Some observations on the operation of the display on the ILI9488 and ILI9341 controller and esp32-s3

[brightproject]

Recently I conducted several experiments with the ESP32-S3-N16-R8 microcontroller and displays based on the ILI9488 and ILI9341.

I designed a printed circuit board according to the following settings:

#define ESP32_PARALLEL
// #define TFT_PARALLEL_8_BIT

#define DISABLE_ALL_LIBRARY_WARNINGS

// #define ILI9341_DRIVER // DB8 = TFT_D0, DB9 = TFT_D1, DB10 =TFT_D2, DB11 =TFT_D3, DB12 =TFT_D4, DB13 = DB5, DB14 = TFT_D6, DB15 = TFT_D7, IM2 = 0, IM1 = 0, IM0 = 1
#define ILI9488_DRIVER // DB0 = TFT_D0, DB1 = TFT_D1, DB2 =TFT_D2, DB3 =TFT_D3, DB4 =TFT_D4, DB5 = DB5, DB6 = TFT_D6, DB7 = TFT_D7, IM2 = 0, IM1 = 1, IM0 = 1

#define TFT_CS 5
#define TFT_DC 7

#define TFT_RST -1// in scheme +3.3V

#define TFT_WR 4
#define TFT_RD 10

#define TFT_D0 14
#define TFT_D1 13
#define TFT_D2 47
#define TFT_D3 38
#define TFT_D4 39
#define TFT_D5 40
#define TFT_D6 41
#define TFT_D7 42

The code compiles with warnings:

.pio/libdeps/esp32-s3-devkitc1-n16r8/TFT_eSPI/Processors/TFT_eSPI_ESP32_S3.h:398:34: note: in expansion of macro 'GPIO_DIR_MASK'
       #define GPIO_OUT_CLR_MASK (GPIO_DIR_MASK | (1 << TFT_WR))

Accordingly, there is no image on the display.
I had to re-solder the connection diagram, according to the following:

#define ESP32_PARALLEL
// #define TFT_PARALLEL_8_BIT

#define DISABLE_ALL_LIBRARY_WARNINGS

// #define ILI9341_DRIVER // DB8 = TFT_D0, DB9 = TFT_D1, DB10 =TFT_D2, DB11 =TFT_D3, DB12 =TFT_D4, DB13 = DB5, DB14 = TFT_D6, DB15 = TFT_D7, IM2 = 0, IM1 = 0, IM0 = 1
#define ILI9488_DRIVER // DB0 = TFT_D0, DB1 = TFT_D1, DB2 =TFT_D2, DB3 =TFT_D3, DB4 =TFT_D4, DB5 = DB5, DB6 = TFT_D6, DB7 = TFT_D7, IM2 = 0, IM1 = 1, IM0 = 1

#define TFT_CS 5
#define TFT_DC 7

#define TFT_RST -1// in scheme +3.3V

#define TFT_WR 4
#define TFT_RD 10

#define TFT_D0 14
#define TFT_D1 13
#define TFT_D4 16
#define TFT_D5 3
#define TFT_D6 9
#define TFT_D7 8

It turns out that the pins

47, 38, 39, 40, 41, 42

Cannot be reassigned by the mask for display output.

https://www.espressif.com/sites/default/files/documentation/esp32-s3-wroom-1_wroom-1u_datasheet_en.pdf

I also did a little experiment connecting the ILI9341 display via SPI.
The platformio settings are as follows:

[env:esp32-s3-devkitc1-n16r8]
; platform = espressif32 @ 5.3.0
platform = espressif32 @ 6.6.0
board = esp32-s3-devkitc1-n16r8
framework = arduino
monitor_speed = 115200
monitor_port = COM6
upload_speed = 921600
upload_port = COM22
monitor_filters = esp32_exception_decoder

lib_deps = bodmer/TFT_e SPI@^2.5.43

If version 5.3.0
Processing esp32-s3-devkitc1-n16r8 (platform: espressif32 @ 5.3.0; board: esp32-s3-devkitc1-n16r8; framework: arduino)
then the output example with a display connected via SPI is 1363 msec.
And if you update to version 6.6.0
Processing esp32-s3-devkitc1-n16r8 (platform: espressif32 @ 6.6.0; board: esp32-s3-devkitc1-n16r8; framework: arduino)
then the output will be 842 msec.

The slowdown of the example occurs at the last step, where the pixels are output

  tft.fillRect(1,15,478-1,304-15,TFT_BLACK);

  for (int i=0; i<10000; i++)
  {
    tft.drawPixel(2+random(476), 16+random(289),random(0xFFFF));
  }

It seems that the 8-bit parallel interface should be faster than the SPI, but on the 480x320 display I see something different.

For display ILI9341 320х240, output via 8-bit parallel interface occurs in 542 ms.
There are no delays.
For display IILI9488 480х320, output via 8-bit parallel interface occurs in 1510 ms.
There is a slowdown again on the pixel output in the very last test.

ili9488_8_bit_parallel.mp4

I just decided to write about my observations, maybe it will help someone or they will give me some tips on how to increase the speed of the displays.