OV - Compare Loops

Loops

▶quants.c: 910 - 8.47 %

ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)
Run orig_default						Run gcc_default						Run gcc_2
Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/arm/quants.c: 910-948 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/simd-mappings.h: 51-51					Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/arm/quants.c: 910-910 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/arm/quants.c: 928-929 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/simd-mappings.h: 51-51 /usr/lib/gcc/aarch64-amazon-linux/11/include/arm_neon.h: 13617-13617 /usr/lib/gcc/aarch64-amazon-linux/11/include/arm_neon.h: 16103-16103 /usr/lib/gcc/aarch64-amazon-linux/11/include/arm_neon.h: 20765-20765 /usr/lib/gcc/aarch64-amazon-linux/11/include/arm_neon.h: 30781-30781 /usr/lib/gcc/aarch64-amazon-linux/11/include/arm_neon.h: 31056-31056 /usr/lib/gcc/aarch64-amazon-linux/11/include/arm_neon.h: 34664-34664					Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/arm/quants.c: 910-910 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/arm/quants.c: 928-928 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/simd-mappings.h: 51-51 /usr/lib/gcc/aarch64-amazon-linux/11/include/arm_neon.h: 13617-13617 /usr/lib/gcc/aarch64-amazon-linux/11/include/arm_neon.h: 16065-16065 /usr/lib/gcc/aarch64-amazon-linux/11/include/arm_neon.h: 16103-16103 /usr/lib/gcc/aarch64-amazon-linux/11/include/arm_neon.h: 20765-20765 /usr/lib/gcc/aarch64-amazon-linux/11/include/arm_neon.h: 30781-30781 /usr/lib/gcc/aarch64-amazon-linux/11/include/arm_neon.h: 31056-31056 /usr/lib/gcc/aarch64-amazon-linux/11/include/arm_neon.h: 34664-34664
2293	0.10	0.08	0.83	10.53	24.67	1955	0.09	0.07	2.50	16.22	25.68	1929	0.10	4.62	5.14	11.11	26.69

Sum on 1 analyzed binary loop (libggml-cpu.so - 2293)						Sum on 1 analyzed binary loop (libggml-cpu.so - 1955)						Sum on 1 analyzed binary loop (libggml-cpu.so - 1929)
Analysis					Count	Analysis					Count	Analysis					Count
Loop Computation Issues						Loop Computation Issues						Loop Computation Issues
Presence of a large number of scalar integer instructions						Presence of a large number of scalar integer instructions					1	Presence of a large number of scalar integer instructions					1
Data Access Issues						Data Access Issues						Data Access Issues
Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access					1
Vectorization Roadblocks						Vectorization Roadblocks						Vectorization Roadblocks
Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access					1

▶vec.cpp: 385 - 0.81 %

ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)
Run orig_default						Run gcc_default						Run gcc_2
Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/vec.h: 1008-1034 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/vec.cpp: 385-387					Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/vec.h: 1009-1023 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/vec.h: 1031-1034 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/vec.cpp: 385-387					Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/vec.h: 1009-1023 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/vec.h: 1031-1034 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/vec.cpp: 385-387
908	0.02	0.01	0.05	68.18	79.76	766	0.02	0.01	0.33	80	97.59	779	0.02	0.38	0.42	90	98.62

Sum on 1 analyzed binary loop (libggml-cpu.so - 908)						Sum on 1 analyzed binary loop (libggml-cpu.so - 766)						Sum on 1 analyzed binary loop (libggml-cpu.so - 779)
Analysis					Count	Analysis					Count	Analysis					Count
Loop Computation Issues						Loop Computation Issues						Loop Computation Issues
Presence of expensive FP instructions					1	Presence of expensive FP instructions					1	Presence of expensive FP instructions					1
Data Access Issues						Data Access Issues						Data Access Issues
Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access
Vectorization Roadblocks						Vectorization Roadblocks						Vectorization Roadblocks
Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access

▶ops.cpp: 6210 - 0.70 %

ASM Loop ID	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)
Run orig_default		Run gcc_default						Run gcc_2
Loop Source Regions		Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 6210-6211 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 6220-6230 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 6238-6245 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 6413-6413					Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 6211-6211 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 6223-6231
		789	0.04	0.01	0.41	2.33	10.1	794	0.03	0.26	0.29	0	12.76

No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.		Sum on 1 analyzed binary loop (libggml-cpu.so - 789)						Sum on 1 analyzed binary loop (libggml-cpu.so - 794)
Analysis	Count	Analysis					Count	Analysis					Count
		Loop Computation Issues						Loop Computation Issues
		Presence of expensive FP instructions					1	Presence of expensive FP instructions					1
		Presence of a large number of scalar integer instructions					1	Presence of a large number of scalar integer instructions					1
		Control Flow Issues						Control Flow Issues
		Presence of calls					1	Presence of calls					1
		Presence of more than 4 paths					1	Presence of more than 4 paths					0
		Vectorization Roadblocks						Vectorization Roadblocks
		Presence of calls					1	Presence of calls					1
		Presence of more than 4 paths					1	Presence of more than 4 paths					1

▶ops.cpp: 6238 - 0.58 %

ASM Loop ID	Vector Length Use (%)	ASM Loop ID	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)
Run orig_default		Run gcc_default		Run gcc_2
Loop Source Regions		Loop Source Regions		Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 6238-6245
				818	0.03	0.52	0.58	0	14.58

No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.		No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.		Sum on 1 analyzed binary loop (libggml-cpu.so - 818)
Analysis	Count	Analysis	Count	Analysis					Count
				Loop Computation Issues
				Presence of expensive FP instructions					1
				Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA					1
				Control Flow Issues
				Presence of calls					1
				Data Access Issues
				Presence of constant non-unit stride data access					1
				Presence of indirect access					1
				Vectorization Roadblocks
				Presence of calls					1
				Presence of constant non-unit stride data access					1
				Presence of indirect access					1

▶binary-ops.cpp: 10 - 0.31 %

ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)
Run orig_default						Run gcc_default						Run gcc_2
Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/binary-ops.cpp: 10-10 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/binary-ops.cpp: 31-32					Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/binary-ops.cpp: 10-10 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/binary-ops.cpp: 31-32					Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/binary-ops.cpp: 10-10 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/binary-ops.cpp: 31-32
433	0.02	0.00	0.04	0	12.5	410	0.02	0.00	0.13	25	50	412	0.02	0.12	0.13	25	50

Sum on 1 analyzed binary loop (libggml-cpu.so - 433)						Sum on 1 analyzed binary loop (libggml-cpu.so - 410)						Sum on 1 analyzed binary loop (libggml-cpu.so - 412)
Analysis					Count	Analysis					Count	Analysis					Count
Loop Computation Issues						Loop Computation Issues						Loop Computation Issues
Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA					1	Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA					1	Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA					1
Data Access Issues						Data Access Issues						Data Access Issues
Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access					1
Vectorization Roadblocks						Vectorization Roadblocks						Vectorization Roadblocks
Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access					1

▶ops.cpp: 4325 - 0.27 %

ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)
Run orig_default						Run gcc_default						Run gcc_2
Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 4325-4326					Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 4325-4326					Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 4325-4326
1278	0.02	0.00	0.03	94.12	95.59	1132	0.02	0.00	0.10	0	13.28	1153	0.01	0.12	0.14	90.91	75.04

Sum on 1 analyzed binary loop (libggml-cpu.so - 1278)						Sum on 1 analyzed binary loop (libggml-cpu.so - 1132)						Sum on 1 analyzed binary loop (libggml-cpu.so - 1153)
Analysis					Count	Analysis					Count	Analysis					Count
Loop Computation Issues						Loop Computation Issues						Loop Computation Issues
Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA					1	Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA					1	Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA					1
Data Access Issues						Data Access Issues						Data Access Issues
Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access						Presence of constant non-unit stride data access
Vectorization Roadblocks						Vectorization Roadblocks						Vectorization Roadblocks
Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access						Presence of constant non-unit stride data access

▶binary-ops.cpp: 18 - 0.23 %

ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)
Run orig_default						Run gcc_default						Run gcc_2
Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/binary-ops.cpp: 18-18 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/binary-ops.cpp: 31-32					Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/binary-ops.cpp: 18-18 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/binary-ops.cpp: 31-32					Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/binary-ops.cpp: 18-18 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/binary-ops.cpp: 31-32
538	0.01	0.00	0.03	0	11.84	494	0.02	0.00	0.08	25	50	498	0.01	0.11	0.12	25	50

Sum on 1 analyzed binary loop (libggml-cpu.so - 538)						Sum on 1 analyzed binary loop (libggml-cpu.so - 494)						Sum on 1 analyzed binary loop (libggml-cpu.so - 498)
Analysis					Count	Analysis					Count	Analysis					Count
Loop Computation Issues						Loop Computation Issues						Loop Computation Issues
Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA					1	Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA					1	Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA					1
Data Access Issues						Data Access Issues						Data Access Issues
Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access					1
Vectorization Roadblocks						Vectorization Roadblocks						Vectorization Roadblocks
Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access					1

▶ggml-cpu.c: 3228 - 0.21 %

ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)
Run orig_default						Run gcc_default						Run gcc_2
Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/./ggml-impl.h: 389-404 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.c: 3228-3229					Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/./ggml-impl.h: 389-404 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.c: 3228-3229					Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/./ggml-impl.h: 389-404 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.c: 3228-3229
0	0.01	0.00	0.01	86.36	95.61	4	0.01	0.00	0.09	93.18	94.03	0	0.01	0.09	0.10	92.11	93.09

Sum on 1 analyzed binary loop (libggml-cpu.so - 0)						Sum on 1 analyzed binary loop (libggml-cpu.so - 4)						Sum on 1 analyzed binary loop (libggml-cpu.so - 0)
Analysis					Count	Analysis					Count	Analysis					Count
Loop Computation Issues						Loop Computation Issues						Loop Computation Issues
Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA					1	Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA					1	Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA
Data Access Issues						Data Access Issues						Data Access Issues
Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access
Vectorization Roadblocks						Vectorization Roadblocks						Vectorization Roadblocks
Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access

▶<unknown>: 0 - 0.13 %

ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)
Run orig_default						Run gcc_default						Run gcc_2
Loop Source Regions						Loop Source Regions						Loop Source Regions
2483	0.00	0.00	0.00	0	0	1648	0.00	0.00	0.00	0	0	4199	0.00	0.00	0.00	0	0
2473	0.00	0.00	0.00	0	0	3728	0.00	0.00	0.00	0	0	3796	0.00	0.00	0.00	0	0
2634	0.00	0.00	0.00	0	0	4140	0.00	0.00	0.00	0	0	4388	0.00	0.00	0.00	0	0
2650	0.01	0.00	0.00	0	0	3907	0.00	0.00	0.00	0	0	53	0.00	0.00	0.00	0	0
2649	0.01	0.00	0.00	0	0	3779	0.00	0.00	0.00	0	0	1905	0.00	0.00	0.00	0	0
2782	0.00	0.00	0.00	0	0	60	0.00	0.00	0.00	0	0	1491	0.01	0.00	0.00	0	0
2480	0.00	0.00	0.00	0	0	56	0.00	0.00	0.01	0	0	61	0.00	0.01	0.01	0	0
2767	0.02	0.00	0.00	0	0	1932	0.01	0.00	0.01	0	0
2886	0.00	0.00	0.00	0	0	1466	0.00	0.00	0.00	0	0
1282	0.00	0.00	0.00	0	0	368	0.00	0.00	0.01	0	0
2268	0.01	0.00	0.00	0	0	1131	0.01	0.00	0.01	0	0
901	0.01	0.00	0.01	0	0	19	0.00	0.00	0.00	0	0
1754	0.00	0.00	0.00	0	0
396	0.01	0.00	0.00	0	0
1437	0.00	0.00	0.01	0	0
66	0.01	0.00	0.00	0	0
314	0.02	0.00	0.00	0	0

No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.						No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.						No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.
Analysis					Count	Analysis					Count	Analysis					Count

▶ops.cpp: 6220 - 0.13 %

ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)	ASM Loop ID	Vector Length Use (%)	ASM Loop ID	Vector Length Use (%)
Run orig_default						Run gcc_default		Run gcc_2
Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 6220-6220 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 6229-6230 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 6238-6245					Loop Source Regions		Loop Source Regions
1432	0.03	0.01	0.13	0	13.17

Sum on 1 analyzed binary loop (libggml-cpu.so - 1432)						No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.		No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.
Analysis					Count	Analysis	Count	Analysis	Count
Loop Computation Issues
Presence of expensive FP instructions					1
Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA					1
Control Flow Issues
Presence of calls					1
Data Access Issues
Presence of constant non-unit stride data access					1
Vectorization Roadblocks
Presence of calls					1
Presence of constant non-unit stride data access					1

▶vec.h: 411 - 0.07 %

ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)	ASM Loop ID	Vector Length Use (%)
Run orig_default						Run gcc_default						Run gcc_2
Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/vec.h: 411-458					Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/vec.h: 411-458					Loop Source Regions
1753	0.01	0.00	0.01	100	100	1470	0.01	0.00	0.06	100	100

Sum on 1 analyzed binary loop (libggml-cpu.so - 1753)						Sum on 1 analyzed binary loop (libggml-cpu.so - 1470)						No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.
Analysis					Count	Analysis					Count	Analysis	Count
Data Access Issues						Data Access Issues
Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access					1
Vectorization Roadblocks						Vectorization Roadblocks
Presence of constant non-unit stride data access					1	Presence of constant non-unit stride data access					1

▶vec.cpp: 231 - 0.05 %

ASM Loop ID	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)	ASM Loop ID	Vector Length Use (%)
Run orig_default		Run gcc_default						Run gcc_2
Loop Source Regions		Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/vec.cpp: 231-262					Loop Source Regions
		763	0.01	0.00	0.05	100	100

No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.		No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.						No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.
Analysis	Count	Analysis					Count	Analysis	Count

▶hashtable.h: 2386 - 0.04 %

ASM Loop ID	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)
Run orig_default		Run gcc_default						Run gcc_2
Loop Source Regions		Loop Source Regions	/usr/include/c++/11/bits/hashtable_policy.h: 287-287 /usr/include/c++/11/bits/hashtable_policy.h: 434-434 /usr/include/c++/11/bits/hashtable.h: 2386-2391 /usr/include/c++/11/bits/hashtable.h: 2402-2403					Loop Source Regions	/usr/include/c++/11/bits/hashtable_policy.h: 287-287 /usr/include/c++/11/bits/hashtable_policy.h: 434-434 /usr/include/c++/11/bits/hashtable.h: 2386-2391 /usr/include/c++/11/bits/hashtable.h: 2402-2403
		4180	0.02	0.00	0.01	0	25	4246	0.02	0.02	0.02	0	25

No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.		No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.						No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.
Analysis	Count	Analysis					Count	Analysis					Count

▶ops.cpp: 6446 - 0.03 %

ASM Loop ID	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)
Run orig_default		Run gcc_default						Run gcc_2
Loop Source Regions		Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 6446-6447 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 6453-6456					Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 6446-6447 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 6453-6456
		799	0.01	0.00	0.01	37.5	40.63	828	0.00	0.02	0.02	42.55	63.3

No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.		No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.						No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.
Analysis	Count	Analysis					Count	Analysis					Count

▶vec.h: 646 - 0.03 %

ASM Loop ID	Vector Length Use (%)	ASM Loop ID	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)
Run orig_default		Run gcc_default		Run gcc_2
Loop Source Regions		Loop Source Regions		Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/vec.h: 646-653
				1156	0.01	0.01	0.01	100	100
				1489	0.00	0.01	0.01	100	100

No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.		No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.		No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.
Analysis	Count	Analysis	Count	Analysis					Count

▶llama-vocab.cpp: 3216 - 0.03 %

ASM Loop ID	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)
Run orig_default		Run gcc_default						Run gcc_2
Loop Source Regions		Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/src/llama-vocab.cpp: 3216-3216 /usr/include/c++/11/bits/hashtable_policy.h: 287-287 /usr/include/c++/11/bits/hashtable_policy.h: 1916-1916 /usr/include/c++/11/ext/new_allocator.h: 145-145 /usr/include/c++/11/bits/basic_string.h: 195-195 /usr/include/c++/11/bits/basic_string.h: 211-211 /usr/include/c++/11/bits/basic_string.h: 239-239 /usr/include/c++/11/bits/basic_string.h: 245-245					Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/src/llama-vocab.cpp: 3216-3216 /usr/include/c++/11/bits/hashtable_policy.h: 287-287 /usr/include/c++/11/bits/hashtable_policy.h: 1916-1916 /usr/include/c++/11/ext/new_allocator.h: 145-145 /usr/include/c++/11/bits/basic_string.h: 195-195 /usr/include/c++/11/bits/basic_string.h: 211-211 /usr/include/c++/11/bits/basic_string.h: 239-239 /usr/include/c++/11/bits/basic_string.h: 245-245
		3732	0.02	0.00	0.01	0	25	3800	0.01	0.01	0.01	0	25

No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.		No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.						No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.
Analysis	Count	Analysis					Count	Analysis					Count

▶ggml-impl.h: 354 - 0.02 %

ASM Loop ID	Vector Length Use (%)	ASM Loop ID	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)
Run orig_default		Run gcc_default		Run gcc_2
Loop Source Regions		Loop Source Regions		Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-impl.h: 354-361 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-impl.h: 369-377 /home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-quants.c: 408-412
				584	0.02	0.02	0.02	39.13	29.21

No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.		No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.		No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.
Analysis	Count	Analysis	Count	Analysis					Count

▶ops.cpp: 8825 - 0.02 %

ASM Loop ID	Vector Length Use (%)	ASM Loop ID	Max Time Over Threads (s)	Time w.r.t. Wall Time (s)	Cov (%)	Vect. Ratio (%)	Vector Length Use (%)	ASM Loop ID	Vector Length Use (%)
Run orig_default		Run gcc_default						Run gcc_2
Loop Source Regions		Loop Source Regions	/home/eoseret/Tools/QaaS/qaas_runs/ip-172-31-46-37.ec2.internal/176-409-8358/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 8825-8826					Loop Source Regions
		1463	0.01	0.00	0.02	0	0

No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.		No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.						No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.
Analysis	Count	Analysis					Count	Analysis	Count

Report Configuration

Loops

▶quants.c: 910 - 8.47 %

▶vec.cpp: 385 - 0.81 %

▶ops.cpp: 6210 - 0.70 %

▶ops.cpp: 6238 - 0.58 %

▶binary-ops.cpp: 10 - 0.31 %

▶ops.cpp: 4325 - 0.27 %

▶binary-ops.cpp: 18 - 0.23 %

▶ggml-cpu.c: 3228 - 0.21 %

▶<unknown>: 0 - 0.13 %

▶ops.cpp: 6220 - 0.13 %

▶vec.h: 411 - 0.07 %

▶vec.cpp: 231 - 0.05 %

▶hashtable.h: 2386 - 0.04 %

▶ops.cpp: 6446 - 0.03 %

▶vec.h: 646 - 0.03 %

▶llama-vocab.cpp: 3216 - 0.03 %

▶ggml-impl.h: 354 - 0.02 %

▶ops.cpp: 8825 - 0.02 %