Set VM_FRAME_FLAG_FINISH at once on MJIT

Performance is probably improved?

$ benchmark-driver -v --rbenv 'before --jit;after --jit' --repeat-count=12 --alternate --output=all benchmark.yml
before --jit: ruby 3.0.0dev (2020-11-27T04:37:47Z master 69e77e81dc) +JIT [x86_64-linux]
after --jit: ruby 3.0.0dev (2020-11-27T05:28:19Z master df6b05c6dd) +JIT [x86_64-linux]
last_commit=Set VM_FRAME_FLAG_FINISH at once
Calculating -------------------------------------
                                 before --jit           after --jit
Optcarrot Lan_Master.nes    80.89292998533379     82.19497327502751 fps
                            80.93130641142331     85.13943315260148
                            81.06214830270119     87.43757879797808
                            82.29172808453910     87.89942441487113
                            84.61206450455929     87.91309779491075
                            85.44545883567997     87.98026086648694
                            86.02923132404449     88.03081060383973
                            86.07411817365879     88.14650206137341
                            86.34348799602836     88.32791633649961
                            87.90257338977324     88.57599644892220
                            88.58006509876580     88.67426384743277
                            89.26611118140011     88.81669430874207

This should have no bad impact on VM because this function is ALWAYS_INLINE.

4 files changed, 12 insertions, 13 deletions

diff --git a/mjit_compile.c b/mjit_compile.c
index c1d8f3468a..a18a9550e5 100644
--- a/mjit_compile.c
+++ b/mjit_compile.c
@@ -268,7 +268,7 @@ compile_inlined_cancel_handler(FILE *f, const struct rb_iseq_constant_body *body
     fprintf(f, "    calling.argc = %d;\n", inline_context->orig_argc);
     fprintf(f, "    calling.recv = reg_cfp->self;\n");
     fprintf(f, "    reg_cfp->self = orig_self;\n");
-    fprintf(f, "    vm_call_iseq_setup_normal(ec, reg_cfp, &calling, (const rb_callable_method_entry_t *)0x%"PRIxVALUE", 0, %d, %d);\n\n",
+    fprintf(f, "    vm_call_iseq_setup_normal(ec, reg_cfp, &calling, (const rb_callable_method_entry_t *)0x%"PRIxVALUE", 0, %d, %d, 0);\n\n",
             inline_context->me, inline_context->param_size, inline_context->local_size); // fastpath_applied_iseq_p checks rb_simple_iseq_p, which ensures has_opt == FALSE
 
     // Start usual cancel from here.
diff --git a/template/call_iseq_optimized.inc.tmpl b/template/call_iseq_optimized.inc.tmpl
index f8883a1d95..c6ea86650f 100644
--- a/template/call_iseq_optimized.inc.tmpl
+++ b/template/call_iseq_optimized.inc.tmpl
@@ -19,7 +19,7 @@ static VALUE
 <%= fname(param, local) %>(rb_execution_context_t *ec, rb_control_frame_t *cfp, struct rb_calling_info *calling, struct rb_call_data *cd)
 {
     RB_DEBUG_COUNTER_INC(ccf_iseq_fix);
-    return vm_call_iseq_setup_normal(ec, cfp, calling, vm_cc_cme(cd->cc), 0, <%= param %>, <%= local %>);
+    return vm_call_iseq_setup_normal(ec, cfp, calling, vm_cc_cme(cd->cc), 0, <%= param %>, <%= local %>, 0);
 }
 
 %   }
diff --git a/tool/ruby_vm/views/_mjit_compile_send.erb b/tool/ruby_vm/views/_mjit_compile_send.erb
index 01ae593bc0..93493ed065 100644
--- a/tool/ruby_vm/views/_mjit_compile_send.erb
+++ b/tool/ruby_vm/views/_mjit_compile_send.erb
@@ -81,14 +81,13 @@
             }
             else { // VM_METHOD_TYPE_ISEQ
 %               # fastpath_applied_iseq_p checks rb_simple_iseq_p, which ensures has_opt == FALSE
-                fprintf(f, "        vm_call_iseq_setup_normal(ec, reg_cfp, &calling, cc_cme, 0, %d, %d);\n", iseq->body->param.size, iseq->body->local_table_size);
+                fprintf(f, "        vm_call_iseq_setup_normal(ec, reg_cfp, &calling, cc_cme, 0, %d, %d, VM_FRAME_FLAG_FINISH);\n", // Set VM_FRAME_FLAG_FINISH like vm_call0_body
+                        iseq->body->param.size, iseq->body->local_table_size);
                 if (iseq->body->catch_except_p) {
-                    fprintf(f, "        VM_ENV_FLAGS_SET(ec->cfp->ep, VM_FRAME_FLAG_FINISH);\n");
                     fprintf(f, "        val = vm_exec(ec, TRUE);\n");
                 }
                 else {
                     fprintf(f, "        if ((val = mjit_exec(ec)) == Qundef) {\n");
-                    fprintf(f, "            VM_ENV_FLAGS_SET(ec->cfp->ep, VM_FRAME_FLAG_FINISH);\n"); // This is vm_call0_body's code after vm_call_iseq_setup
                     fprintf(f, "            val = vm_exec(ec, FALSE);\n");
                     fprintf(f, "        }\n");
                 }
diff --git a/vm_insnhelper.c b/vm_insnhelper.c
index b6ad18cd51..34ca736522 100644
--- a/vm_insnhelper.c
+++ b/vm_insnhelper.c
@@ -2010,7 +2010,7 @@ vm_base_ptr(const rb_control_frame_t *cfp)
 #include "vm_args.c"
 
 static inline VALUE vm_call_iseq_setup_2(rb_execution_context_t *ec, rb_control_frame_t *cfp, struct rb_calling_info *calling, struct rb_call_data *cd, int opt_pc, int param_size, int local_size);
-ALWAYS_INLINE(static VALUE vm_call_iseq_setup_normal(rb_execution_context_t *ec, rb_control_frame_t *cfp, struct rb_calling_info *calling, const rb_callable_method_entry_t *me, int opt_pc, int param_size, int local_size));
+ALWAYS_INLINE(static VALUE vm_call_iseq_setup_normal(rb_execution_context_t *ec, rb_control_frame_t *cfp, struct rb_calling_info *calling, const rb_callable_method_entry_t *me, int opt_pc, int param_size, int local_size, VALUE extra_type));
 static inline VALUE vm_call_iseq_setup_tailcall(rb_execution_context_t *ec, rb_control_frame_t *cfp, struct rb_calling_info *calling, struct rb_call_data *cd, int opt_pc);
 static VALUE vm_call_super_method(rb_execution_context_t *ec, rb_control_frame_t *reg_cfp, struct rb_calling_info *calling, struct rb_call_data *cd);
 static VALUE vm_call_method_nome(rb_execution_context_t *ec, rb_control_frame_t *cfp, struct rb_calling_info *calling, struct rb_call_data *cd);
@@ -2036,7 +2036,7 @@ vm_call_iseq_setup_normal_0start(rb_execution_context_t *ec, rb_control_frame_t
     const rb_iseq_t *iseq = def_iseq_ptr(vm_cc_cme(cc)->def);
     int param = iseq->body->param.size;
     int local = iseq->body->local_table_size;
-    return vm_call_iseq_setup_normal(ec, cfp, calling, vm_cc_cme(cc), 0, param, local);
+    return vm_call_iseq_setup_normal(ec, cfp, calling, vm_cc_cme(cc), 0, param, local, 0);
 }
 
 MJIT_STATIC bool
@@ -2189,7 +2189,7 @@ vm_call_iseq_setup_normal_opt_start(rb_execution_context_t *ec, rb_control_frame
     }
 #endif
 
-    return vm_call_iseq_setup_normal(ec, cfp, calling, vm_cc_cme(cc), opt_pc, param - delta, local);
+    return vm_call_iseq_setup_normal(ec, cfp, calling, vm_cc_cme(cc), opt_pc, param - delta, local, 0);
 }
 
 static VALUE
@@ -2247,7 +2247,7 @@ vm_call_iseq_setup_kwparm_kwarg(rb_execution_context_t *ec, rb_control_frame_t *
 
     int param = iseq->body->param.size;
     int local = iseq->body->local_table_size;
-    return vm_call_iseq_setup_normal(ec, cfp, calling, vm_cc_cme(cc), 0, param, local);
+    return vm_call_iseq_setup_normal(ec, cfp, calling, vm_cc_cme(cc), 0, param, local, 0);
 }
 
 static VALUE
@@ -2277,7 +2277,7 @@ vm_call_iseq_setup_kwparm_nokwarg(rb_execution_context_t *ec, rb_control_frame_t
 
     int param = iseq->body->param.size;
     int local = iseq->body->local_table_size;
-    return vm_call_iseq_setup_normal(ec, cfp, calling, vm_cc_cme(cc), 0, param, local);
+    return vm_call_iseq_setup_normal(ec, cfp, calling, vm_cc_cme(cc), 0, param, local, 0);
 }
 
 static inline int
@@ -2395,7 +2395,7 @@ vm_call_iseq_setup_2(rb_execution_context_t *ec, rb_control_frame_t *cfp, struct
     const struct rb_callcache *cc = cd->cc;
 
     if (LIKELY(!(vm_ci_flag(ci) & VM_CALL_TAILCALL))) {
-        return vm_call_iseq_setup_normal(ec, cfp, calling, vm_cc_cme(cc), opt_pc, param_size, local_size);
+        return vm_call_iseq_setup_normal(ec, cfp, calling, vm_cc_cme(cc), opt_pc, param_size, local_size, 0);
     }
     else {
         return vm_call_iseq_setup_tailcall(ec, cfp, calling, cd, opt_pc);
@@ -2404,14 +2404,14 @@ vm_call_iseq_setup_2(rb_execution_context_t *ec, rb_control_frame_t *cfp, struct
 
 static inline VALUE
 vm_call_iseq_setup_normal(rb_execution_context_t *ec, rb_control_frame_t *cfp, struct rb_calling_info *calling, const rb_callable_method_entry_t *me,
-                          int opt_pc, int param_size, int local_size)
+                          int opt_pc, int param_size, int local_size, VALUE extra_type)
 {
     const rb_iseq_t *iseq = def_iseq_ptr(me->def);
     VALUE *argv = cfp->sp - calling->argc;
     VALUE *sp = argv + param_size;
     cfp->sp = argv - 1 /* recv */;
 
-    vm_push_frame(ec, iseq, VM_FRAME_MAGIC_METHOD | VM_ENV_FLAG_LOCAL, calling->recv,
+    vm_push_frame(ec, iseq, VM_FRAME_MAGIC_METHOD | VM_ENV_FLAG_LOCAL | extra_type, calling->recv,
                   calling->block_handler, (VALUE)me,
                   iseq->body->iseq_encoded + opt_pc, sp,
                   local_size - param_size,