ConvHL would work, but it would modify your frame buffer, and you have to convert it back afterwards, which is time consuming and inconvenient.
You should try to leave the byteswapping to the hardware.
The ByteSwap configuration on the ILI9341 display does not work in SPI mode, unfortunately, as stated in the datasheet.
And sometimes you just have to stick to a particular byte order. Say if you are using Chrom-Art (DMA2D) for instance. While just copying a rectangle just works fine, the BS (ByteSwap) bit for DMA2D (if available at all) doesn't do a great job when blending fonts (A4) onto the background (RGB565). (It probably needs an additional bit for that in the future).
Anyway, the next method may help you solve your problem, by just leaving the byte swapping to the hardware, using the next trick.
For this to happen, you can temporarily configure the SPI bus for 16-bit mode, so when you transfer your buffer, the bytes would be send swapped 2-by-2, automatically.
So let's say you function to transfer your buffer would look something like this, sending byte for byte (but the wrong order for the display):
void ILI9341_WriteData(uint8_t* buff, uint32_t buff_size)
{
SCB_CleanDCache_by_Addr((uint32_t*)(((uint32_t)buff) & ~(uint32_t)(__SCB_DCACHE_LINE_SIZE - 1U)), buff_size + __SCB_DCACHE_LINE_SIZE); /* Mem-to-Peri */
HAL_GPIO_WritePin(ILI9341_DC_PORT, ILI9341_DC_PIN, GPIO_PIN_SET);
#if defined SW_CS_PIN
HAL_GPIO_WritePin(ILI9341_CS_PORT, ILI9341_CS_PIN, GPIO_PIN_RESET);
#endif
uint16_t buff_size_max = 65536 - 4;
// Split data in smaller almost 64k chunks, because HAL can't send 64K at once.
while (buff_size > 0)
{
uint16_t chunk_size = buff_size > buff_size_max ? buff_size_max : buff_size;
HAL_SPI_Transmit(&ILI9341_SPI_PORT, buff, chunk_size, HAL_MAX_DELAY);
buff += chunk_size;
buff_size -= chunk_size;
}
#if defined SW_CS_PIN
HAL_GPIO_WritePin(ILI9341_CS_PORT, ILI9341_CS_PIN, GPIO_PIN_SET);
#endif
}We could make a function like this, which also sends the data, but in the right byte order for the display:
void ILI9341_WriteData16(uint8_t* buff, uint32_t buff_size)
{
SCB_CleanDCache_by_Addr((uint32_t*)(((uint32_t)buff) & ~(uint32_t)(__SCB_DCACHE_LINE_SIZE - 1U)), buff_size + __SCB_DCACHE_LINE_SIZE); /* Mem-to-Peri */
HAL_GPIO_WritePin(ILI9341_DC_PORT, ILI9341_DC_PIN, GPIO_PIN_SET);
#if defined SW_CS_PIN
HAL_GPIO_WritePin(ILI9341_CS_PORT, ILI9341_CS_PIN, GPIO_PIN_RESET);
#endif
// Set SPI to 16-bit
ILI9341_SPI_PORT.Init.DataSize = SPI_DATASIZE_16BIT;
HAL_SPI_Init(&ILI9341_SPI_PORT);
// Split data in smaller, less than 64k chunks, because HAL can't send 64K at once
buff_size /= 2;
uint32_t buff_size_max = 65536 - 4;
while (buff_size > 0)
{
uint16_t chunk_size = buff_size > buff_size_max ? buff_size_max : buff_size;
HAL_SPI_Transmit(&ILI9341_SPI_PORT, buff, chunk_size, HAL_MAX_DELAY);
buff += chunk_size * 2;
buff_size -= chunk_size;
}
#if defined SW_CS_PIN
HAL_GPIO_WritePin(ILI9341_CS_PORT, ILI9341_CS_PIN, GPIO_PIN_SET);
#endif
// Set SPI back to 8-bit
ILI9341_SPI_PORT.Init.DataSize = SPI_DATASIZE_8BIT;
HAL_SPI_Init(&ILI9341_SPI_PORT);
}Note, ILI9341_SPI_PORT is a just definition in a header file for your display, i.e. ILI9341_conf.h, used to confgigure the display IO:
#define ILI9341_SPI_PORT hspi4In the code above, we only modify the DataSize of the SPI bus and leave the rest the same as when initally MX_SPI4_Init was called for example.
I have tried this for you, using an ILI9341 display, connected to a Nucleo-H7A3ZI-Q development board, and it works fine this way. You now do not have to loose time swapping bytes forth and back.
I added GPIO CS control too, in case you do it by software.
And never forget to set your window for the display.
You now can use DMA2D and nicely blend images.
Hope this helps.
