ESP32 ULP lsh bug

CMakeLists.txt

idf_component_register(
    SRCS "main.cpp"
    INCLUDE_DIRS "include"
    REQUIRES ulp soc
)
component_compile_options(-std=gnu++17)

set(ulp_app_name		"ulp_prog")
set(ulp_s_sources		"ulp/ulp_prog.S")
set(ulp_exp_dep_srcs	"main.cpp")

ulp_embed_binary(${ulp_app_name} "${ulp_s_sources}" "${ulp_exp_dep_srcs}")

main.cpp

#include <main.h>

// ULP header (autogen)
#include <ulp_prog.h>

#include <freertos/FreeRTOS.h>
#include <freertos/task.h>

#include <esp32/ulp.h>
#include <esp_sleep.h>
#include <esp_log.h>

#include <cstring>
#include <bitset>

static const char* TAG = "lsh-test";

void app_main(void) {
  ESP_LOGD(TAG, "[%s]: Initializing program memory", __func__);
  ESP_ERROR_CHECK(ulp_load_binary(0, ulp_bin_start,
    (ulp_bin_end - ulp_bin_start) / sizeof(uint32_t)));

  ESP_LOGD(TAG, "[%s]: ulp_wakeup_period=%uus", __func__, 10 * 1000);
  ESP_ERROR_CHECK(ulp_set_wakeup_period(0, 10 * 1000));

  // Start ULP on the io_run entry
  ESP_ERROR_CHECK(ulp_run(&ulp_entry - RTC_SLOW_MEM));

  vTaskDelay(1000 / portTICK_PERIOD_MS);
  
  ESP_LOGI(TAG, "[%s]: bits: \t\t\t%s", __func__,
    std::bitset<16>(0x226).to_string().c_str());
  ESP_LOGI(TAG, "[%s]: Left shift << 8: \t%s", __func__,
    std::bitset<16>(ulp_state1).to_string().c_str());
  ESP_LOGI(TAG, "[%s]: Left shift << 15: \t%s", __func__,
    std::bitset<16>(ulp_state2).to_string().c_str());
  ESP_LOGI(TAG, "[%s]: Left shift << 16: \t%s", __func__,
    std::bitset<16>(ulp_state3).to_string().c_str());
  ESP_LOGI(TAG, "[%s]: Left shift << 17: \t%s", __func__,
    std::bitset<16>(ulp_state4).to_string().c_str());
  ESP_LOGI(TAG, "[%s]: Left shift << 32: \t%s", __func__,
    std::bitset<16>(ulp_state5).to_string().c_str());
}

main.h

#ifndef MAIN_H
#define MAIN_H

#pragma once

#include <stdint.h>

extern const uint8_t ulp_bin_start[] asm("_binary_ulp_prog_bin_start");
extern const uint8_t ulp_bin_end[] asm("_binary_ulp_prog_bin_end");

extern "C" {
  void app_main(void);
}

auto app_main(void) -> void;

#endif /* MAIN_H */

ulp_prog.S

// ld visible variables - shares them & their memory with the MCU
.global entry

// Exposed mcu active flag
.global state1
.global state2
.global state3
.global state4
.global state5

// Zero initialized variable storage
.bss

// Base pointer used by all addr_* offsets
base:

state1:
.long 0
state2:
.long 0
state3:
.long 0
state4:
.long 0
state5:
.long 0

// Offsets, to simplify variable access (by loading evtb address, and
// using this as the `offset` value in `ld` & `st`), greatly decreasing the
// number of steps needed to load & store values
// One has to have the `base` address loaded in some register though, which I
// tend to keep in r2 for most of the program.
.set addr_state1,   (state1 - base)
.set addr_state2,   (state2 - base)
.set addr_state3,   (state3 - base)
.set addr_state4,   (state4 - base)
.set addr_state5,   (state5 - base)

.text

entry:
  move  r2, base
  move  r0, 0x226 //random bits, 0000001000100110

  lsh   r1, r0, 8
  st    r1, r2, addr_state1

  lsh   r1, r0, 15
  st    r1, r2, addr_state2

  lsh   r1, r0, 16
  st    r1, r2, addr_state3

  lsh   r1, r0, 17
  st    r1, r2, addr_state4

  lsh   r1, r0, 32
  st    r1, r2, addr_state5

ulp_halt:
  halt