Help is available by moving the cursor above any 
symbol or by checking MAQAO website.
[ 4 / 4 ] Application profile is long enough (495.58 s)
To have good quality measurements, it is advised that the application profiling time is greater than 10 seconds.
[ 3 / 3 ] Optimization level option is correctly used
[ 2.40 / 3 ] Most of time spent in analyzed modules comes from functions without compilation information
Functions without compilation information (typically not compiled with -g) cumulate 0.01% of the time spent in analyzed modules. Check that -g is present. Remark: if -g is indeed used, this can also be due to some compiler built-in functions (typically math) or statically linked libraries. This warning can be ignored in that case.
[ 2 / 3 ] Security settings from the host restrict profiling. Some metrics will be missing or incomplete.
Current value for kernel.perf_event_paranoid is 2. If possible, set it to 1 or check with your system administrator which flag can be used to achieve this.
[ 0 / 3 ] Compilation of some functions is not optimized for the target processor
Application run on the ARM_NEOVERSE_V2 micro-architecture while the code was specialized for armv8.5-a+crypto+rng+sve2-aes+sve2-sha3+sve2-bitperm+i8mm+bf16+nossbs+nopredres. Architecture specific options are needed to produce efficient code for a specific processor ( -mcpu=native ).
[ 2 / 2 ] Application is correctly profiled ("Others" category represents 0.00 % of the execution time)
To have a representative profiling, it is advised that the category "Others" represents less than 20% of the execution time in order to analyze as much as possible of the user code
[ 1 / 1 ] Lstopo present. The Topology lstopo report will be generated.
[ 0 / 0 ] Fastmath not used
Consider to add ffast-math to compilation flags (or replace -O3 with -Ofast) to unlock potential extra speedup by relaxing floating-point computation consistency. Warning: floating-point accuracy may be reduced and the compliance to IEEE/ISO rules/specifications for math functions will be relaxed, typically 'errno' will no longer be set after calling some math functions.
[ 4 / 4 ] Enough time of the experiment time spent in analyzed loops (98.93%)
If the time spent in analyzed loops is less than 30%, standard loop optimizations will have a limited impact on application performances.
[ 4 / 4 ] Threads activity is good
On average, more than 99.99% of observed threads are actually active
[ 4 / 4 ] CPU activity is good
CPU cores are active 99.99% of time
[ 4 / 4 ] Loop profile is not flat
At least one loop coverage is greater than 4% (44.63%), representing an hotspot for the application
[ 4 / 4 ] Enough time of the experiment time spent in analyzed innermost loops (53.02%)
If the time spent in analyzed innermost loops is less than 15%, standard innermost loop optimizations such as vectorisation will have a limited impact on application performances.
[ 4 / 4 ] Affinity is good (100.00%)
Threads are not migrating to CPU cores: probably successfully pinned
[ 3 / 3 ] Less than 10% (0.00%) is spend in BLAS1 operations
It could be more efficient to inline by hand BLAS1 operations
[ 3 / 3 ] Functions mostly use all threads
Functions running on a reduced number of threads (typically sequential code) cover less than 10% of application walltime (0.05%)
[ 3 / 3 ] Cumulative Outermost/In between loops coverage (45.91%) lower than cumulative innermost loop coverage (53.02%)
Having cumulative Outermost/In between loops coverage greater than cumulative innermost loop coverage will make loop optimization more complex
[ 2 / 2 ] Less than 10% (0.00%) is spend in BLAS2 operations
BLAS2 calls usually could make a poor cache usage and could benefit from inlining.
[ 2 / 2 ] Less than 10% (0.70%) is spend in Libm/SVML (special functions)
| Loop ID | Analysis | Penalty Score |
|---|---|---|
| ►Loop 24 - md-gcc-O2 | Execution Time: 44 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 22 | |
| ○ | [SA] Presence of expensive FP instructions - Perform hoisting, change algorithm, use SVML or proper numerical library or perform value profiling (count the number of distinct input values). There are 5 issues (= instructions) costing 4 points each. | 20 |
| ○ | [SA] Presence of a large number of scalar integer instructions - Simplify loop structure, perform loop splitting or perform unroll and jam. This issue costs 2 points. | 2 |
| ►Control Flow Issues | 22 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ○ | [SA] Too many paths (15 paths) - Simplify control structure. There are 15 issues ( = paths) costing 1 point each with a malus of 4 points. | 19 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 22 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ○ | [SA] Too many paths (15 paths) - Simplify control structure. There are 15 issues ( = paths) costing 1 point each with a malus of 4 points. | 19 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Loop 18 - md-gcc-O2 | Execution Time: 19 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 21 - md-gcc-O2 | Execution Time: 8 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 22 - md-gcc-O2 | Execution Time: 6 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 23 - md-gcc-O2 | Execution Time: 6 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 20 - md-gcc-O2 | Execution Time: 5 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 19 - md-gcc-O2 | Execution Time: 5 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 17 - md-gcc-O2 | Execution Time: 1 % - Vectorization Ratio: 0.00 % - Vector Length Use: 45.83 % | |
| ►Control Flow Issues | 22 | |
| ○ | [SA] Too many paths (16 paths) - Simplify control structure. There are 16 issues ( = paths) costing 1 point each with a malus of 4 points. | 20 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 22 | |
| ○ | [SA] Too many paths (16 paths) - Simplify control structure. There are 16 issues ( = paths) costing 1 point each with a malus of 4 points. | 20 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Loop 38 - md-gcc-O2 | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Control Flow Issues | 1 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ►Data Access Issues | 6 | |
| ○ | [SA] Presence of constant non unit stride data access - Use array restructuring, perform loop interchange or use gather instructions to lower a bit the cost. There are 3 issues ( = data accesses) costing 2 point each. | 6 |
| ►Vectorization Roadblocks | 7 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ○ | [SA] Presence of constant non unit stride data access - Use array restructuring, perform loop interchange or use gather instructions to lower a bit the cost. There are 3 issues ( = data accesses) costing 2 point each. | 6 |
| ►Loop 16 - md-gcc-O2 | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 37.10 % | |
| ►Control Flow Issues | 25 | |
| ○ | [SA] Too many paths (19 paths) - Simplify control structure. There are 19 issues ( = paths) costing 1 point each with a malus of 4 points. | 23 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 25 | |
| ○ | [SA] Too many paths (19 paths) - Simplify control structure. There are 19 issues ( = paths) costing 1 point each with a malus of 4 points. | 23 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
[ 4 / 4 ] Application profile is long enough (253.10 s)
To have good quality measurements, it is advised that the application profiling time is greater than 10 seconds.
[ 3 / 3 ] Optimization level option is correctly used
[ 2.40 / 3 ] Most of time spent in analyzed modules comes from functions without compilation information
Functions without compilation information (typically not compiled with -g) cumulate 0.00% of the time spent in analyzed modules. Check that -g is present. Remark: if -g is indeed used, this can also be due to some compiler built-in functions (typically math) or statically linked libraries. This warning can be ignored in that case.
[ 2 / 3 ] Security settings from the host restrict profiling. Some metrics will be missing or incomplete.
Current value for kernel.perf_event_paranoid is 2. If possible, set it to 1 or check with your system administrator which flag can be used to achieve this.
[ 0 / 3 ] Compilation of some functions is not optimized for the target processor
Application run on the ARM_NEOVERSE_V2 micro-architecture while the code was specialized for armv8.5-a+crypto+rng+sve2-aes+sve2-sha3+sve2-bitperm+i8mm+bf16+nossbs+nopredres. Architecture specific options are needed to produce efficient code for a specific processor ( -mcpu=native ).
[ 2 / 2 ] Application is correctly profiled ("Others" category represents 0.00 % of the execution time)
To have a representative profiling, it is advised that the category "Others" represents less than 20% of the execution time in order to analyze as much as possible of the user code
[ 1 / 1 ] Lstopo present. The Topology lstopo report will be generated.
[ 0 / 0 ] Fastmath not used
Consider to add ffast-math to compilation flags (or replace -O3 with -Ofast) to unlock potential extra speedup by relaxing floating-point computation consistency. Warning: floating-point accuracy may be reduced and the compliance to IEEE/ISO rules/specifications for math functions will be relaxed, typically 'errno' will no longer be set after calling some math functions.
[ 4 / 4 ] Enough time of the experiment time spent in analyzed loops (98.20%)
If the time spent in analyzed loops is less than 30%, standard loop optimizations will have a limited impact on application performances.
[ 4 / 4 ] Threads activity is good
On average, more than 199.96% of observed threads are actually active
[ 4 / 4 ] CPU activity is good
CPU cores are active 99.99% of time
[ 4 / 4 ] Loop profile is not flat
At least one loop coverage is greater than 4% (44.01%), representing an hotspot for the application
[ 4 / 4 ] Enough time of the experiment time spent in analyzed innermost loops (52.81%)
If the time spent in analyzed innermost loops is less than 15%, standard innermost loop optimizations such as vectorisation will have a limited impact on application performances.
[ 4 / 4 ] Affinity is good (99.99%)
Threads are not migrating to CPU cores: probably successfully pinned
[ 3 / 3 ] Less than 10% (0.00%) is spend in BLAS1 operations
It could be more efficient to inline by hand BLAS1 operations
[ 3 / 3 ] Functions mostly use all threads
Functions running on a reduced number of threads (typically sequential code) cover less than 10% of application walltime (1.82%)
[ 3 / 3 ] Cumulative Outermost/In between loops coverage (45.38%) lower than cumulative innermost loop coverage (52.81%)
Having cumulative Outermost/In between loops coverage greater than cumulative innermost loop coverage will make loop optimization more complex
[ 2 / 2 ] Less than 10% (0.00%) is spend in BLAS2 operations
BLAS2 calls usually could make a poor cache usage and could benefit from inlining.
[ 2 / 2 ] Less than 10% (0.69%) is spend in Libm/SVML (special functions)
| Loop ID | Analysis | Penalty Score |
|---|---|---|
| ►Loop 24 - md-gcc-O2 | Execution Time: 44 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 22 | |
| ○ | [SA] Presence of expensive FP instructions - Perform hoisting, change algorithm, use SVML or proper numerical library or perform value profiling (count the number of distinct input values). There are 5 issues (= instructions) costing 4 points each. | 20 |
| ○ | [SA] Presence of a large number of scalar integer instructions - Simplify loop structure, perform loop splitting or perform unroll and jam. This issue costs 2 points. | 2 |
| ►Control Flow Issues | 22 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ○ | [SA] Too many paths (15 paths) - Simplify control structure. There are 15 issues ( = paths) costing 1 point each with a malus of 4 points. | 19 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 22 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ○ | [SA] Too many paths (15 paths) - Simplify control structure. There are 15 issues ( = paths) costing 1 point each with a malus of 4 points. | 19 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Loop 18 - md-gcc-O2 | Execution Time: 18 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 21 - md-gcc-O2 | Execution Time: 8 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 22 - md-gcc-O2 | Execution Time: 7 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 23 - md-gcc-O2 | Execution Time: 6 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 20 - md-gcc-O2 | Execution Time: 5 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 19 - md-gcc-O2 | Execution Time: 5 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 17 - md-gcc-O2 | Execution Time: 1 % - Vectorization Ratio: 0.00 % - Vector Length Use: 45.83 % | |
| ►Control Flow Issues | 22 | |
| ○ | [SA] Too many paths (16 paths) - Simplify control structure. There are 16 issues ( = paths) costing 1 point each with a malus of 4 points. | 20 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 22 | |
| ○ | [SA] Too many paths (16 paths) - Simplify control structure. There are 16 issues ( = paths) costing 1 point each with a malus of 4 points. | 20 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Loop 38 - md-gcc-O2 | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Control Flow Issues | 1 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ►Data Access Issues | 6 | |
| ○ | [SA] Presence of constant non unit stride data access - Use array restructuring, perform loop interchange or use gather instructions to lower a bit the cost. There are 3 issues ( = data accesses) costing 2 point each. | 6 |
| ►Vectorization Roadblocks | 7 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ○ | [SA] Presence of constant non unit stride data access - Use array restructuring, perform loop interchange or use gather instructions to lower a bit the cost. There are 3 issues ( = data accesses) costing 2 point each. | 6 |
| ►Loop 16 - md-gcc-O2 | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 37.10 % | |
| ►Control Flow Issues | 25 | |
| ○ | [SA] Too many paths (19 paths) - Simplify control structure. There are 19 issues ( = paths) costing 1 point each with a malus of 4 points. | 23 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 25 | |
| ○ | [SA] Too many paths (19 paths) - Simplify control structure. There are 19 issues ( = paths) costing 1 point each with a malus of 4 points. | 23 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
[ 4 / 4 ] Application profile is long enough (131.21 s)
To have good quality measurements, it is advised that the application profiling time is greater than 10 seconds.
[ 3 / 3 ] Optimization level option is correctly used
[ 2.40 / 3 ] Most of time spent in analyzed modules comes from functions without compilation information
Functions without compilation information (typically not compiled with -g) cumulate 0.01% of the time spent in analyzed modules. Check that -g is present. Remark: if -g is indeed used, this can also be due to some compiler built-in functions (typically math) or statically linked libraries. This warning can be ignored in that case.
[ 2 / 3 ] Security settings from the host restrict profiling. Some metrics will be missing or incomplete.
Current value for kernel.perf_event_paranoid is 2. If possible, set it to 1 or check with your system administrator which flag can be used to achieve this.
[ 0 / 3 ] Compilation of some functions is not optimized for the target processor
Application run on the ARM_NEOVERSE_V2 micro-architecture while the code was specialized for armv8.5-a+crypto+rng+sve2-aes+sve2-sha3+sve2-bitperm+i8mm+bf16+nossbs+nopredres. Architecture specific options are needed to produce efficient code for a specific processor ( -mcpu=native ).
[ 2 / 2 ] Application is correctly profiled ("Others" category represents 0.00 % of the execution time)
To have a representative profiling, it is advised that the category "Others" represents less than 20% of the execution time in order to analyze as much as possible of the user code
[ 1 / 1 ] Lstopo present. The Topology lstopo report will be generated.
[ 0 / 0 ] Fastmath not used
Consider to add ffast-math to compilation flags (or replace -O3 with -Ofast) to unlock potential extra speedup by relaxing floating-point computation consistency. Warning: floating-point accuracy may be reduced and the compliance to IEEE/ISO rules/specifications for math functions will be relaxed, typically 'errno' will no longer be set after calling some math functions.
[ 4 / 4 ] Enough time of the experiment time spent in analyzed loops (96.06%)
If the time spent in analyzed loops is less than 30%, standard loop optimizations will have a limited impact on application performances.
[ 4 / 4 ] Threads activity is good
On average, more than 399.87% of observed threads are actually active
[ 4 / 4 ] CPU activity is good
CPU cores are active 99.99% of time
[ 4 / 4 ] Loop profile is not flat
At least one loop coverage is greater than 4% (42.34%), representing an hotspot for the application
[ 4 / 4 ] Enough time of the experiment time spent in analyzed innermost loops (52.31%)
If the time spent in analyzed innermost loops is less than 15%, standard innermost loop optimizations such as vectorisation will have a limited impact on application performances.
[ 4 / 4 ] Affinity is good (99.99%)
Threads are not migrating to CPU cores: probably successfully pinned
[ 3 / 3 ] Less than 10% (0.00%) is spend in BLAS1 operations
It could be more efficient to inline by hand BLAS1 operations
[ 3 / 3 ] Functions mostly use all threads
Functions running on a reduced number of threads (typically sequential code) cover less than 10% of application walltime (0.37%)
[ 3 / 3 ] Cumulative Outermost/In between loops coverage (43.75%) lower than cumulative innermost loop coverage (52.31%)
Having cumulative Outermost/In between loops coverage greater than cumulative innermost loop coverage will make loop optimization more complex
[ 2 / 2 ] Less than 10% (0.00%) is spend in BLAS2 operations
BLAS2 calls usually could make a poor cache usage and could benefit from inlining.
[ 2 / 2 ] Less than 10% (0.66%) is spend in Libm/SVML (special functions)
| Loop ID | Analysis | Penalty Score |
|---|---|---|
| ►Loop 24 - md-gcc-O2 | Execution Time: 42 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 22 | |
| ○ | [SA] Presence of expensive FP instructions - Perform hoisting, change algorithm, use SVML or proper numerical library or perform value profiling (count the number of distinct input values). There are 5 issues (= instructions) costing 4 points each. | 20 |
| ○ | [SA] Presence of a large number of scalar integer instructions - Simplify loop structure, perform loop splitting or perform unroll and jam. This issue costs 2 points. | 2 |
| ►Control Flow Issues | 22 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ○ | [SA] Too many paths (15 paths) - Simplify control structure. There are 15 issues ( = paths) costing 1 point each with a malus of 4 points. | 19 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 22 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ○ | [SA] Too many paths (15 paths) - Simplify control structure. There are 15 issues ( = paths) costing 1 point each with a malus of 4 points. | 19 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Loop 18 - md-gcc-O2 | Execution Time: 18 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 21 - md-gcc-O2 | Execution Time: 8 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 22 - md-gcc-O2 | Execution Time: 7 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 23 - md-gcc-O2 | Execution Time: 6 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 20 - md-gcc-O2 | Execution Time: 5 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 19 - md-gcc-O2 | Execution Time: 5 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 17 - md-gcc-O2 | Execution Time: 1 % - Vectorization Ratio: 0.00 % - Vector Length Use: 45.83 % | |
| ►Control Flow Issues | 22 | |
| ○ | [SA] Too many paths (16 paths) - Simplify control structure. There are 16 issues ( = paths) costing 1 point each with a malus of 4 points. | 20 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 22 | |
| ○ | [SA] Too many paths (16 paths) - Simplify control structure. There are 16 issues ( = paths) costing 1 point each with a malus of 4 points. | 20 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Loop 38 - md-gcc-O2 | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Control Flow Issues | 1 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ►Data Access Issues | 6 | |
| ○ | [SA] Presence of constant non unit stride data access - Use array restructuring, perform loop interchange or use gather instructions to lower a bit the cost. There are 3 issues ( = data accesses) costing 2 point each. | 6 |
| ►Vectorization Roadblocks | 7 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ○ | [SA] Presence of constant non unit stride data access - Use array restructuring, perform loop interchange or use gather instructions to lower a bit the cost. There are 3 issues ( = data accesses) costing 2 point each. | 6 |
| ►Loop 16 - md-gcc-O2 | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 37.10 % | |
| ►Control Flow Issues | 25 | |
| ○ | [SA] Too many paths (19 paths) - Simplify control structure. There are 19 issues ( = paths) costing 1 point each with a malus of 4 points. | 23 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 25 | |
| ○ | [SA] Too many paths (19 paths) - Simplify control structure. There are 19 issues ( = paths) costing 1 point each with a malus of 4 points. | 23 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
[ 4 / 4 ] Application profile is long enough (68.96 s)
To have good quality measurements, it is advised that the application profiling time is greater than 10 seconds.
[ 3 / 3 ] Optimization level option is correctly used
[ 2.40 / 3 ] Most of time spent in analyzed modules comes from functions without compilation information
Functions without compilation information (typically not compiled with -g) cumulate 0.01% of the time spent in analyzed modules. Check that -g is present. Remark: if -g is indeed used, this can also be due to some compiler built-in functions (typically math) or statically linked libraries. This warning can be ignored in that case.
[ 2 / 3 ] Security settings from the host restrict profiling. Some metrics will be missing or incomplete.
Current value for kernel.perf_event_paranoid is 2. If possible, set it to 1 or check with your system administrator which flag can be used to achieve this.
[ 0 / 3 ] Compilation of some functions is not optimized for the target processor
Application run on the ARM_NEOVERSE_V2 micro-architecture while the code was specialized for armv8.5-a+crypto+rng+sve2-aes+sve2-sha3+sve2-bitperm+i8mm+bf16+nossbs+nopredres. Architecture specific options are needed to produce efficient code for a specific processor ( -mcpu=native ).
[ 2 / 2 ] Application is correctly profiled ("Others" category represents 0.00 % of the execution time)
To have a representative profiling, it is advised that the category "Others" represents less than 20% of the execution time in order to analyze as much as possible of the user code
[ 1 / 1 ] Lstopo present. The Topology lstopo report will be generated.
[ 0 / 0 ] Fastmath not used
Consider to add ffast-math to compilation flags (or replace -O3 with -Ofast) to unlock potential extra speedup by relaxing floating-point computation consistency. Warning: floating-point accuracy may be reduced and the compliance to IEEE/ISO rules/specifications for math functions will be relaxed, typically 'errno' will no longer be set after calling some math functions.
[ 4 / 4 ] Enough time of the experiment time spent in analyzed loops (93.31%)
If the time spent in analyzed loops is less than 30%, standard loop optimizations will have a limited impact on application performances.
[ 4 / 4 ] Threads activity is good
On average, more than 799.55% of observed threads are actually active
[ 4 / 4 ] CPU activity is good
CPU cores are active 99.99% of time
[ 4 / 4 ] Loop profile is not flat
At least one loop coverage is greater than 4% (39.81%), representing an hotspot for the application
[ 4 / 4 ] Enough time of the experiment time spent in analyzed innermost loops (52.25%)
If the time spent in analyzed innermost loops is less than 15%, standard innermost loop optimizations such as vectorisation will have a limited impact on application performances.
[ 4 / 4 ] Affinity is good (99.99%)
Threads are not migrating to CPU cores: probably successfully pinned
[ 3 / 3 ] Less than 10% (0.00%) is spend in BLAS1 operations
It could be more efficient to inline by hand BLAS1 operations
[ 3 / 3 ] Functions mostly use all threads
Functions running on a reduced number of threads (typically sequential code) cover less than 10% of application walltime (0.78%)
[ 3 / 3 ] Cumulative Outermost/In between loops coverage (41.07%) lower than cumulative innermost loop coverage (52.25%)
Having cumulative Outermost/In between loops coverage greater than cumulative innermost loop coverage will make loop optimization more complex
[ 2 / 2 ] Less than 10% (0.00%) is spend in BLAS2 operations
BLAS2 calls usually could make a poor cache usage and could benefit from inlining.
[ 2 / 2 ] Less than 10% (0.60%) is spend in Libm/SVML (special functions)
| Loop ID | Analysis | Penalty Score |
|---|---|---|
| ►Loop 24 - md-gcc-O2 | Execution Time: 39 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 22 | |
| ○ | [SA] Presence of expensive FP instructions - Perform hoisting, change algorithm, use SVML or proper numerical library or perform value profiling (count the number of distinct input values). There are 5 issues (= instructions) costing 4 points each. | 20 |
| ○ | [SA] Presence of a large number of scalar integer instructions - Simplify loop structure, perform loop splitting or perform unroll and jam. This issue costs 2 points. | 2 |
| ►Control Flow Issues | 22 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ○ | [SA] Too many paths (15 paths) - Simplify control structure. There are 15 issues ( = paths) costing 1 point each with a malus of 4 points. | 19 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 22 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ○ | [SA] Too many paths (15 paths) - Simplify control structure. There are 15 issues ( = paths) costing 1 point each with a malus of 4 points. | 19 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Loop 18 - md-gcc-O2 | Execution Time: 17 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 21 - md-gcc-O2 | Execution Time: 8 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 22 - md-gcc-O2 | Execution Time: 7 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 23 - md-gcc-O2 | Execution Time: 7 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 20 - md-gcc-O2 | Execution Time: 5 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 19 - md-gcc-O2 | Execution Time: 5 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 17 - md-gcc-O2 | Execution Time: 1 % - Vectorization Ratio: 0.00 % - Vector Length Use: 45.83 % | |
| ►Control Flow Issues | 22 | |
| ○ | [SA] Too many paths (16 paths) - Simplify control structure. There are 16 issues ( = paths) costing 1 point each with a malus of 4 points. | 20 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 22 | |
| ○ | [SA] Too many paths (16 paths) - Simplify control structure. There are 16 issues ( = paths) costing 1 point each with a malus of 4 points. | 20 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Loop 38 - md-gcc-O2 | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Control Flow Issues | 1 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ►Data Access Issues | 6 | |
| ○ | [SA] Presence of constant non unit stride data access - Use array restructuring, perform loop interchange or use gather instructions to lower a bit the cost. There are 3 issues ( = data accesses) costing 2 point each. | 6 |
| ►Vectorization Roadblocks | 7 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ○ | [SA] Presence of constant non unit stride data access - Use array restructuring, perform loop interchange or use gather instructions to lower a bit the cost. There are 3 issues ( = data accesses) costing 2 point each. | 6 |
| ►Loop 16 - md-gcc-O2 | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 37.10 % | |
| ►Control Flow Issues | 25 | |
| ○ | [SA] Too many paths (19 paths) - Simplify control structure. There are 19 issues ( = paths) costing 1 point each with a malus of 4 points. | 23 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 25 | |
| ○ | [SA] Too many paths (19 paths) - Simplify control structure. There are 19 issues ( = paths) costing 1 point each with a malus of 4 points. | 23 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
[ 4 / 4 ] Application profile is long enough (37.15 s)
To have good quality measurements, it is advised that the application profiling time is greater than 10 seconds.
[ 3 / 3 ] Optimization level option is correctly used
[ 2.40 / 3 ] Most of time spent in analyzed modules comes from functions without compilation information
Functions without compilation information (typically not compiled with -g) cumulate 0.02% of the time spent in analyzed modules. Check that -g is present. Remark: if -g is indeed used, this can also be due to some compiler built-in functions (typically math) or statically linked libraries. This warning can be ignored in that case.
[ 2 / 3 ] Security settings from the host restrict profiling. Some metrics will be missing or incomplete.
Current value for kernel.perf_event_paranoid is 2. If possible, set it to 1 or check with your system administrator which flag can be used to achieve this.
[ 0 / 3 ] Compilation of some functions is not optimized for the target processor
Application run on the ARM_NEOVERSE_V2 micro-architecture while the code was specialized for armv8.5-a+crypto+rng+sve2-aes+sve2-sha3+sve2-bitperm+i8mm+bf16+nossbs+nopredres. Architecture specific options are needed to produce efficient code for a specific processor ( -mcpu=native ).
[ 2 / 2 ] Application is correctly profiled ("Others" category represents 0.00 % of the execution time)
To have a representative profiling, it is advised that the category "Others" represents less than 20% of the execution time in order to analyze as much as possible of the user code
[ 1 / 1 ] Lstopo present. The Topology lstopo report will be generated.
[ 0 / 0 ] Fastmath not used
Consider to add ffast-math to compilation flags (or replace -O3 with -Ofast) to unlock potential extra speedup by relaxing floating-point computation consistency. Warning: floating-point accuracy may be reduced and the compliance to IEEE/ISO rules/specifications for math functions will be relaxed, typically 'errno' will no longer be set after calling some math functions.
[ 4 / 4 ] Enough time of the experiment time spent in analyzed loops (89.21%)
If the time spent in analyzed loops is less than 30%, standard loop optimizations will have a limited impact on application performances.
[ 4 / 4 ] Threads activity is good
On average, more than 1598.27% of observed threads are actually active
[ 4 / 4 ] CPU activity is good
CPU cores are active 99.98% of time
[ 4 / 4 ] Loop profile is not flat
At least one loop coverage is greater than 4% (36.59%), representing an hotspot for the application
[ 4 / 4 ] Enough time of the experiment time spent in analyzed innermost loops (51.55%)
If the time spent in analyzed innermost loops is less than 15%, standard innermost loop optimizations such as vectorisation will have a limited impact on application performances.
[ 4 / 4 ] Affinity is good (99.97%)
Threads are not migrating to CPU cores: probably successfully pinned
[ 3 / 3 ] Less than 10% (0.00%) is spend in BLAS1 operations
It could be more efficient to inline by hand BLAS1 operations
[ 3 / 3 ] Functions mostly use all threads
Functions running on a reduced number of threads (typically sequential code) cover less than 10% of application walltime (2.04%)
[ 3 / 3 ] Cumulative Outermost/In between loops coverage (37.67%) lower than cumulative innermost loop coverage (51.55%)
Having cumulative Outermost/In between loops coverage greater than cumulative innermost loop coverage will make loop optimization more complex
[ 2 / 2 ] Less than 10% (0.00%) is spend in BLAS2 operations
BLAS2 calls usually could make a poor cache usage and could benefit from inlining.
[ 2 / 2 ] Less than 10% (0.55%) is spend in Libm/SVML (special functions)
| Loop ID | Analysis | Penalty Score |
|---|---|---|
| ►Loop 24 - md-gcc-O2 | Execution Time: 36 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 22 | |
| ○ | [SA] Presence of expensive FP instructions - Perform hoisting, change algorithm, use SVML or proper numerical library or perform value profiling (count the number of distinct input values). There are 5 issues (= instructions) costing 4 points each. | 20 |
| ○ | [SA] Presence of a large number of scalar integer instructions - Simplify loop structure, perform loop splitting or perform unroll and jam. This issue costs 2 points. | 2 |
| ►Control Flow Issues | 22 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ○ | [SA] Too many paths (15 paths) - Simplify control structure. There are 15 issues ( = paths) costing 1 point each with a malus of 4 points. | 19 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 22 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ○ | [SA] Too many paths (15 paths) - Simplify control structure. There are 15 issues ( = paths) costing 1 point each with a malus of 4 points. | 19 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Loop 18 - md-gcc-O2 | Execution Time: 15 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 21 - md-gcc-O2 | Execution Time: 9 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 22 - md-gcc-O2 | Execution Time: 7 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 23 - md-gcc-O2 | Execution Time: 7 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 20 - md-gcc-O2 | Execution Time: 5 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 19 - md-gcc-O2 | Execution Time: 4 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 17 - md-gcc-O2 | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 45.83 % | |
| ►Control Flow Issues | 22 | |
| ○ | [SA] Too many paths (16 paths) - Simplify control structure. There are 16 issues ( = paths) costing 1 point each with a malus of 4 points. | 20 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 22 | |
| ○ | [SA] Too many paths (16 paths) - Simplify control structure. There are 16 issues ( = paths) costing 1 point each with a malus of 4 points. | 20 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Loop 38 - md-gcc-O2 | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Control Flow Issues | 1 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ►Data Access Issues | 6 | |
| ○ | [SA] Presence of constant non unit stride data access - Use array restructuring, perform loop interchange or use gather instructions to lower a bit the cost. There are 3 issues ( = data accesses) costing 2 point each. | 6 |
| ►Vectorization Roadblocks | 7 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ○ | [SA] Presence of constant non unit stride data access - Use array restructuring, perform loop interchange or use gather instructions to lower a bit the cost. There are 3 issues ( = data accesses) costing 2 point each. | 6 |
| ►Loop 16 - md-gcc-O2 | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 37.10 % | |
| ►Control Flow Issues | 25 | |
| ○ | [SA] Too many paths (19 paths) - Simplify control structure. There are 19 issues ( = paths) costing 1 point each with a malus of 4 points. | 23 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 25 | |
| ○ | [SA] Too many paths (19 paths) - Simplify control structure. There are 19 issues ( = paths) costing 1 point each with a malus of 4 points. | 23 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
[ 4 / 4 ] Application profile is long enough (19.98 s)
To have good quality measurements, it is advised that the application profiling time is greater than 10 seconds.
[ 3 / 3 ] Optimization level option is correctly used
[ 2.40 / 3 ] Most of time spent in analyzed modules comes from functions without compilation information
Functions without compilation information (typically not compiled with -g) cumulate 0.04% of the time spent in analyzed modules. Check that -g is present. Remark: if -g is indeed used, this can also be due to some compiler built-in functions (typically math) or statically linked libraries. This warning can be ignored in that case.
[ 2 / 3 ] Security settings from the host restrict profiling. Some metrics will be missing or incomplete.
Current value for kernel.perf_event_paranoid is 2. If possible, set it to 1 or check with your system administrator which flag can be used to achieve this.
[ 0 / 3 ] Compilation of some functions is not optimized for the target processor
Application run on the ARM_NEOVERSE_V2 micro-architecture while the code was specialized for armv8.5-a+crypto+rng+sve2-aes+sve2-sha3+sve2-bitperm+i8mm+bf16+nossbs+nopredres. Architecture specific options are needed to produce efficient code for a specific processor ( -mcpu=native ).
[ 2 / 2 ] Application is correctly profiled ("Others" category represents 0.00 % of the execution time)
To have a representative profiling, it is advised that the category "Others" represents less than 20% of the execution time in order to analyze as much as possible of the user code
[ 1 / 1 ] Lstopo present. The Topology lstopo report will be generated.
[ 0 / 0 ] Fastmath not used
Consider to add ffast-math to compilation flags (or replace -O3 with -Ofast) to unlock potential extra speedup by relaxing floating-point computation consistency. Warning: floating-point accuracy may be reduced and the compliance to IEEE/ISO rules/specifications for math functions will be relaxed, typically 'errno' will no longer be set after calling some math functions.
[ 4 / 4 ] Enough time of the experiment time spent in analyzed loops (86.19%)
If the time spent in analyzed loops is less than 30%, standard loop optimizations will have a limited impact on application performances.
[ 4 / 4 ] Threads activity is good
On average, more than 3192.94% of observed threads are actually active
[ 4 / 4 ] CPU activity is good
CPU cores are active 99.95% of time
[ 4 / 4 ] Loop profile is not flat
At least one loop coverage is greater than 4% (33.25%), representing an hotspot for the application
[ 4 / 4 ] Enough time of the experiment time spent in analyzed innermost loops (52.08%)
If the time spent in analyzed innermost loops is less than 15%, standard innermost loop optimizations such as vectorisation will have a limited impact on application performances.
[ 4 / 4 ] Affinity is good (99.95%)
Threads are not migrating to CPU cores: probably successfully pinned
[ 3 / 3 ] Less than 10% (0.00%) is spend in BLAS1 operations
It could be more efficient to inline by hand BLAS1 operations
[ 3 / 3 ] Functions mostly use all threads
Functions running on a reduced number of threads (typically sequential code) cover less than 10% of application walltime (4.42%)
[ 3 / 3 ] Cumulative Outermost/In between loops coverage (34.11%) lower than cumulative innermost loop coverage (52.08%)
Having cumulative Outermost/In between loops coverage greater than cumulative innermost loop coverage will make loop optimization more complex
[ 2 / 2 ] Less than 10% (0.00%) is spend in BLAS2 operations
BLAS2 calls usually could make a poor cache usage and could benefit from inlining.
[ 2 / 2 ] Less than 10% (0.54%) is spend in Libm/SVML (special functions)
| Loop ID | Analysis | Penalty Score |
|---|---|---|
| ►Loop 24 - md-gcc-O2 | Execution Time: 33 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 22 | |
| ○ | [SA] Presence of expensive FP instructions - Perform hoisting, change algorithm, use SVML or proper numerical library or perform value profiling (count the number of distinct input values). There are 5 issues (= instructions) costing 4 points each. | 20 |
| ○ | [SA] Presence of a large number of scalar integer instructions - Simplify loop structure, perform loop splitting or perform unroll and jam. This issue costs 2 points. | 2 |
| ►Control Flow Issues | 22 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ○ | [SA] Too many paths (15 paths) - Simplify control structure. There are 15 issues ( = paths) costing 1 point each with a malus of 4 points. | 19 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 22 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ○ | [SA] Too many paths (15 paths) - Simplify control structure. There are 15 issues ( = paths) costing 1 point each with a malus of 4 points. | 19 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Loop 18 - md-gcc-O2 | Execution Time: 14 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 21 - md-gcc-O2 | Execution Time: 10 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 22 - md-gcc-O2 | Execution Time: 8 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 23 - md-gcc-O2 | Execution Time: 8 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 20 - md-gcc-O2 | Execution Time: 4 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 19 - md-gcc-O2 | Execution Time: 4 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 17 - md-gcc-O2 | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 45.83 % | |
| ►Control Flow Issues | 22 | |
| ○ | [SA] Too many paths (16 paths) - Simplify control structure. There are 16 issues ( = paths) costing 1 point each with a malus of 4 points. | 20 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 22 | |
| ○ | [SA] Too many paths (16 paths) - Simplify control structure. There are 16 issues ( = paths) costing 1 point each with a malus of 4 points. | 20 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Loop 38 - md-gcc-O2 | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Control Flow Issues | 1 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ►Data Access Issues | 6 | |
| ○ | [SA] Presence of constant non unit stride data access - Use array restructuring, perform loop interchange or use gather instructions to lower a bit the cost. There are 3 issues ( = data accesses) costing 2 point each. | 6 |
| ►Vectorization Roadblocks | 7 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ○ | [SA] Presence of constant non unit stride data access - Use array restructuring, perform loop interchange or use gather instructions to lower a bit the cost. There are 3 issues ( = data accesses) costing 2 point each. | 6 |
| ○Loop 36 - md-gcc-O2 | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % |
[ 4 / 4 ] Application profile is long enough (14.27 s)
To have good quality measurements, it is advised that the application profiling time is greater than 10 seconds.
[ 3 / 3 ] Optimization level option is correctly used
[ 2.40 / 3 ] Most of time spent in analyzed modules comes from functions without compilation information
Functions without compilation information (typically not compiled with -g) cumulate 0.06% of the time spent in analyzed modules. Check that -g is present. Remark: if -g is indeed used, this can also be due to some compiler built-in functions (typically math) or statically linked libraries. This warning can be ignored in that case.
[ 2 / 3 ] Security settings from the host restrict profiling. Some metrics will be missing or incomplete.
Current value for kernel.perf_event_paranoid is 2. If possible, set it to 1 or check with your system administrator which flag can be used to achieve this.
[ 0 / 3 ] Compilation of some functions is not optimized for the target processor
Application run on the ARM_NEOVERSE_V2 micro-architecture while the code was specialized for armv8.5-a+crypto+rng+sve2-aes+sve2-sha3+sve2-bitperm+i8mm+bf16+nossbs+nopredres. Architecture specific options are needed to produce efficient code for a specific processor ( -mcpu=native ).
[ 2 / 2 ] Application is correctly profiled ("Others" category represents 0.00 % of the execution time)
To have a representative profiling, it is advised that the category "Others" represents less than 20% of the execution time in order to analyze as much as possible of the user code
[ 1 / 1 ] Lstopo present. The Topology lstopo report will be generated.
[ 0 / 0 ] Fastmath not used
Consider to add ffast-math to compilation flags (or replace -O3 with -Ofast) to unlock potential extra speedup by relaxing floating-point computation consistency. Warning: floating-point accuracy may be reduced and the compliance to IEEE/ISO rules/specifications for math functions will be relaxed, typically 'errno' will no longer be set after calling some math functions.
[ 4 / 4 ] Enough time of the experiment time spent in analyzed loops (83.28%)
If the time spent in analyzed loops is less than 30%, standard loop optimizations will have a limited impact on application performances.
[ 4 / 4 ] Threads activity is good
On average, more than 4779.51% of observed threads are actually active
[ 4 / 4 ] CPU activity is good
CPU cores are active 99.83% of time
[ 4 / 4 ] Loop profile is not flat
At least one loop coverage is greater than 4% (31.22%), representing an hotspot for the application
[ 4 / 4 ] Enough time of the experiment time spent in analyzed innermost loops (51.30%)
If the time spent in analyzed innermost loops is less than 15%, standard innermost loop optimizations such as vectorisation will have a limited impact on application performances.
[ 4 / 4 ] Affinity is good (99.94%)
Threads are not migrating to CPU cores: probably successfully pinned
[ 3 / 3 ] Less than 10% (0.00%) is spend in BLAS1 operations
It could be more efficient to inline by hand BLAS1 operations
[ 3 / 3 ] Functions mostly use all threads
Functions running on a reduced number of threads (typically sequential code) cover less than 10% of application walltime (8.60%)
[ 3 / 3 ] Cumulative Outermost/In between loops coverage (31.99%) lower than cumulative innermost loop coverage (51.30%)
Having cumulative Outermost/In between loops coverage greater than cumulative innermost loop coverage will make loop optimization more complex
[ 2 / 2 ] Less than 10% (0.00%) is spend in BLAS2 operations
BLAS2 calls usually could make a poor cache usage and could benefit from inlining.
[ 2 / 2 ] Less than 10% (0.46%) is spend in Libm/SVML (special functions)
| Loop ID | Analysis | Penalty Score |
|---|---|---|
| ►Loop 24 - md-gcc-O2 | Execution Time: 31 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 22 | |
| ○ | [SA] Presence of expensive FP instructions - Perform hoisting, change algorithm, use SVML or proper numerical library or perform value profiling (count the number of distinct input values). There are 5 issues (= instructions) costing 4 points each. | 20 |
| ○ | [SA] Presence of a large number of scalar integer instructions - Simplify loop structure, perform loop splitting or perform unroll and jam. This issue costs 2 points. | 2 |
| ►Control Flow Issues | 22 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ○ | [SA] Too many paths (15 paths) - Simplify control structure. There are 15 issues ( = paths) costing 1 point each with a malus of 4 points. | 19 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 22 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ○ | [SA] Too many paths (15 paths) - Simplify control structure. There are 15 issues ( = paths) costing 1 point each with a malus of 4 points. | 19 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Loop 18 - md-gcc-O2 | Execution Time: 13 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 21 - md-gcc-O2 | Execution Time: 10 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 22 - md-gcc-O2 | Execution Time: 8 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 23 - md-gcc-O2 | Execution Time: 8 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 20 - md-gcc-O2 | Execution Time: 4 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 19 - md-gcc-O2 | Execution Time: 4 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Loop Computation Issues | 4 | |
| ○ | [SA] Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA - Reorganize arithmetic expressions to exhibit potential for FMA. This issue costs 4 points. | 4 |
| ►Loop 17 - md-gcc-O2 | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 45.83 % | |
| ►Control Flow Issues | 22 | |
| ○ | [SA] Too many paths (16 paths) - Simplify control structure. There are 16 issues ( = paths) costing 1 point each with a malus of 4 points. | 20 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 22 | |
| ○ | [SA] Too many paths (16 paths) - Simplify control structure. There are 16 issues ( = paths) costing 1 point each with a malus of 4 points. | 20 |
| ○ | [SA] Non innermost loop (InBetween) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Loop 38 - md-gcc-O2 | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % | |
| ►Control Flow Issues | 1 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ►Data Access Issues | 6 | |
| ○ | [SA] Presence of constant non unit stride data access - Use array restructuring, perform loop interchange or use gather instructions to lower a bit the cost. There are 3 issues ( = data accesses) costing 2 point each. | 6 |
| ►Vectorization Roadblocks | 7 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 1 issues (= calls) costing 1 point each. | 1 |
| ○ | [SA] Presence of constant non unit stride data access - Use array restructuring, perform loop interchange or use gather instructions to lower a bit the cost. There are 3 issues ( = data accesses) costing 2 point each. | 6 |
| ○Loop 36 - md-gcc-O2 | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 50.00 % |