| | 1 | = PDAF_print_info = |
| | 2 | |
| | 3 | This page documents the routine `PDAF_print_info` of PDAF. |
| | 4 | |
| | 5 | This routine is called to display memory and timing information measured by PDAF. |
| | 6 | |
| | 7 | == Displaying memory information == |
| | 8 | |
| | 9 | Information about the memory required by PDAF through allocated arrays can be obtained by inserting into the program the line |
| | 10 | {{{ |
| | 11 | CALL PDAF_print_info(2) |
| | 12 | }}} |
| | 13 | The function displays the following information |
| | 14 | * Memory required for the ensemble array, state vector, and transform matrix |
| | 15 | * Memory required by the analysis step |
| | 16 | * Memory required to perform the ensemble transformation |
| | 17 | |
| | 18 | The output will look like this: |
| | 19 | {{{ |
| | 20 | PDAF Memory overview |
| | 21 | --------------------------------------------- |
| | 22 | Allocated memory (MB) |
| | 23 | state and U: 0.59617 MB (persistent) |
| | 24 | ensemble array: 0.64087 MB (persistent) |
| | 25 | analysis step: 6.05578 MB (temporary) |
| | 26 | resampling: 2.81129 MB (temporary) |
| | 27 | }}} |
| | 28 | |
| | 29 | Currently only the memory required by the first process of the filter processes is displayed. Thus the total required memory should be the displayed memory multiplied by the number of processes in `COMM_filter`. |
| | 30 | |
| | 31 | == Displaying timing information == |
| | 32 | |
| | 33 | Timing information can be displayed by adding |
| | 34 | {{{ |
| | 35 | CALL PDAF_print_info(X) |
| | 36 | }}} |
| | 37 | to the code. Where X is the timer level to be shown. Available choices are |
| | 38 | * X=1: Basic timers |
| | 39 | * X=3: Timers showing the time spent int he different call-back routines (this veriant was added with PDAF 1.15) |
| | 40 | * X=4: More detailed timers about parts of the filter algorithm (before PDAF 1.15, this was timer level 3) |
| | 41 | * X=5: Very detailed timers about various operations in the filter algorithm (before PDAF 1.15, this was timer level 4) |
| | 42 | |
| | 43 | For X=1, the output will look like |
| | 44 | {{{ |
| | 45 | PDAF PDAF Timing information |
| | 46 | PDAF --------------------------------------------- |
| | 47 | PDAF Initialize PDAF: 0.078 s |
| | 48 | PDAF Ensemble forecast: 0.003 s |
| | 49 | PDAF LESTKF analysis: 25.183 s |
| | 50 | PDAF Prepoststep: 0.017 s |
| | 51 | }}} |
| | 52 | |
| | 53 | We recommend to use X=3 for optimizing the user routines. The output will look like |
| | 54 | {{{ |
| | 55 | PDAF PDAF Timing information - call-back routines |
| | 56 | PDAF ---------------------------------------------------- |
| | 57 | PDAF Initialize PDAF: 0.078 s |
| | 58 | PDAF init_ens_pdaf: 0.077 s |
| | 59 | PDAF Ensemble forecast: 0.003 s |
| | 60 | PDAF MPI communication in PDAF: 0.000 s |
| | 61 | PDAF distribute_state_pdaf: 0.001 s |
| | 62 | PDAF collect_state_pdaf: 0.000 s |
| | 63 | PDAF LESTKF analysis: 25.183 s |
| | 64 | PDAF PDAF-internal operations: 24.762 s |
| | 65 | PDAF init_n_domains_pdaf: 0.000 s |
| | 66 | PDAF init_dim_obs_f_pdaf: 0.000 s |
| | 67 | PDAF obs_op_f_pdaf: 0.003 s |
| | 68 | PDAF init_dim_l_pdaf: 0.001 s |
| | 69 | PDAF init_dim_obs_l_pdaf: 0.001 s |
| | 70 | PDAF g2l_state_pdaf: 0.001 s |
| | 71 | PDAF g2l_obs_pdaf: 0.021 s |
| | 72 | PDAF init_obs_l_pdaf: 0.000 s |
| | 73 | PDAF prodRinvA_l_pdaf: 0.006 s |
| | 74 | PDAF l2g_state_pdaf: 0.014 s |
| | 75 | PDAF prepoststep_pdaf: 0.017 s |
| | 76 | }}} |
| | 77 | This output will show you in which routine most time is spent. If it is a call-back routine, you have the possibility to check for optimizations to this routine to speed up the computation. If most time is spend inside PDAF, as in the example, you can't further optimize. |
| | 78 | |
| | 79 | |
