Fix #27-#30: dtype感知算力(Cube/AIV速率表) / GM首读下限与整芯片字节列口径+设计文档 / Fixpipe输出落点R4(整case驻留L2否则直写GM) / MergeBatch Cube公式复核注释
- docs/05_L2驻留GM读写与dtype算力口径_设计分析.md: R1-R6公理、S_A/S_B/S_C场景、输出落点R4、dtype速率表(白皮书出处+待标定假设)、逐分支GM/L2归属表 (issue#29/#30 先文档后代码) - hardware: CUBE_DTYPE_FACTOR(f16/bf16=1, fp8=2x, fp4=4x, fp32=1/2假设) + AIV_DTYPE_FACTOR + q_cube/aiv_elem_rate (issue#28) - 全分支 t_mmad/t_comp/drain/尾轮主导项/θ_c/R16 语义按输入dtype取算力; 混精度取慢侧; StreamK归约保持fp32(AIV fp32部分和) - Fixpipe输出落点R4: to_l2 <=> V_in+V_out(+workspace)<=L2; 否则直写GM计入共享总线; ASW场景改S_A整case全驻留(原单batch驻留判定漏计整case输出累积逐出) - 字节列统一整芯片口径(gm/l2/fix/cube_flops), dma_cmd_count注明单核; GM>=V_in不变量入测试; MergeBatch每步flops=2(b0M)(b0N)K公式注释显式化(#27复核与CSV一致无数值改动) - tests 39->49 全过; 双压力seed7/6000+seed2024/4000: 0违规/0占位/0NaN/0GM<输入; examples三件套重生成且可复现0diff
This commit is contained in:
@@ -54,6 +54,9 @@ python -m unittest discover -s tests -v
|
||||
- **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);
|
||||
- **GM 读取下限**: 每输入字节至少从 GM 读一次 (R1); L2 只吸收"驻留后的再次读取"; 整 case 输入+输出 ≤ L2 时 GM 恰读一次、输出全驻留 L2 (issue#29 设计文档 docs/05);
|
||||
- **Fixpipe 输出落点**: 整 case (输入+输出+workspace) 可驻留 L2 → 输出写 L2 写口 5.2TB/s、GM 写流量 0; 否则输入优先保 L2, 输出直写 GM 计入共享总线 (issue#30);
|
||||
- **算力按输入 dtype**: Cube BF16/FP16 486TFLOPS 为基准, fp8=2x/fp4=4x (白皮书), fp32=1/2 (假设待标定); AIV 逐元素通量同理 (issue#28);
|
||||
- **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
|
||||
@@ -89,6 +92,7 @@ BMM_Theory/
|
||||
│ │ ├── 06_ASW_Basic分支.md # 尾轮策略已内化为其必要环节
|
||||
│ │ └── 07_尾轮处理策略.md # 尾轮完整推导 (参考)
|
||||
│ └── 03_测评报告/ # 外部测评报告 (v1.0/v2.0 及后续复评)
|
||||
│ └── 05_L2驻留GM读写与dtype算力口径_设计分析.md # GM/L2/输出落点/dtype算力统一口径 (issue#27-#30)
|
||||
├── examples/ # 示例输入输出
|
||||
└── tests/ # 单元测试 (固化文档边界 case + issue 回归)
|
||||
```
|
||||
|
||||
@@ -80,22 +80,14 @@ 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 (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 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_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_gm(case, single_m, single_n)
|
||||
# Step 5: L2 场景判定 (issue#24/#30, 设计文档 docs/05 §4, 芯片 L2 128MB)
|
||||
# S_A 整case全驻留: V_in+V_out <= L2 -> 输出驻留 L2 (R4, GM 写=0);
|
||||
# S_B 整case超L2但单batch输入可驻留: batch 内共享块重复读命中 L2, 输出直写 GM;
|
||||
# S_C 单batch输入超L2: 分组执行 (工作集), 组间共享块落空回 GM (r_gm).
|
||||
scene = self._l2_scene(case, single_m, single_n)
|
||||
l2_out = ("resident(整case全驻留S_A: 输出驻留L2异步回写, GM写=0)"
|
||||
if scene["to_l2"] else
|
||||
"direct_gm(输出直写GM, 输入优先驻留L2)")
|
||||
|
||||
# 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)
|
||||
@@ -125,7 +117,9 @@ 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}, GM首读倍率={r_in:.2f}; 尾轮: {tail['reason']}",
|
||||
note=(f"L2场景: {scene['label']} (V_in={case.input_bytes/1048576:.1f}MB, "
|
||||
f"V_out={case.output_bytes/1048576:.1f}MB, L2={s.l2_bytes/1048576:.0f}MB), "
|
||||
f"GM首读倍率={scene['r_gm']:.2f}; 尾轮: {tail['reason']}"),
|
||||
)
|
||||
|
||||
# ------------------------------------------------------------------
|
||||
@@ -143,6 +137,7 @@ class AswBasicBranch(Branch):
|
||||
single_m, single_n = clamp_base_mn_l0c(single_m, single_n, False, s)
|
||||
# base_k 由 L0A/L0B 容量按 dtype 反推 (issue#5: 原硬编码 min(K,64) 在 fp32 下溢出)
|
||||
base_k = clamp_base_k(single_m, single_n, dt, case.k, s)
|
||||
out_l2 = case.input_bytes + case.output_bytes <= s.l2_bytes # R4 (issue#30)
|
||||
return ImplPlan(
|
||||
case_id=case.case_id, branch=self.name + "_降核",
|
||||
used_core_num=used,
|
||||
@@ -155,7 +150,9 @@ class AswBasicBranch(Branch):
|
||||
l1_form="标准核内流水",
|
||||
base_m=single_m, base_n=single_n,
|
||||
base_k=base_k,
|
||||
l2_policy_in="allocate", l2_policy_out="direct_gm",
|
||||
l2_policy_in="allocate",
|
||||
l2_policy_out=("resident(整case全驻留S_A)"
|
||||
if out_l2 else "direct_gm(输出直写GM)"),
|
||||
swizzle_w=0, workspace_bytes=0,
|
||||
tail_strategy="不涉及(每核一块无尾轮)",
|
||||
fixpipe_unitflag=True, out_dtype_bytes=case.dtype_out_bytes,
|
||||
@@ -190,6 +187,31 @@ class AswBasicBranch(Branch):
|
||||
w = d
|
||||
return w
|
||||
|
||||
def _l2_scene(self, case: BmmCase, sm: int, sn: int) -> dict:
|
||||
"""L2 场景判定 (make_plan 与 evaluate 共用, 设计文档 docs/05 §4.1).
|
||||
|
||||
判定顺序 S_A -> S_B -> S_C (L2 为整芯片 128MB):
|
||||
S_A 整case全驻留: V_in + V_out <= L2 -> 输出驻留 L2 (R4);
|
||||
S_B 单batch输入驻留: a_b + bb_b <= L2 -> 输入共享块重复读命中 L2;
|
||||
S_C 单batch输入超L2: 分组执行, 组间落空回 GM (r_gm).
|
||||
"""
|
||||
s = self.spec
|
||||
m, n, k = case.m, case.n, case.k
|
||||
dt = case.dtype_in_bytes
|
||||
a_b = m * k * dt
|
||||
bb_b = k * n * dt
|
||||
if case.input_bytes + case.output_bytes <= s.l2_bytes:
|
||||
return {"code": "S_A",
|
||||
"label": "A_整case全驻留(输入+输出<=L2)",
|
||||
"to_l2": True, "r_gm": 1.0}
|
||||
if a_b + bb_b <= s.l2_bytes:
|
||||
return {"code": "S_B",
|
||||
"label": "B_单batch输入驻留(整case超L2, 输出直写GM)",
|
||||
"to_l2": False, "r_gm": 1.0}
|
||||
return {"code": "S_C",
|
||||
"label": "C_单batch输入超L2分组执行(组间落空回GM)",
|
||||
"to_l2": False, "r_gm": self._l2_group_r_gm(case, sm, sn)}
|
||||
|
||||
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.
|
||||
|
||||
@@ -227,7 +249,8 @@ class AswBasicBranch(Branch):
|
||||
|
||||
# 主导项判定
|
||||
bw_eff = s.bw_l2_pc # L2 命中
|
||||
t_mmad = 2 * sm * sn * case.k / s.q16
|
||||
qc = s.q_cube(case.dtype_a, case.dtype_b) # issue#28
|
||||
t_mmad = 2 * sm * sn * case.k / qc
|
||||
t_mte2 = case.k * (sm + sn) * dt / bw_eff
|
||||
t_fix = sm * sn * out_b / s.bw_pc
|
||||
t_block = max(t_mmad, t_mte2, t_fix)
|
||||
@@ -268,15 +291,17 @@ class AswBasicBranch(Branch):
|
||||
|
||||
# ------------------------------------------------------------------
|
||||
def evaluate(self, case: BmmCase, plan: ImplPlan) -> HardwareTiming:
|
||||
"""MTE2 两段块级模型 (issue#24, 用户澄清):
|
||||
"""MTE2 两段块级模型 (issue#24 用户口径 + #28/#29/#30):
|
||||
|
||||
- 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).
|
||||
L2 读口 (5.2TB/s 独享) —— 场景 S_A/S_B (单 batch 工作集可驻留);
|
||||
- 单 batch 输入超 L2 -> 场景 S_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);
|
||||
- 输出落点 R4 (issue#30): 仅 S_A (整 case 输入+输出 <= L2) 驻留 L2
|
||||
(5.2 写口, GM 写 = 0); S_B/S_C 直写 GM —— GM 读写共享总线累加由
|
||||
assemble 的 MTE2 链处理 (issue#23);
|
||||
- 字节列整芯片口径 (issue#29); Cube 算力按输入 dtype (issue#28).
|
||||
"""
|
||||
s = self.spec
|
||||
b = case.batch_c
|
||||
@@ -284,37 +309,33 @@ class AswBasicBranch(Branch):
|
||||
dt = case.dtype_in_bytes
|
||||
out_b = case.dtype_out_bytes
|
||||
used = max(plan.used_core_num, 1)
|
||||
qc = s.q_cube(case.dtype_a, case.dtype_b)
|
||||
|
||||
flops = 2.0 * b * m * n * k
|
||||
t_mmad = flops / (used * s.q16)
|
||||
t_mmad = flops / (used * qc)
|
||||
|
||||
# ---- 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
|
||||
# ---- 字节量 (整芯片口径) ----
|
||||
a_b = m * k * dt # 单 batch A 字节
|
||||
bb_b = k * n * dt # 单 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:
|
||||
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)
|
||||
# ---- L2 场景 (S_A/S_B/S_C, 与 make_plan 同源) ----
|
||||
scene = self._l2_scene(case, plan.single_core_m, plan.single_core_n)
|
||||
to_l2, r_gm = scene["to_l2"], scene["r_gm"]
|
||||
|
||||
# ---- 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_")):
|
||||
if scene["code"] in ("S_A", "S_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
|
||||
l2_read = 0.0 # 场景 S_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 ----
|
||||
# ---- Fixpipe (R4) ----
|
||||
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)
|
||||
|
||||
@@ -13,7 +13,7 @@ from __future__ import annotations
|
||||
|
||||
from ..hardware import NpuSpec, ASCEND950PR
|
||||
from ..models import BmmCase, ImplPlan, HardwareTiming, align_down
|
||||
from ..timing import assemble_timing
|
||||
from ..timing import assemble_timing, output_to_l2
|
||||
from .base import Branch, BranchResult, ConditionCheck
|
||||
|
||||
|
||||
@@ -171,6 +171,12 @@ class IterBatchBranch(Branch):
|
||||
form_name = {"a": "a_单batch全驻留", "b": "b_双batch乒乓",
|
||||
"c": "c_一侧驻留+对侧切K", "d": "d_两侧都切K"}[form]
|
||||
|
||||
# 输出落点 (issue#30, R4): 整 case 输入+输出 <= L2 才驻留 L2
|
||||
out_l2 = output_to_l2(case, s)
|
||||
l2_out = ("resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)"
|
||||
if out_l2 else
|
||||
"direct_gm(整case超L2: 输入优先驻留L2, 输出直写GM不占L2)")
|
||||
|
||||
return ImplPlan(
|
||||
case_id=case.case_id, branch=self.name,
|
||||
used_core_num=s.aic_num,
|
||||
@@ -181,54 +187,63 @@ class IterBatchBranch(Branch):
|
||||
k_l1=k_l1, b_l1=2 if form == "b" else 1, l1_form=form_name,
|
||||
base_m=base_m, base_n=base_n, base_k=base_k,
|
||||
l2_policy_in="allocate(GM->L1随路驻留L2)",
|
||||
l2_policy_out="direct_gm(输出仅写一次,直写GM不占L2)",
|
||||
l2_policy_out=l2_out,
|
||||
swizzle_w=0, workspace_bytes=0,
|
||||
tail_strategy="不涉及(核内不切M/N)",
|
||||
fixpipe_unitflag=True,
|
||||
out_dtype_bytes=case.dtype_out_bytes,
|
||||
note=form_desc,
|
||||
note=form_desc + f"; 输出落点: {'L2驻留' if out_l2 else '直写GM'}",
|
||||
)
|
||||
|
||||
# ------------------------------------------------------------------
|
||||
# 时延评估 (v1.1 §4 端到端模型, IterBatch 侧)
|
||||
# 口径: dtype 感知算力 (issue#28); 字节列整芯片 (issue#29);
|
||||
# 输出落点 R4 (issue#30).
|
||||
# ------------------------------------------------------------------
|
||||
def evaluate(self, case: BmmCase, plan: ImplPlan) -> HardwareTiming:
|
||||
s = self.spec
|
||||
m, n, k = case.m, case.n, case.k
|
||||
dt = case.dtype_in_bytes
|
||||
out_b = case.dtype_out_bytes
|
||||
b = case.batch_c
|
||||
b_core, k_l1 = plan.b_core, plan.k_l1
|
||||
qc = s.q_cube(case.dtype_a, case.dtype_b)
|
||||
out_l2 = output_to_l2(case, s)
|
||||
w_fix = s.bw_l2_pc if out_l2 else s.bw_pc
|
||||
|
||||
k_truncated = k_l1 >= k
|
||||
n_k = 1 if k_truncated else -(-k // k_l1)
|
||||
|
||||
t_load = min(k_l1, k) * (m + n) * dt / s.bw_pc
|
||||
t_comp_chunk = 2.0 * m * n * min(k_l1, k) / s.q16
|
||||
t_write = m * n * out_b / s.bw_pc
|
||||
t_comp_chunk = 2.0 * m * n * min(k_l1, k) / qc
|
||||
t_write = m * n * out_b / w_fix
|
||||
|
||||
# 搬移: 每 batch n_K 次 GM->L1, 每次含 T_cmd
|
||||
# 搬移: 每 batch n_K 次 GM->L1 (各 (batch,K段) 数据互不重叠, GM 每字节一次),
|
||||
# 每次含 T_cmd; dma_cmds 为单核命令数 (各核并行, issue#29 口径)
|
||||
dma_cmds = b_core * n_k
|
||||
t_mte2_data = dma_cmds * t_load
|
||||
t_dma_cmd = dma_cmds * s.t_cmd
|
||||
t_mte2 = t_mte2_data + t_dma_cmd
|
||||
|
||||
# Cube: 无冗余
|
||||
# Cube: 无冗余; 每核 flops = b_core*2MNK (整芯片列 = b*2MNK)
|
||||
flops_pc = b_core * 2.0 * m * n * k
|
||||
t_mmad = flops_pc / s.q16
|
||||
flops_chip = b * 2.0 * m * n * k
|
||||
t_mmad = flops_pc / qc
|
||||
|
||||
# Fixpipe: b_core 个 batch 输出, 按 C dtype
|
||||
# Fixpipe (R4): b_core 个 batch 输出, 按 C dtype
|
||||
fix_bytes_pc = b_core * m * n * out_b
|
||||
t_fix = fix_bytes_pc / s.bw_pc
|
||||
fix_bytes_chip = b * m * n * out_b
|
||||
t_fix = fix_bytes_pc / w_fix
|
||||
|
||||
# drain: 末 batch 排空 = T_comp + T_write
|
||||
t_drain = t_comp_chunk + t_write
|
||||
|
||||
gm_bytes_pc = b_core * (m * k + k * n) * dt
|
||||
|
||||
return assemble_timing(
|
||||
t_mte2_gm=t_mte2_data, t_mte2_l2=0.0, t_dma_cmd=t_dma_cmd,
|
||||
t_mmad=t_mmad, t_fixpipe=t_fix, t_reduce=0.0, t_drain=t_drain,
|
||||
gm_read_bytes=gm_bytes_pc, l2_read_bytes=0.0,
|
||||
dma_cmd_count=dma_cmds, cube_flops=flops_pc,
|
||||
fixpipe_bytes=fix_bytes_pc,
|
||||
gm_read_bytes=b * n_k * min(k_l1, k) * (m + n) * dt,
|
||||
l2_read_bytes=0.0,
|
||||
dma_cmd_count=dma_cmds, cube_flops=flops_chip,
|
||||
fixpipe_bytes=fix_bytes_chip,
|
||||
fixpipe_to_gm=(not out_l2),
|
||||
)
|
||||
|
||||
@@ -18,7 +18,7 @@ import math
|
||||
|
||||
from ..hardware import NpuSpec, ASCEND950PR
|
||||
from ..models import BmmCase, ImplPlan, HardwareTiming, align_down
|
||||
from ..timing import MoveInPlan, assemble_timing
|
||||
from ..timing import assemble_timing, output_to_l2
|
||||
from .base import Branch, BranchResult, ConditionCheck
|
||||
|
||||
MIN_B0 = 2 # b0: 合并搬移有收益的最小合并数
|
||||
@@ -142,9 +142,17 @@ class MergeBatchBranch(Branch):
|
||||
))
|
||||
b_l1 = max(b_l1, b0)
|
||||
|
||||
# Step 5: 输出落点 (issue#30): 整 case 输入+输出可驻留 L2 才写 L2 (R4)
|
||||
out_l2 = output_to_l2(case, s)
|
||||
l2_out = ("resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)"
|
||||
if out_l2 else
|
||||
"direct_gm(整case超L2: 输入优先驻留L2, 输出直写GM不占L2)")
|
||||
|
||||
note = (f"b0={b0} (L0C上限{b0_l0c:.1f}/算存比上限{b0_ai:.1f}/b_core={b_core}); "
|
||||
f"{'K截断' if k_truncated else 'L1绑定'}; "
|
||||
f"合并后单次DMA搬入 A'[{b0*m},{k_l1}]+B'[{k_l1},{b0*n}]")
|
||||
f"合并后单次DMA搬入 A'[{b0*m},{k_l1}]+B'[{k_l1},{b0*n}]; "
|
||||
f"输出落点: {'L2驻留' if out_l2 else '直写GM'} (整case V_in+V_out"
|
||||
f"={case.input_bytes/1048576:.1f}MB vs L2={s.l2_bytes/1048576:.0f}MB)")
|
||||
|
||||
return ImplPlan(
|
||||
case_id=case.case_id, branch=self.name,
|
||||
@@ -157,7 +165,7 @@ class MergeBatchBranch(Branch):
|
||||
# L0C 双缓冲约束已由进入条件 2 保证 (2*(b0*M)*(b0*N)*4B <= L0C)
|
||||
base_m=b0 * m, base_n=b0 * n, base_k=max(min(k_l0, k_l1, k), s.fractal),
|
||||
l2_policy_in="allocate(GM->L1随路驻留L2)",
|
||||
l2_policy_out="direct_gm(输出仅写一次,直写GM不占L2)" if not case.out_nd else "direct_gm",
|
||||
l2_policy_out=l2_out,
|
||||
swizzle_w=0, workspace_bytes=0,
|
||||
tail_strategy="不涉及(核内不切M/N)",
|
||||
fixpipe_unitflag=True,
|
||||
@@ -166,55 +174,71 @@ class MergeBatchBranch(Branch):
|
||||
)
|
||||
|
||||
# ------------------------------------------------------------------
|
||||
# 时延评估 (v1.1 §4 端到端模型)
|
||||
# 时延评估 (v1.1 §4 端到端模型; issue#27/#28/#29/#30 口径)
|
||||
# - Cube flops (issue#27 复核): 每合并步计算全网格 (b0*M)x(b0*N)xK
|
||||
# 含交叉项冗余, flops_step = 2*(b0M)*(b0N)*K, 每核步数 = b_core/b0
|
||||
# -> 每核 flops = b_core*b0*2MNK (与 CSV 及文档公式一致, 无修改);
|
||||
# - 算力按输入 dtype (issue#28): q_cube(dtype_a, dtype_b);
|
||||
# - 字节列 = 整芯片口径 (issue#29); GM 首读 = V_in 一次 (合并按组搬入
|
||||
# 各 (batch,K段) 数据互不重叠, 无 L2 重复读);
|
||||
# - 输出落点按整 case 驻留判定 (issue#30, R4).
|
||||
# ------------------------------------------------------------------
|
||||
def evaluate(self, case: BmmCase, plan: ImplPlan) -> HardwareTiming:
|
||||
s = self.spec
|
||||
m, n, k = case.m, case.n, case.k
|
||||
dt = case.dtype_in_bytes
|
||||
out_b = case.dtype_out_bytes
|
||||
b = case.batch_c
|
||||
b_core, b0, k_l1 = plan.b_core, plan.merge_b0, plan.k_l1
|
||||
qc = s.q_cube(case.dtype_a, case.dtype_b)
|
||||
out_l2 = output_to_l2(case, s, 0.0)
|
||||
w_fix = s.bw_l2_pc if out_l2 else s.bw_pc
|
||||
|
||||
k_truncated = k_l1 >= k
|
||||
# 每 K 分块搬移/计算时延 (未合并基准, v1.1 §4.1 符号)
|
||||
t_load = k_l1 * (m + n) * dt / s.bw_pc
|
||||
t_comp_chunk = 2.0 * m * n * k_l1 / s.q16
|
||||
t_write = m * n * out_b / s.bw_pc # 单 batch 输出写回 (单核带宽份额)
|
||||
t_comp_chunk = 2.0 * m * n * k_l1 / qc
|
||||
t_write = m * n * out_b / w_fix # 单 batch 输出写回 (R4 落点带宽)
|
||||
|
||||
if k_truncated:
|
||||
# K 截断: n_K=1, 合并后单次搬移量 b0 倍
|
||||
# K 截断: n_K=1, 每核 b_core/b0 次合并搬入, 每次搬 b0 个 batch 全 K
|
||||
n_move = b_core / b0
|
||||
t_mte2_data = n_move * b0 * t_load
|
||||
dma_cmds = n_move
|
||||
gm_chip = b * (m + n) * k * dt # = V_in, GM 每字节一次
|
||||
else:
|
||||
# L1 绑定: k_L1^m = k_L1/b0, n_K^m = b0*n_K, 搬移次数与 IterBatch 相同
|
||||
# (每 (合并组, K段) 数据互不重叠, GM 仍每字节一次, 末段含 padding 上取)
|
||||
n_k = -(-k // k_l1)
|
||||
n_move = b_core * n_k
|
||||
t_mte2_data = n_move * t_load
|
||||
dma_cmds = n_move
|
||||
gm_chip = b * (m + n) * n_k * k_l1 * dt # >= V_in (padding 上取)
|
||||
|
||||
t_dma_cmd = dma_cmds * s.t_cmd
|
||||
t_mte2 = t_mte2_data + t_dma_cmd
|
||||
|
||||
# Cube: 合并计算含冗余 (b0^2 输出, 有效 b0) -> 计算量 = b_core*b0*2MNK
|
||||
# Cube (每核口径): 合并计算含冗余 (b0^2 输出, 有效 b0), 每核 b_core/b0 步,
|
||||
# 每步 flops = 2*(b0*M)*(b0*N)*K -> 每核 = b_core*b0*2MNK (issue#27 复核一致)
|
||||
flops_pc = b_core * b0 * 2.0 * m * n * k
|
||||
t_mmad = flops_pc / s.q16
|
||||
flops_chip = b * b0 * 2.0 * m * n * k # 整芯片口径列
|
||||
t_mmad = flops_pc / qc
|
||||
|
||||
# Fixpipe: 只写对角块, 写出量 = b_core*MN*outB (C 矩阵 dtype, 随路转换)
|
||||
# Fixpipe (R4): 只写对角块, 写出量 = b_core*MN*outB (C 矩阵 dtype, 随路转换)
|
||||
fix_bytes_pc = b_core * m * n * out_b
|
||||
t_fix = fix_bytes_pc / s.bw_pc
|
||||
fix_bytes_chip = b * m * n * out_b
|
||||
t_fix = fix_bytes_pc / w_fix
|
||||
|
||||
# drain: 末合并 batch 排空 = b0*(T_comp + T_write)
|
||||
t_drain = b0 * (t_comp_chunk + t_write)
|
||||
|
||||
gm_bytes_pc = b_core * (m * k + k * n) * dt
|
||||
|
||||
return assemble_timing(
|
||||
t_mte2_gm=t_mte2_data, t_mte2_l2=0.0, t_dma_cmd=t_dma_cmd,
|
||||
t_mmad=t_mmad, t_fixpipe=t_fix, t_reduce=0.0, t_drain=t_drain,
|
||||
gm_read_bytes=gm_bytes_pc, l2_read_bytes=0.0,
|
||||
dma_cmd_count=dma_cmds, cube_flops=flops_pc,
|
||||
fixpipe_bytes=fix_bytes_pc,
|
||||
gm_read_bytes=gm_chip, l2_read_bytes=0.0,
|
||||
dma_cmd_count=dma_cmds, cube_flops=flops_chip,
|
||||
fixpipe_bytes=fix_bytes_chip,
|
||||
fixpipe_to_gm=(not out_l2),
|
||||
)
|
||||
|
||||
# ------------------------------------------------------------------
|
||||
@@ -243,7 +267,9 @@ class MergeBatchBranch(Branch):
|
||||
|
||||
plan = self.make_plan(case)
|
||||
b0 = plan.merge_b0
|
||||
t_comp = 2.0 * m * n * min(k_l1_iter, k) / s.q16
|
||||
# T_comp 按输入 dtype 算力 (issue#28); T_write 保持 v1.1 直写 GM 语义
|
||||
qc = s.q_cube(case.dtype_a, case.dtype_b)
|
||||
t_comp = 2.0 * m * n * min(k_l1_iter, k) / qc
|
||||
t_write = m * n * out_b / s.bw_pc
|
||||
drain_pen = (b0 - 1) * (t_comp + t_write)
|
||||
|
||||
|
||||
@@ -10,7 +10,7 @@ from __future__ import annotations
|
||||
|
||||
from ..hardware import NpuSpec, ASCEND950PR
|
||||
from ..models import BmmCase, ImplPlan, HardwareTiming
|
||||
from ..timing import assemble_timing
|
||||
from ..timing import assemble_timing, output_to_l2
|
||||
from .base import Branch, ConditionCheck
|
||||
|
||||
|
||||
@@ -68,32 +68,42 @@ class SpecialBranch(Branch):
|
||||
|
||||
# ------------------------------------------------------------------
|
||||
def evaluate(self, case: BmmCase, plan: ImplPlan) -> HardwareTiming:
|
||||
"""AIV 通路时延: 瓶颈在搬移 (AIV 算力远剩)."""
|
||||
"""AIV 通路时延: 瓶颈在搬移 (AIV 算力远剩).
|
||||
|
||||
口径: AIV 逐元素通量按输入 dtype (issue#28); 输出落点 R4 (issue#30):
|
||||
整 case 输入+输出 <= L2 -> 输出写 L2 (异步回写不占算子时延), 否则直写
|
||||
GM 与读共享总线 (assemble 的 MTE2 链累加, 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
|
||||
out_l2 = output_to_l2(case, s)
|
||||
w_fix = s.bw_l2 if out_l2 else s.bw_gm
|
||||
cube_flops = 0.0
|
||||
t_compute = 0.0
|
||||
|
||||
if k == 0:
|
||||
# 纯写值: 仅写出
|
||||
t_in, t_compute, t_out = 0.0, 0.0, b * m * n * out_b / s.bw_gm
|
||||
in_bytes, cube_flops = 0.0, 0.0
|
||||
# 纯写值: 仅写出 (R1 下 GM 读 = 0, 输入本身为空)
|
||||
t_in = 0.0
|
||||
t_out = b * m * n * out_b / w_fix
|
||||
in_bytes = 0.0
|
||||
else:
|
||||
# 逐元素乘: 搬入 A+B, 搬出 C, AIV 算力远剩
|
||||
# 逐元素乘: 搬入 A+B (GM 每字节一次), 搬出 C
|
||||
in_bytes = b * (m * k + k * n) * dt
|
||||
out_bytes = b * m * n * out_b
|
||||
t_in = in_bytes / s.bw_gm
|
||||
t_out = out_bytes / s.bw_gm
|
||||
# AIV 求积吞吐 (近似按 Q_AIV)
|
||||
t_compute = b * m * n / s.q_aiv
|
||||
t_out = out_bytes / w_fix
|
||||
# AIV 逐元素吞吐按输入 dtype 取通量 (issue#28: bf16/fp16 x2, int8 x4...)
|
||||
t_compute = b * m * n / s.aiv_elem_rate(case.dtype_in)
|
||||
cube_flops = float(b * m * n)
|
||||
|
||||
t_total = max(t_in, t_out, t_compute)
|
||||
return assemble_timing(
|
||||
t_mte2_gm=t_in, t_mte2_l2=0.0, t_dma_cmd=0.0,
|
||||
t_mmad=t_compute, t_fixpipe=t_out, t_reduce=0.0, t_drain=0.0,
|
||||
gm_read_bytes=in_bytes, l2_read_bytes=0.0, dma_cmd_count=0.0,
|
||||
cube_flops=cube_flops,
|
||||
fixpipe_bytes=b * m * n * out_b,
|
||||
fixpipe_to_gm=(not out_l2),
|
||||
)
|
||||
|
||||
@@ -12,7 +12,7 @@ import math
|
||||
|
||||
from ..hardware import NpuSpec, ASCEND950PR
|
||||
from ..models import BmmCase, ImplPlan, HardwareTiming, ceil_div
|
||||
from ..timing import assemble_timing, eval_streamk_reduce
|
||||
from ..timing import assemble_timing, eval_streamk_reduce, output_to_l2
|
||||
from .base import Branch, ConditionCheck
|
||||
|
||||
|
||||
@@ -32,10 +32,14 @@ class StreamKBranch(Branch):
|
||||
p = self._p_value(case)
|
||||
return int(self.spec.aic_num // max(1, math.ceil(p)))
|
||||
|
||||
def _theta_c(self) -> float:
|
||||
"""归约代价系数 theta_c = Q16/2 * (8B/W_L2 + 1/Q_AIV) ≈ 12."""
|
||||
def _theta_c(self, case: BmmCase) -> float:
|
||||
"""归约代价系数 theta_c = Q_cube(dtype)/2 * (8B/W_L2 + 1/Q_AIV_fp32).
|
||||
|
||||
Q_AIV 用 fp32 档 (部分和为 fp32, issue#28); Cube 侧按输入 dtype.
|
||||
"""
|
||||
s = self.spec
|
||||
return s.q16 / 2 * (8 / s.bw_l2 + 1 / s.q_aiv)
|
||||
qc = s.q_cube(case.dtype_a, case.dtype_b)
|
||||
return qc / 2 * (8 / s.bw_l2 + 1 / s.q_aiv)
|
||||
|
||||
# ------------------------------------------------------------------
|
||||
def check_conditions(self, case: BmmCase) -> list:
|
||||
@@ -61,7 +65,7 @@ class StreamKBranch(Branch):
|
||||
|
||||
# 条件 3: K > grid_K^2/(grid_K-1) * theta_c (归约代价可接受)
|
||||
if grid_k >= 2:
|
||||
theta_c = self._theta_c()
|
||||
theta_c = self._theta_c(case)
|
||||
k_thresh = grid_k * grid_k / (grid_k - 1) * theta_c
|
||||
c3 = case.k > k_thresh
|
||||
checks.append(ConditionCheck(
|
||||
@@ -139,7 +143,9 @@ class StreamKBranch(Branch):
|
||||
(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).
|
||||
追加为 drain: t_drain = waves * eval_streamk_reduce(tile, grid_k, out);
|
||||
最终输出写回落点按 R4 (issue#30): 整 case 输入+输出+workspace <= L2
|
||||
时写 L2, 否则直写 GM (drain 内按 GM 带宽).
|
||||
"""
|
||||
s = self.spec
|
||||
b = case.batch_c
|
||||
@@ -152,6 +158,8 @@ class StreamKBranch(Branch):
|
||||
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
|
||||
qc = s.q_cube(case.dtype_a, case.dtype_b)
|
||||
out_l2 = output_to_l2(case, s, float(plan.workspace_bytes))
|
||||
|
||||
# ---- MTE2: GM 段 + L2 段 (同 ASW 块级规则) ----
|
||||
a_b = m * k * dt
|
||||
@@ -166,13 +174,14 @@ class StreamKBranch(Branch):
|
||||
t_l2 = l2_read / s.bw_l2
|
||||
t_mte2 = t_gm + t_l2
|
||||
|
||||
# ---- MMAD: 芯片总量 ----
|
||||
# ---- MMAD: 芯片总量, dtype 感知算力 (issue#28) ----
|
||||
flops = 2.0 * b * m * n * k
|
||||
t_mmad = flops / (s.aic_num * s.q16)
|
||||
t_mmad = flops / (s.aic_num * qc)
|
||||
|
||||
# ---- 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_group = eval_streamk_reduce(tile, grid_k, out_b, s,
|
||||
out_to_gm=not out_l2)
|
||||
t_reduce = waves * t_reduce_group
|
||||
fix_bytes = 0.0
|
||||
t_fix = 0.0
|
||||
|
||||
@@ -12,7 +12,7 @@ from __future__ import annotations
|
||||
|
||||
from ..hardware import NpuSpec, ASCEND950PR
|
||||
from ..models import BmmCase, ImplPlan, HardwareTiming
|
||||
from ..timing import assemble_timing
|
||||
from ..timing import assemble_timing, output_to_l2
|
||||
from .base import Branch, ConditionCheck
|
||||
|
||||
|
||||
@@ -71,7 +71,12 @@ class ToMatmulBranch(Branch):
|
||||
|
||||
# ------------------------------------------------------------------
|
||||
def evaluate(self, case: BmmCase, plan: ImplPlan) -> HardwareTiming:
|
||||
"""折叠后按 Matmul 粗估 (详细切分待 MM 理论体系打通后接入)."""
|
||||
"""折叠后按 Matmul 粗估 (详细切分待 MM 理论体系打通后接入).
|
||||
|
||||
口径: Cube 算力按输入 dtype (issue#28); GM 首读 = V_in 一次 (R1);
|
||||
输出落点 R4 (issue#30): 整 case 输入+输出 <= L2 时写 L2 写口, 否则
|
||||
直写 GM 与读共享总线 (assemble 累加).
|
||||
"""
|
||||
s = self.spec
|
||||
if case.batch_b == 1:
|
||||
fold_m, fold_n, kk = case.batch_a * case.m, case.n, case.k
|
||||
@@ -79,13 +84,15 @@ class ToMatmulBranch(Branch):
|
||||
fold_m, fold_n, kk = case.m, case.batch_b * case.n, case.k
|
||||
dt = case.dtype_in_bytes
|
||||
out_b = case.dtype_out_bytes
|
||||
qc = s.q_cube(case.dtype_a, case.dtype_b)
|
||||
out_l2 = output_to_l2(case, s)
|
||||
|
||||
flops = 2.0 * fold_m * fold_n * kk
|
||||
t_mmad = flops / (s.aic_num * s.q16)
|
||||
t_mmad = flops / (s.aic_num * qc)
|
||||
in_bytes = (fold_m * kk + kk * fold_n) * dt
|
||||
out_bytes = fold_m * fold_n * out_b
|
||||
t_mte2 = in_bytes / s.bw_gm
|
||||
t_fix = out_bytes / s.bw_gm
|
||||
t_fix = out_bytes / (s.bw_l2 if out_l2 else s.bw_gm)
|
||||
|
||||
return assemble_timing(
|
||||
t_mte2_gm=t_mte2, t_mte2_l2=0.0, t_dma_cmd=0.0,
|
||||
@@ -93,4 +100,5 @@ class ToMatmulBranch(Branch):
|
||||
t_drain=0.0,
|
||||
gm_read_bytes=in_bytes, l2_read_bytes=0.0, dma_cmd_count=0.0,
|
||||
cube_flops=flops, fixpipe_bytes=out_bytes,
|
||||
fixpipe_to_gm=(not out_l2),
|
||||
)
|
||||
|
||||
@@ -4,12 +4,36 @@
|
||||
换芯片时逻辑结构不变, 只需新增一份同结构参数表.
|
||||
|
||||
单位约定: 算力 FLOP/s, 带宽 Byte/s, 容量 Byte, 时延 秒.
|
||||
|
||||
dtype 感知算力 (issue#28, 设计文档 docs/05 §2):
|
||||
Cube 算力按输入 dtype 分档, 基准 = BF16 (fp16 同速); 白皮书: FP8/MXFP8/HiF8
|
||||
提供 2x FP16 张量 TFLOPS, MXFP4 提供 4x FP16; FP32/TF32 同代比值为假设值
|
||||
(按 DaVinci 惯例 = 1/2, 白皮书未给同代比值), int8 假设同 FP8, 均待实测标定.
|
||||
AIV 逐元素通量按 lane 位宽等比假设 (16bit x2 / 8bit x4 / 4bit x8, 待标定).
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
from dataclasses import dataclass
|
||||
|
||||
# Cube 精度因子 (相对 BF16/FP16 基准档; A/B 不一致时取较慢一侧 = min 因子)
|
||||
CUBE_DTYPE_FACTOR = {
|
||||
"fp32": 0.5, "f32": 0.5, "tf32": 0.5, # 假设 = 1/2, 待实测标定
|
||||
"fp16": 1.0, "f16": 1.0, "bf16": 1.0,
|
||||
"fp8": 2.0, "fp8_e4m3": 2.0, "fp8_e5m2": 2.0, "int8": 2.0, # 白皮书 FP8=2xFP16; int8 假设同 FP8
|
||||
"fp4": 4.0, "fp4_e2m1": 4.0, # 白皮书 MXFP4=4xFP16; 普通 fp4 假设同 MXFP4
|
||||
}
|
||||
|
||||
# AIV 逐元素通量因子 (相对 fp32 128 lane/拍/核; 位宽等比假设, 待实测标定)
|
||||
AIV_DTYPE_FACTOR = {
|
||||
"fp32": 1.0, "f32": 1.0, "tf32": 1.0,
|
||||
"fp16": 2.0, "f16": 2.0, "bf16": 2.0,
|
||||
"fp8": 4.0, "fp8_e4m3": 4.0, "fp8_e5m2": 4.0, "int8": 4.0,
|
||||
"fp4": 8.0, "fp4_e2m1": 8.0,
|
||||
}
|
||||
|
||||
_RATE_FALLBACK = 1.0 # 未知 dtype 按基准档 (models.dtype_bytes 已先行校验, 正常不会到达)
|
||||
|
||||
|
||||
@dataclass(frozen=True)
|
||||
class NpuSpec:
|
||||
@@ -55,13 +79,50 @@ class NpuSpec:
|
||||
# ---- 派生量 (属性) ----
|
||||
@property
|
||||
def q16(self) -> float:
|
||||
"""单核 Cube BF16 峰值算力 (FLOP/s)."""
|
||||
"""单核 Cube BF16/FP16 峰值算力 (FLOP/s) = 15.1875T."""
|
||||
return self.cube_peak_tflops * 1e12 / self.aic_num
|
||||
|
||||
@staticmethod
|
||||
def _dtype_key(dtype) -> str:
|
||||
return str(dtype).strip().lower()
|
||||
|
||||
def cube_factor(self, dtype_a, dtype_b=None) -> float:
|
||||
"""按输入 dtype 取 Cube 精度因子 (issue#28).
|
||||
|
||||
A/B 不一致时取**较慢一侧** (因子较小者, 等价字节较大者);
|
||||
Cube 乘法两侧的实际吞吐受较慢精度限制.
|
||||
"""
|
||||
keys = [self._dtype_key(dtype_a)]
|
||||
if dtype_b is not None:
|
||||
keys.append(self._dtype_key(dtype_b))
|
||||
factors = [CUBE_DTYPE_FACTOR.get(k, _RATE_FALLBACK) for k in keys]
|
||||
return min(factors)
|
||||
|
||||
def q_cube(self, dtype_a, dtype_b=None) -> float:
|
||||
"""单核 Cube 峰值算力 (FLOP/s), 按输入 dtype 分档 (issue#28)."""
|
||||
return self.q16 * self.cube_factor(dtype_a, dtype_b)
|
||||
|
||||
def aiv_elem_factor(self, dtype) -> float:
|
||||
"""按 dtype 取 AIV 逐元素通量因子 (相对 fp32 基准)."""
|
||||
return AIV_DTYPE_FACTOR.get(self._dtype_key(dtype), _RATE_FALLBACK)
|
||||
|
||||
@property
|
||||
def aiv_elem_rate_fp32(self) -> float:
|
||||
"""AIV fp32 逐元素吞吐 (元素/s), 64 核合计."""
|
||||
return self.aiv_num * self.aiv_fp32_per_cycle * self.aiv_freq_ghz * 1e9
|
||||
|
||||
def aiv_elem_rate(self, dtype) -> float:
|
||||
"""AIV 逐元素吞吐 (元素/s, 64 核合计), 按 dtype 分档 (issue#28)."""
|
||||
return self.aiv_elem_rate_fp32 * self.aiv_elem_factor(dtype)
|
||||
|
||||
@property
|
||||
def q_aiv(self) -> float:
|
||||
"""AIV 向量求和吞吐 (元素/s), 64 核合计."""
|
||||
return self.aiv_num * self.aiv_fp32_per_cycle * self.aiv_freq_ghz * 1e9
|
||||
"""AIV fp32 逐元素吞吐 (元素/s, 64 核合计) = aiv_elem_rate_fp32.
|
||||
|
||||
注意: 该值只适用于 fp32 数据 (如 StreamK 对 fp32 部分和求和);
|
||||
逐元素运算按输入 dtype 用 aiv_elem_rate(dtype) (issue#28).
|
||||
"""
|
||||
return self.aiv_elem_rate_fp32
|
||||
|
||||
@property
|
||||
def bw_pc(self) -> float:
|
||||
@@ -75,7 +136,11 @@ class NpuSpec:
|
||||
|
||||
@property
|
||||
def r16(self) -> float:
|
||||
"""16bit 位宽平衡点算存比 R_16 = Cube峰值 / (GM带宽/2B) ≈ 607.5 FLOP/元素."""
|
||||
"""16bit 位宽平衡点算存比 R_16 = Cube峰值 / (GM带宽/2B) ≈ 607.5 FLOP/元素.
|
||||
|
||||
按 issue#28 速率表该比值对全 dtype 不变 (Cube 因子与元素字节数互成反比),
|
||||
故入口条件沿用单一 R_16.
|
||||
"""
|
||||
return self.cube_peak_tflops * 1e12 / (self.bw_gm / 2)
|
||||
|
||||
@property
|
||||
|
||||
@@ -133,6 +133,12 @@ class BmmCase:
|
||||
# 返回 float 以兼容 fp4 (0.5B)
|
||||
return max(dtype_bytes(self.dtype_a), dtype_bytes(self.dtype_b))
|
||||
|
||||
@property
|
||||
def dtype_in(self) -> str:
|
||||
"""输入侧"较慢"dtype (元素字节较大者) — Cube/AIV 速率取慢侧 (issue#28)."""
|
||||
return (self.dtype_a if dtype_bytes(self.dtype_a) >= dtype_bytes(self.dtype_b)
|
||||
else self.dtype_b)
|
||||
|
||||
@property
|
||||
def dtype_out_bytes(self) -> float:
|
||||
return dtype_bytes(self.dtype_c)
|
||||
@@ -261,23 +267,30 @@ class ImplPlan:
|
||||
|
||||
@dataclass
|
||||
class HardwareTiming:
|
||||
"""各硬件流水级时延 (秒) 与数据量明细."""
|
||||
"""各硬件流水级时延 (秒) 与数据量明细.
|
||||
|
||||
数据量列口径 (issue#29, 设计文档 docs/05 §4.4):
|
||||
gm_read_bytes / l2_read_bytes / fixpipe_bytes / cube_flops = **整芯片**总量
|
||||
(跨分支可比, GM 首读下限断言: gm_read_bytes >= case.input_bytes);
|
||||
dma_cmd_count = **单核** GM->L1 DMA 命令数 (各核引擎并行, 墙钟 T_cmd =
|
||||
单核命令数 x T_cmd), 与字节列口径不同属.
|
||||
"""
|
||||
|
||||
# 搬入 (MTE2)
|
||||
gm_read_bytes: float = 0.0 # GM->L1 直读数据量 (不驻留/未命中 L2 的部分)
|
||||
l2_read_bytes: float = 0.0 # L2->L1 数据量 (驻留 L2 后重复读命中部分)
|
||||
gm_read_bytes: float = 0.0 # GM->L1 直读数据量 (整芯片, 首读/落空重读)
|
||||
l2_read_bytes: float = 0.0 # L2->L1 数据量 (整芯片, 驻留 L2 后重复读命中部分)
|
||||
t_mte2_gm: float = 0.0 # GM->L1 时延 (按 GM 带宽, 不累加 L2->L1)
|
||||
t_mte2_l2: float = 0.0 # L2->L1 时延 (按 L2 带宽)
|
||||
t_mte2: float = 0.0 # 搬入合计 = t_mte2_gm + t_mte2_l2 (两者发生在不同数据上)
|
||||
dma_cmd_count: float = 0.0 # GM->L1 DMA 命令次数 (T_cmd 分析用)
|
||||
t_dma_cmd: float = 0.0 # DMA 命令固定开销合计
|
||||
dma_cmd_count: float = 0.0 # 单核 GM->L1 DMA 命令次数 (T_cmd 分析用)
|
||||
t_dma_cmd: float = 0.0 # DMA 命令固定开销合计 (墙钟)
|
||||
|
||||
# 计算 (Cube MMAD)
|
||||
cube_flops: float = 0.0 # Cube 实际计算量 (MergeBatch 含冗余)
|
||||
cube_flops: float = 0.0 # Cube 实际计算量 (整芯片, MergeBatch 含冗余)
|
||||
t_mmad: float = 0.0
|
||||
|
||||
# 搬出 (Fixpipe)
|
||||
fixpipe_bytes: float = 0.0 # 写出数据量 (按 C 矩阵 dtype / StreamK 临时矩阵按 4B)
|
||||
fixpipe_bytes: float = 0.0 # 写出数据量 (整芯片, 按 C 矩阵 dtype / StreamK 临时矩阵按 4B)
|
||||
t_fixpipe: float = 0.0
|
||||
|
||||
# 归约 (StreamK 专用)
|
||||
|
||||
@@ -24,7 +24,19 @@ from __future__ import annotations
|
||||
from dataclasses import dataclass
|
||||
|
||||
from .hardware import NpuSpec, ASCEND950PR
|
||||
from .models import HardwareTiming
|
||||
from .models import BmmCase, HardwareTiming
|
||||
|
||||
|
||||
def output_to_l2(case: BmmCase, spec: NpuSpec = ASCEND950PR,
|
||||
workspace_bytes: float = 0.0) -> bool:
|
||||
"""Fixpipe 输出落点决策 (issue#30, 设计文档 docs/05 §4.2 R4).
|
||||
|
||||
to_l2 (输出写 L2 写口 5.2TB/s, GM 写流量 = 0, 异步回写不占算子时延)
|
||||
⟺ 整 case 输入 V_in + 输出 V_out [+ StreamK workspace] ≤ L2
|
||||
否则输出**直写 GM**: 输入优先驻留 L2 (输入存在重复读), 输出计入 GM
|
||||
读写共享总线 (与读累加, issue#23).
|
||||
"""
|
||||
return (case.input_bytes + case.output_bytes + workspace_bytes) <= spec.l2_bytes
|
||||
|
||||
|
||||
@dataclass
|
||||
@@ -56,9 +68,11 @@ def eval_mte2(move: MoveInPlan, spec: NpuSpec = ASCEND950PR,
|
||||
return t_gm, t_l2, t_gm + t_l2 + t_cmd, t_cmd
|
||||
|
||||
|
||||
def eval_mmad(flops_per_core: float, spec: NpuSpec = ASCEND950PR) -> float:
|
||||
"""Cube 计算时延: 单核计算量 / 单核算力."""
|
||||
return flops_per_core / spec.q16 if flops_per_core > 0 else 0.0
|
||||
def eval_mmad(flops_per_core: float, spec: NpuSpec = ASCEND950PR,
|
||||
dtype_a=None, dtype_b=None) -> float:
|
||||
"""Cube 计算时延: 单核计算量 / 单核 dtype 感知算力 (issue#28)."""
|
||||
rate = spec.q_cube(dtype_a, dtype_b)
|
||||
return flops_per_core / rate if flops_per_core > 0 else 0.0
|
||||
|
||||
|
||||
def eval_fixpipe(bytes_per_core: float, to_l2: bool,
|
||||
@@ -75,21 +89,24 @@ def eval_fixpipe(bytes_per_core: float, to_l2: bool,
|
||||
|
||||
|
||||
def eval_streamk_reduce(tile_elems: float, grid_k: int, out_dtype_bytes: int,
|
||||
spec: NpuSpec = ASCEND950PR) -> float:
|
||||
spec: NpuSpec = ASCEND950PR, out_to_gm: bool = False) -> float:
|
||||
"""StreamK 单 tile 归约时延 (v0.98 §七).
|
||||
|
||||
部分和 dtype = L0C dtype (4B, 防精度丢失), 驻留 L2, AIV 归约:
|
||||
部分和 dtype = L0C dtype (4B, 防精度丢失), 驻留 L2, AIV 归约 (fp32 求和,
|
||||
AIV 按 fp32 通量, 不随输入 dtype 变 — issue#28):
|
||||
AIC 写部分和 grid_k x tile x 4B / W_L2
|
||||
AIV 读回 grid_k x tile x 4B / W_L2
|
||||
AIV 求和 grid_k x tile / Q_AIV
|
||||
写回 tile x outB / W_L2
|
||||
AIV 求和 grid_k x tile / Q_AIV(fp32)
|
||||
写回 tile x outB / W (最终输出落点, issue#30):
|
||||
整 case 可驻留 (S_A) 时 W = W_L2; 否则直写 GM (W = W_GM)
|
||||
"""
|
||||
b4 = 4
|
||||
w_l2 = spec.bw_l2
|
||||
w_out = spec.bw_gm if out_to_gm else w_l2
|
||||
t_write_partial = grid_k * tile_elems * b4 / w_l2
|
||||
t_read_back = grid_k * tile_elems * b4 / w_l2
|
||||
t_sum = grid_k * tile_elems / spec.q_aiv
|
||||
t_write_out = tile_elems * out_dtype_bytes / w_l2
|
||||
t_write_out = tile_elems * out_dtype_bytes / w_out
|
||||
return t_write_partial + t_read_back + t_sum + t_write_out
|
||||
|
||||
|
||||
|
||||
197
BMM/BMM_Theory/docs/05_L2驻留GM读写与dtype算力口径_设计分析.md
Normal file
197
BMM/BMM_Theory/docs/05_L2驻留GM读写与dtype算力口径_设计分析.md
Normal file
@@ -0,0 +1,197 @@
|
||||
# 05 L2 驻留、GM 读写与 dtype 算力口径 — 设计分析
|
||||
|
||||
> 关联 issue: #27 (MergeBatch Cube 复核) / #28 (dtype 感知算力) /
|
||||
> #29 (GM 读取量下限 + L2 驻留工作集) / #30 (Fixpipe 输出落点)
|
||||
>
|
||||
> 状态: 设计文档先行, 代码按本文档落地 (2026-06 评审轮)。
|
||||
> 本文档是 GM/L2/输出计账与算力口径的**单一语义来源**; 各分支 evaluate/生成注释以本文档为准。
|
||||
|
||||
---
|
||||
|
||||
## 0. 本轮的四个问题与结论摘要
|
||||
|
||||
| # | 问题 | 结论 | 落地 |
|
||||
|---|---|---|---|
|
||||
| #27 | MergeBatch Cube 时延 "每次 b0 个 batch, flops=2·b0M·b0N·K, 共 b_core/b0 次" | **公式与代码一致** (总 flops = b_core·b0·2MNK, 证明见 §6); 无计算改动 | 注释显式化; 暴露的"字节列每核/芯片口径混用"随 #29 统一 |
|
||||
| #28 | 估算时延恒用 BF16 Cube / fp32 AIV 算力 | **确认错误** | §2 dtype 感知速率表 + 全链路替换 |
|
||||
| #29 | GM 读取量"小于输入数据量" | 计算层面各分支 GM ≥ 输入一次 (逐分支表见 §5); **真问题**: ① 字节列每核/芯片口径混用导致比对失真; ② 缺整 case 驻留边界与统一驻留判定 | §3 公理化 + §4 场景模型 + 芯片口径列统一 + GM≥输入不变量测试 |
|
||||
| #30 | Fixpipe 输出落点 (默认写 L2, 容量不足才写 GM) | **确认错误** (Iter/Merge/特殊恒直写 GM; ASW 场景 A 按单 batch 判定, 整 case 输出累积必逐出) | §4.2 输出落点规则 |
|
||||
|
||||
---
|
||||
|
||||
## 1. 硬件事实 (Ascend950PR, 与 hardware/ascend950pr.py 对应)
|
||||
|
||||
- GM (HBM): **1.6TB/s, 读写共享总线** —— 同一时刻读与写累加计时 (issue#23 已落地)。
|
||||
- L2: 128MB, **5.2TB/s, 读口/写口各自独享** —— L2 重复读(读口)与 Fixpipe→L2 写(写口)互不竞争 (issue#23 已落地)。
|
||||
- 输入首访一律走 GM 带宽计一次, 不叠加 L2 (issue#24 已落地); L2 只吸收"驻留后的再次读取"。
|
||||
- Cube 算力分精度 (白皮书: FP8/MXFP8/HiF8 = 2×FP16, MXFP4 = 4×FP16; 16bit 档 = 基准):
|
||||
BF16 486 TFLOPS/芯片 (=15.1875 TFLOPS/核)。FP32/TF32 同代比值白皮书未给 → **假设 ½, 待实测标定** (见 §2 假设表)。
|
||||
- AIV (Vector): 64 核 × 128 fp32 lane/拍 × 1.65GHz = 13.5T 元素/s (fp32 档)。
|
||||
|
||||
---
|
||||
|
||||
## 2. dtype 感知算力 (issue #28)
|
||||
|
||||
### 2.1 速率表 (因子, 相对基准档)
|
||||
|
||||
Cube 因子 (相对 BF16, 同倍率作用于整芯片与单核):
|
||||
|
||||
| dtype | 因子 | 依据 |
|
||||
|---|---|---|
|
||||
| bf16 / fp16 | 1.0 | 基准档 (950 同速) |
|
||||
| fp32 / tf32 | 0.5 | **假设**: 白皮书仅述 FP16/FP32 单核较上代均 +100%, 未给同代比值; 按 DaVinci 惯例 FP32 累加通量减半。待 msProf 实测标定 |
|
||||
| fp8 / fp8_e4m3 / fp8_e5m2 | 2.0 | 白皮书: FP8 = 2×FP16 |
|
||||
| int8 | 2.0 | **假设** 同 FP8 (8bit 整数张量档), 待实测 |
|
||||
| fp4 / fp4_e2m1 | 4.0 | 白皮书: MXFP4 = 4×FP16; 普通 fp4 按 MXFP4 假设, 待实测 |
|
||||
|
||||
AIV (Vector) 逐元素通量因子 (相对 fp32 lane 基准):
|
||||
|
||||
| dtype | 因子 | 依据 |
|
||||
|---|---|---|
|
||||
| fp32 / tf32 | 1.0 | 基准 (128 lane/拍/核) |
|
||||
| fp16 / bf16 | 2.0 | **假设** 16bit 双元素/lane, 待实测 |
|
||||
| fp8 / int8 | 4.0 | **假设** 8bit 四元素/lane, 待实测 |
|
||||
| fp4 | 8.0 | **假设** 待实测 |
|
||||
|
||||
A/B dtype 不一致 (混精度): Cube 取**较慢一侧**的因子 (max(字节数) 侧, 即 min(因子))。
|
||||
StreamK 归约是对 **fp32 部分和**求和 → AIV 归约恒按 fp32 档 (因子 1.0), 不随输入 dtype 变。
|
||||
|
||||
### 2.2 替换点 (全部时延估算)
|
||||
|
||||
| 位置 | 现状 | 改为 |
|
||||
|---|---|---|
|
||||
| 各分支 t_mmad / t_comp_chunk / drain / 尾轮决策主导项 | flops / spec.q16 | flops / spec.q_cube(dtype_a, dtype_b) |
|
||||
| MergeBatch beats_iterbatch 阈值中的 T_comp | q16 | 同上 (该分支 A/B 同 dtype, 简化为 dtype_a) |
|
||||
| special K=1 t_compute (逐元素乘) | b·m·n / spec.q_aiv (fp32) | b·m·n / spec.aiv_elem_rate(dtype_in) |
|
||||
| StreamK θ_c (归约代价系数) | q16(bf16) | q_cube(输入 dtype) (AIV 侧仍 fp32) |
|
||||
| R16 平衡点 (MergeBatch 进入条件 5 / b0 算存比上限) | r16(bf16 基准) | r16_for(dtype) |
|
||||
| 入口条件/生成侧 | — | 同步按 §2.1 因子 (dtype 只影响时延, 不改变分支结构) |
|
||||
|
||||
---
|
||||
|
||||
## 3. 存储口径公理 (issue #29)
|
||||
|
||||
记号:
|
||||
|
||||
- V_in = 输入存储总量 = batch_a·M·K·dt + batch_b·K·N·dt (广播前实际存储), V_out = batch_c·M·N·out_dt;
|
||||
- a_b = M·K·dt (单 batch A 字节), b_b = K·N·dt (单 batch B 字节)。
|
||||
|
||||
- **R1 (GM 首读下限)**: 每个输入字节最初在 GM, 至少从 GM 读一次 → **GM_read ≥ V_in**。
|
||||
违反即模型 bug (统一用测试锁死)。
|
||||
- **R2 (L2 只吸收驻留后的重复读)**: 同一字节的第 2..n 次访问, 仅当两次访问之间该字节仍驻留 L2
|
||||
(所在"复用窗口"的工作集 ≤ L2 减去其它占用) 才按 L2 读口计时; 否则按 GM 重读计时。
|
||||
- **R3 (整 case 全驻留边界)**: V_in + V_out (StreamK 另加部分和 workspace) ≤ L2 时,
|
||||
全程零逐出: GM_read = V_in (每字节一次), 全部重复读走 L2, 输出驻留 L2 (见 §4.2)。
|
||||
- **R4 (输出落点)**: 见 §4.2。
|
||||
- **R5 (输入工作集场景)**: 单 batch 输入 ≤ L2 → batch 内 tile 共享块重复读按 L2 计
|
||||
(A 行块被 n_cnt 个列 tile 复用 → (n_cnt−1) 次 L2; B 列块被 m_cnt 个行 tile 复用 → (m_cnt−1) 次 L2);
|
||||
单 batch 输入 > L2 → 分组执行 (工作集思想, §4.3)。
|
||||
- **R6 (已知简化, 文档标注)**: ① 组内窗口复用未另计 (保守); ② BatchA≠BatchB (均 ≥2, 如 2 vs 64)
|
||||
的广播/混合 batch 结构按"每 batch 独立矩阵"计 GM = b·(a_b+b_b), 是 V_in 的**保守放大**
|
||||
(真实为 V_in + 共享矩阵跨 batch 重复读 ≤ 该值), 文档标注不做特殊建模; ③ K 切分 (k_L1<K) 的
|
||||
chunk 搬运按整段 k_L1 计 (含末段 padding 上取), GM ≥ V_in 恒成立。
|
||||
|
||||
### 3.1 逐分支 GM/L2 流量归属表 (现行模型正确性核查结果)
|
||||
|
||||
| 分支 | 结构 | GM 直读 | L2 重复读 | 核查 |
|
||||
|---|---|---|---|---|
|
||||
| IterBatch (切 B) | 逐 batch 整驻留/切 K; 每 (batch, K段) 数据互不重叠 | V_in (+K padding) | 0 (L1 形态吸收核内复用) | ✓ GM≥V_in |
|
||||
| MergeBatch | 合并组 (b0 batch) 一次搬入; K 段/整 K 均不跨段重读同一字节 | V_in (+K padding) | 0 | ✓ GM≥V_in |
|
||||
| ASW (切 M/N) | batch 内 tile 共享行/列块 | 场景 B: V_in; 场景 C: r_gm·V_in | (n_cnt−1)·a_b·b + (m_cnt−1)·b_b·b (场景 A/B) | ✓ GM≥V_in (场景 C 只放大) |
|
||||
| StreamK | 组内 K 段零重复读; 组间共享块同 ASW | V_in (命中时) | 组间共享块 (命中时) | ✓ |
|
||||
| 特殊分支 K=1 | AIV 通路, 每元素一次 | V_in | 0 | ✓ |
|
||||
| 转Matmul | 折叠单次 GEMM | V_in | 0 (Matmul 精切未展开) | ✓ |
|
||||
|
||||
→ **计算层面 GM 均 ≥ V_in**, 用户看到的 "GM < 输入" 来自**字节列每核/芯片口径混用**
|
||||
(MergeBatch/IterBatch 记每核值 ×32 才是整芯片; ASW/StreamK 记整芯片值), 不是少算。
|
||||
统一为整芯片口径 (修正项 ①), 并加不变量测试 (修正项 ②), 补齐 R3 整 case 边界 (修正项 ③)。
|
||||
|
||||
---
|
||||
|
||||
## 4. 场景模型 (输入侧) 与输出落点 (issue #30)
|
||||
|
||||
### 4.1 场景判定 (对切 M/N 及切 B 分支共用; L2 为整芯片 128MB)
|
||||
|
||||
- **S_A 整 case 全驻留**: V_in + V_out [+ workspace] ≤ L2 → 输出驻留 L2;
|
||||
- **S_B 单 batch 输入可驻留**: 不满足 S_A, 但 a_b + b_b ≤ L2 → 输入 batch 内共享块重复读命中 L2;
|
||||
- **S_C 单 batch 输入超 L2**: a_b + b_b > L2 → 分组执行 (输入工作集切分), 组间落空回 GM。
|
||||
|
||||
判定顺序: S_A → S_B → S_C。(原 #24 场景 A 的"单 batch 输入+输出可驻留"判定被 S_A 取代:
|
||||
整 case 输出跨 batch 累积时单 batch 判定不可靠, 见 #30。)
|
||||
|
||||
### 4.2 R4 输出落点规则 (取代一切"恒直写 GM / 单 batch 判定驻留")
|
||||
|
||||
> to_l2(输出写 L2 写口 5.2TB/s, GM 写流量=0, 异步回写不占算子时延 — #23 口径延续)
|
||||
> ⟺ **V_in + V_out [+ StreamK workspace] ≤ L2** (整 case 判定)
|
||||
> 否则输出**直写 GM**: 计入 GM 读写共享总线, 与读累加进 MTE2 搬移链 (#23 口径)。
|
||||
|
||||
理由: 输出字节只写一次无复用价值, L2 应优先保输入工作集 (输入存在重复读);
|
||||
只有当"整 case 输入+输出"可同时全驻留时才值得让输出占 L2 (省掉整条 GM 写时延)。
|
||||
各分支应用:
|
||||
|
||||
| 分支 | 原行为 | 新行为 |
|
||||
|---|---|---|
|
||||
| IterBatch / MergeBatch / 特殊分支 / 转Matmul | 恒 direct_gm | S_A 时 to_l2; 否则直写 GM |
|
||||
| ASW | 场景 A (单 batch in+out 驻留) 时 to_l2 | S_A (整 case) 时 to_l2; S_B/S_C 直写 GM |
|
||||
| StreamK | 归约内最终写回恒走 L2 | 最终写回按 S_A(+workspace); 不满足时写 GM (drain 内按 GM 带宽) |
|
||||
|
||||
### 4.3 场景 C 分组工作集 (沿用 #24 公式, 语义重申)
|
||||
|
||||
单 batch 输入超 L2 → 以 L2 为单 batch 活动窗口预算 D = L2 / (K·dt) (元素),
|
||||
组 = m_grp × n_grp 个 tile, 组间共享块复用落空回 GM:
|
||||
r_gm = (⌈n_cnt/n_grp⌉·M + ⌈m_cnt/m_grp⌉·N)/(M+N), GM = r_gm·V_in;
|
||||
组内窗口复用未另计 (保守, R6①)。分组追求"最小 L2 替换": 一次只让一组的工作集占 L2。
|
||||
|
||||
### 4.4 字节列与计数列语义 (修正 ①, 整芯片口径)
|
||||
|
||||
输出 CSV 中数据量列一律为**整芯片**口径: gm_read_bytes / l2_read_bytes /
|
||||
fixpipe_bytes / cube_flops = 整芯片量; 时延列 t_* 为墙钟时延 (已含核数折算);
|
||||
dma_cmd_count 为**单核**命令数 (各核 DMA 引擎并行执行, 墙钟 T_cmd = 单核命令数 × t_cmd),
|
||||
与字节列口径不同属, 单独标注。
|
||||
|
||||
---
|
||||
|
||||
## 5. 各分支 GM/输出计账公式落地 (与代码对应)
|
||||
|
||||
统一由 case 属性提供 V_in / V_out; 输出落点判断函数 `output_to_l2(case, spec, workspace)` (S_A)。
|
||||
|
||||
| 分支 | GM 读 (芯片) | L2 读 (芯片) | 输出 |
|
||||
|---|---|---|---|
|
||||
| MergeBatch | V_in (K截断); V_in·ceil(K/k_L1)·k_L1/K (L1绑定 padding, 模型按整段计) | 0 | S_A→L2 写口; else GM |
|
||||
| IterBatch | V_in (a/b 形态); 同上 padding 式 (c/d) | 0 | 同 MergeBatch |
|
||||
| ASW 切M/N | S_A/B: V_in; S_C: r_gm·V_in | S_A/B: b·[(n_cnt−1)·a_b + (m_cnt−1)·b_b] | S_A→L2; else GM |
|
||||
| StreamK | V_in (组间共享命中) | b·[(n_cnt−1)·a_b + (m_cnt−1)·b_b] (命中时) | 最终写回: S_A(+ws)→L2; else GM |
|
||||
| 特殊分支 | K=1: V_in; K=0: 0 | 0 | S_A→L2 写口; else GM (K=0 只有输出) |
|
||||
| 转Matmul | V_in (折叠后) | 0 | S_A→L2; else GM |
|
||||
|
||||
时延计算: 全核并发时 t = 芯片字节 / 芯片带宽; 降核 (used < C) 时按 used×单核份额
|
||||
(线性假设, issue#26 标注), Cube t = 芯片 flops / (used × q_cube(dtype))。
|
||||
|
||||
---
|
||||
|
||||
## 6. MergeBatch Cube 复核 (issue #27) — 证明
|
||||
|
||||
合并语义: A'[b0·M, K] @ B'[K, b0·N] 作为单次 GEMM, Cube 硬件计算**全网格**
|
||||
(b0·M)×(b0·N) 输出 (含交叉项, 冗余比例 (b0²−b0)/b0² = (b0−1)/b0), 只保留 b0 个对角块。
|
||||
|
||||
- 每次合并计算 (步): flops_step = 2·(b0·M)·(b0·N)·K;
|
||||
- 每核合并组数: b_core/b0;
|
||||
- 总: flops = (b_core/b0)·2·(b0·M)·(b0·N)·K = **b_core·b0·2MNK** —— 与现行代码
|
||||
`flops_pc = b_core·b0·2·m·n·k` 逐项相等, t_mmad = flops/Q16 不变。
|
||||
|
||||
CSV 证据 (merge_demo_k_trunc, B=2048 M=N=32 K=256, b0=4, b_core=64):
|
||||
flops_step = 2·128·128·256 = 8,388,608; 步数 16; 合计 134,217,728 = 代码值 = CSV cube_flops 列;
|
||||
t_mmad = 134,217,728 / 15.1875e12 = 8.837us = CSV t_mmad。**结论: 无需数值修改**;
|
||||
若意图是"只算对角块"则与全网格 GEMM 语义冲突 (冗余消失, v0.98 §五 理论前提需另行裁决)。
|
||||
|
||||
---
|
||||
|
||||
## 7. 变更影响与验证口径
|
||||
|
||||
1. 字节列整芯片化 (IterBatch/MergeBatch/special 数值 ×C 或按 b 直接计算);
|
||||
2. S_A 判定下输出写 L2: 小 case (整 case 可全驻留) GM 写流量清零、输出时延降为 L2 写口;
|
||||
3. S_A 之外输出 GM 写流量如实计入共享总线 → 修正单 batch 驻留判定导致的漏计;
|
||||
4. dtype 感知算力: bf16 案例数值零回退; fp32/fp8/fp4/… 案例时延按 §2.1 表修正;
|
||||
5. 新增回归: GM≥V_in 不变量 (全分支随机集)、S_A 输出落点开关、dtype 速率比例、
|
||||
fp32 Cube 时延 = bf16 的 2 倍、字节列整芯片口径等;
|
||||
6. examples 三件套重生成, 与模型改动前 diff 隔离后入库; 压力回归 0 违规/0 NaN。
|
||||
@@ -2,44 +2,44 @@
|
||||
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"
|
||||
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"
|
||||
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)"
|
||||
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]"
|
||||
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
|
||||
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"
|
||||
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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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]; 输出落点: L2驻留 (整case V_in+V_out=64.0MB vs L2=128MB)"
|
||||
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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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; 输出落点: L2驻留
|
||||
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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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; 输出落点: L2驻留
|
||||
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(整case超L2: 输入优先驻留L2, 输出直写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; 输出落点: 直写GM"
|
||||
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_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)"
|
||||
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)"
|
||||
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),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"
|
||||
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, 输入优先驻留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(整case超L2, 输出直写GM) (V_in=64.0MB, V_out=256.0MB, L2=128MB), 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,resident(整case全驻留S_A),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,resident(整case全驻留S_A),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,resident(整case全驻留S_A),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)"
|
||||
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,resident(整case全驻留S_A),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)"
|
||||
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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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; 输出落点: L2驻留
|
||||
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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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; 输出落点: L2驻留"
|
||||
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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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; 输出落点: L2驻留
|
||||
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, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,47,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=3591.0MB, V_out=8.0MB, L2=128MB), 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),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)"
|
||||
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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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; 输出落点: L2驻留"
|
||||
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(整case超L2: 输入优先驻留L2, 输出直写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; 输出落点: 直写GM"
|
||||
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, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,576,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=1536.0MB, V_out=1024.0MB, L2=128MB), 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),"direct_gm(输出直写GM, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,564,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=2304.0MB, V_out=1024.0MB, L2=128MB), 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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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; 输出落点: L2驻留"
|
||||
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, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,2304,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=16384.0MB, V_out=4096.0MB, L2=128MB), 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, 输入优先驻留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(组间落空回GM) (V_in=5376.0MB, V_out=2048.0MB, L2=128MB), 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),"direct_gm(输出直写GM, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,144,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=2048.0MB, V_out=256.0MB, L2=128MB), 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),"direct_gm(输出直写GM, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,576,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=96.0MB, V_out=1024.0MB, L2=128MB), 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),"direct_gm(输出直写GM, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,18,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=512.0MB, V_out=32.0MB, L2=128MB), 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(整case全驻留S_A: 输出驻留L2异步回写, GM写=0)",4,0,方案B,205,1,1,205,1,12,24,True,2,"L2场景: A_整case全驻留(输入+输出<=L2) (V_in=32.1MB, V_out=32.0MB, L2=128MB), 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, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,561,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=272.0MB, V_out=1024.0MB, L2=128MB), 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),"direct_gm(输出直写GM, 输入优先驻留L2)",4,0,方案B,820,1,1,820,1,4,94,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=128.1MB, V_out=128.0MB, L2=128MB), 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, 输入优先驻留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(整case超L2, 输出直写GM) (V_in=516.0MB, V_out=2048.0MB, L2=128MB), 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, 输入优先驻留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(组间落空回GM) (V_in=8960.0MB, V_out=8192.0MB, L2=128MB), 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, 输入优先驻留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(组间落空回GM) (V_in=2056.0MB, V_out=32.0MB, L2=128MB), 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, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,35015,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=1088.0MB, V_out=65536.0MB, L2=128MB), 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, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,17578,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=7680.0MB, V_out=32768.0MB, L2=128MB), 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, 输入优先驻留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(组间落空回GM) (V_in=17408.0MB, V_out=16384.0MB, L2=128MB), 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),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"
|
||||
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, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,492,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=7168.0MB, V_out=896.0MB, L2=128MB), 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, 输入优先驻留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(组间落空回GM) (V_in=7168.0MB, V_out=4096.0MB, L2=128MB), 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),"direct_gm(输出直写GM, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,144,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=16.0MB, V_out=256.0MB, L2=128MB), 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, 输入优先驻留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(组间落空回GM) (V_in=28672.0MB, V_out=16384.0MB, L2=128MB), 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)"
|
||||
|
||||
|
@@ -1,45 +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,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"
|
||||
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-05,8388608,1.6131938461538462e-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",0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,16777216,3.2263876923076923e-06,0.0,3.2263876923076923e-06,0.0,3.2263876923076923e-06,FIXPIPE,True,,写出Bound(L2写口),"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-08,0.0,0.0,8388608.0,3.103030303030303e-07,16777216,3.2263876923076923e-06,0.0,3.2263876923076923e-06,0.0,3.2263876923076923e-06,FIXPIPE,True,,写出Bound(L2写口),"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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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]; 输出落点: L2驻留 (整case V_in+V_out=64.0MB vs L2=128MB)",67108864,0.0,4.194304e-05,0.0,4.2743039999999997e-05,16.0,8.000000000000001e-07,4294967296.0,8.837381267489712e-06,4194304,8.065969230769231e-07,0.0,4.2743039999999997e-05,1.8849638999683443e-07,4.293153638999683e-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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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; 输出落点: L2驻留,16777216,0.0,1.048576e-05,0.0,1.0685759999999999e-05,4,2.0000000000000002e-07,536870912.0,1.104672658436214e-06,1048576,2.0164923076923077e-07,0.0,1.0685759999999999e-05,3.265804723013612e-07,1.101234047230136e-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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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; 输出落点: L2驻留,8388608,0.0,5.24288e-06,0.0,5.4428799999999995e-06,4,2.0000000000000002e-07,268435456.0,5.52336329218107e-07,1048576,2.0164923076923077e-07,0.0,5.4428799999999995e-06,1.8849638999683443e-07,5.631376389996834e-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(整case超L2: 输入优先驻留L2, 输出直写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; 输出落点: 直写GM",134217728,0.0,8.388608e-05,0.0,8.468608e-05,16,8.000000000000001e-07,4294967296.0,8.837381267489712e-06,524288,3.2768e-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),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"
|
||||
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, 输入优先驻留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(整case超L2, 输出直写GM) (V_in=64.0MB, V_out=256.0MB, L2=128MB), 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,resident(整case全驻留S_A),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,resident(整case全驻留S_A),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,resident(整case全驻留S_A),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,resident(整case全驻留S_A),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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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; 输出落点: L2驻留,2228224,0.0,1.39264e-06,0.0,1.44264e-06,1,5.0000000000000004e-08,16777216.0,3.452102057613169e-08,65536,1.2603076923076923e-08,0.0,1.44264e-06,4.712409749920861e-08,1.4897640974992086e-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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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; 输出落点: L2驻留",25067520,0.0,1.56672e-05,0.0,1.5967200000000002e-05,6,3.0000000000000004e-07,268435456.0,5.52336329218107e-07,524288,1.0082461538461538e-07,0.0,1.5967200000000002e-05,1.798661348528015e-07,1.6147066134852802e-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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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; 输出落点: L2驻留,13107200,0.0,8.192e-06,0.0,8.392e-06,4,2.0000000000000002e-07,100663296.0,2.0712612345679012e-07,393216,7.561846153846153e-08,0.0,8.392e-06,7.068614624881291e-08,8.462686146248813e-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),"direct_gm(输出直写GM, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,47,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=3591.0MB, V_out=8.0MB, L2=128MB), GM首读倍率=1.00; 尾轮: r=0 无尾轮",3765436416.0,337641472,0.00235339776,6.493105230769231e-05,0.0024183288123076924,0.0,0.0,60129542144.0,0.00012372333774485596,8388608,5.24288e-06,0.0,0.0024235716923076923,0.0,0.0024235716923076923,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-06,0.0,0.0,1.731729603729604e-06,0.00033558528,1.731729603729604e-06,0.0003373170096037296,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-06,0.0,0.0,1.7726896037296038e-06,0.00058748928,1.7726896037296038e-06,0.0005892619696037297,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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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; 输出落点: L2驻留",58720256,0.0,3.6700159999999995e-05,0.0,3.705016e-05,7,3.5000000000000004e-07,1879048192.0,3.866354304526749e-06,262144,5.041230769230769e-08,0.0,3.705016e-05,6.027486369104147e-07,3.7652908636910414e-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(整case超L2: 输入优先驻留L2, 输出直写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; 输出落点: 直写GM",1879048192,0.0,0.0011744051199999998,0.0,0.00118560512,224,1.1200000000000001e-05,962072674304.0,0.0019795734039176954,134217728,8.388608e-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, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,576,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=1536.0MB, V_out=1024.0MB, L2=128MB), GM首读倍率=1.00; 尾轮: r=0 无尾轮",1610612736.0,17716740096,0.00100663296,0.003407065403076923,0.004413698363076923,0.0,0.0,1649267441664.0,0.0033935544067160493,1073741824,0.00067108864,0.0,0.005084787003076923,0.0,0.005084787003076923,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),"direct_gm(输出直写GM, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,564,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=2304.0MB, V_out=1024.0MB, L2=128MB), GM首读倍率=1.00; 尾轮: r=0 无尾轮",2415919104.0,23085449216,0.00150994944,0.004439509464615385,0.005949458904615385,0.0,0.0,2199023255552.0,0.004524739208954733,1073741824,0.00067108864,0.0,0.006620547544615385,0.0,0.006620547544615385,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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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; 输出落点: L2驻留",67108864,0.0,4.194304e-05,0.0,4.234304e-05,8,4.0000000000000003e-07,34359738368.0,7.06990501399177e-05,67108864,1.290555076923077e-05,0.0,7.06990501399177e-05,2.1742932036720483e-05,9.244198217663819e-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"
|
||||
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, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,2304,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=16384.0MB, V_out=4096.0MB, L2=128MB), 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, 输入优先驻留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(组间落空回GM) (V_in=5376.0MB, V_out=2048.0MB, L2=128MB), 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),"direct_gm(输出直写GM, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,144,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=2048.0MB, V_out=256.0MB, L2=128MB), GM首读倍率=1.00; 尾轮: r=0 无尾轮",2147483648.0,23622320128,0.00134217728,0.004542753870769231,0.005884931150769231,0.0,0.0,2199023255552.0,0.004524739208954733,268435456,0.00016777216,0.0,0.006052703310769231,0.0,0.006052703310769231,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),"direct_gm(输出直写GM, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,576,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=96.0MB, V_out=1024.0MB, L2=128MB), GM首读倍率=1.00; 尾轮: r=0 无尾轮",100663296.0,1509949440,6.291456e-05,0.0002903748923076923,0.0003532894523076923,0.0,0.0,137438953472.0,0.0002827962005596708,1073741824,0.00067108864,0.0,0.0010243780923076921,0.0,0.0010243780923076921,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),"direct_gm(输出直写GM, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,18,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=512.0MB, V_out=32.0MB, L2=128MB), GM首读倍率=1.00; 尾轮: r=0 无尾轮",536870912.0,2684354560,0.00033554432,0.0005162220307692308,0.0008517663507692308,0.0,0.0,274877906944.0,0.0005655924011193416,33554432,2.097152e-05,0.0,0.0008727378707692308,0.0,0.0008727378707692308,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(整case全驻留S_A: 输出驻留L2异步回写, GM写=0)",4,0,方案B,205,1,1,205,1,12,24,True,2,"L2场景: A_整case全驻留(输入+输出<=L2) (V_in=32.1MB, V_out=32.0MB, L2=128MB), 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, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,561,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=272.0MB, V_out=1024.0MB, L2=128MB), 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),"direct_gm(输出直写GM, 输入优先驻留L2)",4,0,方案B,820,1,1,820,1,4,94,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=128.1MB, V_out=128.0MB, L2=128MB), 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-05,134217728,8.388608e-05,0.0,0.00018660745846153846,0.0,0.00018660745846153846,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, 输入优先驻留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(整case超L2, 输出直写GM) (V_in=516.0MB, V_out=2048.0MB, L2=128MB), 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, 输入优先驻留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(组间落空回GM) (V_in=8960.0MB, V_out=8192.0MB, L2=128MB), 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, 输入优先驻留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(组间落空回GM) (V_in=2056.0MB, V_out=32.0MB, L2=128MB), 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, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,35015,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=1088.0MB, V_out=65536.0MB, L2=128MB), 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, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,17578,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=7680.0MB, V_out=32768.0MB, L2=128MB), 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, 输入优先驻留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(组间落空回GM) (V_in=17408.0MB, V_out=16384.0MB, L2=128MB), 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-06,0.0,0.0,2.566079254079254e-07,0.00011042185846153846,2.566079254079254e-07,0.00011067846638694638,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),"direct_gm(输出直写GM, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,492,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=7168.0MB, V_out=896.0MB, L2=128MB), GM首读倍率=1.00; 尾轮: r=0 无尾轮",7516192768.0,70464307200,0.00469762048,0.013550828307692308,0.018248448787692308,0.0,0.0,6734508720128.0,0.013857013827423869,939524096,0.00058720256,0.0,0.018835651347692307,0.0,0.018835651347692307,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, 输入优先驻留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(组间落空回GM) (V_in=7168.0MB, V_out=4096.0MB, L2=128MB), 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),"direct_gm(输出直写GM, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,144,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=16.0MB, V_out=256.0MB, L2=128MB), 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-05,268435456,0.00016777216,0.0,0.0002524648369230769,0.0,0.0002524648369230769,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, 输入优先驻留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(组间落空回GM) (V_in=28672.0MB, V_out=16384.0MB, L2=128MB), 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-07,0.0,0.0,268435456.0,9.92969696969697e-06,536870912,0.00033554432,0.0,0.00033624063999999996,0.0,0.00033624063999999996,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"
|
||||
|
||||
|
@@ -1,45 +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,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
|
||||
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-05,8388608,1.6131938461538462e-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",0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,16777216,3.2263876923076923e-06,0.0,3.2263876923076923e-06,0.0,3.2263876923076923e-06,FIXPIPE,True,,写出Bound(L2写口),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-08,0.0,0.0,8388608.0,3.103030303030303e-07,16777216,3.2263876923076923e-06,0.0,3.2263876923076923e-06,0.0,3.2263876923076923e-06,FIXPIPE,True,,写出Bound(L2写口),"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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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]; 输出落点: L2驻留 (整case V_in+V_out=64.0MB vs L2=128MB)",67108864,0.0,4.194304e-05,0.0,4.2743039999999997e-05,16.0,8.000000000000001e-07,4294967296.0,8.837381267489712e-06,4194304,8.065969230769231e-07,0.0,4.2743039999999997e-05,1.8849638999683443e-07,4.293153638999683e-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=42.93us vs T_IterBatch=45.19us -> 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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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; 输出落点: L2驻留,16777216,0.0,1.048576e-05,0.0,1.0685759999999999e-05,4,2.0000000000000002e-07,536870912.0,1.104672658436214e-06,1048576,2.0164923076923077e-07,0.0,1.0685759999999999e-05,3.265804723013612e-07,1.101234047230136e-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=11.26us vs T_IterBatch=11.01us -> 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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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; 输出落点: L2驻留,8388608,0.0,5.24288e-06,0.0,5.4428799999999995e-06,4,2.0000000000000002e-07,268435456.0,5.52336329218107e-07,1048576,2.0164923076923077e-07,0.0,5.4428799999999995e-06,1.8849638999683443e-07,5.631376389996834e-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(整case超L2: 输入优先驻留L2, 输出直写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; 输出落点: 直写GM",134217728,0.0,8.388608e-05,0.0,8.468608e-05,16,8.000000000000001e-07,4294967296.0,8.837381267489712e-06,524288,3.2768e-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),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"
|
||||
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, 输入优先驻留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(整case超L2, 输出直写GM) (V_in=64.0MB, V_out=256.0MB, L2=128MB), 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,resident(整case全驻留S_A),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,resident(整case全驻留S_A),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,resident(整case全驻留S_A),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,resident(整case全驻留S_A),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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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; 输出落点: L2驻留,2228224,0.0,1.39264e-06,0.0,1.44264e-06,1,5.0000000000000004e-08,16777216.0,3.452102057613169e-08,65536,1.2603076923076923e-08,0.0,1.44264e-06,4.712409749920861e-08,1.4897640974992086e-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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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; 输出落点: L2驻留",25067520,0.0,1.56672e-05,0.0,1.5967200000000002e-05,6,3.0000000000000004e-07,268435456.0,5.52336329218107e-07,524288,1.0082461538461538e-07,0.0,1.5967200000000002e-05,1.798661348528015e-07,1.6147066134852802e-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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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; 输出落点: L2驻留,13107200,0.0,8.192e-06,0.0,8.392e-06,4,2.0000000000000002e-07,100663296.0,2.0712612345679012e-07,393216,7.561846153846153e-08,0.0,8.392e-06,7.068614624881291e-08,8.462686146248813e-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),"direct_gm(输出直写GM, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,47,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=3591.0MB, V_out=8.0MB, L2=128MB), GM首读倍率=1.00; 尾轮: r=0 无尾轮",3765436416.0,337641472,0.00235339776,6.493105230769231e-05,0.0024183288123076924,0.0,0.0,60129542144.0,0.00012372333774485596,8388608,5.24288e-06,0.0,0.0024235716923076923,0.0,0.0024235716923076923,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-06,0.0,0.0,1.731729603729604e-06,0.00033558528,1.731729603729604e-06,0.0003373170096037296,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-06,0.0,0.0,1.7726896037296038e-06,0.00058748928,1.7726896037296038e-06,0.0005892619696037297,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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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; 输出落点: L2驻留",58720256,0.0,3.6700159999999995e-05,0.0,3.705016e-05,7,3.5000000000000004e-07,1879048192.0,3.866354304526749e-06,262144,5.041230769230769e-08,0.0,3.705016e-05,6.027486369104147e-07,3.7652908636910414e-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(整case超L2: 输入优先驻留L2, 输出直写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; 输出落点: 直写GM",1879048192,0.0,0.0011744051199999998,0.0,0.00118560512,224,1.1200000000000001e-05,962072674304.0,0.0019795734039176954,134217728,8.388608e-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, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,576,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=1536.0MB, V_out=1024.0MB, L2=128MB), GM首读倍率=1.00; 尾轮: r=0 无尾轮",1610612736.0,17716740096,0.00100663296,0.003407065403076923,0.004413698363076923,0.0,0.0,1649267441664.0,0.0033935544067160493,1073741824,0.00067108864,0.0,0.005084787003076923,0.0,0.005084787003076923,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),"direct_gm(输出直写GM, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,564,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=2304.0MB, V_out=1024.0MB, L2=128MB), GM首读倍率=1.00; 尾轮: r=0 无尾轮",2415919104.0,23085449216,0.00150994944,0.004439509464615385,0.005949458904615385,0.0,0.0,2199023255552.0,0.004524739208954733,1073741824,0.00067108864,0.0,0.006620547544615385,0.0,0.006620547544615385,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),"resident(整case输入+输出<=L2: 输出驻留L2异步回写, GM写=0)",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; 输出落点: L2驻留",67108864,0.0,4.194304e-05,0.0,4.234304e-05,8,4.0000000000000003e-07,34359738368.0,7.06990501399177e-05,67108864,1.290555076923077e-05,0.0,7.06990501399177e-05,2.1742932036720483e-05,9.244198217663819e-05,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, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,2304,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=16384.0MB, V_out=4096.0MB, L2=128MB), 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, 输入优先驻留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(组间落空回GM) (V_in=5376.0MB, V_out=2048.0MB, L2=128MB), 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),"direct_gm(输出直写GM, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,144,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=2048.0MB, V_out=256.0MB, L2=128MB), GM首读倍率=1.00; 尾轮: r=0 无尾轮",2147483648.0,23622320128,0.00134217728,0.004542753870769231,0.005884931150769231,0.0,0.0,2199023255552.0,0.004524739208954733,268435456,0.00016777216,0.0,0.006052703310769231,0.0,0.006052703310769231,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),"direct_gm(输出直写GM, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,576,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=96.0MB, V_out=1024.0MB, L2=128MB), GM首读倍率=1.00; 尾轮: r=0 无尾轮",100663296.0,1509949440,6.291456e-05,0.0002903748923076923,0.0003532894523076923,0.0,0.0,137438953472.0,0.0002827962005596708,1073741824,0.00067108864,0.0,0.0010243780923076921,0.0,0.0010243780923076921,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),"direct_gm(输出直写GM, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,18,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=512.0MB, V_out=32.0MB, L2=128MB), GM首读倍率=1.00; 尾轮: r=0 无尾轮",536870912.0,2684354560,0.00033554432,0.0005162220307692308,0.0008517663507692308,0.0,0.0,274877906944.0,0.0005655924011193416,33554432,2.097152e-05,0.0,0.0008727378707692308,0.0,0.0008727378707692308,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(整case全驻留S_A: 输出驻留L2异步回写, GM写=0)",4,0,方案B,205,1,1,205,1,12,24,True,2,"L2场景: A_整case全驻留(输入+输出<=L2) (V_in=32.1MB, V_out=32.0MB, L2=128MB), 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, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,561,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=272.0MB, V_out=1024.0MB, L2=128MB), 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),"direct_gm(输出直写GM, 输入优先驻留L2)",4,0,方案B,820,1,1,820,1,4,94,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=128.1MB, V_out=128.0MB, L2=128MB), 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-05,134217728,8.388608e-05,0.0,0.00018660745846153846,0.0,0.00018660745846153846,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, 输入优先驻留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(整case超L2, 输出直写GM) (V_in=516.0MB, V_out=2048.0MB, L2=128MB), 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, 输入优先驻留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(组间落空回GM) (V_in=8960.0MB, V_out=8192.0MB, L2=128MB), 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, 输入优先驻留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(组间落空回GM) (V_in=2056.0MB, V_out=32.0MB, L2=128MB), 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, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,35015,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=1088.0MB, V_out=65536.0MB, L2=128MB), 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, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,17578,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=7680.0MB, V_out=32768.0MB, L2=128MB), 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, 输入优先驻留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(组间落空回GM) (V_in=17408.0MB, V_out=16384.0MB, L2=128MB), 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-06,0.0,0.0,2.566079254079254e-07,0.00011042185846153846,2.566079254079254e-07,0.00011067846638694638,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),"direct_gm(输出直写GM, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,492,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=7168.0MB, V_out=896.0MB, L2=128MB), GM首读倍率=1.00; 尾轮: r=0 无尾轮",7516192768.0,70464307200,0.00469762048,0.013550828307692308,0.018248448787692308,0.0,0.0,6734508720128.0,0.013857013827423869,939524096,0.00058720256,0.0,0.018835651347692307,0.0,0.018835651347692307,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, 输入优先驻留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(组间落空回GM) (V_in=7168.0MB, V_out=4096.0MB, L2=128MB), 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),"direct_gm(输出直写GM, 输入优先驻留L2)",4,0,A0,1,1,1,0,0,0,144,True,2,"L2场景: B_单batch输入驻留(整case超L2, 输出直写GM) (V_in=16.0MB, V_out=256.0MB, L2=128MB), 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-05,268435456,0.00016777216,0.0,0.0002524648369230769,0.0,0.0002524648369230769,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, 输入优先驻留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(组间落空回GM) (V_in=28672.0MB, V_out=16384.0MB, L2=128MB), 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-07,0.0,0.0,268435456.0,9.92969696969697e-06,536870912,0.00033554432,0.0,0.00033624063999999996,0.0,0.00033624063999999996,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"
|
||||
|
||||
|
@@ -137,11 +137,13 @@ class TestTimingSanity(unittest.TestCase):
|
||||
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
|
||||
# C 指定 fp16 输出时, 写出量按 2B 而非 L0C 的 4B;
|
||||
# 字节列为整芯片口径 (issue#29): fixpipe_bytes = B*MN*outB
|
||||
case = mkcase(128, 64, 64, 512, dtype_c="fp16")
|
||||
ib = IterBatchBranch().analyze(case)
|
||||
expect = ib.plan.b_core * 64 * 64 * 2
|
||||
expect = case.batch_c * 64 * 64 * 2
|
||||
self.assertAlmostEqual(ib.timing.fixpipe_bytes, expect)
|
||||
self.assertAlmostEqual(ib.timing.fixpipe_bytes, case.output_bytes)
|
||||
|
||||
|
||||
class TestIssueRegression(unittest.TestCase):
|
||||
@@ -423,5 +425,172 @@ class TestZeroCmdHandling(unittest.TestCase):
|
||||
self.assertEqual(t.t_total, t.t_total)
|
||||
|
||||
|
||||
class TestIssue27to30(unittest.TestCase):
|
||||
"""第四轮评审 (issue#27-#30, 设计文档 docs/05).
|
||||
|
||||
#27 MergeBatch Cube 公式复核 (每步 2*(b0M)(b0N)K x b_core/b0 步);
|
||||
#28 dtype 感知算力 (Cube/AIV 速率表);
|
||||
#29 GM 首读下限 GM>=V_in + 字节列整芯片口径;
|
||||
#30 Fixpipe 输出落点 (整case驻留 -> L2, 否则直写 GM).
|
||||
"""
|
||||
|
||||
def setUp(self):
|
||||
from bmm_theory.hardware import ASCEND950PR
|
||||
self.s = ASCEND950PR
|
||||
self.router = BranchRouter()
|
||||
|
||||
# ---------------- #27 ----------------
|
||||
def test_issue27_merge_cube_flops_matches_formula(self):
|
||||
from bmm_theory.branches.merge_batch import MergeBatchBranch
|
||||
case = mkcase(2048, 32, 32, 256)
|
||||
mb = MergeBatchBranch().analyze(case)
|
||||
self.assertTrue(mb.capable)
|
||||
p, t = mb.plan, mb.timing
|
||||
b0, bc = p.merge_b0, p.b_core
|
||||
# 整芯片列 = B*b0*2MNK; 等价 (b_core/b0) 步 x 每步 2*(b0M)(b0N)K x C 核
|
||||
step_flops = 2.0 * (b0 * 32) * (b0 * 32) * 256
|
||||
self.assertAlmostEqual(t.cube_flops,
|
||||
step_flops * (bc / b0) * self.s.aic_num)
|
||||
self.assertAlmostEqual(t.cube_flops,
|
||||
case.batch_c * b0 * 2.0 * 32 * 32 * 256)
|
||||
# t_mmad = 每核 flops / 单核算力
|
||||
self.assertAlmostEqual(t.t_mmad,
|
||||
(bc * b0 * 2.0 * 32 * 32 * 256) / self.s.q16)
|
||||
|
||||
def test_issue27_merge_gm_chip_col_equals_input(self):
|
||||
# K 截断合并: GM 每字节一次 -> gm 整芯片列 = V_in (issue#29 口径)
|
||||
from bmm_theory.branches.merge_batch import MergeBatchBranch
|
||||
case = mkcase(2048, 32, 32, 256)
|
||||
t = MergeBatchBranch().analyze(case).timing
|
||||
self.assertAlmostEqual(t.gm_read_bytes, case.input_bytes)
|
||||
self.assertAlmostEqual(t.l2_read_bytes, 0.0)
|
||||
|
||||
# ---------------- #28 ----------------
|
||||
def test_issue28_cube_dtype_rate(self):
|
||||
from bmm_theory.branches.iter_batch import IterBatchBranch
|
||||
base = mkcase(128, 64, 64, 512) # IterBatch 形态 b, 各 dtype 均可行
|
||||
t16 = IterBatchBranch().analyze(base).timing
|
||||
t32 = IterBatchBranch().analyze(
|
||||
mkcase(128, 64, 64, 512, dtype_a="fp32", dtype_b="fp32")).timing
|
||||
t8 = IterBatchBranch().analyze(
|
||||
mkcase(128, 64, 64, 512, dtype_a="fp8", dtype_b="fp8")).timing
|
||||
# 同计算量, 时延反比于 dtype 算力: fp32=1/2, fp8=2x (相对 bf16)
|
||||
self.assertAlmostEqual(t32.t_mmad / t16.t_mmad, 2.0, places=6)
|
||||
self.assertAlmostEqual(t8.t_mmad / t16.t_mmad, 0.5, places=6)
|
||||
|
||||
def test_issue28_aiv_dtype_rate_k1(self):
|
||||
# K=1 AIV 逐元素: bf16 通量 2x fp32 -> fp32 时延为 bf16 的 2 倍
|
||||
base = mkcase(64, 8192, 512, 1)
|
||||
t_bf16 = BranchRouter().route(base)["timing"]
|
||||
t_fp32 = BranchRouter().route(
|
||||
mkcase(64, 8192, 512, 1, dtype_a="fp32", dtype_b="fp32"))["timing"]
|
||||
self.assertAlmostEqual(t_fp32.t_mmad / t_bf16.t_mmad, 2.0, places=6)
|
||||
|
||||
# ---------------- #29 ----------------
|
||||
def test_issue29_gm_floor_and_chip_columns(self):
|
||||
# 各分支代表 case: GM 整芯片列 >= V_in (R1), 且迭代/合并/转matmul/special
|
||||
# 等于 V_in (无重复读结构)
|
||||
from bmm_theory.branches.iter_batch import IterBatchBranch
|
||||
from bmm_theory.branches.merge_batch import MergeBatchBranch
|
||||
from bmm_theory.branches.stream_k import StreamKBranch
|
||||
from bmm_theory.branches.asw_basic import AswBasicBranch
|
||||
checks = [
|
||||
(IterBatchBranch().analyze(mkcase(128, 64, 64, 512)).timing,
|
||||
mkcase(128, 64, 64, 512)),
|
||||
(MergeBatchBranch().analyze(mkcase(2048, 32, 32, 256)).timing,
|
||||
mkcase(2048, 32, 32, 256)),
|
||||
(StreamKBranch().analyze(mkcase(4, 128, 128, 10240)).timing,
|
||||
mkcase(4, 128, 128, 10240)),
|
||||
(AswBasicBranch().analyze(mkcase(8, 4096, 4096, 1024)).timing,
|
||||
mkcase(8, 4096, 4096, 1024)),
|
||||
]
|
||||
for t, case in checks:
|
||||
self.assertGreaterEqual(t.gm_read_bytes + 1e-6, case.input_bytes,
|
||||
f"{case.case_id} GM < V_in")
|
||||
# 转Matmul / 特殊分支 (K=1) 亦等于 V_in
|
||||
r = self.router.route(mkcase(1, 2048, 2048, 2048))
|
||||
t = r["timing"]
|
||||
self.assertGreaterEqual(t.gm_read_bytes + 1e-6,
|
||||
BmmCase(case_id="x", batch_a=1, batch_b=1,
|
||||
m=2048, n=2048, k=2048).input_bytes)
|
||||
r = self.router.route(mkcase(128, 256, 256, 1))
|
||||
self.assertAlmostEqual(r["timing"].gm_read_bytes, 128 * (256 + 256) * 2)
|
||||
|
||||
def test_issue29_gm_floor_random_smoke(self):
|
||||
# 随机多 dtype 冒烟: 所有可行方案 GM 整芯片列 >= V_in, 无 NaN
|
||||
import random
|
||||
rng = random.Random(20260609)
|
||||
dtypes = ["bf16", "fp16", "fp32", "int8", "fp8", "fp4"]
|
||||
for _ in range(400):
|
||||
dt = rng.choice(dtypes)
|
||||
kw = dict(dtype_a=dt, dtype_b=dt, dtype_c=rng.choice(["bf16", "fp16"]))
|
||||
b = rng.choice([1, 2, 4, 8, 16, 32, 64, 128, 256, 2048])
|
||||
m = rng.choice([1, 8, 32, 64, 128, 256, 1024, 4096])
|
||||
n = rng.choice([1, 8, 32, 64, 128, 256, 1024, 4096])
|
||||
k = rng.choice([0, 1, 32, 64, 128, 256, 512, 1024, 4096])
|
||||
c = BmmCase(case_id="gm", batch_a=b, batch_b=b, m=m, n=n, k=k, **kw)
|
||||
r = self.router.route(c)
|
||||
t = r["timing"]
|
||||
self.assertIsNotNone(t, c.case_id)
|
||||
self.assertEqual(t.t_total, t.t_total) # 非 NaN
|
||||
self.assertGreaterEqual(t.gm_read_bytes + 1e-6, c.input_bytes,
|
||||
f"{dt} b={b} m={m} n={n} k={k} gm={t.gm_read_bytes}")
|
||||
|
||||
# ---------------- #30 ----------------
|
||||
def test_issue30_iter_output_l2_when_whole_case_fits(self):
|
||||
from bmm_theory.branches.iter_batch import IterBatchBranch
|
||||
# 整 case 输入+输出 16.8MB+1.0MB <= 128MB -> 输出写 L2 写口, GM 写=0
|
||||
case = mkcase(128, 64, 64, 512)
|
||||
self.assertLess(case.input_bytes + case.output_bytes, self.s.l2_bytes)
|
||||
t = IterBatchBranch().analyze(case).timing
|
||||
self.assertAlmostEqual(t.fixpipe_bytes, case.output_bytes)
|
||||
self.assertAlmostEqual(t.t_fixpipe, case.output_bytes / self.s.bw_l2)
|
||||
|
||||
def test_issue30_iter_output_gm_when_case_overflows(self):
|
||||
from bmm_theory.branches.iter_batch import IterBatchBranch
|
||||
# 输出 268MB > L2 余量 -> 直写 GM (输入优先驻留)
|
||||
case = mkcase(128, 1024, 1024, 64)
|
||||
self.assertGreater(case.input_bytes + case.output_bytes, self.s.l2_bytes)
|
||||
t = IterBatchBranch().analyze(case).timing
|
||||
self.assertAlmostEqual(t.fixpipe_bytes, case.output_bytes)
|
||||
self.assertAlmostEqual(t.t_fixpipe, case.output_bytes / self.s.bw_gm)
|
||||
|
||||
def test_issue30_asw_scene_whole_case_labels(self):
|
||||
# S_A (整case全驻留): l2_policy_out=resident, t_fixpipe 走 L2 写口
|
||||
case_sa = mkcase(2, 4096, 4096, 512)
|
||||
self.assertLess(case_sa.input_bytes + case_sa.output_bytes, self.s.l2_bytes)
|
||||
r = self.router.route(case_sa)
|
||||
self.assertEqual(r["branch"], "ASW_Basic")
|
||||
self.assertTrue(r["plan"].l2_policy_out.startswith("resident"), r["plan"].l2_policy_out)
|
||||
self.assertIn("A_整case全驻留", r["plan"].note)
|
||||
self.assertAlmostEqual(r["timing"].t_fixpipe,
|
||||
case_sa.output_bytes / self.s.bw_l2)
|
||||
# S_B (整case超L2, 单batch输入可驻留): 输出直写 GM
|
||||
case_sb = mkcase(64, 4096, 4096, 1024)
|
||||
self.assertGreater(case_sb.input_bytes + case_sb.output_bytes, self.s.l2_bytes)
|
||||
self.assertLess(4096 * 1024 * 2 * 2, self.s.l2_bytes) # 单batch输入 16.8MB
|
||||
r = self.router.route(case_sb)
|
||||
self.assertIn("ASW", r["branch"])
|
||||
self.assertTrue(r["plan"].l2_policy_out.startswith("direct_gm"),
|
||||
r["plan"].l2_policy_out)
|
||||
self.assertAlmostEqual(r["timing"].t_fixpipe,
|
||||
case_sb.output_bytes / self.s.bw_gm)
|
||||
|
||||
def test_issue30_to_matmul_output_l2_toggle(self):
|
||||
# 转Matmul 粗估: 整case可驻留 -> L2; 大 case -> GM
|
||||
small = mkcase(1, 2048, 2048, 2048)
|
||||
self.assertLess(small.input_bytes + small.output_bytes, self.s.l2_bytes)
|
||||
r = self.router.route(small)
|
||||
self.assertEqual(r["branch"], "转Matmul")
|
||||
self.assertAlmostEqual(r["timing"].t_fixpipe,
|
||||
small.output_bytes / self.s.bw_l2)
|
||||
big = mkcase(1, 8192, 8192, 4096)
|
||||
self.assertGreater(big.input_bytes + big.output_bytes, self.s.l2_bytes)
|
||||
r = self.router.route(big)
|
||||
self.assertEqual(r["branch"], "转Matmul")
|
||||
self.assertAlmostEqual(r["timing"].t_fixpipe,
|
||||
big.output_bytes / self.s.bw_gm)
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
unittest.main()
|
||||
|
||||
Reference in New Issue
Block a user