diff --git a/Cargo.toml b/Cargo.toml
index 19443ab..5a46cc0 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -8,4 +8,7 @@ description = "A font renderer written (mostly) in pure, safe Rust"
 repository = "https://github.com/google/font-rs"
 
 [features]
+default = ["std"]
+
 sse = []
+std = []
diff --git a/src/accumulate.rs b/src/accumulate.rs
index d41c536..8f85f8a 100644
--- a/src/accumulate.rs
+++ b/src/accumulate.rs
@@ -12,22 +12,29 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use std::mem;
+#[cfg(not(feature = "std"))]
+use alloc::vec::Vec;
+
+#[cfg(all(feature = "sse", not(feature = "std")))]
+use core::arch::x86_64::*;
 
-#[cfg(target_arch = "x86_64")]
+#[cfg(all(feature = "sse", feature = "std"))]
 use std::arch::x86_64::*;
 
-#[cfg(target_arch = "x86")]
-use std::arch::x86::*;
+#[cfg(all(feature = "sse", not(feature = "std")))]
+use core::mem;
 
-macro_rules! _mm_shuffle {
-    ($z:expr, $y:expr, $x:expr, $w:expr) => {
-        ($z << 6) | ($y << 4) | ($x << 2) | $w
-    };
-}
+#[cfg(all(feature = "sse", feature = "std"))]
+use std::mem;
 
 #[cfg(feature = "sse")]
 pub fn accumulate(src: &[f32]) -> Vec<u8> {
+    macro_rules! _mm_shuffle {
+        ($z:expr, $y:expr, $x:expr, $w:expr) => {
+            ($z << 6) | ($y << 4) | ($x << 2) | $w
+        };
+    }
+
     // SIMD instructions force us to align data since we iterate each 4 elements
     // So:
     // n (0) => 0
@@ -68,9 +75,9 @@ pub fn accumulate(src: &[f32]) -> Vec<u8> {
 
 #[cfg(not(feature = "sse"))]
 pub fn accumulate(src: &[f32]) -> Vec<u8> {
-    let mut acc = 0.0;
-    src.iter()
-        .map(|c| {
+    let mut acc = 0.0f32;
+    src.into_iter()
+        .map(|&c| {
             // This would translate really well to SIMD
             acc += c;
             let y = acc.abs();
@@ -88,8 +95,8 @@ mod tests {
     //  accumulate fn
     fn accumulate_simple_impl(src: &[f32]) -> Vec<u8> {
         let mut acc = 0.0;
-        src.iter()
-            .map(|c| {
+        src.into_iter()
+            .map(|&c| {
                 acc += c;
                 let y = acc.abs();
                 let y = if y < 1.0 { y } else { 1.0 };
diff --git a/src/float32.rs b/src/float32.rs
new file mode 100644
index 0000000..079bda2
--- /dev/null
+++ b/src/float32.rs
@@ -0,0 +1,28 @@
+
+use core::intrinsics::*;
+
+/// Abstraction for floating point intrinsics in no_std mode
+pub trait Float32 {
+    fn abs(self) -> f32;
+    fn floor(self) -> f32;
+    fn ceil(self) -> f32;
+    fn sqrt(self) -> f32;
+}
+
+impl Float32 for f32 {
+    fn abs(self) -> f32 {
+        unsafe { fabsf32(self) }
+    }
+
+    fn floor(self) -> f32 {
+        unsafe { floorf32(self) }
+    }
+
+    fn ceil(self) -> f32 {
+        unsafe { ceilf32(self) }
+    }
+
+    fn sqrt(self) -> f32 {
+        unsafe { sqrtf32(self) }
+    }
+}
diff --git a/src/font.rs b/src/font.rs
index 8f741fc..d676e62 100644
--- a/src/font.rs
+++ b/src/font.rs
@@ -14,15 +14,33 @@
 
 //! A simple renderer for TrueType fonts
 
-use std::collections::HashMap;
+#[cfg(not(feature = "std"))]
+use alloc::collections::BTreeMap;
+#[cfg(not(feature = "std"))]
+use alloc::vec::Vec;
+#[cfg(not(feature = "std"))]
+use alloc::str;
+#[cfg(not(feature = "std"))]
+use alloc::fmt;
+#[cfg(not(feature = "std"))]
+use alloc::fmt::{Debug, Display, Formatter};
+
+#[cfg(feature = "std")]
+use std::collections::BTreeMap;
+#[cfg(feature = "std")]
+use std::str;
+#[cfg(feature = "std")]
 use std::fmt;
+#[cfg(feature = "std")]
 use std::fmt::{Debug, Display, Formatter};
-use std::result::Result;
 
 use geom::{affine_pt, Affine, Point};
 use raster::Raster;
 
-#[derive(PartialEq, Eq, Hash)]
+#[cfg(not(feature = "std"))]
+use float32::Float32;
+
+#[derive(PartialEq, Eq, PartialOrd, Ord)]
 struct Tag(u32);
 
 impl Tag {
@@ -34,13 +52,13 @@ impl Tag {
 impl Display for Tag {
     fn fmt(&self, f: &mut Formatter) -> fmt::Result {
         let &Tag(tag) = self;
-        let buf = vec![
+        let buf = [
             ((tag >> 24) & 0xff) as u8,
             ((tag >> 16) & 0xff) as u8,
             ((tag >> 8) & 0xff) as u8,
             (tag & 0xff) as u8,
         ];
-        f.write_str(&String::from_utf8(buf).unwrap())
+        f.write_str(str::from_utf8(&buf).unwrap())
     }
 }
 
@@ -672,7 +690,7 @@ impl<'a> CompoundGlyph<'a> {
 
 pub struct Font<'a> {
     _version: u32,
-    _tables: HashMap<Tag, &'a [u8]>,
+    _tables: BTreeMap<Tag, &'a [u8]>,
     head: Head<'a>,
     maxp: Maxp<'a>,
     cmap: Option<Cmap<'a>>,
@@ -739,7 +757,7 @@ impl<'a> Font<'a> {
                 }
             }
             _ => {
-                println!("unhandled glyph case");
+                unreachable!("unhandled glyph case");
             }
         }
     }
@@ -776,8 +794,7 @@ impl<'a> Font<'a> {
                 })
             }
             _ => {
-                println!("glyph {} error", glyph_id);
-                None
+                unreachable!("glyph {} error", glyph_id);
             }
         }
     }
@@ -950,7 +967,7 @@ pub fn parse(data: &[u8]) -> Result<Font, FontError> {
     let _search_range = get_u16(data, 6).unwrap();
     let _entry_selector = get_u16(data, 8).unwrap();
     let _range_shift = get_u16(data, 10).unwrap();
-    let mut tables = HashMap::new();
+    let mut tables = BTreeMap::new();
     for i in 0..num_tables {
         let header = &data[12 + i * 16..12 + (i + 1) * 16];
         let tag = get_u32(header, 0).unwrap();
diff --git a/src/geom.rs b/src/geom.rs
index c292cad..ac97ca6 100644
--- a/src/geom.rs
+++ b/src/geom.rs
@@ -14,6 +14,10 @@
 
 //! Geometry primitive data structures and manipulations
 
+#[cfg(not(feature = "std"))]
+use alloc::fmt::{Debug, Formatter, Result};
+
+#[cfg(feature = "std")]
 use std::fmt::{Debug, Formatter, Result};
 
 #[derive(Copy, Clone)]
diff --git a/src/lib.rs b/src/lib.rs
index b7b189f..ca79947 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -14,9 +14,19 @@
 
 //! A very high performance font renderer.
 
+#![cfg_attr(not(feature = "std"), no_std)]
+#![cfg_attr(not(feature = "std"), feature(alloc, core_intrinsics))]
+
+#[cfg(not(feature = "std"))]
+#[macro_use]
+extern crate alloc;
+
 #[macro_use]
 pub mod macros;
 pub mod accumulate;
 pub mod font;
 pub mod geom;
 pub mod raster;
+
+#[cfg(not(feature = "std"))]
+mod float32;
diff --git a/src/raster.rs b/src/raster.rs
index 7caaa9d..10d6e80 100644
--- a/src/raster.rs
+++ b/src/raster.rs
@@ -14,11 +14,15 @@
 
 //! An antialiased rasterizer for quadratic Beziers
 
-use std::cmp::min;
+#[cfg(not(feature = "std"))]
+use alloc::vec::Vec;
 
 use accumulate::accumulate;
 use geom::Point;
 
+#[cfg(not(feature = "std"))]
+use float32::Float32;
+
 // TODO: sort out crate structure. Right now we want this when compiling raster as a binary,
 // but need it commented out when compiling showttf
 //mod geom;
@@ -59,7 +63,8 @@ impl Raster {
         if p0.y < 0.0 {
             x -= p0.y * dxdy;
         }
-        for y in y0..min(self.h, p1.y.ceil() as usize) {
+        let ymin = self.h.min(p1.y.ceil() as usize);
+        for y in y0..ymin {
             let linestart = y * self.w;
             let dy = ((y + 1) as f32).min(p1.y) - (y as f32).max(p0.y);
             let xnext = x + dxdy * dy;