Raspberry Pi Forums

matherp

The big issue with the PicoW is heap usage, for example I implemented TLS in MMBasic and had to allow a heap of over 32K to get it to work - I've since removed it because TLS without SSL is pretty much useless and I can't imagine how much memory SSL on top of TLS would need. So, in answer to the que...

matherp

Doesn't help as there is no way of knowing how and if my application is calling a specific instance of malloc. Does CDC use malloc? Timer alarms seem to but what do I need to avoid calling or using to not invoke malloc. Does stdio_init_all() do anything that uses malloc? etc. I, and others, have ask...

matherp

I've previously asked what, if anything, in the sdk uses malloc (ignoring PicoW functionality) and got no answer. Looking at the map for my application malloc uses over 1K of precious RAM for data structures .data.__malloc_sbrk_base 0x2000d5e0 0x4 c:/progra~2/armgnu~1/125863~1.3re/bin/../lib/gcc/arm...

matherp

I was seeing similar after SDK upgrade from 1.5 to 1.5.1 and have reverted to 1.5 to fix the problem. If you are running 1.5.1 it might be worth trying the same

matherp

All working in MMBasic see https://www.thebackshed.com/forum/ViewT ... &TID=16391

matherp

Before I try and re-invent any wheels, has anyone used the PicoW to read stream data such as internet radio. I can open the radio site as a TCP client (many of them use port 80) and send the command to start the stream. I've also got the PicoW wired to a VS1053 and can send MP3 files to it for playb...

matherp

MMBasic runs the screen at RGB121. This works very nicely as you get 1 nibble per pixel and a framebuffer is only 38400 bytes (320x240). This also allows clever things like 2 display layers with transparency of one over the other. Have a look at https://www.youtube.com/watch?v=iTqL_GKOkGM and other ...

matherp

Check the cmake file and main function in picomite.c on GitHub uktailwind picomite for how to underclock the flash which allows overclocking the CPU. Mmbasic runs stably at 378MHz on nearly all Pico clones and all genuine Picos using this technique

matherp

If you have a system clock, when you acquire the mutex, write the system time into your variable, so other processes can see how long you've had the mutex and maybe guess when it will be released. That's a really nice idea. I can even have the main blocking task on the 2nd CPU "learn" how...

matherp

Such functionality is pointless, I have some code that several layers down after quite a lot of processing will acquire the mutex in a standard blocked fashion. I don't want to acquire the mutex at the top level as that is wasteful and holds it for longer than necessary nor do I want to do all the ...

matherp

Thanks - that is helpful. It would be good though to have an example program in the examples download. There doesn't seem to be a mutex test without trying to acquire the lock. I understand that getting the lock has to be an atomic transaction hence mutex_try_enter() but what I really want is a mute...

matherp

I need to control access to SPI between the two cores. It is perfectly OK in my application for one core to block while the other is using the SPI. So I need a lock that I try to acquire and block until I get it, do my SPI work, and then release the block. I know this should be easy but I can't find...

matherp

How do I use systick on the second core? Do I just refer to it as systick_hw-> or is there a different name for the second core? so: uint32_t delaytime= 0x7FFFFF systick_hw->csr = 0x5; systick_hw->rvr = 0xFFFFFF; systick_hw->cvr=0; while(1){ do { if (multicore_fifo_rvalid()){ multicore_fifo_pop_bloc...

matherp

The issue is caused because I set the core1 stack size to zero. Obviously I want it t be zero so I don't waste memory as I am specifically allocating stack for core1. If I set it to 0x10 the warning goes away and the linker seems to optimize out core1 stack usage. Bizarrely the errors also disappear...

matherp

I'm compiling with -Wall set (GCC 12.3.1 arm-none eabi) and have eliminated all compiler warnings except as follows: Is this a compiler quirk or am I missing something? Thanks my code excerpt uint32_t core1stack[64]; multicore_launch_core1_with_stack(QVgaCore, core1stack, 256); sdk 1.5.0 void multic...

matherp

I look forward to your completed code. This is an area where we are often let down by supplied libraries. I had to do something similar with the STM32H7 and achieved massive improvements using some special floating point registers which can load and store 128 bytes at a time. The other trick I used ...

matherp

Thanks all for the useful input.

I actually solved it before kilograham's post using my own "scratch" resister

unsigned int _excep_code __attribute__ ((persistent));

which I set to a given status when a watchdog had been enabled.

Duh.. Re-inventing wheels

matherp

When you program a RP2040 using the usual uf2 download the user program sees that a watchdog timeout caused the reboot - watchdog_caused_reboot() returns true.
Is there any way to distinguish this from a "genuine" watchdog timeout?

matherp

Can't help with the scratch registers but I think this does what you want

Code: Select all

uint32_t __uninitialized_ram(mycount);
if(watchdog_caused_reboot())mycount++;
else mycount=0;

matherp

My suspicion would be that the change isn't in the time code but is in the web support - the issues only affect the web version of MMBasic. My best guess FWIW is that something is disabling interrupts but that there is a route out of the routine that exits without re-enabling them

matherp

The hex figure should be 0x7FFFF/1000 = 0x20C4 = 8388

matherp

Parameters delay_ms Number of milliseconds before watchdog will reboot without watchdog_update being called. Maximum of 0x7fffff, which is approximately 8.3 seconds This doesn't compute 0x7fffff milliseconds is more than 8300 seconds. The statement above occurs in more than one place. 8.3 seconds i...

matherp

Thought I'd bump this thread as there is definitely some issue that I can't see anyway of diagnosing but has definitely been introduced in 1.5.1 MMBasic runs on the PicoW and Geoff Graham has produced a brilliant web-interfaced watering controller based on it See https://geoffg.net/retic.html for wh...

matherp

Thank you

matherp

Sorry I'm a github beginner.

Please can someone give me the full github commands to download 1.5 rather than 1.5.1.

This needs to be a full download self-consistent with all the various relevant addins (TinyUSB version)

Thanks

Search found 288 matches

Re: Memory Usage of Pico W

Re: Getting rid of malloc?

Getting rid of malloc?

Re: Pico W freezes regularly for about 1h

Re: PicoW - handling stream data (internet radio)

PicoW - handling stream data (internet radio)

Re: VGA but with fewer colors

Re: UART issues with overclocked Pico

Re: spinlocks, mutexes, or semaphores how to use?

Re: spinlocks, mutexes, or semaphores how to use?

Re: spinlocks, mutexes, or semaphores how to use?

spinlocks, mutexes, or semaphores how to use?

Using systick on the second core

Re: Compiler warnings in multicore_launch_core1_with_stack

Compiler warnings in multicore_launch_core1_with_stack

Re: memcpy replacement for ARM M0+ core

Re: Any way to distinguish HW watchdog from just programmed?

Any way to distinguish HW watchdog from just programmed?

Re: RP2040 watchdog and Scratch registers

Re: 1.5.1 broken?

Re: Watchdog minor documentation error

Watchdog minor documentation error

Re: 1.5.1 broken?

Re: Downloading SDK 1.5

Downloading SDK 1.5