Remove linked list impl - use heapless, linked list init now const fn

This commit is contained in:
Emil Fresk 2021-08-16 15:37:39 +02:00
parent 74b3964978
commit bc3eb5c547
8 changed files with 16 additions and 625 deletions

View file

@ -44,7 +44,7 @@ cortex-m = "0.7.0"
cortex-m-rtic-macros = { path = "macros", version = "0.6.0-alpha.5" } cortex-m-rtic-macros = { path = "macros", version = "0.6.0-alpha.5" }
rtic-monotonic = "0.1.0-alpha.2" rtic-monotonic = "0.1.0-alpha.2"
rtic-core = "0.3.1" rtic-core = "0.3.1"
heapless = "0.7.1" heapless = "0.7.5"
bare-metal = "1.0.0" bare-metal = "1.0.0"
[dependencies.dwt-systick-monotonic] [dependencies.dwt-systick-monotonic]

View file

@ -327,7 +327,7 @@ pub fn codegen(
impl #internal_spawn_handle_ident { impl #internal_spawn_handle_ident {
pub fn cancel(self) -> Result<#ty, ()> { pub fn cancel(self) -> Result<#ty, ()> {
rtic::export::interrupt::free(|_| unsafe { rtic::export::interrupt::free(|_| unsafe {
let tq = &mut *#tq.get_mut_unchecked().as_mut_ptr(); let tq = #tq.get_mut_unchecked();
if let Some((_task, index)) = tq.cancel_marker(self.marker) { if let Some((_task, index)) = tq.cancel_marker(self.marker) {
// Get the message // Get the message
let msg = #inputs let msg = #inputs
@ -359,7 +359,7 @@ pub fn codegen(
let marker = *#tq_marker.get_mut_unchecked(); let marker = *#tq_marker.get_mut_unchecked();
*#tq_marker.get_mut_unchecked() = #tq_marker.get_mut_unchecked().wrapping_add(1); *#tq_marker.get_mut_unchecked() = #tq_marker.get_mut_unchecked().wrapping_add(1);
let tq = &mut *#tq.get_mut_unchecked().as_mut_ptr(); let tq = #tq.get_mut_unchecked();
tq.update_marker(self.marker, marker, instant, || #pend).map(|_| #name::#m::SpawnHandle { marker }) tq.update_marker(self.marker, marker, instant, || #pend).map(|_| #name::#m::SpawnHandle { marker })
}) })
@ -420,7 +420,7 @@ pub fn codegen(
*#tq_marker.get_mut_unchecked() = #tq_marker.get_mut_unchecked().wrapping_add(1); *#tq_marker.get_mut_unchecked() = #tq_marker.get_mut_unchecked().wrapping_add(1);
let tq = &mut *#tq.get_mut_unchecked().as_mut_ptr(); let tq = #tq.get_mut_unchecked();
tq.enqueue_unchecked( tq.enqueue_unchecked(
nr, nr,

View file

@ -77,18 +77,10 @@ pub fn codegen(app: &App, analysis: &Analysis, extra: &Extra) -> Vec<TokenStream
);)); );));
} }
// Initialize monotonic's interrupts and timer queues // Initialize monotonic's interrupts
for (_, monotonic) in &app.monotonics { for (_, monotonic) in &app.monotonics {
let priority = &monotonic.args.priority; let priority = &monotonic.args.priority;
let binds = &monotonic.args.binds; let binds = &monotonic.args.binds;
let monotonic_name = monotonic.ident.to_string();
let tq = util::tq_ident(&monotonic_name);
let tq = util::mark_internal_ident(&tq);
// Initialize timer queues
stmts.push(
quote!(#tq.get_mut_unchecked().as_mut_ptr().write(rtic::export::TimerQueue::new());),
);
// Compile time assert that this priority is supported by the device // Compile time assert that this priority is supported by the device
stmts.push(quote!(let _ = [(); ((1 << #nvic_prio_bits) - #priority as usize)];)); stmts.push(quote!(let _ = [(); ((1 << #nvic_prio_bits) - #priority as usize)];));

View file

@ -62,21 +62,20 @@ pub fn codegen(app: &App, analysis: &Analysis, _extra: &Extra) -> Vec<TokenStrea
{ {
// For future use // For future use
// let doc = &format!("Timer queue for {}", monotonic_name); // let doc = &format!("Timer queue for {}", monotonic_name);
let cap: u8 = app let cap: usize = app
.software_tasks .software_tasks
.iter() .iter()
.map(|(_name, task)| task.args.capacity) .map(|(_name, task)| task.args.capacity as usize)
.sum(); .sum();
let n = util::capacity_literal(cap as usize); let n = util::capacity_literal(cap);
let tq_ty = let tq_ty = quote!(rtic::export::TimerQueue<#mono_type, #t, #n>);
quote!(core::mem::MaybeUninit<rtic::export::TimerQueue<#mono_type, #t, #n>>);
// For future use // For future use
// let doc = format!(" RTIC internal: {}:{}", file!(), line!()); // let doc = format!(" RTIC internal: {}:{}", file!(), line!());
items.push(quote!( items.push(quote!(
#[doc(hidden)] #[doc(hidden)]
static #tq: rtic::RacyCell<#tq_ty> = static #tq: rtic::RacyCell<#tq_ty> =
rtic::RacyCell::new(core::mem::MaybeUninit::uninit()); rtic::RacyCell::new(rtic::export::TimerQueue(rtic::export::SortedLinkedList::new_u16()));
)); ));
let mono = util::monotonic_ident(&monotonic_name); let mono = util::monotonic_ident(&monotonic_name);
@ -138,7 +137,7 @@ pub fn codegen(app: &App, analysis: &Analysis, _extra: &Extra) -> Vec<TokenStrea
unsafe fn #bound_interrupt() { unsafe fn #bound_interrupt() {
while let Some((task, index)) = rtic::export::interrupt::free(|_| while let Some((task, index)) = rtic::export::interrupt::free(|_|
if let Some(mono) = #m_ident.get_mut_unchecked().as_mut() { if let Some(mono) = #m_ident.get_mut_unchecked().as_mut() {
(&mut *#tq.get_mut_unchecked().as_mut_ptr()).dequeue(|| #disable_isr, mono) #tq.get_mut_unchecked().dequeue(|| #disable_isr, mono)
} else { } else {
// We can only use the timer queue if `init` has returned, and it // We can only use the timer queue if `init` has returned, and it
// writes the `Some(monotonic)` we are accessing here. // writes the `Some(monotonic)` we are accessing here.

View file

@ -13,6 +13,7 @@ pub use cortex_m::{
peripheral::{scb::SystemHandler, syst::SystClkSource, DWT, NVIC}, peripheral::{scb::SystemHandler, syst::SystClkSource, DWT, NVIC},
Peripherals, Peripherals,
}; };
pub use heapless::sorted_linked_list::SortedLinkedList;
pub use heapless::spsc::Queue; pub use heapless::spsc::Queue;
pub use heapless::BinaryHeap; pub use heapless::BinaryHeap;
pub use rtic_monotonic as monotonic; pub use rtic_monotonic as monotonic;

View file

@ -43,8 +43,6 @@ pub use rtic_monotonic::{self, embedded_time as time, Monotonic};
#[doc(hidden)] #[doc(hidden)]
pub mod export; pub mod export;
#[doc(hidden)] #[doc(hidden)]
mod linked_list;
#[doc(hidden)]
mod tq; mod tq;
/// Sets the given `interrupt` as pending /// Sets the given `interrupt` as pending

View file

@ -1,597 +0,0 @@
use core::fmt;
use core::marker::PhantomData;
use core::mem::MaybeUninit;
use core::ops::{Deref, DerefMut};
use core::ptr;
#[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord)]
struct LinkedIndex(u16);
impl LinkedIndex {
#[inline]
const unsafe fn new_unchecked(value: u16) -> Self {
LinkedIndex(value)
}
#[inline]
const fn none() -> Self {
LinkedIndex(u16::MAX)
}
#[inline]
const fn option(self) -> Option<u16> {
if self.0 == u16::MAX {
None
} else {
Some(self.0)
}
}
}
/// A node in the linked list.
pub struct Node<T> {
val: MaybeUninit<T>,
next: LinkedIndex,
}
/// Iterator for the linked list.
pub struct Iter<'a, T, Kind, const N: usize>
where
T: PartialEq + PartialOrd,
Kind: kind::Kind,
{
list: &'a LinkedList<T, Kind, N>,
index: LinkedIndex,
}
impl<'a, T, Kind, const N: usize> Iterator for Iter<'a, T, Kind, N>
where
T: PartialEq + PartialOrd,
Kind: kind::Kind,
{
type Item = &'a T;
fn next(&mut self) -> Option<Self::Item> {
let index = self.index.option()?;
let node = self.list.node_at(index as usize);
self.index = node.next;
Some(self.list.read_data_in_node_at(index as usize))
}
}
/// Comes from [`LinkedList::find_mut`].
pub struct FindMut<'a, T, Kind, const N: usize>
where
T: PartialEq + PartialOrd,
Kind: kind::Kind,
{
list: &'a mut LinkedList<T, Kind, N>,
is_head: bool,
prev_index: LinkedIndex,
index: LinkedIndex,
maybe_changed: bool,
}
impl<'a, T, Kind, const N: usize> FindMut<'a, T, Kind, N>
where
T: PartialEq + PartialOrd,
Kind: kind::Kind,
{
fn pop_internal(&mut self) -> T {
if self.is_head {
// If it is the head element, we can do a normal pop
unsafe { self.list.pop_unchecked() }
} else {
// Somewhere in the list
// Re-point the previous index
self.list.node_at_mut(self.prev_index.0 as usize).next =
self.list.node_at_mut(self.index.0 as usize).next;
// Release the index into the free queue
self.list.node_at_mut(self.index.0 as usize).next = self.list.free;
self.list.free = self.index;
self.list.extract_data_in_node_at(self.index.0 as usize)
}
}
/// This will pop the element from the list.
///
/// Complexity is O(1).
#[inline]
pub fn pop(mut self) -> T {
self.pop_internal()
}
/// This will resort the element into the correct position in the list in needed.
/// Same as calling `drop`.
///
/// Complexity is worst-case O(N).
#[inline]
pub fn finish(self) {
drop(self)
}
}
impl<T, Kind, const N: usize> Drop for FindMut<'_, T, Kind, N>
where
T: PartialEq + PartialOrd,
Kind: kind::Kind,
{
fn drop(&mut self) {
// Only resort the list if the element has changed
if self.maybe_changed {
let val = self.pop_internal();
unsafe { self.list.push_unchecked(val) };
}
}
}
impl<T, Kind, const N: usize> Deref for FindMut<'_, T, Kind, N>
where
T: PartialEq + PartialOrd,
Kind: kind::Kind,
{
type Target = T;
fn deref(&self) -> &Self::Target {
self.list.read_data_in_node_at(self.index.0 as usize)
}
}
impl<T, Kind, const N: usize> DerefMut for FindMut<'_, T, Kind, N>
where
T: PartialEq + PartialOrd,
Kind: kind::Kind,
{
fn deref_mut(&mut self) -> &mut Self::Target {
self.maybe_changed = true;
self.list.read_mut_data_in_node_at(self.index.0 as usize)
}
}
impl<T, Kind, const N: usize> fmt::Debug for FindMut<'_, T, Kind, N>
where
T: PartialEq + PartialOrd + core::fmt::Debug,
Kind: kind::Kind,
{
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
f.debug_struct("FindMut")
.field("prev_index", &self.prev_index)
.field("index", &self.index)
.field(
"prev_value",
&self
.list
.read_data_in_node_at(self.prev_index.option().unwrap() as usize),
)
.field(
"value",
&self
.list
.read_data_in_node_at(self.index.option().unwrap() as usize),
)
.finish()
}
}
/// The linked list.
pub struct LinkedList<T, Kind, const N: usize>
where
T: PartialEq + PartialOrd,
Kind: kind::Kind,
{
list: MaybeUninit<[Node<T>; N]>,
head: LinkedIndex,
free: LinkedIndex,
_kind: PhantomData<Kind>,
}
impl<T, Kind, const N: usize> LinkedList<T, Kind, N>
where
T: PartialEq + PartialOrd,
Kind: kind::Kind,
{
/// Internal helper to not do pointer arithmetic all over the place.
#[inline]
fn node_at(&self, index: usize) -> &Node<T> {
// Safety: The entire `self.list` is initialized in `new`, which makes this safe.
unsafe { &*(self.list.as_ptr() as *const Node<T>).add(index) }
}
/// Internal helper to not do pointer arithmetic all over the place.
#[inline]
fn node_at_mut(&mut self, index: usize) -> &mut Node<T> {
// Safety: The entire `self.list` is initialized in `new`, which makes this safe.
unsafe { &mut *(self.list.as_mut_ptr() as *mut Node<T>).add(index) }
}
/// Internal helper to not do pointer arithmetic all over the place.
#[inline]
fn write_data_in_node_at(&mut self, index: usize, data: T) {
unsafe {
self.node_at_mut(index).val.as_mut_ptr().write(data);
}
}
/// Internal helper to not do pointer arithmetic all over the place.
#[inline]
fn read_data_in_node_at(&self, index: usize) -> &T {
unsafe { &*self.node_at(index).val.as_ptr() }
}
/// Internal helper to not do pointer arithmetic all over the place.
#[inline]
fn read_mut_data_in_node_at(&mut self, index: usize) -> &mut T {
unsafe { &mut *self.node_at_mut(index).val.as_mut_ptr() }
}
/// Internal helper to not do pointer arithmetic all over the place.
#[inline]
fn extract_data_in_node_at(&mut self, index: usize) -> T {
unsafe { self.node_at(index).val.as_ptr().read() }
}
/// Internal helper to not do pointer arithmetic all over the place.
/// Safety: This can overwrite existing allocated nodes if used improperly, meaning their
/// `Drop` methods won't run.
#[inline]
unsafe fn write_node_at(&mut self, index: usize, node: Node<T>) {
(self.list.as_mut_ptr() as *mut Node<T>)
.add(index)
.write(node)
}
/// Create a new linked list.
pub fn new() -> Self {
let mut list = LinkedList {
list: MaybeUninit::uninit(),
head: LinkedIndex::none(),
free: unsafe { LinkedIndex::new_unchecked(0) },
_kind: PhantomData,
};
let len = N as u16;
let mut free = 0;
if len == 0 {
list.free = LinkedIndex::none();
return list;
}
// Initialize indexes
while free < len - 1 {
unsafe {
list.write_node_at(
free as usize,
Node {
val: MaybeUninit::uninit(),
next: LinkedIndex::new_unchecked(free + 1),
},
);
}
free += 1;
}
// Initialize final index
unsafe {
list.write_node_at(
free as usize,
Node {
val: MaybeUninit::uninit(),
next: LinkedIndex::none(),
},
);
}
list
}
/// Push unchecked
///
/// Complexity is O(N).
///
/// # Safety
///
/// Assumes that the list is not full.
pub unsafe fn push_unchecked(&mut self, value: T) {
let new = self.free.0;
// Store the data and update the next free spot
self.write_data_in_node_at(new as usize, value);
self.free = self.node_at(new as usize).next;
if let Some(head) = self.head.option() {
// Check if we need to replace head
if self
.read_data_in_node_at(head as usize)
.partial_cmp(self.read_data_in_node_at(new as usize))
!= Kind::ordering()
{
self.node_at_mut(new as usize).next = self.head;
self.head = LinkedIndex::new_unchecked(new);
} else {
// It's not head, search the list for the correct placement
let mut current = head;
while let Some(next) = self.node_at(current as usize).next.option() {
if self
.read_data_in_node_at(next as usize)
.partial_cmp(self.read_data_in_node_at(new as usize))
!= Kind::ordering()
{
break;
}
current = next;
}
self.node_at_mut(new as usize).next = self.node_at(current as usize).next;
self.node_at_mut(current as usize).next = LinkedIndex::new_unchecked(new);
}
} else {
self.node_at_mut(new as usize).next = self.head;
self.head = LinkedIndex::new_unchecked(new);
}
}
/// Pushes an element to the linked list and sorts it into place.
///
/// Complexity is O(N).
pub fn push(&mut self, value: T) -> Result<(), T> {
if !self.is_full() {
Ok(unsafe { self.push_unchecked(value) })
} else {
Err(value)
}
}
/// Get an iterator over the sorted list.
pub fn iter(&self) -> Iter<'_, T, Kind, N> {
Iter {
list: self,
index: self.head,
}
}
/// Find an element in the list.
pub fn find_mut<F>(&mut self, mut f: F) -> Option<FindMut<'_, T, Kind, N>>
where
F: FnMut(&T) -> bool,
{
let head = self.head.option()?;
// Special-case, first element
if f(self.read_data_in_node_at(head as usize)) {
return Some(FindMut {
is_head: true,
prev_index: LinkedIndex::none(),
index: self.head,
list: self,
maybe_changed: false,
});
}
let mut current = head;
while let Some(next) = self.node_at(current as usize).next.option() {
if f(self.read_data_in_node_at(next as usize)) {
return Some(FindMut {
is_head: false,
prev_index: unsafe { LinkedIndex::new_unchecked(current) },
index: unsafe { LinkedIndex::new_unchecked(next) },
list: self,
maybe_changed: false,
});
}
current = next;
}
None
}
/// Peek at the first element.
pub fn peek(&self) -> Option<&T> {
self.head
.option()
.map(|head| self.read_data_in_node_at(head as usize))
}
/// Pop unchecked
///
/// # Safety
///
/// Assumes that the list is not empty.
pub unsafe fn pop_unchecked(&mut self) -> T {
let head = self.head.0;
let current = head;
self.head = self.node_at(head as usize).next;
self.node_at_mut(current as usize).next = self.free;
self.free = LinkedIndex::new_unchecked(current);
self.extract_data_in_node_at(current as usize)
}
/// Pops the first element in the list.
///
/// Complexity is O(1).
pub fn pop(&mut self) -> Result<T, ()> {
if !self.is_empty() {
Ok(unsafe { self.pop_unchecked() })
} else {
Err(())
}
}
/// Checks if the linked list is full.
#[inline]
pub fn is_full(&self) -> bool {
self.free.option().is_none()
}
/// Checks if the linked list is empty.
#[inline]
pub fn is_empty(&self) -> bool {
self.head.option().is_none()
}
}
impl<T, Kind, const N: usize> Drop for LinkedList<T, Kind, N>
where
T: PartialEq + PartialOrd,
Kind: kind::Kind,
{
fn drop(&mut self) {
let mut index = self.head;
while let Some(i) = index.option() {
let node = self.node_at_mut(i as usize);
index = node.next;
unsafe {
ptr::drop_in_place(node.val.as_mut_ptr());
}
}
}
}
impl<T, Kind, const N: usize> fmt::Debug for LinkedList<T, Kind, N>
where
T: PartialEq + PartialOrd + core::fmt::Debug,
Kind: kind::Kind,
{
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
f.debug_list().entries(self.iter()).finish()
}
}
/// Min sorted linked list.
pub struct Min;
/// Max sorted linked list.
pub struct Max;
/// Sealed traits and implementations for `linked_list`
pub mod kind {
use super::{Max, Min};
use core::cmp::Ordering;
/// The linked list kind: min first or max first
pub unsafe trait Kind {
#[doc(hidden)]
fn ordering() -> Option<Ordering>;
}
unsafe impl Kind for Min {
#[inline]
fn ordering() -> Option<Ordering> {
Some(Ordering::Less)
}
}
unsafe impl Kind for Max {
#[inline]
fn ordering() -> Option<Ordering> {
Some(Ordering::Greater)
}
}
}
#[cfg(test)]
mod tests {
// Note this useful idiom: importing names from outer (for mod tests) scope.
use super::*;
#[test]
fn test_peek() {
let mut ll: LinkedList<u32, Max, 3> = LinkedList::new();
ll.push(1).unwrap();
assert_eq!(ll.peek().unwrap(), &1);
ll.push(2).unwrap();
assert_eq!(ll.peek().unwrap(), &2);
ll.push(3).unwrap();
assert_eq!(ll.peek().unwrap(), &3);
let mut ll: LinkedList<u32, Min, 3> = LinkedList::new();
ll.push(2).unwrap();
assert_eq!(ll.peek().unwrap(), &2);
ll.push(1).unwrap();
assert_eq!(ll.peek().unwrap(), &1);
ll.push(3).unwrap();
assert_eq!(ll.peek().unwrap(), &1);
}
#[test]
fn test_full() {
let mut ll: LinkedList<u32, Max, 3> = LinkedList::new();
ll.push(1).unwrap();
ll.push(2).unwrap();
ll.push(3).unwrap();
assert!(ll.is_full())
}
#[test]
fn test_empty() {
let ll: LinkedList<u32, Max, 3> = LinkedList::new();
assert!(ll.is_empty())
}
#[test]
fn test_zero_size() {
let ll: LinkedList<u32, Max, 0> = LinkedList::new();
assert!(ll.is_empty());
assert!(ll.is_full());
}
#[test]
fn test_rejected_push() {
let mut ll: LinkedList<u32, Max, 3> = LinkedList::new();
ll.push(1).unwrap();
ll.push(2).unwrap();
ll.push(3).unwrap();
// This won't fit
let r = ll.push(4);
assert_eq!(r, Err(4));
}
#[test]
fn test_updating() {
let mut ll: LinkedList<u32, Max, 3> = LinkedList::new();
ll.push(1).unwrap();
ll.push(2).unwrap();
ll.push(3).unwrap();
let mut find = ll.find_mut(|v| *v == 2).unwrap();
*find += 1000;
find.finish();
assert_eq!(ll.peek().unwrap(), &1002);
let mut find = ll.find_mut(|v| *v == 3).unwrap();
*find += 1000;
find.finish();
assert_eq!(ll.peek().unwrap(), &1003);
// Remove largest element
ll.find_mut(|v| *v == 1003).unwrap().pop();
assert_eq!(ll.peek().unwrap(), &1002);
}
}

View file

@ -1,9 +1,9 @@
use crate::{ use crate::{
linked_list::{LinkedList, Min},
time::{Clock, Instant}, time::{Clock, Instant},
Monotonic, Monotonic,
}; };
use core::cmp::Ordering; use core::cmp::Ordering;
use heapless::sorted_linked_list::{LinkedIndexU16, Min, SortedLinkedList};
#[inline(always)] #[inline(always)]
fn unwrapper<T, E>(val: Result<T, E>) -> T { fn unwrapper<T, E>(val: Result<T, E>) -> T {
@ -14,7 +14,9 @@ fn unwrapper<T, E>(val: Result<T, E>) -> T {
} }
} }
pub struct TimerQueue<Mono, Task, const N: usize>(pub LinkedList<NotReady<Mono, Task>, Min, N>) pub struct TimerQueue<Mono, Task, const N: usize>(
pub SortedLinkedList<NotReady<Mono, Task>, LinkedIndexU16, Min, N>,
)
where where
Mono: Monotonic, Mono: Monotonic,
Task: Copy; Task: Copy;
@ -24,10 +26,6 @@ where
Mono: Monotonic, Mono: Monotonic,
Task: Copy, Task: Copy,
{ {
pub fn new() -> Self {
TimerQueue(LinkedList::new())
}
/// # Safety /// # Safety
/// ///
/// Writing to memory with a transmute in order to enable /// Writing to memory with a transmute in order to enable