diff --git a/.github/workflows/rust.yml b/.github/workflows/rust.yml
index ced6eb9..1393b11 100644
--- a/.github/workflows/rust.yml
+++ b/.github/workflows/rust.yml
@@ -37,6 +37,16 @@ jobs:
     - name: Check with all features
       run: cargo check --all-features
 
+  # Make sure the code and its dependencies compile without std.
+  no_std:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v3
+    - name: Install thumbv6m-none-eabi target
+      run: rustup target add thumbv6m-none-eabi
+    - name: Check no_std build
+      run: cargo check --target thumbv6m-none-eabi --no-default-features --features trace-log,checker,enable-serde
+
   # Lint dependency graph for security advisories, duplicate versions, and
   # incompatible licences.
   cargo_deny:
diff --git a/Cargo.toml b/Cargo.toml
index 60f35b9..dfff957 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -13,11 +13,15 @@ repository = "https://github.com/bytecodealliance/regalloc2"
 [dependencies]
 log = { version = "0.4.8", default-features = false }
 smallvec = { version = "1.6.1", features = ["union"] }
-fxhash = "0.2.1"
-slice-group-by = "0.3.0"
+rustc-hash = { version = "1.1.0", default-features = false }
+slice-group-by = { version = "0.3.0", default-features = false }
+hashbrown = "0.13.2"
 
 # Optional serde support, enabled by feature below.
-serde = { version = "1.0.136", features = ["derive"], optional = true }
+serde = { version = "1.0.136", features = [
+    "derive",
+    "alloc",
+], default-features = false, optional = true }
 
 # The below are only needed for fuzzing.
 libfuzzer-sys = { version = "0.4.2", optional = true }
@@ -29,7 +33,10 @@ debug-assertions = true
 overflow-checks = true
 
 [features]
-default = []
+default = ["std"]
+
+# Enables std-specific features such as the Error trait for RegAllocError.
+std = []
 
 # Enables generation of DefAlloc edits for the checker.
 checker = []
diff --git a/src/cfg.rs b/src/cfg.rs
index a882f97..cbb6cea 100644
--- a/src/cfg.rs
+++ b/src/cfg.rs
@@ -6,6 +6,8 @@
 //! Lightweight CFG analyses.
 
 use crate::{domtree, postorder, Block, Function, Inst, ProgPoint, RegAllocError};
+use alloc::vec;
+use alloc::vec::Vec;
 use smallvec::{smallvec, SmallVec};
 
 #[derive(Clone, Debug)]
diff --git a/src/checker.rs b/src/checker.rs
index 10fd458..6df858b 100644
--- a/src/checker.rs
+++ b/src/checker.rs
@@ -96,14 +96,16 @@
 #![allow(dead_code)]
 
 use crate::{
-    Allocation, AllocationKind, Block, Edit, Function, Inst, InstOrEdit, InstPosition, MachineEnv,
-    Operand, OperandConstraint, OperandKind, OperandPos, Output, PReg, PRegSet, VReg,
+    Allocation, AllocationKind, Block, Edit, Function, FxHashMap, FxHashSet, Inst, InstOrEdit,
+    InstPosition, MachineEnv, Operand, OperandConstraint, OperandKind, OperandPos, Output, PReg,
+    PRegSet, VReg,
 };
-use fxhash::{FxHashMap, FxHashSet};
+use alloc::vec::Vec;
+use alloc::{format, vec};
+use core::default::Default;
+use core::hash::Hash;
+use core::result::Result;
 use smallvec::{smallvec, SmallVec};
-use std::default::Default;
-use std::hash::Hash;
-use std::result::Result;
 
 /// A set of errors detected by the regalloc checker.
 #[derive(Clone, Debug)]
@@ -230,7 +232,7 @@ impl CheckerValue {
     }
 
     fn from_reg(reg: VReg) -> CheckerValue {
-        CheckerValue::VRegs(std::iter::once(reg).collect())
+        CheckerValue::VRegs(core::iter::once(reg).collect())
     }
 
     fn remove_vreg(&mut self, reg: VReg) {
@@ -373,8 +375,8 @@ impl Default for CheckerState {
     }
 }
 
-impl std::fmt::Display for CheckerValue {
-    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
+impl core::fmt::Display for CheckerValue {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
         match self {
             CheckerValue::Universe => {
                 write!(f, "top")
diff --git a/src/domtree.rs b/src/domtree.rs
index 1bf4139..33c7108 100644
--- a/src/domtree.rs
+++ b/src/domtree.rs
@@ -12,6 +12,9 @@
 //   TR-06-33870
 //   https://www.cs.rice.edu/~keith/EMBED/dom.pdf
 
+use alloc::vec;
+use alloc::vec::Vec;
+
 use crate::Block;
 
 // Helper
diff --git a/src/fuzzing/func.rs b/src/fuzzing/func.rs
index 81057ce..6996a4c 100644
--- a/src/fuzzing/func.rs
+++ b/src/fuzzing/func.rs
@@ -8,6 +8,9 @@ use crate::{
     OperandConstraint, OperandKind, OperandPos, PReg, PRegSet, RegClass, VReg,
 };
 
+use alloc::vec::Vec;
+use alloc::{format, vec};
+
 use super::arbitrary::Result as ArbitraryResult;
 use super::arbitrary::{Arbitrary, Unstructured};
 
@@ -275,7 +278,7 @@ pub struct Options {
     pub reftypes: bool,
 }
 
-impl std::default::Default for Options {
+impl core::default::Default for Options {
     fn default() -> Self {
         Options {
             reused_inputs: false,
@@ -404,7 +407,7 @@ impl Func {
             }
             vregs_by_block.push(vregs.clone());
             vregs_by_block_to_be_defined.push(vec![]);
-            let mut max_block_params = u.int_in_range(0..=std::cmp::min(3, vregs.len() / 3))?;
+            let mut max_block_params = u.int_in_range(0..=core::cmp::min(3, vregs.len() / 3))?;
             for &vreg in &vregs {
                 if block > 0 && opts.block_params && bool::arbitrary(u)? && max_block_params > 0 {
                     block_params[block].push(vreg);
@@ -591,8 +594,8 @@ impl Func {
     }
 }
 
-impl std::fmt::Debug for Func {
-    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
+impl core::fmt::Debug for Func {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
         write!(f, "{{\n")?;
         for vreg in self.reftype_vregs() {
             write!(f, "  REF: {}\n", vreg)?;
@@ -653,7 +656,7 @@ impl std::fmt::Debug for Func {
 }
 
 pub fn machine_env() -> MachineEnv {
-    fn regs(r: std::ops::Range<usize>) -> Vec<PReg> {
+    fn regs(r: core::ops::Range<usize>) -> Vec<PReg> {
         r.map(|i| PReg::new(i, RegClass::Int)).collect()
     }
     let preferred_regs_by_class: [Vec<PReg>; 2] = [regs(0..24), vec![]];
diff --git a/src/index.rs b/src/index.rs
index d847484..a624c3e 100644
--- a/src/index.rs
+++ b/src/index.rs
@@ -50,11 +50,11 @@ macro_rules! define_index {
     };
 }
 
-pub trait ContainerIndex: Clone + Copy + std::fmt::Debug + PartialEq + Eq {}
+pub trait ContainerIndex: Clone + Copy + core::fmt::Debug + PartialEq + Eq {}
 
 pub trait ContainerComparator {
     type Ix: ContainerIndex;
-    fn compare(&self, a: Self::Ix, b: Self::Ix) -> std::cmp::Ordering;
+    fn compare(&self, a: Self::Ix, b: Self::Ix) -> core::cmp::Ordering;
 }
 
 define_index!(Inst);
@@ -146,6 +146,9 @@ impl Iterator for InstRangeIter {
 
 #[cfg(test)]
 mod test {
+    use alloc::vec;
+    use alloc::vec::Vec;
+
     use super::*;
 
     #[test]
diff --git a/src/indexset.rs b/src/indexset.rs
index 20e70eb..1fd58a2 100644
--- a/src/indexset.rs
+++ b/src/indexset.rs
@@ -5,8 +5,10 @@
 
 //! Index sets: sets of integers that represent indices into a space.
 
-use fxhash::FxHashMap;
-use std::cell::Cell;
+use alloc::vec::Vec;
+use core::cell::Cell;
+
+use crate::FxHashMap;
 
 const SMALL_ELEMS: usize = 12;
 
@@ -151,10 +153,10 @@ impl AdaptiveMap {
 
 enum AdaptiveMapIter<'a> {
     Small(&'a [u32], &'a [u64]),
-    Large(std::collections::hash_map::Iter<'a, u32, u64>),
+    Large(hashbrown::hash_map::Iter<'a, u32, u64>),
 }
 
-impl<'a> std::iter::Iterator for AdaptiveMapIter<'a> {
+impl<'a> core::iter::Iterator for AdaptiveMapIter<'a> {
     type Item = (u32, u64);
 
     #[inline]
@@ -292,7 +294,7 @@ impl Iterator for SetBitsIter {
         // Build an `Option<NonZeroU64>` so that on the nonzero path,
         // the compiler can optimize the trailing-zeroes operator
         // using that knowledge.
-        std::num::NonZeroU64::new(self.0).map(|nz| {
+        core::num::NonZeroU64::new(self.0).map(|nz| {
             let bitidx = nz.trailing_zeros();
             self.0 &= self.0 - 1; // clear highest set bit
             bitidx as usize
@@ -300,8 +302,8 @@ impl Iterator for SetBitsIter {
     }
 }
 
-impl std::fmt::Debug for IndexSet {
-    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
+impl core::fmt::Debug for IndexSet {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
         let vals = self.iter().collect::<Vec<_>>();
         write!(f, "{:?}", vals)
     }
diff --git a/src/ion/data_structures.rs b/src/ion/data_structures.rs
index f8bafc7..ecc903e 100644
--- a/src/ion/data_structures.rs
+++ b/src/ion/data_structures.rs
@@ -17,14 +17,16 @@ use crate::cfg::CFGInfo;
 use crate::index::ContainerComparator;
 use crate::indexset::IndexSet;
 use crate::{
-    define_index, Allocation, Block, Edit, Function, Inst, MachineEnv, Operand, PReg, ProgPoint,
-    RegClass, VReg,
+    define_index, Allocation, Block, Edit, Function, FxHashSet, Inst, MachineEnv, Operand, PReg,
+    ProgPoint, RegClass, VReg,
 };
-use fxhash::FxHashSet;
+use alloc::collections::BTreeMap;
+use alloc::string::String;
+use alloc::vec::Vec;
+use core::cmp::Ordering;
+use core::fmt::Debug;
+use hashbrown::{HashMap, HashSet};
 use smallvec::SmallVec;
-use std::cmp::Ordering;
-use std::collections::{BTreeMap, HashMap, HashSet};
-use std::fmt::Debug;
 
 /// A range from `from` (inclusive) to `to` (exclusive).
 #[derive(Clone, Copy, Debug, PartialEq, Eq)]
@@ -64,13 +66,13 @@ impl CodeRange {
     }
 }
 
-impl std::cmp::PartialOrd for CodeRange {
+impl core::cmp::PartialOrd for CodeRange {
     #[inline(always)]
     fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
         Some(self.cmp(other))
     }
 }
-impl std::cmp::Ord for CodeRange {
+impl core::cmp::Ord for CodeRange {
     #[inline(always)]
     fn cmp(&self, other: &Self) -> Ordering {
         if self.to <= other.from {
@@ -278,7 +280,7 @@ const fn no_bloat_capacity<T>() -> usize {
     //
     // So if `size_of([T; N]) == size_of(pointer) + size_of(capacity)` then we
     // get the maximum inline capacity without bloat.
-    std::mem::size_of::<usize>() * 2 / std::mem::size_of::<T>()
+    core::mem::size_of::<usize>() * 2 / core::mem::size_of::<T>()
 }
 
 #[derive(Clone, Debug)]
@@ -431,7 +433,7 @@ pub struct Env<'a, F: Function> {
 
     // For debug output only: a list of textual annotations at every
     // ProgPoint to insert into the final allocated program listing.
-    pub debug_annotations: std::collections::HashMap<ProgPoint, Vec<String>>,
+    pub debug_annotations: hashbrown::HashMap<ProgPoint, Vec<String>>,
     pub annotations_enabled: bool,
 
     // Cached allocation for `try_to_allocate_bundle_to_reg` to avoid allocating
@@ -492,7 +494,7 @@ impl SpillSlotList {
 
 #[derive(Clone, Debug)]
 pub struct PrioQueue {
-    pub heap: std::collections::BinaryHeap<PrioQueueEntry>,
+    pub heap: alloc::collections::BinaryHeap<PrioQueueEntry>,
 }
 
 #[derive(Clone, Copy, Debug, PartialEq, Eq, PartialOrd, Ord)]
@@ -531,28 +533,28 @@ impl LiveRangeKey {
     }
 }
 
-impl std::cmp::PartialEq for LiveRangeKey {
+impl core::cmp::PartialEq for LiveRangeKey {
     #[inline(always)]
     fn eq(&self, other: &Self) -> bool {
         self.to > other.from && self.from < other.to
     }
 }
-impl std::cmp::Eq for LiveRangeKey {}
-impl std::cmp::PartialOrd for LiveRangeKey {
+impl core::cmp::Eq for LiveRangeKey {}
+impl core::cmp::PartialOrd for LiveRangeKey {
     #[inline(always)]
-    fn partial_cmp(&self, other: &Self) -> Option<std::cmp::Ordering> {
+    fn partial_cmp(&self, other: &Self) -> Option<core::cmp::Ordering> {
         Some(self.cmp(other))
     }
 }
-impl std::cmp::Ord for LiveRangeKey {
+impl core::cmp::Ord for LiveRangeKey {
     #[inline(always)]
-    fn cmp(&self, other: &Self) -> std::cmp::Ordering {
+    fn cmp(&self, other: &Self) -> core::cmp::Ordering {
         if self.to <= other.from {
-            std::cmp::Ordering::Less
+            core::cmp::Ordering::Less
         } else if self.from >= other.to {
-            std::cmp::Ordering::Greater
+            core::cmp::Ordering::Greater
         } else {
-            std::cmp::Ordering::Equal
+            core::cmp::Ordering::Equal
         }
     }
 }
@@ -562,7 +564,7 @@ pub struct PrioQueueComparator<'a> {
 }
 impl<'a> ContainerComparator for PrioQueueComparator<'a> {
     type Ix = LiveBundleIndex;
-    fn compare(&self, a: Self::Ix, b: Self::Ix) -> std::cmp::Ordering {
+    fn compare(&self, a: Self::Ix, b: Self::Ix) -> core::cmp::Ordering {
         self.prios[a.index()].cmp(&self.prios[b.index()])
     }
 }
@@ -570,7 +572,7 @@ impl<'a> ContainerComparator for PrioQueueComparator<'a> {
 impl PrioQueue {
     pub fn new() -> Self {
         PrioQueue {
-            heap: std::collections::BinaryHeap::new(),
+            heap: alloc::collections::BinaryHeap::new(),
         }
     }
 
diff --git a/src/ion/dump.rs b/src/ion/dump.rs
index ba4f74f..abcaaf3 100644
--- a/src/ion/dump.rs
+++ b/src/ion/dump.rs
@@ -1,5 +1,9 @@
 //! Debugging output.
 
+use alloc::string::ToString;
+use alloc::{format, vec};
+use alloc::{string::String, vec::Vec};
+
 use super::Env;
 use crate::{Block, Function, ProgPoint};
 
diff --git a/src/ion/liveranges.rs b/src/ion/liveranges.rs
index da4668e..5a9b00e 100644
--- a/src/ion/liveranges.rs
+++ b/src/ion/liveranges.rs
@@ -22,13 +22,15 @@ use crate::ion::data_structures::{
     BlockparamIn, BlockparamOut, FixedRegFixupLevel, MultiFixedRegFixup,
 };
 use crate::{
-    Allocation, Block, Function, Inst, InstPosition, Operand, OperandConstraint, OperandKind,
-    OperandPos, PReg, ProgPoint, RegAllocError, VReg,
+    Allocation, Block, Function, FxHashMap, FxHashSet, Inst, InstPosition, Operand,
+    OperandConstraint, OperandKind, OperandPos, PReg, ProgPoint, RegAllocError, VReg,
 };
-use fxhash::{FxHashMap, FxHashSet};
+use alloc::collections::VecDeque;
+use alloc::vec;
+use alloc::vec::Vec;
+use hashbrown::HashSet;
 use slice_group_by::GroupByMut;
 use smallvec::{smallvec, SmallVec};
-use std::collections::{HashSet, VecDeque};
 
 /// A spill weight computed for a certain Use.
 #[derive(Clone, Copy, Debug)]
@@ -42,7 +44,7 @@ pub fn spill_weight_from_constraint(
 ) -> SpillWeight {
     // A bonus of 1000 for one loop level, 4000 for two loop levels,
     // 16000 for three loop levels, etc. Avoids exponentiation.
-    let loop_depth = std::cmp::min(10, loop_depth);
+    let loop_depth = core::cmp::min(10, loop_depth);
     let hot_bonus: f32 = (0..loop_depth).fold(1000.0, |a, _| a * 4.0);
     let def_bonus: f32 = if is_def { 2000.0 } else { 0.0 };
     let constraint_bonus: f32 = match constraint {
@@ -91,7 +93,7 @@ impl SpillWeight {
     }
 }
 
-impl std::ops::Add<SpillWeight> for SpillWeight {
+impl core::ops::Add<SpillWeight> for SpillWeight {
     type Output = SpillWeight;
     fn add(self, other: SpillWeight) -> Self {
         SpillWeight(self.0 + other.0)
diff --git a/src/ion/merge.rs b/src/ion/merge.rs
index 3d57d55..54aeaa5 100644
--- a/src/ion/merge.rs
+++ b/src/ion/merge.rs
@@ -18,6 +18,7 @@ use super::{
 use crate::{
     ion::data_structures::BlockparamOut, Function, Inst, OperandConstraint, OperandKind, PReg,
 };
+use alloc::format;
 use smallvec::smallvec;
 
 impl<'a, F: Function> Env<'a, F> {
@@ -132,7 +133,7 @@ impl<'a, F: Function> Env<'a, F> {
             // `to` bundle is empty -- just move the list over from
             // `from` and set `bundle` up-link on all ranges.
             trace!(" -> to bundle{} is empty; trivial merge", to.index());
-            let list = std::mem::replace(&mut self.bundles[from.index()].ranges, smallvec![]);
+            let list = core::mem::replace(&mut self.bundles[from.index()].ranges, smallvec![]);
             for entry in &list {
                 self.ranges[entry.index.index()].bundle = to;
 
@@ -170,7 +171,7 @@ impl<'a, F: Function> Env<'a, F> {
         // Two non-empty lists of LiveRanges: concatenate and
         // sort. This is faster than a mergesort-like merge into a new
         // list, empirically.
-        let from_list = std::mem::replace(&mut self.bundles[from.index()].ranges, smallvec![]);
+        let from_list = core::mem::replace(&mut self.bundles[from.index()].ranges, smallvec![]);
         for entry in &from_list {
             self.ranges[entry.index.index()].bundle = to;
         }
@@ -213,7 +214,7 @@ impl<'a, F: Function> Env<'a, F> {
         }
 
         if self.bundles[from.index()].spillset != self.bundles[to.index()].spillset {
-            let from_vregs = std::mem::replace(
+            let from_vregs = core::mem::replace(
                 &mut self.spillsets[self.bundles[from.index()].spillset.index()].vregs,
                 smallvec![],
             );
diff --git a/src/ion/mod.rs b/src/ion/mod.rs
index 8edbfe0..550af90 100644
--- a/src/ion/mod.rs
+++ b/src/ion/mod.rs
@@ -16,7 +16,9 @@
 use crate::cfg::CFGInfo;
 use crate::ssa::validate_ssa;
 use crate::{Function, MachineEnv, Output, PReg, ProgPoint, RegAllocError, RegClass};
-use std::collections::HashMap;
+use alloc::vec;
+use alloc::vec::Vec;
+use hashbrown::HashMap;
 
 pub(crate) mod data_structures;
 pub use data_structures::Stats;
@@ -82,7 +84,7 @@ impl<'a, F: Function> Env<'a, F> {
 
             stats: Stats::default(),
 
-            debug_annotations: std::collections::HashMap::new(),
+            debug_annotations: hashbrown::HashMap::new(),
             annotations_enabled,
 
             conflict_set: Default::default(),
diff --git a/src/ion/moves.rs b/src/ion/moves.rs
index a196e5d..32ae768 100644
--- a/src/ion/moves.rs
+++ b/src/ion/moves.rs
@@ -22,12 +22,13 @@ use crate::ion::data_structures::{
 use crate::ion::reg_traversal::RegTraversalIter;
 use crate::moves::{MoveAndScratchResolver, ParallelMoves};
 use crate::{
-    Allocation, Block, Edit, Function, Inst, InstPosition, OperandConstraint, OperandKind,
-    OperandPos, PReg, ProgPoint, RegClass, SpillSlot, VReg,
+    Allocation, Block, Edit, Function, FxHashMap, Inst, InstPosition, OperandConstraint,
+    OperandKind, OperandPos, PReg, ProgPoint, RegClass, SpillSlot, VReg,
 };
-use fxhash::FxHashMap;
+use alloc::vec::Vec;
+use alloc::{format, vec};
+use core::fmt::Debug;
 use smallvec::{smallvec, SmallVec};
-use std::fmt::Debug;
 
 impl<'a, F: Function> Env<'a, F> {
     pub fn is_start_of_block(&self, pos: ProgPoint) -> bool {
@@ -494,9 +495,9 @@ impl<'a, F: Function> Env<'a, F> {
                             // this case returns the index of the first
                             // entry that is greater as an `Err`.
                             if label_vreg.vreg() < vreg.index() {
-                                std::cmp::Ordering::Less
+                                core::cmp::Ordering::Less
                             } else {
-                                std::cmp::Ordering::Greater
+                                core::cmp::Ordering::Greater
                             }
                         })
                         .unwrap_err();
@@ -515,8 +516,8 @@ impl<'a, F: Function> Env<'a, F> {
                             continue;
                         }
 
-                        let from = std::cmp::max(label_from, range.from);
-                        let to = std::cmp::min(label_to, range.to);
+                        let from = core::cmp::max(label_from, range.from);
+                        let to = core::cmp::min(label_to, range.to);
 
                         self.debug_locations.push((label, from, to, alloc));
                     }
@@ -629,7 +630,7 @@ impl<'a, F: Function> Env<'a, F> {
         }
 
         // Handle multi-fixed-reg constraints by copying.
-        for fixup in std::mem::replace(&mut self.multi_fixed_reg_fixups, vec![]) {
+        for fixup in core::mem::replace(&mut self.multi_fixed_reg_fixups, vec![]) {
             let from_alloc = self.get_alloc(fixup.pos.inst(), fixup.from_slot as usize);
             let to_alloc = Allocation::reg(PReg::from_index(fixup.to_preg.index()));
             trace!(
diff --git a/src/ion/process.rs b/src/ion/process.rs
index 0c2161b..e33a4c6 100644
--- a/src/ion/process.rs
+++ b/src/ion/process.rs
@@ -22,12 +22,11 @@ use crate::{
         CodeRange, BUNDLE_MAX_NORMAL_SPILL_WEIGHT, MAX_SPLITS_PER_SPILLSET,
         MINIMAL_BUNDLE_SPILL_WEIGHT, MINIMAL_FIXED_BUNDLE_SPILL_WEIGHT,
     },
-    Allocation, Function, Inst, InstPosition, OperandConstraint, OperandKind, PReg, ProgPoint,
-    RegAllocError,
+    Allocation, Function, FxHashSet, Inst, InstPosition, OperandConstraint, OperandKind, PReg,
+    ProgPoint, RegAllocError,
 };
-use fxhash::FxHashSet;
+use core::fmt::Debug;
 use smallvec::{smallvec, SmallVec};
-use std::fmt::Debug;
 
 #[derive(Clone, Debug, PartialEq, Eq)]
 pub enum AllocRegResult {
@@ -159,7 +158,7 @@ impl<'a, F: Function> Env<'a, F> {
                     trace!("   -> conflict bundle {:?}", conflict_bundle);
                     if self.conflict_set.insert(conflict_bundle) {
                         conflicts.push(conflict_bundle);
-                        max_conflict_weight = std::cmp::max(
+                        max_conflict_weight = core::cmp::max(
                             max_conflict_weight,
                             self.bundles[conflict_bundle.index()].cached_spill_weight(),
                         );
@@ -172,7 +171,7 @@ impl<'a, F: Function> Env<'a, F> {
                     }
 
                     if first_conflict.is_none() {
-                        first_conflict = Some(ProgPoint::from_index(std::cmp::max(
+                        first_conflict = Some(ProgPoint::from_index(core::cmp::max(
                             preg_key.from,
                             key.from,
                         )));
@@ -334,7 +333,7 @@ impl<'a, F: Function> Env<'a, F> {
                     self.bundles[bundle.index()].prio,
                     final_weight
                 );
-                std::cmp::min(BUNDLE_MAX_NORMAL_SPILL_WEIGHT, final_weight)
+                core::cmp::min(BUNDLE_MAX_NORMAL_SPILL_WEIGHT, final_weight)
             } else {
                 0
             }
@@ -824,7 +823,7 @@ impl<'a, F: Function> Env<'a, F> {
                 // (up to the Before of the next inst), *unless*
                 // the original LR was only over the Before (up to
                 // the After) of this inst.
-                let to = std::cmp::min(ProgPoint::before(u.pos.inst().next()), lr_to);
+                let to = core::cmp::min(ProgPoint::before(u.pos.inst().next()), lr_to);
 
                 // If the last bundle was at the same inst, add a new
                 // LR to the same bundle; otherwise, create a LR and a
@@ -863,7 +862,7 @@ impl<'a, F: Function> Env<'a, F> {
 
                 // Otherwise, create a new LR.
                 let pos = ProgPoint::before(u.pos.inst());
-                let pos = std::cmp::max(lr_from, pos);
+                let pos = core::cmp::max(lr_from, pos);
                 let cr = CodeRange { from: pos, to };
                 let lr = self.create_liverange(cr);
                 new_lrs.push((vreg, lr));
@@ -1036,7 +1035,7 @@ impl<'a, F: Function> Env<'a, F> {
                     self.get_or_create_spill_bundle(bundle, /* create_if_absent = */ false)
                 {
                     let mut list =
-                        std::mem::replace(&mut self.bundles[bundle.index()].ranges, smallvec![]);
+                        core::mem::replace(&mut self.bundles[bundle.index()].ranges, smallvec![]);
                     for entry in &list {
                         self.ranges[entry.index.index()].bundle = spill;
                     }
@@ -1107,7 +1106,7 @@ impl<'a, F: Function> Env<'a, F> {
                     lowest_cost_evict_conflict_cost,
                     lowest_cost_split_conflict_cost,
                 ) {
-                    (Some(a), Some(b)) => Some(std::cmp::max(a, b)),
+                    (Some(a), Some(b)) => Some(core::cmp::max(a, b)),
                     _ => None,
                 };
                 match self.try_to_allocate_bundle_to_reg(bundle, preg_idx, scan_limit_cost) {
@@ -1291,7 +1290,7 @@ impl<'a, F: Function> Env<'a, F> {
                 );
                 let bundle_start = self.bundles[bundle.index()].ranges[0].range.from;
                 let mut split_at_point =
-                    std::cmp::max(lowest_cost_split_conflict_point, bundle_start);
+                    core::cmp::max(lowest_cost_split_conflict_point, bundle_start);
                 let requeue_with_reg = lowest_cost_split_conflict_reg;
 
                 // Adjust `split_at_point` if it is within a deeper loop
diff --git a/src/ion/redundant_moves.rs b/src/ion/redundant_moves.rs
index 41ed255..3aed48f 100644
--- a/src/ion/redundant_moves.rs
+++ b/src/ion/redundant_moves.rs
@@ -1,7 +1,6 @@
 //! Redundant-move elimination.
 
-use crate::{Allocation, VReg};
-use fxhash::FxHashMap;
+use crate::{Allocation, FxHashMap, VReg};
 use smallvec::{smallvec, SmallVec};
 
 #[derive(Copy, Clone, Debug, PartialEq, Eq)]
diff --git a/src/ion/reg_traversal.rs b/src/ion/reg_traversal.rs
index 0b457cb..729fd33 100644
--- a/src/ion/reg_traversal.rs
+++ b/src/ion/reg_traversal.rs
@@ -78,7 +78,7 @@ impl<'a> RegTraversalIter<'a> {
     }
 }
 
-impl<'a> std::iter::Iterator for RegTraversalIter<'a> {
+impl<'a> core::iter::Iterator for RegTraversalIter<'a> {
     type Item = PReg;
 
     fn next(&mut self) -> Option<PReg> {
diff --git a/src/ion/spill.rs b/src/ion/spill.rs
index 3583080..3068a65 100644
--- a/src/ion/spill.rs
+++ b/src/ion/spill.rs
@@ -138,7 +138,7 @@ impl<'a, F: Function> Env<'a, F> {
             let mut success = false;
             // Never probe the same element more than once: limit the
             // attempt count to the number of slots in existence.
-            for _attempt in 0..std::cmp::min(self.slots_by_size[size].slots.len(), MAX_ATTEMPTS) {
+            for _attempt in 0..core::cmp::min(self.slots_by_size[size].slots.len(), MAX_ATTEMPTS) {
                 // Note: this indexing of `slots` is always valid
                 // because either the `slots` list is empty and the
                 // iteration limit above consequently means we don't
diff --git a/src/ion/stackmap.rs b/src/ion/stackmap.rs
index 108835a..534e86b 100644
--- a/src/ion/stackmap.rs
+++ b/src/ion/stackmap.rs
@@ -12,6 +12,8 @@
 
 //! Stackmap computation.
 
+use alloc::vec::Vec;
+
 use super::{Env, ProgPoint, VRegIndex};
 use crate::{ion::data_structures::u64_key, Function};
 
diff --git a/src/lib.rs b/src/lib.rs
index 4b5a468..8aa9d4c 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -11,6 +11,12 @@
  */
 
 #![allow(dead_code)]
+#![no_std]
+
+#[cfg(feature = "std")]
+extern crate std;
+
+extern crate alloc;
 
 // Even when trace logging is disabled, the trace macro has a significant
 // performance cost so we disable it in release builds.
@@ -28,6 +34,11 @@ macro_rules! trace_enabled {
     };
 }
 
+use core::hash::BuildHasherDefault;
+use rustc_hash::FxHasher;
+type FxHashMap<K, V> = hashbrown::HashMap<K, V, BuildHasherDefault<FxHasher>>;
+type FxHashSet<V> = hashbrown::HashSet<V, BuildHasherDefault<FxHasher>>;
+
 pub(crate) mod cfg;
 pub(crate) mod domtree;
 pub mod indexset;
@@ -38,6 +49,8 @@ pub mod ssa;
 
 #[macro_use]
 mod index;
+
+use alloc::vec::Vec;
 pub use index::{Block, Inst, InstRange, InstRangeIter};
 
 pub mod checker;
@@ -142,8 +155,8 @@ impl PReg {
     }
 }
 
-impl std::fmt::Debug for PReg {
-    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
+impl core::fmt::Debug for PReg {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
         write!(
             f,
             "PReg(hw = {}, class = {:?}, index = {})",
@@ -154,8 +167,8 @@ impl std::fmt::Debug for PReg {
     }
 }
 
-impl std::fmt::Display for PReg {
-    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
+impl core::fmt::Display for PReg {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
         let class = match self.class() {
             RegClass::Int => "i",
             RegClass::Float => "f",
@@ -311,8 +324,8 @@ impl VReg {
     }
 }
 
-impl std::fmt::Debug for VReg {
-    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
+impl core::fmt::Debug for VReg {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
         write!(
             f,
             "VReg(vreg = {}, class = {:?})",
@@ -322,8 +335,8 @@ impl std::fmt::Debug for VReg {
     }
 }
 
-impl std::fmt::Display for VReg {
-    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
+impl core::fmt::Display for VReg {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
         write!(f, "v{}", self.vreg())
     }
 }
@@ -382,8 +395,8 @@ impl SpillSlot {
     }
 }
 
-impl std::fmt::Display for SpillSlot {
-    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
+impl core::fmt::Display for SpillSlot {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
         write!(f, "stack{}", self.index())
     }
 }
@@ -413,8 +426,8 @@ pub enum OperandConstraint {
     Reuse(usize),
 }
 
-impl std::fmt::Display for OperandConstraint {
-    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
+impl core::fmt::Display for OperandConstraint {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
         match self {
             Self::Any => write!(f, "any"),
             Self::Reg => write!(f, "reg"),
@@ -796,14 +809,14 @@ impl Operand {
     }
 }
 
-impl std::fmt::Debug for Operand {
-    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
-        std::fmt::Display::fmt(self, f)
+impl core::fmt::Debug for Operand {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
+        core::fmt::Display::fmt(self, f)
     }
 }
 
-impl std::fmt::Display for Operand {
-    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
+impl core::fmt::Display for Operand {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
         match (self.kind(), self.pos()) {
             (OperandKind::Def, OperandPos::Late) | (OperandKind::Use, OperandPos::Early) => {
                 write!(f, "{:?}", self.kind())?;
@@ -836,14 +849,14 @@ pub struct Allocation {
     bits: u32,
 }
 
-impl std::fmt::Debug for Allocation {
-    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
-        std::fmt::Display::fmt(self, f)
+impl core::fmt::Debug for Allocation {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
+        core::fmt::Display::fmt(self, f)
     }
 }
 
-impl std::fmt::Display for Allocation {
-    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
+impl core::fmt::Display for Allocation {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
         match self.kind() {
             AllocationKind::None => write!(f, "none"),
             AllocationKind::Reg => write!(f, "{}", self.as_reg().unwrap()),
@@ -1173,8 +1186,8 @@ pub struct ProgPoint {
     bits: u32,
 }
 
-impl std::fmt::Debug for ProgPoint {
-    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
+impl core::fmt::Debug for ProgPoint {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
         write!(
             f,
             "progpoint{}{}",
@@ -1422,9 +1435,9 @@ impl Output {
                 // binary_search_by returns the index of where it would have
                 // been inserted in Err.
                 if pos < ProgPoint::before(inst_range.first()) {
-                    std::cmp::Ordering::Less
+                    core::cmp::Ordering::Less
                 } else {
-                    std::cmp::Ordering::Greater
+                    core::cmp::Ordering::Greater
                 }
             })
             .unwrap_err();
@@ -1463,12 +1476,13 @@ pub enum RegAllocError {
     TooManyLiveRegs,
 }
 
-impl std::fmt::Display for RegAllocError {
-    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
+impl core::fmt::Display for RegAllocError {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
         write!(f, "{:?}", self)
     }
 }
 
+#[cfg(feature = "std")]
 impl std::error::Error for RegAllocError {}
 
 /// Run the allocator.
diff --git a/src/moves.rs b/src/moves.rs
index 8194b72..82c68f5 100644
--- a/src/moves.rs
+++ b/src/moves.rs
@@ -4,8 +4,8 @@
  */
 
 use crate::{ion::data_structures::u64_key, Allocation, PReg};
+use core::fmt::Debug;
 use smallvec::{smallvec, SmallVec};
-use std::fmt::Debug;
 
 /// A list of moves to be performed in sequence, with auxiliary data
 /// attached to each.
diff --git a/src/postorder.rs b/src/postorder.rs
index 96e9787..3e04eb0 100644
--- a/src/postorder.rs
+++ b/src/postorder.rs
@@ -6,6 +6,8 @@
 //! Fast postorder computation.
 
 use crate::Block;
+use alloc::vec;
+use alloc::vec::Vec;
 use smallvec::{smallvec, SmallVec};
 
 pub fn calculate<'a, SuccFn: Fn(Block) -> &'a [Block]>(
diff --git a/src/ssa.rs b/src/ssa.rs
index 8b9c6f8..ac6263e 100644
--- a/src/ssa.rs
+++ b/src/ssa.rs
@@ -5,7 +5,8 @@
 
 //! SSA-related utilities.
 
-use std::collections::HashSet;
+use alloc::vec;
+use hashbrown::HashSet;
 
 use crate::cfg::CFGInfo;
 use crate::{Block, Function, Inst, OperandKind, RegAllocError, VReg};