diff --git a/CMakeLists.txt b/CMakeLists.txt
index efe9802f2d..270fc6fd4e 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -132,6 +132,7 @@ set(C_HEADERS
 set(ZIG_STD_SRC
     "${CMAKE_SOURCE_DIR}/std/bootstrap.zig"
     "${CMAKE_SOURCE_DIR}/std/builtin.zig"
+    "${CMAKE_SOURCE_DIR}/std/compiler_rt.zig"
     "${CMAKE_SOURCE_DIR}/std/test_runner.zig"
     "${CMAKE_SOURCE_DIR}/std/test_runner_libc.zig"
     "${CMAKE_SOURCE_DIR}/std/test_runner_nolibc.zig"
diff --git a/README.md b/README.md
index 803d1da590..bfae0f4409 100644
--- a/README.md
+++ b/README.md
@@ -81,7 +81,7 @@ and `ZIG_LIBC_STATIC_LIB_DIR` should be set to (example below).
 ```
 mkdir build
 cd build
-cmake .. -DCMAKE_INSTALL_PREFIX=$(pwd) -DZIG_LIBC_LIB_DIR=$(dirname $(cc -print-file-name=crt1.o)) -DZIG_LIBC_INCLUDE_DIR=/usr/include -DZIG_LIBC_STATIC_LIB_DIR=$(dirname $(cc -print-file-name=crtbeginT.o))
+cmake .. -DCMAKE_INSTALL_PREFIX=$(pwd) -DZIG_LIBC_LIB_DIR=$(dirname $(cc -print-file-name=crt1.o)) -DZIG_LIBC_INCLUDE_DIR=/usr/include -DZIG_LIBC_STATIC_LIB_DIR=$(dirname $(cc -print-file-name=crtbegin.o))
 make
 make install
 ./run_tests
diff --git a/src/all_types.hpp b/src/all_types.hpp
index adb9b34f95..037d34ddd9 100644
--- a/src/all_types.hpp
+++ b/src/all_types.hpp
@@ -1038,6 +1038,8 @@ enum BuiltinFnId {
     BuiltinFnIdCUndef,
     BuiltinFnIdCompileVar,
     BuiltinFnIdConstEval,
+    BuiltinFnIdCtz,
+    BuiltinFnIdClz,
 };
 
 struct BuiltinFnEntry {
@@ -1161,6 +1163,7 @@ struct CodeGen {
     uint32_t error_value_count;
     TypeTableEntry *err_tag_type;
     LLVMValueRef int_overflow_fns[2][3][4]; // [0-signed,1-unsigned][0-add,1-sub,2-mul][0-8,1-16,2-32,3-64]
+    LLVMValueRef int_builtin_fns[2][4]; // [0-ctz,1-clz][0-8,1-16,2-32,3-64]
 
     const char **clang_argv;
     int clang_argv_len;
diff --git a/src/analyze.cpp b/src/analyze.cpp
index fc438506c4..3e57d646c0 100644
--- a/src/analyze.cpp
+++ b/src/analyze.cpp
@@ -4304,6 +4304,29 @@ static TypeTableEntry *analyze_builtin_fn_call_expr(CodeGen *g, ImportTableEntry
 
                 return resolved_type;
             }
+        case BuiltinFnIdCtz:
+        case BuiltinFnIdClz:
+            {
+                AstNode *type_node = node->data.fn_call_expr.params.at(0);
+                TypeTableEntry *int_type = analyze_type_expr(g, import, context, type_node);
+                if (int_type->id == TypeTableEntryIdInvalid) {
+                    return int_type;
+                } else if (int_type->id == TypeTableEntryIdInt) {
+                    AstNode **expr_node = node->data.fn_call_expr.params.at(1)->parent_field;
+                    TypeTableEntry *resolved_type = analyze_expression(g, import, context, int_type, *expr_node);
+                    if (resolved_type->id == TypeTableEntryIdInvalid) {
+                        return resolved_type;
+                    }
+
+                    // TODO const expr eval
+
+                    return resolved_type;
+                } else {
+                    add_node_error(g, type_node,
+                        buf_sprintf("expected integer type, got '%s'", buf_ptr(&int_type->name)));
+                    return g->builtin_types.entry_invalid;
+                }
+            }
 
     }
     zig_unreachable();
diff --git a/src/codegen.cpp b/src/codegen.cpp
index bf208a527b..7cfd699ce4 100644
--- a/src/codegen.cpp
+++ b/src/codegen.cpp
@@ -227,6 +227,24 @@ static LLVMValueRef get_int_overflow_fn(CodeGen *g, TypeTableEntry *type_entry,
     return *fn;
 }
 
+static LLVMValueRef get_int_builtin_fn(CodeGen *g, TypeTableEntry *int_type, BuiltinFnId fn_id) {
+    // [0-ctz,1-clz][0-8,1-16,2-32,3-64]
+    int index0 = (fn_id == BuiltinFnIdCtz) ? 0 : 1;
+    int index1 = bits_index(int_type->data.integral.bit_count);
+    LLVMValueRef *fn = &g->int_builtin_fns[index0][index1];
+    if (!*fn) {
+        const char *fn_name = (fn_id == BuiltinFnIdCtz) ? "cttz" : "ctlz";
+        Buf *llvm_name = buf_sprintf("llvm.%s.i%d", fn_name, int_type->data.integral.bit_count);
+        LLVMTypeRef param_types[] = {
+            int_type->type_ref,
+            LLVMInt1Type(),
+        };
+        LLVMTypeRef fn_type = LLVMFunctionType(int_type->type_ref, param_types, 2, false);
+        *fn = LLVMAddFunction(g->module, buf_ptr(llvm_name), fn_type);
+    }
+    return *fn;
+}
+
 static LLVMValueRef get_handle_value(CodeGen *g, AstNode *source_node, LLVMValueRef ptr, TypeTableEntry *type) {
     if (handle_is_ptr(type)) {
         return ptr;
@@ -249,6 +267,22 @@ static LLVMValueRef gen_builtin_fn_call_expr(CodeGen *g, AstNode *node) {
         case BuiltinFnIdCDefine:
         case BuiltinFnIdCUndef:
             zig_unreachable();
+        case BuiltinFnIdCtz:
+        case BuiltinFnIdClz:
+            {
+                int fn_call_param_count = node->data.fn_call_expr.params.length;
+                assert(fn_call_param_count == 2);
+                TypeTableEntry *int_type = get_type_for_type_node(node->data.fn_call_expr.params.at(0));
+                assert(int_type->id == TypeTableEntryIdInt);
+                LLVMValueRef fn_val = get_int_builtin_fn(g, int_type, builtin_fn->id);
+                LLVMValueRef operand = gen_expr(g, node->data.fn_call_expr.params.at(1));
+                LLVMValueRef params[] {
+                    operand,
+                    LLVMConstNull(LLVMInt1Type()),
+                };
+                add_debug_source_node(g, node);
+                return LLVMBuildCall(g->builder, fn_val, params, 2, "");
+            }
         case BuiltinFnIdAddWithOverflow:
         case BuiltinFnIdSubWithOverflow:
         case BuiltinFnIdMulWithOverflow:
@@ -1025,11 +1059,12 @@ static LLVMValueRef gen_prefix_op_expr(CodeGen *g, AstNode *node) {
         case PrefixOpDereference:
             {
                 LLVMValueRef expr = gen_expr(g, expr_node);
+                assert(expr_type->id == TypeTableEntryIdPointer);
                 if (!type_has_bits(expr_type)) {
                     return nullptr;
                 } else {
-                    add_debug_source_node(g, node);
-                    return LLVMBuildLoad(g->builder, expr, "");
+                    TypeTableEntry *child_type = expr_type->data.pointer.child_type;
+                    return get_handle_value(g, node, expr, child_type);
                 }
             }
         case PrefixOpMaybe:
@@ -3590,6 +3625,8 @@ static void define_builtin_fns(CodeGen *g) {
     create_builtin_fn_with_arg_count(g, BuiltinFnIdCUndef, "c_undef", 1);
     create_builtin_fn_with_arg_count(g, BuiltinFnIdCompileVar, "compile_var", 1);
     create_builtin_fn_with_arg_count(g, BuiltinFnIdConstEval, "const_eval", 1);
+    create_builtin_fn_with_arg_count(g, BuiltinFnIdCtz, "ctz", 2);
+    create_builtin_fn_with_arg_count(g, BuiltinFnIdClz, "clz", 2);
 }
 
 static void init(CodeGen *g, Buf *source_path) {
diff --git a/src/link.cpp b/src/link.cpp
index 318d6ffa34..5d684c46e6 100644
--- a/src/link.cpp
+++ b/src/link.cpp
@@ -205,6 +205,9 @@ static void construct_linker_job_linux(LinkJob *lj) {
     if (!g->link_libc && (g->out_type == OutTypeExe || g->out_type == OutTypeLib)) {
         Buf *builtin_o_path = build_o(g, "builtin");
         lj->args.append(buf_ptr(builtin_o_path));
+
+        Buf *compiler_rt_o_path = build_o(g, "compiler_rt");
+        lj->args.append(buf_ptr(compiler_rt_o_path));
     }
 
     auto it = g->link_table.entry_iterator();
diff --git a/std/compiler_rt.zig b/std/compiler_rt.zig
new file mode 100644
index 0000000000..e239580139
--- /dev/null
+++ b/std/compiler_rt.zig
@@ -0,0 +1,269 @@
+const CHAR_BIT = 8;
+const du_int = c_ulonglong;
+const di_int = c_longlong;
+const si_int = c_int;
+const su_int = c_uint;
+
+const udwords = [2]su_int;
+const low = if (@compile_var("is_big_endian")) 1 else 0;
+const high = 1 - low;
+
+export fn __udivdi3(a: du_int, b: du_int) -> du_int {
+    return __udivmoddi4(a, b, null);
+}
+
+fn du_int_to_udwords(x: du_int) -> udwords {
+    // TODO ability to take address of params
+    const x2 = x;
+    return *(&udwords)(&x2);
+}
+
+export fn __udivmoddi4(a: du_int, b: du_int, maybe_rem: ?&du_int) -> du_int {
+    const n_uword_bits = @sizeof(su_int) * CHAR_BIT;
+    const n_udword_bits = @sizeof(du_int) * CHAR_BIT;
+    var n = du_int_to_udwords(a);
+    var d = du_int_to_udwords(b);
+    var q: udwords = undefined;
+    var r: udwords = undefined;
+    var sr: c_uint = undefined;
+    /* special cases, X is unknown, K != 0 */
+    if (n[high] == 0) {
+        if (d[high] == 0) {
+            /* 0 X
+             * ---
+             * 0 X
+             */
+            if (const rem ?= maybe_rem) {
+                *rem = n[low] % d[low];
+            }
+            return n[low] / d[low];
+        }
+        /* 0 X
+         * ---
+         * K X
+         */
+        if (const rem ?= maybe_rem) {
+            *rem = n[low];
+        }
+        return 0;
+    }
+    /* n[high] != 0 */
+    if (d[low] == 0) {
+        if (d[high] == 0) {
+            /* K X
+             * ---
+             * 0 0
+             */ 
+            if (var rem ?= maybe_rem) {
+                *rem = n[high] % d[low];
+            }
+            return n[high] / d[low];
+        }
+        /* d[high] != 0 */
+        if (n[low] == 0) {
+            /* K 0
+             * ---
+             * K 0
+             */
+            if (var rem ?= maybe_rem) {
+                r[high] = n[high] % d[high];
+                r[low] = 0;
+                *rem = *(&du_int)(&r[0]);
+            }
+            return n[high] / d[high];
+        }
+        /* K K
+         * ---
+         * K 0
+         */
+        // if d is a power of 2
+        if ((d[high] & (d[high] - 1)) == 0) {
+            if (var rem ?= maybe_rem) {
+                r[low] = n[low];
+                r[high] = n[high] & (d[high] - 1);
+                *rem = *(&du_int)(&r[0]);
+            }
+            return n[high] >> @ctz(@typeof(d[high]), d[high]);
+        }
+        /* K K
+         * ---
+         * K 0
+         */
+        sr = @clz(su_int, d[high]) - @clz(su_int, n[high]);
+        /* 0 <= sr <= n_uword_bits - 2 or sr large */
+        if (sr > n_uword_bits - 2) {
+            if (var rem ?= maybe_rem) {
+                *rem = *(&du_int)(&n[0]);
+            }
+            return 0;
+        }
+        sr += 1;
+        /* 1 <= sr <= n_uword_bits - 1 */
+        /* q.all = n.all << (n_udword_bits - sr); */
+        q[low] = 0;
+        q[high] = n[low] << (n_uword_bits - sr);
+        /* r.all = n.all >> sr; */
+        r[high] = n[high] >> sr;
+        r[low] = (n[high] << (n_uword_bits - sr)) | (n[low] >> sr);
+    } else {
+        /* d[low] != 0 */
+        if (d[high] == 0) {
+            /* K X
+             * ---
+             * 0 K
+             */
+            /* if d is a power of 2 */
+            if ((d[low] & (d[low] - 1)) == 0) {
+                if (var rem ?= maybe_rem) {
+                    *rem = n[low] & (d[low] - 1);
+                }
+                if (d[low] == 1) {
+                    return *(&du_int)(&n[0]);
+                }
+                sr = @ctz(@typeof(d[low]), d[low]);
+                q[high] = n[high] >> sr;
+                q[low] = (n[high] << (n_uword_bits - sr)) | (n[low] >> sr);
+                return *(&du_int)(&q[0]);
+            }
+            /* K X
+             * ---
+             * 0 K
+             */
+            sr = 1 + n_uword_bits + @clz(su_int, d[low]) - @clz(su_int, n[high]);
+            /* 2 <= sr <= n_udword_bits - 1
+             * q.all = n.all << (n_udword_bits - sr);
+             * r.all = n.all >> sr;
+             */
+            if (sr == n_uword_bits) {
+                q[low] = 0;
+                q[high] = n[low];
+                r[high] = 0;
+                r[low] = n[high];
+            } else if (sr < n_uword_bits) {
+                // 2 <= sr <= n_uword_bits - 1
+                q[low] = 0;
+                q[high] = n[low] << (n_uword_bits - sr);
+                r[high] = n[high] >> sr;
+                r[low] = (n[high] << (n_uword_bits - sr)) | (n[low] >> sr);
+            } else {
+                // n_uword_bits + 1 <= sr <= n_udword_bits - 1
+                q[low] = n[low] << (n_udword_bits - sr);
+                q[high] = (n[high] << (n_udword_bits - sr)) |
+                    (n[low] >> (sr - n_uword_bits));
+                r[high] = 0;
+                r[low] = n[high] >> (sr - n_uword_bits);
+            }
+        } else {
+            /* K X
+             * ---
+             * K K
+             */
+            sr = @clz(su_int, d[high]) - @clz(su_int, n[high]);
+            /* 0 <= sr <= n_uword_bits - 1 or sr large */
+            if (sr > n_uword_bits - 1) {
+                if (var rem ?= maybe_rem) {
+                    *rem = *(&du_int)(&n[0]);
+                }
+                return 0;
+            }
+            sr += 1;
+            /* 1 <= sr <= n_uword_bits */
+            /*  q.all = n.all << (n_udword_bits - sr); */
+            q[low] = 0;
+            if (sr == n_uword_bits) {
+                q[high] = n[low];
+                r[high] = 0;
+                r[low] = n[high];
+            } else {
+                q[high] = n[low] << (n_uword_bits - sr);
+                r[high] = n[high] >> sr;
+                r[low] = (n[high] << (n_uword_bits - sr)) | (n[low] >> sr);
+            }
+        }
+    }
+    /* Not a special case
+     * q and r are initialized with:
+     * q.all = n.all << (n_udword_bits - sr);
+     * r.all = n.all >> sr;
+     * 1 <= sr <= n_udword_bits - 1
+     */
+    var carry: su_int = 0;
+    while (sr > 0) {
+        /* r:q = ((r:q)  << 1) | carry */
+        r[high] = (r[high] << 1) | (r[low]  >> (n_uword_bits - 1));
+        r[low]  = (r[low]  << 1) | (q[high] >> (n_uword_bits - 1));
+        q[high] = (q[high] << 1) | (q[low]  >> (n_uword_bits - 1));
+        q[low]  = (q[low]  << 1) | carry;
+        /* carry = 0;
+            * if (r.all >= d.all)
+            * {
+            *      r.all -= d.all;
+            *      carry = 1;
+            * }
+            */
+        const s: di_int = (di_int)(*(&du_int)(&d[0]) - *(&du_int)(&r[0]) - 1) >> (n_udword_bits - 1);
+        carry = su_int(s & 1);
+        *(&du_int)(&r[0]) -= *(&du_int)(&d[0]) & c_ulonglong(s);
+
+        sr -= 1;
+    }
+    *(&du_int)(&q[0]) = (*(&du_int)(&q[0]) << 1) | carry;
+    if (var rem ?= maybe_rem) {
+        *rem = *(&du_int)(&r[0]);
+    }
+    return *(&du_int)(&q[0]);
+}
+
+export fn __umoddi3(a: du_int, b: du_int) -> du_int {
+    var r: du_int = undefined;
+    __udivmoddi4(a, b, &r);
+    return r;
+}
+
+
+#attribute("test")
+fn test_umoddi3() {
+    test_one_umoddi3(0, 1, 0);
+    test_one_umoddi3(2, 1, 0);
+    test_one_umoddi3(0x8000000000000000, 1, 0x0);
+    test_one_umoddi3(0x8000000000000000, 2, 0x0);
+    test_one_umoddi3(0xFFFFFFFFFFFFFFFF, 2, 0x1);
+}
+
+fn test_one_umoddi3(a: du_int, b: du_int, expected_r: du_int) {
+    const r = __umoddi3(a, b);
+    assert(r == expected_r);
+}
+
+#attribute("test")
+fn test_udivmoddi4() {
+    const cases = [][4]du_int {
+        []du_int{0x0000000000000000, 0x0000000000000001, 0x0000000000000000, 0x0000000000000000},
+        []du_int{0x0000000080000000, 0x0000000100000001, 0x0000000000000000, 0x0000000080000000},
+        []du_int{0x7FFFFFFF00000001, 0x0000000000000001, 0x7FFFFFFF00000001, 0x0000000000000000},
+        []du_int{0x7FFFFFFF7FFFFFFF, 0xFFFFFFFFFFFFFFFF, 0x0000000000000000, 0x7FFFFFFF7FFFFFFF},
+        []du_int{0x8000000000000002, 0xFFFFFFFFFFFFFFFE, 0x0000000000000000, 0x8000000000000002},
+        []du_int{0x80000000FFFFFFFD, 0xFFFFFFFFFFFFFFFD, 0x0000000000000000, 0x80000000FFFFFFFD},
+        []du_int{0xFFFFFFFD00000010, 0xFFFFFFFF80000000, 0x0000000000000000, 0xFFFFFFFD00000010},
+        []du_int{0xFFFFFFFDFFFFFFFF, 0xFFFFFFFF7FFFFFFF, 0x0000000000000000, 0xFFFFFFFDFFFFFFFF},
+        []du_int{0xFFFFFFFE0747AE14, 0xFFFFFFFF0747AE14, 0x0000000000000000, 0xFFFFFFFE0747AE14},
+        []du_int{0xFFFFFFFF00000001, 0xFFFFFFFF078644FA, 0x0000000000000000, 0xFFFFFFFF00000001},
+        []du_int{0xFFFFFFFF80000000, 0xFFFFFFFF00000010, 0x0000000000000001, 0x000000007FFFFFF0},
+        []du_int{0xFFFFFFFFFFFFFFFF, 0xFFFFFFFFFFFFFFFF, 0x0000000000000001, 0x0000000000000000},
+    };
+
+    for (cases) |case| {
+        test_one_udivmoddi4(case[0], case[1], case[2], case[3]);
+    }
+}
+
+fn test_one_udivmoddi4(a: du_int, b: du_int, expected_q: du_int, expected_r: du_int) {
+    var r: du_int = undefined;
+    const q = __udivmoddi4(a, b, &r);
+    assert(q == expected_q);
+    assert(r == expected_r);
+}
+
+fn assert(b: bool) {
+    if (!b) unreachable{};
+}
diff --git a/test/self_hosted.zig b/test/self_hosted.zig
index a35b2a5ff5..38338cfaf0 100644
--- a/test/self_hosted.zig
+++ b/test/self_hosted.zig
@@ -479,6 +479,20 @@ struct ArrayDotLenConstExpr {
 const some_array = []u8 {0, 1, 2, 3};
 
 
+#attribute("test")
+fn count_leading_zeroes() {
+    assert(@clz(u8, 0b00001010) == 4);
+    assert(@clz(u8, 0b10001010) == 0);
+    assert(@clz(u8, 0b00000000) == 8);
+}
+
+#attribute("test")
+fn count_trailing_zeroes() {
+    assert(@ctz(u8, 0b10100000) == 5);
+    assert(@ctz(u8, 0b10001010) == 1);
+    assert(@ctz(u8, 0b00000000) == 8);
+}
+
 
 
 fn assert(b: bool) {