avr: convert initialization from asm to Go

This increases code size by 1 instruction (2 bytes) because LLVM isn't
yet smart enough to recognize that it doesn't need to clear a register
to use 0: it can just use r1 which is always 0 according to the
convention. It makes initialization a lot easier to read, however.

src/runtime/runtime.go

@@ -22,9 +22,15 @@ var hasScheduler bool
 // Entry point for Go. Initialize all packages and call main.main().
 //go:export main
 func main() int {
+	// Initialize memory etc.
+	preinit()
+
 	// Run initializers of all packages.
 	initAll()
 
+	// Enable interrupts etc.
+	postinit()
+
 	// This branch must be optimized away. Only one of the targets must remain,
 	// or there will be link errors.
 	if hasScheduler {

src/runtime/runtime_avr.go

@@ -4,6 +4,7 @@ package runtime
 
 import (
 	"device/avr"
+	"unsafe"
 )
 
 const BOARD = "arduino"
@@ -28,6 +29,25 @@ const (
 	WDT_PERIOD_2S
 )
 
+var (
+	_extern__sbss unsafe.Pointer // defined by the linker
+	_extern__ebss unsafe.Pointer // defined by the linker
+)
+
+func preinit() {
+	// Initialize .bss: zero-initialized global variables.
+	ptr := uintptr(unsafe.Pointer(&_extern__sbss))
+	for ptr != uintptr(unsafe.Pointer(&_extern__ebss)) {
+		*(*uint8)(unsafe.Pointer(ptr)) = 0
+		ptr += 1
+	}
+}
+
+func postinit() {
+	// Enable interrupts after initialization.
+	avr.Asm("sei")
+}
+
 func init() {
 	initUART()
 }

src/runtime/runtime_nrf.go

@@ -16,6 +16,12 @@ func _start() {
 	main()
 }
 
+func preinit() {
+}
+
+func postinit() {
+}
+
 func init() {
 	initUART()
 	initLFCLK()

src/runtime/runtime_unix.go

@@ -24,6 +24,12 @@ type timespec struct {
 
 const CLOCK_MONOTONIC_RAW = 4
 
+func preinit() {
+}
+
+func postinit() {
+}
+
 func putchar(c byte) {
 	_Cfunc_putchar(int(c))
 }

targets/avr.S

@@ -11,30 +11,12 @@ isr:
 reset:
     clr  r1          ; r1 is expected to be 0 by the C calling convention
 
-    ; Zero .bss
-clear_bss:
-    ldi  xl, lo8(_sbss)
-    ldi  xh, hi8(_sbss)
-clear_bss_loop:
-    ldi  yl, lo8(_ebss)
-    ldi  yh, hi8(_ebss)
-    cp   xl, yl         ; if x == y
-    cpc  xh, yh
-    breq clear_bss_end
-    st   x+, r1         ; zero byte in *x
-    rjmp clear_bss_loop
-clear_bss_end:
-
     ; Set up the stack pointer.
     ldi  xl, lo8(_stack_top)
     ldi  xh, hi8(_stack_top)
     out  0x3d, xl; SPL
     out  0x3e, xh; SPH
 
-    ; Enable interrupts.
-    ; TODO: make sure interrupts are started after all initializers have run.
-    sei
-
     ; main will be placed right after here by the linker script so there's no
     ; need to jump.