diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index abf1a68..a3509c1 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -16,8 +16,8 @@ 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 fast_cal.cc)
 
 target_link_libraries(klotski core)
 target_link_libraries(klotski common)
diff --git a/src/fast_cal.cc b/src/fast_cal.cc
new file mode 100644
index 0000000..8a15483
--- /dev/null
+++ b/src/fast_cal.cc
@@ -0,0 +1,77 @@
+#include <iostream>
+#include <unordered_map>
+#include <queue>
+#include "core.h"
+#include "fast_cal.h"
+#include "raw_code.h"
+
+struct fast_cal_t {
+    uint64_t code;
+    uint64_t mask;
+//    uint32_t step;
+};
+
+std::unordered_map<uint64_t, fast_cal_t> cal_data;
+std::queue<fast_cal_t*> cal_temp;
+
+void add_new_case(uint64_t code, uint64_t mask) {
+
+    auto exist_case = cal_data.find(code);
+    if (exist_case != cal_data.end()) { // find it
+        // mask update
+//        std::cout << "match exist case" << std::endl;
+        return;
+    }
+
+//    std::cout << RawCode(code).dump_case();
+//    std::cout << "~~~~~~~" << std::endl;
+//    for (int n = 0; n < 20; ++n, mask >>= 3) {
+//        std::cout << "+."[!(mask & 0b111)] << " \n"[!(~n & 0b11)];
+//    }
+//    std::cout << std::endl;
+
+    auto new_case = fast_cal_t {
+        .code = code,
+        .mask = mask,
+    };
+
+    cal_data[code] = new_case;
+
+    // TODO: avoid redundancy map search
+    cal_temp.emplace(&cal_data[code]);
+
+}
+
+void fast_cal(uint64_t start_raw_code) {
+    std::cout << RawCode(start_raw_code).dump_case() << std::endl;
+
+    auto core = Core();
+
+    auto setup = fast_cal_t {
+        .code = start_raw_code,
+        .mask = 0,
+    };
+
+    cal_data[start_raw_code] = setup;
+    cal_temp.emplace(&cal_data[start_raw_code]);
+
+//    core.next_step(cal_temp.front()->code, add_new_case);
+//    cal_temp.pop();
+
+//    std::cout << "start cal second case" << std::endl;
+//    std::cout << RawCode(cal_temp.front()->code).dump_case() << std::endl << std::endl;
+
+//    core.next_step(cal_temp.front()->code, add_new_case);
+//    cal_temp.pop();
+
+
+    while (!cal_temp.empty()) {
+        core.next_step(cal_temp.front()->code, add_new_case);
+        cal_temp.pop();
+    }
+
+    std::cout << "size = " << cal_data.size() << std::endl;
+    std::cout << "queue size = " << cal_temp.size() << std::endl;
+
+
+}
diff --git a/src/fast_cal.h b/src/fast_cal.h
new file mode 100644
index 0000000..e68670a
--- /dev/null
+++ b/src/fast_cal.h
@@ -0,0 +1,5 @@
+#pragma once
+
+#include <cstdint>
+
+void fast_cal(uint64_t start_raw_code);
diff --git a/src/main.cc b/src/main.cc
index 4551eeb..5498e2b 100644
--- a/src/main.cc
+++ b/src/main.cc
@@ -5,6 +5,7 @@
 #include "common_code.h"
 #include "short_code.h"
 #include "raw_code.h"
+#include "fast_cal.h"
 
 //#include "core_demo.h"
 
@@ -37,14 +38,14 @@
 //    }
 //}
 
-void release(uint64_t code, uint64_t mask) {
-    std::cout << RawCode(code).dump_case();
-    std::cout << "~~~~~~~" << std::endl;
-    for (int n = 0; n < 20; ++n, mask >>= 3) {
-        std::cout << "+."[!(mask & 0b111)] << " \n"[!(~n & 0b11)];
-    }
-    std::cout << std::endl;
-}
+//void release(uint64_t code, uint64_t mask) {
+//    std::cout << RawCode(code).dump_case();
+//    std::cout << "~~~~~~~" << std::endl;
+//    for (int n = 0; n < 20; ++n, mask >>= 3) {
+//        std::cout << "+."[!(mask & 0b111)] << " \n"[!(~n & 0b11)];
+//    }
+//    std::cout << std::endl;
+//}
 
 int main() {
 
@@ -153,11 +154,11 @@ int main() {
 //    std::cout << CommonCode(RawCode(0x0E58FC85FFEBC4DB)).to_string() << std::endl;
 
 
-    auto c = Core();
+//    auto c = Core();
 
 //    auto raw_code = RawCode(CommonCode("4fea134")).unwrap();
-    auto raw_code = RawCode(CommonCode("1A9bf0c")).unwrap();
-    c.next_step(raw_code, release);
+//    auto raw_code = RawCode(CommonCode("1A9bf0c")).unwrap();
+//    c.next_step(raw_code, release);
 
 //    std::vector<uint64_t> all_cases_raw;
 //    for (int head = 0; head < 16; ++head) {
@@ -195,5 +196,7 @@ int main() {
 //    std::cout << (clock() - start_time) * 1000 / CLOCKS_PER_SEC << "ms" << std::endl;
 //    std::cout << "complete benchmark" << std::endl;
 
+    fast_cal(CommonCode("1a9bf0c").to_raw_code().unwrap());
+
     return 0;
 }