diff --git a/src/core/CMakeLists.txt b/src/core/CMakeLists.txt
index 7c60b63..443f683 100644
--- a/src/core/CMakeLists.txt
+++ b/src/core/CMakeLists.txt
@@ -18,6 +18,9 @@ set(KLOTSKI_CORE_SRC
         short_code/internal/serialize.cc
 
         core/internal/core.cc
+
+        fast_cal/internal/cal_core.cc
+        fast_cal/internal/fast_cal.cc
 )
 
 add_library(klotski_core STATIC ${KLOTSKI_CORE_SRC})
diff --git a/src/core/fast_cal/fast_cal.h b/src/core/fast_cal/fast_cal.h
new file mode 100644
index 0000000..e8384da
--- /dev/null
+++ b/src/core/fast_cal/fast_cal.h
@@ -0,0 +1,65 @@
+/// Klotski Engine by Dnomd343 @2024
+
+// TODO: only copy from old implementation, the interfaces will change in future.
+
+#pragma once
+
+#include <queue>
+#include <vector>
+#include <cstdint>
+#include <functional>
+#include <unordered_map>
+
+#include "core/core.h"
+#include "raw_code/raw_code.h"
+
+using klotski::core::Core;
+using klotski::codec::RawCode;
+
+// TODO: using prime number
+const uint32_t FC_MAP_RESERVE = 65536 * 8;
+
+/// FastCal not found -> return invalid raw code
+const RawCode FC_NOT_FOUND = RawCode::unsafe_create(0);
+
+class FastCal {
+public:
+    typedef std::function<bool(uint64_t)> match_t;
+
+    /// setting root code
+    void set_root(const RawCode &code);
+    explicit FastCal(const RawCode &code);
+
+    /// backtrack functions
+    int step_num(const RawCode &code);
+    std::vector<RawCode> backtrack(const RawCode &code);
+
+    /// BFS search functions
+    void build();
+    RawCode solve();
+    std::vector<RawCode> furthest();
+    std::vector<RawCode> solve_multi();
+    RawCode target(const match_t &match);
+    std::vector<RawCode> target_multi(const match_t &match);
+
+    /// static BFS search functions
+    static std::vector<RawCode> resolve(const RawCode &start);
+    static std::vector<std::vector<RawCode>> to_furthest(const RawCode &start);
+    static std::vector<std::vector<RawCode>> resolve_multi(const RawCode &start);
+    static std::vector<RawCode> search(const RawCode &start, const match_t &match);
+    static std::vector<std::vector<RawCode>> search_multi(const RawCode &start, const match_t &match);
+
+private:
+    struct fast_cal_t {
+        uint64_t code;
+        uint64_t mask;
+        fast_cal_t *last;
+    };
+
+    uint64_t root;
+    std::queue<fast_cal_t*> cache;
+    std::unordered_map<uint64_t, fast_cal_t> cases;
+
+    inline Core init(uint64_t code);
+    void new_case(uint64_t code, uint64_t mask);
+};
diff --git a/src/core/fast_cal/internal/cal_core.cc b/src/core/fast_cal/internal/cal_core.cc
new file mode 100644
index 0000000..6eed096
--- /dev/null
+++ b/src/core/fast_cal/internal/cal_core.cc
@@ -0,0 +1,105 @@
+#include "fast_cal/fast_cal.h"
+
+Core FastCal::init(uint64_t code) { // initialize process
+    /// reset working data
+    cases.clear();
+    cases.reserve(FC_MAP_RESERVE); // hashmap pre-reserve
+    std::queue<fast_cal_t*>{}.swap(cache);
+
+    /// insert root node
+    cache.emplace(&cases.emplace(code, fast_cal_t {
+        .code = code,
+        .mask = 0,
+        .last = nullptr, // without parent node
+    }).first->second);
+
+    /// import klotski core
+    return Core(
+        [this](auto &&code, auto &&mask) { // lambda as function pointer
+            new_case(std::forward<decltype(code)>(code), std::forward<decltype(mask)>(mask));
+        }
+    );
+}
+
+/// callback function for new case
+void FastCal::new_case(uint64_t code, uint64_t mask) {
+    auto current = cases.find(code);
+    if (current != cases.end()) { // find existed case
+        current->second.mask |= mask; // update mask info
+        return;
+    }
+    cache.emplace(&cases.emplace(code, fast_cal_t { // record new case
+        .code = code,
+        .mask = mask,
+        .last = cache.front(), // link parent case
+    }).first->second);
+}
+
+/// build total search tree
+void FastCal::build() {
+    auto core = init(root);
+    while (!cache.empty()) {
+        core.next_cases(cache.front()->code, cache.front()->mask);
+        cache.pop();
+    }
+}
+
+/// found first matched target
+RawCode FastCal::target(const match_t &match) {
+    auto core = init(root);
+    while (!cache.empty()) {
+        if (match(cache.front()->code)) {
+            return RawCode::unsafe_create(cache.front()->code); // match target
+        }
+        core.next_cases(cache.front()->code, cache.front()->mask);
+        cache.pop();
+    }
+    return FC_NOT_FOUND; // target not found
+}
+
+/// found all of the furthest cases
+std::vector<RawCode> FastCal::furthest() {
+    auto core = init(root);
+    auto layer_end = cache.back();
+    std::vector<RawCode> layer_cases;
+    /// start BFS search
+    while (!cache.empty()) {
+        core.next_cases(cache.front()->code, cache.front()->mask);
+        layer_cases.emplace_back(
+            RawCode::unsafe_create(cache.front()->code) // record layer cases
+        );
+        if (cache.front() == layer_end) { // reach layer ending
+            if (cache.size() == 1) {
+                break; // stop loop at last layer
+            }
+            layer_cases.clear();
+            layer_end = cache.back(); // reset layer ending
+        }
+        cache.pop();
+    }
+    return layer_cases; // release the latest layer cases
+}
+
+/// found multi-targets matched in first same layer
+std::vector<RawCode> FastCal::target_multi(const match_t &match) {
+    auto core = init(root);
+    auto layer_end = cache.back();
+    std::vector<RawCode> matched; // matched list
+    /// start BFS search
+    while (!cache.empty()) {
+        if (match(cache.front()->code)) { // match target
+            matched.emplace_back(
+                RawCode::unsafe_create(cache.front()->code) // record matched cases
+            );
+        }
+        core.next_cases(cache.front()->code, cache.front()->mask);
+        if (cache.front() == layer_end) { // reach layer ending
+            if (!matched.empty()) {
+                return matched; // stop at first matched layer
+            }
+            layer_end = cache.back(); // reset layer ending
+        }
+        cache.pop();
+    }
+    return std::vector<RawCode>{}; // no target found
+}
diff --git a/src/core/fast_cal/internal/fast_cal.cc b/src/core/fast_cal/internal/fast_cal.cc
new file mode 100644
index 0000000..85479e2
--- /dev/null
+++ b/src/core/fast_cal/internal/fast_cal.cc
@@ -0,0 +1,91 @@
+#include <algorithm>
+
+#include "utils/common.h"
+#include "fast_cal/fast_cal.h"
+#include "raw_code/raw_code.h"
+
+FastCal::FastCal(const RawCode &code) {
+    this->root = (uint64_t)code;
+}
+
+void FastCal::set_root(const RawCode &code) {
+    this->root = (uint64_t)code;
+}
+
+/// klotski resolved -> 2x2 block at address 13 (aka 0xD)
+auto resolved = [](uint64_t code) {
+    return ((code >> (3 * 0xD)) & 0b111) == BLOCK_2x2; // check 2x2 block address
+};
+
+RawCode FastCal::solve() {
+    return FastCal::target(resolved);
+}
+
+std::vector<RawCode> FastCal::solve_multi() {
+    return FastCal::target_multi(resolved);
+}
+
+std::vector<RawCode> FastCal::resolve(const RawCode &start) {
+    return FastCal::search(start, resolved);
+}
+
+std::vector<std::vector<RawCode>> FastCal::resolve_multi(const RawCode &start) {
+    return FastCal::search_multi(start, resolved);
+}
+
+/// backtrack of FastCal tree
+int FastCal::step_num(const RawCode &code) {
+    auto tmp = cases.find((uint64_t)code);
+    if (tmp == cases.end()) {
+        return -1; // code not exist
+    }
+    int num = 0; // step number
+    auto node = &tmp->second; // backtrack entry
+    while ((node = node->last) != nullptr) {
+        ++num;
+    }
+    return num;
+}
+
+std::vector<RawCode> FastCal::backtrack(const RawCode &code) {
+    auto tmp = cases.find((uint64_t)code);
+    if (tmp == cases.end()) {
+        return std::vector<RawCode>{}; // code not exist
+    }
+    auto node = &tmp->second; // backtrack entry
+    std::vector<RawCode> path; // backtrack path
+    while (node != nullptr) {
+        path.emplace_back(RawCode::unsafe_create(node->code)); // record path info
+        node = node->last;
+    }
+    std::reverse(path.begin(), path.end()); // reverse path cases
+    return path;
+}
+
+/// static BFS search functions
+std::vector<std::vector<RawCode>> FastCal::to_furthest(const RawCode &start) {
+    auto fc = FastCal(start);
+    std::vector<std::vector<RawCode>> result;
+    for (const auto &furthest : fc.furthest()) {
+        result.emplace_back(fc.backtrack(furthest)); // backtrack every furthest cases
+    }
+    return result;
+}
+
+std::vector<RawCode> FastCal::search(const RawCode &start, const match_t &match) {
+    auto fc = FastCal(start);
+    auto result = fc.target(match);
+    if (result == FC_NOT_FOUND) {
+        return std::vector<RawCode>{}; // target not matched
+    }
+    return fc.backtrack(result); // backtrack target path
+}
+
+std::vector<std::vector<RawCode>> FastCal::search_multi(const RawCode &start, const match_t &match) {
+    auto fc = FastCal(start);
+    std::vector<std::vector<RawCode>> result;
+    for (const auto &target : fc.target_multi(match)) {
+        result.emplace_back(fc.backtrack(target)); // backtrack every target
+    }
+    return result;
+}
diff --git a/src/core/main.cc b/src/core/main.cc
index 82706de..0ad642d 100644
--- a/src/core/main.cc
+++ b/src/core/main.cc
@@ -3,6 +3,7 @@
 
 #include "core/core.h"
 #include "raw_code/raw_code.h"
+#include "fast_cal/fast_cal.h"
 #include "all_cases/all_cases.h"
 #include "short_code/short_code.h"
 #include "common_code/common_code.h"
@@ -21,45 +22,28 @@ using klotski::codec::SHORT_CODE_LIMIT;
 int main() {
     const auto start = clock();
 
-    auto core = Core([](const uint64_t code, uint64_t) {
-        std::cout << RawCode::unsafe_create(code);
-        std::cout << std::endl;
-    });
+    // auto core = Core([](const uint64_t code, uint64_t) {
+    //     std::cout << RawCode::unsafe_create(code);
+    //     std::cout << std::endl;
+    // });
+    // core.next_cases(RawCode::from_common_code(0x1A9BF0C00).value().unwrap(), 0x0);
 
-    core.next_cases(RawCode::from_common_code(0x1A9BF0C00).value().unwrap(), 0x0);
+    // auto cal = FastCal(RawCode::from_common_code(0x1A9BF0C00).value());
+    auto cal = FastCal(RawCode::from_common_code("25EEF04").value());
 
-//    std::vector<uint64_t> common_codes;
-//    common_codes.reserve(klotski::cases::ALL_CASES_NUM_);
-
-//    for (uint64_t head = 0; head < 15; ++head) {
-//        for (auto range : AllCases::instance().fetch()[head]) {
-//            common_codes.emplace_back(head << 32 | range);
-//        }
-//    }
-
-//    std::vector<std::string> common_codes_str;
-//    common_codes_str.reserve(klotski::cases::ALL_CASES_NUM_);
-//    for (auto x : common_codes) {
-//        common_codes_str.emplace_back(CommonCode::string_encode(x, false));
-//    }
-
-    // ShortCode::speed_up(true);
+    // auto ret = cal.solve();
     //
-    // for (uint32_t short_code = 0; short_code < SHORT_CODE_LIMIT; ++short_code) {
-    //     ShortCode::unsafe_create(short_code).to_common_code();
+    // for (auto kk : cal.backtrack(ret)) {
+    //     std::cout << kk.to_common_code() << "," << kk.to_common_code().to_short_code() << std::endl;
     // }
 
-//    for (auto common_code : common_codes) {
-//        printf("%llX\n", common_code);
-//        CommonCode::string_encode(common_code, true);
-//        CommonCode::string_encode(common_code, false);
-//        printf("%s\n", CommonCode::string_encode(common_code, false).c_str());
-//        std::cout << CommonCode::string_encode(common_code, false) << std::endl;
-//    }
-
-//    for (auto &common_code_str : common_codes_str) {
-//        CommonCode::string_decode(common_code_str);
-//    }
+    for (const auto solve : cal.solve_multi()) {
+        for (const auto raw_code : cal.backtrack(solve)) {
+            const auto common_code = raw_code.to_common_code();
+            std::cout << common_code << "/" << common_code.to_short_code() << std::endl;
+        }
+        std::cout << "----" << std::endl;
+    }
 
     std::cerr << ((clock() - start) * 1000 / CLOCKS_PER_SEC) << "ms" << std::endl;