matmul-analysis/BMM/BMM_Theory/t_out.csv

1	case_id	batch_a	batch_b	m	n	k	dtype_a	dtype_b	dtype_c	trans_a	trans_b	has_bias	out_nd	deterministic_level	plan_case_id	plan_branch	plan_npu	plan_op	plan_used_core_num	plan_split_b	plan_m_cnt	plan_n_cnt	plan_grid_k	plan_core_map	plan_b_core	plan_merge_b0	plan_single_core_m	plan_single_core_n	plan_single_core_k	plan_k_l1	plan_b_l1	plan_l1_form	plan_base_m	plan_base_n	plan_base_k	plan_l2_policy_in	plan_l2_policy_out	plan_swizzle_w	plan_workspace_bytes	plan_tail_strategy	plan_tail_m_cnt	plan_tail_n_cnt	plan_tail_k_cnt	plan_tail_m_main	plan_tail_n_main	plan_tail_block_cnt	plan_tail_wave_num	plan_fixpipe_unitflag	plan_out_dtype_bytes	plan_note	gm_read_bytes	l2_read_bytes	t_mte2_gm	t_mte2_l2	t_mte2	dma_cmd_count	t_dma_cmd	cube_flops	t_mmad	fixpipe_bytes	t_fixpipe	t_reduce	t_steady	t_drain	t_total	bottleneck	feasible	violations	bound_type	advice
2	x	51	51	255	42	682	fp32	fp32	fp32	False	False	False	True	0	x	StreamK	Ascend950PR	batch_mat_mul_v3	32	1	1	1	3	B/M/N切出51块, 每块3核切K归约 (归约组内核c负责K段[c*K/3,(c+1)*K/3))	1	1	255	42	228	128	1	K段标准分块流水	128	42	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	6554520	grid_K=3路切K+归约	1	1	3	0	0	0	0	True	4	P=8.33, grid_K=3, 部分和驻留L2按4B写出, AIV归约后按C dtype=4B写最终	465578.6666666667	0.0	9.311573333333333e-06	0.0	9.311573333333333e-06	0.0	0.0	29797034.666666668	1.9619446694101507e-06	262144	1.6131938461538462e-06	3.674316083916084e-07	9.311573333333333e-06	3.674316083916084e-07	9.679004941724942e-06	MTE2_GM	True		访存Bound(GM)	切B分支条件不满足, 落 StreamK
3	y	2	2	64	16	32	fp16	fp16	fp16	False	False	False	True	0	y	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	64	16	32	32	1	标准核内流水	64	16	32	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=0.03<C, 降核是理性选择 (强切则 tile 跌破搬移效率下限反而更慢)	10240.0	0.0	2.048e-07	0.0	2.048e-07	0.0	0.0	131072.0	8.630255144032922e-09	4096	8.192e-08	0.0	2.048e-07	0.0	2.048e-07	MTE2_GM	True		访存Bound(GM)	ASW_Basic兜底 (StreamK未过: 2_单核K段下限: K/grid_K >= 256B/dtype 且 grid_K>=2; 3_归约代价可接受: K > grid_K^2/(grid_K-1)*theta_c) [自检违规: dValue=64B < 下限 128B, K 段连续维搬移效率崩塌] —— 方案生成存在缺陷, 需人工复核
4	z	256	256	128	1024	8	bf16	bf16	bf16	False	False	False	True	0	z	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)	8	1	128	1024	8	8	2	b_双batch乒乓	128	256	16	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=36KB <= L1	147456	0.0	2.94912e-06	0.0	3.34912e-06	8	4.0000000000000003e-07	16777216.0	1.104672658436214e-06	2097152	4.194304e-05	0.0	4.194304e-05	5.380964082304527e-06	4.7324004082304525e-05	FIXPIPE	True		写出Bound	仅 IterBatch 条件满足