OV - - Loops

MAQAO

options

Loops Index

36 loops have been discarded from the report because their ratio ((Max Inclusive Time Over Threads * 100) / Max Thread Active Time) is lower than the threshold set by object_coverage_threshold (0.1%). It represents about 0.04% of the application. To include them, change the value of object_coverage_threshold in the experiment directory configuration file, then rerun the command with the additionnal parameter --force-static-analysis

▶Filters

Loop id	Source Location	Source Function	Level	Max Thread Time / Walltime aocc_0 (%)	Exclusive Coverage aocc_0 (%)	Inclusive Coverage aocc_0 (%)	Max Exclusive Time Over Threads aocc_0 (s)	Max Inclusive Time Over Threads aocc_0 (s)	Exclusive Time w.r.t. Wall Time aocc_0 (s)	Inclusive Time w.r.t. Wall Time aocc_0 (s)	Nb Threads aocc_0	GFLOPS aocc_0	Vectorization Ratio (%)	Vector Length Use (%)	Speedup If No Scalar Integer	Speedup If FP Vectorized	Speedup If Fully Vectorized	Speedup If Perfect Load Balancing aocc_0	Stride 0	Stride 1	Stride n	Stride Unknown	Stride Indirect	Array Access Efficiency
2537	libggml-cpu.so - quants.c:108-1042 [...]	ggml_vec_dot_q8_0_q8_0	Single	82.80	61.76	61.76	57.34	57.34	37.65	37.65	192	53.84	NA	NA	NA	NA	NA	1.53	NA	NA	NA	NA	NA	0.00
2373	libggml-cpu.so - sgemm.cpp:138-1044 [...]	void (anonymous namespace)::tinyBLAS_Q0_AVX<block_q8_0, block_q8_0, float>::gemm4xN<2>(long, long, long, long)	Innermost	0.22	0.13	0.13	0.15	0.15	0.08	0.08	189	91.50	NA	NA	NA	NA	NA	1.94	NA	NA	NA	NA	NA	0.00
2364	libggml-cpu.so - sgemm.cpp:138-1044 [...]	void (anonymous namespace)::tinyBLAS_Q0_AVX<block_q8_0, block_q8_0, float>::gemm4xN<4>(long, long, long, long)	Innermost	0.20	0.12	0.12	0.14	0.14	0.07	0.07	189	200.87	NA	NA	NA	NA	NA	1.89	NA	NA	NA	NA	NA	0.00
84	libggml-cpu.so - ggml-cpu.c:1164-1198 [...]	ggml_compute_forward_mul_mat	InBetween	0.19	0.08	0.13	0.13	0.20	0.05	0.08	192	70.88	0	12.34	1	1	8	2.55	2	0	0	0	0	100.00
82	libggml-cpu.so - ggml-cpu.c:1193-1194	ggml_compute_forward_mul_mat	Innermost	0.13	0.05	0.05	0.09	0.09	0.03	0.03	190	53.36	0	12.5	1	1	8	3.21	1	0	0	0	0	100.00
99	libggml-cpu.so - ggml-cpu.c:533-2891 [...]	ggml_graph_compute_thread	Single	0.12	0.05	0.05	0.08	0.08	0.03	0.03	184	25.47	0	9.58	1	1	14.77	2.77	NA	NA	NA	NA	NA	0.00
1723	libggml-cpu.so - vec.h:491-497	ggml_compute_forward_flash_attn_ext	Innermost	0.32	0.04	0.04	0.22	0.22	0.02	0.02	33	1781.82	100	75	1	1	1.45	1.73	0	2	0	0	0	100.00
761	libggml-cpu.so - vec.cpp:311-316	ggml_vec_dot_f16	Single	0.25	0.03	0.03	0.17	0.17	0.02	0.02	33	1005.58	100	66.67	1	1	1.57	1.72	0	2	0	0	0	100.00
2527	libggml-cpu.so - quants.c:298-355 [...]	quantize_row_q8_0	Single	0.12	0.03	0.03	0.08	0.08	0.02	0.02	139	0.46	58.66	28.88	1	1.14	2.72	3.61	0	2	0	0	0	100.00
1462	libggml-cpu.so - ops.cpp:6220-6245 [...]	ggml_compute_forward_rope_f32(ggml_compute_params const, ggml_tensor, bool)	Innermost	0.10	0.02	0.02	0.07	0.07	0.01	0.01	192	369.05	6.67	7.5	1.25	1.23	13.33	5.01	1	2	0	0	0	100.00
1710	libggml-cpu.so - ops.cpp:8759-8881 [...]	ggml_compute_forward_flash_attn_ext	InBetween	0.10	0.01	0.04	0.07	0.27	0.00	0.03	32	3130.72	23.21	14.54	2.36	1.47	2.1	2.55	NA	NA	NA	NA	NA	0.00
765	libggml-cpu.so - vec.h:1084-1116 [...]	ggml_vec_swiglu_f32	Single	0.29	0.00	0.00	0.20	0.20	0.00	0.00	7	4609.13	98	98.13	1	1	1	6.67	0.5	0	0	3	0	56.25
398	libggml-cpu.so - binary-ops.cpp:10-32 [...]	ggml_compute_forward_add_non_quantized	Innermost	0.17	0.00	0.00	0.12	0.12	0.00	0.00	1	226.14	0	6.25	1	1.33	16	1	0	3	0	0	0	100.00
484	libggml-cpu.so - binary-ops.cpp:18-32 [...]	ggml_compute_forward_mul	Innermost	0.14	0.00	0.00	0.10	0.10	0.00	0.00	1	190.44	0	6.25	1	1.33	16	1	0	3	0	0	0	100.00
1195	libggml-cpu.so - ops.cpp:4325-4326	ggml_compute_forward_rms_norm	Innermost	0.13	0.00	0.00	0.09	0.09	0.00	0.00	1	528.99	0	7.81	1	2.95	14	1	0	1	0	0	0	100.00

×