Skylake GCC Ofast Manual Unroll | Skylake Clang O3 + ffast-math Manual Unroll | Skylake ICPX Ofast Manual Unroll |
---|---|---|
[ 3 / 3 ] Host configuration allows retrieval of all necessary metrics. | [ 3 / 3 ] Host configuration allows retrieval of all necessary metrics. | [ 3 / 3 ] Host configuration allows retrieval of all necessary metrics. |
Not available for this run | [ 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. | Not available for this run |
[ 0 / 3 ] Compilation of some functions is not optimized for the target processor Application run on the SKYLAKE micro-architecture while the code was specialized for skylake-avx512. | [ 0 / 3 ] Compilation of some functions is not optimized for the target processor Architecture specific options are needed to produce efficient code for a specific processor ( -x(target) or -ax(target) ). | [ 3 / 3 ] Architecture specific option -x Host is used |
[ 3 / 3 ] Most of time spent in analyzed modules comes from functions compiled with -g and -fno-omit-frame-pointer -g option gives access to debugging informations, such are source locations. -fno-omit-frame-pointer improve the accuracy of callchains found during the application profiling. | [ 0 / 3 ] Most of time spent in analyzed modules comes from functions without compilation information Functions without compilation information (typically not compiled with -g and -grecord-gcc-switches) cumulate 100.00% of the time spent in analyzed modules. Check that -g and -grecord-gcc-switches are present. Remark: if -g and -grecord-gcc-switches are 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. | [ 3 / 3 ] Most of time spent in analyzed modules comes from functions compiled with -g and -fno-omit-frame-pointer -g option gives access to debugging informations, such are source locations. -fno-omit-frame-pointer improves the accuracy of callchains found during the application profiling. |
[ 4 / 4 ] Application profile is long enough (105.15 s) To have good quality measurements, it is advised that the application profiling time is greater than 10 seconds. | [ 4 / 4 ] Application profile is long enough (94.40 s) To have good quality measurements, it is advised that the application profiling time is greater than 10 seconds. | [ 4 / 4 ] Application profile is long enough (28.13 s) To have good quality measurements, it is advised that the application profiling time is greater than 10 seconds. |
[ 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 | [ 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 | [ 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 |
[ 3 / 3 ] Optimization level option is correctly used | [ 0 / 3 ] Some functions are compiled with a low optimization level (O0 or O1) 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 ] Optimization level option is correctly used |
[ 1 / 1 ] Lstopo present. The Topology lstopo report will be generated. | [ 1 / 1 ] Lstopo present. The Topology lstopo report will be generated. | [ 1 / 1 ] Lstopo present. The Topology lstopo report will be generated. |
Skylake GCC Ofast Manual Unroll | Skylake Clang O3 + ffast-math Manual Unroll | Skylake ICPX Ofast Manual Unroll |
---|---|---|
[ 4 / 4 ] CPU activity is good CPU cores are active 96.76% of time | [ 4 / 4 ] CPU activity is good CPU cores are active 96.38% of time | [ 3 / 4 ] CPU activity is below 90% (87.83%) 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 ] Affinity is good (96.62%) Threads are not migrating to CPU cores: probably successfully pinned | [ 4 / 4 ] Affinity is good (96.21%) Threads are not migrating to CPU cores: probably successfully pinned | [ 3 / 4 ] Affinity stability is lower than 90% (87.18%) 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 ] Functions mostly use all threads Functions running on a reduced number of threads (typically sequential code) cover less than 10% of application walltime (0.02%) | [ 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 (0.02%) | [ 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 (0.04%) |
[ 3 / 3 ] Cumulative Outermost/In between loops coverage (1.41%) lower than cumulative innermost loop coverage (98.44%) Having cumulative Outermost/In between loops coverage greater than cumulative innermost loop coverage will make loop optimization more complex | [ 3 / 3 ] Cumulative Outermost/In between loops coverage (3.21%) lower than cumulative innermost loop coverage (96.62%) Having cumulative Outermost/In between loops coverage greater than cumulative innermost loop coverage will make loop optimization more complex | [ 3 / 3 ] Cumulative Outermost/In between loops coverage (8.63%) lower than cumulative innermost loop coverage (90.77%) Having cumulative Outermost/In between loops coverage greater than cumulative innermost loop coverage will make loop optimization more complex |
[ 2 / 4 ] A significant amount of threads are idle (47.31%) 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. | [ 1 / 4 ] A significant amount of threads are idle (50.20%) 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. | [ 0 / 4 ] A significant amount of threads are idle (78.92%) 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 / 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 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 BLAS2 operations BLAS2 calls usually could make a poor cache usage and could benefit from inlining. |
[ 4 / 4 ] Enough time of the experiment time spent in analyzed innermost loops (98.44%) 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 ] Enough time of the experiment time spent in analyzed innermost loops (96.62%) 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 ] Enough time of the experiment time spent in analyzed innermost loops (90.77%) 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 / 3 ] Less than 10% (0.00%) is spend in BLAS1 operations It could be more efficient to inline by hand BLAS1 operations | [ 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 ] Less than 10% (0.00%) is spend in BLAS1 operations It could be more efficient to inline by hand BLAS1 operations |
[ 2 / 2 ] Less than 10% (0.00%) is spend in Libm/SVML (special functions) | [ 2 / 2 ] Less than 10% (0.00%) is spend in Libm/SVML (special functions) | [ 2 / 2 ] Less than 10% (0.00%) is spend in Libm/SVML (special functions) |
[ 4 / 4 ] Loop profile is not flat At least one loop coverage is greater than 4% (96.17%), representing an hotspot for the application | [ 4 / 4 ] Loop profile is not flat At least one loop coverage is greater than 4% (94.40%), representing an hotspot for the application | [ 4 / 4 ] Loop profile is not flat At least one loop coverage is greater than 4% (80.25%), representing an hotspot for the application |
[ 4 / 4 ] Enough time of the experiment time spent in analyzed loops (99.85%) If the time spent in analyzed loops is less than 30%, standard loop optimizations will have a limited impact on application performances. | [ 4 / 4 ] Enough time of the experiment time spent in analyzed loops (99.83%) If the time spent in analyzed loops is less than 30%, standard loop optimizations will have a limited impact on application performances. | [ 4 / 4 ] Enough time of the experiment time spent in analyzed loops (99.40%) If the time spent in analyzed loops is less than 30%, standard loop optimizations will have a limited impact on application performances. |
Analysis | r0 | r1 | r2 | |
---|---|---|---|---|
Loop Computation Issues | Less than 10% of the FP ADD/SUB/MUL arithmetic operations are performed using FMA | 2 | 3 | 1 |
Presence of a large number of scalar integer instructions | 2 | 2 | 2 | |
Low iteration count | 0 | 0 | 1 | |
Control Flow Issues | Presence of 2 to 4 paths | 0 | 1 | 0 |
Non-innermost loop | 1 | 1 | 1 | |
Low iteration count | 0 | 0 | 1 | |
Data Access Issues | Presence of indirect access | 1 | 1 | 1 |
Presence of expensive instructions: scatter/gather | 0 | 0 | 1 | |
Presence of special instructions executing on a single port | 1 | 2 | 2 | |
More than 20% of the loads are accessing the stack | 0 | 1 | 0 | |
Vectorization Roadblocks | Presence of 2 to 4 paths | 0 | 1 | 0 |
Non-innermost loop | 1 | 1 | 1 | |
Presence of indirect access | 1 | 1 | 1 | |
Inefficient Vectorization | Presence of expensive instructions: scatter/gather | 0 | 0 | 1 |
Presence of special instructions executing on a single port | 1 | 2 | 2 | |
Use of masked instructions | 0 | 0 | 1 |