Optimize UC1698 and flush_cb()

Firmware/CoverUI/YardForce/CortexMCycleCounter.cpp

@@ -1,3 +1,4 @@
+#include <Arduino.h>
 #include <stdint.h>
 #include "include/CortexMCycleCounter.hpp"
 
@@ -22,13 +23,21 @@ CortexMCycleCounter::CortexMCycleCounter()
 
 void CortexMCycleCounter::start()
 {
+#ifdef STM32F030x8 // No DWT MCUs like STM32f0x
+    start_ = micros();
+#else // DWT capable MCU
     start_ = ARM_CM_DWT_CYCCNT;
+#endif
 }
 
 void CortexMCycleCounter::stop()
 {
+#ifdef STM32F030x8 // No DWT MCUs like STM32f0x
+    const uint32_t stop = micros();
+#else // DWT capable MCU
     const uint32_t stop = ARM_CM_DWT_CYCCNT;
-
+#endif
+
     current_ = stop - start_;
 
     // Min/Max

Firmware/CoverUI/YardForce/UC1698.cpp

@@ -23,12 +23,11 @@ namespace yardforce
             {
                 initConnection_();
 
-                // Reset via RST control line
-                digitalWrite(PIN_CS, 0); // Chip select
+                CLR_CS; // Chip select
                 digitalWrite(PIN_RST, 0);
                 delay(1);
                 digitalWrite(PIN_RST, 1);
-                digitalWrite(PIN_CS, 1);      // Chip deselect
+                SET_CS;                       // Chip deselect
                 delay(UC1698_RESET_DELAY_MS); // Wait for MTP read ready (after RST)
 
                 // Check if it's an UC1698
@@ -39,7 +38,14 @@ namespace yardforce
                     return false;
                 }
 
+#ifdef BENCHMARK
+                cycle_cnt_init_.start();
+#endif
                 initDisplay_();
+#ifdef BENCHMARK
+                cycle_cnt_init_.stop();
+#endif
+
                 return true;
             }
 
@@ -70,13 +76,13 @@ namespace yardforce
             {
                 static uint8_t status[UC1698_GET_STATUS_SIZE];
 
-                digitalWrite(PIN_CS, 0); // Chip select
-                digitalWrite(PIN_CD, 0); // Command
+                CLR_CS; // Chip select
+                CLR_CD; // Command
                 for (unsigned int i = 0; i < UC1698_GET_STATUS_SIZE; i++)
                 {
                     status[i] = read_();
                 }
-                digitalWrite(PIN_CS, 1); // Chip deselect
+                SET_CS; // Chip deselect
 
                 return status;
             }
@@ -149,47 +155,47 @@ namespace yardforce
             // Write command
             void UC1698::writeCommand(uint8_t data)
             {
-                digitalWrite(PIN_CS, 0); // Chip select
-                digitalWrite(PIN_CD, 0); // Command
+                CLR_CS; // Chip select
+                CLR_CD; // Command
                 writeSeq_(data);
-                digitalWrite(PIN_CS, 1); // Chip deselect
+                SET_CS; // Chip deselect
             }
 
             void UC1698::writeCommand(const uint8_t *data_array, unsigned int length)
             {
-                digitalWrite(PIN_CS, 0); // Chip select
-                digitalWrite(PIN_CD, 0); // Command
-                // while (length > 0)
+                CLR_CS; // Chip select
+                CLR_CD; // Command
                 for (; length > 0; length--, data_array++)
                 {
                     writeSeq_(*data_array);
                 }
-                digitalWrite(PIN_CS, 1); // Chip deselect
+                SET_CS; // Chip deselect
             }
 
             void UC1698::writeData(uint16_t data)
             {
-                digitalWrite(PIN_CS, 0); // Chip select
-                digitalWrite(PIN_CD, 1); // Data
-
+                CLR_CS; // Chip select
+                SET_CD; // Data
                 uint8_t part = (uint8_t)(data >> 8);
                 writeSeq_(part);
 
                 part = (uint8_t)data;
                 writeSeq_(part);
 
-                digitalWrite(PIN_CS, 1); // Chip deselect
+                SET_CS; // Chip deselect
             }
 
             void UC1698::writeData(const uint8_t *data_array, unsigned int length)
             {
-                digitalWrite(PIN_CS, 0); // Chip select
-                digitalWrite(PIN_CD, 1); // Data
+                CLR_CS; // Chip select
+                SET_CD; // Data
                 for (; length > 0; length--, data_array++)
                 {
                     writeSeq_(*data_array);
+                    NOOP;
+                    NOOP; // >=45ns. This NOOPs = 2*21ns + the next cycle. See Thpw80 of UC1698 datasheet page 60
                 }
-                digitalWrite(PIN_CS, 1); // Chip deselect
+                SET_CS; // Chip deselect
             }
 
             // Initialize control and data lines
@@ -205,7 +211,7 @@ namespace yardforce
                 digitalWrite(PIN_WR, 1);
                 pinMode(PIN_CD, OUTPUT);
                 digitalWrite(PIN_CD, 1);
-                GPIO_DATA_MODE_INPUT;
+                GPIO_DATA_MODE_OUTPUT;
             }
 
             // Initialize control and data lines
@@ -219,22 +225,10 @@ namespace yardforce
                 writeCommand(0b11010000 | 1); // [20] Set Color Pattern to R-G-B
                 writeCommand(0b11010110);     // [21] Set Color Mode to RRRRR-GGGGGG-BBBBB, 64k-color
 
-                fillScreen(false); // Clear Screen
+                // fillScreen(true); // Clear Screen
                 setDisplayEnable(true);
             }
 
-            // NOP delay to fullfil ns timing requirements. Only useful for values > 100ns, otherwise use single NOOP's
-            void UC1698::nopDelay_(unsigned int t_ns)
-            {
-                int16_t i = t_ns / NOP_CYCLE_NS;
-                i -= 6; // Already consumed (or more)
-
-                for (; i >= 0; i--)
-                {
-                    NOOP;
-                }
-            }
-
             // [2*] Read data from port
             uint8_t UC1698::read_()
             {
@@ -245,19 +239,25 @@ namespace yardforce
                 NOOP;
                 uint8_t data = ((GPIO_TypeDef *)GPIO_DATA)->IDR & GPIO_DATA_MASK;
                 digitalWrite(PIN_RD, 1); // Read end
+                GPIO_DATA_MODE_OUTPUT;
 
                 return data;
             }
 
             // Write data sequence
             void UC1698::writeSeq_(uint8_t data)
             {
-                GPIO_DATA_MODE_OUTPUT;
+#ifdef BENCHMARK
+                cycle_cnt_writeSeq_.start();
+#endif
 
-                digitalWrite(PIN_WR, 0); // Trigger write
+                CLR_WR; // Trigger write
                 ((GPIO_TypeDef *)GPIO_DATA)->ODR = (((GPIO_TypeDef *)GPIO_DATA)->ODR & ~GPIO_DATA_MASK) | data;
-                NOOP;                    // >=30ns. This NOOP = 21ns + the next cycle for the rising edge. See Tds80 of UC1698 datasheet page 60
-                digitalWrite(PIN_WR, 1); // Write end
+                NOOP;   // >=30ns. This NOOP = 21ns + the next cycle for the rising edge. See Tds80 of UC1698 datasheet page 60
+                SET_WR; // Write end
+#ifdef BENCHMARK
+                cycle_cnt_writeSeq_.stop();
+#endif
             }
         } // namespace controller
     }     // namespace display

Firmware/CoverUI/YardForce/include/CortexMCycleCounter.hpp

@@ -24,13 +24,13 @@ class CortexMCycleCounter
     void stop();  // Stop  measurement
 
 private:
-    unsigned int start_ = 0;
-    unsigned int current_ = 0;
-    unsigned int min_ = 0xffffffff;
-    unsigned int max_ = 0;
-    unsigned int avg_ = 0;
-    unsigned int sum_ = 0;
-    unsigned int cnt_ = 0;
+    unsigned long start_ = 0;
+    unsigned long current_ = 0;
+    unsigned long min_ = 0xffffffff;
+    unsigned long max_ = 0;
+    unsigned long avg_ = 0;
+    unsigned long sum_ = 0;
+    unsigned long cnt_ = 0;
     static bool s_dwt_initialized_;
 }; // CortexMCycleCounter
 

Firmware/CoverUI/YardForce/include/UC1698.hpp

@@ -15,21 +15,29 @@
 
 #include <Arduino.h>
 
-#define PIN_CS PC7   // Chip Select pin
 #define PIN_RST PA12 // ReSTart pin
 #define PIN_RD PC9   // Read control line
-#define PIN_WR PC8   // Write control line
-#define PIN_CD PA8   // Command/Data (A0)
+
+#define PIN_CS PC7                                                          // Chip Select pin
+#define SET_CS GPIOC->ODR = GPIOC->ODR | 0b10000000                         // Set CS (PC7) = high
+#define CLR_CS GPIOC->ODR = GPIOC->ODR & 0b11111111111111111111111101111111 // Clear CS (PC7) = low
+
+#define PIN_CD PA8                                                          // Command/Data (A0)
+#define SET_CD GPIOA->ODR = GPIOA->ODR | 0b100000000                        // Set CD (PA8) = high
+#define CLR_CD GPIOA->ODR = GPIOA->ODR & 0b11111111111111111111111011111111 // Clear CD (PA8) = low
+
+#define PIN_WR PC8                                                          // Write control line. Only for Mode set
+#define SET_WR GPIOC->ODR = GPIOC->ODR | 0b100000000                        // Set WR (PC8) = high
+#define CLR_WR GPIOC->ODR = GPIOC->ODR & 0b11111111111111111111111011111111 // Clear WR (PC8) = low
 
 #define GPIO_DATA GPIOB_BASE  // GPIO port/base of the data lines. Assumed that D0-D7 are wired to ONE port!
 #define GPIO_DATA_MASK 0x00ff // Mask which forms the data. Assumed that D0-D7 are all connected consecutive via LSB 0
 
 // Shorthands to set GPIO Mode data pins to input/output
-#define GPIO_DATA_MODE_INPUT ((GPIO_TypeDef *)GPIO_DATA)->MODER = (((GPIO_TypeDef *)GPIO_DATA)->MODER & 0xffff0000) & 0xffff0000;         // Mode 0 = Input
-#define GPIO_DATA_MODE_OUTPUT ((GPIO_TypeDef *)GPIO_DATA)->MODER = (((GPIO_TypeDef *)GPIO_DATA)->MODER & 0xffff0000) | 0b0101010101010101 // Mode 01 = Output
+#define GPIO_DATA_MODE_INPUT ((GPIO_TypeDef *)GPIO_DATA)->MODER = (((GPIO_TypeDef *)GPIO_DATA)->MODER & 0xffff0000) & 0xffff0000;         // D0-D7 Mode 0 = Input
+#define GPIO_DATA_MODE_OUTPUT ((GPIO_TypeDef *)GPIO_DATA)->MODER = (((GPIO_TypeDef *)GPIO_DATA)->MODER & 0xffff0000) | 0b0101010101010101 // D0-D7 Mode 01 = Output
 
-#define NOP_CYCLE_NS 21 // For STM32F030x it's 1/48MHz which is 20.9ns (but we use int for not wasting time in slow FP calculations)
-#define NOOP asm("nop\n")
+#define NOOP asm("nop\n") // For STM32F030x it's 1/48MHz which is 20.9ns
 
 #define UC1698_RESET_DELAY_MS 150
 #define UC1698_GET_STATUS_SIZE 3
@@ -41,6 +49,13 @@
 #define UC1698_DISPLAY_HEIGHT 128
 #endif
 
+// Enable for benchmarking specific code
+// #define BENCHMARK
+
+#ifdef BENCHMARK
+#include "include/CortexMCycleCounter.hpp"
+#endif
+
 namespace yardforce
 {
     namespace display
@@ -58,6 +73,9 @@ namespace yardforce
                 void drawPixelTriplet(bool pixel1State, bool pixel2State, bool pixel3State);
                 void fillScreen(bool t_color_black); // Fill screen (t_color: false = white, true = black)
 
+                void writeData(uint16_t data);
+                void writeData(const uint8_t *data_array, unsigned int length);
+
             private:
                 void initConnection_(); // Initialize control and data lines
                 void initDisplay_();    // Initialize control and data lines
@@ -72,13 +90,13 @@ namespace yardforce
                 void writeCommand(uint8_t data);
                 void writeCommand(const uint8_t *data_array, unsigned int length);
 
-                void writeData(uint16_t data);
-                void writeData(const uint8_t *data_array, unsigned int length);
-
-                void nopDelay_(unsigned int t_ns); // NOP delay to fullfil ns timing requirements. Only useful for values > 100ns, otherwise use single NOOP's
-
                 uint8_t read_();
                 void writeSeq_(uint8_t data);
+
+#ifdef BENCHMARK
+                CortexMCycleCounter cycle_cnt_init_;
+                CortexMCycleCounter cycle_cnt_writeSeq_;
+#endif
             };
         } // namespace controller
     }     // namespace display

Firmware/CoverUI/YardForce/include/model/Display_SAxPRO.cpp

@@ -60,20 +60,30 @@ namespace yardforce
          */
         void DisplaySAXPRO::flush_cb(lv_disp_drv_t *disp_drv, const lv_area_t *area, lv_color_t *t_color_p)
         {
-            unsigned int x, y;
-            lv_color_t *color_p;
+#ifdef BENCHMARK
+            cycle_cnt_flush_cb_.start();
+#endif
+            size_t cols16b = ((area->x2 - area->x1) + 1) / 3; // Num of 16 bit columns for this flush area
+            size_t cols8b = cols16b * 2;
+            uint8_t row_buffer[cols8b]; // Buffer for one row 2*8-bit (every 16 bit col holds 3 pixel)
+
             uc1698.setWindowProgramArea(area->x1, area->x2, area->y1, area->y2);
 
-            for (y = area->y1; y <= area->y2; y++)
+            for (size_t y = area->y1; y <= area->y2; y++)
             {
-                for (x = area->x1; x <= area->x2; x += 3) // FIXME: Might overflow buffer if area->x2 is not dividable by 3
+                for (size_t x = 0; x < cols8b; x += 2) // ATTENTION: Will only work with a correct rounder_cb()!
                 {
                     // Color is inverted (0 = black but pixel off / >0 = white but pixel on) but UC1698 "[16] Set Inverse Display" is set
-                    uc1698.drawPixelTriplet(t_color_p->full, (t_color_p + 1)->full, (t_color_p + 2)->full);
+                    row_buffer[x] = (0b11111000 * t_color_p->full | (0b00000111 * (t_color_p + 1)->full));
+                    row_buffer[x + 1] = (0b11100000 * (t_color_p + 1)->full) | (0b00011111 * (t_color_p + 2)->full);
                     t_color_p += 3;
                 }
+                uc1698.writeData(row_buffer, cols8b);
             }
             lv_disp_flush_ready(disp_drv);
+#ifdef BENCHMARK
+            cycle_cnt_flush_cb_.stop();
+#endif
         }
 
         void DisplaySAXPRO::set_undocked()
@@ -87,6 +97,9 @@ namespace yardforce
 
         void DisplaySAXPRO::mainScreen()
         {
+#ifdef BENCHMARK
+            volatile auto perf_test = cycle_cnt_flush_cb_;
+#endif
             // On the left side of the status bar we do have functional status symbols like heartbeat and ROS
             v_led_ros = new lvgl::WidgetLedSymbol(FA_SYMBOL_ROS, LV_ALIGN_TOP_LEFT, 0, 0); // Leftmost
 

Firmware/CoverUI/YardForce/include/model/Display_SAxPRO.hpp

@@ -31,6 +31,13 @@
 #define BACKLIGHT_TIMEOUT_MS 120000 // 2 minutes
 #define STATUS_TICKER_LENGTH 100
 
+// Enable for benchmarking specific code
+// #define BENCHMARK
+
+#ifdef BENCHMARK
+#include "include/CortexMCycleCounter.hpp"
+#endif
+
 namespace yardforce
 {
     namespace display
@@ -57,6 +64,11 @@ namespace yardforce
 
             static void set_undocked();
         };
+
+#ifdef BENCHMARK
+        static CortexMCycleCounter cycle_cnt_flush_cb_;
+#endif
+
     } // namespace display
 } // namespace yardforce