arm: automatically determine stack sizes

This is a big change that will determine the stack size for many
goroutines automatically. Functions that aren't recursive and don't call
function pointers can in many cases have an automatically determined
worst case stack size. This is useful, as the stack size is usually much
lower than the previous hardcoded default of 1024 bytes: somewhere
around 200-500 bytes is common.

A side effect of this change is that the default stack sizes (including
the stack size for other architectures such as AVR) can now be changed
in the config JSON file, making it tunable per application.

src/device/arm/cortexm.s

@@ -1,9 +1,11 @@
 .syntax unified
+.cfi_sections .debug_frame
 
 .section .text.HardFault_Handler
 .global  HardFault_Handler
 .type    HardFault_Handler, %function
 HardFault_Handler:
+    .cfi_startproc
     // Put the old stack pointer in the first argument, for easy debugging. This
     // is especially useful on Cortex-M0, which supports far fewer debug
     // facilities.
@@ -19,6 +21,7 @@ HardFault_Handler:
 
     // Continue handling this error in Go.
     bl handleHardFault
+    .cfi_endproc
 .size HardFault_Handler, .-HardFault_Handler
 
 // This is a convenience function for semihosting support.
@@ -27,5 +30,8 @@ HardFault_Handler:
 .global  SemihostingCall
 .type    SemihostingCall, %function
 SemihostingCall:
+    .cfi_startproc
     bkpt 0xab
     bx   lr
+    .cfi_endproc
+.size SemihostingCall, .-SemihostingCall

src/internal/task/task.go

@@ -18,3 +18,8 @@ type Task struct {
 	// state is the underlying running state of the task.
 	state state
 }
+
+// getGoroutineStackSize is a compiler intrinsic that returns the stack size for
+// the given function and falls back to the default stack size. It is replaced
+// with a load from a special section just before codegen.
+func getGoroutineStackSize(fn uintptr) uintptr

src/internal/task/task_coroutine.go

@@ -67,7 +67,7 @@ func createTask() *Task {
 // start invokes a function in a new goroutine. Calls to this are inserted by the compiler.
 // The created goroutine starts running immediately.
 // This is implemented inside the compiler.
-func start(fn uintptr, args unsafe.Pointer)
+func start(fn uintptr, args unsafe.Pointer, stackSize uintptr)
 
 // Current returns the current active task.
 // This is implemented inside the compiler.

src/internal/task/task_none.go

@@ -17,7 +17,7 @@ func Current() *Task {
 }
 
 //go:noinline
-func start(fn uintptr, args unsafe.Pointer) {
+func start(fn uintptr, args unsafe.Pointer, stackSize uintptr) {
 	// The compiler will error if this is reachable.
 	runtimePanic("scheduler is disabled")
 }

src/internal/task/task_stack.go

@@ -54,7 +54,7 @@ func (t *Task) Resume() {
 }
 
 // initialize the state and prepare to call the specified function with the specified argument bundle.
-func (s *state) initialize(fn uintptr, args unsafe.Pointer) {
+func (s *state) initialize(fn uintptr, args unsafe.Pointer, stackSize uintptr) {
 	// Create a stack.
 	stack := make([]uintptr, stackSize/unsafe.Sizeof(uintptr(0)))
 
@@ -67,9 +67,9 @@ func runqueuePushBack(*Task)
 
 // start creates and starts a new goroutine with the given function and arguments.
 // The new goroutine is scheduled to run later.
-func start(fn uintptr, args unsafe.Pointer) {
+func start(fn uintptr, args unsafe.Pointer, stackSize uintptr) {
 	t := &Task{}
-	t.state.initialize(fn, args)
+	t.state.initialize(fn, args, stackSize)
 	runqueuePushBack(t)
 }
 

src/internal/task/task_stack_avr.go

@@ -4,8 +4,6 @@ package task
 
 import "unsafe"
 
-const stackSize = 256
-
 // calleeSavedRegs is the list of registers that must be saved and restored when
 // switching between tasks. Also see scheduler_avr.S that relies on the
 // exact layout of this struct.

src/internal/task/task_stack_cortexm.go

@@ -4,8 +4,6 @@ package task
 
 import "unsafe"
 
-const stackSize = 1024
-
 // calleeSavedRegs is the list of registers that must be saved and restored when
 // switching between tasks. Also see scheduler_cortexm.S that relies on the
 // exact layout of this struct.

src/runtime/scheduler_cortexm.S

@@ -41,11 +41,7 @@ tinygo_getSystemStackPointer:
     .cfi_endproc
 .size tinygo_getSystemStackPointer, .-tinygo_getSystemStackPointer
 
-
-// switchToScheduler and switchToTask are also in the same section, to make sure
-// relative branches work.
-.section .text.tinygo_swapTask
-
+.section .text.tinygo_switchToScheduler
 .global  tinygo_switchToScheduler
 .type    tinygo_switchToScheduler, %function
 tinygo_switchToScheduler:
@@ -62,6 +58,7 @@ tinygo_switchToScheduler:
     .cfi_endproc
 .size tinygo_switchToScheduler, .-tinygo_switchToScheduler
 
+.section .text.tinygo_switchToTask
 .global  tinygo_switchToTask
 .type    tinygo_switchToTask, %function
 tinygo_switchToTask:
@@ -72,11 +69,11 @@ tinygo_switchToTask:
     // and then we can invoke swapTask.
     msr PSP, r0
 
-    // Continue executing in the swapTask function, which swaps the stack
-    // pointer.
+    b.n tinygo_swapTask
     .cfi_endproc
 .size tinygo_switchToTask, .-tinygo_switchToTask
 
+.section .text.tinygo_swapTask
 .global  tinygo_swapTask
 .type    tinygo_swapTask, %function
 tinygo_swapTask: