esp-idf/examples/system/xip_from_psram
2024-09-14 11:13:20 +08:00
..
main doc(psram): clarify why psram speed is faster than flash 2024-09-14 11:13:20 +08:00
CMakeLists.txt feat(xip_psram): support xip psram example on p4 2024-07-25 10:52:59 +08:00
partitions.csv system: add an example and pytest to showing moving Flash content to PSRAM feature. 2022-09-19 12:25:30 +08:00
pytest_xip_from_psram.py doc(psram): clarify why psram speed is faster than flash 2024-09-14 11:13:20 +08:00
README.md doc(psram): clarify why psram speed is faster than flash 2024-09-14 11:13:20 +08:00
sdkconfig.ci.esp32p4_200m feat(psram): add psram noinit segment support on S2/S3/P4/C5 2024-08-21 12:13:16 +08:00
sdkconfig.ci.esp32s3_f4r8 ci: fix invalid kconfig options in system test apps 2023-05-09 11:27:55 +08:00
sdkconfig.ci.generic system: add an example and pytest to showing moving Flash content to PSRAM feature. 2022-09-19 12:25:30 +08:00
sdkconfig.defaults feat(xip_psram): support xip psram example on p4 2024-07-25 10:52:59 +08:00

Supported Targets ESP32-C5 ESP32-C61 ESP32-P4 ESP32-S2 ESP32-S3

XIP (Execute-In-Place) From PSRAM Example

This example illustrates a typical usage of XIP (Execute-In-Place) From PSRAM. With this feature:

  • You can optimize internal RAM usage
  • System can avoid disabling the Cache during an SPI1 Flash operation.

Overview

Here we define two sets of operations related to external memory: SET1: Operations where CPU fetches data and instructions from external memory. SET2: ESP Flash driver operations and other operations from drivers based on ESP Flash (NVS, Partition drivers, etc.).

By default, during SET2 operations, concurrent access requests to the Flash and PSRAM (SET1 operations) will be disabled otherwise both the SET1 and SET2 operations are not guaranteed to be safe (this is an undefined behaviour).

Only ISRs in internal RAM will get executed during SET2 operations. Besides, if any functions or data are accessed in these ISRs (usually this happens in ISR callbacks), they need to be placed into internal RAM as well. For interrupt handlers which need to execute when the cache is disabled (e.g., for low latency operations), you need to set the ESP_INTR_FLAG_IRAM flag when the interrupt handler is registered.

When CONFIG_SPIRAM_XIP_FROM_PSRAM is enabled, the flash.text sections (for instructions) and the .rodata section (read only data) will be moved to PSRAM. Corresponding virtual memory range will be mapped to PSRAM. Under this condition, ESP-IDF won't disable concurrent accesses to external memory (SET1 operations) anymore.

By using this feature, during SET2 operations, placement of ISRs, ISR callbacks, and related data are no longer limited to internal RAM.

Example Process

To show this feature, in this example we go through the following steps:

General Steps:

  1. Create a partition for Flash Erase Operation
  2. Create an esp_timer in one-shot mode

PSRAM Steps: 3. Do a Flash erase operation, and start the timer 4. ESP-Timer callback is dispatched and it calls a function in PSRAM during the flash erase operation 5. The Flash erase operation finishes 6. Show the result about the callback(in PSRAM) response and execute time

IRAM Steps: 7. Do a Flash erase operation, and start the timer 8. ESP-Timer callback is dispatched and it calls a function in IRAM during the flash erase operation 9. The flash erase operation finishes 10. Show the result about the callback(in IRAM) response and execute time

Timeline

Initialization and config -> Flash erase start -> ESP-Timer callback(in PSRAM) appear -> Flash erase finish -> Flash erase start -> ESP-Timer callback(in IRAM) appear -> Flash erase finish

                   ISR         CPU
                    |           |
                    |           |
                    |           |
                    |           * <----flash operation starts
                    |           *
  callback starts   * --------> *
     (in PSRAM)     *           *
  callback finishes * <-------- *
                    |           *
                    |           *
                    |           * <----flash operation finishes
                    |           |
                    |           |
                    |           |
                    |           |
                    |           |
                    |           |
                    |           * <----flash operation starts
                    |           *
  callback starts   * --------> *
     (in IRAM)      *           *
  callback finishes * <-------- *
                    |           *
                    |           *
                    |           * <----flash operation finishes
                    |           |
                    |           |

Example Result

The ISR which call a function in IRAM happening during Flash erase operations. CPU fetches instructions and data from internal RAM. The ISR which call a function in PSRAM happening during Flash erase operations and its response time is longer than calling a function in IRAM. That's because fetching instructions from PSRAM takes more time than fetching from IRAM.

Configure the project

Open the project configuration menu (idf.py menuconfig).

  1. Set the Partition Table -> Custom partition table CSV to y.
  2. Set the Component config -> High resolution timer -> Support ISR dispatch method to y.
  3. Set the Component config -> ESP PSRAM -> SPI RAM config to y then set Cache fetch instructions from SPI RAM and Cache load read only data from SPI RAM to y.

Build and Flash

Run idf.py -p PORT flash monitor to build and flash the project..

(To exit the serial monitor, type Ctrl-].)

See the Getting Started Guide for full steps to configure and use ESP-IDF to build projects.

Example Output

I (742) esp_psram: Reserving pool of 32K of internal memory for DMA/internal allocations
I (742) example: found partition 'storage1' at offset 0x110000 with size 0x10000

I (1152) example: callback(in PSRAM) response time: 7 us
I (1362) example: callback(in IRAM) response time: 5 us

Troubleshooting

For any technical queries, please open an issue on GitHub. We will get back to you soon.