The dedicated GPIO is designed for CPU interaction with GPIO matrix and IO MUX. Any GPIO that is configured as "dedicated" can be access by CPU instructions directly, which makes it easy to achieve a high GPIO flip speed, and simulate serial/parallel interface in a bit-banging way. As toggling a GPIO in this "CPU Dedicated" way costs few overhead, it would be great for cases like performance measurement using an oscilloscope.
A GPIO bundle is a group of GPIOs, which can be manipulated at the same time in one CPU cycle. The maximal number of GPIOs that a bundle can contain is limited by each CPU. What's more, the GPIO bundle has a strong relevance to the CPU which it derives from. **Any operations on the GPIO bundle should be put inside a task which is running on the same CPU core to the GPIO bundle belongs to.** Likewise, only those ISRs who are installed on the same CPU core are allowed to do operations on that GPIO bundle.
Dedicated GPIO is more of a CPU peripheral, so it has a strong relationship with CPU core. It's highly recommended to install and operate GPIO bundle in a pin-to-core task. For example, if GPIOA is connected to CPU0, and the dedicated GPIO instruction is issued from CPU1, then it's impossible to control GPIOA.
To install a GPIO bundle, one needs to call :cpp:func:`dedic_gpio_new_bundle` to allocate the software resources and connect the dedicated channels to user selected GPIOs. Configurations for a GPIO bundle are covered in :cpp:type:`dedic_gpio_bundle_config_t` structure:
-:cpp:member:`gpio_array`: An array that contains GPIO number.
-:cpp:member:`array_size`: Element number of :cpp:member:`gpio_array`.
-:cpp:member:`flags`: Extra flags to control the behavior of GPIO Bundle.
-:cpp:member:`in_en` and :cpp:member:`out_en` are used to select whether to enable the input and output function (note, they can be enabled together).
-:cpp:member:`in_invert` and :cpp:member:`out_invert` are used to select whether to invert the GPIO signal.
The following code shows how to install a output only GPIO bundle:
..highlight:: c
::
// configure GPIO
const int bundleA_gpios[] = {0, 1};
gpio_config_t io_conf = {
.mode = GPIO_MODE_OUTPUT,
};
for (int i = 0; i < sizeof(bundleA_gpios) / sizeof(bundleA_gpios[0]); i++) {
:cpp:func:`dedic_gpio_new_bundle` doesn't cover any GPIO pad configuration (e.g., pull up/down, drive ability, output/input enable), so before installing a dedicated GPIO bundle, you have to configure the GPIO separately using GPIO driver API (e.g., :cpp:func:`gpio_config`). For more information about GPIO driver, please refer to :doc:`GPIO API Reference <gpio>`.
Using the above functions might not get a high GPIO flip speed because of the overhead of function calls and the bit operations involved inside. Users can try :ref:`manipulate_gpios_by_writing_assembly_code` instead to reduce the overhead but should take care of the thread safety by themselves.
Code examples for manipulating dedicated GPIOs from assembly are provided in the :example:`peripherals/dedicated_gpio` directory of ESP-IDF examples. These examples show how to emulate a UART, an I2C and an SPI bus in assembly thanks to dedicated GPIOs.
Writing assembly code in application could make your code hard to port between targets, because those customized instructions are not guaranteed to remain the same format on different targets.
Dedicated GPIO can also trigger interrupt on specific input event. All supported events are defined in :cpp:type:`dedic_gpio_intr_type_t`.
One can enable and register interrupt callback by calling :cpp:func:`dedic_gpio_bundle_set_interrupt_and_callback`. The prototype of the callback function is defined in :cpp:type:`dedic_gpio_isr_callback_t`. Keep in mind, the callback should return true if there's some high priority task woken up.