ThreadX Porting #292
Replies: 7 comments 28 replies
-
|
for other platform, arduino core will use systick and provide a callback symbol but i couldn't find in arduino-pico. |
Beta Was this translation helpful? Give feedback.
-
|
The board package has included all required toolchain to build pico-sdk, right? I got below errors |
Beta Was this translation helpful? Give feedback.
-
|
You really need to use a GIT copy. The Arduino IDE installs the binaries in a completely different folder. Do the git clone and run |
Beta Was this translation helpful? Give feedback.
-
|
follow below steps i can get everything in |
Beta Was this translation helpful? Give feedback.
-
|
Hi. I'm still looking for a solution for this: ) |
Beta Was this translation helpful? Give feedback.
-
|
For Windows, the path is actually When placed there, restarting the Arduino IDE should make it appear. You can also add in the cmsys_core to the cmake file and rebuild before running the IDE to get the new symbols right off the bat. |
Beta Was this translation helpful? Give feedback.
-
|
Sorry, back to this topic after one year, I still want to add ThreadX to Arduino and RP2040. I just realized the pico-SDK was built into a static library in Arduino, what I need is to takeover systick and pendsv handler and override them with OS port defined in .S file, whether they are cmiss-compatible symbols (need link cmsis_core interface library and this is preferred to keep minimal code change) or private symbol (use isr_systick/isr_pendsv) in ctr0.S Any suggestions? |
Beta Was this translation helpful? Give feedback.
-
|
More inputs:
last but not least, i think threadX should not work with FreeRTOS enabled, how can I turn off/on FreeRTOS in configuration? |
Beta Was this translation helpful? Give feedback.
-
FreeRTOS already overrides this in SW, (see their rp2040 port.c)
I'm fine w/a PR to support this and it's pretty trivial to do, yes? In fact, we overrive/ignore depending on the FreeRTOS linking state so the same logic could be done here. See the
It's already off by default. The user needs to manually add |
Beta Was this translation helpful? Give feedback.
-
|
Thanks.
|
Beta Was this translation helpful? Give feedback.
-
|
For the Systick is used to give a CPU cycle count, which is useful for profiling. There is a fallback PIO based counter, used by FreeRTOS, which you can use. Look at the |
Beta Was this translation helpful? Give feedback.
-
|
Thanks again. I was able to call the SDK API to register PendSV in the low level initialization function in ASM. I checked the right function address is placed into ram_vector_table[14], which is used by NVIC as vector. So I think PendSV will be fine. PUSH {r0, lr}
LDR r0, =-2
LDR r1, =PendSV_Handler
BL exception_set_exclusive_handler
POP {r0, r1}
MOV lr, r1I'm still trying to figure out how to do with Systick. I think my project will not compatible with FreeRTOS, and as you said, by default it is off. So PIO is not used. You have below code to register a callback if (!__isFreeRTOS) {
// Enable SYSTICK exception
exception_set_exclusive_handler(SYSTICK_EXCEPTION, _SystickHandler);
systick_hw->csr = 0x7;
systick_hw->rvr = 0x00FFFFFF;
} else {private:
static void _SystickHandler() {
rp2040._epoch += 1LL << 24;
}Can I override this systick and will system works just fine? Because when I register my callback to replace yours, my rp2040 looks dead and USB is not funcitoning, I try to add some blink code in my task and it seems ThreadX multitask environement is not running. Or I should not override your _SystickHandler? It's quite hard to debug in arduino environment without a debugger. |
Beta Was this translation helpful? Give feedback.
-
|
Just FYI. my port of ThreadX is verified and works on a single cortex-m0 configuration on rp2040. |
Beta Was this translation helpful? Give feedback.
-
|
Cool! I think the systick stuff is simpler than you think. Add a bool Then at arduino-pico/cores/rp2040/RP2040Support.h Line 224 in 98c4b92 arduino-pico/cores/rp2040/RP2040Support.h Line 252 in 98c4b92 arduino-pico/cores/rp2040/RP2040Support.h Line 266 in 98c4b92 __isFreeRTOS || __isThreadX and it will still give cycle counts using the PIO hw.
|
Beta Was this translation helpful? Give feedback.
-
|
Thanks for your guidance. I was able to get ThreadX environment running on top of Arduino-pico platform under specific configuration/modification. I also observed some strange behaviors that need further debug/discuss.
class RP2040 {
public:
RP2040() { /* noop */ }
~RP2040() { /* noop */ }
void begin() {
_epoch = 0;
int off = 0;
_ccountPgm = new PIOProgram(&ccount_program);
_ccountPgm->prepare(&_pio, &_sm, &off);
ccount_program_init(_pio, _sm, off);
pio_sm_set_enabled(_pio, _sm, true);
}
......
inline uint32_t getCycleCount() {
return ccount_read(_pio, _sm);
}
inline uint64_t getCycleCount64() {
return ccount_read(_pio, _sm);
}
void setup()
{
pinMode(LED_BUILTIN, OUTPUT); // If I move this line, USB is not recognized by windows, and board seems dead.
Serial.begin(115200);
Serial.print("\r\nArduino Environment is Up\r\n");
/* Enter the ThreadX kernel. */
tx_kernel_enter();
}The library I made for testing is here, appreciate if you can have a try.. |
Beta Was this translation helpful? Give feedback.
-
|
Thank you I fully understand and agree with your comments.. These are further steps need to be done to deeply integrate with (any) arduino platform code to support all system-level features running a RTOS alongside. Current my plan with the ThreadX arduino library is generic and only provide basic CPU time slicing capabilities to use multithreading. So the first thing need to do is at least IT IS WORKING with minimum features, then I can see how much can invest on more support. Now after move USB Serial initlization code into the initialization task (rather than in main), everythings works fine and stable now. What I was asking about using PIO and avoid to occupy Systick in the core, is that good for you to make such a change? Any side effect you would be concern about? After this I can make first release and other may leverage the port to have engagement in the maker community just like this: |
Beta Was this translation helpful? Give feedback.
-
|
Understood, but I would warn you that leaving those changed undone will result in very hard to debug problems later on (believe me, I know since I did the same thing for FreeRTOS and spent quite a while tracking things down!) For SysTick, I really do not want to dedicate a PIO unit to a cycle counter. There is a limited number of PIO units and program memory, and there are some people using the SW USB that need every single word of PIO memory already. Systick is there and unused in the main SDK, and there's a simple way to fallback to PIO mode if systick is used elsewhere (like in FreeRTOS). |
Beta Was this translation helpful? Give feedback.
-
|
I'm looking into other options.
Here are modifications in void begin() {
_epoch = 0;
if (!(__isFreeRTOS || __isAzureRTOS)) {
// Enable SYSTICK exception
exception_set_exclusive_handler(SYSTICK_EXCEPTION, _SystickHandler);
systick_hw->csr = 0x7;
systick_hw->rvr = 0x00FFFFFF;
} else { inline uint32_t getCycleCount() {
if (!(__isFreeRTOS || __isAzureRTOS)) {
uint32_t epoch;
uint32_t ctr;
do {
epoch = (uint32_t)_epoch;
ctr = systick_hw->cvr;
} while (epoch != (uint32_t)_epoch);
return epoch + (1 << 24) - ctr; /* CTR counts down from 1<<24-1 */
} else {
return ccount_read(_pio, _sm);
}
}
inline uint64_t getCycleCount64() {
if (!(__isFreeRTOS || __isAzureRTOS)) {
uint64_t epoch;
uint64_t ctr;
do {
epoch = _epoch;
ctr = systick_hw->cvr;
} while (epoch != _epoch);
return epoch + (1LL << 24) - ctr;
} else {
return ccount_read(_pio, _sm);
}
}And bool __isAzureRTOS = true; // just after bool __isFreeRTOS;
// Let rest of core know if we're using FreeRTOS
__isFreeRTOS = initFreeRTOS ? true : false;
__isAzureRTOS = true;But if I initialize the variable in definition, everything will be ok. bool __isAzureRTOS = true;I couldn't figure out what's going on there. Any ideas? Btw, I think think systick is more important resources than PIO because using the chip has one only single instance and it is common. |
Beta Was this translation helpful? Give feedback.
-
|
Sorry, for the USB bit you're breaking something and will need to look at GDB or a trace to see what's failing. I'd do the The PIO counter override trivially just starts the PIO program which runs w/o any core usage whatsoever until you manually read it. Basically, I think it all comes down to the USB being busted by the threadx code, but w/o tracing it's hard to guess what's wrong. The FreeRTOS methods in the core for USB setup/run are working well so please implement the same in threadx and I have a feeling you'll be in good shape. |
Beta Was this translation helpful? Give feedback.
-
|
Actually, now that I think about it, you were using a call in the main app That is definitely is going to kill everything. The USB stack is interrupt driven by default and has timing restrictions in some bits. Your OS code is using PendSV to swap tasks, which overrides the interrupt state IIRC, meaning the USB core could be in the process of being processed and then suddenly swapped out. Not only would this be a timing problem, the rest of the core can't use a USB stack that's in an incoherent state so you could end up crashing on a |
Beta Was this translation helpful? Give feedback.
-
|
Oh yes. Actually the when I start ThreadX environmenet in void arduino::serialEventRun(void) {
if (serialEvent && Serial.available()) {
serialEvent();
}
}PendSV should be the least priority configured by the initalization code, so I'm not sure how it will affect USB or other interrupt. In RTOS based system, urgent/frequent/time-sensing interrupt could be run independ of RTOS as long as user do not call RTOS specific API or use RTOS data in that ISR. That's my understanding.. But anyway, I see a deep integration of FreeRTOS in your core, I probably need to think about whether it is worth and how can do this in a similiar way.. But few things want to double check with you before start.
Appreciate! |
Beta Was this translation helpful? Give feedback.
-
If you are taking over control of the processor and interrupt handling and adding threads, you have no choice but to integrate it or it will be not useful at all.
Yes
The FreeRTOS SMP lib included here is the upstream project plus some minor fixes to let it integrate properly and make the rest of the Arduino stuff work properly. I can't imagine how a different threading library doing essentially the same thing could work without the same kinds of changes. At a very basic level, any multithreading breaks a basic assumption of the Pico-SDK mutex. The owner of the mutex is the core ID, but now you can have the same core ID but different processes so you need to not use the SDK mutexes and implement your own, for example. There are other real-world issues, too. |
Beta Was this translation helpful? Give feedback.
-
|
Here's an idea: Since ThreadX isn't SMP safe/supported, why not avoid the whole issue and run ThreadX on That way you basically have your own system with no interrupts, no exceptions, etc. from other code that need to run. Everything running on core1 would be your own code... |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Hi, I'm trying to get threadx support to arduino-pico. I need three symbol to support RTOS port.
I use the value above but it seems the system is not working. Not sure if i'm using the right value.
Beta Was this translation helpful? Give feedback.
All reactions