diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index 8cd5289..abf1a68 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -1,5 +1,6 @@
 cmake_minimum_required(VERSION 3.0)
 
+include_directories(core)
 include_directories(common)
 include_directories(all_cases)
 
@@ -7,6 +8,7 @@ include_directories(raw_code)
 include_directories(short_code)
 include_directories(common_code)
 
+add_subdirectory(core)
 add_subdirectory(common)
 add_subdirectory(all_cases)
 
@@ -14,8 +16,11 @@ add_subdirectory(raw_code)
 add_subdirectory(short_code)
 add_subdirectory(common_code)
 
-#add_executable(klotski main.cc)
-add_executable(klotski main.cc core_demo.cc)
+add_executable(klotski main.cc)
+#add_executable(klotski main.cc core_demo.cc)
+
+target_link_libraries(klotski core)
+target_link_libraries(klotski common)
 target_link_libraries(klotski all_cases)
 
 target_link_libraries(klotski raw_code)
diff --git a/src/core/CMakeLists.txt b/src/core/CMakeLists.txt
new file mode 100644
index 0000000..fae268d
--- /dev/null
+++ b/src/core/CMakeLists.txt
@@ -0,0 +1,3 @@
+cmake_minimum_required(VERSION 3.0)
+
+add_library(core core.cc)
diff --git a/src/core/bfs.h b/src/core/bfs.h
new file mode 100644
index 0000000..d925f60
--- /dev/null
+++ b/src/core/bfs.h
@@ -0,0 +1,16 @@
+#pragma once
+
+#define BFS_INIT \
+int next_addr; \
+cache_size = 1; \
+int current = 0;  \
+cache[0].code = code; \
+cache[0].addr = addr; \
+cache[0].filter = 0;
+
+#define BFS_LOAD \
+code = cache[current].code; \
+addr = cache[current].addr; \
+int filter = cache[current++].filter;
+
+#define BFS_STOP (current == cache_size)
diff --git a/src/core/core.cc b/src/core/core.cc
new file mode 100644
index 0000000..35b4bd6
--- /dev/null
+++ b/src/core/core.cc
@@ -0,0 +1,162 @@
+#include "bfs.h"
+#include "core.h"
+#include "raw_code.h"
+
+#define ALLOW_UP    (filter != -UP)
+#define ALLOW_DOWN  (filter != -DOWN)
+#define ALLOW_LEFT  (filter != -LEFT)
+#define ALLOW_RIGHT (filter != -RIGHT)
+
+#define NOT_COLUMN_0 ((addr & 3) != 0)
+#define NOT_COLUMN_2 ((addr & 3) != 2)
+#define NOT_COLUMN_3 ((addr & 3) != 1)
+
+#define MOVE_UP    (next_addr = addr + UP)
+#define MOVE_DOWN  (next_addr = addr + DOWN)
+#define MOVE_LEFT  (next_addr = addr + LEFT)
+#define MOVE_RIGHT (next_addr = addr + RIGHT)
+
+#define TOP_LIMIT(ADDR)    (addr >= ADDR * 3)
+#define BOTTOM_LIMIT(ADDR) (addr <= ADDR * 3)
+
+#define RELEASE(NEXT_CODE, FILTER) { \
+    cache_t next_case = { \
+        .code = NEXT_CODE, \
+        .mask = F_1x1 << next_addr, \
+        .filter = FILTER, \
+        .addr = next_addr \
+    }; \
+    cache_insert(next_case); \
+}
+
+#define release_1x1(FILTER) RELEASE(NEXT_CODE_1x1, FILTER)
+#define release_1x2(FILTER) RELEASE(NEXT_CODE_1x2, FILTER)
+#define release_2x1(FILTER) RELEASE(NEXT_CODE_2x1, FILTER)
+#define release_2x2(FILTER) RELEASE(NEXT_CODE_2x2, FILTER)
+
+#define NEXT_CODE_1x1 (code & ~(F_1x1 << addr) | (C_1x1 << next_addr))
+#define NEXT_CODE_1x2 (code & ~(F_1x2 << addr) | (C_1x2 << next_addr))
+#define NEXT_CODE_2x1 (code & ~(F_2x1 << addr) | (C_2x1 << next_addr))
+#define NEXT_CODE_2x2 (code & ~(F_2x2 << addr) | (C_2x2 << next_addr))
+
+inline void Core::cache_insert(Core::cache_t &next_case) { // try to insert into cache
+    auto *p = cache;
+    for (; p < cache + cache_size; ++p) {
+        if (p->code == next_case.code) {
+            return; // already exist -> insert failed
+        }
+    }
+    *p = next_case;
+    ++cache_size;
+}
+
+void Core::move_1x1(uint64_t code, int addr) { // try to move target 1x1 block
+    BFS_INIT
+    while (!BFS_STOP) { // bfs search process
+        BFS_LOAD
+        if (ALLOW_UP && TOP_LIMIT(4) && !(code >> MOVE_UP & F_1x1)) {
+            release_1x1(UP) // 1x1 block move up
+        }
+        if (ALLOW_DOWN && BOTTOM_LIMIT(15) && !(code >> MOVE_DOWN & F_1x1)) {
+            release_1x1(DOWN) // 1x1 block move down
+        }
+        if (ALLOW_LEFT && NOT_COLUMN_0 && !(code >> MOVE_LEFT & F_1x1)) {
+            release_1x1(LEFT) // 1x1 block move left
+        }
+        if (ALLOW_RIGHT && NOT_COLUMN_3 && !(code >> MOVE_RIGHT & F_1x1)) {
+            release_1x1(RIGHT) // 1x1 block move right
+        }
+    }
+}
+
+void Core::move_1x2(uint64_t code, int addr) { // try to move target 1x2 block
+    BFS_INIT
+    while (!BFS_STOP) { // bfs search process
+        BFS_LOAD
+        if (ALLOW_UP && TOP_LIMIT(4) && !(code >> MOVE_UP & F_1x2)) {
+            release_1x2(UP) // 1x2 block move up
+        }
+        if (ALLOW_DOWN && BOTTOM_LIMIT(14) && !(code >> MOVE_DOWN & F_1x2)) {
+            release_1x2(DOWN) // 1x2 block move down
+        }
+        if (ALLOW_LEFT && NOT_COLUMN_0 && !(code >> MOVE_LEFT & F_1x1)) {
+            release_1x2(LEFT) // 1x2 block move left
+        }
+        if (ALLOW_RIGHT && NOT_COLUMN_2 && !(code >> MOVE_RIGHT & F_1x1_R)) {
+            release_1x2(RIGHT) // 1x2 block move right
+        }
+    }
+}
+
+void Core::move_2x1(uint64_t code, int addr) { // try to move target 2x1 block
+    BFS_INIT
+    while (!BFS_STOP) { // bfs search process
+        BFS_LOAD
+        if (ALLOW_UP && TOP_LIMIT(4) && !(code >> MOVE_UP & F_1x1)) {
+            release_2x1(UP) // 2x1 block move up
+        }
+        if (ALLOW_DOWN && BOTTOM_LIMIT(11) && !(code >> MOVE_DOWN & F_1x1_D)) {
+            release_2x1(DOWN) // 2x1 block move down
+        }
+        if (ALLOW_LEFT && NOT_COLUMN_0 && !(code >> MOVE_LEFT & F_2x1)) {
+            release_2x1(LEFT) // 2x1 block move left
+        }
+        if (ALLOW_RIGHT && NOT_COLUMN_3 && !(code >> MOVE_RIGHT & F_2x1)) {
+            release_2x1(RIGHT) // 2x1 block move right
+        }
+    }
+}
+
+void Core::move_2x2(uint64_t code, int addr) { // try to move target 2x2 block
+    BFS_INIT
+    while (!BFS_STOP) { // bfs search process
+        BFS_LOAD
+        if (ALLOW_UP && TOP_LIMIT(4) && !(code >> MOVE_UP & F_1x2)) {
+            release_2x2(UP) // 2x2 block move up
+        }
+        if (ALLOW_DOWN && BOTTOM_LIMIT(10) && !(code >> MOVE_DOWN & F_1x2_D)) {
+            release_2x2(DOWN) // 2x2 block move down
+        }
+        if (ALLOW_LEFT && NOT_COLUMN_0 && !(code >> MOVE_LEFT & F_2x1)) {
+            release_2x2(LEFT) // 2x2 block move left
+        }
+        if (ALLOW_RIGHT && NOT_COLUMN_2 && !(code >> MOVE_RIGHT & F_2x1_R)) {
+            release_2x2(RIGHT) // 2x2 block move right
+        }
+    }
+}
+
+#include <iostream>
+
+void Core::next_step(uint64_t raw_code) {
+
+    auto temp_code = raw_code;
+    for (int addr = 0; temp_code; addr += 3, temp_code >>= 3) {
+
+        switch (temp_code & 0b111) {
+            case B_1x1:
+                move_1x1(raw_code, addr);
+                break;
+            case B_1x2:
+                move_1x2(raw_code, addr);
+                break;
+            case B_2x1:
+                move_2x1(raw_code, addr);
+                break;
+            case B_2x2:
+                move_2x2(raw_code, addr);
+                break;
+            default:
+                continue; // B_space or B_fill
+        }
+        if (cache_size != 1) {
+            for (int i = 1; i < cache_size; ++i) {
+                std::cout << RawCode(cache[i].code).dump_case();
+                printf("MASK -> %016lX\n", cache[i].mask);
+            }
+//            std::cout << "found: " << cache_size - 1 << std::endl;
+        }
+    }
+}
+
+
diff --git a/src/core/core.h b/src/core/core.h
new file mode 100644
index 0000000..7341bdf
--- /dev/null
+++ b/src/core/core.h
@@ -0,0 +1,43 @@
+#pragma once
+
+#include <cstdint>
+
+#define UP    (-12) // -4 * 3-bits
+#define LEFT   (-3) // +1 * 3-bits
+#define DOWN  (+12) // +4 * 3-bits
+#define RIGHT  (+3) // +1 * 3-bits
+
+#define F_1x1 (uint64_t)0x7 // 111
+#define F_1x2 (uint64_t)0x3F // 111 111
+#define F_2x1 (uint64_t)0x7007 // 111 000 000 000 111
+#define F_2x2 (uint64_t)0x3F03F // 111 111 000 000 111 111
+
+#define F_1x1_R (uint64_t)0x38 // 111 000
+#define F_1x1_D (uint64_t)0x7000 // 111 000 000 000 000
+#define F_2x1_R (uint64_t)0x38038 // 111 000 000 000 111 000
+#define F_1x2_D (uint64_t)0x3F000 // 111 111 000 000 000 000
+
+class Core {
+public:
+    struct cache_t {
+        uint64_t code; // raw code
+        uint64_t mask; // only 000 or 111
+        int filter;
+        int addr;
+    };
+
+    int cache_size;
+    cache_t cache[16];
+
+    inline void cache_insert(Core::cache_t &next_case);
+
+    void move_1x1(uint64_t code, int addr);
+    void move_1x2(uint64_t code, int addr);
+    void move_2x1(uint64_t code, int addr);
+    void move_2x2(uint64_t code, int addr);
+
+    void next_step(uint64_t raw_code);
+
+};
+
+
diff --git a/src/core_demo.cc b/src/core_demo.cc
index 442b546..673fe6e 100644
--- a/src/core_demo.cc
+++ b/src/core_demo.cc
@@ -191,8 +191,8 @@ void next_step(uint64_t raw_code) {
         }
         if (cache_size != 1) {
             for (int i = 1; i < cache_size; ++i) {
-                std::cout << RawCode(cache[i].code).dump_case();
-                printf("MASK -> %016lX\n", cache[i].mask);
+//                std::cout << RawCode(cache[i].code).dump_case();
+//                printf("MASK -> %016lX\n", cache[i].mask);
             }
 //            std::cout << "found: " << cache_size - 1 << std::endl;
         }
diff --git a/src/main.cc b/src/main.cc
index c32997e..5c47f61 100644
--- a/src/main.cc
+++ b/src/main.cc
@@ -1,11 +1,12 @@
 #include <iostream>
+#include "core.h"
 #include "all_cases.h"
 #include "basic_ranges.h"
 #include "common_code.h"
 #include "short_code.h"
 #include "raw_code.h"
 
-#include "core_demo.h"
+//#include "core_demo.h"
 
 #include <thread>
 
@@ -145,8 +146,9 @@ int main() {
 //    next_step(CommonCode("4FEA134").to_raw_code().unwrap(), 0); // mask unset
     auto raw_code = RawCode(CommonCode("4fea134")).unwrap();
 
-//    for (int i = 0; i < 1000000000; ++i) {
-        next_step(raw_code);
+    auto c = Core();
+//    for (int i = 0; i < 100000000; ++i) {
+        c.next_step(raw_code);
 //    }
 
     return 0;
diff --git a/src/raw_code/raw_code.h b/src/raw_code/raw_code.h
index 7272c31..eec1f47 100644
--- a/src/raw_code/raw_code.h
+++ b/src/raw_code/raw_code.h
@@ -15,16 +15,6 @@
 #define C_2x1 (uint64_t)0x7002 // 111 000 000 000 010
 #define C_2x2 (uint64_t)0x3F03C // 111 111 000 000 111 100
 
-#define F_1x1 (uint64_t)0x7 // 111
-#define F_1x2 (uint64_t)0x3F // 111 111
-#define F_2x1 (uint64_t)0x7007 // 111 000 000 000 111
-#define F_2x2 (uint64_t)0x3F03F // 111 111 000 000 111 111
-
-#define F_1x1_R (uint64_t)0x38 // 111 000
-#define F_1x1_D (uint64_t)0x7000 // 111 000 000 000 000
-#define F_2x1_R (uint64_t)0x38038 // 111 000 000 000 111 000
-#define F_1x2_D (uint64_t)0x3F000 // 111 111 000 000 000 000
-
 class CommonCode;
 
 class RawCode {