diff --git a/Cargo.toml b/Cargo.toml
index 663d9b1..dfd1fde 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -2,3 +2,6 @@
 authors = ["Jorge Aparicio <japaricious@gmail.com>"]
 name = "rustc_builtins"
 version = "0.1.0"
+
+[dev-dependencies]
+quickcheck = "0.3.1"
diff --git a/ci/install.sh b/ci/install.sh
index d60571d..9c971c6 100644
--- a/ci/install.sh
+++ b/ci/install.sh
@@ -49,19 +49,19 @@ install_c_toolchain() {
             ;;
         mips-unknown-linux-gnu)
             apt-get install -y --no-install-recommends \
-                    gcc-mips-linux-gnu libc6-dev-mips-cross
+                    gcc gcc-mips-linux-gnu libc6-dev libc6-dev-mips-cross
             ;;
         mipsel-unknown-linux-gnu)
             apt-get install -y --no-install-recommends \
-                    gcc-mipsel-linux-gnu libc6-dev-mipsel-cross
+                    gcc gcc-mipsel-linux-gnu libc6-dev libc6-dev-mipsel-cross
             ;;
         powerpc64-unknown-linux-gnu)
             apt-get install -y --no-install-recommends \
-                    gcc-powerpc64-linux-gnu libc6-dev-ppc64-cross
+                    gcc gcc-powerpc64-linux-gnu libc6-dev libc6-dev-ppc64-cross
             ;;
         powerpc64le-unknown-linux-gnu)
             apt-get install -y --no-install-recommends \
-                    gcc-powerpc64le-linux-gnu libc6-dev-ppc64el-cross
+                    gcc gcc-powerpc64le-linux-gnu libc6-dev libc6-dev-ppc64el-cross
             ;;
     esac
 }
diff --git a/src/arm.rs b/src/arm.rs
index 7beb91a..51206ed 100644
--- a/src/arm.rs
+++ b/src/arm.rs
@@ -1,3 +1,37 @@
+use core::intrinsics;
+
+// NOTE This function and the one below are implemented using assembly because they using a custom
+// calling convention which can't be implemented using a normal Rust function
+// TODO use `global_asm!`
+#[naked]
+#[no_mangle]
+pub unsafe extern "aapcs" fn __aeabi_uidivmod() {
+    asm!("push    { lr }
+          sub     sp, sp, #4
+          mov     r2, sp
+          bl      __udivmodsi4
+          ldr     r1, [sp]
+          add     sp, sp, #4
+          pop     { pc }");
+    intrinsics::unreachable();
+}
+
+// TODO use `global_asm!`
+#[naked]
+#[no_mangle]
+pub unsafe extern "aapcs" fn __aeabi_uldivmod() {
+    asm!("push	{r11, lr}
+          sub	sp, sp, #16
+          add	r12, sp, #8
+          str	r12, [sp]
+          bl	__udivmoddi4
+          ldr	r2, [sp, #8]
+          ldr	r3, [sp, #12]
+          add	sp, sp, #16
+          pop	{r11, pc}");
+    intrinsics::unreachable();
+}
+
 extern "C" {
     fn memcpy(dest: *mut u8, src: *const u8, n: usize) -> *mut u8;
     fn memmove(dest: *mut u8, src: *const u8, n: usize) -> *mut u8;
diff --git a/src/div.rs b/src/div.rs
new file mode 100644
index 0000000..44d59b7
--- /dev/null
+++ b/src/div.rs
@@ -0,0 +1,271 @@
+use {Int, LargeInt, U64};
+
+/// Returns `n / d`
+#[no_mangle]
+pub extern "C" fn __udivsi3(n: u32, d: u32) -> u32 {
+    let u32_bits = u32::bits() as u32;
+
+    // Special cases
+    if d == 0 {
+        panic!("Division by zero");
+    }
+
+    if n == 0 {
+        return 0;
+    }
+
+    let mut sr = d.leading_zeros().wrapping_sub(n.leading_zeros());
+
+    // d > n
+    if sr > u32_bits - 1 {
+        return 0;
+    }
+
+    // d == 1
+    if sr == u32_bits - 1 {
+        return n;
+    }
+
+    sr = sr + 1;
+
+    // 1 <= sr <= u32_bits - 1
+    let mut q = n << (u32_bits - sr);
+    let mut r = n >> sr;
+
+    let mut carry = 0;
+    for _ in 0..sr {
+        // r:q = ((r:q) << 1) | carry
+        r = (r << 1) | (q >> (u32_bits - 1));
+        q = (q << 1) | carry;
+
+        // carry = 0;
+        // if r > d {
+        //     r -= d;
+        //     carry = 1;
+        // }
+
+        let s = (d.wrapping_sub(r).wrapping_sub(1)) as i32 >> (u32_bits - 1);
+        carry = (s & 1) as u32;
+        r -= d & s as u32;
+    }
+
+    (q << 1) | carry
+}
+
+/// Returns `n / d` and sets `*rem = n % d`
+#[no_mangle]
+pub extern "C" fn __udivmodsi4(a: u32, b: u32, rem: Option<&mut u32>) -> u32 {
+    let d = __udivsi3(a, b);
+    if let Some(rem) = rem {
+        *rem = a - (d * b);
+    }
+    return d;
+}
+
+/// Returns `n / d` and sets `*rem = n % d`
+#[no_mangle]
+pub extern "C" fn __udivmoddi4(n: u64, d: u64, rem: Option<&mut u64>) -> u64 {
+    let u32_bits = u32::bits() as u32;
+    let u64_bits = u64::bits() as u32;
+
+    // NOTE X is unknown, K != 0
+    if n.high() == 0 {
+        if d.high() == 0 {
+            // 0 X
+            // ---
+            // 0 X
+
+            if let Some(rem) = rem {
+                *rem = u64::from(n.low() % d.low());
+            }
+            return u64::from(n.low() / d.low());
+        } else
+        // d.high() != 0
+        {
+            // 0 X
+            // ---
+            // K X
+
+            if let Some(rem) = rem {
+                *rem = u64::from(n.low());
+            }
+            return 0;
+        };
+    }
+
+    let mut sr;
+    let mut q = U64 { low: 0, high: 0 };
+    let mut r = U64 { low: 0, high: 0 };
+
+    // n.high() != 0
+    if d.low() == 0 {
+        if d.high() == 0 {
+            // K X
+            // ---
+            // 0 0
+
+            panic!("Division by zero");
+        }
+
+        // d.high() != 0
+        if n.low() == 0 {
+            // K 0
+            // ---
+            // K 0
+
+            if let Some(rem) = rem {
+                *rem = U64 {
+                    low: 0,
+                    high: n.high() % d.high(),
+                }[..];
+            }
+            return u64::from(n.high() / d.high());
+        }
+
+        // n.low() != 0
+        // K K
+        // ---
+        // K 0
+
+        if d.high().is_power_of_two() {
+            if let Some(rem) = rem {
+                *rem = U64 {
+                    low: n.low(),
+                    high: n.high() & (d.high() - 1),
+                }[..];
+            }
+
+            return u64::from(n.high() >> d.high().trailing_zeros());
+        }
+
+        sr = d.high().leading_zeros().wrapping_sub(n.high().leading_zeros());
+
+        // D > N
+        if sr > u32_bits - 2 {
+            if let Some(rem) = rem {
+                *rem = n;
+            }
+            return 0;
+        }
+
+        sr = sr + 1;
+
+        // 1 <= sr <= u32_bits - 1
+        // q = n << (u64_bits - sr);
+        q.low = 0;
+        q.high = n.low() << (u32_bits - sr);
+        // r = n >> sr
+        r.high = n.high() >> sr;
+        r.low = (n.high() << (u32_bits - sr)) | (n.low() >> sr);
+    } else
+    // d.low() != 0
+    {
+        if d.high() == 0 {
+            // K X
+            // ---
+            // 0 K
+            if d.low().is_power_of_two() {
+                if let Some(rem) = rem {
+                    *rem = u64::from(n.low() & (d.low() - 1));
+                }
+
+                if d.low() == 1 {
+                    return n;
+                } else {
+                    let sr = d.low().trailing_zeros();
+                    return U64 {
+                        low: (n.high() << (u32_bits - sr)) | (n.low() >> sr),
+                        high: n.high() >> sr,
+                    }[..];
+                };
+            }
+
+            sr = 1 + u32_bits + d.low().leading_zeros() - n.high().leading_zeros();
+
+            // 2 <= sr <= u64_bits - 1
+            // q = n << (u64_bits - sr)
+            // r = n >> sr;
+            if sr == u32_bits {
+                q.low = 0;
+                q.high = n.low();
+                r.high = 0;
+                r.low = n.high();
+            } else if sr < u32_bits
+            // 2 <= sr <= u32_bits - 1
+            {
+                q.low = 0;
+                q.high = n.low() << (u32_bits - sr);
+                r.high = n.high() >> sr;
+                r.low = (n.high() << (u32_bits - sr)) | (n.low() >> sr);
+            } else
+            // u32_bits + 1 <= sr <= u64_bits - 1
+            {
+                q.low = n.low() << (u64_bits - sr);
+                q.high = (n.high() << (u64_bits - sr)) | (n.low() >> (sr - u32_bits));
+                r.high = 0;
+                r.low = n.high() >> (sr - u32_bits);
+            }
+
+        } else
+        // d.high() != 0
+        {
+            // K X
+            // ---
+            // K K
+
+            sr = d.high().leading_zeros().wrapping_sub(n.high().leading_zeros());
+
+            // D > N
+            if sr > u32_bits - 1 {
+                if let Some(rem) = rem {
+                    *rem = n;
+                    return 0;
+                }
+            }
+
+            sr += 1;
+
+            // 1 <= sr <= u32_bits
+            // q = n << (u64_bits - sr)
+            q.low = 0;
+            if sr == u32_bits {
+                q.high = n.low();
+                r.high = 0;
+                r.low = n.high();
+            } else {
+                q.high = n.low() << (u32_bits - sr);
+                r.high = n.high() >> sr;
+                r.low = (n.high() << (u32_bits - sr)) | (n.low() >> sr);
+            }
+        }
+    }
+
+    // Not a special case
+    // q and r are initialized with
+    // q = n << (u64_bits - sr)
+    // r = n >> sr
+    // 1 <= sr <= u64_bits - 1
+    let mut carry = 0;
+
+    for _ in 0..sr {
+        // r:q = ((r:q) << 1) | carry
+        r[..] = (r[..] << 1) | (q[..] >> 63);
+        q[..] = (q[..] << 1) | carry as u64;
+
+        // carry = 0
+        // if r >= d {
+        //     r -= d;
+        //     carry = 1;
+        // }
+
+        let s = (d.wrapping_sub(r[..]).wrapping_sub(1)) as i64 >> (u64_bits - 1);
+        carry = (s & 1) as u32;
+        r[..] -= d & s as u64;
+    }
+
+    q[..] = (q[..] << 1) | carry as u64;
+    if let Some(rem) = rem {
+        *rem = r[..];
+    }
+    q[..]
+}
diff --git a/src/lib.rs b/src/lib.rs
index 3ef1354..3a24025 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -8,13 +8,20 @@
 
 #[cfg(test)]
 extern crate core;
+#[cfg(test)]
+#[macro_use]
+extern crate quickcheck;
 
 #[cfg(target_arch = "arm")]
 pub mod arm;
 
+pub mod div;
+
 #[cfg(test)]
 mod test;
 
+use core::ops::{Index, IndexMut, RangeFull};
+
 /// Trait for some basic operations on integers
 trait Int {
     fn bits() -> usize;
@@ -22,16 +29,24 @@ trait Int {
 
 // TODO: Once i128/u128 support lands, we'll want to add impls for those as well
 impl Int for u32 {
-    fn bits() -> usize { 32 }
+    fn bits() -> usize {
+        32
+    }
 }
 impl Int for i32 {
-    fn bits() -> usize { 32 }
+    fn bits() -> usize {
+        32
+    }
 }
 impl Int for u64 {
-    fn bits() -> usize { 64 }
+    fn bits() -> usize {
+        64
+    }
 }
 impl Int for i64 {
-    fn bits() -> usize { 64 }
+    fn bits() -> usize {
+        64
+    }
 }
 
 /// Trait to convert an integer to/from smaller parts
@@ -74,6 +89,36 @@ impl LargeInt for i64 {
     }
 }
 
+/// Union-like access to the 32-bit words that make an `u64`: `x.low` and `x.high`. The whole `u64`
+/// can be accessed via the expression `x[..]`, which can be used in lvalue or rvalue position.
+#[cfg(target_endian = "little")]
+#[repr(C)]
+struct U64 {
+    low: u32,
+    high: u32,
+}
+
+#[cfg(target_endian = "big")]
+#[repr(C)]
+struct U64 {
+    high: u32,
+    low: u32,
+}
+
+impl Index<RangeFull> for U64 {
+    type Output = u64;
+
+    fn index(&self, _: RangeFull) -> &u64 {
+        unsafe { &*(self as *const _ as *const u64) }
+    }
+}
+
+impl IndexMut<RangeFull> for U64 {
+    fn index_mut(&mut self, _: RangeFull) -> &mut u64 {
+        unsafe { &mut *(self as *const _ as *mut u64) }
+    }
+}
+
 macro_rules! absv_i2 {
     ($intrinsic:ident : $ty:ty) => {
         #[no_mangle]
diff --git a/src/test.rs b/src/test.rs
index 9abfec7..fa8691c 100644
--- a/src/test.rs
+++ b/src/test.rs
@@ -1,5 +1,7 @@
 use std::panic;
 
+use quickcheck::TestResult;
+
 macro_rules! absv_i2 {
     ($intrinsic:ident: $ty:ident) => {
         #[test]
@@ -23,3 +25,32 @@ absv_i2!(__absvsi2: i32);
 absv_i2!(__absvdi2: i64);
 // TODO(rust-lang/35118)?
 // absv_i2!(__absvti2: i128);
+
+quickcheck! {
+    fn udivmoddi4(n: (u32, u32), d: (u32, u32)) -> TestResult {
+        let n = ::U64 { low: n.0, high: n.1 }[..];
+        let d = ::U64 { low: d.0, high: d.1 }[..];
+
+        if d == 0 {
+            TestResult::discard()
+        } else {
+            let mut r = 0;
+            let q = ::div::__udivmoddi4(n, d, Some(&mut r));
+
+            TestResult::from_bool(q * d + r == n)
+        }
+    }
+}
+
+quickcheck! {
+    fn udivmodsi4(n: u32, d: u32) -> TestResult {
+        if d == 0 {
+            TestResult::discard()
+        } else {
+            let mut r = 0;
+            let q = ::div::__udivmodsi4(n, d, Some(&mut r));
+
+            TestResult::from_bool(q * d + r == n)
+        }
+    }
+}