Optimize node layoutand improve connection branch predition

crates/core/src/redpiler/backend/direct/compile.rs

@@ -1,5 +1,7 @@
+use crate::redpiler::backend::direct::node::NonMaxU8;
 use crate::redpiler::compile_graph::{CompileGraph, LinkType, NodeIdx};
 use crate::redpiler::{CompilerOptions, TaskMonitor};
+use itertools::Itertools;
 use mchprs_blocks::blocks::Block;
 use mchprs_world::TickEntry;
 use petgraph::visit::EdgeRef;
@@ -68,6 +70,10 @@ fn compile_node(
     let updates = if node.ty != CNodeType::Constant {
         graph
             .edges_directed(node_idx, Direction::Outgoing)
+            .sorted_by_key(|edge| nodes_map[&edge.target()])
+            .into_group_map_by(|edge| std::mem::discriminant(&graph[edge.target()].ty))
+            .into_values()
+            .flatten()
             .map(|edge| unsafe {
                 let idx = edge.target();
                 let idx = nodes_map[&idx];
@@ -99,7 +105,7 @@ fn compile_node(
             facing_diode,
         } => NodeType::Comparator {
             mode: *mode,
-            far_input: *far_input,
+            far_input: far_input.map(|value| NonMaxU8::new(value).unwrap()),
             facing_diode: *facing_diode,
         },
         CNodeType::Lamp => NodeType::Lamp,

crates/core/src/redpiler/backend/direct/node.rs

@@ -1,5 +1,6 @@
 use mchprs_blocks::blocks::ComparatorMode;
 use smallvec::SmallVec;
+use std::num::NonZeroU8;
 use std::ops::{Index, IndexMut};
 
 #[derive(Debug, Copy, Clone)]
@@ -114,7 +115,7 @@ pub enum NodeType {
     Torch,
     Comparator {
         mode: ComparatorMode,
-        far_input: Option<u8>,
+        far_input: Option<NonMaxU8>,
         facing_diode: bool,
     },
     Lamp,
@@ -132,15 +133,25 @@ pub struct NodeInput {
     pub ss_counts: [u8; 16],
 }
 
-// struct is 128 bytes to fit nicely into cachelines
-// which are usualy 64 bytes, it can vary but is almost always a power of 2
+#[derive(Debug, Clone, Copy)]
+pub struct NonMaxU8(NonZeroU8);
+
+impl NonMaxU8 {
+    pub fn new(value: u8) -> Option<Self> {
+        NonZeroU8::new(value + 1).map(|x| Self(x))
+    }
+
+    pub fn get(self) -> u8 {
+        self.0.get() - 1
+    }
+}
+
 #[derive(Debug, Clone)]
-#[repr(align(128))]
 pub struct Node {
     pub ty: NodeType,
     pub default_inputs: NodeInput,
     pub side_inputs: NodeInput,
-    pub updates: SmallVec<[ForwardLink; 18]>,
+    pub updates: SmallVec<[ForwardLink; 10]>,
     pub is_io: bool,
 
     /// Powered or lit

crates/core/src/redpiler/backend/direct/tick.rs

@@ -40,7 +40,7 @@ impl DirectBackend {
                 let (mut input_power, side_input_power) = get_all_input(node);
                 if let Some(far_override) = far_input {
                     if input_power < 15 {
-                        input_power = far_override;
+                        input_power = far_override.get();
                     }
                 }
                 let old_strength = node.output_power;

crates/core/src/redpiler/backend/direct/update.rs

@@ -34,9 +34,8 @@ pub(super) fn update_node(scheduler: &mut TickScheduler, node: &mut Node, node_i
             if node.pending_tick {
                 return;
             }
-            let should_be_off = get_bool_input(node);
-            let lit = node.powered;
-            if lit == should_be_off {
+            let should_be_powered = !get_bool_input(node);
+            if node.powered != should_be_powered {
                 schedule_tick(scheduler, node_id, node, 1, TickPriority::Normal);
             }
         }
@@ -51,7 +50,7 @@ pub(super) fn update_node(scheduler: &mut TickScheduler, node: &mut Node, node_i
             let (mut input_power, side_input_power) = get_all_input(node);
             if let Some(far_override) = far_input {
                 if input_power < 15 {
-                    input_power = far_override;
+                    input_power = far_override.get();
                 }
             }
             let old_strength = node.output_power;