Update version

RP2040::memcpyDMA implements a DMA-controlled memory copy call identical in
function to standard memcpy, but using an onboard DMA engine.  For large
memory transfers this can be significantly faster than using the CPU-based
memcpy.  Only 4-byte aligned source, destination, and counts are allowed.
If any inputs are not 4-byte aligned, then standard memcpy will occur so
it will behave correctly for any inputs.

cores/rp2040/RP2040Support.h

@@ -27,6 +27,7 @@
 #include <hardware/structs/rosc.h>
 #include <hardware/structs/systick.h>
 #include <pico/multicore.h>
+#include <hardware/dma.h>
 #include <pico/rand.h>
 #include <pico/util/queue.h>
 #include <pico/bootrom.h>
@@ -311,6 +312,36 @@ public:
         return id;
     }
 
+#pragma GCC push_options
+#pragma GCC optimize ("Os")
+    void *memcpyDMA(void *dest, const void *src, size_t n) {
+        // Allocate a DMA channel on 1st call, reuse it every call after
+        if (memcpyDMAChannel < 1) {
+            memcpyDMAChannel = dma_claim_unused_channel(true);
+            dma_channel_config c = dma_channel_get_default_config(memcpyDMAChannel);
+            channel_config_set_transfer_data_size(&c, DMA_SIZE_32);
+            channel_config_set_read_increment(&c, true);
+            channel_config_set_write_increment(&c, true);
+            channel_config_set_irq_quiet(&c, true);
+            dma_channel_set_config(memcpyDMAChannel, &c, false);
+        }
+        // If there's any misalignment or too small, use regular memcpy which can handle it
+        if ((n < 64) || (((uint32_t)dest) | ((uint32_t)src) | n) & 3) {
+            return memcpy(dest, src, n);
+        }
+
+        int words = n / 4;
+        dma_channel_set_read_addr(memcpyDMAChannel, src, false);
+        dma_channel_set_write_addr(memcpyDMAChannel, dest, false);
+        dma_channel_set_trans_count(memcpyDMAChannel, words, false);
+        dma_channel_start(memcpyDMAChannel);
+        while (dma_channel_is_busy(memcpyDMAChannel)) {
+            /* busy wait dma */
+        }
+        return dest;
+    }
+#pragma GCC pop_options
+
     // Multicore comms FIFO
     _MFIFO fifo;
 
@@ -336,4 +367,5 @@ private:
     PIO _pio;
     int _sm;
     PIOProgram *_ccountPgm;
+    int memcpyDMAChannel = -1;
 };

cores/rp2040/RP2040Version.h

@@ -1,5 +1,5 @@
 #pragma once
 #define ARDUINO_PICO_MAJOR 3
 #define ARDUINO_PICO_MINOR 9
-#define ARDUINO_PICO_REVISION 4
-#define ARDUINO_PICO_VERSION_STR "3.9.4"
+#define ARDUINO_PICO_REVISION 5
+#define ARDUINO_PICO_VERSION_STR "3.9.5"

cores/rp2040/lwip_wrap.cpp

@@ -247,19 +247,19 @@ extern "C" {
     extern void __real_tcp_setprio(struct tcp_pcb *pcb, u8_t prio);
     void __wrap_tcp_setprio(struct tcp_pcb *pcb, u8_t prio) {
         LWIPMutex m;
-        return __real_tcp_setprio(pcb, prio);
+        __real_tcp_setprio(pcb, prio);
     }
 
     extern void __real_tcp_backlog_delayed(struct tcp_pcb* pcb);
     void __wrap_tcp_backlog_delayed(struct tcp_pcb* pcb) {
         LWIPMutex m;
-        return __real_tcp_backlog_delayed(pcb);
+        __real_tcp_backlog_delayed(pcb);
     }
 
     extern void __real_tcp_backlog_accepted(struct tcp_pcb* pcb);
     void __wrap_tcp_backlog_accepted(struct tcp_pcb* pcb) {
         LWIPMutex m;
-        return __real_tcp_backlog_accepted(pcb);
+        __real_tcp_backlog_accepted(pcb);
     }
     extern struct udp_pcb *__real_udp_new(void);
     struct udp_pcb *__wrap_udp_new(void) {

docs/bearssl-client-secure-class.rst

@@ -120,7 +120,7 @@ Valid values for min and max are `BR_TLS10`, `BR_TLS11`, `BR_TLS12`.  Min and ma
 
 
 ESP32 Compatibility
-===================
+~~~~~~~~~~~~~~~~~~~
 Simple ESP32 ``WiFiClientSecure`` compatibility is built-in, allow for some sketches to run without any modification.
 The following methods are implemented:
 

docs/conf.py

@@ -54,9 +54,9 @@ author = u'Earle F. Philhower, III'
 # built documents.
 #
 # The short X.Y version.
-version = u'3.9.4'
+version = u'3.9.5'
 # The full version, including alpha/beta/rc tags.
-release = u'3.9.4'
+release = u'3.9.5'
 
 # The language for content autogenerated by Sphinx. Refer to documentation
 # for a list of supported languages.

docs/rp2040.rst

@@ -99,3 +99,13 @@ bootloader.
 void rp2040.rebootToBootloader()
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 Will reboot the RP2040 into USB UF2 upload mode.
+
+DMA-based MEMCPY
+----------------
+The onboard DMA engines can copy 4-byte aligned quantities faster than the CPU.
+
+void \*rp2040.memcpyDMA(void \*dest, const void \*src, size_t count);
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+Uses a DMA engine to transfer data from src to dest in 4-byte chunks without CPU
+intervention.  If any arguments are not 4-byte aligned, or if the count is not a
+multiple of 4, then it will fall back to CPU-managed memcpy.

include/bearssl/bearssl_git.h

@@ -1,2 +1,2 @@
 // Do not edit -- Automatically generated by tools/sdk/ssl/bearssl/Makefile
-#define BEARSSL_GIT f294aa0
+#define BEARSSL_GIT 5166f2b

lib/platform_wrap.txt

@@ -166,6 +166,7 @@
 
 -Wl,--wrap=tcp_arg
 -Wl,--wrap=tcp_new
+-Wl,--wrap=tcp_bind
 -Wl,--wrap=tcp_listen
 -Wl,--wrap=tcp_listen_with_backlog
 -Wl,--wrap=tcp_accept

libraries/FatFSUSB/src/FatFSUSB.h

@@ -19,6 +19,10 @@
 
 #pragma once
 
+#ifdef USE_TINYUSB
+#error FatFSUSB is not compatible with Adafruit TinyUSB
+#endif
+
 #include <Arduino.h>
 
 class FatFSUSBClass {

libraries/Joystick/src/Joystick.h

@@ -18,11 +18,14 @@
     Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
 */
 
-#ifndef _JOYSTICK_h
-#define _JOYSTICK_h
+#pragma once
+
+#ifdef USE_TINYUSB
+#error Joystick is not compatible with Adafruit TinyUSB
+#endif
 
 #include <HID_Joystick.h>
-#include "class/hid/hid.h"
+#include <class/hid/hid.h>
 
 //======================================================================
 class Joystick_ : public HID_Joystick {
@@ -31,5 +34,3 @@ public:
     virtual void send_now(void) override;
 };
 extern Joystick_ Joystick;
-
-#endif /* _JOYSTICK_h */

libraries/Keyboard/src/Keyboard.h

@@ -20,8 +20,11 @@
     Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
 */
 
-#ifndef KEYBOARD_h
-#define KEYBOARD_h
+#pragma once
+
+#ifdef USE_TINYUSB
+#error Keyboard is not compatible with Adafruit TinyUSB
+#endif
 
 #include <HID_Keyboard.h>
 
@@ -34,5 +37,3 @@ public:
     Keyboard_(void);
 };
 extern Keyboard_ Keyboard;
-
-#endif

libraries/Mouse/src/Mouse.h

@@ -19,8 +19,11 @@
     Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
 */
 
-#ifndef MOUSE_h
-#define MOUSE_h
+#pragma once
+
+#ifdef USE_TINYUSB
+#error Mouse is not compatible with Adafruit TinyUSB
+#endif
 
 #include <HID_Mouse.h>
 
@@ -30,5 +33,3 @@ public:
     virtual void move(int x, int y, signed char wheel = 0) override;
 };
 extern Mouse_ Mouse;
-
-#endif

libraries/MouseAbsolute/src/MouseAbsolute.h

@@ -1,5 +1,5 @@
 /*
-    Mouse.h
+    MouseAbsolute.h
 
     Copyright (c) 2015, Arduino LLC
     Original code (pre-library): Copyright (c) 2011, Peter Barrett
@@ -19,8 +19,11 @@
     Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
 */
 
-#ifndef MOUSEABSOLUTE_h
-#define MOUSEABSOLUTE_h
+#pragma once
+
+#ifdef USE_TINYUSB
+#error MouseAbsolute is not compatible with Adafruit TinyUSB
+#endif
 
 #include <HID_Mouse.h>
 
@@ -30,5 +33,3 @@ public:
     virtual void move(int x, int y, signed char wheel = 0) override;
 };
 extern MouseAbsolute_ MouseAbsolute;
-
-#endif

libraries/SingleFileDrive/src/SingleFileDrive.h

@@ -19,6 +19,10 @@
 
 #pragma once
 
+#ifdef USE_TINYUSB
+#error SingleFileDrive is not compatible with Adafruit TinyUSB
+#endif
+
 #include <Arduino.h>
 
 class SingleFileDrive {

libraries/rp2040/examples/DMAMemcpy/DMAMemcpy.ino

@@ -0,0 +1,51 @@
+// Simple speed and functionality test for DMA memcpy
+// Released to the public domain by Earle F. Philhower, III, 2024
+
+
+uint32_t src[1024];
+uint32_t dest[1024];
+
+void setup() {
+  // Set up a simple test pattern in src, verify after each pass
+  srand(0xc0de);
+  for (int i = 0; i < 1024; i++) {
+    src[i] = ((~i) & 1024) | ((i) << 22);
+  }
+}
+
+void verify(const char *name, uint32_t *src) {
+  for (int i = 0; i < 1024; i++) {
+    uint32_t expected = ((~i) & 1024) | ((i) << 22);
+    if (expected != src[i]) {
+      Serial.printf("ERROR, mismatch @ %d on %s memcpy\n", i, name);
+      while (true) {
+        // Idle forever, this is fatal!
+      }
+    }
+  }
+}
+
+void loop() {
+  uint64_t start, stop;
+
+  start = rp2040.getCycleCount();
+  for (int i = 0; i < 1000; i++) {
+    memcpy(dest, src, 4 * 1024);
+    memcpy(src, dest, 4 * 1024);
+  }
+  stop = rp2040.getCycleCount();
+  verify("CPU", src);
+  verify("CPU", dest);
+  Serial.printf("CPU: %lld clock cycles for 4K\n", (stop - start) / 1000);
+
+  start = rp2040.getCycleCount();
+  for (int i = 0; i < 1000; i++) {
+    rp2040.memcpyDMA(dest, src, 4 * 1024);
+    rp2040.memcpyDMA(src, dest, 4 * 1024);
+  }
+  stop = rp2040.getCycleCount();
+  verify("DMA", src);
+  verify("DMA", dest);
+  Serial.printf("DMA: %lld clock cycles for 4K\n\n\n", (stop - start) / 1000);
+  delay(1000);
+}

package.json

@@ -1,6 +1,6 @@
 {
   "name": "framework-arduinopico",
-  "version": "1.30904.0",
+  "version": "1.30905.0",
   "description": "Arduino Wiring-based Framework (RPi Pico RP2040)",
   "keywords": [
     "framework",

platform.txt

@@ -20,7 +20,7 @@
 # https://github.com/arduino/Arduino/wiki/Arduino-IDE-1.5---3rd-party-Hardware-specification
 
 name=Raspberry Pi RP2040 Boards
-version=3.9.4
+version=3.9.5
 
 # Required discoveries and monitors
 # ---------------------------------