PPPA_G~1 (1158360), страница 3
Текст из файла (страница 3)
C WRITE (3,*) B
C CLOSE (3)
END
INTERVAL ( LINE= 24 SOURCE= Jacobi.cdv ) LEVEL=0 EXE_COUNT=1
--- Main characteristics ---
| Parallelization efficiency | 0.767783 | |
| Execution time | 2.102363 | |
| Processors | 4 | |
| Total time | 8.409452 | |
| Productive time | 6.456632 | ( CPU time= 6.188302 I/O time= 0.009660 Sys.CPU= 0.258670 ) |
| Lost time | 1.952820 | |
| - Insufficient parallelism | 1.031156 | ( User = 0.226167 Sys = 0.804989 ) |
| - Communication | 0.575413 | |
| - Idle time | 0.346251 | |
| Real synchronization | 0.506832 | |
| Synchronization | 3.576560 | |
| Time variation | 2.150352 | |
| Overlap | 0.125000 | ( Reduction=0.042000 Shadow=0.083000 ) |
| Load Imbalance | 0.066160 |
| Nop | Communicat. | Real synchro. | Synchro | Variation | |
| I/O | 11 | 0.000000 | 0.000000 | 0.462173 | 0.502587 |
| Start reduction | 4 | 0.007683 | 0.000000 | 0.470711 | 0.470848 |
| Wait reduction | 4 | 0.496737 | 0.478112 | 0.470921 | 0.005110 |
| Start shadow | 4 | 0.016004 | 0.000000 | 0.033208 | 0.047597 |
| Wait shadow | 4 | 0.054989 | 0.028720 | 0.047595 | 0.010037 |
Note: there are only non-distributed data print operators in the program, this fact explains the absence of input/output communication losses. As such data have the same value on each processor such operations are executed by input/output processor without interprocessor exchanges.
--- Comparative characteristics ---
| T min | Npr | T max | Npr | T mid | |
| Lost_time | 0.473324 | 3 | 0.498466 | 1 | 0.488205 |
| User insufficient_par. | 0.029247 | 1 | 0.074063 | 2 | 0.056542 |
| Sys insufficient par. | 0.088352 | 1 | 0.245586 | 3 | 0.201247 |
| Idle | 0.000000 | 1 | 0.124955 | 2 | 0.086563 |
| Communication | 0.035971 | 3 | 0.380868 | 1 | 0.143853 |
| Real synchronization | 0.015710 | 3 | 0.363574 | 1 | 0.126708 |
| Synchronization | 0.122760 | 3 | 0.363574 | 1 | 0.894140 |
| Time_variation | 0.093413 | 3 | 3.400139 | 1 | 0.537588 |
| Execution_time | 1.977408 | 2 | 2.102363 | 1 | 2.015800 |
| User CPUtime | 1.529593 | 4 | 1.574446 | 1 | 1.547076 |
| Sys CPU_time | 0.022260 | 1 | 0.081105 | 3 | 0.064668 |
| I/O_time | 0.000736 | 4 | 0.007191 | 1 | 0.002415 |
| Load_Imbalance | 0.000000 | 3 | 0.031577 | 1 | 0.016540 |
| Processors | 4 | 1 | 4 | 1 | 4 |
| Communication | Real Synchro | Synchronization | Variation | |
| I/O Tmin I/O Tmax I/O Tmid | 0.000000 1 0.000000 1 0.000000 | 0.000000 1 0.000000 1 0.000000 | 0.001516 3 0.256436 1 0.115543 | 0.018253 3 0.247435 1 0.125647 |
| Start reduction Tmin Start reduction Tmax Start reduction Tmid | 0.000515 3 0.002447 1 0.001921 | 0.000000 1 0.000000 0 0.000000 | 0.007081 3 0.342945 1 0.117678 | 0.008501 3 0.342493 1 0.117712 |
| Wait reduction Tmin Wait reduction Tmin Wait reduction Tmin | 0.013249 3 0.348629 1 0.124184 | 0.006362 3 0.344202 1 0.119528 | 0.008525 3 0.342496 1 0.117730 | 0.000000 4 0.002535 3 0.001278 |
| Start shadow Tmin Start shadow Tmax Start shadow Tmid | 0.002377 1 0.006030 2 0.004001 | 0.000000 1 0.000000 0 0.000000 | 0.000000 4 0.018813 1 0.008302 | 0.002635 4 0.028053 1 0.011899 |
| Wait shadow Tmin Wait shadow Tmax Wait shadow Tmid | 0.004735 4 0.027415 1 0.013747 | 0.000000 2 0.019372 1 0.007180 | 0.002648 4 0.028052 1 0.011899 | 0.000299 3 0.005288 1 0.002509 |
Appendix. The list of characteristics.
Main characteristics and their components
-
Efficiency coefficient (Parallelization efficiency) is ratio of productive time to total processor time.
-
Time of execution (Execution_ time).
-
The number of used processors (Processors).
-
Total processor time (Total_time) is production of the time of execution (Execution_time) by the number of used processors (Processors).
-
Productive time (Productive_time) is the sum of productive processor time (CPU_time), input/output time (I/O_time) and productive system time (Sys CPU).
-
Lost time (Lost_time).
-
Insufficient parallelism (Insufficient_par) and its components.
-
Communications and all components (Communication).
-
Idle (Idle).
-
Dissynchronization losses and all components (Real synchronization).
-
Potential synchronization losses and all components (Synchronization).
-
Potential time variation losses and all components (Time_variation).
-
Overlapping time and all components (Overlap).
-
Imbalance (Load_Imbalance).
Characteristics of program execution on each processor
-
Lost time (Lost time) is the sum of insufficient parallelism losses (User Insufficient_par), system insufficient parallelism losses (Sys Insufficient_par), communications losses (Communication) and idle (Idle).
-
Insufficient parallelism losses (User Insufficient_par).
-
System insufficient parallelism losses (Sys Insufficient_par).
-
Time of losses because of the given processor idle (Idle) is difference between maximal interval execution time (on any processor) and interval execution time on the given processor.
-
Total communication time (Communication).
-
Real time of losses because of dissynchrinization (Real synchronization).
-
Potential time of losses because of dissynchrinization (Synchronization).
-
Potential time of losses because of time variation (Variation).
-
Time of reduction overlapping (Reduction overlap).
-
Time of shadow edges renewing overlap (Shadow renewing overlap).
-
Time of interval execution (Execution_time).
-
Productive processor time (User CPU_time).
-
Productive system time (Sys CPU_time)
-
Input/output time (I/O_time).
-
Losses because of the given processor load imbalance (Load_Imbalance) is difference between maximal time (on any processor) and the time on the given processor.
-
Number of processors used for interval (Processors).
-
Communication times for all types of collective operations (Start_reduction, Wait_reduction, Start_shadow, Wait_shadow, Remote_access, Redistribution и I/O).
-
Real dissynchronization losses for all types of collective operations.
-
Potential dissynchronization losses for all types of collective operations.
-
Potential time variation losses for all types of collective operations














