Implement LVGL 8.4

[mutatrum]

This PR adds LVGL 8.4 and refactors all current screens to use it.

The font is from https://int10h.org/oldschool-pc-fonts/fontlist/font?portfolio_6x8, converted to an LVGL font. There are also other fonts available, but currently unused.

The logo bitmap has been re-converted into an LVGL image.

This is a very basic implementation, almost pixel perfect close to the current screens. I think it makes sense to have this running for a while, to see if it's stable before venturing further into fancier usage of displays.

I haven't tested the self-test yet.

Fixes #299

[mutatrum]

Comparison of the flash size, before and after:

┏━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┳━━━━━━━━━━┳━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Memory Type/Section ┃ Used [bytes] ┃ Used [%] ┃ Remain [bytes] ┃ Total [bytes] ┃
┡━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━╇━━━━━━━━━━╇━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ Flash Code          │       736098 │     8.78 │        7652478 │       8388576 │
│    .text            │       736098 │     8.78 │                │               │
│ Flash Data          │       192460 │     0.57 │       33361940 │      33554400 │
│    .rodata          │       192204 │     0.57 │                │               │
│    .appdesc         │          256 │      0.0 │                │               │
│ DIRAM               │       123780 │    36.22 │         217980 │        341760 │
│    .text            │        79679 │    23.31 │                │               │
│    .bss             │        21536 │      6.3 │                │               │
│    .data            │        22372 │     6.55 │                │               │
│ IRAM                │        16383 │    99.99 │              1 │         16384 │
│    .text            │        15356 │    93.73 │                │               │
│    .vectors         │         1027 │     6.27 │                │               │
│ RTC FAST            │          284 │     3.47 │           7908 │          8192 │
│    .rtc_reserved    │           24 │     0.29 │                │               │
│    .force_fast      │            4 │     0.05 │                │               │
└─────────────────────┴──────────────┴──────────┴────────────────┴───────────────┘
Total image size: 1046996 bytes (.bin may be padded larger)

┏━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┳━━━━━━━━━━┳━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Memory Type/Section ┃ Used [bytes] ┃ Used [%] ┃ Remain [bytes] ┃ Total [bytes] ┃
┡━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━╇━━━━━━━━━━╇━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ Flash Code          │       836930 │     9.98 │        7551646 │       8388576 │
│    .text            │       836930 │     9.98 │                │               │
│ Flash Data          │       214292 │     0.64 │       33340108 │      33554400 │
│    .rodata          │       214036 │     0.64 │                │               │
│    .appdesc         │          256 │      0.0 │                │               │
│ DIRAM               │       155796 │    45.59 │         185964 │        341760 │
│    .text            │        79679 │    23.31 │                │               │
│    .bss             │        54272 │    15.88 │                │               │
│    .data            │        21652 │     6.34 │                │               │
│ IRAM                │        16383 │    99.99 │              1 │         16384 │
│    .text            │        15356 │    93.73 │                │               │
│    .vectors         │         1027 │     6.27 │                │               │
│ RTC FAST            │          284 │     3.47 │           7908 │          8192 │
│    .rtc_reserved    │           24 │     0.29 │                │               │
│    .force_fast      │            4 │     0.05 │                │               │
└─────────────────────┴──────────────┴──────────┴────────────────┴───────────────┘
Total image size: 1168940 bytes (.bin may be padded larger)

[mutatrum]

As a bare minimum for leveraging LVGL, I added scrolling labels for long values. The main one would the the pool url, which is not restricted to 20 characters.

[mutatrum]

After 17 hours uptime, nominal hashrate and no memory leaks as far as I can tell.

[mutatrum]

Using LVGL 9.2 with this display is blocked by espressif/esp-bsp#392.

[eandersson]

@mutatrum can you rebase this patch.

[mutatrum]

Ok, it's clean now. Squashed everything into a single commit. Some notes:

• This overwrites sdkconfig.defaults as it was generated by idf.py save-defconfig. There were several stale options and some missing options. If you change anything in the menu config, re-run save-defconfig to get an updated state, do not manually edit this.
• This enables the built-in PSRAM. This is not stricktly neccesary for LVGL, it does run without this enabled but it's running fairly low on memory. I'm not 100% convinced this is a safe thing to do, as I don't know if all the BitAxes ever built use the ESP32S3-WROOM-1-n16r8. If you have a model with different type memory (quad vs octal, or no memory), it'll crash. I don't know if it's possible to auto-detect this.
• I flipped screen one and two, it now shows IP address and pool as the first screen, as the asic is still ramping up at this time.
• The font used is: https://int10h.org/oldschool-pc-fonts/fontlist/font?portfolio_6x8, converted with lv_font_conf tool. I think this is exactly the same font.

[mutatrum]

I opted to do an almost 1-to-1 implementation first to get a clean change-set. The next step would be to inverse the control of displays, and have it refresh values live. This means that a lot of code from system.c would move to display.c, and the display class pulls data periodically out of the global state. This also make more sense if there are going to be different screen implementations.

[eandersson]

There is a crash if the lcd panel isn't installed (or broken) that will need to be fixed.

E (1370) display: Install SSD1306 panel driver
E (1370) i2c.master: I2C transaction unexpected nack detected
E (1380) i2c.master: s_i2c_synchronous_transaction(872): I2C transaction failed
E (1380) i2c.master: i2c_master_transmit(1074): I2C transaction failed
E (1390) lcd_panel.io.i2c: panel_io_i2c_tx_buffer(179): i2c transaction failed
E (1400) lcd_panel.ssd1306: panel_ssd1306_init(147): io tx param SSD1306_CMD_SET_MULTIPLEX failed
ESP_ERROR_CHECK failed: esp_err_t 0x103 (ESP_ERR_INVALID_STATE) at 0x420128a3
0x420128a3: display_init at /app/main/display.c:69 (discriminator 1)

file: "./main/display.c" line 69
func: display_init
expression: esp_lcd_panel_init(panel_handle)

abort() was called at PC 0x4037d797 on core 0
0x4037d797: _esp_error_check_failed at /opt/esp/idf/components/esp_system/esp_err.c:49

[mutatrum]

Without a display, it only fails on the init call, which was a bit unexpected. It now continues and starts hashing without a display:

I (2203) display: Install panel IO
I (2203) display: Install SSD1306 panel driver
I (2213) lcd_panel.ssd1306: ssd1306 init
E (2223) i2c.master: I2C transaction unexpected nack detected
E (2223) i2c.master: s_i2c_synchronous_transaction(872): I2C transaction failed
E (2233) i2c.master: i2c_master_transmit(1074): I2C transaction failed
E (2243) lcd_panel.io.i2c: panel_io_i2c_tx_buffer(204): i2c transaction failed
E (2253) lcd_panel.ssd1306: panel_ssd1306_init(150): io tx param SSD1306_CMD_SET_MULTIPLEX failed
E (2253) display: display_init(82): Display init failed
W (2263) SystemModule: OLED init failed!
I (2273) SystemModule: SYSTEM_task started

Changes implemented.

[eandersson]

Free Heap Memory changed drastically for me, maybe related to the sdk changes?

[eandersson]

Ah yep. If I use the old sdkconfig heap is the same as in previous versions

CONFIG_PARTITION_TABLE_CUSTOM=y
CONFIG_PARTITION_TABLE_CUSTOM_FILENAME="partitions.csv"
CONFIG_PARTITION_TABLE_FILENAME="partitions.csv"
CONFIG_PARTITION_TABLE_OFFSET=0x8000
CONFIG_PARTITION_TABLE_MD5=y
CONFIG_ESPTOOLPY_FLASHSIZE_16MB=y
CONFIG_ESPTOOLPY_FLASHSIZE="16MB"
CONFIG_HTTPD_WS_SUPPORT=y
CONFIG_SPIFFS_OBJ_NAME_LEN=64
CONFIG_HTTPD_MAX_URI_LEN=2048
CONFIG_ESP32S3_DEFAULT_CPU_FREQ_240=y
CONFIG_ESP_WIFI_11KV_SUPPORT=y
CONFIG_LV_USE_THEME_DEFAULT=n
CONFIG_LV_USE_THEME_BASIC=n
CONFIG_LV_USE_SNAPSHOT=n
CONFIG_LV_BUILD_EXAMPLES=n

[mutatrum]

This enables the built-in PSRAM. This is not stricktly neccesary for LVGL, it does run without this enabled but it's running fairly low on memory. I'm not 100% convinced this is a safe thing to do, as I don't know if all the BitAxes ever built use the ESP32S3-WROOM-1-n16r8. If you have a model with different type memory (quad vs octal, or no memory), it'll crash. I don't know if it's possible to auto-detect this.

No it's not the sdk version, it's enabled because the n16r8 has 8MB of PSRAM on board. This is enabled by:

CONFIG_SPIRAM=y
CONFIG_SPIRAM_MODE_OCT=y

As said above, not 100% sure about this. I think this is the model that's on the BOM, so I would assume everyone picked this variant. But no way to know for sure. I have a few other S3's, but they are all the same variant so I cannot test otherwise.

[eandersson]

This enables the built-in PSRAM. This is not stricktly neccesary for LVGL, it does run without this enabled but it's running fairly low on memory. I'm not 100% convinced this is a safe thing to do, as I don't know if all the BitAxes ever built use the ESP32S3-WROOM-1-n16r8. If you have a model with different type memory (quad vs octal, or no memory), it'll crash. I don't know if it's possible to auto-detect this.

No, it's enabled because the n16r8 has 8MB of PSRAM on board. This is enabled by:

CONFIG_SPIRAM=y
CONFIG_SPIRAM_MODE_OCT=y

As said above, not 100% sure about this. I think this is the model that's on the BOM, so I would assume everyone picked this variant. But no way to know for sure. I have a few other S3's, but they are all the same variant so I cannot test otherwise.

Yea - I would def prefer reducing the scope of the changes to make it easier to test, reduce the chance of red herrings etc and you can open a second PR if you want to make additional changes.

[eandersson]

Oher than that this PR looks solid now.

[mutatrum]

Current free ram after an hour: 92k.

I'll run it for a while to see if this causes issues.

[mutatrum]

Current free ram after an hour: 92k.

Still 92k free ran after 16 hours. Not a lot but stable.

[mutatrum]

Probably better to bump this over to at least the next version. I think they'll merge the fix for LVGL 9.2 soon, and I want to move to a minimal lv_conf.h so it only compiles LVGL features that are actually used. Which is almost nothing at the moment.