25.4 Lua Script Performance

There are many ways to measure and improve the performance of a Lua Script running on a T-Series device. A set of benchmarking example scripts are available in the Lua Script Examples and in Kipling.

• The analog input benchmark script runs at approximately 12500 loop iterations per second.

• The DIO benchmark script runs at approximately 16000 loop iterations per second. 

Test Methodology

The performance tests on this page were obtained by measuring the total time it took to run 100000 iterations of a given chunk of Lua code and then dividing that total time by 100,000 to convert to microseconds (μs per iteration). Test results were obtained using a T4 running firmware version 1.0023, default Lua throttling settings, and default processor speed.  A T7-Pro running firmware version 1.0282 produced similar results. Execution times can vary by up to 10% with different firmware versions and device types.

Using local to Improve Lookup Time

Define Variables as local

When performing operations in a scripted language like Lua, it takes time to look up variables depending on their scope. To speed up a script, define variables as local. For example, the average iteration time for the following script was 17.2 µs:

x = 0
for i = 1, 100000 do
    x = x + 1
end

The same script with variable x defined as a local variable took approximately 8.5 µs per iteration:

local x = 0
for i = 1, 100000 do
  x = x + 1
end

Create local References to Functions

It is also faster to use local references to functions. The benchmarking example scripts all define local references to the functions from the Lua LabJack Library, which provides the global MB and LJ objects. For example, the average iteration time for the following script was 23.7 µs:

for i = 1, 100000 do
  LJ.DIO_S_W(5,1)
end

The same script with a local reference to LJ.DIO_S_W as dioStateWrite took approximately 18.1 µs per iteration.

local dioStateWrite = LJ.DIO_S_W
for i = 1, 100000 do
  dioStateWrite(5,1)
end

Adjusting LJ.setLuaThrottle

After a script is successfully downloaded to a device and no RAM errors are encountered, the T-Series device uses its internal Lua compiler to translate (precompile) the source code into a set of instructions similar to machine code but intended for the Lua interpreter. When it is time for the code to be executed, the T-Series device steps through one or more instructions at a time as defined by the LJ.setLuaThrottle function. In practical terms, this setting adjusts the number of Lua machine instruction codes that are executed per service interval. This number can be increased at the expense of potentially preventing other device features from executing as quickly, such as processing packets from a host computer.

The following script took approximately 8.4 µs per iteration when tested:

LJ.setLuaThrottle(10)
local x = 1
for i = 1, 100000 do
  x = x + 1
end

The same loop took approximately 10.8 µs per iteration when the LabJack was configured to execute 100 instructions per service interval:

LJ.setLuaThrottle(100)
local x = 1
for i = 1, 100000 do
  x = x + 1
end

Benchmark Comparison

There are three ways to toggle a digital I/O line using the LabJack Lua Library functions:

• MB.writeName (a Modbus Name Function)

• MB.W (a Modbus Address Function)

• LJ.DIO_S_W (a Shortcut Function)

Each of the three methods are shown with benchmark results below. Note that these benchmark tests use local function references.

MB.writeName - Modbus Name Function

Approximately 24 μs per iteration.

LJ.setLuaThrottle(1000)
local modbus_write_name = MB.writeName
for i = 1, 100000 do
  modbus_write_name("FIO5", 1)
end

MB.W - Modbus Address Function

Approximately 17 μs per iteration.

LJ.setLuaThrottle(1000)
local mbW = MB.W
for i = 1, 100000 do
  mbW(2005, 0, 1)
end

LJ.DIO_S_W - Shortcut Function

Approximately 11 μs per iteration.

LJ.setLuaThrottle(1000)
local dioStateWrite = LJ.DIO_S_W
for i = 1, 100000 do
  dioStateWrite(5,1)
end

Comparison with External Communications

Since T-Series devices implement a Modbus Server that can be controlled directly by Lua or by a host computer, the performance of these functions can be useful to compare to round trip command-response times issued by host computers through the USB, Ethernet, or WiFi communication interfaces.

Further Reading

• Lua Performance Tips: chapter 2 by Roberto Lerusalimschy.