STM32 DMA2D

STM32 DMA2D is also called Chrom-ART. It will be referred to as DMA2D on this page.

What is DMA2D

Many STM32s come with a dedicated hardware peripheral for accelerating pixel data manipulation. It works asynchronously with the CPU. If you know what DMA is, it's similar to a memory-to-memory DMA engine and can even be used like one, but allows specifying source and destination stride and converting the data it's moving from one pixel format to another. It can blend foregrounds together with backgrounds controlled by a constant alpha value or from a pixel alpha channel.

LVGL integrates DMA2D as a draw unit. During rendering, the LVGL DMA2D draw unit will take tasks that it can perform. The ones it cannot perform will be done in parallel by the software draw unit. The draw tasks that the LVGL DMA2D draw unit can render are fills and images. In general, the DMA2D draw unit can handle fills and images that use standard color formats, no gradient, no transforms, and no radius.

Usage

To use DMA2D in your project now, set LV_USE_DRAW_DMA2D to 1 in lv_conf.h. You will need to specify the header for LVGL to include internally for DMA2D definitions. Set LV_DRAW_DMA2D_HAL_INCLUDE to the corresponding header. E.g., if your STM32 model is an STM32U5, the header name will likely be "stm32u5xx_hal.h". If you're using a framework, ensure it will not be in contention with LVGL over the DMA2D peripheral. LVGL uses direct memory register access to manipulate the DMA2D peripheral. It does not use the STM32 HAL.

Nothing more is required for DMA2D to start working immediately, but you may want to make some other considerations.

Since DMA2D is implemented in LVGL as a draw unit, it can work in parallel with other draw units (typically only the software draw unit). By default, any time that DMA2D is working on a draw task that LVGL depends on the completion of before continuing, LVGL will spin waiting for DMA2D to complete. If you would like the CPU to sleep or be scheduled to other RTOS tasks while a DMA2D transfer is ongoing, do the following:

  1. An RTOS/OS must be present. Set LV_USE_OS to one of the supported values in lv_conf_template.h corresponding to the RTOS you're using.

  2. Set LV_USE_DRAW_DMA2D_INTERRUPT to 1 in lv_conf.h.

  3. You must call lv_draw_dma2d_transfer_complete_interrupt_handler() when you receive the global interrupt that signals a DMA2D transfer has completed.

Interop with LTDC and NeoChrom

DMA2D usage can be freely mixed with LTDC usage as long as LV_ST_LTDC_USE_DMA2D_FLUSH is not enabled. LTDC will use the DMA2D peripheral for flushing, if that is enabled.

NeoChrom and DMA2D may be enabled at the same time. They are both draw units and they will both independently accept draw tasks.

API

lv_draw_dma2d.h

lv_draw_dma2d_private.h