diff --git a/lib/runtime-c-api/src/lib.rs b/lib/runtime-c-api/src/lib.rs
index 054cd14e1..fb95c7dc6 100644
--- a/lib/runtime-c-api/src/lib.rs
+++ b/lib/runtime-c-api/src/lib.rs
@@ -95,6 +95,8 @@ pub mod instance;
 pub mod memory;
 pub mod module;
 pub mod table;
+#[cfg(all(unix, target_arch = "x86_64"))]
+pub mod trampoline;
 pub mod value;
 
 #[allow(non_camel_case_types)]
diff --git a/lib/runtime-c-api/src/trampoline.rs b/lib/runtime-c-api/src/trampoline.rs
new file mode 100644
index 000000000..923622aa5
--- /dev/null
+++ b/lib/runtime-c-api/src/trampoline.rs
@@ -0,0 +1,84 @@
+//! Trampoline emitter for transforming function calls.
+
+use std::ffi::c_void;
+use std::mem;
+use wasmer_runtime_core::trampoline::*;
+
+#[repr(C)]
+pub struct wasmer_trampoline_buffer_builder_t;
+
+#[repr(C)]
+pub struct wasmer_trampoline_buffer_t;
+
+#[repr(C)]
+pub struct wasmer_trampoline_callable_t;
+
+/// Creates a new trampoline builder.
+#[no_mangle]
+#[allow(clippy::cast_ptr_alignment)]
+pub extern "C" fn wasmer_trampoline_buffer_builder_new() -> *mut wasmer_trampoline_buffer_builder_t
+{
+    Box::into_raw(Box::new(TrampolineBufferBuilder::new())) as *mut _
+}
+
+/// Adds a context trampoline to the builder.
+#[no_mangle]
+#[allow(clippy::cast_ptr_alignment)]
+pub unsafe extern "C" fn wasmer_trampoline_buffer_builder_add_context_trampoline(
+    builder: *mut wasmer_trampoline_buffer_builder_t,
+    func: *const wasmer_trampoline_callable_t,
+    ctx: *const c_void,
+) -> usize {
+    let builder = &mut *(builder as *mut TrampolineBufferBuilder);
+    builder.add_context_trampoline(func as *const CallTarget, ctx as *const CallContext)
+}
+
+/// Adds a callinfo trampoline to the builder.
+#[no_mangle]
+#[allow(clippy::cast_ptr_alignment)]
+pub unsafe extern "C" fn wasmer_trampoline_buffer_builder_add_callinfo_trampoline(
+    builder: *mut wasmer_trampoline_buffer_builder_t,
+    func: *const wasmer_trampoline_callable_t,
+    ctx: *const c_void,
+    num_params: u32,
+) -> usize {
+    let builder = &mut *(builder as *mut TrampolineBufferBuilder);
+    builder.add_callinfo_trampoline(mem::transmute(func), ctx as *const CallContext, num_params)
+}
+
+/// Finalizes the trampoline builder into an executable buffer.
+#[no_mangle]
+#[allow(clippy::cast_ptr_alignment)]
+pub unsafe extern "C" fn wasmer_trampoline_buffer_builder_build(
+    builder: *mut wasmer_trampoline_buffer_builder_t,
+) -> *mut wasmer_trampoline_buffer_t {
+    let builder = Box::from_raw(builder as *mut TrampolineBufferBuilder);
+    Box::into_raw(Box::new(builder.build())) as *mut _
+}
+
+/// Destroys the trampoline buffer if not null.
+#[no_mangle]
+#[allow(clippy::cast_ptr_alignment)]
+pub unsafe extern "C" fn wasmer_trampoline_buffer_destroy(buffer: *mut wasmer_trampoline_buffer_t) {
+    if !buffer.is_null() {
+        Box::from_raw(buffer);
+    }
+}
+
+/// Returns the callable pointer for the trampoline with index `idx`.
+#[no_mangle]
+#[allow(clippy::cast_ptr_alignment)]
+pub unsafe extern "C" fn wasmer_trampoline_buffer_get_trampoline(
+    buffer: *const wasmer_trampoline_buffer_t,
+    idx: usize,
+) -> *const wasmer_trampoline_callable_t {
+    let buffer = &*(buffer as *const TrampolineBuffer);
+    buffer.get_trampoline(idx) as _
+}
+
+/// Returns the context added by `add_context_trampoline`, from within the callee function.
+#[no_mangle]
+#[allow(clippy::cast_ptr_alignment)]
+pub unsafe extern "C" fn wasmer_trampoline_get_context() -> *mut c_void {
+    get_context() as *const c_void as *mut c_void
+}
diff --git a/lib/runtime-c-api/tests/test-import-function-callinfo.c b/lib/runtime-c-api/tests/test-import-function-callinfo.c
new file mode 100644
index 000000000..371f9eeba
--- /dev/null
+++ b/lib/runtime-c-api/tests/test-import-function-callinfo.c
@@ -0,0 +1,132 @@
+#include <stdio.h>
+#include "../wasmer.h"
+#include <assert.h>
+#include <stdint.h>
+#include <string.h>
+
+static bool print_str_called = false;
+static int memory_len = 0;
+static int ptr_len = 0;
+static char actual_str[14] = {};
+static int actual_context_data_value = 0;
+
+typedef struct {
+  int value;
+} context_data;
+
+struct print_str_context {
+    int call_count;
+};
+
+void print_str(struct print_str_context *local_context, uint64_t *args)
+{
+    local_context->call_count++;
+
+    wasmer_instance_context_t *ctx = (void *) args[0];
+    int32_t ptr = args[1];
+    int32_t len = args[2];
+    
+    const wasmer_memory_t *memory = wasmer_instance_context_memory(ctx, 0);
+    uint32_t mem_len = wasmer_memory_length(memory);
+    uint8_t *mem_bytes = wasmer_memory_data(memory);
+    for (int32_t idx = 0; idx < len; idx++)
+    {
+        actual_str[idx] = mem_bytes[ptr + idx];
+    }
+    actual_str[13] = '\0';
+    printf("In print_str, memory len: %d, ptr_len: %d\n, str %s", mem_len, len, actual_str);
+    print_str_called = true;
+    memory_len = mem_len;
+    ptr_len = len;
+
+    actual_context_data_value = ((context_data *) wasmer_instance_context_data_get(ctx))->value;
+}
+
+int main()
+{
+    wasmer_value_tag params_sig[] = {WASM_I32, WASM_I32};
+    wasmer_value_tag returns_sig[] = {};
+    struct print_str_context local_context = {
+        .call_count = 0
+    };
+
+    printf("Creating trampoline buffer\n");
+    wasmer_trampoline_buffer_builder_t *tbb = wasmer_trampoline_buffer_builder_new();
+    unsigned long print_str_idx = wasmer_trampoline_buffer_builder_add_callinfo_trampoline(
+        tbb,
+        (wasmer_trampoline_callable_t *) print_str,
+        (void *) &local_context,
+        3
+    );
+    wasmer_trampoline_buffer_t *tb = wasmer_trampoline_buffer_builder_build(tbb);
+    const wasmer_trampoline_callable_t *print_str_callable = wasmer_trampoline_buffer_get_trampoline(tb, print_str_idx);
+
+    printf("Creating new func\n");
+    wasmer_import_func_t *func = wasmer_import_func_new((void (*)(void *)) print_str_callable, params_sig, 2, returns_sig, 0);
+    wasmer_import_t import;
+
+    char *module_name = "env";
+    wasmer_byte_array module_name_bytes;
+    module_name_bytes.bytes = (const uint8_t *) module_name;
+    module_name_bytes.bytes_len = strlen(module_name);
+    char *import_name = "print_str";
+    wasmer_byte_array import_name_bytes;
+    import_name_bytes.bytes = (const uint8_t *) import_name;
+    import_name_bytes.bytes_len = strlen(import_name);
+
+    import.module_name = module_name_bytes;
+    import.import_name = import_name_bytes;
+    import.tag = WASM_FUNCTION;
+    import.value.func = func;
+    wasmer_import_t imports[] = {import};
+
+    // Read the wasm file bytes
+    FILE *file = fopen("assets/wasm_sample_app.wasm", "r");
+    fseek(file, 0, SEEK_END);
+    long len = ftell(file);
+    uint8_t *bytes = malloc(len);
+    fseek(file, 0, SEEK_SET);
+    fread(bytes, 1, len, file);
+    fclose(file);
+
+    printf("Instantiating\n");
+    wasmer_instance_t *instance = NULL;
+    wasmer_result_t compile_result = wasmer_instantiate(&instance, bytes, len, imports, 1);
+    printf("Compile result:  %d\n", compile_result);
+
+    assert(compile_result == WASMER_OK);
+
+    context_data* context_data = malloc(sizeof(context_data));
+    int context_data_value = 42;
+    context_data->value = context_data_value;
+    wasmer_instance_context_data_set(instance, context_data);
+
+    wasmer_value_t params[] = {};
+    wasmer_value_t results[] = {};
+    wasmer_result_t call_result = wasmer_instance_call(instance, "hello_wasm", params, 0, results, 0);
+    printf("Call result:  %d\n", call_result);
+
+    int error_len = wasmer_last_error_length();
+    printf("Error len: `%d`\n", error_len);
+    char *error_str = malloc(error_len);
+    wasmer_last_error_message(error_str, error_len);
+    printf("Error str: `%s`\n", error_str);
+
+    assert(call_result == WASMER_OK);
+
+    assert(print_str_called);
+    assert(memory_len == 17);
+    assert(ptr_len == 13);
+    assert(0 == strcmp(actual_str, "Hello, World!"));
+    assert(context_data_value == actual_context_data_value);
+    assert(local_context.call_count == 1);
+
+    printf("Destroying trampoline buffer\n");
+    wasmer_trampoline_buffer_destroy(tb);
+    printf("Destroying func\n");
+    wasmer_import_func_destroy(func);
+    printf("Destroy instance\n");
+    wasmer_instance_destroy(instance);
+    free(context_data);
+    return 0;
+}
diff --git a/lib/runtime-c-api/tests/test-import-function.c b/lib/runtime-c-api/tests/test-import-function.c
index 92e3e821d..a6bea2a54 100644
--- a/lib/runtime-c-api/tests/test-import-function.c
+++ b/lib/runtime-c-api/tests/test-import-function.c
@@ -14,8 +14,15 @@ typedef struct {
   int value;
 } context_data;
 
+struct print_str_context {
+    int call_count;
+};
+
 void print_str(wasmer_instance_context_t *ctx, int32_t ptr, int32_t len)
 {
+    struct print_str_context *local_context = wasmer_trampoline_get_context();
+    local_context->call_count++;
+
     const wasmer_memory_t *memory = wasmer_instance_context_memory(ctx, 0);
     uint32_t mem_len = wasmer_memory_length(memory);
     uint8_t *mem_bytes = wasmer_memory_data(memory);
@@ -36,9 +43,22 @@ int main()
 {
     wasmer_value_tag params_sig[] = {WASM_I32, WASM_I32};
     wasmer_value_tag returns_sig[] = {};
+    struct print_str_context local_context = {
+        .call_count = 0
+    };
+
+    printf("Creating trampoline buffer\n");
+    wasmer_trampoline_buffer_builder_t *tbb = wasmer_trampoline_buffer_builder_new();
+    unsigned long print_str_idx = wasmer_trampoline_buffer_builder_add_context_trampoline(
+        tbb,
+        (wasmer_trampoline_callable_t *) print_str,
+        (void *) &local_context
+    );
+    wasmer_trampoline_buffer_t *tb = wasmer_trampoline_buffer_builder_build(tbb);
+    const wasmer_trampoline_callable_t *print_str_callable = wasmer_trampoline_buffer_get_trampoline(tb, print_str_idx);
 
     printf("Creating new func\n");
-    wasmer_import_func_t *func = wasmer_import_func_new((void (*)(void *)) print_str, params_sig, 2, returns_sig, 0);
+    wasmer_import_func_t *func = wasmer_import_func_new((void (*)(void *)) print_str_callable, params_sig, 2, returns_sig, 0);
     wasmer_import_t import;
 
     char *module_name = "env";
@@ -95,7 +115,10 @@ int main()
     assert(ptr_len == 13);
     assert(0 == strcmp(actual_str, "Hello, World!"));
     assert(context_data_value == actual_context_data_value);
+    assert(local_context.call_count == 1);
 
+    printf("Destroying trampoline buffer\n");
+    wasmer_trampoline_buffer_destroy(tb);
     printf("Destroying func\n");
     wasmer_import_func_destroy(func);
     printf("Destroy instance\n");
diff --git a/lib/runtime-c-api/wasmer.h b/lib/runtime-c-api/wasmer.h
index 6752f71a1..7c0f5249a 100644
--- a/lib/runtime-c-api/wasmer.h
+++ b/lib/runtime-c-api/wasmer.h
@@ -133,6 +133,18 @@ typedef struct {
 
 } wasmer_serialized_module_t;
 
+typedef struct {
+
+} wasmer_trampoline_buffer_builder_t;
+
+typedef struct {
+
+} wasmer_trampoline_callable_t;
+
+typedef struct {
+
+} wasmer_trampoline_buffer_t;
+
 /**
  * Creates a new Module from the given wasm bytes.
  * Returns `wasmer_result_t::WASMER_OK` upon success.
@@ -584,6 +596,47 @@ uint32_t wasmer_table_length(wasmer_table_t *table);
  */
 wasmer_result_t wasmer_table_new(wasmer_table_t **table, wasmer_limits_t limits);
 
+/**
+ * Adds a callinfo trampoline to the builder.
+ */
+uintptr_t wasmer_trampoline_buffer_builder_add_callinfo_trampoline(wasmer_trampoline_buffer_builder_t *builder,
+                                                                   const wasmer_trampoline_callable_t *func,
+                                                                   const void *ctx,
+                                                                   uint32_t num_params);
+
+/**
+ * Adds a context trampoline to the builder.
+ */
+uintptr_t wasmer_trampoline_buffer_builder_add_context_trampoline(wasmer_trampoline_buffer_builder_t *builder,
+                                                                  const wasmer_trampoline_callable_t *func,
+                                                                  const void *ctx);
+
+/**
+ * Finalizes the trampoline builder into an executable buffer.
+ */
+wasmer_trampoline_buffer_t *wasmer_trampoline_buffer_builder_build(wasmer_trampoline_buffer_builder_t *builder);
+
+/**
+ * Creates a new trampoline builder.
+ */
+wasmer_trampoline_buffer_builder_t *wasmer_trampoline_buffer_builder_new(void);
+
+/**
+ * Destroys the trampoline buffer if not null.
+ */
+void wasmer_trampoline_buffer_destroy(wasmer_trampoline_buffer_t *buffer);
+
+/**
+ * Returns the callable pointer for the trampoline with index `idx`.
+ */
+const wasmer_trampoline_callable_t *wasmer_trampoline_buffer_get_trampoline(const wasmer_trampoline_buffer_t *buffer,
+                                                                            uintptr_t idx);
+
+/**
+ * Returns the context added by `add_context_trampoline`, from within the callee function.
+ */
+void *wasmer_trampoline_get_context(void);
+
 /**
  * Returns true for valid wasm bytes and false for invalid bytes
  */
diff --git a/lib/runtime-c-api/wasmer.hh b/lib/runtime-c-api/wasmer.hh
index 99e21fcc8..1c00b74c3 100644
--- a/lib/runtime-c-api/wasmer.hh
+++ b/lib/runtime-c-api/wasmer.hh
@@ -131,6 +131,18 @@ struct wasmer_serialized_module_t {
 
 };
 
+struct wasmer_trampoline_buffer_builder_t {
+
+};
+
+struct wasmer_trampoline_callable_t {
+
+};
+
+struct wasmer_trampoline_buffer_t {
+
+};
+
 extern "C" {
 
 /// Creates a new Module from the given wasm bytes.
@@ -458,6 +470,33 @@ uint32_t wasmer_table_length(wasmer_table_t *table);
 /// and `wasmer_last_error_message` to get an error message.
 wasmer_result_t wasmer_table_new(wasmer_table_t **table, wasmer_limits_t limits);
 
+/// Adds a callinfo trampoline to the builder.
+uintptr_t wasmer_trampoline_buffer_builder_add_callinfo_trampoline(wasmer_trampoline_buffer_builder_t *builder,
+                                                                   const wasmer_trampoline_callable_t *func,
+                                                                   const void *ctx,
+                                                                   uint32_t num_params);
+
+/// Adds a context trampoline to the builder.
+uintptr_t wasmer_trampoline_buffer_builder_add_context_trampoline(wasmer_trampoline_buffer_builder_t *builder,
+                                                                  const wasmer_trampoline_callable_t *func,
+                                                                  const void *ctx);
+
+/// Finalizes the trampoline builder into an executable buffer.
+wasmer_trampoline_buffer_t *wasmer_trampoline_buffer_builder_build(wasmer_trampoline_buffer_builder_t *builder);
+
+/// Creates a new trampoline builder.
+wasmer_trampoline_buffer_builder_t *wasmer_trampoline_buffer_builder_new();
+
+/// Destroys the trampoline buffer if not null.
+void wasmer_trampoline_buffer_destroy(wasmer_trampoline_buffer_t *buffer);
+
+/// Returns the callable pointer for the trampoline with index `idx`.
+const wasmer_trampoline_callable_t *wasmer_trampoline_buffer_get_trampoline(const wasmer_trampoline_buffer_t *buffer,
+                                                                            uintptr_t idx);
+
+/// Returns the context added by `add_context_trampoline`, from within the callee function.
+void *wasmer_trampoline_get_context();
+
 /// Returns true for valid wasm bytes and false for invalid bytes
 bool wasmer_validate(const uint8_t *wasm_bytes, uint32_t wasm_bytes_len);
 
diff --git a/lib/runtime-core/src/lib.rs b/lib/runtime-core/src/lib.rs
index 20708c17a..662f5c970 100644
--- a/lib/runtime-core/src/lib.rs
+++ b/lib/runtime-core/src/lib.rs
@@ -8,6 +8,10 @@ extern crate field_offset;
 #[macro_use]
 extern crate serde_derive;
 
+#[allow(unused_imports)]
+#[macro_use]
+extern crate lazy_static;
+
 #[macro_use]
 mod macros;
 #[doc(hidden)]
@@ -29,12 +33,16 @@ mod sig_registry;
 pub mod structures;
 mod sys;
 pub mod table;
+#[cfg(all(unix, target_arch = "x86_64"))]
+pub mod trampoline_x64;
 pub mod typed_func;
 pub mod types;
 pub mod units;
 pub mod vm;
 #[doc(hidden)]
 pub mod vmcalls;
+#[cfg(all(unix, target_arch = "x86_64"))]
+pub use trampoline_x64 as trampoline;
 
 use self::error::CompileResult;
 #[doc(inline)]
diff --git a/lib/runtime-core/src/trampoline_x64.rs b/lib/runtime-core/src/trampoline_x64.rs
new file mode 100644
index 000000000..4defd1fa7
--- /dev/null
+++ b/lib/runtime-core/src/trampoline_x64.rs
@@ -0,0 +1,245 @@
+//! Trampoline generator for carrying context with function pointer.
+//!
+//! This makes use of the `mm0` register to pass the context as an implicit "parameter" because `mm0` is
+//! not used to pass parameters and is almost never used by modern compilers. It's still better to call
+//! `get_context()` as early as possible in the callee function though, as a good practice.
+//!
+//! Variadic functions are not supported because `rax` is used by the trampoline code.
+
+use crate::loader::CodeMemory;
+use std::{mem, slice};
+
+lazy_static! {
+    /// Reads the context pointer from `mm0`.
+    ///
+    /// This function generates code at runtime since `asm!` macro is not yet stable.
+    static ref GET_CONTEXT: extern "C" fn () -> *const CallContext = {
+        static CODE: &'static [u8] = &[
+            0x48, 0x0f, 0x7e, 0xc0, // movq %mm0, %rax
+            0xc3, // retq
+        ];
+        let mut mem = CodeMemory::new(4096);
+        mem[..CODE.len()].copy_from_slice(CODE);
+        mem.make_executable();
+        let ptr = mem.as_ptr();
+        mem::forget(mem);
+        unsafe {
+            mem::transmute(ptr)
+        }
+    };
+}
+
+/// An opaque type for pointers to a callable memory location.
+pub enum CallTarget {}
+
+/// An opaque type for context pointers.
+pub enum CallContext {}
+
+/// An opaque type for generated trampolines' call entries.
+pub enum Trampoline {}
+
+/// Trampoline Buffer Builder.
+pub struct TrampolineBufferBuilder {
+    code: Vec<u8>,
+    offsets: Vec<usize>,
+}
+
+/// Trampoline Buffer.
+pub struct TrampolineBuffer {
+    code: CodeMemory,
+    offsets: Vec<usize>,
+}
+
+fn value_to_bytes<T: Copy>(ptr: &T) -> &[u8] {
+    unsafe { slice::from_raw_parts(ptr as *const T as *const u8, mem::size_of::<T>()) }
+}
+
+/// Calls `GET_CONTEXT` and returns the current context.
+pub fn get_context() -> *const CallContext {
+    GET_CONTEXT()
+}
+
+impl TrampolineBufferBuilder {
+    pub fn new() -> TrampolineBufferBuilder {
+        TrampolineBufferBuilder {
+            code: vec![],
+            offsets: vec![],
+        }
+    }
+
+    /// Adds a context trampoline.
+    ///
+    /// This generates a transparent trampoline function that forwards any call to `target` with
+    /// unmodified params/returns. When called from the trampoline, `target` will have access to
+    /// the `context` specified here through `get_context()`.
+    ///
+    /// Note that since `rax` is overwritten internally, variadic functions are not supported as `target`.
+    pub fn add_context_trampoline(
+        &mut self,
+        target: *const CallTarget,
+        context: *const CallContext,
+    ) -> usize {
+        let idx = self.offsets.len();
+        self.offsets.push(self.code.len());
+        self.code.extend_from_slice(&[
+            0x48, 0xb8, // movabsq ?, %rax
+        ]);
+        self.code.extend_from_slice(value_to_bytes(&context));
+        self.code.extend_from_slice(&[
+            0x48, 0x0f, 0x6e, 0xc0, // movq %rax, %mm0
+        ]);
+        self.code.extend_from_slice(&[
+            0x48, 0xb8, // movabsq ?, %rax
+        ]);
+        self.code.extend_from_slice(value_to_bytes(&target));
+        self.code.extend_from_slice(&[
+            0xff, 0xe0, // jmpq *%rax
+        ]);
+        idx
+    }
+
+    /// Adds a callinfo trampoline.
+    ///
+    /// This generates a trampoline function that collects `num_params` parameters into an array
+    /// and passes the array into `target` as the second argument when called. The first argument
+    /// of `target` is the `context` specified here.
+    ///
+    /// Note that non-integer parameters/variadic functions are not supported.
+    pub fn add_callinfo_trampoline(
+        &mut self,
+        target: unsafe extern "C" fn(*const CallContext, *const u64) -> u64,
+        context: *const CallContext,
+        num_params: u32,
+    ) -> usize {
+        let idx = self.offsets.len();
+        self.offsets.push(self.code.len());
+
+        let mut stack_offset: u32 = num_params.checked_mul(8).unwrap();
+        if stack_offset % 16 == 0 {
+            stack_offset += 8;
+        }
+
+        self.code.extend_from_slice(&[0x48, 0x81, 0xec]); // sub ?, %rsp
+        self.code.extend_from_slice(value_to_bytes(&stack_offset));
+        for i in 0..num_params {
+            match i {
+                0..=5 => {
+                    // mov %?, ?(%rsp)
+                    let prefix: &[u8] = match i {
+                        0 => &[0x48, 0x89, 0xbc, 0x24], // rdi
+                        1 => &[0x48, 0x89, 0xb4, 0x24], // rsi
+                        2 => &[0x48, 0x89, 0x94, 0x24], // rdx
+                        3 => &[0x48, 0x89, 0x8c, 0x24], // rcx
+                        4 => &[0x4c, 0x89, 0x84, 0x24], // r8
+                        5 => &[0x4c, 0x89, 0x8c, 0x24], // r9
+                        _ => unreachable!(),
+                    };
+                    self.code.extend_from_slice(prefix);
+                    self.code.extend_from_slice(value_to_bytes(&(i * 8u32)));
+                }
+                _ => {
+                    self.code.extend_from_slice(&[
+                        0x48, 0x8b, 0x84, 0x24, // mov ?(%rsp), %rax
+                    ]);
+                    self.code.extend_from_slice(value_to_bytes(
+                        &((i - 6) * 8u32 + stack_offset + 8/* ret addr */),
+                    ));
+                    // mov %rax, ?(%rsp)
+                    self.code.extend_from_slice(&[0x48, 0x89, 0x84, 0x24]);
+                    self.code.extend_from_slice(value_to_bytes(&(i * 8u32)));
+                }
+            }
+        }
+        self.code.extend_from_slice(&[
+            0x48, 0xbf, // movabsq ?, %rdi
+        ]);
+        self.code.extend_from_slice(value_to_bytes(&context));
+        self.code.extend_from_slice(&[
+            0x48, 0x89, 0xe6, // mov %rsp, %rsi
+        ]);
+
+        self.code.extend_from_slice(&[
+            0x48, 0xb8, // movabsq ?, %rax
+        ]);
+        self.code.extend_from_slice(value_to_bytes(&target));
+        self.code.extend_from_slice(&[
+            0xff, 0xd0, // callq *%rax
+        ]);
+        self.code.extend_from_slice(&[
+            0x48, 0x81, 0xc4, // add ?, %rsp
+        ]);
+        self.code.extend_from_slice(value_to_bytes(&stack_offset));
+        self.code.extend_from_slice(&[
+            0xc3, //retq
+        ]);
+        idx
+    }
+
+    /// Consumes the builder and builds the trampoline buffer.
+    pub fn build(self) -> TrampolineBuffer {
+        get_context(); // ensure lazy initialization is completed
+
+        let mut code = CodeMemory::new(self.code.len());
+        code[..self.code.len()].copy_from_slice(&self.code);
+        code.make_executable();
+        TrampolineBuffer {
+            code,
+            offsets: self.offsets,
+        }
+    }
+}
+
+impl TrampolineBuffer {
+    /// Returns the trampoline pointer at index `idx`.
+    pub fn get_trampoline(&self, idx: usize) -> *const Trampoline {
+        &self.code[self.offsets[idx]] as *const u8 as *const Trampoline
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    #[test]
+    fn test_context_trampoline() {
+        struct TestContext {
+            value: i32,
+        }
+        extern "C" fn do_add(a: i32, b: f32) -> f32 {
+            let ctx = unsafe { &*(get_context() as *const TestContext) };
+            a as f32 + b + ctx.value as f32
+        }
+        let mut builder = TrampolineBufferBuilder::new();
+        let ctx = TestContext { value: 3 };
+        let idx = builder.add_context_trampoline(
+            do_add as usize as *const _,
+            &ctx as *const TestContext as *const _,
+        );
+        let buf = builder.build();
+        let t = buf.get_trampoline(idx);
+        let ret = unsafe { mem::transmute::<_, extern "C" fn(i32, f32) -> f32>(t)(1, 2.0) as i32 };
+        assert_eq!(ret, 6);
+    }
+    #[test]
+    fn test_callinfo_trampoline() {
+        struct TestContext {
+            value: i32,
+        }
+        unsafe extern "C" fn do_add(ctx: *const CallContext, args: *const u64) -> u64 {
+            let ctx = &*(ctx as *const TestContext);
+            let args: &[u64] = slice::from_raw_parts(args, 8);
+            (args.iter().map(|x| *x as i32).fold(0, |a, b| a + b) + ctx.value) as u64
+        }
+        let mut builder = TrampolineBufferBuilder::new();
+        let ctx = TestContext { value: 100 };
+        let idx =
+            builder.add_callinfo_trampoline(do_add, &ctx as *const TestContext as *const _, 8);
+        let buf = builder.build();
+        let t = buf.get_trampoline(idx);
+        let ret = unsafe {
+            mem::transmute::<_, extern "C" fn(i32, i32, i32, i32, i32, i32, i32, i32) -> i32>(t)(
+                1, 2, 3, 4, 5, 6, 7, 8,
+            ) as i32
+        };
+        assert_eq!(ret, 136);
+    }
+}