Enable gprof onboard profiling (#2669)

Adds a menu item to enable onboard profiling. This requires significant RAM and really only makes sense on devices with PSRAM to store the state. When the menu item is selected, allocates RAM and tracks function calls and periodically samples the PC to generate a histogram of application usage. The onboard gmon.out file can be written over Semihosting or some other way to transfer to a PC for analysis. Adds a profiling example with command lines.
earlephilhower · Dec 6, 2024 · 0061d3f · 0061d3f
1 parent 48bc91a
commit 0061d3f
Show file tree

Hide file tree

Showing 17 changed files with 1,259 additions and 28 deletions.
diff --git a/.github/workflows/pull-request.yml b/.github/workflows/pull-request.yml
@@ -20,7 +20,7 @@ jobs:
       uses: codespell-project/actions-codespell@v2
       with:
         skip: ./ArduinoCore-API,./libraries/ESP8266SdFat,./libraries/Adafruit_TinyUSB_Arduino,./libraries/LittleFS/lib,./tools/pyserial,./pico-sdk,./.github,./docs/i2s.rst,./cores/rp2040/api,./libraries/FreeRTOS,./tools/libbearssl/bearssl,./include,./libraries/WiFi/examples/BearSSL_Server,./ota/uzlib,./libraries/http-parser/lib,./libraries/WebServer/examples/HelloServerBearSSL/HelloServerBearSSL.ino,./libraries/HTTPUpdateServer/examples/SecureBearSSLUpdater/SecureBearSSLUpdater.ino,./.git,./libraries/FatFS/lib/fatfs,./libraries/FatFS/src/diskio.h,./libraries/FatFS/src/ff.cpp,./libraries/FatFS/src/ffconf.h,./libraries/FatFS/src/ffsystem.cpp,./libraries/FatFS/src/ff.h,./libraries/lwIP_WINC1500/src/driver,./libraries/lwIP_WINC1500/src/common,./libraries/lwIP_WINC1500/src/bus_wrapper,./libraries/lwIP_WINC1500/src/spi_flash
-        ignore_words_list: ser,dout,shiftIn,acount
+        ignore_words_list: ser,dout,shiftIn,acount,froms
     - name: Get submodules for following tests
       run: git submodule update --init
     - name: Check package references

diff --git a/README.md b/README.md
@@ -137,6 +137,8 @@ Read the [Contributing Guide](https://github.com/earlephilhower/arduino-pico/blo
 * printf (i.e. debug) output over USB serial
 * Transparent use of PSRAM globals and heap (RP2350 only)
 * ARM or RISC-V (Hazard3) support for the RP2350
+* Semihosted serial and file system access
+* GPROF profiling support
 
 The RP2040 PIO state machines (SMs) are used to generate jitter-free:
 * Servos

diff --git a/boards.txt b/boards.txt
diff --git a/cores/rp2040/RP2040Support.cpp b/cores/rp2040/RP2040Support.cpp
@@ -18,9 +18,11 @@
     Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
 */
 
+#include <Arduino.h>
+#include <pico/runtime.h>
+
 #ifdef PICO_RP2040
 
-#include <Arduino.h>
 #include <hardware/structs/psm.h>
 
 extern "C" void boot_double_tap_check();
@@ -35,3 +37,17 @@ void RP2040::enableDoubleResetBootloader() {
 }
 
 #endif
+
+#ifdef __PROFILE
+Stream *__profileFile;
+int __writeProfileCB(const void *data, int len) {
+    return __profileFile->write((const char *)data, len);
+}
+
+#ifdef __PROFILE
+extern "C" void runtime_init_setup_profiling();
+#define PICO_RUNTIME_INIT_PROFILING "11011" // Towards the end, after PSRAM
+PICO_RUNTIME_INIT_FUNC_RUNTIME(runtime_init_setup_profiling, PICO_RUNTIME_INIT_PROFILING);
+#endif
+
+#endif
diff --git a/cores/rp2040/RP2040Support.h b/cores/rp2040/RP2040Support.h
@@ -18,6 +18,8 @@
     Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
 */
 
+#pragma once
+
 #include <hardware/clocks.h>
 #include <hardware/irq.h>
 #include <hardware/pio.h>
@@ -45,6 +47,13 @@
 
 extern "C" volatile bool __otherCoreIdled;
 
+extern "C" {
+#ifdef __PROFILE
+    typedef int (*profileWriteCB)(const void *data, int len);
+    extern void _writeProfile(profileWriteCB writeCB);
+#endif
+}
+
 class _MFIFO {
 public:
     _MFIFO() { /* noop */ };
@@ -180,7 +189,7 @@ class RP2040 {
 
     void begin() {
         _epoch = 0;
-#if !defined(__riscv)
+#if !defined(__riscv) && !defined(__PROFILE)
         if (!__isFreeRTOS) {
             // Enable SYSTICK exception
             exception_set_exclusive_handler(SYSTICK_EXCEPTION, _SystickHandler);
@@ -193,7 +202,7 @@ class RP2040 {
             _ccountPgm->prepare(&_pio, &_sm, &off);
             ccount_program_init(_pio, _sm, off);
             pio_sm_set_enabled(_pio, _sm, true);
-#if !defined(__riscv)
+#if !defined(__riscv) && !defined(__PROFILE)
         }
 #endif
     }
@@ -217,7 +226,7 @@ class RP2040 {
     // Get CPU cycle count.  Needs to do magic to extens 24b HW to something longer
     volatile uint64_t _epoch = 0;
     inline uint32_t getCycleCount() {
-#if !defined(__riscv)
+#if !defined(__riscv) && !defined(__PROFILE)
         if (!__isFreeRTOS) {
             uint32_t epoch;
             uint32_t ctr;
@@ -229,13 +238,13 @@ class RP2040 {
         } else {
 #endif
             return ccount_read(_pio, _sm);
-#if !defined(__riscv)
+#if !defined(__riscv) && !defined(__PROFILE)
         }
 #endif
     }
 
     inline uint64_t getCycleCount64() {
-#if !defined(__riscv)
+#if !defined(__riscv) && !defined(__PROFILE)
         if (!__isFreeRTOS) {
             uint64_t epoch;
             uint64_t ctr;
@@ -247,7 +256,7 @@ class RP2040 {
         } else {
 #endif
             return ccount_read(_pio, _sm);
-#if !defined(__riscv)
+#if !defined(__riscv) && !defined(__PROFILE)
         }
 #endif
     }
@@ -473,6 +482,21 @@ class RP2040 {
 #endif
     }
 
+#ifdef __PROFILE
+    void writeProfiling(Stream *f) {
+        extern Stream *__profileFile;
+        extern int __writeProfileCB(const void *data, int len);
+        __profileFile = f;
+        _writeProfile(__writeProfileCB);
+    }
+
+    size_t getProfileMemoryUsage() {
+        extern int __profileMemSize;
+        return (size_t) __profileMemSize;
+    }
+#endif
+
+
 
 private:
     static void _SystickHandler() {

diff --git a/cores/rp2040/_freertos.h b/cores/rp2040/_freertos.h
@@ -30,32 +30,36 @@ extern bool __isFreeRTOS;
 // FreeRTOS has been set up
 extern volatile bool __freeRTOSinitted;
 
+#ifdef __cplusplus
 extern "C" {
-    struct QueueDefinition; /* Using old naming convention so as not to break kernel aware debuggers. */
-    typedef struct QueueDefinition   * QueueHandle_t;
-    typedef QueueHandle_t SemaphoreHandle_t;
-    typedef int32_t BaseType_t;
+#endif // __cplusplus
+struct QueueDefinition; /* Using old naming convention so as not to break kernel aware debuggers. */
+typedef struct QueueDefinition   * QueueHandle_t;
+typedef QueueHandle_t SemaphoreHandle_t;
+typedef int32_t BaseType_t;
 
-    extern bool __freertos_check_if_in_isr() __attribute__((weak));
+extern bool __freertos_check_if_in_isr() __attribute__((weak));
 
-    extern SemaphoreHandle_t __freertos_mutex_create() __attribute__((weak));
-    extern SemaphoreHandle_t _freertos_recursive_mutex_create() __attribute__((weak));
+extern SemaphoreHandle_t __freertos_mutex_create() __attribute__((weak));
+extern SemaphoreHandle_t _freertos_recursive_mutex_create() __attribute__((weak));
 
-    extern void __freertos_mutex_take(SemaphoreHandle_t mtx) __attribute__((weak));
+extern void __freertos_mutex_take(SemaphoreHandle_t mtx) __attribute__((weak));
 
-    extern int __freertos_mutex_take_from_isr(SemaphoreHandle_t mtx, BaseType_t* pxHigherPriorityTaskWoken) __attribute__((weak));
-    extern int __freertos_mutex_try_take(SemaphoreHandle_t mtx) __attribute__((weak));
-    extern void __freertos_mutex_give(SemaphoreHandle_t mtx) __attribute__((weak));
-    extern void __freertos_mutex_give_from_isr(SemaphoreHandle_t mtx, BaseType_t* pxHigherPriorityTaskWoken) __attribute__((weak));
+extern int __freertos_mutex_take_from_isr(SemaphoreHandle_t mtx, BaseType_t* pxHigherPriorityTaskWoken) __attribute__((weak));
+extern int __freertos_mutex_try_take(SemaphoreHandle_t mtx) __attribute__((weak));
+extern void __freertos_mutex_give(SemaphoreHandle_t mtx) __attribute__((weak));
+extern void __freertos_mutex_give_from_isr(SemaphoreHandle_t mtx, BaseType_t* pxHigherPriorityTaskWoken) __attribute__((weak));
 
-    extern void __freertos_recursive_mutex_take(SemaphoreHandle_t mtx) __attribute__((weak));
-    extern int __freertos_recursive_mutex_try_take(SemaphoreHandle_t mtx) __attribute__((weak));
-    extern void __freertos_recursive_mutex_give(SemaphoreHandle_t mtx) __attribute__((weak));
+extern void __freertos_recursive_mutex_take(SemaphoreHandle_t mtx) __attribute__((weak));
+extern int __freertos_recursive_mutex_try_take(SemaphoreHandle_t mtx) __attribute__((weak));
+extern void __freertos_recursive_mutex_give(SemaphoreHandle_t mtx) __attribute__((weak));
 
-    extern void __freertos_idle_other_core() __attribute__((weak));
-    extern void __freertos_resume_other_core() __attribute__((weak));
+extern void __freertos_idle_other_core() __attribute__((weak));
+extern void __freertos_resume_other_core() __attribute__((weak));
 
-    extern void __freertos_task_exit_critical() __attribute__((weak));
-    extern void __freertos_task_enter_critical() __attribute__((weak));
+extern void __freertos_task_exit_critical() __attribute__((weak));
+extern void __freertos_task_enter_critical() __attribute__((weak));
+#ifdef __cplusplus
 }
 extern SemaphoreHandle_t __get_freertos_mutex_for_ptr(mutex_t *m, bool recursive = false);
+#endif // __cplusplus