Improving the Performance of your LabVIEW Applications

Transcription

1 Improving the Performance of your LabVIEW Applications 1

2 Improving Performance in LabVIEW Purpose of Optimization Profiling Tools Memory Optimization Execution Optimization 2

3 Optimization Cycle Benchmark Evaluate performance Identify problem areas Optimize Improve efficiency Improve speed 3

4 Why Should You Profile Your VIs? The 80/20 rule of software performance 80 percent of the execution time is spent in 20 percent of the code Performance improvements are most effective in the 20 percent Guessing which 20 percent is difficult 4

5 Improving Performance in LabVIEW Purpose of Optimization Profiling Tools Memory Optimization Execution Optimization 5

6 Windows Task Manager Gives user a rough idea of whether memory or CPU is the bottleneck Can be helpful in identifying memory leaks View» Select Columns allows you to add additional stats 6

7 Perfmon Allows you to monitor Processors Disk I/O Network Tx/Rx Memory/Paging Access by typing perfmon into the Windows Run dialog 7

8 Benchmarking Code Execution Timing Template LabVIEW Shipping Example 8

9 Benchmarking Code Execution Code Calibration Analysis Benchmark Project LabVIEW Real-Time Shipping Example 9

10 VI Profiler Timing and memory statistics for VIs Tools» Profile» Performance and Memory 10

11 Demo VI Profiling 11

12 LabVIEW Execution Trace Tool Detailed execution traces Thread and VI views Measurement of execution time Priorities Multiple Sessions Threads VIs Time 12

13 Benchmarking Summary OS Level Task Manager, Perfmon LabVIEW Level VI Profiler, Benchmark VIs VI Level Execution Trace Tool 13

14 Improving Performance in LabVIEW Purpose of Optimization Profiling Tools Memory Optimization Execution Optimization 14

15 VI Components Panel Diagram Code Diagram compiled into machine code Data Control/indicator values Default data Block diagram constant data And so on 15

16 VIs in Memory When a VI is loaded into memory We always load the data We load the code if it matches our platform (x86 Windows, x86 Linux, x86 Mac, PowerPC Mac) We load the panel and diagram only if we need to (for instance, we need to recompile the VI) 16

17 Panel and Diagram Data Front panel controls and indicators need their own copy of the data to display, called operate data This VI uses about 8 KB of data if the panel is open, and about 4 KB otherwise 17

18 LabVIEW Terminology Panel data is called Operate Data Controls and indicators have their own copy of the data, so that front panel editing of data doesn t interfere with computations in the diagram Diagram data is called Execution Data Every wire represents a buffer of data Transfer Data is used to copy between them Avoids multithreading issues 18

19 Memory Performance The following features affect the impact memory allocation has on your application: Wire Semantics and In Placeness Type coercion Passing values to subvis Shift registers Front panel indicators 19

20 Wire Semantics Every wire is a buffer Branches create copies 20

21 Optimizations by LabVIEW The theoretical 5 copies become 1 copy operation. Copy Output is inplace with input 21

22 The In Place Algorithm Determines when a copy needs to be made Weights arrays and clusters higher than other types Algorithm runs before execution Does not know the size of an array or cluster Relies on the sequential aspects of the program Branches might require copies 22

23 Bottom Up In place information is propagated bottom up Branched wire Copy because of increment No copies required Increments array in place 23

24 Showing Buffer Allocations 24

25 The In Place Element Structure Allows you to explicitly modify data in place 25

26 Four types of border nodes Array index/replace Unbundle/rebundle Variant to/from G data Simple in place relationship 26

27 Example of In Place Optimization Operate on each element of an array of waveforms 27

28 Make the first SubVI in place changes into 28

29 SubVI 2 is made in place changes into 29

30 SubVI 3 is made in place changes into 30

31 Final Result: Dots are Hidden 31

32 Type Coercion Changing the data type of a wire to match the required data type Dots indicate automatic coercion Requires a copy Coercion Dot 32

33 Passing Values to SubVIs SubVI does not have to make a copy Best when Inputs are required Inputs and outputs are not in structures Inputs and outputs are in place 33

34 Passing Values to SubVIs Parameters within structures cause unnecessary copies Move 34

35 Shift Registers Single left input always inplace with right output Even better when body inplace Inplace Always Inplace 35

36 Building Arrays There are a number of ways to build arrays and some are better than others Bad Reallocates array memory on every loop iteration No compile time optimization 36

37 Building Arrays There are a number of ways to build arrays and some are better than others Better Memory allocated Keep graphics at load time Values populated below the in text place and to the right. 37

38 Building Arrays There are a number of ways to build arrays and some are better than others Best Memory preallocated Indexing tunnel eliminates need for copies 38

39 Demo Effects of Memory Optimization 39

40 Improving Performance in LabVIEW Purpose of Optimization Profiling Tools Memory Optimization Execution Optimization 40

41 Code Execution Code Execution What LV Does For You What You Can Do Execution Systems Threading Constant Folding Using Primitives UI Updating Execution Properties Reentrant VIs 41

42 VIs Are Compiled

43 VIs Are Compiled: Clumps Clump 0 Clump 1 Clump 0 Clump 2 43

44 VIs Are Compiled: Clumps Clump 0 Clump 0 Start of diagram: Reads controls, then schedules Clumps 1 and 2 Then sleeps... Clump 1 Top For Loop Indicator is updated Clump 0 Scheduled Sleep... Clump 2 Clump 0 Sleeping Bottom For Loop Indicator is updated Clump 0 Scheduled Sleep... Clump 1 Sleeping Completion of diagram: Divide nodes, display of indicators, then VI exit. Clump 2 Sleeping 44

45 Single Threaded LabVIEW CPU Thread User Interface Loop co-routines Code Execution 45

46 Multi-Threaded LabVIEW CPU Thread messages Thread Thread UI Loop Exec Thread Exec Exec Thread Exec 46

47 Multi-Core LabVIEW Thread Thread Thread UI Loop Exec Exec CPU0 CPU1 messages Thread Thread Exec Exec 47

48 Constant Folding LabVIEW compiler replaces unnecessary code with a constant. This code always results in the same array and does not need to be computed at run time. Tools» Options» Block Diagram» Show Constant Folding of Wires/Structures 48

49 Avoid Unnecessary Code in Loops Only put code that needs to be repeated in loops Opening/closing references should be outside the loop 49

50 Using Primitives Primitives are faster than chained operations. Is less efficient than 50

51 UI Updating Most overlooked performance bottleneck Not as obvious Like memory management, LabVIEW tries to optimize UI drawing but it can not perform miracles 51

52 UI Thread Front panel updates occur in the UI thread Execution takes place in other threads Shared data must be protected LabVIEW creates an extra copy called a transfer buffer Thread Thread Thread UI Loop Exec Exec 52

53 UI Management Only update indicators as often as you need Usually more than 10 times per second is excessive Data is copied from execution buffer to transfer buffer from transfer buffer to the indicator s operate data 53

54 UI Property Node Effect on Performance Control and Indicator Property nodes are slow Require thread swap Cannot be parallel Many trigger UI redraw UI Property nodes in an un-throttled loop are bad The Value property has the performance characteristics of a UI Property Node, not a terminal 54

55 Defer Panel Updates When performing multiple control property changes on a graph, use Defer Panel Updates 55

56 Execution Properties File» VI Properties» Execution <CTRL-I> 56

57 Reentrant VIs Reentrancy allows one subvi to be called simultaneously from different places Requires extra memory for each instance Use reentrant VIs in two different cases To allow a subvi to be called in parallel To allow a subvi instance to maintain it s own state 57

58 SubVI Calls SubVI calls are relatively inexpensive There is some overhead Subroutine priority is most useful for small VIs which are called often Calling a subvi in a different execution system can be very expensive However, calling a subvi does not inherently cause new memory copies 58

59 Demo Effects of Execution Optimization 59

60 Improving Performance in LabVIEW Purpose of Optimization Profiling Tools Memory Optimization Execution Optimization Decrease memory usage, increase speed 80/20 Rule 60

61 Improving Performance in LabVIEW Purpose of Optimization Profiling Tools Memory Optimization Execution Optimization Windows Tools (Task Manager, perfmon) VI Profiler Real-Time Execution Trace Toolkit 61

62 Improving Performance in LabVIEW Purpose of Optimization Profiling Tools Memory Optimization Execution Optimization Hide the dots Take advantage of in placeness 62

63 Improving Performance in LabVIEW Purpose of Optimization Profiling Tools Memory Optimization Execution Optimization Know where LabVIEW helps you Primitives are generally faster Limit screen updates Use appropriate execution settings 63

64 Next Steps In LabVIEW LabVIEW Help On the Web ni.com/multicore ni.com/devzone 64