From f0a260be663077f86608162a36e928ddeee651f3 Mon Sep 17 00:00:00 2001
From: Ayke van Laethem <aykevanlaethem@gmail.com>
Date: Thu, 14 Apr 2022 19:34:15 +0200
Subject: [PATCH] ws2812: convert AVR assembly to C inline assembly

See https://github.com/tinygo-org/drivers/pull/401 for details.
I haven't converted it to autogenerated assembly because AVR is
different from many other architectures (8-bit, among others) and it
didn't seem worth the effort as many chips run at 16MHz anyway.

I ran the two AVR smoke tests for the ws2812 driver and the resulting
binary is exactly the same.
---
 ws2812/ws2812_avr.go | 72 ++++++++++++++++++++++++--------------------
 1 file changed, 40 insertions(+), 32 deletions(-)

diff --git a/ws2812/ws2812_avr.go b/ws2812/ws2812_avr.go
index 481c10b..57857bb 100644
--- a/ws2812/ws2812_avr.go
+++ b/ws2812/ws2812_avr.go
@@ -6,11 +6,49 @@ package ws2812
 // This file implements the WS2812 protocol for AVR microcontrollers.
 
 import (
-	"device/avr"
 	"machine"
 	"runtime/interrupt"
+	"unsafe"
 )
 
+/*
+#include <stdint.h>
+
+__attribute__((always_inline))
+void ws2812_writeByte16(char c, uint8_t *port, uint8_t maskSet, uint8_t maskClear) {
+	// See:
+	// https://wp.josh.com/2014/05/13/ws2812-neopixels-are-not-so-finicky-once-you-get-to-know-them/
+	// T0H: 4  cycles or 250ns
+	// T0L: 14 cycles or 875ns -> together 18 cycles or 1125ns
+	// T1H: 9  cycles or 562ns
+	// T1L: 8  cycles or 500ns -> together 17 cycles or 1062ns
+	char i = 8;
+	__asm__ __volatile__(
+		"1:\n"
+		"\t  st    %[port], %[maskSet]   ; [2]   set output high\n"
+		"\t  lsl   %[value]              ; [1]   shift off the next bit, store it in C\n"
+		"\t  brcs  2f                    ; [1/2] branch if this bit is high (long pulse)\n"
+		"\t  st    %[port], %[maskClear] ; [2]   set output low (short pulse)\n"
+		"\t2:\n"
+		"\t  nop                         ; [4]   wait before changing the output again\n"
+		"\t  nop\n"
+		"\t  nop\n"
+		"\t  nop\n"
+		"\t  st    %[port], %[maskClear] ; [2]   set output low (end of pulse)\n"
+		"\t  nop                         ; [3]\n"
+		"\t  nop\n"
+		"\t  nop\n"
+		"\t  subi  %[i], 1               ; [1]   subtract one (for the loop)\n"
+		"\t  brne  1b                    ; [1/2] send the next bit, if not at the end of the loop\n"
+	: [value]"+r"(c),
+	  [i]"+r"(i)
+	: [maskSet]"r"(maskSet),
+	  [maskClear]"r"(maskClear),
+	  [port]"m"(*port));
+}
+*/
+import "C"
+
 // Send a single byte using the WS2812 protocol.
 func (d Device) WriteByte(c byte) error {
 	// On AVR, the port is always the same for setting and clearing a register
@@ -23,37 +61,7 @@ func (d Device) WriteByte(c byte) error {
 
 	switch machine.CPUFrequency() {
 	case 16e6: // 16MHz
-		// See:
-		// https://wp.josh.com/2014/05/13/ws2812-neopixels-are-not-so-finicky-once-you-get-to-know-them/
-		// T0H: 4  cycles or 250ns
-		// T0L: 14 cycles or 875ns -> together 18 cycles or 1125ns
-		// T1H: 9  cycles or 562ns
-		// T1L: 8  cycles or 500ns -> together 17 cycles or 1062ns
-		avr.AsmFull(`
-	send_bit:
-		st    {portSet}, {maskSet}     ; [2]   set output high
-		lsl   {value}                  ; [1]   shift off the next bit, store it in C
-		brcs  skip_store               ; [1/2] branch if this bit is high (long pulse)
-		st    {portClear}, {maskClear} ; [2]   set output low (short pulse)
-	skip_store:
-		nop                            ; [4]   wait before changing the output again
-		nop
-		nop
-		nop
-		st    {portClear}, {maskClear} ; [2]   set output low (end of pulse)
-		nop                            ; [3]
-		nop
-		nop
-		subi  {i}, 1                   ; [1]   subtract one (for the loop)
-		brne  send_bit                 ; [1/2] send the next bit, if not at the end of the loop
-	`, map[string]interface{}{
-			"value":     c,
-			"i":         byte(8),
-			"maskSet":   maskSet,
-			"portSet":   port,
-			"maskClear": maskClear,
-			"portClear": port,
-		})
+		C.ws2812_writeByte16(C.char(c), (*uint8)(unsafe.Pointer(port)), maskSet, maskClear)
 		interrupt.Restore(mask)
 		return nil
 	default: