Trice consists of 2 parts: Tiny & super fast embedded device real-time trace C code (TRICE
macros) and PC tool trice
for managing and visualization. The aim is to replace printf
in a conveinient and much faster way. The PC trice tool itself is written in Go and therefore usable on all platforms Go supports.
It is also easy to receive the COBS packges, exchange the IDs with the format string and to print out in your own environment if you want.
- Copy 3 files to your embedded project:
./pkg/src/trice.h
./pkg/src/trice.c
./test/.../triceConfig.h
- In your source.c:
#include "trice.h"
- In a function:
TRICE( "Coming soon: %d!\n", 2022 );
- In project root:
- Create empty file:
touch til.json
. - Run
trice u
should:- patch source.c to
TRICE( Id(12345), "Coming soon: %d!\n", 2022 );
- extend
til.json
- patch source.c to
- Create empty file:
- Modify
triceConfig.h
acording your needs.- With
#define TRICE_MODE 0
(direct mode) just provide a putchar() function. - Recommended is an indirect mode which allows to use
TRICE
macros also inside interrupts.
- With
- Compile & load your app.
- In project root:
trice l -p COM1
should showComing soon: 2022!
after app start. - Look in
./pkg/src/triceCheck.c
for examples.
- In release v0.41.0 now the
TRICE
macro works additionally. To use it, simply use it likeprintf
:- No need for parameter count and bit width.
- The internal used parameter bit width is 32 bit, but you can use also
TRICE8
,TRICE16
,TRICE32
,TRICE64
, - 0 to 12 parameters possible (extendable).
- The internal used parameter bit width is 32 bit, but you can use also
- No strings supported ("%s"). Use
TRICE_S
than. - Many usage options inside
pkg/src/triceCheck.c
visible.
- No need for parameter count and bit width.
- Needs better tests and updated documentation.
In release v0.39.0 now encryption works again. To implement it well and open for future, the additional COBS package descriptor is now 4 bytes long. That means the trice tool version 0.39.0 does not work with older target code. Please update your target code or stay with an older release. Probably the COBS encoding will not change in the next time anymore.
In release v0.38.0 now target timestamps possible. To implement it well and open for future, an additional COBS package descriptor byte was added. That means the trice tool version 0.38.0 does not work with older target code. Please update your target code or stay with an older release.
The TRICE technique changed heavily between release 0.33.0 and 0.34.0. The flex
and esc
encodings are replaced by a COBS encoding which will be the default now. The stuff works already well but is not in its final state and is not documented vet. It lacks also automated tests. The internal speed goes to its limit (~6 clocks per trice on M0+ possible) by using a double buffer instead of a fifo. Also porting is easier now. The documentation is outdated but gets updated soon. But first the tests. If you have a project with flex
or esc
encoding, please update the target code or stay with version 0.33.0.
- Printf-like trace macros
TRICE
and PCtrice
tool (written in Go) for automatic ID managing & logging. - Communication without string transfer, just with IDs. Prerequisite: byte transmission to PC, low bandwidth is ok:
- "log in (a) trice" (S>G)
- Main idea: Logging strings not into an embedded device to display them later on a PC but keep usage comfortable and simple.
- Real fast: 12 CPU clocks per (short) trice possible!!!
- With a 48MHz clock this is 250ns. Light travels about 80 meters in that time.
- TRICE in your code reduces the needed FLASH memory because the instrumentation code is very small (can be less 200 bytes FLASH and about 100 bytes RAM) and no printf library code nor log strings are inside the embedded device anymore.
- Using trice not only for dynamic debugging but also as logging technique
is possible and gives the advantage to have very short messages (no strings) for transmission,
but keep in mind that the file til.json is the key to read all output if your devices in the field for 10 or more years.
- Optionally add til.json as a (compressed) resource.
- You can see TRICE also as a kind of data compression what could be interesting for IoT things, especially NB-IoT, where you have very low data rates.
- Storing trices in FLASH for later log analysis saves memory because a typical
TRICE
occupies only 4 or 8 bytes. - Also it is possible to encrypt the trice transfer packets to get a reasonable protection for many cases.
- This way you can deliver firmware images with encrypted TRICE output only readable with the appropriate key and til.json.
- XTEA is implemented as one option.
- You can even translate the til.json in different languages, so changing a language is just changing the til.json file.
- Using trice with an RTOS gives the option for detailed task timing analysis. Because of the very short execution time of a trice you could add to the scheduler:
Trice16i( "tim:@tick %5u ", clock );
Trice8i( "sig:task %u -> %u\n", previousTaskID, nexTaskID );
The execution of this code block produces totally 8 log bytes to visualize the output on PC, what looks similar to this for 3 task switches:
First are the PC reception timestamps and after the port info are the used trice ids just for easy location inside the source code. See the diferences between the (blue) ticks in this 3 lines. These are 28 or 36 processor clocks only. The code producing this is:
The same is possible for interrupt timing analysis.
- Mixed case trice macros are short trices and the letter i at the end says inside critical section. (FLEX encoding)
Trice16( "tim: myFunc %d\n", sysTick );
before and after a function call lets you easy measure the function execution time.- As graphical visualization you could use a tool similar to https://github.com/sqshq/sampler.
For example change the source code line
printf( "MSG: %d Kelvin\n", k );
into
Trice16( "MSG: %d Kelvin\n", k );
trice update
(run it automatically in the tool chain) changes it to
Trice16( Id(12345), "MSG: %d Kelvin\n", k );
or (if -addParamCount
is used)
Trice16_1( Id(12345), "MSG: %d Kelvin\n", k );
and adds the ID 12345 together with "MSG: %d Kelvin\n" into a trice ID list, a JSON reference file named til.json.
- The 12345 is a randomly or policy generated ID not used so far.
- With the
16
in Trice16 you adjust the parameter size to 16 bit what allows more runtime efficient code compared to32
or64
. - The optional appended _1 sets the expected parameter count to 1, allowing a compile time parameter count check.
- During compilation the
Trice16[_1]
macro is translated to only a 12345 reference and the variable k. The format string never sees the target.
This is a slightly simplified view:
- When the program flow passes the line
Trice16( Id(12345), "MSG: %d Kelvin\n", k );
the ID 12345 and the 16 bit temperature value are transferred as one combined 32 bit value into the triceFifo, what goes really fast. Different encodings are possible. The program flow is nearly undisturbed, so TRICE macros are usable also inside interrupts or in the scheduler. - For visualization a background service is needed. In the simplest case it is just an UART triggered interrupt for triceFIFO reading. Or you can use RTT.
- So the whole target instrumentation are the trice macros, the trice fifo and the UART ISR.
- During runtime the PC trice tool receives the trice as a 4 byte package
0x30 0x39 0x00 0x0e
from the UART port. - The
0x30 0x39
is the ID 12345 and a map lookup delivers the format string "MSG: %d Kelvin\n" and also the format information "TRICE16_1". Now the trice tool is able to executeprintf("MSG: %d Kelvin\n", 0x000e);
and the full log information is displayed in the MSG color. - Only the parameter count and size affect encoding size but not the format string length.
- Manages
TRICE
macro IDs inside a C or C++ source tree and extracts the strings in an ID-string list during target device compile time. - Displays
TRICE
macros like printf() output in real-time during target device runtime. The received IDs and parameters are printed out. - Can receive trices on several PCs and display them on a remote display server.
- Written in Go, simply usage, no installer, needs to be in $PATH.
Right now only event logging is implemented.
According to the design aim "Keep embedded device code small and fast" there is no structuring code inside the target device, but you can add channel information to the trice log strings:
trice32( Id(12345), "Verbose: bla bla")
These can be understood as tags too. But only one tag per trice right now. Look into lineTransformerANSI.go for options or extensions.
Also you can at compile time disable trice code generation on file level with #define TRICE_OFF
before including trice.h
.
Because one trice consists typically only of 4 to 8 bytes there is usually no need to dynamically switch trices on and off inside the embedded device. This can be done on the display side inside the trice tool with the command line switches -ban
or -pick
. For example -pick err,wrn
disables all output despite error and warning messages.
Switching trices on and off inside the target increases the overhead and demands some kind of command interface.
If needed, always an if
is usable.
The trice tool can also perform further tasks like JSON encoding with additional log information and transferring this information to some webserver in the future.
Yes, you can simply start trice ds
inside a console, option: third_party/alacritty, locally or on a remote PC and connect with several trice tool instances like with trice log -p COM15 -ds
for example.
- Of course
git
, but it is not forbidden to compile til.json as a resource into the embedded device and get it later back if you have enough flash memory.
- Get trice or download latest release assets for your system: Source code and compressed binaries.
- A port to Darwin should be easy possible.
- Place the extracted
trice
binary somewhere in your $PATH.
- Install Go.
- On Windows you need to install TDM-GCC - recommendation: Minimal online installer.
- GCC is only needed for ./pkg/src/src.go, what gives the option to test the C-code on the host.
- Make sure TDM-GCC is found first in the path.
- Other gcc variants could work also but not tested.
- Open a console inside the
trice
directory. - Check and install:
go vet ./...
go test ./...
go install ./...
Afterwards you should find an executable trice
inside $GOPATH/bin/
trice help
- It is sufficient for most cases just to use the
trice32
macro with max 4 parameters as a replacement forprintf
and to use the default settings. - Compare the not instrumented test project MDK-ARM_LL_generatedDemo_STM32F030R8-NUCLEO-64 with one of the instrumented test projects in test to see what to to.
- Recommendation: FLEX encoding
- Or follow these steps for instrumentation information even your target processor is not an ARM (any bit width will do):
- Install the free STCubeMX.
- Choose from test examples the for you best fitting project
MyExample
. - Open the
MyExample.ioc
file with STCubeMX and generate without changing any setting. - Make an empty directory
MyProject
inside thetest
folder and copy theMyExample.ioc
there and rename it toMyProject.ioc
. - Open
MyProject.ioc
with STCubeMX, change in projects settingsMyExample
toMyProject
and generate. - Now compare the directories
MyExample
andMyProject
to see the trice instrumentation as differences.
- For compiler adaption see triceConfigCompiler.h.
- For hardware adaption see triceUART_LL_STM32
No need to read all this stuff - is is just for help and reference.
- fix color issues under windows
- Command Line Examples
- Common.md
- TriceEncodings.md
- ID management
- OneWireOption
- SeggerRTT
Yes please: May be you create a graphical display server, have a cool idea, a port to other hardware, some correction or simply like to ⭐ it. ☺
git clone https://github.com/rokath/trice.git
Maybe you find this project interesting too: baical.net