Skip to content

RP2040 multi-core module

lbuild module: modm:platform:multicore

This module adds the ability to run code on CPU1 and provides mailbox, spinlock and mutex implementations, and additionally a stack of size modm:platform:cortex-m:main_stack_size is added to the CORE1 memory and the modm::atomic::Lock is made multicore safe. You can use attributes for placing objects into core-coupled memories:

modm_core0_data uint8_t data = 0xab;
modm_core1_bss uint16_t zero;
modm_core0_noinit uint32_t uninitialized;

void function()
    data = 0xcd;
    zero = 100;
    uninitialized = 200;

We provide a symmetric multiprocessing (SMP) abstraction, where the same binary is used for both cores. By default CPU0 is used to boot the device and code running on CPU1 needs to be explicitly started:

void function()
    uint8_t cpuid = Core::cpuId();
function(); // cpuid == 0
multicore::Core1::run(function); // cpuid == 1
// Before CPU1 executes code, this function is called
// so that you can customize the CPU1 environment
extern "C" void modm_initialize_core1()
    // set a custom vector table for CPU1
    SCB->VTOR = custom_vector_table;

When using the modm:processing:fiber module you can now schedule fibers on both cores, each with their own scheduler:

modm_core0_noinit modm::fiber::Stack<256> stack0;
modm_core0_data modm::Fiber fiber0(stack0, []() { function(); }, /* core= */0);

modm_core1_noinit modm::fiber::Stack<256> stack1;
modm_core1_data modm::Fiber fiber1(stack1, []() { function(); }, /* core= */1);

This module is only available for rp2040.

Limited availability: Check with 'lbuild discover' if this module is available for your target!