Help is available by moving the cursor above any 
symbol or by checking MAQAO website.
[ 4 / 4 ] Application profile is long enough (213.01 s)
To have good quality measurements, it is advised that the application profiling time is greater than 10 seconds.
[ 0 / 3 ] Most of time spent in analyzed modules comes from functions without source/debug info
Functions without source/debug info (typically not compiled with -g) cumulate 30.83% 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.
[ 0 / 3 ] Architecture specific options are unknown for some functions (compilation options info are not available)
Architecture specific options are needed to produce efficient code for a specific processor ( -march=(target) ).
[ 0 / 3 ] Most of time spent in analyzed modules comes from functions without compilation options informations
Functions without compilation options information cumulate 30.83% of the time spent in analyzed modules. Check that -g is present.
[ 0 / 3 ] Optimization level is unknown for some functions (compilation options info are not available)
To have better performances, it is advised to help the compiler by using a proper optimization level (-O2 of higher). Warning, depending on compilers, faster optimization levels can decrease numeric accuracy.
[ 3 / 3 ] Host configuration allows retrieval of all necessary metrics.
[ 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.
[ 4 / 4 ] Enough time of the experiment time spent in analyzed loops (30.64%)
If the time spent in analyzed loops is less than 30%, standard loop optimizations will have a limited impact on application performances.
[ 2 / 4 ] A significant amount of threads are idle (40.58%)
On average, more than 10% of observed threads are idle. Such threads are probably IO/sync waiting. Some hints: use faster filesystems to read/write data, improve parallel load balancing and/or scheduling.
[ 2 / 4 ] CPU activity is below 90% (63.50%)
CPU cores are idle more than 10% of time. Threads supposed to run on these cores are probably IO/sync waiting. Some hints: use faster filesystems to read/write data, improve parallel load balancing and/or scheduling.
[ 4 / 4 ] Loop profile is not flat
At least one loop coverage is greater than 4% (16.05%), representing an hotspot for the application
[ 0 / 4 ] Too little time of the experiment time spent in analyzed innermost loops (14.40%)
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.
[ 1 / 4 ] Affinity stability is lower than 90% (43.17%)
Threads are often migrating to other CPU cores/threads. For OpenMP, typically set (OMP_PLACES=cores OMP_PROC_BIND=close) or (OMP_PLACES=threads OMP_PROC_BIND=spread). With OpenMPI + OpenMP, use --bind-to core --map-by node:PE=$OMP_NUM_THREADS --report-bindings. With IntelMPI + OpenMP, set I_MPI_PIN_DOMAIN=omp:compact or I_MPI_PIN_DOMAIN=omp:scatter and use -print-rank-map.
[ 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 (0.44%)
[ 0 / 3 ] Cumulative Outermost/In between loops coverage (16.24%) greater than cumulative innermost loop coverage (14.40%)
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% (0.00%) is spend in Libm/SVML (special functions)
| Loop ID | Analysis | Penalty Score |
|---|---|---|
| ►Loop 547 - libdmumps.so | Execution Time: 16 % - Vectorization Ratio: 5.13 % - Vector Length Use: 12.20 % | |
| ►Control Flow Issues | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [SA] Non innermost loop (Outermost) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [SA] Non innermost loop (Outermost) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ○Loop 548 - libdmumps.so | Execution Time: 8 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop 1765 - libdmumps.so | Execution Time: 2 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop Computation Issues | 6 | |
| ○ | [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 |
| ○ | [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 |
| ►Data Access Issues | 4 | |
| ○ | [SA] Presence of indirect accesses - Use array restructuring or gather instructions to lower the cost. There are 1 issues ( = indirect data accesses) costing 4 point each. | 4 |
| ►Vectorization Roadblocks | 4 | |
| ○ | [SA] Presence of indirect accesses - Use array restructuring or gather instructions to lower the cost. There are 1 issues ( = indirect data accesses) costing 4 point each. | 4 |
| ○Loop 1940 - libdmumps.so | Execution Time: 1 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop 854 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Data Access Issues | 4 | |
| ○ | [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 2 issues ( = data accesses) costing 2 point each. | 4 |
| ►Vectorization Roadblocks | 4 | |
| ○ | [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 2 issues ( = data accesses) costing 2 point each. | 4 |
| ►Loop 1936 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop Computation Issues | 2 | |
| ○ | [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 | 2 | |
| ○ | [SA] Several paths (2 paths) - Simplify control structure or force the compiler to use masked instructions. There are 2 issues ( = paths) costing 1 point each. | 2 |
| ►Data Access Issues | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Vectorization Roadblocks | 4 | |
| ○ | [SA] Several paths (2 paths) - Simplify control structure or force the compiler to use masked instructions. There are 2 issues ( = paths) costing 1 point each. | 2 |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Loop 552 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 33.33 % - Vector Length Use: 16.67 % | |
| ►Data Access Issues | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Vectorization Roadblocks | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ○Loop 1942 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop 840 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Data Access Issues | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Vectorization Roadblocks | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ○Loop 1947 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % |
[ 4 / 4 ] Application profile is long enough (161.68 s)
To have good quality measurements, it is advised that the application profiling time is greater than 10 seconds.
[ 0 / 3 ] Most of time spent in analyzed modules comes from functions without source/debug info
Functions without source/debug info (typically not compiled with -g) cumulate 31.60% 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.
[ 0 / 3 ] Architecture specific options are unknown for some functions (compilation options info are not available)
Architecture specific options are needed to produce efficient code for a specific processor ( -march=(target) ).
[ 0 / 3 ] Most of time spent in analyzed modules comes from functions without compilation options informations
Functions without compilation options information cumulate 31.79% of the time spent in analyzed modules. Check that -g is present.
[ 0 / 3 ] Optimization level is unknown for some functions (compilation options info are not available)
To have better performances, it is advised to help the compiler by using a proper optimization level (-O2 of higher). Warning, depending on compilers, faster optimization levels can decrease numeric accuracy.
[ 3 / 3 ] Host configuration allows retrieval of all necessary metrics.
[ 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.
[ 4 / 4 ] Enough time of the experiment time spent in analyzed loops (30.69%)
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 106.97% of observed threads are actually active
[ 2 / 4 ] CPU activity is below 90% (58.36%)
CPU cores are idle more than 10% of time. Threads supposed to run on these cores are probably IO/sync waiting. Some hints: use faster filesystems to read/write data, improve parallel load balancing and/or scheduling.
[ 4 / 4 ] Loop profile is not flat
At least one loop coverage is greater than 4% (9.85%), representing an hotspot for the application
[ 4 / 4 ] Enough time of the experiment time spent in analyzed innermost loops (20.14%)
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.
[ 3 / 4 ] Affinity stability is lower than 90% (70.49%)
Threads are often migrating to other CPU cores/threads. For OpenMP, typically set (OMP_PLACES=cores OMP_PROC_BIND=close) or (OMP_PLACES=threads OMP_PROC_BIND=spread). With OpenMPI + OpenMP, use --bind-to core --map-by node:PE=$OMP_NUM_THREADS --report-bindings. With IntelMPI + OpenMP, set I_MPI_PIN_DOMAIN=omp:compact or I_MPI_PIN_DOMAIN=omp:scatter and use -print-rank-map.
[ 3 / 3 ] Less than 10% (0.00%) is spend in BLAS1 operations
It could be more efficient to inline by hand BLAS1 operations
[ 0 / 3 ] Too many functions do not use all threads
Functions running on a reduced number of threads (typically sequential code) cover at least 10% of application walltime (14.77%). Check both "Max Inclusive Time Over Threads" and "Nb Threads" in Functions or Loops tabs and consider parallelizing sequential regions or improving parallelization of regions running on a reduced number of threads
[ 3 / 3 ] Cumulative Outermost/In between loops coverage (10.55%) lower than cumulative innermost loop coverage (20.14%)
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% (0.00%) is spend in Libm/SVML (special functions)
| Loop ID | Analysis | Penalty Score |
|---|---|---|
| ►Loop 547 - libdmumps.so | Execution Time: 9 % - Vectorization Ratio: 5.13 % - Vector Length Use: 12.20 % | |
| ►Control Flow Issues | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [SA] Non innermost loop (Outermost) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [SA] Non innermost loop (Outermost) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ○Loop 548 - libdmumps.so | Execution Time: 5 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ○Loop 731 - libdmumps.so | Execution Time: 3 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop 753 - libdmumps.so | Execution Time: 1 % - Vectorization Ratio: 33.33 % - Vector Length Use: 16.67 % | |
| ►Data Access Issues | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Vectorization Roadblocks | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Loop 1765 - libdmumps.so | Execution Time: 1 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop Computation Issues | 6 | |
| ○ | [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 |
| ○ | [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 |
| ►Data Access Issues | 4 | |
| ○ | [SA] Presence of indirect accesses - Use array restructuring or gather instructions to lower the cost. There are 1 issues ( = indirect data accesses) costing 4 point each. | 4 |
| ►Vectorization Roadblocks | 4 | |
| ○ | [SA] Presence of indirect accesses - Use array restructuring or gather instructions to lower the cost. There are 1 issues ( = indirect data accesses) costing 4 point each. | 4 |
| ►Loop 746 - libdmumps.so | Execution Time: 1 % - Vectorization Ratio: 33.33 % - Vector Length Use: 16.67 % | |
| ►Data Access Issues | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Vectorization Roadblocks | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Loop 748 - libdmumps.so | Execution Time: 1 % - Vectorization Ratio: 12.50 % - Vector Length Use: 11.72 % | |
| ►Control Flow Issues | 2 | |
| ○ | [SA] Several paths (2 paths) - Simplify control structure or force the compiler to use masked instructions. There are 2 issues ( = paths) costing 1 point each. | 2 |
| ►Data Access Issues | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Vectorization Roadblocks | 4 | |
| ○ | [SA] Several paths (2 paths) - Simplify control structure or force the compiler to use masked instructions. There are 2 issues ( = paths) costing 1 point each. | 2 |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ○Loop 740 - libdmumps.so | Execution Time: 1 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ○Loop 1940 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop 730 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 11.48 % | |
| ►Loop Computation Issues | 2 | |
| ○ | [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 | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [SA] Non innermost loop (Outermost) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [SA] Non innermost loop (Outermost) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
[ 4 / 4 ] Application profile is long enough (161.57 s)
To have good quality measurements, it is advised that the application profiling time is greater than 10 seconds.
[ 0 / 3 ] Most of time spent in analyzed modules comes from functions without source/debug info
Functions without source/debug info (typically not compiled with -g) cumulate 22.35% 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.
[ 0 / 3 ] Architecture specific options are unknown for some functions (compilation options info are not available)
Architecture specific options are needed to produce efficient code for a specific processor ( -march=(target) ).
[ 0 / 3 ] Most of time spent in analyzed modules comes from functions without compilation options informations
Functions without compilation options information cumulate 23.12% of the time spent in analyzed modules. Check that -g is present.
[ 0 / 3 ] Optimization level is unknown for some functions (compilation options info are not available)
To have better performances, it is advised to help the compiler by using a proper optimization level (-O2 of higher). Warning, depending on compilers, faster optimization levels can decrease numeric accuracy.
[ 3 / 3 ] Host configuration allows retrieval of all necessary metrics.
[ 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 / 4 ] Too little time of the experiment time spent in analyzed loops (19.38%)
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 155.26% of observed threads are actually active
[ 3 / 4 ] CPU activity is below 90% (77.67%)
CPU cores are idle more than 10% of time. Threads supposed to run on these cores are probably IO/sync waiting. Some hints: use faster filesystems to read/write data, improve parallel load balancing and/or scheduling.
[ 0 / 4 ] Loop profile is flat
No hotspot found in the application (greatest loop coverage is 3.25%), and the twenty hottest loops cumulated coverage is lower than 20% of the application profiled time (18.47%)
[ 4 / 4 ] Enough time of the experiment time spent in analyzed innermost loops (16.52%)
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.
[ 2 / 4 ] Affinity stability is lower than 90% (56.27%)
Threads are often migrating to other CPU cores/threads. For OpenMP, typically set (OMP_PLACES=cores OMP_PROC_BIND=close) or (OMP_PLACES=threads OMP_PROC_BIND=spread). With OpenMPI + OpenMP, use --bind-to core --map-by node:PE=$OMP_NUM_THREADS --report-bindings. With IntelMPI + OpenMP, set I_MPI_PIN_DOMAIN=omp:compact or I_MPI_PIN_DOMAIN=omp:scatter and use -print-rank-map.
[ 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 (0.00%)
[ 3 / 3 ] Cumulative Outermost/In between loops coverage (2.86%) lower than cumulative innermost loop coverage (16.52%)
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% (0.00%) is spend in Libm/SVML (special functions)
| Loop ID | Analysis | Penalty Score |
|---|---|---|
| ○Loop 731 - libdmumps.so | Execution Time: 3 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop 753 - libdmumps.so | Execution Time: 2 % - Vectorization Ratio: 33.33 % - Vector Length Use: 16.67 % | |
| ►Data Access Issues | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Vectorization Roadblocks | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Loop 547 - libdmumps.so | Execution Time: 2 % - Vectorization Ratio: 5.13 % - Vector Length Use: 12.20 % | |
| ►Control Flow Issues | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [SA] Non innermost loop (Outermost) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [SA] Non innermost loop (Outermost) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Loop 746 - libdmumps.so | Execution Time: 2 % - Vectorization Ratio: 33.33 % - Vector Length Use: 16.67 % | |
| ►Data Access Issues | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Vectorization Roadblocks | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ○Loop 548 - libdmumps.so | Execution Time: 1 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop 748 - libdmumps.so | Execution Time: 1 % - Vectorization Ratio: 12.50 % - Vector Length Use: 11.72 % | |
| ►Control Flow Issues | 2 | |
| ○ | [SA] Several paths (2 paths) - Simplify control structure or force the compiler to use masked instructions. There are 2 issues ( = paths) costing 1 point each. | 2 |
| ►Data Access Issues | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Vectorization Roadblocks | 4 | |
| ○ | [SA] Several paths (2 paths) - Simplify control structure or force the compiler to use masked instructions. There are 2 issues ( = paths) costing 1 point each. | 2 |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ○Loop 740 - libdmumps.so | Execution Time: 1 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop 885 - libdmumps.so | Execution Time: 1 % - Vectorization Ratio: 4.56 % - Vector Length Use: 9.67 % | |
| ►Loop Computation Issues | 2 | |
| ○ | [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 | 39 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 39 issues (= calls) costing 1 point each. | 39 |
| ►Data Access Issues | 2 | |
| ○ | [SA] More than 20% of the loads are accessing the stack - Perform loop splitting to decrease pressure on registers. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 1039 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 39 issues (= calls) costing 1 point each. | 39 |
| ○ | [SA] Too many paths (at least 1000 paths) - Simplify control structure. There are at least 1000 issues ( = paths) costing 1 point. | 1000 |
| ►Loop 754 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 20.00 % - Vector Length Use: 13.75 % | |
| ►Data Access Issues | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Vectorization Roadblocks | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Loop 1765 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop Computation Issues | 6 | |
| ○ | [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 |
| ○ | [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 |
| ►Data Access Issues | 4 | |
| ○ | [SA] Presence of indirect accesses - Use array restructuring or gather instructions to lower the cost. There are 1 issues ( = indirect data accesses) costing 4 point each. | 4 |
| ►Vectorization Roadblocks | 4 | |
| ○ | [SA] Presence of indirect accesses - Use array restructuring or gather instructions to lower the cost. There are 1 issues ( = indirect data accesses) costing 4 point each. | 4 |
[ 4 / 4 ] Application profile is long enough (167.76 s)
To have good quality measurements, it is advised that the application profiling time is greater than 10 seconds.
[ 0 / 3 ] Most of time spent in analyzed modules comes from functions without source/debug info
Functions without source/debug info (typically not compiled with -g) cumulate 21.59% 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.
[ 0 / 3 ] Architecture specific options are unknown for some functions (compilation options info are not available)
Architecture specific options are needed to produce efficient code for a specific processor ( -march=(target) ).
[ 0 / 3 ] Most of time spent in analyzed modules comes from functions without compilation options informations
Functions without compilation options information cumulate 23.01% of the time spent in analyzed modules. Check that -g is present.
[ 0 / 3 ] Optimization level is unknown for some functions (compilation options info are not available)
To have better performances, it is advised to help the compiler by using a proper optimization level (-O2 of higher). Warning, depending on compilers, faster optimization levels can decrease numeric accuracy.
[ 3 / 3 ] Host configuration allows retrieval of all necessary metrics.
[ 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 / 4 ] Too little time of the experiment time spent in analyzed loops (16.13%)
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 330.38% of observed threads are actually active
[ 3 / 4 ] CPU activity is below 90% (82.64%)
CPU cores are idle more than 10% of time. Threads supposed to run on these cores are probably IO/sync waiting. Some hints: use faster filesystems to read/write data, improve parallel load balancing and/or scheduling.
[ 0 / 4 ] Loop profile is flat
No hotspot found in the application (greatest loop coverage is 3.04%), and the twenty hottest loops cumulated coverage is lower than 20% of the application profiled time (15.71%)
[ 4 / 4 ] Enough time of the experiment time spent in analyzed innermost loops (15.13%)
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.
[ 2 / 4 ] Affinity stability is lower than 90% (66.45%)
Threads are often migrating to other CPU cores/threads. For OpenMP, typically set (OMP_PLACES=cores OMP_PROC_BIND=close) or (OMP_PLACES=threads OMP_PROC_BIND=spread). With OpenMPI + OpenMP, use --bind-to core --map-by node:PE=$OMP_NUM_THREADS --report-bindings. With IntelMPI + OpenMP, set I_MPI_PIN_DOMAIN=omp:compact or I_MPI_PIN_DOMAIN=omp:scatter and use -print-rank-map.
[ 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 (0.00%)
[ 3 / 3 ] Cumulative Outermost/In between loops coverage (0.99%) lower than cumulative innermost loop coverage (15.13%)
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% (0.00%) is spend in Libm/SVML (special functions)
| Loop ID | Analysis | Penalty Score |
|---|---|---|
| ►Loop 753 - libdmumps.so | Execution Time: 3 % - Vectorization Ratio: 33.33 % - Vector Length Use: 16.67 % | |
| ►Data Access Issues | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Vectorization Roadblocks | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ○Loop 731 - libdmumps.so | Execution Time: 2 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop 885 - libdmumps.so | Execution Time: 2 % - Vectorization Ratio: 4.56 % - Vector Length Use: 9.67 % | |
| ►Loop Computation Issues | 2 | |
| ○ | [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 | 39 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 39 issues (= calls) costing 1 point each. | 39 |
| ►Data Access Issues | 2 | |
| ○ | [SA] More than 20% of the loads are accessing the stack - Perform loop splitting to decrease pressure on registers. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 1039 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 39 issues (= calls) costing 1 point each. | 39 |
| ○ | [SA] Too many paths (at least 1000 paths) - Simplify control structure. There are at least 1000 issues ( = paths) costing 1 point. | 1000 |
| ►Loop 746 - libdmumps.so | Execution Time: 1 % - Vectorization Ratio: 33.33 % - Vector Length Use: 16.67 % | |
| ►Data Access Issues | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Vectorization Roadblocks | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Loop 748 - libdmumps.so | Execution Time: 1 % - Vectorization Ratio: 12.50 % - Vector Length Use: 11.72 % | |
| ►Control Flow Issues | 2 | |
| ○ | [SA] Several paths (2 paths) - Simplify control structure or force the compiler to use masked instructions. There are 2 issues ( = paths) costing 1 point each. | 2 |
| ►Data Access Issues | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Vectorization Roadblocks | 4 | |
| ○ | [SA] Several paths (2 paths) - Simplify control structure or force the compiler to use masked instructions. There are 2 issues ( = paths) costing 1 point each. | 2 |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ○Loop 740 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop 547 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 5.13 % - Vector Length Use: 12.20 % | |
| ►Control Flow Issues | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [SA] Non innermost loop (Outermost) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [SA] Non innermost loop (Outermost) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Loop 179 - libmumps_common.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 8.25 % | |
| ►Control Flow Issues | 3 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ►Data Access Issues | 4 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ○ | [SA] More than 20% of the loads are accessing the stack - Perform loop splitting to decrease pressure on registers. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ○Loop 548 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop 328 - libmumps_common.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 10.25 % | |
| ►Control Flow Issues | 13 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 9 issues (= calls) costing 1 point each. | 9 |
| ○ | [SA] Several paths (4 paths) - Simplify control structure or force the compiler to use masked instructions. There are 4 issues ( = paths) costing 1 point each. | 4 |
| ►Data Access Issues | 4 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ○ | [SA] More than 20% of the loads are accessing the stack - Perform loop splitting to decrease pressure on registers. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 15 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 9 issues (= calls) costing 1 point each. | 9 |
| ○ | [SA] Several paths (4 paths) - Simplify control structure or force the compiler to use masked instructions. There are 4 issues ( = paths) costing 1 point each. | 4 |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
[ 4 / 4 ] Application profile is long enough (50.40 s)
To have good quality measurements, it is advised that the application profiling time is greater than 10 seconds.
[ 0 / 3 ] Most of time spent in analyzed modules comes from functions without source/debug info
Functions without source/debug info (typically not compiled with -g) cumulate 20.01% 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.
[ 0 / 3 ] Architecture specific options are unknown for some functions (compilation options info are not available)
Architecture specific options are needed to produce efficient code for a specific processor ( -march=(target) ).
[ 0 / 3 ] Most of time spent in analyzed modules comes from functions without compilation options informations
Functions without compilation options information cumulate 21.15% of the time spent in analyzed modules. Check that -g is present.
[ 0 / 3 ] Optimization level is unknown for some functions (compilation options info are not available)
To have better performances, it is advised to help the compiler by using a proper optimization level (-O2 of higher). Warning, depending on compilers, faster optimization levels can decrease numeric accuracy.
[ 3 / 3 ] Host configuration allows retrieval of all necessary metrics.
[ 2 / 2 ] Application is correctly profiled ("Others" category represents 2.54 % 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 / 4 ] Too little time of the experiment time spent in analyzed loops (15.34%)
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 480.35% of observed threads are actually active
[ 2 / 4 ] CPU activity is below 90% (60.14%)
CPU cores are idle more than 10% of time. Threads supposed to run on these cores are probably IO/sync waiting. Some hints: use faster filesystems to read/write data, improve parallel load balancing and/or scheduling.
[ 0 / 4 ] Loop profile is flat
No hotspot found in the application (greatest loop coverage is 2.99%), and the twenty hottest loops cumulated coverage is lower than 20% of the application profiled time (14.62%)
[ 0 / 4 ] Too little time of the experiment time spent in analyzed innermost loops (14.29%)
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.
[ 2 / 4 ] Affinity stability is lower than 90% (57.04%)
Threads are often migrating to other CPU cores/threads. For OpenMP, typically set (OMP_PLACES=cores OMP_PROC_BIND=close) or (OMP_PLACES=threads OMP_PROC_BIND=spread). With OpenMPI + OpenMP, use --bind-to core --map-by node:PE=$OMP_NUM_THREADS --report-bindings. With IntelMPI + OpenMP, set I_MPI_PIN_DOMAIN=omp:compact or I_MPI_PIN_DOMAIN=omp:scatter and use -print-rank-map.
[ 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 (0.00%)
[ 3 / 3 ] Cumulative Outermost/In between loops coverage (1.05%) lower than cumulative innermost loop coverage (14.29%)
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% (0.00%) is spend in Libm/SVML (special functions)
| Loop ID | Analysis | Penalty Score |
|---|---|---|
| ○Loop 731 - libdmumps.so | Execution Time: 2 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop 753 - libdmumps.so | Execution Time: 2 % - Vectorization Ratio: 33.33 % - Vector Length Use: 16.67 % | |
| ►Data Access Issues | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Vectorization Roadblocks | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Loop 885 - libdmumps.so | Execution Time: 1 % - Vectorization Ratio: 4.56 % - Vector Length Use: 9.67 % | |
| ►Loop Computation Issues | 2 | |
| ○ | [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 | 39 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 39 issues (= calls) costing 1 point each. | 39 |
| ►Data Access Issues | 2 | |
| ○ | [SA] More than 20% of the loads are accessing the stack - Perform loop splitting to decrease pressure on registers. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 1039 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 39 issues (= calls) costing 1 point each. | 39 |
| ○ | [SA] Too many paths (at least 1000 paths) - Simplify control structure. There are at least 1000 issues ( = paths) costing 1 point. | 1000 |
| ►Loop 746 - libdmumps.so | Execution Time: 1 % - Vectorization Ratio: 33.33 % - Vector Length Use: 16.67 % | |
| ►Data Access Issues | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Vectorization Roadblocks | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Loop 748 - libdmumps.so | Execution Time: 1 % - Vectorization Ratio: 12.50 % - Vector Length Use: 11.72 % | |
| ►Control Flow Issues | 2 | |
| ○ | [SA] Several paths (2 paths) - Simplify control structure or force the compiler to use masked instructions. There are 2 issues ( = paths) costing 1 point each. | 2 |
| ►Data Access Issues | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Vectorization Roadblocks | 4 | |
| ○ | [SA] Several paths (2 paths) - Simplify control structure or force the compiler to use masked instructions. There are 2 issues ( = paths) costing 1 point each. | 2 |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ○Loop 740 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop 179 - libmumps_common.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 8.25 % | |
| ►Control Flow Issues | 3 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ►Data Access Issues | 4 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ○ | [SA] More than 20% of the loads are accessing the stack - Perform loop splitting to decrease pressure on registers. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ○Loop 548 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop 547 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 5.13 % - Vector Length Use: 12.20 % | |
| ►Control Flow Issues | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [SA] Non innermost loop (Outermost) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [SA] Non innermost loop (Outermost) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Loop 730 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 11.48 % | |
| ►Loop Computation Issues | 2 | |
| ○ | [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 | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [SA] Non innermost loop (Outermost) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [SA] Non innermost loop (Outermost) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
[ 4 / 4 ] Application profile is long enough (11.84 s)
To have good quality measurements, it is advised that the application profiling time is greater than 10 seconds.
[ 0 / 3 ] Most of time spent in analyzed modules comes from functions without source/debug info
Functions without source/debug info (typically not compiled with -g) cumulate 10.17% 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.
[ 0 / 3 ] Architecture specific options are unknown for some functions (compilation options info are not available)
Architecture specific options are needed to produce efficient code for a specific processor ( -march=(target) ).
[ 0 / 3 ] Most of time spent in analyzed modules comes from functions without compilation options informations
Functions without compilation options information cumulate 10.46% of the time spent in analyzed modules. Check that -g is present.
[ 0 / 3 ] Optimization level is unknown for some functions (compilation options info are not available)
To have better performances, it is advised to help the compiler by using a proper optimization level (-O2 of higher). Warning, depending on compilers, faster optimization levels can decrease numeric accuracy.
[ 3 / 3 ] Host configuration allows retrieval of all necessary metrics.
[ 2 / 2 ] Application is correctly profiled ("Others" category represents 3.60 % 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 / 4 ] Too little time of the experiment time spent in analyzed loops (9.04%)
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 407.01% of observed threads are actually active
[ 0 / 4 ] CPU activity is below 90% (25.53%)
CPU cores are idle more than 10% of time. Threads supposed to run on these cores are probably IO/sync waiting. Some hints: use faster filesystems to read/write data, improve parallel load balancing and/or scheduling.
[ 0 / 4 ] Loop profile is flat
No hotspot found in the application (greatest loop coverage is 1.28%), and the twenty hottest loops cumulated coverage is lower than 20% of the application profiled time (8.20%)
[ 0 / 4 ] Too little time of the experiment time spent in analyzed innermost loops (8.02%)
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.
[ 0 / 4 ] Affinity stability is lower than 90% (24.98%)
Threads are often migrating to other CPU cores/threads. For OpenMP, typically set (OMP_PLACES=cores OMP_PROC_BIND=close) or (OMP_PLACES=threads OMP_PROC_BIND=spread). With OpenMPI + OpenMP, use --bind-to core --map-by node:PE=$OMP_NUM_THREADS --report-bindings. With IntelMPI + OpenMP, set I_MPI_PIN_DOMAIN=omp:compact or I_MPI_PIN_DOMAIN=omp:scatter and use -print-rank-map.
[ 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 (0.00%)
[ 3 / 3 ] Cumulative Outermost/In between loops coverage (1.02%) lower than cumulative innermost loop coverage (8.02%)
Having cumulative Outermost/In between loops coverage greater than cumulative innermost loop coverage will make loop optimization more complex
[ 2 / 2 ] Less than 10% (0.01%) is spend in BLAS2 operations
BLAS2 calls usually could make a poor cache usage and could benefit from inlining.
[ 2 / 2 ] Less than 10% (0.00%) is spend in Libm/SVML (special functions)
| Loop ID | Analysis | Penalty Score |
|---|---|---|
| ○Loop 731 - libdmumps.so | Execution Time: 1 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ○Loop 548 - libdmumps.so | Execution Time: 1 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop 753 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 33.33 % - Vector Length Use: 16.67 % | |
| ►Data Access Issues | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Vectorization Roadblocks | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Loop 547 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 5.13 % - Vector Length Use: 12.20 % | |
| ►Control Flow Issues | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [SA] Non innermost loop (Outermost) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [SA] Non innermost loop (Outermost) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Loop 1797 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 10.94 % | |
| ►Loop Computation Issues | 6 | |
| ○ | [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 |
| ○ | [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 |
| ►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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ○ | [SA] Presence of indirect accesses - Use array restructuring or gather instructions to lower the cost. There are 1 issues ( = indirect data accesses) costing 4 point each. | 4 |
| ►Vectorization Roadblocks | 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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ○ | [SA] Presence of indirect accesses - Use array restructuring or gather instructions to lower the cost. There are 1 issues ( = indirect data accesses) costing 4 point each. | 4 |
| ►Loop 179 - libmumps_common.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 8.25 % | |
| ►Control Flow Issues | 3 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ►Data Access Issues | 4 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ○ | [SA] More than 20% of the loads are accessing the stack - Perform loop splitting to decrease pressure on registers. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Loop 1765 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop Computation Issues | 6 | |
| ○ | [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 |
| ○ | [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 |
| ►Data Access Issues | 4 | |
| ○ | [SA] Presence of indirect accesses - Use array restructuring or gather instructions to lower the cost. There are 1 issues ( = indirect data accesses) costing 4 point each. | 4 |
| ►Vectorization Roadblocks | 4 | |
| ○ | [SA] Presence of indirect accesses - Use array restructuring or gather instructions to lower the cost. There are 1 issues ( = indirect data accesses) costing 4 point each. | 4 |
| ►Loop 748 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 12.50 % - Vector Length Use: 11.72 % | |
| ►Control Flow Issues | 2 | |
| ○ | [SA] Several paths (2 paths) - Simplify control structure or force the compiler to use masked instructions. There are 2 issues ( = paths) costing 1 point each. | 2 |
| ►Data Access Issues | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Vectorization Roadblocks | 4 | |
| ○ | [SA] Several paths (2 paths) - Simplify control structure or force the compiler to use masked instructions. There are 2 issues ( = paths) costing 1 point each. | 2 |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Loop 746 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 33.33 % - Vector Length Use: 16.67 % | |
| ►Data Access Issues | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Vectorization Roadblocks | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Loop 885 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 4.56 % - Vector Length Use: 9.67 % | |
| ►Loop Computation Issues | 2 | |
| ○ | [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 | 39 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 39 issues (= calls) costing 1 point each. | 39 |
| ►Data Access Issues | 2 | |
| ○ | [SA] More than 20% of the loads are accessing the stack - Perform loop splitting to decrease pressure on registers. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 1039 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 39 issues (= calls) costing 1 point each. | 39 |
| ○ | [SA] Too many paths (at least 1000 paths) - Simplify control structure. There are at least 1000 issues ( = paths) costing 1 point. | 1000 |
[ 0 / 4 ] Application profile is too short (6.77 s)
If the overall application profiling time is less than 10 seconds, many of the measurements at function or loop level will very likely be under the measurement quality threshold (0,1 seconds).
Rerun to increase runtime duration: for example use a larger dataset, include a repetition loop or change profile_start or profile_stop settings.
[ 0 / 3 ] Most of time spent in analyzed modules comes from functions without source/debug info
Functions without source/debug info (typically not compiled with -g) cumulate 8.82% 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.
[ 0 / 3 ] Architecture specific options are unknown for some functions (compilation options info are not available)
Architecture specific options are needed to produce efficient code for a specific processor ( -march=(target) ).
[ 0 / 3 ] Most of time spent in analyzed modules comes from functions without compilation options informations
Functions without compilation options information cumulate 9.09% of the time spent in analyzed modules. Check that -g is present.
[ 0 / 3 ] Optimization level is unknown for some functions (compilation options info are not available)
To have better performances, it is advised to help the compiler by using a proper optimization level (-O2 of higher). Warning, depending on compilers, faster optimization levels can decrease numeric accuracy.
[ 3 / 3 ] Host configuration allows retrieval of all necessary metrics.
[ 2 / 2 ] Application is correctly profiled ("Others" category represents 2.01 % 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 / 4 ] Too little time of the experiment time spent in analyzed loops (7.55%)
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 520.64% of observed threads are actually active
[ 0 / 4 ] CPU activity is below 90% (16.32%)
CPU cores are idle more than 10% of time. Threads supposed to run on these cores are probably IO/sync waiting. Some hints: use faster filesystems to read/write data, improve parallel load balancing and/or scheduling.
[ 0 / 4 ] Loop profile is flat
No hotspot found in the application (greatest loop coverage is 0.95%), and the twenty hottest loops cumulated coverage is lower than 20% of the application profiled time (6.69%)
[ 0 / 4 ] Too little time of the experiment time spent in analyzed innermost loops (6.88%)
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.
[ 0 / 4 ] Affinity stability is lower than 90% (15.80%)
Threads are often migrating to other CPU cores/threads. For OpenMP, typically set (OMP_PLACES=cores OMP_PROC_BIND=close) or (OMP_PLACES=threads OMP_PROC_BIND=spread). With OpenMPI + OpenMP, use --bind-to core --map-by node:PE=$OMP_NUM_THREADS --report-bindings. With IntelMPI + OpenMP, set I_MPI_PIN_DOMAIN=omp:compact or I_MPI_PIN_DOMAIN=omp:scatter and use -print-rank-map.
[ 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 (0.00%)
[ 3 / 3 ] Cumulative Outermost/In between loops coverage (0.67%) lower than cumulative innermost loop coverage (6.88%)
Having cumulative Outermost/In between loops coverage greater than cumulative innermost loop coverage will make loop optimization more complex
[ 2 / 2 ] Less than 10% (0.01%) is spend in BLAS2 operations
BLAS2 calls usually could make a poor cache usage and could benefit from inlining.
[ 2 / 2 ] Less than 10% (0.00%) is spend in Libm/SVML (special functions)
| Loop ID | Analysis | Penalty Score |
|---|---|---|
| ○Loop 731 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop 1797 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 10.94 % | |
| ►Loop Computation Issues | 6 | |
| ○ | [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 |
| ○ | [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 |
| ►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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ○ | [SA] Presence of indirect accesses - Use array restructuring or gather instructions to lower the cost. There are 1 issues ( = indirect data accesses) costing 4 point each. | 4 |
| ►Vectorization Roadblocks | 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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ○ | [SA] Presence of indirect accesses - Use array restructuring or gather instructions to lower the cost. There are 1 issues ( = indirect data accesses) costing 4 point each. | 4 |
| ○Loop 548 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop 753 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 33.33 % - Vector Length Use: 16.67 % | |
| ►Data Access Issues | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Vectorization Roadblocks | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Loop 179 - libmumps_common.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 8.25 % | |
| ►Control Flow Issues | 3 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ►Data Access Issues | 4 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ○ | [SA] More than 20% of the loads are accessing the stack - Perform loop splitting to decrease pressure on registers. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Loop 1765 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop Computation Issues | 6 | |
| ○ | [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 |
| ○ | [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 |
| ►Data Access Issues | 4 | |
| ○ | [SA] Presence of indirect accesses - Use array restructuring or gather instructions to lower the cost. There are 1 issues ( = indirect data accesses) costing 4 point each. | 4 |
| ►Vectorization Roadblocks | 4 | |
| ○ | [SA] Presence of indirect accesses - Use array restructuring or gather instructions to lower the cost. There are 1 issues ( = indirect data accesses) costing 4 point each. | 4 |
| ►Loop 885 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 4.56 % - Vector Length Use: 9.67 % | |
| ►Loop Computation Issues | 2 | |
| ○ | [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 | 39 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 39 issues (= calls) costing 1 point each. | 39 |
| ►Data Access Issues | 2 | |
| ○ | [SA] More than 20% of the loads are accessing the stack - Perform loop splitting to decrease pressure on registers. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 1039 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 39 issues (= calls) costing 1 point each. | 39 |
| ○ | [SA] Too many paths (at least 1000 paths) - Simplify control structure. There are at least 1000 issues ( = paths) costing 1 point. | 1000 |
| ►Loop 547 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 5.13 % - Vector Length Use: 12.20 % | |
| ►Control Flow Issues | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [SA] Non innermost loop (Outermost) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [SA] Non innermost loop (Outermost) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Loop 748 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 12.50 % - Vector Length Use: 11.72 % | |
| ►Control Flow Issues | 2 | |
| ○ | [SA] Several paths (2 paths) - Simplify control structure or force the compiler to use masked instructions. There are 2 issues ( = paths) costing 1 point each. | 2 |
| ►Data Access Issues | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Vectorization Roadblocks | 4 | |
| ○ | [SA] Several paths (2 paths) - Simplify control structure or force the compiler to use masked instructions. There are 2 issues ( = paths) costing 1 point each. | 2 |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Loop 746 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 33.33 % - Vector Length Use: 16.67 % | |
| ►Data Access Issues | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Vectorization Roadblocks | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
[ 0 / 4 ] Application profile is too short (4.91 s)
If the overall application profiling time is less than 10 seconds, many of the measurements at function or loop level will very likely be under the measurement quality threshold (0,1 seconds).
Rerun to increase runtime duration: for example use a larger dataset, include a repetition loop or change profile_start or profile_stop settings.
[ 0 / 3 ] Most of time spent in analyzed modules comes from functions without source/debug info
Functions without source/debug info (typically not compiled with -g) cumulate 7.10% 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.
[ 0 / 3 ] Architecture specific options are unknown for some functions (compilation options info are not available)
Architecture specific options are needed to produce efficient code for a specific processor ( -march=(target) ).
[ 0 / 3 ] Most of time spent in analyzed modules comes from functions without compilation options informations
Functions without compilation options information cumulate 7.31% of the time spent in analyzed modules. Check that -g is present.
[ 0 / 3 ] Optimization level is unknown for some functions (compilation options info are not available)
To have better performances, it is advised to help the compiler by using a proper optimization level (-O2 of higher). Warning, depending on compilers, faster optimization levels can decrease numeric accuracy.
[ 3 / 3 ] Host configuration allows retrieval of all necessary metrics.
[ 2 / 2 ] Application is correctly profiled ("Others" category represents 0.88 % 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 / 4 ] Too little time of the experiment time spent in analyzed loops (6.24%)
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 538.11% of observed threads are actually active
[ 0 / 4 ] CPU activity is below 90% (12.56%)
CPU cores are idle more than 10% of time. Threads supposed to run on these cores are probably IO/sync waiting. Some hints: use faster filesystems to read/write data, improve parallel load balancing and/or scheduling.
[ 0 / 4 ] Loop profile is flat
No hotspot found in the application (greatest loop coverage is 0.74%), and the twenty hottest loops cumulated coverage is lower than 20% of the application profiled time (5.43%)
[ 0 / 4 ] Too little time of the experiment time spent in analyzed innermost loops (5.65%)
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.
[ 0 / 4 ] Affinity stability is lower than 90% (11.12%)
Threads are often migrating to other CPU cores/threads. For OpenMP, typically set (OMP_PLACES=cores OMP_PROC_BIND=close) or (OMP_PLACES=threads OMP_PROC_BIND=spread). With OpenMPI + OpenMP, use --bind-to core --map-by node:PE=$OMP_NUM_THREADS --report-bindings. With IntelMPI + OpenMP, set I_MPI_PIN_DOMAIN=omp:compact or I_MPI_PIN_DOMAIN=omp:scatter and use -print-rank-map.
[ 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 (0.00%)
[ 3 / 3 ] Cumulative Outermost/In between loops coverage (0.59%) lower than cumulative innermost loop coverage (5.65%)
Having cumulative Outermost/In between loops coverage greater than cumulative innermost loop coverage will make loop optimization more complex
[ 2 / 2 ] Less than 10% (0.01%) is spend in BLAS2 operations
BLAS2 calls usually could make a poor cache usage and could benefit from inlining.
[ 2 / 2 ] Less than 10% (0.00%) is spend in Libm/SVML (special functions)
| Loop ID | Analysis | Penalty Score |
|---|---|---|
| ○Loop 548 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop 1797 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 10.94 % | |
| ►Loop Computation Issues | 6 | |
| ○ | [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 |
| ○ | [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 |
| ►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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ○ | [SA] Presence of indirect accesses - Use array restructuring or gather instructions to lower the cost. There are 1 issues ( = indirect data accesses) costing 4 point each. | 4 |
| ►Vectorization Roadblocks | 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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ○ | [SA] Presence of indirect accesses - Use array restructuring or gather instructions to lower the cost. There are 1 issues ( = indirect data accesses) costing 4 point each. | 4 |
| ○Loop 731 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop 1765 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop Computation Issues | 6 | |
| ○ | [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 |
| ○ | [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 |
| ►Data Access Issues | 4 | |
| ○ | [SA] Presence of indirect accesses - Use array restructuring or gather instructions to lower the cost. There are 1 issues ( = indirect data accesses) costing 4 point each. | 4 |
| ►Vectorization Roadblocks | 4 | |
| ○ | [SA] Presence of indirect accesses - Use array restructuring or gather instructions to lower the cost. There are 1 issues ( = indirect data accesses) costing 4 point each. | 4 |
| ►Loop 179 - libmumps_common.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 8.25 % | |
| ►Control Flow Issues | 3 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ►Data Access Issues | 4 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ○ | [SA] More than 20% of the loads are accessing the stack - Perform loop splitting to decrease pressure on registers. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ○Loop 1940 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 0.00 % - Vector Length Use: 12.50 % | |
| ►Loop 547 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 5.13 % - Vector Length Use: 12.20 % | |
| ►Control Flow Issues | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [SA] Non innermost loop (Outermost) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 5 | |
| ○ | [SA] Several paths (3 paths) - Simplify control structure or force the compiler to use masked instructions. There are 3 issues ( = paths) costing 1 point each. | 3 |
| ○ | [SA] Non innermost loop (Outermost) - Collapse loop with innermost ones. This issue costs 2 points. | 2 |
| ►Loop 753 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 33.33 % - Vector Length Use: 16.67 % | |
| ►Data Access Issues | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Vectorization Roadblocks | 2 | |
| ○ | [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 1 issues ( = data accesses) costing 2 point each. | 2 |
| ►Loop 885 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 4.56 % - Vector Length Use: 9.67 % | |
| ►Loop Computation Issues | 2 | |
| ○ | [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 | 39 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 39 issues (= calls) costing 1 point each. | 39 |
| ►Data Access Issues | 2 | |
| ○ | [SA] More than 20% of the loads are accessing the stack - Perform loop splitting to decrease pressure on registers. This issue costs 2 points. | 2 |
| ►Vectorization Roadblocks | 1039 | |
| ○ | [SA] Presence of calls - Inline either by compiler or by hand and use SVML for libm calls. There are 39 issues (= calls) costing 1 point each. | 39 |
| ○ | [SA] Too many paths (at least 1000 paths) - Simplify control structure. There are at least 1000 issues ( = paths) costing 1 point. | 1000 |
| ►Loop 2844 - libdmumps.so | Execution Time: 0 % - Vectorization Ratio: 29.17 % - Vector Length Use: 16.15 % | |
| ►Control Flow Issues | 2 | |
| ○ | [SA] Several paths (2 paths) - Simplify control structure or force the compiler to use masked instructions. There are 2 issues ( = paths) costing 1 point each. | 2 |
| ►Vectorization Roadblocks | 2 | |
| ○ | [SA] Several paths (2 paths) - Simplify control structure or force the compiler to use masked instructions. There are 2 issues ( = paths) costing 1 point each. | 2 |