diff --git a/3rdparty/capo/make.hpp b/3rdparty/capo/make.hpp
new file mode 100755
index 0000000..908849c
--- /dev/null
+++ b/3rdparty/capo/make.hpp
@@ -0,0 +1,22 @@
+/*
+    The Capo Library
+    Code covered by the MIT License
+    Author: mutouyun (http://orzz.org)
+*/
+
+#pragma once
+
+#include <type_traits>  // std::decay
+#include <utility>      // std::forward
+
+namespace capo
+{
+    template <template <typename...> class Ret, typename... T>
+    using return_t = Ret<typename std::decay<T>::type...>;
+
+    template <template <typename...> class Ret, typename... T>
+    inline return_t<Ret, T...> make(T&&... args)
+    {
+        return return_t<Ret, T...>(std::forward<T>(args)...);
+    }
+}
\ No newline at end of file
diff --git a/3rdparty/capo/noncopyable.hpp b/3rdparty/capo/noncopyable.hpp
new file mode 100755
index 0000000..cb70a11
--- /dev/null
+++ b/3rdparty/capo/noncopyable.hpp
@@ -0,0 +1,25 @@
+/*
+    The Capo Library
+    Code covered by the MIT License
+    Author: mutouyun (http://orzz.org)
+*/
+
+#pragma once
+
+namespace capo {
+
+////////////////////////////////////////////////////////////////
+
+class noncopyable
+{
+protected:
+    noncopyable(void) = default;
+    ~noncopyable(void) = default;
+public:
+    noncopyable(const noncopyable&) = delete;
+    noncopyable& operator=(const noncopyable&) = delete;
+};
+
+////////////////////////////////////////////////////////////////
+
+} // namespace capo
\ No newline at end of file
diff --git a/3rdparty/capo/scope_guard.hpp b/3rdparty/capo/scope_guard.hpp
new file mode 100755
index 0000000..306b689
--- /dev/null
+++ b/3rdparty/capo/scope_guard.hpp
@@ -0,0 +1,103 @@
+/*
+    The Capo Library
+    Code covered by the MIT License
+    Author: mutouyun (http://orzz.org)
+*/
+
+#pragma once
+
+#include "capo/noncopyable.hpp"
+#include "capo/unused.hpp"
+#include "capo/make.hpp"
+
+#include <utility>      // std::forward, std::move
+#include <algorithm>    // std::swap
+#include <functional>   // std::function
+
+namespace capo {
+
+////////////////////////////////////////////////////////////////
+/// Execute guard function when the enclosing scope exits
+////////////////////////////////////////////////////////////////
+
+template <typename F = std::function<void()>>
+class scope_guard : capo::noncopyable
+{
+    F destructor_;
+    mutable bool dismiss_;
+
+public:
+    template <typename F_>
+    scope_guard(F_&& destructor)
+        : destructor_(std::forward<F_>(destructor))
+        , dismiss_(false)
+    {}
+
+    scope_guard(scope_guard&& rhs)
+        : destructor_(std::move(rhs.destructor_))
+        , dismiss_(true) // dismiss rhs
+    {
+        std::swap(dismiss_, rhs.dismiss_);
+    }
+
+    ~scope_guard(void)
+    {
+        try { do_exit(); }
+        /*
+            In the realm of exceptions, it is fundamental that you can do nothing
+            if your "undo/recover" action fails.
+        */
+        catch(...) { /* Do nothing */ }
+    }
+
+    void dismiss(void) const noexcept
+    {
+        dismiss_ = true;
+    }
+
+    void do_exit(void)
+    {
+        if (!dismiss_)
+        {
+            dismiss_ = true;
+            destructor_();
+        }
+    }
+
+    void swap(scope_guard& rhs)
+    {
+        std::swap(destructor_, rhs.destructor_);
+        std::swap(dismiss_,    rhs.dismiss_);
+    }
+};
+
+namespace detail_scope_guard {
+
+struct helper
+{
+    template <typename F>
+    auto operator=(F&& destructor) -> decltype(capo::make<scope_guard>(std::forward<F>(destructor)))
+    {
+        return capo::make<scope_guard>(std::forward<F>(destructor));
+    }
+};
+
+} // namespace detail_scope_guard
+
+#define CAPO_SCOPE_GUARD_V_(L)  CAPO_UNUSED_ scope_guard_##L##__
+#define CAPO_SCOPE_GUARD_L_(L)  auto CAPO_SCOPE_GUARD_V_(L) = capo::detail_scope_guard::helper{}
+
+/*
+    Do things like this:
+    -->
+    CAPO_SCOPE_GUARD_ = [ptr]
+    {
+        if (ptr) free(ptr);
+    };
+*/
+
+#define CAPO_SCOPE_GUARD_ CAPO_SCOPE_GUARD_L_(__LINE__)
+
+////////////////////////////////////////////////////////////////
+
+} // namespace capo
\ No newline at end of file
diff --git a/3rdparty/capo/type_name.hpp b/3rdparty/capo/type_name.hpp
new file mode 100755
index 0000000..aee3033
--- /dev/null
+++ b/3rdparty/capo/type_name.hpp
@@ -0,0 +1,320 @@
+/*
+    The Capo Library
+    Code covered by the MIT License
+    Author: mutouyun (http://orzz.org)
+*/
+
+#pragma once
+
+#include "capo/scope_guard.hpp"
+
+#include <typeinfo>     // typeid
+#include <sstream>      // std::ostringstream, std::string
+#include <type_traits>  // std::is_array
+
+#if defined(__GNUC__)
+#   include <cxxabi.h>  // abi::__cxa_demangle
+#endif/*__GNUC__*/
+
+namespace capo {
+namespace detail_type_name {
+
+////////////////////////////////////////////////////////////////
+/// Ready for output
+////////////////////////////////////////////////////////////////
+
+// Forward declarations
+template <typename T, bool IsBase = false>
+struct check;
+
+/*
+    Output state management
+*/
+
+class output
+{
+    bool is_compact_ = true;
+
+    template <typename T>
+    bool check_empty(const T&) { return false; }
+    bool check_empty(const char* val)
+    {
+        return (!val) || (val[0] == 0);
+    }
+
+    template <typename T>
+    void out(const T& val)
+    {
+        if (check_empty(val)) return;
+        if (!is_compact_) sr_ += " ";
+        std::ostringstream ss;
+        ss << val;
+        sr_ += ss.str();
+        is_compact_ = false;
+    }
+
+    std::string& sr_;
+
+public:
+    output(std::string& sr) : sr_(sr) {}
+
+    output& operator()(void) { return (*this); }
+
+    template <typename T1, typename... T>
+    output& operator()(const T1& val, const T&... args)
+    {
+        out(val);
+        return operator()(args...);
+    }
+
+    output& compact(void)
+    {
+        is_compact_ = true;
+        return (*this);
+    }
+};
+
+// ()
+
+template <bool>
+struct bracket
+{
+    output& out_;
+
+    bracket(output& out, const char* = nullptr) : out_(out)
+    { out_("(").compact(); }
+
+    ~bracket(void)
+    { out_.compact()(")"); }
+};
+
+template <>
+struct bracket<false>
+{
+    bracket(output& out, const char* str = nullptr)
+    { out(str); }
+};
+
+// [N]
+
+template <size_t N = 0>
+struct bound
+{
+    output& out_;
+
+    bound(output& out) : out_(out) {}
+    ~bound(void)
+    {
+        if (N == 0) out_("[]");
+        else        out_("[").compact()
+                        ( N ).compact()
+                        ("]");
+    }
+};
+
+// (P1, P2, ...)
+
+template <bool, typename... P>
+struct parameter;
+
+template <bool IsStart>
+struct parameter<IsStart>
+{
+    output& out_;
+
+    parameter(output& out) : out_(out) {}
+    ~parameter(void)
+    { bracket<IsStart> { out_ }; }
+};
+
+template <bool IsStart, typename P1, typename... P>
+struct parameter<IsStart, P1, P...>
+{
+    output& out_;
+
+    parameter(output& out) : out_(out) {}
+    ~parameter(void)
+    {
+        [this](bracket<IsStart>&&)
+        {
+            check<P1> { out_ };
+            parameter<false, P...> { out_.compact() };
+        } (bracket<IsStart> { out_, "," });
+    }
+};
+
+////////////////////////////////////////////////////////////////
+/// Template specializations for checking
+////////////////////////////////////////////////////////////////
+
+/*
+    CV-qualifiers, references, pointers
+*/
+
+template <typename T, bool IsBase>
+struct check
+{
+    output out_;
+    check(const output& out) : out_(out)
+    {
+#   if defined(__GNUC__)
+        const char* typeid_name = typeid(T).name();
+        char* real_name = abi::__cxa_demangle(typeid_name, nullptr, nullptr, nullptr);
+        CAPO_SCOPE_GUARD_ = [real_name]
+        {
+            if (real_name) ::free(real_name);
+        };
+        out_(real_name ? real_name : typeid_name);
+#   else /*__GNUC__*/
+        out_(typeid(T).name());
+#   endif/*__GNUC__*/
+    }
+};
+
+#pragma push_macro("CAPO_CHECK_TYPE__")
+#undef  CAPO_CHECK_TYPE__
+#define CAPO_CHECK_TYPE__(OPT)                                 \
+    template <typename T, bool IsBase>                         \
+    struct check<T OPT, IsBase> : check<T, true>               \
+    {                                                          \
+        using base_t = check<T, true>;                         \
+        using base_t::out_;                                    \
+        check(const output& out) : base_t(out) { out_(#OPT); } \
+    };
+
+CAPO_CHECK_TYPE__(const)
+CAPO_CHECK_TYPE__(volatile)
+CAPO_CHECK_TYPE__(const volatile)
+CAPO_CHECK_TYPE__(&)
+CAPO_CHECK_TYPE__(&&)
+CAPO_CHECK_TYPE__(*)
+
+#pragma pop_macro("CAPO_CHECK_TYPE__")
+
+/*
+    Arrays
+*/
+
+#pragma push_macro("CAPO_CHECK_TYPE_ARRAY__")
+#pragma push_macro("CAPO_CHECK_TYPE_ARRAY_CV__")
+#pragma push_macro("CAPO_CHECK_TYPE_PLACEHOLDER__")
+
+#undef CAPO_CHECK_TYPE_ARRAY__
+#undef CAPO_CHECK_TYPE_ARRAY_CV__
+#undef CAPO_CHECK_TYPE_PLACEHOLDER__
+
+#define CAPO_CHECK_TYPE_ARRAY__(CV_OPT, BOUND_OPT, ...)                                    \
+    template <typename T, bool IsBase __VA_ARGS__>                                         \
+    struct check<T CV_OPT [BOUND_OPT], IsBase> : check<T CV_OPT, !std::is_array<T>::value> \
+    {                                                                                      \
+        using base_t = check<T CV_OPT, !std::is_array<T>::value>;                          \
+        using base_t::out_;                                                                \
+                                                                                           \
+        bound<BOUND_OPT> bound_   = out_;                                                  \
+        bracket<IsBase>  bracket_ = out_;                                                  \
+                                                                                           \
+        check(const output& out) : base_t(out) {}                                          \
+    };
+
+#define CAPO_CHECK_TYPE_ARRAY_CV__(BOUND_OPT, ...)               \
+    CAPO_CHECK_TYPE_ARRAY__(, BOUND_OPT, ,##__VA_ARGS__)         \
+    CAPO_CHECK_TYPE_ARRAY__(const, BOUND_OPT, ,##__VA_ARGS__)    \
+    CAPO_CHECK_TYPE_ARRAY__(volatile, BOUND_OPT, ,##__VA_ARGS__) \
+    CAPO_CHECK_TYPE_ARRAY__(const volatile, BOUND_OPT, ,##__VA_ARGS__)
+
+#define CAPO_CHECK_TYPE_PLACEHOLDER__
+
+CAPO_CHECK_TYPE_ARRAY_CV__(CAPO_CHECK_TYPE_PLACEHOLDER__)
+#if defined(__GNUC__)
+CAPO_CHECK_TYPE_ARRAY_CV__(0)
+#endif/*__GNUC__*/
+CAPO_CHECK_TYPE_ARRAY_CV__(N, size_t N)
+
+#pragma pop_macro("CAPO_CHECK_TYPE_PLACEHOLDER__")
+#pragma pop_macro("CAPO_CHECK_TYPE_ARRAY_CV__")
+#pragma pop_macro("CAPO_CHECK_TYPE_ARRAY__")
+
+/*
+    Functions
+*/
+
+template <typename T, bool IsBase, typename... P>
+struct check<T(P...), IsBase> : check<T, true>
+{
+    using base_t = check<T, true>;
+    using base_t::out_;
+
+    parameter<true, P...> parameter_ = out_;
+    bracket<IsBase>       bracket_   = out_;
+
+    check(const output& out) : base_t(out) {}
+};
+
+/*
+    Pointers to members
+*/
+
+template <typename T, bool IsBase, typename C>
+struct check<T C::*, IsBase> : check<T, true>
+{
+    using base_t = check<T, true>;
+    using base_t::out_;
+
+    check(const output& out) : base_t(out)
+    {
+        check<C> { out_ };
+        out_.compact()("::*");
+    }
+};
+
+/*
+    Pointers to member functions
+*/
+
+#pragma push_macro("CAPO_CHECK_TYPE_MEM_FUNC__")
+#undef  CAPO_CHECK_TYPE_MEM_FUNC__
+#define CAPO_CHECK_TYPE_MEM_FUNC__(...)                           \
+    template <typename T, bool IsBase, typename C, typename... P> \
+    struct check<T(C::*)(P...) __VA_ARGS__, IsBase>               \
+    {                                                             \
+        scope_guard<> at_destruct_cv_;                            \
+        check<T(P...), true> base_;                               \
+        output& out_ = base_.out_;                                \
+                                                                  \
+        check(const output& out)                                  \
+            : at_destruct_cv_([this]{ out_(#__VA_ARGS__); })      \
+            , base_(out)                                          \
+        {                                                         \
+            check<C> { out_ };                                    \
+            out_.compact()("::*");                                \
+        }                                                         \
+    };
+
+CAPO_CHECK_TYPE_MEM_FUNC__()
+CAPO_CHECK_TYPE_MEM_FUNC__(const)
+CAPO_CHECK_TYPE_MEM_FUNC__(volatile)
+CAPO_CHECK_TYPE_MEM_FUNC__(const volatile)
+
+#pragma pop_macro("CAPO_CHECK_TYPE_MEM_FUNC__")
+
+} // namespace detail_type_name
+
+////////////////////////////////////////////////////////////////
+/// Get the name of the given type
+////////////////////////////////////////////////////////////////
+
+/*
+    type_name<const volatile void *>()
+    -->
+    void const volatile *
+*/
+
+template <typename T>
+inline std::string type_name(void)
+{
+    std::string str;
+    detail_type_name::check<T> { str };
+    return str;
+}
+
+} // namespace capo
\ No newline at end of file
diff --git a/3rdparty/capo/unused.hpp b/3rdparty/capo/unused.hpp
new file mode 100755
index 0000000..b79d7ef
--- /dev/null
+++ b/3rdparty/capo/unused.hpp
@@ -0,0 +1,23 @@
+/*
+    The Capo Library
+    Code covered by the MIT License
+    Author: mutouyun (http://orzz.org)
+*/
+
+#pragma once
+
+////////////////////////////////////////////////////////////////
+
+#ifdef CAPO_UNUSED_
+#   error "CAPO_UNUSED_ has been defined."
+#endif
+
+#if defined(_MSC_VER)
+#   define CAPO_UNUSED_ __pragma(warning(suppress: 4100 4101 4189))
+#elif defined(__GNUC__)
+#   define CAPO_UNUSED_ __attribute__((__unused__))
+#else
+#   define CAPO_UNUSED_
+#endif
+
+////////////////////////////////////////////////////////////////
\ No newline at end of file
diff --git a/demo/chat/CMakeLists.txt b/demo/chat/CMakeLists.txt
old mode 100644
new mode 100755
index 8dfabce..c5bcea8
--- a/demo/chat/CMakeLists.txt
+++ b/demo/chat/CMakeLists.txt
@@ -6,6 +6,3 @@ file(GLOB HEAD_FILES ./*.h)
 add_executable(${PROJECT_NAME} ${SRC_FILES} ${HEAD_FILES})
 
 target_link_libraries(${PROJECT_NAME} ipc)
-if(NOT MSVC)
-  target_link_libraries(${PROJECT_NAME} pthread rt)
-endif()
diff --git a/include/ipc.h b/include/ipc.h
old mode 100644
new mode 100755
index 187b01b..f72acef
--- a/include/ipc.h
+++ b/include/ipc.h
@@ -27,10 +27,10 @@ struct IPC_EXPORT chan_impl {
     static std::size_t recv_count(handle_t h);
     static bool wait_for_recv(handle_t h, std::size_t r_count, std::size_t tm);
 
-    static bool   send(handle_t h, void const * data, std::size_t size);
+    static bool   send(handle_t h, void const * data, std::size_t size, std::size_t tm);
     static buff_t recv(handle_t h, std::size_t tm);
 
-    static bool   try_send(handle_t h, void const * data, std::size_t size);
+    static bool   try_send(handle_t h, void const * data, std::size_t size, std::size_t tm);
     static buff_t try_recv(handle_t h);
 };
 
@@ -38,16 +38,18 @@ template <typename Flag>
 class chan_wrapper {
 private:
     using detail_t = chan_impl<Flag>;
-    handle_t h_ = nullptr;
+
+    handle_t h_    = nullptr;
+    unsigned mode_ = ipc::sender;
 
 public:
     chan_wrapper() = default;
 
-    explicit chan_wrapper(char const * name, unsigned mode = sender) {
+    explicit chan_wrapper(char const * name, unsigned mode = ipc::sender) {
         this->connect(name, mode);
     }
 
-    chan_wrapper(chan_wrapper&& rhs) {
+    chan_wrapper(chan_wrapper&& rhs) noexcept {
         swap(rhs);
     }
 
@@ -55,7 +57,7 @@ public:
         disconnect();
     }
 
-    void swap(chan_wrapper& rhs) {
+    void swap(chan_wrapper& rhs) noexcept {
         std::swap(h_, rhs.h_);
     }
 
@@ -68,22 +70,26 @@ public:
         return detail_t::name(h_);
     }
 
-    handle_t handle() const {
+    handle_t handle() const noexcept {
         return h_;
     }
 
-    bool valid() const {
+    bool valid() const noexcept {
         return (handle() != nullptr);
     }
 
-    chan_wrapper clone() const {
-        return chan_wrapper { name() };
+    unsigned mode() const noexcept {
+        return mode_;
     }
 
-    bool connect(char const * name, unsigned mode = sender | receiver) {
+    chan_wrapper clone() const {
+        return chan_wrapper { name(), mode_ };
+    }
+
+    bool connect(char const * name, unsigned mode = ipc::sender | ipc::receiver) {
         if (name == nullptr || name[0] == '\0') return false;
         this->disconnect();
-        h_ = detail_t::connect(name, mode);
+        h_ = detail_t::connect(name, mode_ = mode);
         return valid();
     }
 
@@ -105,13 +111,13 @@ public:
         return chan_wrapper(name).wait_for_recv(r_count, tm);
     }
 
-    bool send    (void        const * data, std::size_t size) { return detail_t::send(h_, data, size)             ; }
-    bool send    (buff_t      const & buff)                   { return     this->send(buff.data(), buff.size())   ; }
-    bool send    (std::string const & str)                    { return     this->send(str.c_str(), str.size() + 1); }
+    bool send    (void        const * data, std::size_t size, std::size_t tm = default_timeout) { return detail_t::send(h_, data, size, tm)             ; }
+    bool send    (buff_t      const & buff                  , std::size_t tm = default_timeout) { return     this->send(buff.data(), buff.size(), tm)   ; }
+    bool send    (std::string const & str                   , std::size_t tm = default_timeout) { return     this->send(str.c_str(), str.size() + 1, tm); }
 
-    bool try_send(void        const * data, std::size_t size) { return detail_t::try_send(h_, data, size)             ; }
-    bool try_send(buff_t      const & buff)                   { return     this->try_send(buff.data(), buff.size())   ; }
-    bool try_send(std::string const & str)                    { return     this->try_send(str.c_str(), str.size() + 1); }
+    bool try_send(void        const * data, std::size_t size, std::size_t tm = default_timeout) { return detail_t::try_send(h_, data, size, tm)             ; }
+    bool try_send(buff_t      const & buff                  , std::size_t tm = default_timeout) { return     this->try_send(buff.data(), buff.size(), tm)   ; }
+    bool try_send(std::string const & str                   , std::size_t tm = default_timeout) { return     this->try_send(str.c_str(), str.size() + 1, tm); }
 
     buff_t recv(std::size_t tm = invalid_value) {
         return detail_t::recv(h_, tm);
diff --git a/include/rw_lock.h b/include/rw_lock.h
old mode 100644
new mode 100755
index 052ce03..2a898b1
--- a/include/rw_lock.h
+++ b/include/rw_lock.h
@@ -63,8 +63,10 @@ inline void yield(K& k) noexcept {
     if (k < 4)  { /* Do nothing */ }
     else
     if (k < 16) { IPC_LOCK_PAUSE_(); }
+    else
+    if (k < 32) { std::this_thread::yield(); }
     else {
-        std::this_thread::yield();
+        std::this_thread::sleep_for(std::chrono::milliseconds(1));
         return;
     }
     ++k;
diff --git a/include/tls_pointer.h b/include/tls_pointer.h
index fa6cced..41b0b6b 100755
--- a/include/tls_pointer.h
+++ b/include/tls_pointer.h
@@ -1,87 +1,112 @@
 #pragma once
 
-#include <cstdint>
-#include <utility>
-#include <limits>
+#include <memory>   // std::unique_ptr
+#include <utility>  // std::forward
+#include <cstddef>  // std::size_t
 
 #include "export.h"
+#include "platform/detail.h"
 
 namespace ipc {
 namespace tls {
 
-using key_t        = std::uint64_t;
-using destructor_t = void (*)(void*);
+using key_t = std::size_t;
+using destructor_t = void (*)(void *);
 
-enum : key_t {
-    invalid_value = (std::numeric_limits<key_t>::max)()
+struct key_info {
+    key_t key_;
 };
 
-IPC_EXPORT key_t create (destructor_t destructor = nullptr);
-IPC_EXPORT void  release(key_t key);
+IPC_EXPORT bool create (key_info * pkey, destructor_t destructor = nullptr);
+IPC_EXPORT void release(key_info const * pkey);
 
-IPC_EXPORT bool  set(key_t key, void* ptr);
-IPC_EXPORT void* get(key_t key);
+IPC_EXPORT bool   set(key_info const * pkey, void * ptr);
+IPC_EXPORT void * get(key_info const * pkey);
 
 ////////////////////////////////////////////////////////////////
 /// Thread-local pointer
 ////////////////////////////////////////////////////////////////
 
-/*
- * <Remarks>
- *
+/**
+ * @remarks
  * You need to set the ipc::tls::pointer's storage manually:
  * ```
  *     tls::pointer<int> p;
  *     if (!p) p = new int(123);
  * ```
  * It would be like an ordinary pointer.
- * Or you could just call create to 'new' this pointer automatically.
+ * Or you could just call create_once to 'new' this pointer automatically.
  * ```
  *     tls::pointer<int> p;
- *     p.create(123);
+ *     p.create_once(123);
  * ```
 */
 
 template <typename T>
-class pointer {
-    key_t key_;
+class pointer : public key_info {
+
+    pointer(pointer const &) = delete;
+    pointer & operator=(pointer const &) = delete;
+
+    void destruct() const {
+        delete static_cast<T*>(tls::get(this));
+    }
 
 public:
     using value_type = T;
 
-    pointer()
-        : key_(tls::create([](void* p) { delete static_cast<T*>(p); })) {
+    pointer() {
+        tls::create(this, [](void* p) {
+            delete static_cast<T*>(p);
+        });
     }
 
     ~pointer() {
-        tls::release(key_);
+        destruct();
+        tls::release(this);
     }
 
     template <typename... P>
     T* create(P&&... params) {
-        auto ptr = static_cast<T*>(get(key_));
-        if (ptr == nullptr) {
-            ptr = new T(std::forward<P>(params)...);
-            if (!set(key_, ptr)) {
-                delete ptr;
+        destruct();
+        auto ptr = detail::unique_ptr(new T(std::forward<P>(params)...));
+        if (!tls::set(this, ptr.get())) {
+            return nullptr;
+        }
+        return ptr.release();
+    }
+
+    template <typename... P>
+    T* create_once(P&&... params) {
+        auto p = static_cast<T*>(tls::get(this));
+        if (p == nullptr) {
+            auto ptr = detail::unique_ptr(new T(std::forward<P>(params)...));
+            if (!tls::set(this, ptr.get())) {
                 return nullptr;
             }
+            p = ptr.release();
         }
-        return ptr;
+        return p;
     }
 
-    T* operator=(T* ptr) {
-        set(key_, ptr);
-        return ptr;
+    T* operator=(T* p) {
+        set(this, p);
+        return p;
     }
 
-    operator T*() const { return static_cast<T*>(get(key_)); }
+    explicit operator T *() const {
+        return static_cast<T *>(tls::get(this));
+    }
 
-    T&       operator* ()       { return *static_cast<T*>(*this); }
-    const T& operator* () const { return *static_cast<T*>(*this); }
+    explicit operator bool() const {
+        return tls::get(this) != nullptr;
+    }
 
-    T*       operator->()       { return  static_cast<T*>(*this); }
-    const T* operator->() const { return  static_cast<T*>(*this); }
+    T &       operator* ()       { return *static_cast<T *>(*this); }
+    const T & operator* () const { return *static_cast<T *>(*this); }
+
+    T *       operator->()       { return  static_cast<T *>(*this); }
+    const T * operator->() const { return  static_cast<T *>(*this); }
 };
 
 } // namespace tls
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
old mode 100644
new mode 100755
index 36bdfbb..904dd94
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -2,10 +2,13 @@ project(ipc)
 
 add_compile_options(-D__IPC_LIBRARY__)
 
+if(NOT MSVC)
+  add_compile_options(-fPIC)
+endif()
+
 include_directories(
     ${CMAKE_SOURCE_DIR}/include 
-    ${CMAKE_SOURCE_DIR}/src
-)
+    ${CMAKE_SOURCE_DIR}/src)
 
 if(UNIX)
   file(GLOB SRC_FILES ${CMAKE_SOURCE_DIR}/src/platform/*_linux.cpp)
@@ -16,3 +19,8 @@ aux_source_directory(${CMAKE_SOURCE_DIR}/src SRC_FILES)
 file(GLOB HEAD_FILES ${CMAKE_SOURCE_DIR}/include/*.h ${CMAKE_SOURCE_DIR}/src/*.inc ${CMAKE_SOURCE_DIR}/src/memory/*.hpp)
 
 add_library(${PROJECT_NAME} SHARED ${SRC_FILES} ${HEAD_FILES})
+if(NOT MSVC)
+  target_link_libraries(${PROJECT_NAME} PUBLIC
+      pthread
+      $<$<NOT:$<STREQUAL:${CMAKE_SYSTEM_NAME},Windows>>:rt>)
+endif()
diff --git a/src/concept.h b/src/concept.h
old mode 100644
new mode 100755
index 4905925..9325ddc
--- a/src/concept.h
+++ b/src/concept.h
@@ -1,13 +1,13 @@
 #pragma once
 
-#include <type_traits>
+#include <type_traits>  // std::enable_if
 
 namespace ipc {
 
 // concept helpers
 
 template <bool Cond, typename R = void>
-using require = std::enable_if_t<Cond, R>;
+using require = typename std::enable_if<Cond, R>::type;
 
 #ifdef IPC_CONCEPT_
 #   error "IPC_CONCEPT_ has been defined."
diff --git a/src/ipc.cpp b/src/ipc.cpp
old mode 100644
new mode 100755
index 04f4a01..516c850
--- a/src/ipc.cpp
+++ b/src/ipc.cpp
@@ -283,7 +283,7 @@ struct conn_info_head {
     }
 
     auto& recv_cache() {
-        return *recv_cache_.create();
+        return *recv_cache_.create_once();
     }
 };
 
@@ -440,21 +440,21 @@ static bool send(F&& gen_push, ipc::handle_t h, void const * data, std::size_t s
     return true;
 }
 
-static bool send(ipc::handle_t h, void const * data, std::size_t size) {
-    return send([](auto info, auto que, auto msg_id) {
-        return [info, que, msg_id](std::int32_t remain, void const * data, std::size_t size) {
+static bool send(ipc::handle_t h, void const * data, std::size_t size, std::size_t tm) {
+    return send([tm](auto info, auto que, auto msg_id) {
+        return [tm, info, que, msg_id](std::int32_t remain, void const * data, std::size_t size) {
             if (!wait_for(info->wt_waiter_, [&] {
                     return !que->push(info->cc_id_, msg_id, remain, data, size);
-                }, default_timeout)) {
+                }, tm)) {
                 ipc::log("force_push: msg_id = %zd, remain = %d, size = %zd\n", msg_id, remain, size);
                 if (!que->force_push([](void* p) {
-                    auto tmp_msg = static_cast<typename queue_t::value_t*>(p);
-                    if (tmp_msg->storage_) {
-                        clear_storage(*reinterpret_cast<std::size_t*>(&tmp_msg->data_), 
-                                      static_cast<std::int32_t>(data_length) + tmp_msg->remain_);
-                    }
-                    return true;
-                }, info->cc_id_, msg_id, remain, data, size)) {
+                        auto tmp_msg = static_cast<typename queue_t::value_t*>(p);
+                        if (tmp_msg->storage_) {
+                            clear_storage(*reinterpret_cast<std::size_t*>(&tmp_msg->data_), 
+                                        static_cast<std::int32_t>(data_length) + tmp_msg->remain_);
+                        }
+                        return true;
+                    }, info->cc_id_, msg_id, remain, data, size)) {
                     return false;
                 }
             }
@@ -464,12 +464,12 @@ static bool send(ipc::handle_t h, void const * data, std::size_t size) {
     }, h, data, size);
 }
 
-static bool try_send(ipc::handle_t h, void const * data, std::size_t size) {
-    return send([](auto info, auto que, auto msg_id) {
-        return [info, que, msg_id](std::int32_t remain, void const * data, std::size_t size) {
+static bool try_send(ipc::handle_t h, void const * data, std::size_t size, std::size_t tm) {
+    return send([tm](auto info, auto que, auto msg_id) {
+        return [tm, info, que, msg_id](std::int32_t remain, void const * data, std::size_t size) {
             if (!wait_for(info->wt_waiter_, [&] {
                     return !que->push(info->cc_id_, msg_id, remain, data, size);
-                }, 0)) {
+                }, tm)) {
                 return false;
             }
             info->rd_waiter_.broadcast();
@@ -587,8 +587,8 @@ bool chan_impl<Flag>::wait_for_recv(ipc::handle_t h, std::size_t r_count, std::s
 }
 
 template <typename Flag>
-bool chan_impl<Flag>::send(ipc::handle_t h, void const * data, std::size_t size) {
-    return detail_impl<policy_t<Flag>>::send(h, data, size);
+bool chan_impl<Flag>::send(ipc::handle_t h, void const * data, std::size_t size, std::size_t tm) {
+    return detail_impl<policy_t<Flag>>::send(h, data, size, tm);
 }
 
 template <typename Flag>
@@ -597,8 +597,8 @@ buff_t chan_impl<Flag>::recv(ipc::handle_t h, std::size_t tm) {
 }
 
 template <typename Flag>
-bool chan_impl<Flag>::try_send(ipc::handle_t h, void const * data, std::size_t size) {
-    return detail_impl<policy_t<Flag>>::try_send(h, data, size);
+bool chan_impl<Flag>::try_send(ipc::handle_t h, void const * data, std::size_t size, std::size_t tm) {
+    return detail_impl<policy_t<Flag>>::try_send(h, data, size, tm);
 }
 
 template <typename Flag>
diff --git a/src/memory/wrapper.h b/src/memory/wrapper.h
index f511843..05f130d 100755
--- a/src/memory/wrapper.h
+++ b/src/memory/wrapper.h
@@ -166,7 +166,7 @@ public:
     template <typename ... P>
     async_wrapper(P ... pars) {
         get_alloc_ = [this, pars ...]()->ref_t {
-            return *tls_.create(this, pars ...);
+            return *tls_.create_once(this, pars ...);
         };
     }
 
@@ -229,8 +229,8 @@ struct default_mapping_policy {
     };
 
     template <typename F, typename ... P>
-    IPC_CONSTEXPR_ static void foreach(F f, P ... params) {
-        for (std::size_t i = 0; i < classes_size; ++i) f(i, params...);
+    IPC_CONSTEXPR_ static void foreach(F && f, P && ... params) {
+        for (std::size_t i = 0; i < classes_size; ++i) f(i, std::forward<P>(params)...);
     }
 
     IPC_CONSTEXPR_ static std::size_t block_size(std::size_t id) noexcept {
@@ -238,9 +238,9 @@ struct default_mapping_policy {
     }
 
     template <typename F, typename D, typename ... P>
-    IPC_CONSTEXPR_ static auto classify(F f, D d, std::size_t size, P ... params) {
+    IPC_CONSTEXPR_ static auto classify(F && f, D && d, std::size_t size, P && ... params) {
         std::size_t id = (size - base_size - 1) / iter_size;
-        return (id < classes_size) ? f(id, size, params...) : d(size, params...);
+        return (id < classes_size) ? f(id, size, std::forward<P>(params)...) : d(size, std::forward<P>(params)...);
     }
 };
 
@@ -251,41 +251,49 @@ class variable_wrapper {
 
     struct initiator {
 
-        std::aligned_storage_t< sizeof(FixedAlloc), 
-                               alignof(FixedAlloc)> arr_[MappingP::classes_size];
+        using falc_t = std::aligned_storage_t<sizeof(FixedAlloc), alignof(FixedAlloc)>;
+        falc_t arr_[MappingP::classes_size];
 
         initiator() {
-            MappingP::foreach([](std::size_t id, initiator* t) {
-                ::new (&(t->arr_[id])) FixedAlloc(MappingP::block_size(id));
-            }, this);
+            MappingP::foreach([](std::size_t id, falc_t * a) {
+                ::new (&(a[id])) FixedAlloc(MappingP::block_size(id));
+            }, arr_);
         }
 
-        FixedAlloc& at(std::size_t id) {
-            return reinterpret_cast<FixedAlloc&>(arr_[id]);
+        ~initiator() {
+            MappingP::foreach([](std::size_t id, falc_t * a) {
+                initiator::at(a, id).~FixedAlloc();
+            }, arr_);
+        }
+
+        static FixedAlloc & at(falc_t * arr, std::size_t id) noexcept {
+            return reinterpret_cast<FixedAlloc&>(arr[id]);
         }
     } init_;
 
+    using falc_t = typename initiator::falc_t;
+
 public:
     void swap(variable_wrapper & other) {
-        MappingP::foreach([](std::size_t id, initiator* t, initiator* o) {
-            t->at(id).swap(o->at(id));
-        }, &init_, &other.init_);
+        MappingP::foreach([](std::size_t id, falc_t * in, falc_t * ot) {
+            initiator::at(in, id).swap(initiator::at(ot, id));
+        }, init_.arr_, other.init_.arr_);
     }
 
     void* alloc(std::size_t size) {
-        return MappingP::classify([](std::size_t id, std::size_t size, initiator* t) {
-            return t->at(id).alloc(size);
-        }, [](std::size_t size, initiator*) {
+        return MappingP::classify([](std::size_t id, std::size_t size, falc_t * a) {
+            return initiator::at(a, id).alloc(size);
+        }, [](std::size_t size, falc_t *) {
             return DefaultAlloc::alloc(size);
-        }, size, &init_);
+        }, size, init_.arr_);
     }
 
     void free(void* p, std::size_t size) {
-        MappingP::classify([](std::size_t id, std::size_t size, void* p, initiator* t) {
-            t->at(id).free(p, size);
-        }, [](std::size_t size, void* p, initiator*) {
+        MappingP::classify([](std::size_t id, std::size_t size, void* p, falc_t * a) {
+            initiator::at(a, id).free(p, size);
+        }, [](std::size_t size, void* p, falc_t *) {
             DefaultAlloc::free(p, size);
-        }, size, p, &init_);
+        }, size, p, init_.arr_);
     }
 };
 
diff --git a/src/platform/detail.h b/src/platform/detail.h
index d2fb294..fbf539a 100755
--- a/src/platform/detail.h
+++ b/src/platform/detail.h
@@ -5,13 +5,7 @@
 #include <shared_mutex>
 #include <type_traits>
 #include <tuple>
-#include <atomic>
 #include <algorithm>
-#include <utility>
-#include <new>
-
-#include "def.h"
-#include "export.h"
 
 // pre-defined
 
@@ -61,6 +55,8 @@
 namespace std {
 
 // deduction guides for std::unique_ptr
+template <typename T>
+unique_ptr(T* p) -> unique_ptr<T>;
 template <typename T, typename D>
 unique_ptr(T* p, D&& d) -> unique_ptr<T, std::decay_t<D>>;
 
@@ -81,20 +77,25 @@ namespace ipc {
 namespace detail {
 
 // deduction guides for std::unique_ptr
+template <typename T>
+constexpr auto unique_ptr(T* p) noexcept {
+    return std::unique_ptr<T> { p };
+}
+
 template <typename T, typename D>
-constexpr auto unique_ptr(T* p, D&& d) {
+constexpr auto unique_ptr(T* p, D&& d) noexcept {
     return std::unique_ptr<T, std::decay_t<D>> { p, std::forward<D>(d) };
 }
 
 // deduction guides for std::unique_lock
 template <typename T>
-constexpr auto unique_lock(T&& lc) {
+constexpr auto unique_lock(T&& lc) noexcept {
     return std::unique_lock<std::decay_t<T>> { std::forward<T>(lc) };
 }
 
 // deduction guides for std::shared_lock
 template <typename T>
-constexpr auto shared_lock(T&& lc) {
+constexpr auto shared_lock(T&& lc) noexcept {
     return std::shared_lock<std::decay_t<T>> { std::forward<T>(lc) };
 }
 
@@ -110,40 +111,5 @@ constexpr const T& (min)(const T& a, const T& b) {
 
 #endif/*__cplusplus < 201703L*/
 
-template <typename F, typename D>
-constexpr decltype(auto) static_switch(std::size_t /*i*/, std::index_sequence<>, F&& /*f*/, D&& def) {
-    return std::forward<D>(def)();
-}
-
-template <typename F, typename D, std::size_t N, std::size_t...I>
-constexpr decltype(auto) static_switch(std::size_t i, std::index_sequence<N, I...>, F&& f, D&& def) {
-    return (i == N) ? std::forward<F>(f)(std::integral_constant<std::size_t, N>{}) :
-                      static_switch(i, std::index_sequence<I...>{}, std::forward<F>(f), std::forward<D>(def));
-}
-
-template <std::size_t N, typename F, typename D>
-constexpr decltype(auto) static_switch(std::size_t i, F&& f, D&& def) {
-    return static_switch(i, std::make_index_sequence<N>{}, std::forward<F>(f), std::forward<D>(def));
-}
-
-template <typename F, std::size_t...I>
-IPC_CONSTEXPR_ void static_for(std::index_sequence<I...>, F&& f) {
-    IPC_UNUSED_ auto expand = { (std::forward<F>(f)(std::integral_constant<std::size_t, I>{}), 0)... };
-}
-
-template <std::size_t N, typename F>
-IPC_CONSTEXPR_ void static_for(F&& f) {
-    static_for(std::make_index_sequence<N>{}, std::forward<F>(f));
-}
-
-// Minimum offset between two objects to avoid false sharing.
-enum {
-// #if __cplusplus >= 201703L
-//     cache_line_size = std::hardware_destructive_interference_size
-// #else /*__cplusplus < 201703L*/
-    cache_line_size = 64
-// #endif/*__cplusplus < 201703L*/
-};
-
 } // namespace detail
 } // namespace ipc
diff --git a/src/platform/tls_pointer_linux.cpp b/src/platform/tls_pointer_linux.cpp
deleted file mode 100644
index 153a821..0000000
--- a/src/platform/tls_pointer_linux.cpp
+++ /dev/null
@@ -1,29 +0,0 @@
-#include "tls_pointer.h"
-
-#include <pthread.h> // pthread_...
-
-namespace ipc {
-namespace tls {
-
-key_t create(destructor_t destructor) {
-    pthread_key_t k;
-    if (pthread_key_create(&k, destructor) == 0) {
-        return static_cast<key_t>(k);
-    }
-    return invalid_value;
-}
-
-void release(key_t key) {
-    pthread_key_delete(static_cast<pthread_key_t>(key));
-}
-
-bool set(key_t key, void* ptr) {
-    return pthread_setspecific(static_cast<pthread_key_t>(key), ptr) == 0;
-}
-
-void* get(key_t key) {
-    return pthread_getspecific(static_cast<pthread_key_t>(key));
-}
-
-} // namespace tls
-} // namespace ipc
diff --git a/src/platform/tls_pointer_linux.h b/src/platform/tls_pointer_linux.h
new file mode 100755
index 0000000..ffc61e7
--- /dev/null
+++ b/src/platform/tls_pointer_linux.h
@@ -0,0 +1,82 @@
+#include "tls_pointer.h"
+
+#include <pthread.h>    // pthread_...
+
+#include <atomic>       // std::atomic_thread_fence
+#include <cassert>      // assert
+
+#include "log.h"
+#include "utility.h"
+
+namespace ipc {
+namespace tls {
+
+namespace {
+namespace native {
+
+using key_t = pthread_key_t;
+
+bool create(key_t * pkey, void (*destructor)(void*)) {
+    assert(pkey != nullptr);
+    int err = ::pthread_key_create(pkey, destructor);
+    if (err != 0) {
+        ipc::error("[native::create] pthread_key_create failed [%d].\n", err);
+        return false;
+    }
+    return true;
+}
+
+bool release(key_t key) {
+    int err = ::pthread_key_delete(key);
+    if (err != 0) {
+        ipc::error("[native::release] pthread_key_delete failed [%d].\n", err);
+        return false;
+    }
+    return true;
+}
+
+bool set(key_t key, void * ptr) {
+    int err = ::pthread_setspecific(key, ptr);
+    if (err != 0) {
+        ipc::error("[native::set] pthread_setspecific failed [%d].\n", err);
+        return false;
+    }
+    return true;
+}
+
+void * get(key_t key) {
+    return ::pthread_getspecific(key);
+}
+
+} // namespace native
+} // internal-linkage
+
+bool create(key_info * pkey, destructor_t destructor) {
+    assert(pkey != nullptr);
+    native::key_t k;
+    if (!native::create(&k, destructor)) {
+        return false;
+    }
+    pkey->key_ = horrible_cast<key_t>(k);
+    std::atomic_thread_fence(std::memory_order_seq_cst);
+    return true;
+}
+
+void release(key_info const * pkey) {
+    assert(pkey != nullptr);
+    static_cast<void>(
+        native::release(horrible_cast<native::key_t>(pkey->key_)));
+}
+
+bool set(key_info const * pkey, void* ptr) {
+    assert(pkey != nullptr);
+    return native::set(horrible_cast<native::key_t>(pkey->key_), ptr);
+}
+
+void * get(key_info const * pkey) {
+    assert(pkey != nullptr);
+    return native::get(horrible_cast<native::key_t>(pkey->key_));
+}
+
+} // namespace tls
+} // namespace ipc
diff --git a/src/platform/tls_pointer_win.cpp b/src/platform/tls_pointer_win.cpp
old mode 100644
new mode 100755
index 3c4c341..d252d06
--- a/src/platform/tls_pointer_win.cpp
+++ b/src/platform/tls_pointer_win.cpp
@@ -1,18 +1,14 @@
 #include "tls_pointer.h"
 #include "log.h"
 
-#include <Windows.h>     // ::Tls...
-#include <atomic>
+#include <Windows.h>
 
-namespace ipc {
-
-/*
- * <Remarks>
- *
+/**
+ * @remarks
  * Windows doesn't support a per-thread destructor with its TLS primitives.
  * So, here will build it manually by inserting a function to be called on each thread's exit.
  *
- * <Reference>
+ * @see
  * - https://www.codeproject.com/Articles/8113/Thread-Local-Storage-The-C-Way
  * - https://src.chromium.org/viewvc/chrome/trunk/src/base/threading/thread_local_storage_win.cc
  * - https://github.com/mirror/mingw-org-wsl/blob/master/src/libcrt/crt/tlssup.c
@@ -20,186 +16,17 @@ namespace ipc {
  * - http://svn.boost.org/svn/boost/trunk/libs/thread/src/win32/tss_pe.cpp
 */
 
-namespace {
-
-struct tls_data {
-    using destructor_t = void(*)(void*);
-
-    unsigned     index_;
-    DWORD        win_key_;
-    destructor_t destructor_;
-
-    bool valid() const noexcept {
-        return win_key_ != TLS_OUT_OF_INDEXES;
-    }
-
-    void* get() const {
-        return ::TlsGetValue(win_key_);
-    }
-
-    bool set(void* p) {
-        return TRUE == ::TlsSetValue(win_key_, static_cast<LPVOID>(p));
-    }
-
-    void destruct() {
-        void* data = valid() ? get() : nullptr;
-        if (data != nullptr) {
-            if (destructor_ != nullptr) destructor_(data);
-            set(nullptr);
-        }
-    }
-
-    void clear_self() {
-        if (valid()) {
-            destruct();
-            ::TlsFree(win_key_);
-        }
-        delete this;
-    }
-};
-
-struct tls_recs {
-    tls_data* recs_[TLS_MINIMUM_AVAILABLE] {};
-    unsigned index_ = 0;
-
-    bool insert(tls_data* data) noexcept {
-        if (index_ >= TLS_MINIMUM_AVAILABLE) {
-            ipc::error("[tls_recs] insert tls_data failed[index_ >= TLS_MINIMUM_AVAILABLE].\n");
-            return false;
-        }
-        recs_[(data->index_ = index_++)] = data;
-        return true;
-    }
-
-    bool erase(tls_data* data) noexcept {
-        if (data->index_ >= TLS_MINIMUM_AVAILABLE) return false;
-        recs_[data->index_] = nullptr;
-        return true;
-    }
-
-    tls_data*       * begin()       noexcept { return &recs_[0]; }
-    tls_data* const * begin() const noexcept { return &recs_[0]; }
-    tls_data*       * end  ()       noexcept { return &recs_[index_]; }
-    tls_data* const * end  () const noexcept { return &recs_[index_]; }
-};
-
-struct key_gen {
-    DWORD rec_key_;
-
-    key_gen() : rec_key_(::TlsAlloc()) {
-        if (rec_key_ == TLS_OUT_OF_INDEXES) {
-            ipc::error("[record_key] TlsAlloc failed[%lu].\n", ::GetLastError());
-        }
-    }
-
-    ~key_gen() { 
-        ::TlsFree(rec_key_);
-        rec_key_ = TLS_OUT_OF_INDEXES;
-    }
-} gen__;
-
-DWORD& record_key() noexcept {
-    return gen__.rec_key_;
-}
-
-bool record(tls_data* tls_dat) {
-    if (record_key() == TLS_OUT_OF_INDEXES) return false;
-    auto rec = static_cast<tls_recs*>(::TlsGetValue(record_key()));
-    if (rec == nullptr) {
-        if (FALSE == ::TlsSetValue(record_key(), static_cast<LPVOID>(rec = new tls_recs))) {
-            ipc::error("[record] TlsSetValue failed[%lu].\n", ::GetLastError());
-            return false;
-        }
-    }
-    return rec->insert(tls_dat);
-}
-
-void erase_record(tls_data* tls_dat) {
-    if (tls_dat == nullptr) return;
-    if (record_key() == TLS_OUT_OF_INDEXES) return;
-    auto rec = static_cast<tls_recs*>(::TlsGetValue(record_key()));
-    if (rec == nullptr) return;
-    rec->erase(tls_dat);
-    tls_dat->clear_self();
-}
-
-void clear_all_records() {
-    if (record_key() == TLS_OUT_OF_INDEXES) return;
-    auto rec = static_cast<tls_recs*>(::TlsGetValue(record_key()));
-    if (rec == nullptr) return;
-    for (auto tls_dat : *rec) {
-        if (tls_dat != nullptr) tls_dat->destruct();
-    }
-    delete rec;
-    ::TlsSetValue(record_key(), static_cast<LPVOID>(nullptr));
-}
-
-} // internal-linkage
-
+namespace ipc {
 namespace tls {
 
-key_t create(destructor_t destructor) {
-    record_key(); // gen record-key
-    auto tls_dat = new tls_data { unsigned(-1), ::TlsAlloc(), destructor };
-    std::atomic_thread_fence(std::memory_order_seq_cst);
-    if (!tls_dat->valid()) {
-        ipc::error("[tls::create] TlsAlloc failed[%lu].\n", ::GetLastError());
-        tls_dat->clear_self();
-        return invalid_value;
-    }
-    return reinterpret_cast<key_t>(tls_dat);
-}
-
-void release(key_t tls_key) {
-    if (tls_key == invalid_value) {
-        ipc::error("[tls::release] tls_key is invalid_value.\n");
-        return;
-    }
-    auto tls_dat = reinterpret_cast<tls_data*>(tls_key);
-    if (tls_dat == nullptr) {
-        ipc::error("[tls::release] tls_dat is nullptr.\n");
-        return;
-    }
-    erase_record(tls_dat);
-}
-
-bool set(key_t tls_key, void* ptr) {
-    if (tls_key == invalid_value) {
-        ipc::error("[tls::set] tls_key is invalid_value.\n");
-        return false;
-    }
-    auto tls_dat = reinterpret_cast<tls_data*>(tls_key);
-    if (tls_dat == nullptr) {
-        ipc::error("[tls::set] tls_dat is nullptr.\n");
-        return false;
-    }
-    if (!tls_dat->set(ptr)) {
-        ipc::error("[tls::set] TlsSetValue failed[%lu].\n", ::GetLastError());
-        return false;
-    }
-    record(tls_dat);
-    return true;
-}
-
-void* get(key_t tls_key) {
-    if (tls_key == invalid_value) {
-        ipc::error("[tls::get] tls_key is invalid_value.\n");
-        return nullptr;
-    }
-    auto tls_dat = reinterpret_cast<tls_data*>(tls_key);
-    if (tls_dat == nullptr) {
-        ipc::error("[tls::get] tls_dat is nullptr.\n");
-        return nullptr;
-    }
-    return tls_dat->get();
-}
-
-} // namespace tls
+void at_thread_exit();
 
 namespace {
 
 void NTAPI OnTlsCallback(PVOID, DWORD dwReason, PVOID) {
-    if (dwReason == DLL_THREAD_DETACH) clear_all_records();
+    if (dwReason == DLL_THREAD_DETACH) {
+        ipc::tls::at_thread_exit();
+    }
 }
 
 } // internal-linkage
@@ -271,4 +98,5 @@ extern "C" NX_CRTALLOC_(".tls") const IMAGE_TLS_DIRECTORY _tls_used = {
 
 #endif/*_MSC_VER, __GNUC__*/
 
+} // namespace tls
 } // namespace ipc
diff --git a/src/platform/tls_pointer_win.h b/src/platform/tls_pointer_win.h
new file mode 100755
index 0000000..2387042
--- /dev/null
+++ b/src/platform/tls_pointer_win.h
@@ -0,0 +1,76 @@
+#pragma once
+
+#include <unordered_map>    // std::unordered_map
+#include <atomic>           // std::atomic_thread_fence
+#include <cassert>          // assert
+
+#include "log.h"
+#include "utility.h"
+
+namespace ipc {
+namespace tls {
+
+namespace {
+
+inline void tls_destruct(key_info const * pkey, void * p) {
+    assert(pkey != nullptr);
+    auto destructor = horrible_cast<destructor_t>(pkey->key_);
+    if (destructor != nullptr) destructor(p);
+}
+
+struct tls_recs : public std::unordered_map<key_info const *, void *> {
+    ~tls_recs() {
+        for (auto & pair : *this) {
+            tls_destruct(pair.first, pair.second);
+        }
+    }
+};
+
+inline tls_recs * tls_get_recs() {
+    thread_local tls_recs * recs_ptr = nullptr;
+    if (recs_ptr == nullptr) {
+        recs_ptr = new tls_recs;
+    }
+    assert(recs_ptr != nullptr);
+    return recs_ptr;
+}
+
+} // internal-linkage
+
+void at_thread_exit() {
+    delete tls_get_recs();
+}
+
+bool create(key_info * pkey, destructor_t destructor) {
+    assert(pkey != nullptr);
+    pkey->key_ = horrible_cast<key_t>(destructor);
+    std::atomic_thread_fence(std::memory_order_seq_cst);
+    return true;
+}
+
+void release(key_info const * pkey) {
+    assert(pkey != nullptr);
+    assert(tls_get_recs() != nullptr);
+    tls_get_recs()->erase(pkey);
+}
+
+bool set(key_info const * pkey, void * ptr) {
+    assert(pkey != nullptr);
+    assert(tls_get_recs() != nullptr);
+    (*tls_get_recs())[pkey] = ptr;
+    return true;
+}
+
+void * get(key_info const * pkey) {
+    assert(pkey != nullptr);
+    assert(tls_get_recs() != nullptr);
+    auto const recs = tls_get_recs();
+    auto it = recs->find(pkey);
+    if (it == recs->end()) {
+        return nullptr;
+    }
+    return it->second;
+}
+
+} // namespace tls
+} // namespace ipc
diff --git a/src/prod_cons.h b/src/prod_cons.h
old mode 100644
new mode 100755
index 27e45d2..8b9396a
--- a/src/prod_cons.h
+++ b/src/prod_cons.h
@@ -8,6 +8,8 @@
 #include "def.h"
 #include "platform/detail.h"
 #include "circ/elem_def.h"
+#include "log.h"
+#include "utility.h"
 
 namespace ipc {
 
@@ -26,8 +28,8 @@ struct prod_cons_impl<wr<relat::single, relat::single, trans::unicast>> {
         std::aligned_storage_t<DataSize, AlignSize> data_ {};
     };
 
-    alignas(detail::cache_line_size) std::atomic<circ::u2_t> rd_; // read index
-    alignas(detail::cache_line_size) std::atomic<circ::u2_t> wt_; // write index
+    alignas(cache_line_size) std::atomic<circ::u2_t> rd_; // read index
+    alignas(cache_line_size) std::atomic<circ::u2_t> wt_; // write index
 
     constexpr circ::u2_t cursor() const noexcept {
         return 0;
@@ -110,7 +112,7 @@ struct prod_cons_impl<wr<relat::multi , relat::multi, trans::unicast>>
         std::atomic<flag_t> f_ct_ { 0 }; // commit flag
     };
 
-    alignas(detail::cache_line_size) std::atomic<circ::u2_t> ct_; // commit index
+    alignas(cache_line_size) std::atomic<circ::u2_t> ct_; // commit index
 
     template <typename W, typename F, typename E>
     bool push(W* /*wrapper*/, F&& f, E* elems) {
@@ -196,7 +198,7 @@ struct prod_cons_impl<wr<relat::single, relat::multi, trans::broadcast>> {
         std::atomic<rc_t> rc_ { 0 }; // read-counter
     };
 
-    alignas(detail::cache_line_size) std::atomic<circ::u2_t> wt_; // write index
+    alignas(cache_line_size) std::atomic<circ::u2_t> wt_; // write index
 
     circ::u2_t cursor() const noexcept {
         return wt_.load(std::memory_order_acquire);
@@ -290,7 +292,7 @@ struct prod_cons_impl<wr<relat::multi , relat::multi, trans::broadcast>> {
         std::atomic<flag_t> f_ct_ { 0 }; // commit flag
     };
 
-    alignas(detail::cache_line_size) std::atomic<circ::u2_t> ct_; // commit index
+    alignas(cache_line_size) std::atomic<circ::u2_t> ct_; // commit index
 
     circ::u2_t cursor() const noexcept {
         return ct_.load(std::memory_order_acquire);
diff --git a/src/queue.h b/src/queue.h
old mode 100644
new mode 100755
index 3bfa8ac..b0e7700
--- a/src/queue.h
+++ b/src/queue.h
@@ -9,6 +9,7 @@
 #include <thread>
 #include <chrono>
 #include <string>
+#include <cassert>  // assert
 
 #include "def.h"
 #include "shm.h"
@@ -103,6 +104,12 @@ public:
         elems_ = open<elems_t>(name);
     }
 
+    explicit queue_base(elems_t * elems)
+        : queue_base() {
+        assert(elems != nullptr);
+        elems_ = elems;
+    }
+
     /* not virtual */ ~queue_base() {
         base_t::close();
     }
@@ -136,7 +143,7 @@ public:
     }
 
     template <typename T, typename... P>
-    auto push(P&&... params) {
+    bool push(P&&... params) {
         if (elems_ == nullptr) return false;
         return elems_->push(this, [&](void* p) {
             ::new (p) T(std::forward<P>(params)...);
@@ -144,7 +151,7 @@ public:
     }
 
     template <typename T, typename F, typename... P>
-    auto force_push(F&& prep, P&&... params) {
+    bool force_push(F&& prep, P&&... params) {
         if (elems_ == nullptr) return false;
         return elems_->force_push(this, [&](void* p) {
             if (prep(p)) ::new (p) T(std::forward<P>(params)...);
@@ -174,12 +181,12 @@ public:
     using base_t::base_t;
 
     template <typename... P>
-    auto push(P&&... params) {
+    bool push(P&&... params) {
         return base_t::template push<T>(std::forward<P>(params)...);
     }
 
     template <typename... P>
-    auto force_push(P&&... params) {
+    bool force_push(P&&... params) {
         return base_t::template force_push<T>(std::forward<P>(params)...);
     }
 
diff --git a/src/tls_pointer.cpp b/src/tls_pointer.cpp
new file mode 100755
index 0000000..9fd9f2f
--- /dev/null
+++ b/src/tls_pointer.cpp
@@ -0,0 +1,9 @@
+#include "tls_pointer.h"
+
+#if defined(WIN64) || defined(_WIN64) || defined(__WIN64__) || \
+    defined(WIN32) || defined(_WIN32) || defined(__WIN32__) || defined(__NT__) || \
+    defined(WINCE) || defined(_WIN32_WCE)
+#include "platform/tls_pointer_win.h"
+#else /*!WIN*/
+#include "platform/tls_pointer_linux.h"
+#endif/*!WIN*/
diff --git a/src/utility.h b/src/utility.h
new file mode 100755
index 0000000..bd3a6b8
--- /dev/null
+++ b/src/utility.h
@@ -0,0 +1,56 @@
+#pragma once
+
+#include <utility>  // std::forward, std::integer_sequence
+#include <cstddef>  // std::size_t
+#include <new>      // std::hardware_destructive_interference_size
+
+#include "platform/detail.h"
+
+namespace ipc {
+
+template <typename F, typename D>
+constexpr decltype(auto) static_switch(std::size_t /*i*/, std::index_sequence<>, F&& /*f*/, D&& def) {
+    return std::forward<D>(def)();
+}
+
+template <typename F, typename D, std::size_t N, std::size_t...I>
+constexpr decltype(auto) static_switch(std::size_t i, std::index_sequence<N, I...>, F&& f, D&& def) {
+    return (i == N) ? std::forward<F>(f)(std::integral_constant<std::size_t, N>{}) :
+                      static_switch(i, std::index_sequence<I...>{}, std::forward<F>(f), std::forward<D>(def));
+}
+
+template <std::size_t N, typename F, typename D>
+constexpr decltype(auto) static_switch(std::size_t i, F&& f, D&& def) {
+    return static_switch(i, std::make_index_sequence<N>{}, std::forward<F>(f), std::forward<D>(def));
+}
+
+template <typename F, std::size_t...I>
+IPC_CONSTEXPR_ void static_for(std::index_sequence<I...>, F&& f) {
+    IPC_UNUSED_ auto expand = { (std::forward<F>(f)(std::integral_constant<std::size_t, I>{}), 0)... };
+}
+
+template <std::size_t N, typename F>
+IPC_CONSTEXPR_ void static_for(F&& f) {
+    static_for(std::make_index_sequence<N>{}, std::forward<F>(f));
+}
+
+// Minimum offset between two objects to avoid false sharing.
+enum {
+// #if __cplusplus >= 201703L
+//     cache_line_size = std::hardware_destructive_interference_size
+// #else /*__cplusplus < 201703L*/
+    cache_line_size = 64
+// #endif/*__cplusplus < 201703L*/
+};
+
+template <typename T, typename U>
+T horrible_cast(U val) {
+    union {
+        T out;
+        U in;
+    } u;
+    u.in = val;
+    return u.out;
+}
+
+} // namespace ipc
diff --git a/test/CMakeLists.txt b/test/CMakeLists.txt
old mode 100644
new mode 100755
index f7dd881..b863254
--- a/test/CMakeLists.txt
+++ b/test/CMakeLists.txt
@@ -13,8 +13,7 @@ include_directories(
     ${CMAKE_SOURCE_DIR}/src
     ${CMAKE_SOURCE_DIR}/test
     ${CMAKE_SOURCE_DIR}/3rdparty
-    ${CMAKE_SOURCE_DIR}/3rdparty/gtest/include
-)
+    ${CMAKE_SOURCE_DIR}/3rdparty/gtest/include)
 
 file(GLOB SRC_FILES  ${CMAKE_SOURCE_DIR}/test/*.cpp)
 file(GLOB HEAD_FILES ${CMAKE_SOURCE_DIR}/test/*.h)
@@ -24,6 +23,3 @@ add_executable(${PROJECT_NAME} ${SRC_FILES} ${HEAD_FILES})
 link_directories(${CMAKE_SOURCE_DIR}/3rdparty/gperftools)
 target_link_libraries(${PROJECT_NAME} gtest gtest_main ipc)
 #target_link_libraries(${PROJECT_NAME} tcmalloc_minimal)
-if(NOT MSVC)
-  target_link_libraries(${PROJECT_NAME} pthread rt)
-endif()
diff --git a/test/test.h b/test/test.h
index ad59f3d..19bd22b 100755
--- a/test/test.h
+++ b/test/test.h
@@ -8,14 +8,40 @@
 #include <mutex>
 #include <utility>
 
-#if defined(__GNUC__)
-#   include <cxxabi.h>  // abi::__cxa_demangle
-#endif/*__GNUC__*/
-
 #include "gtest/gtest.h"
 
 #include "capo/stopwatch.hpp"
-#include "capo/spin_lock.hpp"
+
+#include "thread_pool.h"
+
+namespace ipc_ut {
+
+template <typename Dur>
+struct unit;
+
+template <> struct unit<std::chrono::nanoseconds> {
+    constexpr static char const * str() noexcept {
+        return "ns";
+    }
+};
+
+template <> struct unit<std::chrono::microseconds> {
+    constexpr static char const * str() noexcept {
+        return "us";
+    }
+};
+
+template <> struct unit<std::chrono::milliseconds> {
+    constexpr static char const * str() noexcept {
+        return "ms";
+    }
+};
+
+template <> struct unit<std::chrono::seconds> {
+    constexpr static char const * str() noexcept {
+        return "sec";
+    }
+};
 
 struct test_stopwatch {
     capo::stopwatch<> sw_;
@@ -27,115 +53,34 @@ struct test_stopwatch {
         }
     }
 
-    template <int Factor, typename ToDur = std::chrono::microseconds>
-    void print_elapsed(int N, int M, int Loops, const char * unit = " us/d") {
+    template <typename ToDur = std::chrono::nanoseconds>
+    void print_elapsed(int N, int Loops, char const * message = "") {
         auto ts = sw_.elapsed<ToDur>();
-        std::cout << "[" << N << ":" << M << ", " << Loops << "] "
-                  << (double(ts) / double(Factor ? (Loops * Factor) : (Loops * N))) << unit << std::endl;
+        std::cout << "[" << N << ", \t" << Loops << "] " << message << "\t"
+                  << (double(ts) / double(Loops)) << " " << unit<ToDur>::str() << std::endl;
     }
 
-    void print_elapsed(int N, int M, int Loops) {
-        print_elapsed<0>(N, M, Loops);
+    template <int Factor, typename ToDur = std::chrono::nanoseconds>
+    void print_elapsed(int N, int M, int Loops, char const * message = "") {
+        auto ts = sw_.elapsed<ToDur>();
+        std::cout << "[" << N << "-" << M << ", \t" << Loops << "] " << message << "\t"
+                  << (double(ts) / double(Factor ? (Loops * Factor) : (Loops * N))) << " " << unit<ToDur>::str() << std::endl;
+    }
+
+    template <typename ToDur = std::chrono::nanoseconds>
+    void print_elapsed(int N, int M, int Loops, char const * message = "") {
+        print_elapsed<0, ToDur>(N, M, Loops, message);
     }
 };
 
-template <typename V>
-struct test_verify;
-
-template <>
-struct test_verify<void> {
-    test_verify   (int)      {}
-    void prepare  (void*)    {}
-    void verify   (int, int) {}
-
-    template <typename U>
-    void push_data(int, U&&) {}
-};
-
-template <typename T>
-struct test_cq;
-
-template <typename T>
-std::string type_name() {
-#if defined(__GNUC__)
-    const char* typeid_name = typeid(T).name();
-    const char* real_name = abi::__cxa_demangle(typeid_name, nullptr, nullptr, nullptr);
-    std::unique_ptr<void, decltype(::free)*> guard { (void*)real_name, ::free };
-    if (real_name == nullptr) real_name = typeid_name;
-    return real_name;
-#else
-    return typeid(T).name();
-#endif/*__GNUC__*/
+inline static thread_pool & sender() {
+    static thread_pool pool;
+    return pool;
 }
 
-template <int N, int M, int Loops, typename V = void, typename T>
-void benchmark_prod_cons(T* cq) {
-    std::cout << "benchmark_prod_cons " << type_name<T>() << " [" << N << ":" << M << ", " << Loops << "]" << std::endl;
-    test_cq<T> tcq { cq };
-
-    std::thread producers[N];
-    std::thread consumers[M];
-    std::atomic_int fini_p { 0 }, fini_c { 0 };
-
-    test_stopwatch sw;
-    test_verify<V> vf { M };
-
-//    capo::spin_lock lc;
-
-    int cid = 0;
-    for (auto& t : consumers) {
-        t = std::thread{[&, cid] {
-            vf.prepare(&t);
-            auto cn = tcq.connect();
-            int i = 0;
-            tcq.recv(cn, [&](auto&& msg) {
-//                if (i % ((Loops * N) / 10) == 0) {
-//                    std::unique_lock<capo::spin_lock> guard { lc };
-//                    std::cout << cid << "-recving: " << (i * 100) / (Loops * N) << "%" << std::endl;
-//                }
-                vf.push_data(cid, std::forward<decltype(msg)>(msg));
-                ++i;
-            });
-//            {
-//                std::unique_lock<capo::spin_lock> guard { lc };
-//                std::cout << cid << "-consumer-disconnect" << std::endl;
-//            }
-            tcq.disconnect(cn);
-            if ((fini_c.fetch_add(1, std::memory_order_relaxed) + 1) != M) {
-//                std::unique_lock<capo::spin_lock> guard { lc };
-//                std::cout << cid << "-consumer-end" << std::endl;
-                return;
-            }
-            sw.print_elapsed(N, M, Loops);
-            vf.verify(N, Loops);
-        }};
-        ++cid;
-    }
-
-    tcq.wait_start(M);
-
-    std::cout << "start producers..." << std::endl;
-    int pid = 0;
-    for (auto& t : producers) {
-        t = std::thread{[&, pid] {
-            auto cn = tcq.connect_send();
-            sw.start();
-            for (int i = 0; i < Loops; ++i) {
-//                if (i % (Loops / 10) == 0) {
-//                    std::unique_lock<capo::spin_lock> guard { lc };
-//                    std::cout << pid << "-sending: " << (i * 100 / Loops) << "%" << std::endl;
-//                }
-                tcq.send(cn, { pid, i });
-            }
-            if ((fini_p.fetch_add(1, std::memory_order_relaxed) + 1) != N) {
-                return;
-            }
-            // quit
-            tcq.send(cn, { -1, -1 });
-            tcq.disconnect(cn);
-        }};
-        ++pid;
-    }
-    for (auto& t : producers) t.join();
-    for (auto& t : consumers) t.join();
+inline static thread_pool & reader() {
+    static thread_pool pool;
+    return pool;
 }
+
+} // namespace ipc_ut
\ No newline at end of file
diff --git a/test/test_circ.cpp b/test/test_circ.cpp
deleted file mode 100755
index 98ab3b8..0000000
--- a/test/test_circ.cpp
+++ /dev/null
@@ -1,385 +0,0 @@
-#include <iostream>
-#include <string>
-#include <type_traits>
-#include <memory>
-#include <new>
-#include <vector>
-#include <unordered_map>
-
-#include "queue.h"
-#include "prod_cons.h"
-#include "policy.h"
-#include "circ/elem_array.h"
-#include "memory/resource.h"
-#include "test.h"
-
-namespace {
-
-struct msg_t {
-    int pid_;
-    int dat_;
-};
-
-template <ipc::relat Rp, ipc::relat Rc, ipc::trans Ts>
-using pc_t = ipc::prod_cons_impl<ipc::wr<Rp, Rc, Ts>>;
-
-template <std::size_t DataSize, typename Policy>
-struct ea_t : public ipc::circ::elem_array<Policy, DataSize, 1> {
-    ea_t() { std::memset(this, 0, sizeof(ipc::circ::elem_array<Policy, DataSize, 1>)); }
-};
-
-using cq_t = ea_t<
-    sizeof(msg_t),
-    pc_t<ipc::relat::single, ipc::relat::multi, ipc::trans::broadcast>
->;
-
-bool operator==(msg_t const & m1, msg_t const & m2) {
-    return (m1.pid_ == m2.pid_) && (m1.dat_ == m2.dat_);
-}
-
-} // internal-linkage
-
-template <std::size_t D, typename P>
-struct test_verify<ea_t<D, P>> {
-    std::vector<std::unordered_map<int, std::vector<int>>> list_;
-
-    test_verify(int M)
-        : list_(static_cast<std::size_t>(M))
-    {}
-
-    void prepare(void* pt) {
-        std::cout << "start consumer: " << pt << std::endl;
-    }
-
-    void push_data(int cid, msg_t const & msg) {
-        list_[cid][msg.pid_].push_back(msg.dat_);
-    }
-
-    void verify(int N, int Loops) {
-        std::cout << "verifying..." << std::endl;
-        for (auto& c_dats : list_) {
-            for (int n = 0; n < N; ++n) {
-                auto& vec = c_dats[n];
-                //for (int d : vec) {
-                //    std::cout << d << " ";
-                //}
-                //std::cout << std::endl;
-                EXPECT_EQ(vec.size(), static_cast<std::size_t>(Loops));
-                int i = 0;
-                for (int d : vec) {
-                    EXPECT_EQ(i, d);
-                    ++i;
-                }
-            }
-        }
-    }
-};
-
-template <ipc::relat Rp>
-struct test_verify<pc_t<Rp, ipc::relat::multi, ipc::trans::unicast>> : test_verify<cq_t> {
-    using test_verify<cq_t>::test_verify;
-
-    void verify(int N, int Loops) {
-        std::cout << "verifying..." << std::endl;
-        for (int n = 0; n < N; ++n) {
-            std::vector<int> datas;
-            std::uint64_t sum = 0;
-            for (auto& c_dats : list_) {
-                for (int d : c_dats[n]) {
-                    datas.push_back(d);
-                    sum += d;
-                }
-            }
-            EXPECT_EQ(datas.size(), static_cast<std::size_t>(Loops));
-            EXPECT_EQ(sum, (Loops * std::uint64_t(Loops - 1)) / 2);
-        }
-    }
-};
-
-template <typename P>
-struct quit_mode;
-
-template <ipc::relat Rp, ipc::relat Rc>
-struct quit_mode<pc_t<Rp, Rc, ipc::trans::unicast>> {
-    using type = volatile bool;
-};
-
-template <ipc::relat Rp, ipc::relat Rc>
-struct quit_mode<pc_t<Rp, Rc, ipc::trans::broadcast>> {
-    struct type {
-        constexpr type(bool) {}
-        constexpr operator bool() const { return false; }
-    };
-};
-
-template <std::size_t D, typename P>
-struct test_cq<ea_t<D, P>> {
-    using ca_t = ea_t<D, P>;
-    using cn_t = decltype(std::declval<ca_t>().connect());
-    using cr_t = decltype(std::declval<ca_t>().cursor());
-
-    typename quit_mode<P>::type quit_ = false;
-    ca_t* ca_;
-    std::unordered_map<cn_t, cr_t> conns_;
-
-    test_cq(ca_t* ca) : ca_(ca) {}
-
-    cn_t connect() {
-        auto it = conns_.emplace(ca_->connect(), ca_->cursor());
-        return it.first->first;
-    }
-
-    ca_t* connect_send() {
-        return ca_;
-    }
-
-    void disconnect(cn_t cc_id) {
-        ca_->disconnect(cc_id);
-    }
-
-    void disconnect(ca_t*) {}
-
-    ca_t       * elems()       noexcept { return ca_; }
-    ca_t const * elems() const noexcept { return ca_; }
-
-    void wait_start(int M) {
-        while (ca_->conn_count() != static_cast<std::size_t>(M)) {
-            std::this_thread::yield();
-        }
-    }
-
-    template <typename F>
-    void recv(cn_t cn, F&& proc) {
-
-        struct que {
-            cn_t cn_;
-            // for ca_->pop
-            cn_t connected_id() const noexcept { return cn_; }
-        } q { cn };
-
-        cr_t& cur = conns_[cn];
-        while (1) {
-            msg_t msg;
-            while (ca_->pop(&q, &cur, [&msg](void* p) {
-                msg = *static_cast<msg_t*>(p);
-            })) {
-                if (msg.pid_ < 0) {
-                    quit_ = true;
-                    return;
-                }
-                proc(msg);
-            }
-            if (quit_) return;
-            std::this_thread::yield();
-        }
-    }
-
-    void send(ca_t* ca, msg_t const & msg) {
-        while (!ca->push(this, [&msg](void* p) {
-            (*static_cast<msg_t*>(p)) = msg;
-        })) {
-            std::this_thread::yield();
-        }
-    }
-};
-
-template <typename... T>
-struct test_cq<ipc::queue<T...>> {
-    using cn_t = ipc::queue<T...>;
-
-    test_cq(void*) {}
-
-    cn_t* connect() {
-        cn_t* queue = new cn_t { "test-ipc-queue" };
-        [&] { EXPECT_TRUE(queue->connect()); } ();
-        return queue;
-    }
-
-    void disconnect(cn_t* queue) {
-        queue->disconnect();
-        delete queue;
-    }
-
-    void wait_start(int M) {
-        cn_t que("test-ipc-queue");
-        while (que.conn_count() != static_cast<std::size_t>(M)) {
-            std::this_thread::yield();
-        }
-    }
-
-    template <typename F>
-    void recv(cn_t* queue, F&& proc) {
-        while(1) {
-            typename cn_t::value_t msg;
-            while (!queue->pop(msg)) {
-                std::this_thread::yield();
-            }
-            if (msg.pid_ < 0) return;
-            proc(msg);
-        }
-    }
-
-    cn_t* connect_send() {
-        return new cn_t { "test-ipc-queue" };
-    }
-
-    void send(cn_t* cn, msg_t const & msg) {
-        while (!cn->push(msg)) {
-            std::this_thread::yield();
-        }
-    }
-};
-
-namespace {
-
-constexpr int LoopCount = 1000000;
-//constexpr int LoopCount = 1000/*0000*/;
-
-TEST(Circ, init) {
-    std::cout << "cq_t::head_size  = " << cq_t::head_size << std::endl;
-    std::cout << "cq_t::data_size  = " << cq_t::data_size << std::endl;
-    std::cout << "cq_t::elem_size  = " << cq_t::elem_size << std::endl;
-    std::cout << "cq_t::block_size = " << cq_t::block_size << std::endl;
-
-    EXPECT_EQ(static_cast<std::size_t>(cq_t::data_size), sizeof(msg_t));
-
-    std::cout << "sizeof(ea_t<sizeof(msg_t)>) = " << sizeof(cq_t) << std::endl;
-}
-
-TEST(Circ, prod_cons_1v1) {
-    ea_t<
-        sizeof(msg_t),
-        pc_t<ipc::relat::single, ipc::relat::single, ipc::trans::unicast>
-    > el_arr_ssu;
-    benchmark_prod_cons<1, 1, LoopCount, cq_t>(&el_arr_ssu);
-    benchmark_prod_cons<1, 1, LoopCount, void>(&el_arr_ssu);
-
-    ea_t<
-        sizeof(msg_t),
-        pc_t<ipc::relat::single, ipc::relat::multi, ipc::trans::unicast>
-    > el_arr_smu;
-    benchmark_prod_cons<1, 1, LoopCount, decltype(el_arr_smu)::policy_t>(&el_arr_smu);
-    benchmark_prod_cons<1, 1, LoopCount, void>(&el_arr_smu);
-
-    ea_t<
-        sizeof(msg_t),
-        pc_t<ipc::relat::multi, ipc::relat::multi, ipc::trans::unicast>
-    > el_arr_mmu;
-    benchmark_prod_cons<1, 1, LoopCount, decltype(el_arr_mmu)::policy_t>(&el_arr_mmu);
-    benchmark_prod_cons<1, 1, LoopCount, void>(&el_arr_mmu);
-
-    ea_t<
-        sizeof(msg_t),
-        pc_t<ipc::relat::single, ipc::relat::multi, ipc::trans::broadcast>
-    > el_arr_smb;
-    benchmark_prod_cons<1, 1, LoopCount, cq_t>(&el_arr_smb);
-    benchmark_prod_cons<1, 1, LoopCount, void>(&el_arr_smb);
-
-    ea_t<
-        sizeof(msg_t),
-        pc_t<ipc::relat::multi, ipc::relat::multi, ipc::trans::broadcast>
-    > el_arr_mmb;
-    benchmark_prod_cons<1, 1, LoopCount, cq_t>(&el_arr_mmb);
-    benchmark_prod_cons<1, 1, LoopCount, void>(&el_arr_mmb);
-    benchmark_prod_cons<2, 1, LoopCount, void>(&el_arr_mmb);
-}
-
-TEST(Circ, prod_cons_1v3) {
-    ea_t<
-        sizeof(msg_t),
-        pc_t<ipc::relat::single, ipc::relat::multi, ipc::trans::unicast>
-    > el_arr_smu;
-    benchmark_prod_cons<1, 3, LoopCount, decltype(el_arr_smu)::policy_t>(&el_arr_smu);
-    benchmark_prod_cons<1, 3, LoopCount, void>(&el_arr_smu);
-
-    ea_t<
-        sizeof(msg_t),
-        pc_t<ipc::relat::multi, ipc::relat::multi, ipc::trans::unicast>
-    > el_arr_mmu;
-    benchmark_prod_cons<1, 3, LoopCount, decltype(el_arr_mmu)::policy_t>(&el_arr_mmu);
-    benchmark_prod_cons<1, 3, LoopCount, void>(&el_arr_mmu);
-
-    ea_t<
-        sizeof(msg_t),
-        pc_t<ipc::relat::single, ipc::relat::multi, ipc::trans::broadcast>
-    > el_arr_smb;
-    benchmark_prod_cons<1, 3, LoopCount, cq_t>(&el_arr_smb);
-    benchmark_prod_cons<1, 3, LoopCount, void>(&el_arr_smb);
-
-    ea_t<
-        sizeof(msg_t),
-        pc_t<ipc::relat::multi, ipc::relat::multi, ipc::trans::broadcast>
-    > el_arr_mmb;
-    benchmark_prod_cons<1, 3, LoopCount, cq_t>(&el_arr_mmb);
-    benchmark_prod_cons<1, 3, LoopCount, void>(&el_arr_mmb);
-}
-
-TEST(Circ, prod_cons_performance) {
-    ea_t<
-        sizeof(msg_t),
-        pc_t<ipc::relat::single, ipc::relat::multi, ipc::trans::unicast>
-    > el_arr_smu;
-    ipc::detail::static_for<8>([&el_arr_smu](auto index) {
-        benchmark_prod_cons<1, decltype(index)::value + 1, LoopCount, void>(&el_arr_smu);
-    });
-
-    ea_t<
-        sizeof(msg_t),
-        pc_t<ipc::relat::single, ipc::relat::multi, ipc::trans::broadcast>
-    > el_arr_smb;
-    ipc::detail::static_for<8>([&el_arr_smb](auto index) {
-        benchmark_prod_cons<1, decltype(index)::value + 1, LoopCount, void>(&el_arr_smb);
-    });
-    // test & verify
-    ipc::detail::static_for<8>([&el_arr_smb](auto index) {
-        benchmark_prod_cons<1, decltype(index)::value + 1, LoopCount, cq_t>(&el_arr_smb);
-    });
-
-    ea_t<
-        sizeof(msg_t),
-        pc_t<ipc::relat::multi, ipc::relat::multi, ipc::trans::unicast>
-    > el_arr_mmu;
-    ipc::detail::static_for<8>([&el_arr_mmu](auto index) {
-        benchmark_prod_cons<1, decltype(index)::value + 1, LoopCount, void>(&el_arr_mmu);
-    });
-    ipc::detail::static_for<8>([&el_arr_mmu](auto index) {
-        benchmark_prod_cons<decltype(index)::value + 1, 1, LoopCount, void>(&el_arr_mmu);
-    });
-    ipc::detail::static_for<8>([&el_arr_mmu](auto index) {
-        benchmark_prod_cons<decltype(index)::value + 1, decltype(index)::value + 1, LoopCount, void>(&el_arr_mmu);
-    });
-
-    ea_t<
-        sizeof(msg_t),
-        pc_t<ipc::relat::multi, ipc::relat::multi, ipc::trans::broadcast>
-    > el_arr_mmb;
-    ipc::detail::static_for<8>([&el_arr_mmb](auto index) {
-        benchmark_prod_cons<1, decltype(index)::value + 1, LoopCount, void>(&el_arr_mmb);
-    });
-    ipc::detail::static_for<8>([&el_arr_mmb](auto index) {
-        benchmark_prod_cons<decltype(index)::value + 1, 1, LoopCount, void>(&el_arr_mmb);
-    });
-    ipc::detail::static_for<8>([&el_arr_mmb](auto index) {
-        benchmark_prod_cons<decltype(index)::value + 1, decltype(index)::value + 1, LoopCount, void>(&el_arr_mmb);
-    });
-}
-
-TEST(Circ, queue) {
-    using queue_t = ipc::queue<msg_t, ipc::policy::choose<
-            ipc::circ::elem_array,
-            ipc::wr<ipc::relat::single, ipc::relat::multi, ipc::trans::broadcast>
-    >>;
-    queue_t queue;
-
-    EXPECT_TRUE(!queue.push(msg_t { 1, 2 }));
-    msg_t msg {};
-    EXPECT_TRUE(!queue.pop(msg));
-    EXPECT_EQ(msg, (msg_t {}));
-    EXPECT_TRUE(sizeof(decltype(queue)::elems_t) <= sizeof(cq_t));
-
-    ipc::detail::static_for<16>([](auto index) {
-        benchmark_prod_cons<1, decltype(index)::value + 1, LoopCount>((queue_t*)nullptr);
-    });
-}
-
-} // internal-linkage
diff --git a/test/test_ipc.cpp b/test/test_ipc.cpp
index 98490be..a9148ec 100755
--- a/test/test_ipc.cpp
+++ b/test/test_ipc.cpp
@@ -1,472 +1,150 @@
-#include <thread>
-#include <vector>
-#include <type_traits>
-#include <iostream>
-#include <shared_mutex>
-#include <mutex>
-#include <typeinfo>
-#include <memory>
-#include <string>
-#include <cstring>
-#include <algorithm>
-#include <array>
-#include <limits>
-#include <utility>
 
-#include "capo/stopwatch.hpp"
-#include "capo/spin_lock.hpp"
-#include "capo/random.hpp"
+#include <vector>
+#include <iostream>
+#include <cstring>
 
 #include "ipc.h"
-#include "rw_lock.h"
+#include "buffer.h"
 #include "memory/resource.h"
 
 #include "test.h"
+#include "thread_pool.h"
+
+#include "capo/random.hpp"
+
+using namespace ipc;
 
 namespace {
 
-std::vector<ipc::buff_t> datas__;
-
-constexpr int DataMin   = 2;
-constexpr int DataMax   = 256;
-constexpr int LoopCount = 100000;
-// constexpr int LoopCount = 1000;
-
-} // internal-linkage
-
-template <typename T>
-struct test_verify {
-    std::vector<std::vector<ipc::buff_t>> list_;
-
-    test_verify(int M)
-        : list_(static_cast<std::size_t>(M))
-    {}
-
-    void prepare(void* /*pt*/) {}
-
-    void push_data(int cid, ipc::buff_t && msg) {
-        list_[cid].emplace_back(std::move(msg));
-    }
-
-    void verify(int /*N*/, int /*Loops*/) {
-        std::cout << "verifying..." << std::endl;
-        for (auto& c_dats : list_) {
-            EXPECT_EQ(datas__.size(), c_dats.size());
-            std::size_t i = 0;
-            for (auto& d : c_dats) {
-                EXPECT_EQ(datas__[i++], d);
-            }
-        }
-    }
-};
-
-template <>
-struct test_cq<ipc::route> {
-    using cn_t = ipc::route;
-
-    std::string conn_name_;
-
-    test_cq(void*)
-        : conn_name_("test-ipc-route") {
-    }
-
-    cn_t connect() {
-        return cn_t { conn_name_.c_str() };
-    }
-
-    void disconnect(cn_t& cn) {
-        cn.disconnect();
-    }
-
-    void wait_start(int M) {
-        cn_t::wait_for_recv(conn_name_.c_str(), static_cast<std::size_t>(M));
-    }
-
-    template <typename F>
-    void recv(cn_t& cn, F&& proc) {
-        do {
-            auto msg = cn.recv();
-            if (msg.size() < 2) {
-                EXPECT_EQ(msg, ipc::buff_t('\0'));
-                return;
-            }
-            proc(std::move(msg));
-        } while(1);
-    }
-
-    cn_t connect_send() {
-        return connect();
-    }
-
-    void send(cn_t& cn, const std::array<int, 2>& info) {
-        int n = info[1];
-        if (n < 0) {
-            /*EXPECT_TRUE*/(cn.send(ipc::buff_t('\0')));
-        }
-        else /*EXPECT_TRUE*/(cn.send(datas__[static_cast<decltype(datas__)::size_type>(n)]));
-    }
-};
-
-template <>
-struct test_cq<ipc::channel> {
-    using cn_t = ipc::channel;
-
-    std::string conn_name_;
-    int m_ = 0;
-
-    test_cq(void*)
-        : conn_name_("test-ipc-channel") {
-    }
-
-    cn_t connect() {
-        return cn_t { conn_name_.c_str() };
-    }
-
-    void disconnect(cn_t& cn) {
-        cn.disconnect();
-    }
-
-    void wait_start(int M) { m_ = M; }
-
-    template <typename F>
-    void recv(cn_t& cn, F&& proc) {
-        do {
-            auto msg = cn.recv();
-            if (msg.size() < 2) {
-                EXPECT_EQ(msg, ipc::buff_t('\0'));
-                return;
-            }
-            proc(std::move(msg));
-        } while(1);
-    }
-
-    cn_t connect_send() {
-        return connect();
-    }
-
-    void send(cn_t& cn, const std::array<int, 2>& info) {
-        thread_local struct s_dummy {
-            s_dummy(cn_t& cn, int m) {
-                cn.wait_for_recv(static_cast<std::size_t>(m));
-//                std::printf("start to send: %d.\n", m);
-            }
-        } _(cn, m_);
-        int n = info[1];
-        if (n < 0) {
-            /*EXPECT_TRUE*/(cn.send(ipc::buff_t('\0')));
-        }
-        else /*EXPECT_TRUE*/(cn.send(datas__[static_cast<decltype(datas__)::size_type>(n)]));
-    }
-};
-
-namespace {
-
-struct Init {
-    Init() {
-        capo::random<> rdm { DataMin, DataMax };
-        capo::random<> bit { 0, (std::numeric_limits<ipc::byte_t>::max)() };
-
-        for (int i = 0; i < LoopCount; ++i) {
-            std::size_t n = static_cast<std::size_t>(rdm());
-            ipc::buff_t buff {
-                new ipc::byte_t[n], n,
-                [](void* p, std::size_t) {
-                    delete [] static_cast<ipc::byte_t*>(p);
-                }
-            };
-            for (std::size_t k = 0; k < buff.size(); ++k) {
-                static_cast<ipc::byte_t*>(buff.data())[k] = static_cast<ipc::byte_t>(bit());
-            }
-            datas__.emplace_back(std::move(buff));
-        }
-    }
-} init__;
-
-template <typename T>
-constexpr T acc(T b, T e) noexcept {
-    return (e + b) * (e - b + 1) / 2;
-}
-
-template <typename Mutex>
-struct lc_wrapper : Mutex {
-    void lock_shared  () { Mutex::lock  (); }
-    void unlock_shared() { Mutex::unlock(); }
-};
-
-template <typename Lc, int W, int R, int Loops = LoopCount>
-void benchmark_lc() {
-    std::thread w_trd[W];
-    std::thread r_trd[R];
-    std::atomic_int fini { 0 };
-//    std::atomic_bool wf { false };
-
-    std::vector<int> datas;
-    Lc lc;
-
-    test_stopwatch sw;
-    std::cout << std::endl << type_name<Lc>() << std::endl;
-
-    for (auto& t : r_trd) {
-        t = std::thread([&] {
-            std::vector<int> seq;
-            std::size_t cnt = 0;
-            while (1) {
-                int x = -1;
-                {
-                    std::shared_lock<Lc> guard { lc };
-//                    EXPECT_TRUE(!wf);
-                    if (cnt < datas.size()) {
-                        x = datas[cnt];
-                    }
-//                    std::this_thread::sleep_for(std::chrono::milliseconds(1));
-                    if (x ==  0) break; // quit
-                    if (x != -1) {
-                        seq.push_back(x);
-                        ++cnt;
-                    }
-                }
-                std::this_thread::yield();
-            }
-            if ((fini.fetch_add(1, std::memory_order_relaxed) + 1) == R) {
-                sw.print_elapsed(W, R, Loops);
-            }
-            std::uint64_t sum = 0;
-            for (int i : seq) sum += static_cast<std::uint64_t>(i);
-            EXPECT_EQ(sum, acc<std::uint64_t>(1, Loops) * W);
+class rand_buf : public buffer {
+public:
+    rand_buf() {
+        int size = capo::random{1, 65536}();
+        *this = buffer(new char[size], size, [](void * p, std::size_t) {
+            delete [] static_cast<char *>(p);
         });
     }
 
-    for (auto& t : w_trd) {
-        t = std::thread([&] {
+    rand_buf(rand_buf &&) = default;
+    rand_buf(rand_buf const & rhs) {
+        if (rhs.empty()) return;
+        void * mem = new char[rhs.size()];
+        std::memcpy(mem, rhs.data(), rhs.size());
+        *this = buffer(mem, rhs.size(), [](void * p, std::size_t) {
+            delete [] static_cast<char *>(p);
+        });
+    }
+
+    rand_buf(buffer && rhs)
+        : buffer(std::move(rhs)) {
+    }
+
+    void set_id(int k) noexcept {
+        *get<char *>() = static_cast<char>(k);
+    }
+
+    int get_id() const noexcept {
+        return static_cast<int>(*get<char *>());
+    }
+
+    using buffer::operator=;
+};
+
+template <relat Rp, relat Rc, trans Ts>
+void test_basic(char const * name) {
+    using que_t = chan<wr<Rp, Rc, Ts>>;
+    rand_buf test1, test2;
+
+    que_t que1 { name };
+    EXPECT_FALSE(que1.send(test1));
+    EXPECT_FALSE(que1.try_send(test2));
+
+    que_t que2 { que1.name(), ipc::receiver };
+    EXPECT_TRUE(que1.send(test1));
+    EXPECT_TRUE(que1.try_send(test2));
+
+    EXPECT_EQ(que2.recv(), test1);
+    EXPECT_EQ(que2.recv(), test2);
+}
+
+template <relat Rp, relat Rc, trans Ts>
+void test_sr(char const * name, int size, int s_cnt, int r_cnt) {
+    using que_t = chan<wr<Rp, Rc, Ts>>;
+
+    ipc_ut::sender().start(static_cast<std::size_t>(s_cnt));
+    ipc_ut::reader().start(static_cast<std::size_t>(r_cnt));
+    ipc_ut::test_stopwatch sw;
+    std::vector<rand_buf> tests(size);
+
+    for (int k = 0; k < s_cnt; ++k) {
+        ipc_ut::sender() << [name, &tests, &sw, r_cnt, k] {
+            que_t que1 { name };
+            EXPECT_TRUE(que1.wait_for_recv(r_cnt));
             sw.start();
-            for (int i = 1; i <= Loops; ++i) {
-                {
-                    std::unique_lock<Lc> guard { lc };
-//                    wf = true;
-                    datas.push_back(i);
-//                    std::this_thread::sleep_for(std::chrono::milliseconds(1));
-//                    wf = false;
-                }
-                std::this_thread::yield();
+            for (auto & buf : tests) {
+                rand_buf data { buf };
+                data.set_id(k);
+                EXPECT_TRUE(que1.send(data));
             }
-        });
+        };
     }
 
-    for (auto& t : w_trd) t.join();
-    lc.lock();
-    datas.push_back(0);
-    lc.unlock();
-    for (auto& t : r_trd) t.join();
-}
-
-template <int W, int R>
-void test_lock_performance() {
-
-    std::cout << std::endl
-              << "test_lock_performance: [" << W << ":" << R << "]"
-              << std::endl;
-
-    benchmark_lc<ipc::rw_lock               , W, R>();
-    benchmark_lc<lc_wrapper< ipc::spin_lock>, W, R>();
-    benchmark_lc<lc_wrapper<capo::spin_lock>, W, R>();
-    benchmark_lc<lc_wrapper<std::mutex>     , W, R>();
-    benchmark_lc<std::shared_timed_mutex    , W, R>();
-}
-
-TEST(IPC, rw_lock) {
-//    test_lock_performance<1, 1>();
-//    test_lock_performance<4, 4>();
-//    test_lock_performance<1, 8>();
-//    test_lock_performance<8, 1>();
-}
-
-template <typename T, int N, int M, bool V = true, int Loops = LoopCount>
-void test_prod_cons() {
-    benchmark_prod_cons<N, M, Loops, std::conditional_t<V, T, void>>((T*)nullptr);
-}
-
-TEST(IPC, route) {
-    // return;
-    std::vector<char const *> const datas = {
-        "hello!",
-        "foo",
-        "bar",
-        "ISO/IEC",
-        "14882:2011",
-        "ISO/IEC 14882:2017 Information technology - Programming languages - C++",
-        "ISO/IEC 14882:2020",
-        "Modern C++ Design: Generic Programming and Design Patterns Applied"
-    };
-
-    std::thread t1 {[&] {
-        ipc::route cc { "my-ipc-route" };
-        for (std::size_t i = 0; i < datas.size(); ++i) {
-            ipc::buff_t dd = cc.recv();
-            std::cout << "recv: " << (char*)dd.data() << std::endl;
-            EXPECT_EQ(dd.size(), std::strlen(datas[i]) + 1);
-            EXPECT_TRUE(std::memcmp(dd.data(), datas[i], dd.size()) == 0);
-        }
-    }};
-
-    std::thread t2 {[&] {
-        ipc::route cc { "my-ipc-route" };
-        while (cc.recv_count() == 0) {
-            std::this_thread::yield();
-        }
-        for (std::size_t i = 0; i < datas.size(); ++i) {
-            std::cout << "sending: " << datas[i] << std::endl;
-            EXPECT_TRUE(cc.send(datas[i]));
-        }
-    }};
-
-    t1.join();
-    t2.join();
-
-    test_prod_cons<ipc::route, 1, 1>(); // test & verify
-}
-
-TEST(IPC, route_rtt) {
-    // return;
-    test_stopwatch sw;
-
-    std::thread t1 {[&] {
-        ipc::route cc { "my-ipc-route-1" };
-        ipc::route cr { "my-ipc-route-2" };
-        for (std::size_t i = 0;; ++i) {
-            auto dd = cc.recv();
-            if (dd.size() < 2) return;
-            //std::cout << "recv: " << i << "-[" << dd.size() << "]" << std::endl;
-            while (!cr.send(ipc::buff_t('a'))) {
-                std::this_thread::yield();
+    for (int k = 0; k < r_cnt; ++k) {
+        ipc_ut::reader() << [name, &tests, s_cnt] {
+            que_t que2 { name, ipc::receiver };
+            std::vector<int> cursors(s_cnt);
+            for (;;) {
+                rand_buf got { que2.recv() };
+                ASSERT_FALSE(got.empty());
+                int & cur = cursors.at(got.get_id());
+                ASSERT_TRUE((cur >= 0) && (cur < static_cast<int>(tests.size())));
+                rand_buf buf { tests.at(cur++) };
+                buf.set_id(got.get_id());
+                EXPECT_EQ(got, buf);
+                int n = 0;
+                for (; n < static_cast<int>(cursors.size()); ++n) {
+                    if (cursors[n] < static_cast<int>(tests.size())) break;
+                }
+                if (n == static_cast<int>(cursors.size())) break;
             }
-        }
-    }};
+        };
+    }
 
-    std::thread t2 {[&] {
-        ipc::route cc { "my-ipc-route-1" };
-        ipc::route cr { "my-ipc-route-2" };
-        while (cc.recv_count() == 0) {
-            std::this_thread::yield();
-        }
-        sw.start();
-        for (std::size_t i = 0; i < LoopCount; ++i) {
-            cc.send(datas__[i]);
-            //std::cout << "sent: " << i << "-[" << datas__[i].size() << "]" << std::endl;
-            /*auto dd = */cr.recv();
-//            if (dd.size() != 1 || dd[0] != 'a') {
-//                EXPECT_TRUE(false);
-//            }
-        }
-        cc.send(ipc::buff_t('\0'));
-        t1.join();
-        sw.print_elapsed(1, 1, LoopCount);
-    }};
-
-    t2.join();
+    ipc_ut::sender().wait_for_done();
+    ipc_ut::reader().wait_for_done();
+    sw.print_elapsed<std::chrono::microseconds>(s_cnt, r_cnt, size, name);
 }
 
-TEST(IPC, route_performance) {
-    // return;
-    ipc::detail::static_for<8>([](auto index) {
-        test_prod_cons<ipc::route, 1, decltype(index)::value + 1, false>();
-    });
-    // test_prod_cons<ipc::route, 1, 8>(); // test & verify
-}
-
-TEST(IPC, channel) {
-    // return;
-    int fail_v = 0;
-
-    std::thread t1 {[&] {
-        ipc::channel cc { "my-ipc-channel" };
-        for (std::size_t i = 0;; ++i) {
-            ipc::buff_t dd = cc.recv();
-            if (dd.size() < 2) return;
-            if (dd != datas__[i]) {
-                std::printf("fail recv: %zd-[%zd, %zd]\n", i, dd.size(), datas__[i].size());
-                // for (std::size_t k = 0; k < dd.size(); ++k) {
-                //     if (dd.data<ipc::byte_t>()[k] != datas__[i].data<ipc::byte_t>()[k]) {
-                //         std::printf("fail check: %zd-%zd, %02x != %02x\n", 
-                //                     i, k, (unsigned)dd        .data<ipc::byte_t>()[k], 
-                //                           (unsigned)datas__[i].data<ipc::byte_t>()[k]);
-                //     }
-                // }
-                ++fail_v;
-            }
-        }
-    }};
-
-    std::thread t2 {[&] {
-        ipc::channel cc { "my-ipc-channel" };
-        cc.wait_for_recv(1);
-        for (std::size_t i = 0; i < static_cast<std::size_t>((std::min)(100, LoopCount)); ++i) {
-            std::printf("sending: %zd-[%zd]\n", i, datas__[i].size());
-            cc.send(datas__[i]);
-        }
-        cc.send(ipc::buff_t('\0'));
-        t1.join();
-    }};
-
-    t2.join();
-
-    EXPECT_EQ(fail_v, 0);
-}
-
-TEST(IPC, channel_rtt) {
-    // return;
-    test_stopwatch sw;
-
-    std::thread t1 {[&] {
-        ipc::channel cc { "my-ipc-channel" };
-        bool recv_2 = false;
-        for (std::size_t i = 0;; ++i) {
-            auto dd = cc.recv();
-            if (dd.size() < 2) return;
-            //if (i % 1000 == 0) {
-            //    std::cout << "recv: " << i << "-[" << dd.size() << "]" << std::endl;
-            //}
-            while (!recv_2) {
-                recv_2 = cc.wait_for_recv(2);
-            }
-            cc.send(ipc::buff_t('a'));
-        }
-    }};
-
-    std::thread t2 {[&] {
-        ipc::channel cc { "my-ipc-channel" };
-        cc.wait_for_recv(1);
-        sw.start();
-        for (std::size_t i = 0; i < LoopCount; ++i) {
-            //if (i % 1000 == 0) {
-            //    std::cout << "send: " << i << "-[" << datas__[i].size() << "]" << std::endl;
-            //}
-            cc.send(datas__[i]);
-            /*auto dd = */cc.recv();
-            //if (dd.size() != 1 || dd.data<char>()[0] != 'a') {
-            //    std::cout << "recv ack fail: " << i << "-[" << dd.size() << "]" << std::endl;
-            //    EXPECT_TRUE(false);
-            //}
-        }
-        cc.send(ipc::buff_t('\0'));
-        t1.join();
-        sw.print_elapsed(1, 1, LoopCount);
-    }};
-
-    t2.join();
-}
-
-TEST(IPC, channel_performance) {
-    // return;
-    ipc::detail::static_for<8>([](auto index) {
-        test_prod_cons<ipc::channel, 1, decltype(index)::value + 1, false>();
-    });
-    ipc::detail::static_for<8>([](auto index) {
-        test_prod_cons<ipc::channel, decltype(index)::value + 1, 1, false>();
-    });
-    ipc::detail::static_for<8>([](auto index) {
-        test_prod_cons<ipc::channel, decltype(index)::value + 1,
-                                     decltype(index)::value + 1, false>();
-    });
-}
+constexpr int LoopCount = 10000;
+constexpr int MultiMax  = 8;
 
 } // internal-linkage
+
+TEST(IPC, basic) {
+    test_basic<relat::single, relat::single, trans::unicast  >("ssu");
+    test_basic<relat::single, relat::multi , trans::unicast  >("smu");
+    test_basic<relat::multi , relat::multi , trans::unicast  >("mmu");
+    test_basic<relat::single, relat::multi , trans::broadcast>("smb");
+    test_basic<relat::multi , relat::multi , trans::broadcast>("mmb");
+}
+
+TEST(IPC, 1v1) {
+    test_sr<relat::single, relat::single, trans::unicast  >("ssu", LoopCount, 1, 1);
+    test_sr<relat::single, relat::multi , trans::unicast  >("smu", LoopCount, 1, 1);
+    test_sr<relat::multi , relat::multi , trans::unicast  >("mmu", LoopCount, 1, 1);
+    test_sr<relat::single, relat::multi , trans::broadcast>("smb", LoopCount, 1, 1);
+    test_sr<relat::multi , relat::multi , trans::broadcast>("mmb", LoopCount, 1, 1);
+}
+
+TEST(IPC, 1vN) {
+    test_sr<relat::single, relat::multi , trans::broadcast>("smb", LoopCount, 1, MultiMax);
+    test_sr<relat::multi , relat::multi , trans::broadcast>("mmb", LoopCount, 1, MultiMax);
+}
+
+TEST(IPC, Nv1) {
+    test_sr<relat::multi , relat::multi , trans::broadcast>("mmb", LoopCount, MultiMax, 1);
+}
+
+TEST(IPC, NvN) {
+    test_sr<relat::multi , relat::multi , trans::broadcast>("mmb", LoopCount, MultiMax, MultiMax);
+}
diff --git a/test/test_mem.cpp b/test/test_mem.cpp
index 1db02db..e1ea259 100755
--- a/test/test_mem.cpp
+++ b/test/test_mem.cpp
@@ -12,18 +12,21 @@
 // #include "gperftools/tcmalloc.h"
 
 #include "test.h"
+#include "thread_pool.h"
 
 namespace {
 
 constexpr int DataMin   = 4;
 constexpr int DataMax   = 256;
-constexpr int LoopCount = 4194304;
+constexpr int LoopCount = 8388608;
+constexpr int ThreadMax = 16;
 
 // constexpr int DataMin   = 256;
 // constexpr int DataMax   = 512;
 // constexpr int LoopCount = 2097152;
 
 std::vector<std::size_t> sizes__;
+std::vector<void*> ptr_cache__[ThreadMax];
 
 template <typename M>
 struct alloc_ix_t {
@@ -33,10 +36,11 @@ struct alloc_ix_t {
     alloc_ix_t() {
         if (inited_) return;
         inited_ = true;
-        M::init(ix_);
+        M::init();
     }
 
-    int index(std::size_t /*pid*/, std::size_t /*k*/, std::size_t n) {
+    template <int ThreadsN>
+    static int index(std::size_t /*pid*/, std::size_t /*k*/, std::size_t n) {
         return ix_[n];
     }
 };
@@ -46,38 +50,36 @@ std::vector<int> alloc_ix_t<M>::ix_(LoopCount);
 template <typename M>
 bool alloc_ix_t<M>::inited_ = false;
 
-template <std::size_t N>
-struct alloc_FIFO : alloc_ix_t<alloc_FIFO<N>> {
-    static void init(std::vector<int>& ix) {
+struct alloc_FIFO : alloc_ix_t<alloc_FIFO> {
+    static void init() {
         for (int i = 0; i < LoopCount; ++i) {
-            ix[static_cast<std::size_t>(i)] = i;
+            ix_[static_cast<std::size_t>(i)] = i;
         }
     }
 };
 
-template <std::size_t N>
-struct alloc_LIFO : alloc_ix_t<alloc_LIFO<N>> {
-    static void init(std::vector<int>& ix) {
+struct alloc_LIFO : alloc_ix_t<alloc_LIFO> {
+    static void init() {
         for (int i = 0; i < LoopCount; ++i) {
-            ix[static_cast<std::size_t>(i)] = i;
+            ix_[static_cast<std::size_t>(i)] = i;
         }
     }
 
-    int index(std::size_t pid, std::size_t k, std::size_t n) {
-        constexpr static int CacheSize = LoopCount / N;
+    template <int ThreadsN>
+    static int index(std::size_t pid, std::size_t k, std::size_t n) {
+        constexpr static int CacheSize = LoopCount / ThreadsN;
         if (k) {
-            return this->ix_[(CacheSize * (2 * pid + 1)) - 1 - n];
+            return ix_[(CacheSize * (2 * pid + 1)) - 1 - n];
         }
-        else return this->ix_[n];
+        else return ix_[n];
     }
 };
 
-template <std::size_t N>
-struct alloc_random : alloc_ix_t<alloc_random<N>> {
-    static void init(std::vector<int>& ix) {
+struct alloc_Random : alloc_ix_t<alloc_Random> {
+    static void init() {
         capo::random<> rdm_index(0, LoopCount - 1);
         for (int i = 0; i < LoopCount; ++i) {
-            ix[static_cast<std::size_t>(i)] = rdm_index();
+            ix_[static_cast<std::size_t>(i)] = rdm_index();
         }
     }
 };
@@ -88,69 +90,50 @@ struct Init {
         for (int i = 0; i < LoopCount; ++i) {
             sizes__.emplace_back(static_cast<std::size_t>(rdm()));
         }
+        for (auto& vec : ptr_cache__) {
+            vec.resize(LoopCount, nullptr);
+        }
     }
 } init__;
 
 template <typename AllocT, int ThreadsN>
-void benchmark_alloc() {
-    std::cout << std::endl
-              << "[Threads: " << ThreadsN << "] "
-              << type_name<AllocT>() << std::endl;
+void benchmark_alloc(char const * message) {
+    std::string msg = std::to_string(ThreadsN) + "\t" + message;
 
     constexpr static int CacheSize = LoopCount / ThreadsN;
+    ipc_ut::sender().start(static_cast<std::size_t>(ThreadsN));
+    ipc_ut::test_stopwatch sw;
 
-    std::atomic_int fini { 0 };
-    test_stopwatch sw;
-
-    std::thread works[ThreadsN];
-    int pid = 0;
-
-    for (auto& w : works) {
-        w = std::thread {[&, pid] {
+    for (int pid = 0; pid < ThreadsN; ++pid) {
+        ipc_ut::sender() << [&, pid] {
             sw.start();
-            for (std::size_t k = 0; k < 100; ++k)
             for (int n = (CacheSize * pid); n < (CacheSize * (pid + 1)); ++n) {
                 std::size_t s = sizes__[n];
                 AllocT::free(AllocT::alloc(s), s);
             }
-            if ((fini.fetch_add(1, std::memory_order_relaxed) + 1) == ThreadsN) {
-                sw.print_elapsed<1, std::chrono::nanoseconds>(DataMin, DataMax, LoopCount * 100, " ns/d");
-            }
-        }};
-        ++pid;
+        };
     }
 
-    for (auto& w : works) w.join();
+    ipc_ut::sender().wait_for_done();
+    sw.print_elapsed<1>(DataMin, DataMax, LoopCount, msg.c_str());
 }
 
-template <typename AllocT, template <std::size_t> class ModeT, int ThreadsN>
-void benchmark_alloc() {
-    std::cout << std::endl
-              << "[Threads: " << ThreadsN << ", Mode: " << type_name<ModeT<ThreadsN>>() << "] "
-              << type_name<AllocT>() << std::endl;
+template <typename AllocT, typename ModeT, int ThreadsN>
+void benchmark_alloc(char const * message) {
+    std::string msg = std::to_string(ThreadsN) + "\t" + message;
 
     constexpr static int CacheSize = LoopCount / ThreadsN;
+    ModeT mode;
+    ipc_ut::sender().start(static_cast<std::size_t>(ThreadsN));
+    ipc_ut::test_stopwatch sw;
 
-    std::vector<void*> ptrs[ThreadsN];
-    for (auto& vec : ptrs) {
-        vec.resize(LoopCount);
-    }
-
-    ModeT<ThreadsN> mode;
-
-    std::atomic_int fini { 0 };
-    test_stopwatch  sw;
-
-    std::thread works[ThreadsN];
-    int pid = 0;
-
-    for (auto& w : works) {
-        w = std::thread {[&, pid] {
-            auto& vec = ptrs[pid];
+    for (int pid = 0; pid < ThreadsN; ++pid) {
+        ipc_ut::sender() << [&, pid] {
+            auto& vec = ptr_cache__[pid];
             sw.start();
             for (std::size_t k = 0; k < 2; ++k)
             for (int n = (CacheSize * pid); n < (CacheSize * (pid + 1)); ++n) {
-                int    m = mode.index(pid, k, n);
+                int    m = mode.template index<ThreadsN>(pid, k, n);
                 void*& p = vec[static_cast<std::size_t>(m)];
                 std::size_t s  = sizes__[static_cast<std::size_t>(m)];
                 if (p == nullptr) {
@@ -161,45 +144,40 @@ void benchmark_alloc() {
                     p = nullptr;
                 }
             }
-            if ((fini.fetch_add(1, std::memory_order_relaxed) + 1) == ThreadsN) {
-                sw.print_elapsed<1>(DataMin, DataMax, LoopCount);
-            }
-        }};
-        ++pid;
+        };
     }
 
-    for (auto& w : works) w.join();
+    ipc_ut::sender().wait_for_done();
+    sw.print_elapsed<1>(DataMin, DataMax, LoopCount, msg.c_str());
 }
 
-template <typename AllocT, template <std::size_t> class ModeT, int ThreadsN>
+template <typename AllocT, typename ModeT, int ThreadsN>
 struct test_performance {
-    static void start() {
-        test_performance<AllocT, ModeT, ThreadsN / 2>::start();
-        benchmark_alloc<AllocT, ModeT, ThreadsN>();
+    static void start(char const * message) {
+        test_performance<AllocT, ModeT, ThreadsN / 2>::start(message);
+        benchmark_alloc<AllocT, ModeT, ThreadsN>(message);
     }
 };
 
-template <typename AllocT, template <std::size_t> class ModeT>
+template <typename AllocT, typename ModeT>
 struct test_performance<AllocT, ModeT, 1> {
-    static void start() {
-        benchmark_alloc<AllocT, ModeT, 1>();
+    static void start(char const * message) {
+        benchmark_alloc<AllocT, ModeT, 1>(message);
     }
 };
 
-template <std::size_t> struct dummy;
-
 template <typename AllocT, int ThreadsN>
-struct test_performance<AllocT, dummy, ThreadsN> {
-    static void start() {
-        test_performance<AllocT, dummy, ThreadsN / 2>::start();
-        benchmark_alloc<AllocT, ThreadsN>();
+struct test_performance<AllocT, void, ThreadsN> {
+    static void start(char const * message) {
+        test_performance<AllocT, void, ThreadsN / 2>::start(message);
+        benchmark_alloc<AllocT, ThreadsN>(message);
     }
 };
 
 template <typename AllocT>
-struct test_performance<AllocT, dummy, 1> {
-    static void start() {
-        benchmark_alloc<AllocT, 1>();
+struct test_performance<AllocT, void, 1> {
+    static void start(char const * message) {
+        benchmark_alloc<AllocT, 1>(message);
     }
 };
 
@@ -217,27 +195,24 @@ struct test_performance<AllocT, dummy, 1> {
 // };
 
 TEST(Memory, static_alloc) {
-    // test_performance<ipc::mem::static_alloc, dummy       , 128>::start();
-    // test_performance<ipc::mem::static_alloc, alloc_FIFO  , 128>::start();
-    // test_performance<ipc::mem::static_alloc, alloc_LIFO  , 128>::start();
-    // test_performance<ipc::mem::static_alloc, alloc_random, 128>::start();
+    test_performance<ipc::mem::static_alloc, void        , ThreadMax>::start("alloc-free");
+    test_performance<ipc::mem::static_alloc, alloc_FIFO  , ThreadMax>::start("alloc-FIFO");
+    test_performance<ipc::mem::static_alloc, alloc_LIFO  , ThreadMax>::start("alloc-LIFO");
+    test_performance<ipc::mem::static_alloc, alloc_Random, ThreadMax>::start("alloc-Rand");
 }
 
 TEST(Memory, pool_alloc) {
-    //test_performance<ipc::mem::async_pool_alloc, dummy       , 128>::start();
-    //test_performance<ipc::mem::async_pool_alloc, alloc_FIFO  , 128>::start();
-
-    //test_performance<ipc::mem::async_pool_alloc, dummy       , 128>::start();
-    //test_performance<ipc::mem::async_pool_alloc, alloc_FIFO  , 128>::start();
-    //test_performance<ipc::mem::async_pool_alloc, alloc_LIFO  , 128>::start();
-    //test_performance<ipc::mem::async_pool_alloc, alloc_random, 128>::start();
+    test_performance<ipc::mem::async_pool_alloc, void        , ThreadMax>::start("alloc-free");
+    test_performance<ipc::mem::async_pool_alloc, alloc_FIFO  , ThreadMax>::start("alloc-FIFO");
+    test_performance<ipc::mem::async_pool_alloc, alloc_LIFO  , ThreadMax>::start("alloc-LIFO");
+    test_performance<ipc::mem::async_pool_alloc, alloc_Random, ThreadMax>::start("alloc-Rand");
 }
 
-TEST(Memory, tc_alloc) {
-    // test_performance<tc_alloc, dummy       , 128>::start();
-    // test_performance<tc_alloc, alloc_FIFO  , 128>::start();
-    // test_performance<tc_alloc, alloc_LIFO  , 128>::start();
-    // test_performance<tc_alloc, alloc_random, 128>::start();
-}
+// TEST(Memory, tc_alloc) {
+//     test_performance<tc_alloc, void       , ThreadMax>::start();
+//     test_performance<tc_alloc, alloc_FIFO  , ThreadMax>::start();
+//     test_performance<tc_alloc, alloc_LIFO  , ThreadMax>::start();
+//     test_performance<tc_alloc, alloc_Random, ThreadMax>::start();
+// }
 
 } // internal-linkage
diff --git a/test/test_queue.cpp b/test/test_queue.cpp
new file mode 100755
index 0000000..93decb0
--- /dev/null
+++ b/test/test_queue.cpp
@@ -0,0 +1,181 @@
+#include <iostream>
+#include <string>
+#include <type_traits>
+#include <memory>
+#include <new>
+#include <vector>
+#include <unordered_map>
+
+#include "prod_cons.h"
+#include "policy.h"
+#include "circ/elem_array.h"
+#include "queue.h"
+
+#include "test.h"
+
+namespace {
+
+struct msg_t {
+    int pid_;
+    int dat_;
+
+    msg_t() = default;
+    msg_t(int p, int d) : pid_(p), dat_(d) {}
+};
+
+template <ipc::relat Rp, ipc::relat Rc, ipc::trans Ts>
+using queue_t = ipc::queue<msg_t, ipc::policy::choose<ipc::circ::elem_array, ipc::wr<Rp, Rc, Ts>>>;
+
+template <ipc::relat Rp, ipc::relat Rc, ipc::trans Ts>
+struct elems_t : public queue_t<Rp, Rc, Ts>::elems_t {};
+
+bool operator==(msg_t const & m1, msg_t const & m2) noexcept {
+    return (m1.pid_ == m2.pid_) && (m1.dat_ == m2.dat_);
+}
+
+bool operator!=(msg_t const & m1, msg_t const & m2) noexcept {
+    return !(m1 == m2);
+}
+
+constexpr int LoopCount = 1000000;
+constexpr int PushRetry = 1000000;
+constexpr int ThreadMax = 8;
+
+template <typename Que>
+void push(Que & que, int p, int d) {
+    for (int n = 0; !que.push(p, d); ++n) {
+        ASSERT_NE(n, PushRetry);
+        std::this_thread::yield();
+    }
+}
+
+template <typename Que>
+msg_t pop(Que & que) {
+    msg_t msg;
+    while (!que.pop(msg)) {
+        std::this_thread::yield();
+    }
+    return msg;
+}
+
+template <ipc::trans Ts>
+struct quitter;
+
+template <>
+struct quitter<ipc::trans::unicast> {
+    template <typename Que>
+    static void emit(Que && que, int r_cnt) {
+        for (int k = 0; k < r_cnt; ++k) {
+            push(que, -1, -1);
+        }
+    }
+};
+
+template <>
+struct quitter<ipc::trans::broadcast> {
+    template <typename Que>
+    static void emit(Que && que, int /*r_cnt*/) {
+        push(que, -1, -1);
+    }
+};
+
+template <ipc::relat Rp, ipc::relat Rc, ipc::trans Ts>
+void test_sr(elems_t<Rp, Rc, Ts> && elems, int s_cnt, int r_cnt, char const * message) {
+    ipc_ut::sender().start(static_cast<std::size_t>(s_cnt));
+    ipc_ut::reader().start(static_cast<std::size_t>(r_cnt));
+    ipc_ut::test_stopwatch sw;
+
+    for (int k = 0; k < s_cnt; ++k) {
+        ipc_ut::sender() << [&elems, &sw, r_cnt, k] {
+            queue_t<Rp, Rc, Ts> que { &elems };
+            while (que.conn_count() != static_cast<std::size_t>(r_cnt)) {
+                std::this_thread::yield();
+            }
+            sw.start();
+            for (int i = 0; i < LoopCount; ++i) {
+                push(que, k, i);
+            }
+        };
+    }
+    for (int k = 0; k < r_cnt; ++k) {
+        ipc_ut::reader() << [&elems, k] {
+            queue_t<Rp, Rc, Ts> que { &elems };
+            ASSERT_TRUE(que.connect());
+            while (pop(que).pid_ >= 0) ;
+            EXPECT_TRUE(que.disconnect());
+        };
+    }
+
+    ipc_ut::sender().wait_for_done();
+    quitter<Ts>::emit(queue_t<Rp, Rc, Ts> { &elems }, r_cnt);
+    ipc_ut::reader().wait_for_done();
+    sw.print_elapsed(s_cnt, r_cnt, LoopCount, message);
+}
+
+} // internal-linkage
+
+TEST(Queue, check_size) {
+    using el_t = elems_t<ipc::relat::single, ipc::relat::multi, ipc::trans::broadcast>;
+
+    std::cout << "cq_t::head_size  = " << el_t::head_size << std::endl;
+    std::cout << "cq_t::data_size  = " << el_t::data_size << std::endl;
+    std::cout << "cq_t::elem_size  = " << el_t::elem_size << std::endl;
+    std::cout << "cq_t::block_size = " << el_t::block_size << std::endl;
+
+    EXPECT_EQ(static_cast<std::size_t>(el_t::data_size), sizeof(msg_t));
+
+    std::cout << "sizeof(elems_t<s, m, b>) = " << sizeof(el_t) << std::endl;
+}
+
+TEST(Queue, prod_cons_1v1_unicast) {
+    test_sr(elems_t<ipc::relat::single, ipc::relat::single, ipc::trans::unicast> {}, 1, 1, "ssu");
+    test_sr(elems_t<ipc::relat::single, ipc::relat::multi , ipc::trans::unicast> {}, 1, 1, "smu");
+    test_sr(elems_t<ipc::relat::multi , ipc::relat::multi , ipc::trans::unicast> {}, 1, 1, "mmu");
+}
+
+TEST(Queue, prod_cons_1v1_broadcast) {
+    test_sr(elems_t<ipc::relat::single, ipc::relat::multi , ipc::trans::broadcast> {}, 1, 1, "smb");
+    test_sr(elems_t<ipc::relat::multi , ipc::relat::multi , ipc::trans::broadcast> {}, 1, 1, "mmb");
+}
+
+TEST(Queue, prod_cons_1vN_unicast) {
+    for (int i = 1; i <= ThreadMax; ++i) {
+        test_sr(elems_t<ipc::relat::single, ipc::relat::multi , ipc::trans::unicast> {}, 1, i, "smu");
+    }
+    for (int i = 1; i <= ThreadMax; ++i) {
+        test_sr(elems_t<ipc::relat::multi , ipc::relat::multi , ipc::trans::unicast> {}, 1, i, "mmu");
+    }
+}
+
+TEST(Queue, prod_cons_1vN_broadcast) {
+    for (int i = 1; i <= ThreadMax; ++i) {
+        test_sr(elems_t<ipc::relat::single, ipc::relat::multi , ipc::trans::broadcast> {}, 1, i, "smb");
+    }
+    for (int i = 1; i <= ThreadMax; ++i) {
+        test_sr(elems_t<ipc::relat::multi , ipc::relat::multi , ipc::trans::broadcast> {}, 1, i, "mmb");
+    }
+}
+
+TEST(Queue, prod_cons_NvN_unicast) {
+    for (int i = 1; i <= ThreadMax; ++i) {
+        test_sr(elems_t<ipc::relat::multi , ipc::relat::multi , ipc::trans::unicast> {}, 1, i, "mmu");
+    }
+    for (int i = 1; i <= ThreadMax; ++i) {
+        test_sr(elems_t<ipc::relat::multi , ipc::relat::multi , ipc::trans::unicast> {}, i, 1, "mmu");
+    }
+    for (int i = 1; i <= ThreadMax; ++i) {
+        test_sr(elems_t<ipc::relat::multi , ipc::relat::multi , ipc::trans::unicast> {}, i, i, "mmu");
+    }
+}
+
+TEST(Queue, prod_cons_NvN_broadcast) {
+    for (int i = 1; i <= ThreadMax; ++i) {
+        test_sr(elems_t<ipc::relat::multi , ipc::relat::multi , ipc::trans::broadcast> {}, 1, i, "mmb");
+    }
+    for (int i = 1; i <= ThreadMax; ++i) {
+        test_sr(elems_t<ipc::relat::multi , ipc::relat::multi , ipc::trans::broadcast> {}, i, 1, "mmb");
+    }
+    for (int i = 1; i <= ThreadMax; ++i) {
+        test_sr(elems_t<ipc::relat::multi , ipc::relat::multi , ipc::trans::broadcast> {}, i, i, "mmb");
+    }
+}
diff --git a/test/test_shm.cpp b/test/test_shm.cpp
index cf83162..67a224a 100755
--- a/test/test_shm.cpp
+++ b/test/test_shm.cpp
@@ -9,83 +9,94 @@ using namespace ipc::shm;
 
 namespace {
 
-handle shm_hd__;
-
 TEST(SHM, acquire) {
-    EXPECT_TRUE(!shm_hd__.valid());
+    handle shm_hd;
+    EXPECT_FALSE(shm_hd.valid());
 
-    EXPECT_TRUE(shm_hd__.acquire("my-test-1", 1024));
-    EXPECT_TRUE(shm_hd__.valid());
-    EXPECT_STREQ(shm_hd__.name(), "my-test-1");
+    EXPECT_TRUE(shm_hd.acquire("my-test-1", 1024));
+    EXPECT_TRUE(shm_hd.valid());
+    EXPECT_STREQ(shm_hd.name(), "my-test-1");
 
-    EXPECT_TRUE(shm_hd__.acquire("my-test-2", 2048));
-    EXPECT_TRUE(shm_hd__.valid());
-    EXPECT_STREQ(shm_hd__.name(), "my-test-2");
+    EXPECT_TRUE(shm_hd.acquire("my-test-2", 2048));
+    EXPECT_TRUE(shm_hd.valid());
+    EXPECT_STREQ(shm_hd.name(), "my-test-2");
 
-    EXPECT_TRUE(shm_hd__.acquire("my-test-3", 4096));
-    EXPECT_TRUE(shm_hd__.valid());
-    EXPECT_STREQ(shm_hd__.name(), "my-test-3");
+    EXPECT_TRUE(shm_hd.acquire("my-test-3", 4096));
+    EXPECT_TRUE(shm_hd.valid());
+    EXPECT_STREQ(shm_hd.name(), "my-test-3");
 }
 
 TEST(SHM, release) {
-    EXPECT_TRUE(shm_hd__.valid());
-    shm_hd__.release();
-    EXPECT_TRUE(!shm_hd__.valid());
+    handle shm_hd;
+    EXPECT_FALSE(shm_hd.valid());
+    shm_hd.release();
+    EXPECT_FALSE(shm_hd.valid());
+    EXPECT_TRUE(shm_hd.acquire("release-test-1", 512));
+    EXPECT_TRUE(shm_hd.valid());
+    shm_hd.release();
+    EXPECT_FALSE(shm_hd.valid());
 }
 
 TEST(SHM, get) {
-    EXPECT_TRUE(shm_hd__.get() == nullptr);
-    EXPECT_TRUE(shm_hd__.acquire("my-test", 1024));
+    handle shm_hd;
+    EXPECT_TRUE(shm_hd.get() == nullptr);
+    EXPECT_TRUE(shm_hd.acquire("get-test", 2048));
 
-    auto mem = shm_hd__.get();
+    auto mem = shm_hd.get();
     EXPECT_TRUE(mem != nullptr);
-    EXPECT_TRUE(mem == shm_hd__.get());
+    EXPECT_TRUE(mem == shm_hd.get());
 
     std::uint8_t buf[1024] = {};
     EXPECT_TRUE(memcmp(mem, buf, sizeof(buf)) == 0);
 
-    handle shm_other(shm_hd__.name(), shm_hd__.size());
-    EXPECT_TRUE(shm_other.get() != shm_hd__.get());
+    handle shm_other(shm_hd.name(), shm_hd.size());
+    EXPECT_TRUE(shm_other.get() != shm_hd.get());
 }
 
 TEST(SHM, hello) {
-    auto mem = shm_hd__.get();
+    handle shm_hd;
+    EXPECT_TRUE(shm_hd.acquire("hello-test", 128));
+    auto mem = shm_hd.get();
     EXPECT_TRUE(mem != nullptr);
 
     constexpr char hello[] = "hello!";
     std::memcpy(mem, hello, sizeof(hello));
-    EXPECT_STREQ((char const *)shm_hd__.get(), hello);
+    EXPECT_STREQ((char const *)shm_hd.get(), hello);
 
-    shm_hd__.release();
-    EXPECT_TRUE(shm_hd__.get() == nullptr);
-    EXPECT_TRUE(shm_hd__.acquire("my-test", 1024));
+    shm_hd.release();
+    EXPECT_TRUE(shm_hd.get() == nullptr);
+    EXPECT_TRUE(shm_hd.acquire("hello-test", 1024));
 
-    mem = shm_hd__.get();
+    mem = shm_hd.get();
     EXPECT_TRUE(mem != nullptr);
     std::uint8_t buf[1024] = {};
     EXPECT_TRUE(memcmp(mem, buf, sizeof(buf)) == 0);
 
     std::memcpy(mem, hello, sizeof(hello));
-    EXPECT_STREQ((char const *)shm_hd__.get(), hello);
+    EXPECT_STREQ((char const *)shm_hd.get(), hello);
 }
 
 TEST(SHM, mt) {
+    handle shm_hd;
+    EXPECT_TRUE(shm_hd.acquire("mt-test", 256));
+    constexpr char hello[] = "hello!";
+    std::memcpy(shm_hd.get(), hello, sizeof(hello));
+
     std::thread {
-        [] {
-            handle shm_mt(shm_hd__.name(), shm_hd__.size());
-
-            shm_hd__.release();
-
+        [&shm_hd] {
+            handle shm_mt(shm_hd.name(), shm_hd.size());
+            shm_hd.release();
             constexpr char hello[] = "hello!";
             EXPECT_STREQ((char const *)shm_mt.get(), hello);
         }
     }.join();
-    EXPECT_TRUE(shm_hd__.get() == nullptr);
-    EXPECT_TRUE(!shm_hd__.valid());
 
-    EXPECT_TRUE(shm_hd__.acquire("my-test", 1024));
+    EXPECT_TRUE(shm_hd.get() == nullptr);
+    EXPECT_FALSE(shm_hd.valid());
+
+    EXPECT_TRUE(shm_hd.acquire("mt-test", 1024));
     std::uint8_t buf[1024] = {};
-    EXPECT_TRUE(memcmp(shm_hd__.get(), buf, sizeof(buf)) == 0);
+    EXPECT_TRUE(memcmp(shm_hd.get(), buf, sizeof(buf)) == 0);
 }
 
 } // internal-linkage
diff --git a/test/test_thread_utility.cpp b/test/test_thread_utility.cpp
new file mode 100755
index 0000000..abda198
--- /dev/null
+++ b/test/test_thread_utility.cpp
@@ -0,0 +1,202 @@
+#include <type_traits>
+#include <iostream>
+#include <shared_mutex> // std::shared_lock
+#include <utility>
+#include <thread>
+#include <future>       // std::async
+#include <atomic>
+#include <string>       // std::string
+
+#include "capo/spin_lock.hpp"
+#include "capo/type_name.hpp"
+
+#include "rw_lock.h"
+#include "tls_pointer.h"
+
+#include "test.h"
+#include "thread_pool.h"
+
+namespace {
+
+constexpr int LoopCount = 100000;
+constexpr int ThreadMax = 8;
+
+template <typename T>
+constexpr T acc(T b, T e) noexcept {
+    return (e + b) * (e - b + 1) / 2;
+}
+
+template <typename Mutex>
+struct lc_wrapper : Mutex {
+    void lock_shared  () { Mutex::lock  (); }
+    void unlock_shared() { Mutex::unlock(); }
+};
+
+template <typename Lc, int Loops = LoopCount>
+void benchmark_lc(int w, int r, char const * message) {
+    ipc_ut::sender().start(static_cast<std::size_t>(w));
+    ipc_ut::reader().start(static_cast<std::size_t>(r));
+    ipc_ut::test_stopwatch sw;
+
+    std::uint64_t data = 0;
+    std::uint64_t sum  = acc<std::uint64_t>(1, Loops) * w;
+    Lc lc;
+
+    for (int k = 0; k < r; ++k) {
+        ipc_ut::reader() << [sum, &lc, &data] {
+            while (1) {
+                {
+                    std::shared_lock<Lc> guard { lc };
+                    if (data == sum) break;
+                }
+                std::this_thread::yield();
+            }
+        };
+    }
+
+    for (int k = 0; k < w; ++k) {
+        ipc_ut::sender() << [&sw, &lc, &data] {
+            sw.start();
+            for (int i = 1; i <= Loops; ++i) {
+                {
+                    std::lock_guard<Lc> guard { lc };
+                    data += i;
+                }
+                std::this_thread::yield();
+            }
+        };
+    }
+
+    ipc_ut::sender().wait_for_done();
+    sw.print_elapsed(w, r, Loops, message);
+    ipc_ut::reader().wait_for_done();
+}
+
+void test_lock_performance(int w, int r) {
+    std::cout << "test_lock_performance: [" << w << "-" << r << "]" << std::endl;
+    benchmark_lc<ipc::rw_lock               >(w, r, "ipc::rw_lock");
+    benchmark_lc<lc_wrapper< ipc::spin_lock>>(w, r, "ipc::spin_lock");
+    benchmark_lc<lc_wrapper<capo::spin_lock>>(w, r, "capo::spin_lock");
+    benchmark_lc<lc_wrapper<std::mutex>     >(w, r, "std::mutex");
+    // benchmark_lc<std::shared_mutex          >(w, r, "std::shared_mutex");
+}
+
+} // internal-linkage
+
+//TEST(Thread, rw_lock) {
+//    for (int i = 1; i <= ThreadMax; ++i) test_lock_performance(i, 0);
+//    for (int i = 1; i <= ThreadMax; ++i) test_lock_performance(1, i);
+//    for (int i = 2; i <= ThreadMax; ++i) test_lock_performance(i, i);
+//}
+
+TEST(Thread, tls_main_thread) {
+    ipc::tls::pointer<int> p;
+    EXPECT_FALSE(p);
+    EXPECT_NE(p.create(123), nullptr);
+    EXPECT_EQ(*p, 123);
+}
+
+namespace {
+
+struct Foo {
+    std::atomic<int> * px_;
+    Foo(std::atomic<int> * px) : px_(px) {}
+    ~Foo() {
+        px_->fetch_add(1, std::memory_order_relaxed);
+    }
+};
+
+} // namespace
+
+TEST(Thread, tls_create_once) {
+    std::atomic<int> x { 0 };
+    Foo { &x };
+    EXPECT_EQ(x, 1);
+    {
+        ipc::tls::pointer<Foo> foo;
+        EXPECT_NE(foo.create(&x), nullptr);
+        EXPECT_EQ(x, 1);
+
+        std::thread {[&foo, &x] {
+            auto foo1 = foo.create_once(&x);
+            auto foo2 = foo.create_once(&x);
+            EXPECT_EQ(foo1, foo2);
+        }}.join();
+        EXPECT_EQ(x, 2);
+    }
+    EXPECT_EQ(x, 3);
+}
+
+TEST(Thread, tls_multi_thread) {
+    std::atomic<int> x { 0 };
+    {
+        ipc::tls::pointer<Foo> foo; // no create
+        EXPECT_EQ(x, 0);
+
+        ipc_ut::thread_pool pool { 10 };
+        for (int i = 0; i < 1000; ++i) {
+          pool << [&foo, &x] {
+            // foo.create_once(&x);
+            std::this_thread::sleep_for(std::chrono::milliseconds(1));
+          };
+        }
+        pool.wait_for_done();
+        EXPECT_EQ(x, 0); // thread_pool hasn't destructed.
+    }
+    // EXPECT_EQ(x, 10);
+}
+
+namespace {
+
+template <typename Tls, typename T>
+void benchmark_tls(char const * message) {
+    ipc_ut::thread_pool pool { static_cast<std::size_t>(ThreadMax) };
+    ipc_ut::test_stopwatch sw;
+    pool.wait_for_started();
+
+    for (int k = 0; k < ThreadMax; ++k) {
+        pool << [&sw] {
+            sw.start();
+            for (int i = 0; i < LoopCount * 10; ++i) {
+                *Tls::template get<T>() = i;
+            }
+        };
+    }
+
+    pool.wait_for_done();
+    sw.print_elapsed(ThreadMax, LoopCount * 10, message);
+}
+
+struct ipc_tls {
+    template <typename T>
+    static T * get() {
+        static ipc::tls::pointer<T> p;
+        return p.create_once();
+    }
+};
+
+struct std_tls {
+    template <typename T>
+    static T * get() {
+        thread_local T p;
+        return &p;
+    }
+};
+
+struct Str {
+    std::string str_;
+
+    Str & operator=(int i) {
+        str_ = std::to_string(i);
+        return *this;
+    }
+};
+
+} // internal-linkage
+
+TEST(Thread, tls_benchmark) {
+    benchmark_tls<std_tls, int>("std_tls: int");
+    benchmark_tls<ipc_tls, int>("ipc_tls: int");
+    benchmark_tls<std_tls, Str>("std_tls: Str");
+    benchmark_tls<ipc_tls, Str>("ipc_tls: Str");
+}
\ No newline at end of file
diff --git a/test/thread_pool.h b/test/thread_pool.h
new file mode 100755
index 0000000..a7d2b28
--- /dev/null
+++ b/test/thread_pool.h
@@ -0,0 +1,119 @@
+#pragma once
+
+#include <thread>             // std::thread
+#include <mutex>              // std::mutex
+#include <condition_variable> // std::condition_variable
+#include <deque>              // std::deque
+#include <functional>         // std::function
+#include <utility>            // std::forward, std::move
+#include <cstddef>            // std::size_t
+#include <cassert>            // assert
+
+namespace ipc_ut {
+
+class thread_pool final {
+
+  std::deque<std::thread> workers_;
+  std::deque<std::function<void()>> jobs_;
+
+  std::mutex lock_;
+  std::condition_variable cv_jobs_;
+  std::condition_variable cv_empty_;
+
+  std::size_t waiting_cnt_ = 0;
+  bool quit_ = false;
+
+  static void proc(thread_pool * pool) {
+    assert(pool != nullptr);
+    std::function<void()> job;
+    for (;;) {
+      {
+        std::unique_lock<std::mutex> guard { pool->lock_ };
+        if (pool->quit_) return;
+        if (pool->jobs_.empty()) {
+          pool->waiting_cnt_ += 1;
+
+          if (pool->waiting_cnt_ == pool->workers_.size()) {
+            pool->cv_empty_.notify_all();
+          }
+          assert(pool->waiting_cnt_ <= pool->workers_.size());
+          do {
+            pool->cv_jobs_.wait(guard);
+            if (pool->quit_) return;
+          } while (pool->jobs_.empty());
+
+          pool->waiting_cnt_ -= 1;
+        }
+        assert(!pool->jobs_.empty());
+        job = std::move(pool->jobs_.front());
+        pool->jobs_.pop_front();
+      }
+      if (job) job();
+    }
+  }
+
+public:
+  thread_pool() = default;
+
+  ~thread_pool() {
+    {
+      std::lock_guard<std::mutex> guard { lock_ };
+      static_cast<void>(guard);
+      quit_ = true;
+    }
+    cv_jobs_.notify_all();
+    cv_empty_.notify_all();
+    for (auto & trd : workers_) trd.join();
+  }
+
+  explicit thread_pool(std::size_t n) : thread_pool() {
+    start(n);
+  }
+
+  void start(std::size_t n) {
+    if (n <= workers_.size()) return;
+    for (std::size_t i = workers_.size(); i < n; ++i) {
+      workers_.push_back(std::thread { &thread_pool::proc, this });
+    }
+  }
+
+  std::size_t size() const noexcept {
+    return workers_.size();
+  }
+
+  std::size_t jobs_size() const noexcept {
+    return jobs_.size();
+  }
+
+  void wait_for_started() {
+    std::unique_lock<std::mutex> guard { lock_ };
+    if (quit_) return;
+    while (!workers_.empty() && (waiting_cnt_ != workers_.size())) {
+      cv_empty_.wait(guard);
+      if (quit_) return;
+    }
+  }
+
+  void wait_for_done() {
+    std::unique_lock<std::mutex> guard { lock_ };
+    if (quit_) return;
+    while (!jobs_.empty() || (waiting_cnt_ != workers_.size())) {
+      assert(waiting_cnt_ <= workers_.size());
+      cv_empty_.wait(guard);
+      if (quit_) return;
+    }
+  }
+
+  template <typename F>
+  thread_pool & operator<<(F && job) {
+    {
+      std::lock_guard<std::mutex> guard { lock_ };
+      static_cast<void>(guard);
+      jobs_.emplace_back(std::forward<F>(job));
+    }
+    cv_jobs_.notify_one();
+    return *this;
+  }
+};
+
+} // namespace ipc_ut
\ No newline at end of file