windows: add ARM64 support

This was actually surprising once I got TinyGo to build on Windows 11
ARM64. All the changes are exactly what you'd expect for a new
architecture, there was no special weirdness just for arm64.

Actually getting TinyGo to build was kind of involved though. The very
short summary is: install arm64 versions of some pieces of software
(like golang, cmake) instead of installing them though choco. In
particular, use the llvm-mingw[1] toolchain instead of using standard
mingw.

[1]: https://github.com/mstorsjo/llvm-mingw/releases

src/internal/task/task_stack_arm64.go

@@ -1,4 +1,4 @@
-//go:build scheduler.tasks && arm64
+//go:build scheduler.tasks && arm64 && !windows
 
 package task
 

src/internal/task/task_stack_arm64_windows.S

@@ -0,0 +1,65 @@
+.section .text.tinygo_startTask,"ax"
+.global  tinygo_startTask
+tinygo_startTask:
+    .cfi_startproc
+    // Small assembly stub for starting a goroutine. This is already run on the
+    // new stack, with the callee-saved registers already loaded.
+    // Most importantly, x19 contains the pc of the to-be-started function and
+    // x20 contains the only argument it is given. Multiple arguments are packed
+    // into one by storing them in a new allocation.
+
+    // Indicate to the unwinder that there is nothing to unwind, this is the
+    // root frame. It avoids the following (bogus) error message in GDB:
+    //     Backtrace stopped: previous frame identical to this frame (corrupt stack?)
+    .cfi_undefined lr
+
+    // Set the first argument of the goroutine start wrapper, which contains all
+    // the arguments.
+    mov   x0, x20
+
+    // Branch to the "goroutine start" function. By using blx instead of bx,
+    // we'll return here instead of tail calling.
+    blr   x19
+
+    // After return, exit this goroutine. This is a tail call.
+    b     tinygo_pause
+    .cfi_endproc
+
+
+.global tinygo_swapTask
+tinygo_swapTask:
+    // This function gets the following parameters:
+    // x0 = newStack uintptr
+    // x1 = oldStack *uintptr
+
+    // Save all callee-saved registers:
+    stp     x19, x20, [sp, #-160]!
+    stp     x21, x22, [sp, #16]
+    stp     x23, x24, [sp, #32]
+    stp     x25, x26, [sp, #48]
+    stp     x27, x28, [sp, #64]
+    stp     x29, x30, [sp, #80]
+    stp     d8,  d9,  [sp, #96]
+    stp     d10, d11, [sp, #112]
+    stp     d12, d13, [sp, #128]
+    stp     d14, d15, [sp, #144]
+
+    // Save the current stack pointer in oldStack.
+    mov x8, sp
+    str x8, [x1]
+
+    // Switch to the new stack pointer.
+    mov sp, x0
+
+    // Restore stack state and return.
+    ldp     d14, d15, [sp, #144]
+    ldp     d12, d13, [sp, #128]
+    ldp     d10, d11, [sp, #112]
+    ldp     d8, d9, [sp, #96]
+    ldp     x29, x30, [sp, #80]
+    ldp     x27, x28, [sp, #64]
+    ldp     x25, x26, [sp, #48]
+    ldp     x23, x24, [sp, #32]
+    ldp     x21, x22, [sp, #16]
+    ldp     x19, x20, [sp], #160
+    ret

src/internal/task/task_stack_arm64_windows.go

@@ -0,0 +1,72 @@
+//go:build scheduler.tasks && arm64 && windows
+
+package task
+
+import "unsafe"
+
+var systemStack uintptr
+
+// calleeSavedRegs is the list of registers that must be saved and restored
+// when switching between tasks. Also see task_stack_arm64_windows.S that
+// relies on the exact layout of this struct.
+type calleeSavedRegs struct {
+	x19 uintptr
+	x20 uintptr
+	x21 uintptr
+	x22 uintptr
+	x23 uintptr
+	x24 uintptr
+	x25 uintptr
+	x26 uintptr
+	x27 uintptr
+	x28 uintptr
+	x29 uintptr
+	pc  uintptr // aka x30 aka LR
+
+	d8  uintptr
+	d9  uintptr
+	d10 uintptr
+	d11 uintptr
+	d12 uintptr
+	d13 uintptr
+	d14 uintptr
+	d15 uintptr
+}
+
+// archInit runs architecture-specific setup for the goroutine startup.
+func (s *state) archInit(r *calleeSavedRegs, fn uintptr, args unsafe.Pointer) {
+	// Store the initial sp for the startTask function (implemented in assembly).
+	s.sp = uintptr(unsafe.Pointer(r))
+
+	// Initialize the registers.
+	// These will be popped off of the stack on the first resume of the goroutine.
+
+	// Start the function at tinygo_startTask (defined in src/internal/task/task_stack_arm64_windows.S).
+	// This assembly code calls a function (passed in x19) with a single argument
+	// (passed in x20). After the function returns, it calls Pause().
+	r.pc = uintptr(unsafe.Pointer(&startTask))
+
+	// Pass the function to call in x19.
+	// This function is a compiler-generated wrapper which loads arguments out of a struct pointer.
+	// See createGoroutineStartWrapper (defined in compiler/goroutine.go) for more information.
+	r.x19 = fn
+
+	// Pass the pointer to the arguments struct in x20.
+	r.x20 = uintptr(args)
+}
+
+func (s *state) resume() {
+	swapTask(s.sp, &systemStack)
+}
+
+func (s *state) pause() {
+	newStack := systemStack
+	systemStack = 0
+	swapTask(newStack, &s.sp)
+}
+
+// SystemStack returns the system stack pointer when called from a task stack.
+// When called from the system stack, it returns 0.
+func SystemStack() uintptr {
+	return systemStack
+}

src/runtime/asm_arm64_windows.S

@@ -0,0 +1,36 @@
+.section .text.tinygo_scanCurrentStack,"ax"
+.global tinygo_scanCurrentStack
+tinygo_scanCurrentStack:
+    // Sources:
+    //   * https://learn.microsoft.com/en-us/cpp/build/arm64-windows-abi-conventions?view=msvc-170
+    //   * https://godbolt.org/z/foc1xncvb
+
+    // Save callee-saved registers.
+    stp     x29, x30, [sp, #-160]!
+    stp     x28, x27, [sp, #16]
+    stp     x26, x25, [sp, #32]
+    stp     x24, x23, [sp, #48]
+    stp     x22, x21, [sp, #64]
+    stp     x20, x19, [sp, #80]
+    stp     d8,  d9,  [sp, #96]
+    stp     d10, d11, [sp, #112]
+    stp     d12, d13, [sp, #128]
+    stp     d14, d15, [sp, #144]
+
+    // Scan the stack.
+    mov     x0, sp
+    bl      tinygo_scanstack
+
+    // Restore stack state and return.
+    ldp     x29, x30, [sp], #160
+    ret
+
+
+.section .text.tinygo_longjmp,"ax"
+.global tinygo_longjmp
+tinygo_longjmp:
+    // Note: the code we jump to assumes x0 is set to a non-zero value if we
+    // jump from here (which is conveniently already the case).
+    ldp x1, x2, [x0] // jumpSP, jumpPC
+    mov sp, x1
+    br  x2