[Core] Use polled waiting on ChibiOS platforms that support it (#17607)

* Use polled waiting on platforms that support it

Due to context switching overhead waiting a very short amount of time on
a sleeping thread is often not accurate and in fact not usable for timing
critical usage i.e. in a driver. Thus we use polled waiting for ranges
in the us range on platforms that support it instead. The fallback is
the thread sleeping mechanism.

This includes:

* ARM platforms with CYCCNT register (ARMv7, ARMv8) this is
  incremented at CPU clock frequency
* GD32VF103 RISC-V port with CSR_MCYCLE register this is incremented at
  CPU clock frequency
* RP2040 ARMv6 port which uses the integrated timer peripheral which is
  incremented with a fixed 1MHz frequency

* Use wait_us() instead of chSysPolledDelayX

...as it is powered by busy waiting now.

* Add chibios waiting methods test bench

keyboards/handwired/onekey/elite_c/config.h

@@ -30,3 +30,6 @@
 #define RGB_CI_PIN B1
 
 #define ADC_PIN F6
+
+#define QMK_WAITING_TEST_BUSY_PIN F6
+#define QMK_WAITING_TEST_YIELD_PIN F7

keyboards/handwired/onekey/keymaps/chibios_waiting_test/config.h

@@ -0,0 +1,12 @@
+// Copyright 2022 Stefan Kerkmann
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#if !defined(QMK_WAITING_TEST_BUSY_PIN)
+#    define QMK_WAITING_TEST_BUSY_PIN A8
+#endif
+
+#if !defined(QMK_WAITING_TEST_YIELD_PIN)
+#    define QMK_WAITING_TEST_YIELD_PIN A9
+#endif

keyboards/handwired/onekey/keymaps/chibios_waiting_test/keymap.c

@@ -0,0 +1,47 @@
+// Copyright 2022 Stefan Kerkmann
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include QMK_KEYBOARD_H
+
+const uint16_t PROGMEM keymaps[][MATRIX_ROWS][MATRIX_COLS] = {LAYOUT_ortho_1x1(KC_A)};
+
+#if defined(__AVR__)
+#    pragma message "AVR uses polled waiting by default, running theses tests will not show any difference"
+static inline void chThdSleepMicroseconds(uint32_t us) {
+    wait_us(us);
+}
+#endif
+
+void keyboard_post_init_user(void) {
+    setPinOutput(QMK_WAITING_TEST_BUSY_PIN);
+    setPinOutput(QMK_WAITING_TEST_YIELD_PIN);
+}
+
+static inline void wait_us_polling_with_strobe(uint32_t us) {
+    writePinHigh(QMK_WAITING_TEST_BUSY_PIN);
+    wait_us(us);
+    writePinLow(QMK_WAITING_TEST_BUSY_PIN);
+}
+
+static inline void wait_us_yield_with_strobe(uint32_t us) {
+    writePinHigh(QMK_WAITING_TEST_YIELD_PIN);
+    chThdSleepMicroseconds(us);
+    writePinLow(QMK_WAITING_TEST_YIELD_PIN);
+}
+
+static const uint32_t waiting_values[] = {0, 1, 5, 10, 25, 50, 100, 150, 200, 500, 1000};
+
+void housekeeping_task_user(void) {
+    static uint32_t last_bench = 0;
+    if (timer_elapsed32(last_bench) > 500) {
+        for (int i = 0; i < (sizeof(waiting_values) / sizeof(waiting_values[0])); i++) {
+            wait_us_polling_with_strobe(waiting_values[i]);
+            wait_us(10);
+        }
+        for (int i = 0; i < (sizeof(waiting_values) / sizeof(waiting_values[0])); i++) {
+            wait_us_yield_with_strobe(waiting_values[i]);
+            wait_us(10);
+        }
+        last_bench = timer_read32();
+    }
+}

keyboards/handwired/onekey/promicro/config.h

@@ -30,3 +30,6 @@
 #define RGB_CI_PIN B1
 
 #define ADC_PIN F6
+
+#define QMK_WAITING_TEST_BUSY_PIN F6
+#define QMK_WAITING_TEST_YIELD_PIN F7

keyboards/handwired/onekey/rp2040/config.h

@@ -6,10 +6,15 @@
 #include "config_common.h"
 
 #define PRODUCT Onekey Raspberry Pi RP2040
-#define MATRIX_COL_PINS { GP4 }
-#define MATRIX_ROW_PINS { GP5 }
+#define MATRIX_COL_PINS \
+    { GP4 }
+#define MATRIX_ROW_PINS \
+    { GP5 }
 #define DEBUG_MATRIX_SCAN_RATE
 
+#define QMK_WAITING_TEST_BUSY_PIN GP8
+#define QMK_WAITING_TEST_YIELD_PIN GP9
+
 #define RP2040_BOOTLOADER_DOUBLE_TAP_RESET
 #define RP2040_BOOTLOADER_DOUBLE_TAP_RESET_LED GP25
 #define RP2040_BOOTLOADER_DOUBLE_TAP_RESET_TIMEOUT 500U

keyboards/handwired/onekey/teensy_2/config.h

@@ -29,3 +29,6 @@
 #define RGB_DI_PIN F6
 
 #define ADC_PIN F6
+
+#define QMK_WAITING_TEST_BUSY_PIN F6
+#define QMK_WAITING_TEST_YIELD_PIN F7

keyboards/handwired/onekey/teensy_2pp/config.h

@@ -29,3 +29,6 @@
 #define RGB_DI_PIN F6
 
 #define ADC_PIN F6
+
+#define QMK_WAITING_TEST_BUSY_PIN F6
+#define QMK_WAITING_TEST_YIELD_PIN F7

keyboards/planck/rev6_drop/matrix.c

@@ -15,14 +15,7 @@
  * along with this program.  If not, see <http://www.gnu.org/licenses/>.
  */
 
-#include <stdint.h>
-#include <stdbool.h>
-#include <string.h>
-#include "hal.h"
-#include "timer.h"
-#include "wait.h"
-#include "debug.h"
-#include "matrix.h"
+#include "quantum.h"
 
 /*
  *     col: { B11, B10, B2, B1, A7, B0 }
@@ -38,9 +31,13 @@ __attribute__((weak)) void matrix_init_user(void) {}
 
 __attribute__((weak)) void matrix_scan_user(void) {}
 
-__attribute__((weak)) void matrix_init_kb(void) { matrix_init_user(); }
+__attribute__((weak)) void matrix_init_kb(void) {
+    matrix_init_user();
+}
 
-__attribute__((weak)) void matrix_scan_kb(void) { matrix_scan_user(); }
+__attribute__((weak)) void matrix_scan_kb(void) {
+    matrix_scan_user();
+}
 
 void matrix_init(void) {
     dprintf("matrix init\n");
@@ -146,9 +143,13 @@ uint8_t matrix_scan(void) {
     return 1;
 }
 
-bool matrix_is_on(uint8_t row, uint8_t col) { return (matrix[row] & (1 << col)); }
+bool matrix_is_on(uint8_t row, uint8_t col) {
+    return (matrix[row] & (1 << col));
+}
 
-matrix_row_t matrix_get_row(uint8_t row) { return matrix[row]; }
+matrix_row_t matrix_get_row(uint8_t row) {
+    return matrix[row];
+}
 
 void matrix_print(void) {
     dprintf("\nr/c 01234567\n");

keyboards/preonic/rev3_drop/matrix.c

@@ -15,14 +15,7 @@
  * along with this program.  If not, see <http://www.gnu.org/licenses/>.
  */
 
-#include <stdint.h>
-#include <stdbool.h>
-#include <string.h>
-#include "hal.h"
-#include "timer.h"
-#include "wait.h"
-#include "debug.h"
-#include "matrix.h"
+#include "quantum.h"
 
 typedef uint16_t matrix_col_t;
 
@@ -40,9 +33,13 @@ __attribute__((weak)) void matrix_init_user(void) {}
 
 __attribute__((weak)) void matrix_scan_user(void) {}
 
-__attribute__((weak)) void matrix_init_kb(void) { matrix_init_user(); }
+__attribute__((weak)) void matrix_init_kb(void) {
+    matrix_init_user();
+}
 
-__attribute__((weak)) void matrix_scan_kb(void) { matrix_scan_user(); }
+__attribute__((weak)) void matrix_scan_kb(void) {
+    matrix_scan_user();
+}
 
 void matrix_init(void) {
     dprintf("matrix init\n");
@@ -150,9 +147,13 @@ uint8_t matrix_scan(void) {
     return 1;
 }
 
-bool matrix_is_on(uint8_t row, uint8_t col) { return (matrix[row] & (1 << col)); }
+bool matrix_is_on(uint8_t row, uint8_t col) {
+    return (matrix[row] & (1 << col));
+}
 
-matrix_row_t matrix_get_row(uint8_t row) { return matrix[row]; }
+matrix_row_t matrix_get_row(uint8_t row) {
+    return matrix[row];
+}
 
 void matrix_print(void) {
     dprintf("\nr/c 01234567\n");

platforms/chibios/_wait.h

@@ -30,6 +30,11 @@
 
 #ifdef WAIT_US_TIMER
 void wait_us(uint16_t duration);
+#elif PORT_SUPPORTS_RT == TRUE
+#    define wait_us(us)                                            \
+        do {                                                       \
+            chSysPolledDelayX(US2RTC(REALTIME_COUNTER_CLOCK, us)); \
+        } while (0)
 #else
 #    define wait_us(us)                     \
         do {                                \

platforms/chibios/bootloaders/rp2040.c

@@ -43,9 +43,8 @@ void __late_init(void) {
     if (magic_location != magic_token) {
         magic_location = magic_token;
         // ChibiOS is not initialized at this point, so sleeping is only
-        // possible via busy waiting. The internal timer peripheral is running
-        // at this point with a precision of 1us.
-        chSysPolledDelayX(MS2RTC(1 * MHZ, RP2040_BOOTLOADER_DOUBLE_TAP_RESET_TIMEOUT));
+        // possible via busy waiting.
+        wait_us(RP2040_BOOTLOADER_DOUBLE_TAP_RESET_TIMEOUT * 1000U);
         magic_location = 0;
         return;
     }

platforms/chibios/chibios_config.h

@@ -21,6 +21,9 @@
 
 #if defined(MCU_RP)
 #    define CPU_CLOCK RP_CORE_CLK
+// ChibiOS uses the RP2040 timer peripheral as its real time counter, this timer
+// is monotonic and running at 1MHz.
+#    define REALTIME_COUNTER_CLOCK 1000000
 
 #    define USE_GPIOV1
 #    define PAL_OUTPUT_TYPE_OPENDRAIN _Static_assert(0, "RP2040 has no Open Drain GPIO configuration, setting this is not possible");
@@ -102,6 +105,11 @@
 #    endif
 #endif
 
+#if defined(MCU_MIMXRT1062)
+#    include "clock_config.h"
+#    define CPU_CLOCK BOARD_BOOTCLOCKRUN_CORE_CLOCK
+#endif
+
 #if defined(HT32)
 #    define CPU_CLOCK HT32_CK_SYS_FREQUENCY
 #    define PAL_MODE_ALTERNATE PAL_HT32_MODE_AF
@@ -109,3 +117,7 @@
 #    define PAL_OUTPUT_TYPE_PUSHPULL PAL_HT32_MODE_DIR
 #    define PAL_OUTPUT_SPEED_HIGHEST 0
 #endif
+
+#if !defined(REALTIME_COUNTER_CLOCK)
+#    define REALTIME_COUNTER_CLOCK CPU_CLOCK
+#endif

platforms/chibios/drivers/serial.c

@@ -20,7 +20,8 @@
 #    error "chSysPolledDelayX method not supported on this platform"
 #else
 #    undef wait_us
-#    define wait_us(x) chSysPolledDelayX(US2RTC(CPU_CLOCK, x))
+// Force usage of polled waiting - in case WAIT_US_TIMER is activated
+#    define wait_us(us) chSysPolledDelayX(US2RTC(REALTIME_COUNTER_CLOCK, us))
 #endif
 
 #ifndef SELECT_SOFT_SERIAL_SPEED

platforms/chibios/drivers/vendor/RP/RP2040/serial_vendor.c

@@ -152,7 +152,7 @@ static inline void enter_rx_state(void) {
     }
     // Wait for ~11 bits, 1 start bit + 8 data bits + 1 stop bit + 1 bit
     // headroom.
-    chSysPolledDelayX(US2RTC(1 * MHZ, (1000000U * 11 / SERIAL_USART_SPEED)));
+    wait_us(1000000U * 11U / SERIAL_USART_SPEED);
     // Disable tx state machine to not interfere with our tx pin manipulation
     pio_sm_set_enabled(pio, tx_state_machine, false);
     gpio_set_drive_strength(SERIAL_USART_TX_PIN, GPIO_DRIVE_STRENGTH_2MA);