OV - Compare Loops

Loops

▶repack.cpp: 153 - 145.62 %

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 (%)	ASM Loop ID	Vector Length Use (%)
Run orig_default						Run icx_default						Run aocc_3						Run icx_9
Loop Source Regions	/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 153-153 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 170-172 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 559-575 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 582-586 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 616-616					Loop Source Regions	/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 153-153 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 170-172 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 1129-1133 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 1139-1139 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 1184-1202 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 559-575 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 582-586 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 616-616					Loop Source Regions	/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 153-153 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 170-172 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 559-575 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 582-586 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 616-616					Loop Source Regions
2570	1.67	1.50	48.32	0	0	2980	0.02	0.00	0.10	0	0	2642	1.64	1.52	49.04	0	0
						2960	1.63	1.54	48.15	0	0

Sum on 1 analyzed binary loop (libggml-cpu.so - 2570)						Sum on 2 analyzed binary loops (libggml-cpu.so - 2980, libggml-cpu.so - 2960)						Sum on 1 analyzed binary loop (libggml-cpu.so - 2642)						No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.
Analysis					Count	Analysis					Count	Analysis					Count	Analysis	Count

▶quants.c: 682 - 58.23 %

ASM Loop ID	Vector Length Use (%)	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 icx_default		Run aocc_3		Run icx_9
Loop Source Regions		Loop Source Regions		Loop Source Regions		Loop Source Regions	/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/quants.c: 682-682 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/quants.c: 690-695
						3839	4.13	3.16	58.23	93.02	21.37

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.		Sum on 1 analyzed binary loop (libggml-cpu.so - 3839)
Analysis	Count	Analysis	Count	Analysis	Count	Analysis					Count
						Loop Computation Issues
						Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA					1
						Data Access Issues
						Presence of indirect access					1
						Presence of special instructions executing on a single port					1
						Vectorization Roadblocks
						Presence of indirect access					1
						Inefficient Vectorization
						Presence of special instructions executing on a single port					1

▶quants.c: 2150 - 40.59 %

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 (%)	ASM Loop ID	Vector Length Use (%)
Run orig_default						Run icx_default						Run aocc_3						Run icx_9
Loop Source Regions	/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/quants.c: 2150-2152 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/quants.c: 2158-2158 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/quants.c: 2166-2216					Loop Source Regions	/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/quants.c: 2150-2152 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/quants.c: 2158-2158 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/quants.c: 2166-2216					Loop Source Regions	/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/simd-mappings.h: 130-130 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/quants.c: 2150-2152 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/quants.c: 2158-2158 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/quants.c: 2166-2216					Loop Source Regions
2545	0.47	0.41	13.30	97.09	43.87	2936	0.43	0.44	13.93	96.81	44.08	2617	0.46	0.41	13.36	97.09	43.87

Sum on 1 analyzed binary loop (libggml-cpu.so - 2545)						Sum on 1 analyzed binary loop (libggml-cpu.so - 2936)						Sum on 1 analyzed binary loop (libggml-cpu.so - 2617)						No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.
Analysis					Count	Analysis					Count	Analysis					Count	Analysis	Count
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
Presence of indirect access					1	Presence of indirect access					1	Presence of indirect access					1
Presence of special instructions executing on a single port					1	Presence of special instructions executing on a single port					1	Presence of special instructions executing on a single port					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
Presence of indirect access					1	Presence of indirect access					1	Presence of indirect access					1
Inefficient Vectorization						Inefficient Vectorization						Inefficient Vectorization
Presence of special instructions executing on a single port					1	Presence of special instructions executing on a single port					1	Presence of special instructions executing on a single port					1

▶quants.c: 716 - 8.01 %

ASM Loop ID	Vector Length Use (%)	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 icx_default		Run aocc_3		Run icx_9
Loop Source Regions		Loop Source Regions		Loop Source Regions		Loop Source Regions	/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/simd-mappings.h: 130-130 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/quants.c: 716-716 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/quants.c: 725-725 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/quants.c: 733-736 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/quants.c: 745-754
						499	0.45	0.43	8.01	46.92	13.2

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.		Sum on 1 analyzed binary loop (libggml-cpu.so - 499)
Analysis	Count	Analysis	Count	Analysis	Count	Analysis					Count
						Loop Computation Issues
						Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA					1
						Presence of a large number of scalar integer instructions					1
						Data Access Issues
						Presence of constant non-unit stride data access					1
						Presence of indirect access					1
						Presence of special instructions executing on a single port					1
						More than 20% of the loads are accessing the stack					1
						Vectorization Roadblocks
						Presence of constant non-unit stride data access					1
						Presence of indirect access					1
						Inefficient Vectorization
						Presence of special instructions executing on a single port					1

▶repack.cpp: 125 - 2.02 %

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 (%)	ASM Loop ID	Vector Length Use (%)
Run orig_default						Run icx_default						Run aocc_3						Run icx_9
Loop Source Regions	/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 125-125 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 137-140 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 755-779 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 822-842					Loop Source Regions	/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 125-125 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 137-140 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 755-760 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 766-766 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 772-779 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 822-842					Loop Source Regions	/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 125-125 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 137-140 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 755-779 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp: 822-842					Loop Source Regions
2594	0.04	0.02	0.71	99.51	83.19	2990	0.04	0.02	0.64	99.48	85.43	2666	0.04	0.02	0.67	99.51	83.19

Sum on 1 analyzed binary loop (libggml-cpu.so - 2594)						Sum on 1 analyzed binary loop (libggml-cpu.so - 2990)						Sum on 1 analyzed binary loop (libggml-cpu.so - 2666)						No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.
Analysis					Count	Analysis					Count	Analysis					Count	Analysis	Count
Data Access Issues						Data Access Issues						Data Access Issues
Presence of special instructions executing on a single port					1	Presence of special instructions executing on a single port					1	Presence of special instructions executing on a single port					1
More than 20% of the loads are accessing the stack					1	More than 20% of the loads are accessing the stack					1	More than 20% of the loads are accessing the stack					1
Inefficient Vectorization						Inefficient Vectorization						Inefficient Vectorization
Presence of special instructions executing on a single port					1	Presence of special instructions executing on a single port					1	Presence of special instructions executing on a single port					1
Use of masked instructions					1	Use of masked instructions					1	Use of masked instructions					1

▶<unknown>: 0 - 1.45 %

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 (%)	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 icx_default						Run aocc_3						Run icx_9
Loop Source Regions						Loop Source Regions						Loop Source Regions						Loop Source Regions
1681	0.00	0.00	0.00	0	0	2189	0.00	0.00	0.00	0	0	2659	0.01	0.00	0.01	0	0	1714	0.01	0.00	0.00	0	0
2842	0.00	0.00	0.00	0	0	2392	0.00	0.00	0.00	0	0	2660	0.00	0.00	0.00	0	0	4118	0.01	0.00	0.00	0	0
2640	0.00	0.00	0.00	0	0	2393	0.00	0.00	0.00	0	0	2759	0.02	0.00	0.01	0	0	4130	0.01	0.00	0.00	0	0
2458	0.00	0.00	0.00	0	0	2395	0.01	0.00	0.00	0	0	2485	0.00	0.00	0.00	0	0	4341	0.01	0.00	0.00	0	0
2628	0.01	0.00	0.00	0	0	2155	0.00	0.00	0.00	0	0	2861	0.00	0.00	0.00	0	0	648	0.01	0.00	0.03	0	0
2638	0.01	0.00	0.00	0	0	2291	0.00	0.00	0.00	0	0	99	0.02	0.00	0.08	0	0	1747	0.04	0.00	0.02	0	0
2741	0.01	0.00	0.01	0	0	2220	0.00	0.00	0.00	0	0	819	0.01	0.00	0.01	0	0	114	0.00	0.00	0.00	0	0
99	0.01	0.00	0.09	0	0	1948	0.01	0.00	0.01	0	0	83	0.00	0.00	0.00	0	0	135	0.01	0.00	0.03	0	0
2587	0.02	0.00	0.09	0	0	182	0.00	0.00	0.00	0	0	1530	0.01	0.00	0.01	0	0	530	0.01	0.00	0.01	0	0
359	0.01	0.00	0.00	0	0	444	0.06	0.00	0.02	0	0	1782	0.01	0.00	0.02	0	0	108	0.02	0.00	0.07	0	0
2528	0.01	0.00	0.01	0	0	278	0.01	0.00	0.05	0	0	82	0.01	0.00	0.01	0	0	2986	0.02	0.00	0.04	0	0
399	0.06	0.00	0.02	0	0	114	0.00	0.00	0.00	0	0	397	0.06	0.00	0.03	0	0	2574	0.03	0.00	0.01	0	0
5	0.01	0.00	0.01	0	0	106	0.00	0.00	0.01	0	0	5	0.01	0.00	0.01	0	0	2572	0.01	0.00	0.00	0	0
1457	0.00	0.00	0.01	0	0	124	0.02	0.00	0.10	0	0	1803	0.00	0.00	0.00	0	0	112	0.00	0.00	0.00	0	0
2563	0.00	0.00	0.00	0	0	108	0.00	0.00	0.00	0	0	1773	0.01	0.00	0.00	0	0	2975	0.01	0.00	0.02	0	0
1731	0.00	0.00	0.00	0	0	873	0.02	0.00	0.02	0	0	483	0.01	0.00	0.01	0	0	3433	0.01	0.00	0.00	0	0
1196	0.01	0.00	0.01	0	0	2099	0.01	0.00	0.02	0	0	2600	0.02	0.00	0.01	0	0	119	0.01	0.00	0.02	0	0
485	0.00	0.00	0.00	0	0	6	0.01	0.00	0.01	0	0	2659	0.01	0.00	0.09	0	0	117	0.00	0.00	0.00	0	0
766	0.01	0.00	0.01	0	0	2111	0.00	0.00	0.00	0	0	357	0.01	0.00	0.01	0	0	649	0.03	0.00	0.01	0	0
82	0.01	0.00	0.02	0	0	1418	0.01	0.00	0.01	0	0	273	0.01	0.00	0.01	0	0	3	0.01	0.00	0.00	0	0
1710	0.01	0.00	0.05	0	0	1707	0.00	0.00	0.01	0	0	2635	0.01	0.00	0.01	0	0	3429	0.00	0.00	0.00	0	0
233	0.17	0.00	0.06	0	0	2954	0.01	0.00	0.01	0	0	232	0.21	0.00	0.08	0	0	3446	0.01	0.00	0.00	0	0
274	0.01	0.00	0.01	0	0	102	0.00	0.00	0.01	0	0							945	0.02	0.00	0.00	0	0
831	0.00	0.00	0.00	0	0	401	0.01	0.00	0.02	0	0
						2918	0.02	0.00	0.01	0	0
						2094	0.01	0.00	0.01	0	0
						558	0.01	0.00	0.01	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.						No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.
Analysis					Count	Analysis					Count	Analysis					Count	Analysis					Count

▶vec.h: 508 - 0.51 %

ASM Loop ID	Vector Length Use (%)	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 icx_default		Run aocc_3		Run icx_9
Loop Source Regions		Loop Source Regions		Loop Source Regions		Loop Source Regions	/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/simd-mappings.h: 130-130 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/vec.h: 508-509 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/./ggml-impl.h: 346-346 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/./ggml-impl.h: 389-389 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/./ggml-impl.h: 399-404
						3443	0.12	0.03	0.51	85.05	20.79

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.		Sum on 1 analyzed binary loop (libggml-cpu.so - 3443)
Analysis	Count	Analysis	Count	Analysis	Count	Analysis					Count
						Loop Computation Issues
						Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA					1
						Data Access Issues
						Presence of indirect access					1
						Presence of special instructions executing on a single port					1
						Vectorization Roadblocks
						Presence of indirect access					1
						Inefficient Vectorization
						Presence of special instructions executing on a single port					1

▶vec.h: 491 - 0.47 %

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 (%)	ASM Loop ID	Vector Length Use (%)
Run orig_default						Run icx_default						Run aocc_3						Run icx_9
Loop Source Regions	/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/vec.h: 491-497					Loop Source Regions	/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/vec.h: 491-497					Loop Source Regions	/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/vec.h: 491-497					Loop Source Regions
1724	0.03	0.01	0.17	100	75	2109	0.03	0.00	0.13	100	75	1796	0.04	0.01	0.17	100	75

Sum on 1 analyzed binary loop (libggml-cpu.so - 1724)						Sum on 1 analyzed binary loop (libggml-cpu.so - 2109)						Sum on 1 analyzed binary loop (libggml-cpu.so - 1796)						No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.
Analysis					Count	Analysis					Count	Analysis					Count	Analysis	Count

▶vec.cpp: 311 - 0.36 %

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 (%)	ASM Loop ID	Vector Length Use (%)
Run orig_default						Run icx_default						Run aocc_3						Run icx_9
Loop Source Regions	/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/vec.cpp: 311-316					Loop Source Regions	/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/vec.cpp: 311-316					Loop Source Regions	/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/vec.cpp: 311-316					Loop Source Regions
762	0.03	0.00	0.12	100	66.67	866	0.03	0.00	0.11	100	66.67	811	0.03	0.00	0.13	100	66.67

Sum on 1 analyzed binary loop (libggml-cpu.so - 762)						Sum on 1 analyzed binary loop (libggml-cpu.so - 866)						Sum on 1 analyzed binary loop (libggml-cpu.so - 811)						No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.
Analysis					Count	Analysis					Count	Analysis					Count	Analysis	Count
Loop Computation Issues						Loop Computation Issues						Loop Computation Issues
Low iteration count						Low iteration count					1	Low iteration count
Control Flow Issues						Control Flow Issues						Control Flow Issues
Low iteration count						Low iteration count					1	Low iteration count

▶ops.cpp: 6220 - 0.35 %

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 (%)	ASM Loop ID	Vector Length Use (%)
Run orig_default						Run icx_default						Run aocc_3						Run icx_9
Loop Source Regions	/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 6220-6220 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 6229-6230 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 6238-6245					Loop Source Regions	/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 6220-6220 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 6229-6230 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 6238-6245					Loop Source Regions	/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 6220-6220 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 6229-6229 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/ops.cpp: 6238-6245					Loop Source Regions
1463	0.02	0.00	0.12	6.67	7.5	1734	0.02	0.00	0.11	15.56	9.17	1536	0.03	0.00	0.12	0	6.25

Sum on 1 analyzed binary loop (libggml-cpu.so - 1463)						Sum on 1 analyzed binary loop (libggml-cpu.so - 1734)						Sum on 1 analyzed binary loop (libggml-cpu.so - 1536)						No Optimizer analysis found for any assembly loop. More loops can be analyzed using option --optimizer-loop-count.
Analysis					Count	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
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
Presence of a large number of scalar integer instructions					1	Presence of a large number of scalar integer instructions					0	Presence of a large number of scalar integer instructions					1
Control Flow Issues						Control Flow Issues						Control Flow Issues
Presence of calls					1	Presence of calls					1	Presence of calls					1
Data Access Issues						Data Access Issues						Data Access Issues
More than 20% of the loads are accessing the stack					1	More than 20% of the loads are accessing the stack					1	More than 20% of the loads are accessing the stack					1
Vectorization Roadblocks						Vectorization Roadblocks						Vectorization Roadblocks
Presence of calls					1	Presence of calls					1	Presence of calls					1

▶vec.cpp: 331 - 0.27 %

ASM Loop ID	Vector Length Use (%)	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 icx_default		Run aocc_3		Run icx_9
Loop Source Regions		Loop Source Regions		Loop Source Regions		Loop Source Regions	/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/vec.cpp: 331-332 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/simd-mappings.h: 130-130
						1733	0.07	0.01	0.27	5.88	8.82

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.		Sum on 1 analyzed binary loop (libggml-cpu.so - 1733)
Analysis	Count	Analysis	Count	Analysis	Count	Analysis					Count
						Loop Computation Issues
						Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA					1
						Data Access Issues
						Presence of indirect access					1
						Vectorization Roadblocks
						Presence of indirect access					1

▶repack.cpp: 1074 - 0.11 %

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 (%)	ASM Loop ID	Vector Length Use (%)
Run orig_default		Run icx_default						Run aocc_3		Run icx_9
Loop Source Regions		Loop Source Regions	/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/repack.cpp: 1074-1074 /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz17.benchmarkcenter.megware.com/176-989-1525/llama.cpp/build/llama.cpp/ggml/src/ggml-cpu/repack.cpp: 1311-1315					Loop Source Regions		Loop Source Regions
		257	0.28	0.00	0.11	43.33	32.81

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 - 257)						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	Analysis	Count
		Loop Computation Issues
		Presence of a large number of scalar integer instructions					1
		Data Access Issues
		Presence of constant non-unit stride data access					1
		Presence of indirect access					1
		Presence of expensive instructions: scatter/gather					1
		Presence of special instructions executing on a single port					1
		More than 20% of the loads are accessing the stack					1
		Vectorization Roadblocks
		Presence of constant non-unit stride data access					1
		Presence of indirect access					1
		Inefficient Vectorization
		Presence of expensive instructions: scatter/gather					1
		Presence of special instructions executing on a single port					1

Report Configuration

Loops

▶repack.cpp: 153 - 145.62 %

▶quants.c: 682 - 58.23 %

▶quants.c: 2150 - 40.59 %

▶quants.c: 716 - 8.01 %

▶repack.cpp: 125 - 2.02 %

▶<unknown>: 0 - 1.45 %

▶vec.h: 508 - 0.51 %

▶vec.h: 491 - 0.47 %

▶vec.cpp: 311 - 0.36 %

▶ops.cpp: 6220 - 0.35 %

▶vec.cpp: 331 - 0.27 %

▶repack.cpp: 1074 - 0.11 %