optimize: rclk via pio

firmware/src/gfx_decoder.cpp

@@ -67,6 +67,12 @@ int32_t gfx_decoder_loadNextFrame() {
         }
         // insert placeholder for unused last stage (after pixel 19)
         sample >>=1;
+        // shift to LSB position
+        sample >>=8;
+        // MSB=1 indicates end of row
+        if (xModule == COL_MODULES - 1) {
+          sample |= 0x80000000;
+        }
 
         ledBuffer[bi][(ROW_COUNT - 1 - y) * COL_MODULES + xModule] = sample;
       }

firmware/src/leds.cpp

@@ -49,7 +49,8 @@ void leds_init() {
   pinMode(COL_SER, OUTPUT);
   pinMode(COL_OE, OUTPUT);
   outputEnable(ROW_OE, false);
-  pinMode(COL_RCLK, OUTPUT);
+  // pinMode(COL_RCLK, OUTPUT);
+  pinMode(RCLK, OUTPUT);
   pinMode(COL_SRCLK, OUTPUT);
   pinMode(COL_SRCLR, OUTPUT);
 
@@ -57,18 +58,18 @@ void leds_init() {
   pinMode(ROW_SER, OUTPUT);
   pinMode(ROW_OE, OUTPUT);
   outputEnable(ROW_OE, false);
-  pinMode(ROW_RCLK, OUTPUT);
+  // pinMode(ROW_RCLK, OUTPUT);
   pinMode(ROW_SRCLK, OUTPUT);
   pinMode(ROW_SRCLR, OUTPUT);
 
   // clear output - cols
   clearShiftReg(COL_SRCLK, COL_SRCLR);
-  pulsePin(COL_RCLK);
-  outputEnable(COL_OE, true);
+  pulsePin(RCLK);
+  outputEnable(COL_OE, true); // this is fine, because we control OE via rows only
 
   // clear output - rows
   clearShiftReg(ROW_SRCLK, ROW_SRCLR);
-  pulsePin(ROW_RCLK);
+  pulsePin(RCLK);
 }
 
 void leds_disable() {
@@ -127,11 +128,8 @@ void leds_render() {
         pulsePin(ROW_SRCLK);
       }
 
-      // set row data
-      // NOTE: values are loaded right-left
-      // Optimized implementation: use PIO, avoid division, modulo, etc...
-      // we use 7/8 stages of each shift register + 1 is unused so we need to do
-      // silly shit
+      // set row data using PIO
+      // latch signal is also sent here
       // TODO: Some ideas for future optimization:
       // - see if we can disable px pusher delays on improved electric interface
       // - improve outer loop which adds 2us of processing on each loop
@@ -151,12 +149,11 @@ void leds_render() {
       // pio_sm_drain_tx_fifo doesn't seem to do the trick
       // if not, we might need to use irqs or something
       busy_wait_us(4);
-
-      // latch rows and columns
-      gpio_set_mask(1 << ROW_RCLK | 1 << COL_RCLK);
-      _NOP();
-      _NOP();
-      gpio_clr_mask(1 << ROW_RCLK | 1 << COL_RCLK);
+      // TODO: Why doesn't this work?!
+      // while (!pio_interrupt_get(pusher_pio, pusher_sm)) {
+      //   tight_loop_contents();
+      // }
+      pio_interrupt_clear(pusher_pio, pusher_sm);
 
       // show for a certain period
       outputEnable(ROW_OE, true);
@@ -196,11 +193,16 @@ void leds_initPusher() {
   // data is inverted
   gpio_set_outover(dataPin, GPIO_OVERRIDE_INVERT);
 
-  // Set SET (latch) pin, connect to pad, set as output
+  // Set sideset (SRCLK) pin, connect to pad, set as output
   sm_config_set_sideset_pins(&config, latchPin);
   pio_gpio_init(pio, latchPin);
   pio_sm_set_consecutive_pindirs(pio, sm, latchPin, 1, true);
 
+  // Set SET (RCLK) pin, connect to pad, set as output
+  sm_config_set_set_pins(&config, RCLK, 1);
+  pio_gpio_init(pio, RCLK);
+  pio_sm_set_consecutive_pindirs(pio, sm, RCLK, 1, true);
+
   // Load our configuration, and jump to the start of the program
   pio_sm_init(pio, sm, offset, &config);
   pio_sm_set_enabled(pio, sm, true);

firmware/src/leds.h

@@ -6,13 +6,14 @@
 
 #define COL_SER 20
 #define COL_OE 21
-#define COL_RCLK 22
+// #define COL_RCLK 22
+#define RCLK 22
 #define COL_SRCLK 26
 #define COL_SRCLR 27
 
 #define ROW_SER 14
 #define ROW_OE 13
-#define ROW_RCLK 12
+// #define ROW_RCLK 12
 #define ROW_SRCLK 11
 #define ROW_SRCLR 10
 

firmware/src/leds.pio

@@ -7,15 +7,26 @@ public entry_point:
   ; push 24 bits to the shift registers
   ; also, return latch bit to 0
   set x, 23   side 0
-  ; ignore the 8 least significant bits
-  out null, 8
 loop:
-  ; lower clock edge
-  nop side 0
-  ; set bit
-  out pins, 1
+  ; TODO: check if delays can be lowered with a PCB
+  ; set bit; lower clock edge
+  out pins, 1    side 0 [1]
   ; loop; latch bit (rising edge)
-  ; TODO: check if this delay can be lowered with a PCB
   jmp x-- loop   side 1 [2]
 end:
-.wrap
+  ; ignore unused bits
+  ; load MSBs into x
+  ; lower clock edge
+  out x, 8    side 0
+  ; MSB=1 indicates end of row
+  jmp x-- end_of_row
+  .wrap
+end_of_row:
+  ; indicate to main processor that row was processed
+  ; clock RCLK (latch onto register output stage)
+  set pins, 1 [3]
+  set pins, 0
+  irq wait 0 rel
+
+  ; wait for next row
+  jmp entry_point

firmware/src/leds.pio.h

@@ -20,17 +20,21 @@
 static const uint16_t leds_px_pusher_program_instructions[] = {
             //     .wrap_target
     0xf037, //  0: set    x, 23           side 0     
-    0x6068, //  1: out    null, 8                    
-    0xb042, //  2: nop                    side 0     
-    0x6001, //  3: out    pins, 1                    
-    0x1a42, //  4: jmp    x--, 2          side 1 [2] 
+    0x7101, //  1: out    pins, 1         side 0 [1] 
+    0x1a41, //  2: jmp    x--, 1          side 1 [2] 
+    0x7028, //  3: out    x, 8            side 0     
+    0x0045, //  4: jmp    x--, 5                     
             //     .wrap
+    0xe301, //  5: set    pins, 1                [3] 
+    0xe000, //  6: set    pins, 0                    
+    0xc030, //  7: irq    wait 0 rel                 
+    0x0000, //  8: jmp    0                          
 };
 
 #if !PICO_NO_HARDWARE
 static const struct pio_program leds_px_pusher_program = {
     .instructions = leds_px_pusher_program_instructions,
-    .length = 5,
+    .length = 9,
     .origin = -1,
 };