MicroPython on ESP32 with SPIRAM support

All ESP32 boards running MicroPython.
Target audience: MicroPython users with an ESP32 board.
Post Reply
loboris
Posts: 344
Joined: Fri Oct 02, 2015 6:19 pm

MicroPython on ESP32 with SPIRAM support

Post by loboris » Sat Jul 08, 2017 10:45 am

MicroPython for ESP32 with SPIRAM support
GitHub repository
The repository can be used to build MicroPython for modules/boards with SPIRAM as well as for regular ESP32 modules/boards without SPIRAM.

As of Sep 18, 2017 full support for psRAM is included into esp-idf and xtensa toolchain.

Building on Linux, MacOS and Windows is supported.

You can support this project by donating via PayPal
MicroPython works great on ESP32, but the most serious issue is still (as on most other MicroPython boards) limited amount of free memory.

ESP32 can use external SPI RAM (psRAM) to expand available RAM up to 16MB.
Currently, there are several modules & development boards which incorporates 4MB of SPIRAM:

ESP-WROVER-KIT boards from Espressif, AnalogLamb or Electrodragon.
ESP-WROVER from Espressif, AnalogLamb or Electrodragon.
ALB32-WROVER from AnalogLamb.
S01 and L01 OEM modules from Pycom.

The repository contains all the tools and sources necessary to build working MicroPython firmware which can fully use the advantages of 4MB (or more) of SPIRAM

It is huge difference between MicroPython running with less than 100KB of free memory and running with 4MB of free memory.

Some basic documentation specific to this MicroPython port is available. It will soon be updated to include the documentation for all added/changed modules.

Some examples can be found in modules_examples directory.

The MicroPython firmware is built as esp-idf component

This means the regular esp-idf menuconfig system can be used for configuration. Besides the ESP32 configuration itself, some MicroPython options can also be configured via menuconfig.
This way many features not available in standard ESP32 MicroPython are enabled, like unicore/dualcore, all Flash speed/mode options etc. No manual sdkconfig.h editing and tweaking is necessary.


Features and some differences from standard MicroPython ESP32 build
  • MicroPython build is based on latest build (1.9.2) from main MicroPython repository.
  • ESP32 build is based on MicroPython's ESP32 build) with added changes needed to build on ESP32 with SPIRAM and with esp-idf build system.
  • Default configuration for SPIRAM build has 2MB of MicroPython heap, 20KB of MicroPython stack, ~200KB of free DRAM heap for C modules and functions
  • MicroPython can be built in unicore (FreeRTOS & MicroPython task running only on the first ESP32 core, or dualcore configuration (MicroPython task running on ESP32 App core).
  • ESP32 Flash can be configured in any mode, QIO, QOUT, DIO, DOUT
  • BUILD.sh script is provided to make building MicroPython firmware as easy as possible
  • Internal filesystem is built with esp-idf wear leveling driver, so there is less danger of damaging the flash with frequent writes. File system parameters (start address, size, ...) can be set via menuconfig.
  • sdcard module is included which uses esp-idf sdmmc driver and can work in SD mode (1-bit and 4-bit) or in SPI mode (sd card can be connected to any pins). On ESP-WROVER-KIT it works without changes, for information on how to connect sdcard on other boards check the dicumentation
  • Native ESP32 VFS support is used for spi Flash (FatFS or SPIFS) & sdcard filesystems.
  • SPIFFS filesystem support, can be used instead of FatFS on SPI Flash. Configurable via menuconfig
  • RTC Class is added to machine module, including methods for synchronization of system time to ntp server, deepsleep, wakeup from deepsleep on external pin level, ...
  • Time zone can be configured via menuconfig and is used when syncronizing time from NTP server
  • files timestamp is correctly set to system time both on internal fat filesysten and on sdcard
  • Built-in ymodem module for fast transfer of text or binary files of any size to/from host. Uses the same uart on which REPL runs.
  • Some additional frozen modules are added, like pye editor, urequests, functools, logging, ...
  • Btree module is included, can be Enabled/Disabled via menuconfig.
  • _threads module is greatly improved, inter-thread notifications and messaging included
  • Neopixel module using ESP32 RMT peripheral with many new features, unlimited number of pixels
  • i2c module uses ESP32 hardware i2c driver
  • spi module uses ESP32 hardware spi driver
  • curl module added, http/https get,post, send mail (including gMail), ftp client (get, put, list)
  • ssh module added, sftp get, put, list, mkdir, execute any command on host
  • display module added with full support for spi TFT displays
  • DHT module implemented using ESP32 RMT peripheral
  • mqtt module added, implemented in C, runs in separate task
  • telnet module added, connect to **REPL via WiFi** using telnet protocol
  • ftp server module added, runs as separate ESP32 task
  • GSM module with PPPoS support, all network functions works the same as with WiFi, SMS, AT commands, ...
  • NVS support in machine module
  • Eclipse project files are included. To include the project into Eclipse goto File->Import->Existing Projects into Workspace->Select root directory->[select MicroPython_BUILD directory]->Finish. After opening, rebuild the index.

How to Build

Clone the repository:

Code: Select all

git clone https://github.com/loboris/MicroPython_ESP32_psRAM_LoBo.git
Xtensa toolchains for Linux, Windows and MacOS and esp-idf are provided as tar archives. They will be automatically unpacked on first run of BUILD.sh script.

Goto MicroPython_BUILD directory

To change some ESP32 & Micropython options run:

Code: Select all

./BUILD.sh menuconfig
To build the MicroPython firmware, run:

Code: Select all

./BUILD.sh
You can use -jN (N=number of cores to use) to make the build process faster (it only takes les 15 seconds seconds on my system with -j4).
If using too high N the build may fail, if that happens, run build again or run without the -j option.

If no errors are detected, you can now flash the MicroPython firmware to your board. Run:

Code: Select all

./BUILD.sh flash
The board stays in bootloader mode. Run your terminal emulator to connect to the board and reset the board.
You can also run ./BUILD.sh monitor to use esp-idf's terminal program, it will reset the board automatically.

BUILD.sh

Included BUILD.sh script makes building MicroPython firmware easy.
Usage:
  • ./BUILD.sh - run the build, create MicroPython firmware
  • ./BUILD.sh -jn - run the build on multicore system, much faster build. Replace n with the number of cores on your system
  • ./BUILD.sh menuconfig - run menuconfig to configure ESP32/MicroPython
  • ./BUILD.sh clean - clean the build
  • ./BUILD.sh flash - flash MicroPython firmware to ESP32
  • ./BUILD.sh erase - erase the whole ESP32 Flash
  • ./BUILD.sh monitor - run esp-idf terminal program
  • ./BUILD.sh makefs - create SPIFFS file system image which can be flashed to ESP32
  • ./BUILD.sh flashfs - flash SPIFFS file system image to ESP32, if not created, create it first
  • ./BUILD.sh copyfs - flash the default SPIFFS file system image to ESP32
  • ./BUILD.sh makefatfs - create FatFS file system image which can be flashed to ESP32
  • ./BUILD.sh flashfatfs - flash FatFS file system image to ESP32, if not created, create it first
  • ./BUILD.sh copyfatfs - flash the default FatFS file system image to ESP32
As default the build process runs silently, without showing compiler output. You can change that by adding verbose as the last parameter to BUILD.sh.

To build with SPIRAM support:
In menuconfig select → Component config → ESP32-specific → Support for external, SPI-connected RAM
In menuconfig select → Component config → ESP32-specific → SPI RAM config → Make RAM allocatable using heap_caps_malloc

After the successful build the firmware files will be placed into firmware directory. flash.sh script will also be created which can be used for flashing the firmware without building it first.

Using SPIFFS filesystem

SPIFFS filesystem can be used on internal spi Flash instead of FatFS.
If you want to use it configure it via menuconfig → MicroPython → File systems → Use SPIFFS

Prepared image file can be flashed to ESP32, if not flashed, filesystem will be formated after first boot.

SFPIFFS image can be prepared on host and flashed to ESP32:

Copy the files to be included on spiffs into components/spiffs_image/image/ directory. Subdirectories can also be added.

Execute:

Code: Select all

./BUILD.sh makefs
to create spiffs image in build directory without flashing to ESP32

Execute:

Code: Select all

./BUILD.sh flashfs
to create spiffs image in build directory and flash it to ESP32

Execute:

Code: Select all

./BUILD.sh copyfs
to flash default spiffs image components/spiffs_image/spiffs_image.img to ESP32


Some examples

Using new machine methods and RTC:

Code: Select all

import machine

rtc = machine.RTC()

rtc.init((2017, 6, 12, 14, 35, 20))

rtc.now()

rtc.ntp_sync(server="<ntp_server>" [,update_period=])
  <ntp_server> can be empty string, then the default server is used ("pool.ntp.org")

rtc.synced()
  returns True if time synchronized to NTP server

rtc.wake_on_ext0(Pin, level)
rtc.wake_on_ext1(Pin, level)
  wake up from deepsleep on pin level

machine.deepsleep(time_ms)
machine.wake_reason()
  returns tuple with reset & wakeup reasons
machine.wake_description()
  returns tuple with strings describing reset & wakeup reasons
Using sdcard module:

Code: Select all

import uos

uos.mountsd()
uos.listdir('/sd')
You can also set working directory to the root sd card directory automatically:

Code: Select all

>>> import uos
>>> uos.mountsd(True)
---------------------
 Mode:  SD (4bit)
 Name: NCard
 Type: SDHC/SDXC
Speed: default speed (25 MHz)
 Size: 15079 MB
  CSD: ver=1, sector_size=512, capacity=30881792 read_bl_len=9
  SCR: sd_spec=2, bus_width=5

>>> uos.listdir()
['overlays', 'bcm2708-rpi-0-w.dtb', ......
Example terminal session

Code: Select all

rst:0x1 (POWERON_RESET),boot:0x3e (SPI_FAST_FLASH_BOOT)
configsip: 0, SPIWP:0xee
clk_drv:0x00,q_drv:0x00,d_drv:0x00,cs0_drv:0x00,hd_drv:0x00,wp_drv:0x00
mode:DIO, clock div:2
load:0x3fff0010,len:4
load:0x3fff0014,len:5656
load:0x40078000,len:0
ho 12 tail 0 room 4
load:0x40078000,len:13220
entry 0x40078fe4
W (36) rtc_clk: Possibly invalid CONFIG_ESP32_XTAL_FREQ setting (40MHz). Detected 40 MHz.
I (59) boot: ESP-IDF ESP32_LoBo_v1.9.1-13-gfecf988-dirty 2nd stage bootloader
I (60) boot: compile time 21:07:29
I (108) boot: Enabling RNG early entropy source...
I (108) boot: SPI Speed      : 40MHz
I (108) boot: SPI Mode       : DIO
I (115) boot: SPI Flash Size : 4MB
I (128) boot: Partition Table:
I (139) boot: ## Label            Usage          Type ST Offset   Length
I (162) boot:  0 nvs              WiFi data        01 02 00009000 00006000
I (185) boot:  1 phy_init         RF data          01 01 0000f000 00001000
I (209) boot:  2 MicroPython      factory app      00 00 00010000 00270000
I (232) boot:  3 internalfs       Unknown data     01 81 00280000 00140000
I (255) boot: End of partition table
I (268) esp_image: segment 0: paddr=0x00010020 vaddr=0x3f400020 size=0x48a74 (297588) map
I (613) esp_image: segment 1: paddr=0x00058a9c vaddr=0x3ffb0000 size=0x07574 ( 30068) load
I (650) esp_image: segment 2: paddr=0x00060018 vaddr=0x400d0018 size=0xc83f4 (820212) map
0x400d0018: _stext at ??:?

I (1525) esp_image: segment 3: paddr=0x00128414 vaddr=0x3ffb7574 size=0x052d0 ( 21200) load
I (1551) esp_image: segment 4: paddr=0x0012d6ec vaddr=0x40080000 size=0x00400 (  1024) load
0x40080000: _iram_start at /home/LoBo2_Razno/ESP32/MicroPython/MicroPython_ESP32_psRAM_LoBo/Tools/esp-idf/components/freertos/./xtensa_vectors.S:1675

I (1553) esp_image: segment 5: paddr=0x0012daf4 vaddr=0x40080400 size=0x1a744 (108356) load
I (1711) esp_image: segment 6: paddr=0x00148240 vaddr=0x400c0000 size=0x0006c (   108) load
I (1712) esp_image: segment 7: paddr=0x001482b4 vaddr=0x50000000 size=0x00400 (  1024) load
I (1794) boot: Loaded app from partition at offset 0x10000
I (1794) boot: Disabling RNG early entropy source...
I (1800) spiram: SPI RAM mode: flash 40m sram 40m
I (1812) spiram: PSRAM initialized, cache is in low/high (2-core) mode.
I (1834) cpu_start: Pro cpu up.
I (1846) cpu_start: Starting app cpu, entry point is 0x400814e4
0x400814e4: call_start_cpu1 at /home/LoBo2_Razno/ESP32/MicroPython/MicroPython_ESP32_psRAM_LoBo/Tools/esp-idf/components/esp32/./cpu_start.c:219

I (0) cpu_start: App cpu up.
I (4612) spiram: SPI SRAM memory test OK
I (4614) heap_init: Initializing. RAM available for dynamic allocation:
I (4615) heap_init: At 3FFAE2A0 len 00001D60 (7 KiB): DRAM
I (4633) heap_init: At 3FFC30C0 len 0001CF40 (115 KiB): DRAM
I (4653) heap_init: At 3FFE0440 len 00003BC0 (14 KiB): D/IRAM
I (4672) heap_init: At 3FFE4350 len 0001BCB0 (111 KiB): D/IRAM
I (4692) heap_init: At 4009AB44 len 000054BC (21 KiB): IRAM
I (4712) cpu_start: Pro cpu start user code
I (4777) cpu_start: Starting scheduler on PRO CPU.
I (2920) cpu_start: Starting scheduler on APP CPU.

FreeRTOS running on BOTH CORES, MicroPython task started on App Core.

uPY stack size = 19456 bytes
uPY  heap size = 2097152 bytes (in SPIRAM using heap_caps_malloc)

Reset reason: Power on reset Wakeup: Power on wake
I (3130) phy: phy_version: 359.0, e79c19d, Aug 31 2017, 17:06:07, 0, 0

Starting WiFi ...
WiFi started
Synchronize time from NTP server ...
Time set

MicroPython ESP32_LoBo_v2.0.2 - 2017-09-19 on ESP32 board with ESP32
Type "help()" for more information.
>>> 
>>> import micropython, machine
>>> micropython.mem_info()
stack: 736 out of 19456
GC: total: 2049088, used: 6848, free: 2042240
 No. of 1-blocks: 37, 2-blocks: 9, max blk sz: 329, max free sz: 127565
>>> machine.heap_info()
Free heap outside of MicroPython heap:
 total=2232108, SPISRAM=2097108, DRAM=135000
>>> 
>>> a = ['esp32'] * 200000
>>> 
>>> a[123456]
'esp32'
>>> 
>>> micropython.mem_info()
stack: 736 out of 19456
GC: total: 2049088, used: 807104, free: 1241984
 No. of 1-blocks: 44, 2-blocks: 13, max blk sz: 50000, max free sz: 77565
>>> 

Tested on ESP-WROVER-KIT v3, Adafruit HUZZAH32 - ESP32 Feather

Image ................... Image
Last edited by loboris on Tue Oct 24, 2017 7:10 pm, edited 5 times in total.

loboris
Posts: 344
Joined: Fri Oct 02, 2015 6:19 pm

Re: MicroPython on ESP-WROVER with 4MB of psRAM

Post by loboris » Sun Jul 16, 2017 12:17 pm

Additional information may be available on ESP32 Forum.
Last edited by loboris on Tue Oct 24, 2017 2:41 pm, edited 1 time in total.

User avatar
pythoncoder
Posts: 5956
Joined: Fri Jul 18, 2014 8:01 am
Location: UK
Contact:

Re: MicroPython on ESP-WROVER with 4MB of psRAM

Post by pythoncoder » Sun Jul 16, 2017 4:16 pm

That sounds like a major achievement. Can you tell more about how a psRAM build works? How does heap allocation "decide" between internal RAM and psRAM? Is bytecode preferentially assigned to RAM? Has performance been quantified compared to a build using internal RAM only?
Peter Hinch
Index to my micropython libraries.

loboris
Posts: 344
Joined: Fri Oct 02, 2015 6:19 pm

Re: MicroPython on ESP-WROVER with 4MB of psRAM

Post by loboris » Sun Jul 16, 2017 7:23 pm

At this time the complete MicroPython heap is allocated on psRAM.
Operations from psRAM are slower then from internal RAM, as it is accessed over SPI interface, with cache in internal RAM (32KB if I remember well). So, if we are operating with the data already in the cache, the speed will be the same. If the new block of data has to be placed in cache,there will be a delay. It works the same way the SPI Flash is accessed.

I've done some simple tests, keep in mind that for now I couldn't enable psRAM operating at 80MHz (only with MicroPython, it works with esp-idf applications) and that the psRAM support in eps-idf is at very early stage (not even officially supported yet).

Here are the rusults:

Running without psRAM support:

Code: Select all

FreeRTOS running on BOTH CORES, MicroPython task started on App Core.

Allocating uPY stack: size=16384 bytes
Allocating uPY heap:  size=94208 bytes

Reset reason: Power on reset Wakeup: Power on wake
OSError: [Errno 2] ENOENT
MicroPython b94fddd-dirty on 2017-07-16; ESP-WROVER module with ESP32+psRAM
Type "help()" for more information.
>>> import micropython,time
>>> t1=time.ticks_us(); x=['esp32'] * 15000; t2=time.ticks_us(); print(t2-t1)
3554
>>> micropython.mem_info()
stack: 736 out of 15360
GC: total: 92032, used: 68640, free: 23392
 No. of 1-blocks: 43, 2-blocks: 6, max blk sz: 3750, max free sz: 1428
>>> 

==================

>>> t1=time.ticks_us(); x=['esp32'] * 1000; t2=time.ticks_us(); print(t2-t1)
383
>>> micropython.mem_info()
stack: 736 out of 15360
GC: total: 92032, used: 12640, free: 79392
 No. of 1-blocks: 43, 2-blocks: 6, max blk sz: 264, max free sz: 4928
>>> 
With psRAM enabled:

Code: Select all

FreeRTOS running on BOTH CORES, MicroPython task started on App Core.

Allocating uPY stack: size=65536 bytes
Allocating uPY heap:  size=4194048 bytes (in SPIRAM using pvPortMallocCaps)

Reset reason: Power on reset Wakeup: Power on wake
OSError: [Errno 2] ENOENT
MicroPython b94fddd-dirty on 2017-07-16; ESP-WROVER module with ESP32+psRAM
Type "help()" for more information.
>>> import micropython,time
>>> t1=time.ticks_us(); x=['esp32'] * 15000; t2=time.ticks_us(); print(t2-t1)
17655
>>> micropython.mem_info()
stack: 736 out of 64512
GC: total: 4097984, used: 68640, free: 4029344
 No. of 1-blocks: 43, 2-blocks: 6, max blk sz: 3750, max free sz: 251800
>>> 

==================

>>> t1=time.ticks_us(); x=['esp32'] * 1000; t2=time.ticks_us(); print(t2-t1)
692
>>> micropython.mem_info()
stack: 736 out of 64512
GC: total: 4097984, used: 12640, free: 4085344
 No. of 1-blocks: 43, 2-blocks: 6, max blk sz: 264, max free sz: 255300
>>> 

==================

>>> t1=time.ticks_us(); x=['esp32'] * 150000; t2=time.ticks_us(); print(t2-t1)
184038
>>> micropython.mem_info()
stack: 736 out of 64512
GC: total: 4097984, used: 608848, free: 3489136
 No. of 1-blocks: 49, 2-blocks: 8, max blk sz: 37500, max free sz: 218037
Note the small difference with small amount of data.
With 80Mhz psRAM clock the times should be half of the measured values.

My measurements with simple memset function (on 2MB block allocated on psRAM) in esp-idf based application shows that the speed is ~7.5 MB/s with 40MHZ psRAM clock and ~14.3 MB/s with 80MHz psRAM clock.
How does heap allocation "decide" between internal RAM and psRAM?
There is a configuration option → Component config → ESP32-specific → Capability allocator can allocate SPI RAM memory (CONFIG_MEMMAP_SPIRAM_ENABLE).
This option is checked in app_main() in main.c to decide if psRAM can be used, see main.c for detailes.

I think all this is very promising, we jus have to wait a little for full esp-idf support and modules availability to take full advantage of ESP32 with psRAM capabilities. Of course, MicroPython itself could/should be optimized to work with psRAM.

User avatar
pythoncoder
Posts: 5956
Joined: Fri Jul 18, 2014 8:01 am
Location: UK
Contact:

Re: MicroPython on ESP-WROVER with 4MB of psRAM

Post by pythoncoder » Mon Jul 17, 2017 8:17 am

Impressive, especially when you get 80MHz working.
Peter Hinch
Index to my micropython libraries.

loboris
Posts: 344
Joined: Fri Oct 02, 2015 6:19 pm

Re: MicroPython on ESP-WROVER with 4MB of psRAM

Post by loboris » Tue Jul 18, 2017 9:00 am

This is the test with SPI Flash and psRAM running at 80MHz:
x=['esp32'] * 150000
40MHz: 184038 usec
80MHz: 115515 usec
Not exactly half the time, but significant improvement (~63%).

Code: Select all

I (44) boot: ESP-IDF b94fddd-dirty 2nd stage bootloader
I (45) boot: compile time 10:43:52
I (71) boot: Enabling RNG early entropy source...
I (72) qio_mode: Enabling QIO for flash chip GD
I (76) boot: SPI Speed      : 80MHz
I (89) boot: SPI Mode       : QIO
I (101) boot: SPI Flash Size : 4MB
I (114) boot: Partition Table:
I (125) boot: ## Label            Usage          Type ST Offset   Length
I (148) boot:  0 nvs              WiFi data        01 02 00009000 00006000
I (171) boot:  1 phy_init         RF data          01 01 0000f000 00001000
I (194) boot:  2 factory          factory app      00 00 00010000 00100000
I (218) boot: End of partition table
I (231) boot: Disabling RNG early entropy source...
I (248) boot: Loading app partition at offset 00010000
I (1535) boot: segment 0: paddr=0x00010018 vaddr=0x00000000 size=0x0ffe8 ( 65512) 
I (1536) boot: segment 1: paddr=0x00020008 vaddr=0x3f400010 size=0x2e4b0 (189616) map
I (1552) boot: segment 2: paddr=0x0004e4c0 vaddr=0x3ffb0000 size=0x03078 ( 12408) load
I (1580) boot: segment 3: paddr=0x00051540 vaddr=0x40080000 size=0x00400 (  1024) load
I (1605) boot: segment 4: paddr=0x00051948 vaddr=0x40080400 size=0x17084 ( 94340) load
I (1652) boot: segment 5: paddr=0x000689d4 vaddr=0x400c0000 size=0x00070 (   112) load
I (1658) boot: segment 6: paddr=0x00068a4c vaddr=0x00000000 size=0x075bc ( 30140) 
I (1683) boot: segment 7: paddr=0x00070010 vaddr=0x400d0018 size=0xb1ce0 (728288) map
I (1710) cpu_start: PSRAM mode: flash 80m sram 80m
I (1726) cpu_start: PSRAM initialized, cache is in even/odd (2-core) mode.
I (1749) cpu_start: Pro cpu up.
I (1761) cpu_start: Starting app cpu, entry point is 0x400810a8
I (0) cpu_start: App cpu up.
I (3230) heap_alloc_caps: SPI SRAM memory test OK
I (3231) heap_alloc_caps: Initializing. RAM available for dynamic allocation:
I (3236) heap_alloc_caps: At 3F800000 len 00400000 (4096 KiB): SPIRAM
I (3258) heap_alloc_caps: At 3FFAE2A0 len 00001D60 (7 KiB): DRAM
I (3279) heap_alloc_caps: At 3FFB9170 len 00006E90 (27 KiB): DRAM
I (3299) heap_alloc_caps: At 3FFC0000 len 00010000 (64 KiB): DMAONLY
I (3321) heap_alloc_caps: At 3FFD0000 len 00010000 (64 KiB): DRAM
I (3342) heap_alloc_caps: At 3FFE0440 len 00003BC0 (14 KiB): D/IRAM
I (3363) heap_alloc_caps: At 3FFE4350 len 0001BCB0 (111 KiB): D/IRAM
I (3385) heap_alloc_caps: At 40097484 len 00008B7C (34 KiB): IRAM
I (3406) cpu_start: Pro cpu start user code
I (3462) cpu_start: Starting scheduler on PRO CPU.
I (1683) cpu_start: Starting scheduler on APP CPU.

FreeRTOS running on BOTH CORES, MicroPython task started on App Core.

Allocating uPY stack: size=65536 bytes
Allocating uPY heap:  size=4194048 bytes (in SPIRAM using malloc)

Reset cause: 0
MicroPython 4ad81dd-dirty on 2017-07-15; ESP-WROVER module with ESP32+psRAM
Type "help()" for more information.
>>> import time,micropython
>>> micropython.mem_info()
stack: 736 out of 64512
GC: total: 4097984, used: 320, free: 4097664
 No. of 1-blocks: 6, 2-blocks: 7, max blk sz: 2, max free sz: 256095
>>> t1=time.ticks_us(); x=['esp32'] * 150000; t2=time.ticks_us(); print(t2-t1)
115515
>>> micropython.mem_info()
stack: 736 out of 64512
GC: total: 4097984, used: 600944, free: 3497040
 No. of 1-blocks: 13, 2-blocks: 7, max blk sz: 37500, max free sz: 218528
>>> 

loboris
Posts: 344
Joined: Fri Oct 02, 2015 6:19 pm

Re: MicroPython on ESP-WROVER with 4MB of psRAM

Post by loboris » Fri Jul 21, 2017 3:48 pm

:!: Important update:
  • The MicroPython firmware is now built as esp-idf component.
  • menuconfig can be used for ESP32 configuration as well as for configuration of some MicroPython options.
  • The repository can be now used to build MicroPython for modules with psRAM as well as for ESP32 modules without psRAM. Standard esp-idf and Xtensa toolchain can be used if building without psRAM support.
  • micropython is no longer submodule, it's core is based on latest build (1.9.1) from main MicroPython repository, with some modifications needed for the new build system.
  • All ESP32 working modes (unicore/dualcore) and speeds, as well as all flash/psRAM speeds and modes are supported.
  • BUILD.sh script improved with more build options.
  • BTree module is included, can be Enabled/Disabled via menuconfig.
  • The repository README-md is updated with new information.
GitHub repository: https://github.com/loboris/MicroPython_ESP32_psRAM_LoBo
ESP32 Forum: https://www.esp32.com/viewtopic.php?f=17&t=2446
Last edited by loboris on Sat Jul 22, 2017 7:49 pm, edited 1 time in total.

User avatar
pythoncoder
Posts: 5956
Joined: Fri Jul 18, 2014 8:01 am
Location: UK
Contact:

Re: MicroPython on ESP-WROVER with 4MB of psRAM

Post by pythoncoder » Sat Jul 22, 2017 6:18 am

Both the board and the work you've done look awesome :D You might want to tell Adafruit about it given that they sell the board and support MicroPython.
Peter Hinch
Index to my micropython libraries.

loboris
Posts: 344
Joined: Fri Oct 02, 2015 6:19 pm

Re: MicroPython on ESP-WROVER with 4MB of psRAM

Post by loboris » Sat Jul 22, 2017 7:57 pm

:D Good news for everybody wanting to work with ESP32+psRAM.

AnalogLamb is offering ALB32-WROVER module with 8MB Flash and 4MB psRAM in ESP-WROOM-32 footprint package. Available for pre-order, will be released on August 8, 2017.
https://www.analoglamb.com/product/alb3 ... 32mb-psram
Image

User avatar
pythoncoder
Posts: 5956
Joined: Fri Jul 18, 2014 8:01 am
Location: UK
Contact:

Re: MicroPython on ESP-WROVER with 4MB of psRAM

Post by pythoncoder » Sun Jul 23, 2017 9:45 am

Interesting. I'd assumed that the Adafruit/Espressif boards had psRAM but looking at the schematic https://dl.espressif.com/dl/schematics/ ... _SCH-3.pdf I can't see any. An unfortunate omission.

AAMOI why the ESD protection diodes on the SD card connector? Is there a possibility of hardware other than an SD card being plugged in?
Peter Hinch
Index to my micropython libraries.

Post Reply