Replace esp_delay_us() with ISR-friendly code in IRAM

Grbl_Esp32/src/NutsBolts.h

@@ -24,6 +24,8 @@
 // #define true 1
 
 #include <cstdint>
+#include <esp_attr.h>
+#include <xtensa/core-macros.h>
 
 enum class DwellMode : uint8_t {
     Dwell      = 0,  // (Default: Must be zero)
@@ -93,6 +95,9 @@ const float INCH_PER_MM = (0.0393701f);
 // a pointer to the result variable. Returns true when it succeeds
 uint8_t read_float(const char* line, uint8_t* char_counter, float* float_ptr);
 
+// Blocking delay for very short time intervals
+void delay_us(int32_t microseconds);
+
 // Non-blocking delay function used for general operation and suspend features.
 bool delay_msec(int32_t milliseconds, DwellMode mode);
 
@@ -118,3 +123,38 @@ void swap(T& a, T& b) {
     a = b;
     b = c;
 }
+
+// Short delays measured using the CPU cycle counter.  There is a ROM
+// routine "esp_delay_us(us)" that almost does what what we need,
+// except that it is in ROM and thus dodgy for use from ISRs.  We
+// duplicate the esp_delay_us() here, but placed in IRAM, inlined,
+// and factored so it can be used in different ways.
+
+inline int32_t IRAM_ATTR getCpuTicks() {
+    return XTHAL_GET_CCOUNT();
+}
+
+extern uint32_t g_ticks_per_us_pro;  // For CPU 0 - typically 240 MHz
+extern uint32_t g_ticks_per_us_app;  // For CPU 1 - typically 240 MHz
+
+inline int32_t IRAM_ATTR usToCpuTicks(int32_t us) {
+    return us * g_ticks_per_us_pro;
+}
+
+inline int32_t IRAM_ATTR usToEndTicks(int32_t us) {
+    return getCpuTicks() + usToCpuTicks(us);
+}
+
+// At the usual ESP32 clock rate of 240MHz, the range of this is
+// just under 18 seconds, but it really should be used only for
+// short delays up to a few tens of microseconds.
+
+inline void IRAM_ATTR spinUntil(int32_t endTicks) {
+    while ((XTHAL_GET_CCOUNT() - endTicks) < 0) {
+        asm volatile("nop");
+    }
+}
+
+inline void IRAM_ATTR delay_us(int32_t us) {
+    spinUntil(usToEndTicks(us));
+}

Grbl_Esp32/src/Spindles/OnOffSpindle.cpp

@@ -52,7 +52,7 @@ namespace Spindles {
         }
         set_output(dev_speed);
         set_enable(state != SpindleState::Disable);
-        spinDelay(state, speed);
+        spindleDelay(state, speed);
 
         sys.report_ovr_counter = 0;  // Set to report change immediately
     }

Grbl_Esp32/src/Spindles/PWMSpindle.cpp

@@ -112,7 +112,7 @@ namespace Spindles {
         // converters on some boards.
         set_output(dev_speed);
         set_enable(state != SpindleState::Disable);
-        spinDelay(state, speed);
+        spindleDelay(state, speed);
 
         sys.report_ovr_counter = 0;  // Set to report change immediately
     }

Grbl_Esp32/src/Spindles/Spindle.cpp

@@ -140,7 +140,7 @@ namespace Spindles {
         // log_debug("rpm " << speed << " speed " << dev_speed); // This will spew quite a bit of data on your output
         return dev_speed;
     }
-    void Spindle::spinDelay(SpindleState state, SpindleSpeed speed) {
+    void Spindle::spindleDelay(SpindleState state, SpindleSpeed speed) {
         uint32_t up = 0, down = 0;
         switch (state) {
             case SpindleState::Unknown:

Grbl_Esp32/src/Spindles/Spindle.h

@@ -59,7 +59,7 @@ namespace Spindles {
 
         static void switchSpindle(uint8_t new_tool, SpindleList spindles, Spindle*& spindle);
 
-        void         spinDelay(SpindleState state, SpindleSpeed speed);
+        void         spindleDelay(SpindleState state, SpindleSpeed speed);
         virtual void init() = 0;  // not in constructor because this also gets called when $$ settings change
 
         // Used by Protocol.cpp to restore the state during a restart

Grbl_Esp32/src/Spindles/VFDSpindle.cpp

@@ -345,7 +345,7 @@ namespace Spindles {
             }
         }
         if (!supports_actual_speed()) {
-            spinDelay(state, speed);
+            spindleDelay(state, speed);
         } else {
             // _sync_dev_speed is set by a callback that handles
             // responses from periodic get_current_speed() requests.
@@ -386,7 +386,7 @@ namespace Spindles {
             }
 
             _syncing = false;
-            // spinDelay() sets these when it is used
+            // spindleDelay() sets these when it is used
             _current_state = state;
             _current_speed = speed;
         }

Grbl_Esp32/src/Stepping.cpp

@@ -73,13 +73,6 @@ namespace Machine {
             _engine = I2S_STREAM;
         }
     }
-    void IRAM_ATTR Stepping::spinDelay(int64_t start_time, uint32_t durationUs) {
-        int64_t endTime = start_time + durationUs;
-        while ((esp_timer_get_time() - endTime) < 0) {
-            NOP();
-        }
-    }
-
     void IRAM_ATTR Stepping::waitPulse() {
         uint64_t pulseEndTime;
         switch (_engine) {
@@ -90,7 +83,7 @@ namespace Machine {
             case I2S_STATIC:
                 i2s_out_push();
             case TIMED:
-                spinDelay(_stepPulseStartTime, _pulseUsecs);
+                spinUntil(_stepPulseEndTime);
                 break;
             case RMT:
                 // RMT generates the trailing edges in hardware
@@ -112,7 +105,7 @@ namespace Machine {
                     // If we are using GPIO stepping as opposed to RMT, record the
                     // time that we turned on the direction pins so we can delay a bit.
                     // If we are using RMT, we can't delay here.
-                    spinDelay(esp_timer_get_time(), _directionDelayUsecs);
+                    delay_us(_directionDelayUsecs);
                     break;
                 case stepper_id_t::RMT:
                     break;
@@ -126,7 +119,7 @@ namespace Machine {
                 break;
             case stepper_id_t::I2S_STATIC:
             case stepper_id_t::TIMED:
-                _stepPulseStartTime = esp_timer_get_time();
+                _stepPulseEndTime = usToEndTicks(_pulseUsecs);
                 break;
             case stepper_id_t::RMT:
                 break;

Grbl_Esp32/src/Stepping.h

@@ -35,9 +35,7 @@ namespace Machine {
         static const int      ticksPerMicrosecond = fStepperTimer / 1000000;
 
         bool    _switchedStepper = false;
-        int64_t _stepPulseStartTime;
+        int32_t _stepPulseEndTime;
-
-        static void spinDelay(int64_t start_time, uint32_t duration);
 
     public:
         // Counts stepper ISR invocations.  This variable can be inspected