diff --git a/ujit_compile.c b/ujit_compile.c
index 67162fbd21..e4aafde936 100644
--- a/ujit_compile.c
+++ b/ujit_compile.c
@@ -34,7 +34,7 @@ typedef struct ctx_struct
 } ctx_t;
 
 // MicroJIT code generation function signature
-typedef void (*codegen_fn)(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx);
+typedef bool (*codegen_fn)(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx);
 
 // Map from YARV opcodes to code generation functions
 static st_table *gen_fns;
@@ -255,7 +255,10 @@ ujit_compile_insn(rb_iseq_t *iseq, unsigned int insn_idx, unsigned int* next_uji
 
         // Call the code generation function
         codegen_fn gen_fn = (codegen_fn)st_gen_fn;
-        gen_fn(cb, ocb, &ctx);
+        if (!gen_fn(cb, ocb, &ctx))
+        {
+            break;
+        }
 
     	// Move to the next instruction
         insn_idx += insn_len(opcode);
@@ -278,7 +281,8 @@ ujit_compile_insn(rb_iseq_t *iseq, unsigned int insn_idx, unsigned int* next_uji
     return code_ptr;
 }
 
-void gen_dup(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
+bool
+gen_dup(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
 {
     x86opnd_t dup_val = ctx_stack_pop(ctx, 1);
     x86opnd_t loc0 = ctx_stack_push(ctx, 1);
@@ -286,27 +290,35 @@ void gen_dup(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
     mov(cb, RAX, dup_val);
     mov(cb, loc0, RAX);
     mov(cb, loc1, RAX);
+    return true;
 }
 
-void gen_nop(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
+bool
+gen_nop(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
 {
     // Do nothing
+    return true;
 }
 
-void gen_pop(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
+bool
+gen_pop(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
 {
     // Decrement SP
     ctx_stack_pop(ctx, 1);
+    return true;
 }
 
-void gen_putnil(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
+bool
+gen_putnil(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
 {
     // Write constant at SP
     x86opnd_t stack_top = ctx_stack_push(ctx, 1);
     mov(cb, stack_top, imm_opnd(Qnil));
+    return true;
 }
 
-void gen_putobject(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
+bool
+gen_putobject(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
 {
     // Load the argument from the bytecode sequence.
     // We need to do this as the argument can chanage due to GC compaction.
@@ -317,9 +329,12 @@ void gen_putobject(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
     // Write argument at SP
     x86opnd_t stack_top = ctx_stack_push(ctx, 1);
     mov(cb, stack_top, RAX);
+
+    return true;
 }
 
-void gen_putobject_int2fix(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
+bool
+gen_putobject_int2fix(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
 {
     int opcode = ctx_get_opcode(ctx);
     int cst_val = (opcode == BIN(putobject_INT2FIX_0_))? 0:1;
@@ -327,9 +342,12 @@ void gen_putobject_int2fix(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
     // Write constant at SP
     x86opnd_t stack_top = ctx_stack_push(ctx, 1);
     mov(cb, stack_top, imm_opnd(INT2FIX(cst_val)));
+
+    return true;
 }
 
-void gen_putself(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
+bool
+gen_putself(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
 {
     // Load self from CFP
     mov(cb, RAX, mem_opnd(64, RDI, 24));
@@ -337,9 +355,12 @@ void gen_putself(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
     // Write it on the stack
     x86opnd_t stack_top = ctx_stack_push(ctx, 1);
     mov(cb, stack_top, RAX);
+
+    return true;
 }
 
-void gen_getlocal_wc0(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
+bool
+gen_getlocal_wc0(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
 {
     // Load environment pointer EP from CFP
     mov(cb, RDX, member_opnd(RDI, rb_control_frame_t, ep));
@@ -354,9 +375,12 @@ void gen_getlocal_wc0(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
     // Write the local at SP
     x86opnd_t stack_top = ctx_stack_push(ctx, 1);
     mov(cb, stack_top, RCX);
+
+    return true;
 }
 
-void gen_setlocal_wc0(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
+bool
+gen_setlocal_wc0(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
 {
     /*
     vm_env_write(const VALUE *ep, int index, VALUE v)
@@ -392,9 +416,12 @@ void gen_setlocal_wc0(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
     int32_t local_idx = (int32_t)ctx_get_arg(ctx, 0);
     const int32_t offs = -8 * local_idx;
     mov(cb, mem_opnd(64, RDX, offs), RCX);
+
+    return true;
 }
 
-void gen_opt_minus(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
+bool
+gen_opt_minus(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
 {
     // Create a size-exit to fall back to the interpreter
     // Note: we generate the side-exit before popping operands from the stack
@@ -429,56 +456,87 @@ void gen_opt_minus(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
     // Push the output on the stack
     x86opnd_t dst = ctx_stack_push(ctx, 1);
     mov(cb, dst, RAX);
+
+    return true;
 }
 
-void gen_opt_send_without_block(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
+bool
+gen_opt_send_without_block(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
 {
-    // Create a size-exit to fall back to the interpreter
-    uint8_t* side_exit = ujit_side_exit(ocb, ctx, ctx->pc);
-
-
-
-    /*
     struct rb_call_data * cd = (struct rb_call_data *)ctx_get_arg(ctx, 0);
     int32_t argc = (int32_t)vm_ci_argc(cd->ci);
     const struct rb_callcache *cc = cd->cc;
 
+    // Callee method ID
     ID mid = vm_ci_mid(cd->ci);
     //fprintf(stderr, "jitting call to \"%s\", argc: %lu\n", rb_id2name(mid), argc);
 
-    // TODO: don't jit calls that aren't simple
-    // have this codegen function return false, make codegen stop?
-    if (vm_ci_flag(cd->ci) & VM_CALL_ARGS_SIMPLE)
+    // Don't JIT calls with keyword splat
+    if (vm_ci_flag(cd->ci) & VM_CALL_KW_SPLAT)
     {
-        //fprintf(stderr, "simple call\n");
+        return false;
     }
 
+    // Don't jit calls that aren't simple
+    if (!(vm_ci_flag(cd->ci) & VM_CALL_ARGS_SIMPLE))
+    {
+        return false;
+    }
+
+    // TODO: stop if the inline cache isn't filled
+
+    // TODO: stop if this isn't a C call
+
+
+
+
+
+
+
+
+
+
+    return false;
+
+    /*
+    // Create a size-exit to fall back to the interpreter
+    uint8_t* side_exit = ujit_side_exit(ocb, ctx, ctx->pc);
+
+    struct rb_calling_info *calling = (struct rb_calling_info*)malloc(sizeof(struct rb_calling_info));
+    calling->block_handler = VM_BLOCK_HANDLER_NONE;
+    calling->kw_splat = 0;
+    calling->argc = argc;
+
     mov(cb, RAX, const_ptr_opnd(cd));
     x86opnd_t ptr_to_cc = member_opnd(RAX, struct rb_call_data, cc);
     mov(cb, RAX, ptr_to_cc);
-    */
 
-
-
-
-
-    /*
     x86opnd_t ptr_to_klass = mem_opnd(64, RAX, offsetof(struct rb_callcache, klass));
     x86opnd_t ptr_to_cme_ = mem_opnd(64, RAX, offsetof(struct rb_callcache, cme_));
-    mov(cb, RBX, ptr_to_klass);
+    x86opnd_t ptr_to_call_ = mem_opnd(64, RAX, offsetof(struct rb_callcache, call_));
+    mov(cb, R9, ptr_to_klass);
     mov(cb, RCX, ptr_to_cme_);
 
     // Points to the receiver operand on the stack
     x86opnd_t recv = ctx_stack_opnd(ctx, argc);
     mov(cb, RDX, recv);
+    //print_int(cb, recv);
+
+    // Store calling->recv
+    mov(cb, R8, const_ptr_opnd(calling));
+    x86opnd_t recv_opnd = mem_opnd(64, R8, offsetof(struct rb_calling_info, recv));
+    mov(cb, recv_opnd, RDX);
+
+
+    //print_int(cb, recv_opnd);
+
     // Pointer to the klass field of the receiver
     x86opnd_t klass_opnd = mem_opnd(64, RDX, offsetof(struct RBasic, klass));
 
-    print_int(cb, klass_opnd);
 
-    cmp(cb, RBX, klass_opnd);
+    cmp(cb, R9, klass_opnd);
     jne_ptr(cb, side_exit);
-
+    //print_int(cb, klass_opnd);
     print_str(cb, "cache klass hit");
 
     //#define METHOD_ENTRY_INVALIDATED(me)         ((me)->flags & IMEMO_FL_USER5)
@@ -486,14 +544,53 @@ void gen_opt_send_without_block(codeblock_t* cb, codeblock_t* ocb, ctx_t* ctx)
     test(cb, flags_opnd, imm_opnd(IMEMO_FL_USER5));
     jnz_ptr(cb, side_exit);
 
-    print_str(cb, "method entry not invalidated!!!1");
+    push(cb, RDI);
+    push(cb, RSI);
+
+    x86opnd_t ptr_to_pc = mem_opnd(64, RDI, offsetof(rb_control_frame_t, pc));
+    mov(cb, ptr_to_pc, const_ptr_opnd(ctx->pc + insn_len(BIN(opt_send_without_block))));
+
+    // Write the adjusted SP back into the CFP
+    if (ctx->stack_diff != 0)
+    {
+        x86opnd_t stack_pointer = ctx_sp_opnd(ctx, 1);
+        lea(cb, RSI, stack_pointer);
+        mov(cb, mem_opnd(64, RDI, 8), RSI);
+    }
+
+    // val = vm_cc_call(cc)(ec, GET_CFP(), &calling, cd);
+    mov(cb, RSI, RDI);
+    mov(cb, RDI, const_ptr_opnd(rb_current_execution_context()));
+    mov(cb, RDX, R8);
+    print_int(cb, RDX);
+    mov(cb, RCX, const_ptr_opnd(cd));
+
+    call(cb, ptr_to_call_);
+
+    pop(cb, RSI);
+    pop(cb, RDI);
+
+    size_t continue_in_jit = cb_new_label(cb, "continue_in_jit");
+    cmp(cb, RAX, imm_opnd(Qundef));
+    jne(cb, continue_in_jit);
+
+    //print_str(cb, "method entry not invalidated!!!1");
+
+    mov(cb, RDI, const_ptr_opnd(rb_current_execution_context()));
+    mov(cb, RDI, mem_opnd(64, RDI, offsetof(rb_execution_context_t, cfp)));
+
+    // Read the PC from the CFP
+    mov(cb, RAX, mem_opnd(64, RDI, 0));
+
+    // Write the post call bytes
+    for (size_t i = 0; i < sizeof(ujit_post_call_bytes); ++i)
+        cb_write_byte(cb, ujit_post_call_bytes[i]);
+
+    cb_write_label(cb, continue_in_jit);
+    cb_link_labels(cb);
+
+    return true;
     */
-
-
-
-
-
-    jmp_ptr(cb, side_exit);
 }
 
 bool
@@ -505,9 +602,11 @@ rb_ujit_enabled_p(void)
 void
 rb_ujit_init(void)
 {
-    if (!ujit_scrape_successful || !PLATFORM_SUPPORTED_P) {
+    if (!ujit_scrape_successful || !PLATFORM_SUPPORTED_P)
+    {
         return;
     }
+
     // Initialize the code blocks
     size_t mem_size = 128 * 1024 * 1024;
     uint8_t* mem_block = alloc_exec_mem(mem_size);