 0.0 1649267441664.0 0.0033935544067160493 1073741824 0.00067108864 0.0002064888123076923 0.0 0.01207959552 0.004413698363076923 0.0 0.01207959552 0.004413698363076923 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0
24 b64_m1024_n8192_k2048 64 64 1024 8192 2048 bf16 bf16 bf16 False False False True 0 b64_m1024_n8192_k2048 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 6 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 2048 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 564 True 2 L2场景C_输入超L2分组执行, r_in=10.56; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮 25501368320.0 2415919104.0 0.0 23085449216 0.0159383552 0.00150994944 0.0 0.004439509464615385 0.0159383552 0.005949458904615385 0.0 0.0 2199023255552.0 0.004524739208954733 1073741824 0.00067108864 0.0002064888123076923 0.0 0.0159383552 0.005949458904615385 0.0 0.0159383552 0.005949458904615385 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0
25 b32_m1024_n1024_k512 32 32 1024 1024 512 bf16 bf16 bf16 False False False True 0 b32_m1024_n1024_k512 IterBatch Ascend950PR batch_mat_mul_v3 32 32 1 1 1 切B轮转分配(核间零重复读零依赖) 1 1 1024 1024 512 64 1 d_两侧都切K 176 176 64 allocate(GM->L1随路驻留L2) direct_gm(输出仅写一次,直写GM不占L2) 0 0 不涉及(核内不切M/N) 1 1 1 0 0 0 0 True 2 两侧都切K: k_L1=64, K段成对流水, batch边界天然无缝; dValueA=128B/dValueB=2048B 2097152 0.0 4.194304e-05 0.0 4.234304e-05 8 4.0000000000000003e-07 1073741824.0 7.06990501399177e-05 2097152 4.194304e-05 0.0 7.06990501399177e-05 8.428607999999999e-05 5.078042126748971e-05 0.0001214794714074074 0.00013506650126748969 MMAD MTE2 True 计算Bound 访存Bound(GM读写共享+L2重复读) 仅 IterBatch 条件满足
26 b128_m4096_n4096_k8192 128 128 4096 4096 8192 bf16 bf16 bf16 False False False True 0 b128_m4096_n4096_k8192 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 24 24 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 8192 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A0 1 1 1 0 0 0 2304 True 2 L2场景C_输入超L2分组执行, r_in=24.00; 尾轮: r=0 无尾轮 L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮 412316860416.0 17179869184.0 0.0 395136991232 0.25769803776 0.01073741824 0.0 0.07598788292923077 0.25769803776 0.08672530116923077 0.0 0.0 35184372088832.0 0.07239582734327572 4294967296 0.00268435456 0.0 0.25769803776 0.08940965572923076 0.0 0.25769803776 0.08940965572923076 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0
27 b32_m8192_n4096_k7168 32 32 8192 4096 7168 bf16 bf16 bf16 False False False True 0 b32_m8192_n4096_k7168 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 47 24 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 7168 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A0 1 1 1 0 0 0 1128 True 2 L2场景C_输入超L2分组执行, r_in=31.67; 尾轮: r=0 无尾轮 L2场景C_单batch输入超L2分组执行, GM首读倍率=1.33; 尾轮: r=0 无尾轮 178509578240.0 7516192768.0 0.0 0.1115684864 0.00469762048 0.0 0.1115684864 0.00469762048 0.0 0.0 15393162788864.0 0.031673174462683126 2147483648 0.00134217728 0.0 0.1115684864 0.031673174462683126 0.0 0.1115684864 0.031673174462683126 MTE2_GM MMAD True 访存Bound(GM) 计算Bound IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0
28 b32_m2048_n2048_k8192 32 32 2048 2048 8192 bf16 bf16 bf16 False False False True 0 b32_m2048_n2048_k8192 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 12 12 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 8192 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 144 True 2 L2场景C_输入超L2分组执行, r_in=12.00; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮 25769803776.0 2147483648.0 0.0 23622320128 0.01610612736 0.00134217728 0.0 0.004542753870769231 0.01610612736 0.005884931150769231 0.0 0.0 2199023255552.0 0.004524739208954733 268435456 0.00016777216 5.162220307692308e-05 0.0 0.01610612736 0.005884931150769231 0.0 0.01610612736 0.005884931150769231 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0
29 b64_m4096_n2048_k128 64 64 4096 2048 128 bf16 bf16 bf16 False False False True 0 b64_m4096_n2048_k128 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 24 12 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 128 128 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 576 True 2 L2场景B_输入驻留输出直写GM, r_in=1.00; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮 100663296.0 0.0 1509949440 6.291456e-05 0.0 0.0002903748923076923 6.291456e-05 0.0003532894523076923 0.0 0.0 137438953472.0 0.0002827962005596708 1073741824 0.00067108864 0.0002064888123076923 0.0 0.00067108864 0.0003532894523076923 0.0 0.00067108864 0.0003532894523076923 FIXPIPE MTE2 True 写出Bound 访存Bound(GM读写共享+L2重复读) IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0
30 b16_m1024_n1024_k8192 16 16 1024 1024 8192 bf16 bf16 bf16 False False False True 0 b16_m1024_n1024_k8192 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 6 6 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 8192 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 18 True 2 L2场景C_输入超L2分组执行, r_in=6.00; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮 3221225472.0 536870912.0 0.0 2684354560 0.00201326592 0.00033554432 0.0 0.0005162220307692308 0.00201326592 0.0008517663507692308 0.0 0.0 274877906944.0 0.0005655924011193416 33554432 2.097152e-05 6.452775384615385e-06 0.0 0.00201326592 0.0008517663507692308 0.0 0.00201326592 0.0008517663507692308 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
31 b4_m32768_n128_k128 4 4 32768 128 128 bf16 bf16 bf16 False False False True 0 b4_m32768_n128_k128 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 187 1 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 128 128 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) resident(输出驻留L2异步回写) 4 0 方案B 205 1 1 205 1 12 24 True 2 L2场景A_全驻留, r_in=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=28.96us), 选方案B工程简洁 (一套tile) L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=28.96us), 选方案B工程简洁 (一套tile) 33685504.0 0.0 24379392 2.105344e-05 0.0 4.688344615384615e-06 2.105344e-05 2.5741784615384616e-05 0.0 0.0 4294967296.0 8.837381267489712e-06 33554432 6.452775384615385e-06 0.0 2.105344e-05 2.5741784615384616e-05 0.0 2.105344e-05 2.5741784615384616e-05 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
32 b8_m32768_n2048_k512 8 8 32768 2048 512 bf16 bf16 bf16 False False False True 0 b8_m32768_n2048_k512 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 187 12 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 512 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A0 1 1 1 0 0 0 561 True 2 L2场景C_输入超L2分组执行, r_in=1.24; 尾轮: r=0 无尾轮 L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮 352321536.0 285212672.0 0.0 6073352192 0.00022020096 0.00017825792 0.0 0.0011679523446153847 0.00022020096 0.0013462102646153848 0.0 0.0 549755813888.0 0.0011311848022386832 1073741824 0.00067108864 0.0 0.0011311848022386832 0.0020172989046153846 0.0 0.0011311848022386832 0.0020172989046153846 MMAD MTE2 True 计算Bound 访存Bound(GM读写共享+L2重复读) ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
33 b4_m131072_n128_k128 4 4 131072 128 128 bf16 bf16 bf16 False False False True 0 b4_m131072_n128_k128 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 745 1 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 128 128 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 方案B 820 1 1 820 1 4 94 True 2 L2场景C_输入超L2分组执行, r_in=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=115.39us), 选方案B工程简洁 (一套tile) L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=115.39us), 选方案B工程简洁 (一套tile) 134610944.0 134348800.0 0.0 97517568 8.413184e-05 8.3968e-05 0.0 1.875337846153846e-05 8.413184e-05 0.00010272137846153847 0.0 0.0 17179869184.0 3.534952506995885e-05 134217728 8.388608e-05 2.581110153846154e-05 0.0 8.413184e-05 0.00010272137846153847 0.0 8.413184e-05 0.00010272137846153847 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
34 b8_m131072_n1024_k256 8 8 131072 1024 256 bf16 bf16 bf16 False False False True 0 b8_m131072_n1024_k256 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 745 6 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 256 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 方案B 820 7 1 820 7 16 1118 True 2 L2场景C_输入超L2分组执行, r_in=1.06; 尾轮: 面积型主导, A1b与方案B严格打平(T=1384.63us), 选方案B工程简洁 (一套tile) L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: 面积型主导, A1b与方案B严格打平(T=1384.63us), 选方案B工程简洁 (一套tile) 574619648.0 541065216.0 0.0 5804916736 0.00035913728 0.00033816576 0.0 0.0011163301415384615 0.00035913728 0.0014544959015384616 0.0 0.0 549755813888.0 0.0011311848022386832 2147483648 0.00134217728 0.0 0.00134217728 0.0027966731815384617 0.0 0.00134217728 0.0027966731815384617 FIXPIPE MTE2 True 写出Bound 访存Bound(GM读写共享+L2重复读) ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
35 b16_m32768_n8192_k7168 16 16 32768 8192 7168 bf16 bf16 bf16 False False False True 0 b16_m32768_n8192_k7168 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 187 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 7168 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 方案B 205 52 1 205 52 16 4395 True 2 L2场景C_输入超L2分组执行, r_in=75.00; 尾轮: 周长型主导, rho=0.50<rho_dv=0.53, A1b被dValue卡死, 方案B反超 L2场景C_单batch输入超L2分组执行, GM首读倍率=3.20; 尾轮: 周长型主导, rho=0.50<rho_dv=0.53, A1b被dValue卡死, 方案B反超 704643072000.0 30064771072.0 0.0 0.44040192 0.01879048192 0.0 0.44040192 0.01879048192 0.0 0.0 61572651155456.0 0.1266926978507325 8589934592 0.00536870912 0.0 0.44040192 0.1266926978507325 0.0 0.44040192 0.1266926978507325 MTE2_GM MMAD True 访存Bound(GM) 计算Bound ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
36 b4_m32768_n128_k8192 4 4 32768 128 8192 bf16 bf16 bf16 False False False True 0 b4_m32768_n128_k8192 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 187 1 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 8192 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 方案B 205 1 1 205 1 12 24 True 2 L2场景C_输入超L2分组执行, r_in=1.14; 尾轮: 周长型主导, rho=0.38<rho_dv=0.53, A1b被dValue卡死, 方案B反超 L2场景C_单batch输入超L2分组执行, GM首读倍率=1.03; 尾轮: 周长型主导, rho=0.38<rho_dv=0.53, A1b被dValue卡死, 方案B反超 2466250752.0 2222981120.0 0.0 0.00154140672 0.0013893632 0.0 0.00154140672 0.0013893632 0.0 0.0 274877906944.0 0.0005655924011193416 33554432 2.097152e-05 0.0 0.00154140672 0.00141033472 0.0 0.00154140672 0.00141033472 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
37 b32_m131072_n8192_k128 32 32 131072 8192 128 bf16 bf16 bf16 False False False True 0 b32_m131072_n8192_k128 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 745 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 128 128 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A0 1 1 1 0 0 0 35015 True 2 L2场景C_输入超L2分组执行, r_in=2.88; 尾轮: r=0 无尾轮 L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮 3288334336.0 1140850688.0 0.0 99321118720 0.00205520896 0.00071303168 0.0 0.019100215138461538 0.00205520896 0.019813246818461538 0.0 0.0 8796093022208.0 0.01809895683581893 68719476736 0.04294967296 0.0 0.04294967296 0.06276291977846153 0.0 0.04294967296 0.06276291977846153 FIXPIPE MTE2 True 写出Bound 访存Bound(GM读写共享+L2重复读) IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0
38 b64_m32768_n8192_k1536 64 64 32768 8192 1536 bf16 bf16 bf16 False False False True 0 b64_m32768_n8192_k1536 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 187 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 1536 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A0 1 1 1 0 0 0 17578 True 2 L2场景C_输入超L2分组执行, r_in=75.00; 尾轮: r=0 无尾轮 L2场景B_单batch输入驻留输出直写GM, GM首读倍率=1.00; 尾轮: r=0 无尾轮 603979776000.0 8053063680.0 0.0 595926712320 0.37748736 0.0050331648 0.0 0.11460129083076923 0.37748736 0.11963445563076923 0.0 0.0 52776558133248.0 0.10859374101491358 34359738368 0.02147483648 0.0 0.37748736 0.14110929211076922 0.0 0.37748736 0.14110929211076922 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0
39 b8_m131072_n8192_k8192 8 8 131072 8192 8192 bf16 bf16 bf16 False False False True 0 b8_m131072_n8192_k8192 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 745 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 8192 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A1b 2 2 1 2 2 24 8754 True 2 L2场景C_输入超L2分组执行, r_in=44.53; 尾轮: 周长型主导, rho=0.75>=rho_dv=0.53, A1b恒优 (尾轮tile缩√rho=0.87倍凑满核) L2场景C_单batch输入超L2分组执行, GM首读倍率=4.76; 尾轮: 周长型主导, rho=0.75>=rho_dv=0.53, A1b恒优 (尾轮tile缩√rho=0.87倍凑满核) 812822560768.0 86973087744.0 0.0 0.50801410048 0.05435817984 0.0 0.50801410048 0.05435817984 0.0 0.0 140737488355328.0 0.2895833093731029 17179869184 0.01073741824 0.0 0.50801410048 0.2895833093731029 0.0 0.50801410048 0.2895833093731029 MTE2_GM MMAD True 访存Bound(GM) 计算Bound ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
40 b8_m16_n7168_k1536 8 8 16 7168 1536 bf16 bf16 bf16 False False False True 0 b8_m16_n7168_k1536 StreamK Ascend950PR batch_mat_mul_v3 32 1 1 2 2 B/M/N切出16块, 每块2核切K归约 (归约组内核c负责K段[c*K/2,(c+1)*K/2)) 1 1 16 3584 768 256 1 K段标准分块流水 16 128 64 allocate(部分和驻留L2) resident(部分和4B驻留L2, 防精度丢失不随C的fp16/fp8转换) 0 3670016 grid_K=2路切K+归约 1 1 2 0 0 0 0 True 4 P=14.00, grid_K=2, 部分和驻留L2按4B写出, AIV归约后按C dtype=2B写最终 786432.0 176553984 0.0 393216 1.572864e-05 0.00011034624 0.0 7.561846153846153e-08 1.572864e-05 0.00011042185846153846 0.0 0.0 100663296.0 2818572288.0 6.628035950617284e-06 5.799531456790123e-06 0.0 0.0 2.3655235431235433e-07 2.0698331002331e-07 1.572864e-05 0.00011042185846153846 2.3655235431235433e-07 2.0698331002331e-07 1.5965192354312353e-05 0.00011062884177156177 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) P<=C/2, B/M/N并行度买不满, 切K (grid_K=2)
41 b64_m1024_n7168_k7168 64 64 1024 7168 7168 bf16 bf16 bf16 False False False True 0 b64_m1024_n7168_k7168 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 6 41 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 7168 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 492 True 2 L2场景C_输入超L2分组执行, r_in=10.38; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮 77980499968.0 7516192768.0 0.0 70464307200 0.04873781248 0.00469762048 0.0 0.013550828307692308 0.04873781248 0.018248448787692308 0.0 0.0 6734508720128.0 0.013857013827423869 939524096 0.00058720256 0.00018067771076923076 0.0 0.04873781248 0.018248448787692308 0.0 0.04873781248 0.018248448787692308 MTE2_GM MTE2 True 访存Bound(GM) 访存Bound(GM读写共享+L2重复读) IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0
42 b32_m8192_n8192_k7168 32 32 8192 8192 7168 bf16 bf16 bf16 False False False True 0 b32_m8192_n8192_k7168 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 47 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 7168 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A0 1 1 1 0 0 0 2209 True 2 L2场景C_输入超L2分组执行, r_in=47.00; 尾轮: r=0 无尾轮 L2场景C_单batch输入超L2分组执行, GM首读倍率=2.00; 尾轮: r=0 无尾轮 353261060096.0 15032385536.0 0.0 0.22078816256 0.00939524096 0.0 0.22078816256 0.00939524096 0.0 0.0 30786325577728.0 0.06334634892536625 4294967296 0.00268435456 0.0 0.22078816256 0.06334634892536625 0.0 0.22078816256 0.06334634892536625 MTE2_GM MMAD True 访存Bound(GM) 计算Bound IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 1_batch关系与每核份额: BatchA==BatchB 且 b_core=B/C>=2*b0; 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0
43 b8_m4096_n4096_k128 8 8 4096 4096 128 bf16 bf16 bf16 False False False True 0 b8_m4096_n4096_k128 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 24 24 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 128 128 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) resident(输出驻留L2异步回写) 4 0 A0 1 1 1 0 0 0 144 True 2 L2场景B_输入驻留输出直写GM, r_in=1.00; 尾轮: r=0 无尾轮 L2场景A_单batch全驻留(输入+输出), GM首读倍率=1.00; 尾轮: r=0 无尾轮 16777216.0 0.0 385875968 1.048576e-05 0.0 7.420691692307692e-05 1.048576e-05 8.469267692307693e-05 0.0 0.0 34359738368.0 7.06990501399177e-05 268435456 0.00016777216 5.162220307692308e-05 0.0 0.00016777216 8.469267692307693e-05 0.0 0.00016777216 8.469267692307693e-05 FIXPIPE MTE2 True 写出Bound 访存Bound(GM读写共享+L2重复读) ASW_Basic兜底 (StreamK未过: 1_并行缺口: P=B*MN*4B/L0C <= C/2; 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受)
44 b128_m8192_n8192_k7168 128 128 8192 8192 7168 bf16 bf16 bf16 False False False True 0 b128_m8192_n8192_k7168 ASW_Basic Ascend950PR batch_mat_mul_v3 32 1 47 47 1 B->M->N线性映射+ASW滑窗蛇形(W=4) 0 1 176 176 7168 256 1 双缓冲驻留当前tile输入 176 176 80 allocate(输入驻留L2吸收重复读) direct_gm(输出直写GM不占L2) 4 0 A0 1 1 1 0 0 0 8836 True 2 L2场景C_输入超L2分组执行, r_in=47.00; 尾轮: r=0 无尾轮 L2场景C_单batch输入超L2分组执行, GM首读倍率=2.00; 尾轮: r=0 无尾轮 1413044240384.0 60129542144.0 0.0 0.88315265024 0.03758096384 0.0 0.88315265024 0.03758096384 0.0 0.0 123145302310912.0 0.253385395701465 17179869184 0.01073741824 0.0 0.88315265024 0.253385395701465 0.0 0.88315265024 0.253385395701465 MTE2_GM MMAD True 访存Bound(GM) 计算Bound IterBatch/MergeBatch 进入条件均不满足, 回落 ASW_Basic; IterBatch未过: 3_L1驻留形态(四选一, 核心要求: 单batch核内零重复读); MergeBatch未过: 2_L0C容量: 2*(b0*M)*(b0*N)*4B <= L0C; 2b_L0AB容量: 合并后 b0*M/b0*N 在最小K粒度下可驻留 L0A/L0B; 5_访存Bound: 2MN/(M+N) < R16/b0
45 special_k1_b64 64 64 8192 512 1 bf16 bf16 bf16 False False False True 0 special_k1_b64 特殊分支 Ascend950PR batch_mat_mul_v3 64 1 1 1 1 AIV 核间按行均分 (无 Cube tile 概念) 0 1 0 0 1 0 1 UB驻留(AIV) AIV单缓冲 0 0 0 allocate direct_gm 0 0 不涉及(AIV逐元素) 1 1 1 0 0 0 0 False 2 K=1逐元素乘: 退化为 C=A⊙B 无累加深度, Cube 16x16x16 粒度浪费 15/16; 走 AIV 通路 GM->UB->Mul->GM, AIV单缓冲 (B<2*AIV 逐batch单缓冲串行) 1114112 0.0 6.9632e-07 0.0 6.9632e-07 0.0 0.0 268435456.0 1.985939393939394e-05 536870912 0.00033554432 0.0 0.00033554432 0.00033624063999999996 0.0 0.00033554432 0.00033624063999999996 FIXPIPE MTE2 True 写出Bound 访存Bound(GM读写共享+L2重复读) K=1逐元素乘, 走AIV向量通路

9
BMM/BMM_Theory/tests/test_branches.py
View File

@@ -134,7 +134,7 @@ class TestTimingSanity(unittest.TestCase):
# MergeBatch case 必为访存 Bound (进入条件 5)
case = mkcase(128, 64, 64, 512)
mb = MergeBatchBranch().analyze(case)
self.assertIn(mb.timing.bottleneck, ("MTE2_GM", "MTE2_L2"))
self.assertIn(mb.timing.bottleneck, ("MTE2", "MMAD"))
def test_fixpipe_dtype_conversion(self):
# C 指定 fp16 输出时, 写出量按 2B 而非 L0C 的 4B
@@ -224,10 +224,9 @@ class TestIssueRegression2(unittest.TestCase):
t = sk.timing
self.assertAlmostEqual(t.t_fixpipe, 0.0) # 归约串行口径下无稳态 fixpipe 账
self.assertAlmostEqual(t.fixpipe_bytes, 0.0)
# 端到端 = max(MTE2, MMAD) + 归约, 不再虚高到 55us/FIXPIPE
expect = max(t.t_mte2, t.t_mmad) + t.t_reduce
self.assertAlmostEqual(t.t_total, expect)
self.assertEqual(t.bottleneck, "MTE2_GM")
# 端到端 = 稳态(MTE2搬移链) + 归约, fixpipe 无重复账
self.assertEqual(t.bottleneck, "MTE2")
self.assertAlmostEqual(t.t_total, t.t_steady + t.t_drain)
def test_issue12_k1_pingpong_still_ok(self):
# issue#12/#17: K=1 且 B>=128 仍走 UB 乒乓 (原行为不变)