diff --git a/emitter.c b/emitter.c
index 7bcae50..fba0d76 100644
--- a/emitter.c
+++ b/emitter.c
@@ -11,6 +11,7 @@ Emitter emitter_create(uint8_t* code_address)
         .pos = 0,
         .loop_counter = 0,
         .cmp_flags_set = false,
+        .rax_contains_copy = false,
     };
 }
 
@@ -52,6 +53,9 @@ void emitter_push_u64(Emitter* emitter, uint64_t value)
     emitter->pos += 1;
 }
 
+inline bool is_8(int value) { return value >= -128 && value <= 127; }
+inline bool is_16(int value) { return value >= -32768 && value <= 32767; }
+
 void emitter_emit_expr(Emitter* emitter, Expr* expr)
 {
     emitter->cmp_flags_set = false;
@@ -61,32 +65,48 @@ void emitter_emit_expr(Emitter* emitter, Expr* expr)
             exit(1);
             break;
         case ExprType_Incr:
-            // add BYTE [rbx], 1
+            // add BYTE [rbx], <value, rel8>
             emitter_push_u8(emitter, 0x80);
             emitter_push_u8(emitter, 0x03);
-            emitter_push_u8(emitter, expr->value);
+            emitter_push_u8(emitter, (uint8_t)expr->value);
             emitter->cmp_flags_set = true;
             break;
         case ExprType_Decr:
-            // sub BYTE [rbx], 1
+            // sub BYTE [rbx], <value: rel8>
             emitter_push_u8(emitter, 0x80);
             emitter_push_u8(emitter, 0x2b);
-            emitter_push_u8(emitter, expr->value);
+            emitter_push_u8(emitter, (uint8_t)expr->value);
             emitter->cmp_flags_set = true;
             break;
         case ExprType_Left:
-            // sub rbx, 1
-            emitter_push_u8(emitter, 0x48);
-            emitter_push_u8(emitter, 0x83);
-            emitter_push_u8(emitter, 0xeb);
-            emitter_push_u8(emitter, expr->value);
+            if (is_8(expr->value)) {
+                // sub rbx, <value: rel8>
+                emitter_push_u8(emitter, 0x48);
+                emitter_push_u8(emitter, 0x83);
+                emitter_push_u8(emitter, 0xeb);
+                emitter_push_u8(emitter, (uint8_t)expr->value);
+            } else {
+                // sub rbx, <value: rel32>
+                emitter_push_u8(emitter, 0x48);
+                emitter_push_u8(emitter, 0x81);
+                emitter_push_u8(emitter, 0xeb);
+                emitter_push_u32(emitter, expr->value);
+            }
             break;
         case ExprType_Right:
-            // add rbx, 1
-            emitter_push_u8(emitter, 0x48);
-            emitter_push_u8(emitter, 0x83);
-            emitter_push_u8(emitter, 0xc3);
-            emitter_push_u8(emitter, expr->value);
+            if (is_8(expr->value)) {
+                // add rbx, <value: rel8>
+                emitter_push_u8(emitter, 0x48);
+                emitter_push_u8(emitter, 0x83);
+                emitter_push_u8(emitter, 0xc3);
+                emitter_push_u8(emitter, (uint8_t)expr->value);
+            } else {
+                // add rbx, <value: rel32>
+                emitter_push_u8(emitter, 0x48);
+                emitter_push_u8(emitter, 0x81);
+                emitter_push_u8(emitter, 0xc3);
+                emitter_push_u32(emitter, expr->value);
+            }
             break;
         case ExprType_Output:
             // movzx edi, BYTE [rbx]
@@ -124,6 +144,30 @@ void emitter_emit_expr(Emitter* emitter, Expr* expr)
             emitter_push_u8(emitter, 0x00);
             emitter->cmp_flags_set = true;
             break;
+        case ExprType_Add:
+            if (!emitter->rax_contains_copy) {
+                // movzx rax, BYTE [rbx]
+                emitter_push_u8(emitter, 0x48);
+                emitter_push_u8(emitter, 0x0f);
+                emitter_push_u8(emitter, 0xb6);
+                emitter_push_u8(emitter, 0x03);
+                emitter->rax_contains_copy = true;
+            }
+            if (is_8(expr->value)) {
+                // add BYTE [rbx + <value: rel8>], al
+                emitter_push_u8(emitter, 0x00);
+                emitter_push_u8(emitter, 0x43);
+                emitter_push_u8(emitter, (uint8_t)expr->value);
+            } else {
+                // add BYTE [rbx + <value: rel32>], al
+                emitter_push_u8(emitter, 0x00);
+                emitter_push_u8(emitter, 0x83);
+                emitter_push_u32(emitter, expr->value);
+            }
+            break;
+    }
+    if (expr->type != ExprType_Add) {
+        emitter->rax_contains_copy = false;
     }
 }
 
diff --git a/emitter.h b/emitter.h
index 72111dd..c3fd6bf 100644
--- a/emitter.h
+++ b/emitter.h
@@ -11,6 +11,7 @@ typedef struct {
     size_t pos;
     int loop_counter;
     bool cmp_flags_set;
+    bool rax_contains_copy;
 } Emitter;
 
 Emitter emitter_create(uint8_t* code_address);
diff --git a/expr.c b/expr.c
index 050d2d7..2e06f04 100644
--- a/expr.c
+++ b/expr.c
@@ -40,6 +40,24 @@ Expr expr_vec_pop(ExprVec* vec)
     return vec->data[vec->length];
 }
 
+void expr_vec_stringify(ExprVec* vec, char* acc, int depth)
+{
+    strcat(acc, color_bold);
+    strcat(acc, expr_bracket_color(depth));
+    strcat(acc, "[");
+    strcat(acc, color_reset);
+    for (size_t i = 0; i < vec->length; ++i) {
+        if (i != 0) {
+            strcat(acc, " ");
+        }
+        expr_stringify(&vec->data[i], acc, depth + 1);
+    }
+    strcat(acc, color_bold);
+    strcat(acc, expr_bracket_color(depth));
+    strcat(acc, "]");
+    strcat(acc, color_reset);
+}
+
 bool expr_vec_equal(const ExprVec* self, const ExprVec* other)
 {
     if (self->length != other->length) {
@@ -53,6 +71,16 @@ bool expr_vec_equal(const ExprVec* self, const ExprVec* other)
     return true;
 }
 
+ExprVec expr_vec_clone(const ExprVec* original)
+{
+    ExprVec vec;
+    expr_vec_construct(&vec);
+    for (size_t i = 0; i < original->length; ++i) {
+        expr_vec_push(&vec, expr_clone(&original->data[i]));
+    }
+    return vec;
+}
+
 void expr_free(Expr* expr)
 {
     switch (expr->type) {
@@ -92,24 +120,6 @@ void expr_stringify_concat_value(Expr* expr, char* acc, int depth)
     strcat(acc, color_reset);
 }
 
-void expr_vec_stringify(ExprVec* vec, char* acc, int depth)
-{
-    strcat(acc, color_bold);
-    strcat(acc, expr_bracket_color(depth));
-    strcat(acc, "[");
-    strcat(acc, color_reset);
-    for (size_t i = 0; i < vec->length; ++i) {
-        if (i != 0) {
-            strcat(acc, " ");
-        }
-        expr_stringify(&vec->data[i], acc, depth + 1);
-    }
-    strcat(acc, color_bold);
-    strcat(acc, expr_bracket_color(depth));
-    strcat(acc, "]");
-    strcat(acc, color_reset);
-}
-
 void expr_stringify(Expr* expr, char* acc, int depth)
 {
     switch (expr->type) {
@@ -163,6 +173,12 @@ void expr_stringify(Expr* expr, char* acc, int depth)
             strcat(acc, "Zero");
             strcat(acc, color_reset);
             break;
+        case ExprType_Add:
+            strcat(acc, color_cyan);
+            strcat(acc, "Add");
+            strcat(acc, color_reset);
+            expr_stringify_concat_value(expr, acc, depth);
+            break;
     }
 }
 
@@ -190,3 +206,15 @@ bool expr_equal(const Expr* self, const Expr* other)
     }
     return true;
 }
+
+Expr expr_clone(const Expr* expr)
+{
+    if (expr->type == ExprType_Loop) {
+        return (Expr) {
+            .type = ExprType_Loop,
+            .exprs = expr_vec_clone(&expr->exprs),
+        };
+    } else {
+        return *expr;
+    }
+}
diff --git a/expr.h b/expr.h
index 8b6c2ad..54536f5 100644
--- a/expr.h
+++ b/expr.h
@@ -14,6 +14,7 @@ typedef enum {
     ExprType_Input,
     ExprType_Loop,
     ExprType_Zero,
+    ExprType_Add,
 } ExprType;
 
 typedef struct Expr Expr;
@@ -29,7 +30,9 @@ void expr_vec_destroy(ExprVec* vec);
 void expr_vec_free(ExprVec* vec);
 void expr_vec_push(ExprVec* vec, Expr expr);
 Expr expr_vec_pop(ExprVec* vec);
+void expr_vec_stringify(ExprVec* vec, char* acc, int depth);
 bool expr_vec_equal(const ExprVec* self, const ExprVec* other);
+ExprVec expr_vec_clone(const ExprVec* original);
 
 struct Expr {
     ExprType type;
@@ -42,8 +45,9 @@ struct Expr {
 void expr_free(Expr* expr);
 const char* expr_bracket_color(int depth);
 void expr_stringify_concat_value(Expr* expr, char* acc, int depth);
-void expr_vec_stringify(ExprVec* vec, char* acc, int depth);
+void expr_stringify_concat_pair(Expr* expr, char* acc, int depth);
 void expr_stringify(Expr* expr, char* acc, int depth);
 bool expr_equal(const Expr* self, const Expr* other);
+Expr expr_clone(const Expr* expr);
 
 #endif
diff --git a/main.c b/main.c
index aac86a8..ac033a1 100644
--- a/main.c
+++ b/main.c
@@ -10,13 +10,27 @@
 #include <string.h>
 #include <sys/mman.h>
 
+#define ADD_OPTIMIZATION_WO_FREE_AST(NAME)                                     \
+    previous_ast = ast;                                                        \
+    ast = optimize_##NAME(&ast);                                               \
+    printf("%s" #NAME ":%s\n", color_bold, color_reset);                       \
+    if (!expr_vec_equal(&ast, &previous_ast)) {                                \
+        ast_string[0] = '\0';                                                  \
+        expr_vec_stringify(&ast, ast_string, 0);                               \
+        puts(ast_string);                                                      \
+    }
+
+#define ADD_OPTIMIZATION(NAME)                                                 \
+    expr_vec_free(&previous_ast);                                              \
+    ADD_OPTIMIZATION_WO_FREE_AST(NAME)
+
 int main(int argc, char** argv)
 {
-    const char* text = "++++++++++[>+<-]";
-    printf("\ntext:%s\n\"%s\"%s\n", color_bright_green, text, color_reset);
-    Parser parser = parser_create(lexer_from_string(text, strlen(text)));
+    // const char* text = "++++++++++[>+<-]";
+    // printf("\ntext:%s\n\"%s\"%s\n", color_bright_green, text, color_reset);
+    // Parser parser = parser_create(lexer_from_string(text, strlen(text)));
 
-    // Parser parser = parser_create(lexer_from_args_or_stdin(argc, argv));
+    Parser parser = parser_create(lexer_from_args_or_stdin(argc, argv));
 
     char* ast_string = malloc(sizeof(char) * 33768);
     ast_string[0] = '\0';
@@ -42,44 +56,11 @@ int main(int argc, char** argv)
         if (!first) {
             expr_vec_free(&previous_ast);
         }
-        previous_ast = ast;
-        ast = optimize_fold_adjecent(&ast);
-        printf("%sfold_adjecent:%s\n", color_bold, color_reset);
-        if (!expr_vec_equal(&ast, &previous_ast)) {
-            ast_string[0] = '\0';
-            expr_vec_stringify(&ast, ast_string, 0);
-            puts(ast_string);
-        }
-
-        expr_vec_free(&previous_ast);
-        previous_ast = ast;
-        ast = optimize_eliminate_negation(&ast);
-        printf("%seliminate_negation:%s\n", color_bold, color_reset);
-        if (!expr_vec_equal(&ast, &previous_ast)) {
-            ast_string[0] = '\0';
-            expr_vec_stringify(&ast, ast_string, 0);
-            puts(ast_string);
-        }
-
-        expr_vec_free(&previous_ast);
-        previous_ast = ast;
-        ast = optimize_eliminate_overflow(&ast);
-        printf("%seliminate_overflow:%s\n", color_bold, color_reset);
-        if (!expr_vec_equal(&ast, &previous_ast)) {
-            ast_string[0] = '\0';
-            expr_vec_stringify(&ast, ast_string, 0);
-            puts(ast_string);
-        }
-
-        expr_vec_free(&previous_ast);
-        previous_ast = ast;
-        ast = optimize_replace_zeroing_loops(&ast);
-        printf("%sreplace_zeroing_loops:%s\n", color_bold, color_reset);
-        if (!expr_vec_equal(&ast, &previous_ast)) {
-            ast_string[0] = '\0';
-            expr_vec_stringify(&ast, ast_string, 0);
-            puts(ast_string);
-        }
+        ADD_OPTIMIZATION_WO_FREE_AST(fold_adjecent);
+        ADD_OPTIMIZATION(eliminate_negation);
+        ADD_OPTIMIZATION(eliminate_overflow);
+        ADD_OPTIMIZATION(replace_zeroing_loops);
+        ADD_OPTIMIZATION(replace_copying_loops);
 
         if (first) {
             first = false;
@@ -112,12 +93,15 @@ int main(int argc, char** argv)
 
     printf("\n%scode:%s\n", color_bold, color_reset);
 
-    for (size_t y = 0; y < 8; ++y) {
-        for (size_t x = 0; x < 8; ++x) {
+    for (size_t y = 0; y < 40; ++y) {
+        for (size_t x = 0; x < 16; ++x) {
             uint8_t v = ((uint8_t*)code)[y * 8 + x];
             if (v == 0) {
                 fputs(color_gray, stdout);
             }
+            if (x == 8) {
+                fputc(' ', stdout);
+            }
             printf("%02x ", v);
             fputs(color_reset, stdout);
         }
@@ -135,7 +119,7 @@ int main(int argc, char** argv)
 
     printf("\n%smemory:%s\n", color_bold, color_reset);
 
-    for (size_t y = 0; y < 8; ++y) {
+    for (size_t y = 0; y < 4; ++y) {
         for (size_t x = 0; x < 8; ++x) {
             uint8_t v = memory[y * 8 + x];
             if (v == 0) {
diff --git a/optimizer.c b/optimizer.c
index d8e9804..3eb584e 100644
--- a/optimizer.c
+++ b/optimizer.c
@@ -4,9 +4,9 @@
 /*
  *  fold adjecent
  *
- *  A(N) :: { Incr(N) | Decr(N) | Right(N) | Left(N) }
+ *  A(n) :: { Incr(n) | Decr(n) | Right(n) | Left(n) }
  *
- *  [A(N1) A(N2)] -> [A(N1 + N2)]
+ *  [A(n1) A(n2)] -> [A(n1 + n2)]
  *
  */
 
@@ -49,20 +49,20 @@ Expr expr_optimize_fold_adjecent(const Expr* expr)
             .exprs = optimize_fold_adjecent(&expr->exprs),
         };
     } else {
-        return *expr;
+        return expr_clone(expr);
     }
 }
 
 /*
  *  eliminate negation
  *
- *  A(N), B(N) :: { Incr(N) | Decr(N) | Right(N) | Left(N) }
+ *  A(n), B(n) :: { Incr(n) | Decr(n) | Right(n) | Left(n) }
  *
- *  [A(N) B(N)] = []
+ *  [A(n) B(n)] = []
  *
- *  [A(N1) B(N2)] ? N1 == N2 -> []
- *  [A(N1) B(N2)] ? N1 < N2 -> [B(N2 - N1)]
- *  [A(N1) B(N2)] ? N1 > N2 -> [A(N1 - N2)]
+ *  [A(n1) B(n2)] ? n1 == n2 -> []
+ *  [A(n1) B(n2)] ? n1 < n2 -> [B(n2 - n1)]
+ *  [A(n1) B(n2)] ? n1 > n2 -> [A(n1 - n2)]
  *
  */
 
@@ -140,18 +140,18 @@ Expr expr_optimize_eliminate_negation(const Expr* expr)
             .exprs = optimize_eliminate_negation(&expr->exprs),
         };
     } else {
-        return *expr;
+        return expr_clone(expr);
     }
 }
 
 /*
  *  eliminate overflow
  *
- *  A(N) :: { Incr(N) | Decr(N) | Right(N) | Left(N) }
+ *  A(n) :: { Incr(n) | Decr(n) | Right(n) | Left(n) }
  *
- *  N > 255
+ *  n > 255
  *
- *  A(N) -> A(N % 256)
+ *  A(n) -> A(n % 256)
  *
  */
 
@@ -175,18 +175,18 @@ Expr expr_optimize_eliminate_overflow(const Expr* expr)
     } else if (expr->value > 255) {
         return (Expr) { .type = expr->type, .value = expr->value % 256 };
     } else {
-        return *expr;
+        return expr_clone(expr);
     }
 }
 
 /*
  *  replace zeroing loops
  *
- *  A(N) :: { Incr(N) | Decr(N) }
+ *  A(n) :: { Incr(n) | Decr(n) }
  *
- *  N % 2 == 1
+ *  n % 2 == 1
  *
- *  Loop[A(N)] -> Zero
+ *  Loop[A(n)] -> Zero
  *
  */
 
@@ -217,6 +217,67 @@ Expr expr_optimize_replace_zeroing_loops(const Expr* expr)
             };
         }
     } else {
-        return *expr;
+        return expr_clone(expr);
     }
 }
+
+/*
+ *  replace copying loops
+ *
+ *  O(n), I(n) :: { Left(n) | Right(n) }
+ *
+ *  O != I
+ *
+ *  [Loop[O(n) Incr(1) I(n) Decr(1)]] -> [Copy(n) Zero]
+ *
+ */
+
+ExprVec optimize_replace_copying_loops(const ExprVec* original)
+{
+    ExprVec result;
+    expr_vec_construct(&result);
+    for (size_t i = 0; i < original->length; ++i) {
+        const Expr* expr = &original->data[i];
+        if (expr->type == ExprType_Loop) {
+            const ExprVec* loop = &expr->exprs;
+            if (loop->length == 4
+                && ((loop->data[0].type == ExprType_Left
+                     && loop->data[2].type == ExprType_Right)
+                    || (loop->data[0].type == ExprType_Right
+                        && loop->data[2].type == ExprType_Left))
+                && loop->data[1].type == ExprType_Incr
+                && loop->data[3].type == ExprType_Decr
+                && loop->data[1].value == loop->data[3].value) {
+                if (loop->data[0].type == ExprType_Right) {
+                    expr_vec_push(
+                        &result,
+                        (Expr) {
+                            .type = ExprType_Add,
+                            .value = loop->data[0].value,
+                        }
+                    );
+                } else {
+                    expr_vec_push(
+                        &result,
+                        (Expr) {
+                            .type = ExprType_Add,
+                            .value = -loop->data[0].value,
+                        }
+                    );
+                }
+                expr_vec_push(&result, (Expr) { .type = ExprType_Zero });
+            } else {
+                expr_vec_push(
+                    &result,
+                    (Expr) {
+                        .type = ExprType_Loop,
+                        .exprs = optimize_replace_copying_loops(&expr->exprs),
+                    }
+                );
+            }
+        } else {
+            expr_vec_push(&result, expr_clone(expr));
+        }
+    }
+    return result;
+}
diff --git a/optimizer.h b/optimizer.h
index 6ed7c65..f72ad66 100644
--- a/optimizer.h
+++ b/optimizer.h
@@ -15,4 +15,6 @@ Expr expr_optimize_eliminate_overflow(const Expr* expr);
 ExprVec optimize_replace_zeroing_loops(const ExprVec* vec);
 Expr expr_optimize_replace_zeroing_loops(const Expr* expr);
 
+ExprVec optimize_replace_copying_loops(const ExprVec* vec);
+
 #endif