Remove __bp__ and speed-up bmethod calls (#8060)

Remove rb_control_frame_t::__bp__ and optimize bmethod calls

This commit removes the __bp__ field from rb_control_frame_t. It was
introduced to help MJIT, but since MJIT was replaced by RJIT, we can use
vm_base_ptr() to compute it from the SP of the previous control frame
instead. Removing the field avoids needing to set it up when pushing new
frames.

Simply removing __bp__ would cause crashes since RJIT and YJIT used a
slightly different stack layout for bmethod calls than the interpreter.
At the moment of the call, the two layouts looked as follows:

                   ┌────────────┐    ┌────────────┐
                   │ frame_base │    │ frame_base │
                   ├────────────┤    ├────────────┤
                   │    ...     │    │    ...     │
                   ├────────────┤    ├────────────┤
                   │    args    │    │    args    │
                   ├────────────┤    └────────────┘<─prev_frame_sp
                   │  receiver  │
    prev_frame_sp─>└────────────┘
                     RJIT & YJIT      interpreter

Essentially, vm_base_ptr() needs to compute the address to frame_base
given prev_frame_sp in the diagrams. The presence of the receiver
created an off-by-one situation.

Make the interpreter use the layout the JITs use for iseq-to-iseq
bmethod calls. Doing so removes unnecessary argument shifting and
vm_exec_core() re-entry from the interpreter, yielding a speed
improvement visible through `benchmark/vm_defined_method.yml`:

     patched:   7578743.1 i/s
      master:   4796596.3 i/s - 1.58x  slower

C-to-iseq bmethod calls now store one more VALUE than before, but that
should have negligible impact on overall performance.

Note that re-entering vm_exec_core() used to be necessary for firing
TracePoint events, but that's no longer the case since
9121e57a5f50bc91bae48b3b91edb283bf96cb6b.

Closes ruby/ruby#6428

1 files changed, 13 insertions, 9 deletions

diff --git a/vm.c b/vm.c
index f8cdc5a5ea..c132f9a377 100644
--- a/vm.c
+++ b/vm.c
@@ -198,7 +198,7 @@ VM_CAPTURED_BLOCK_TO_CFP(const struct rb_captured_block *captured)
 {
     rb_control_frame_t *cfp = ((rb_control_frame_t *)((VALUE *)(captured) - 3));
     VM_ASSERT(!VM_CFP_IN_HEAP_P(GET_EC(), cfp));
-    VM_ASSERT(sizeof(rb_control_frame_t)/sizeof(VALUE) == 8 + VM_DEBUG_BP_CHECK ? 1 : 0);
+    VM_ASSERT(sizeof(rb_control_frame_t)/sizeof(VALUE) == 7 + VM_DEBUG_BP_CHECK ? 1 : 0);
     return cfp;
 }
 
@@ -1398,7 +1398,7 @@ invoke_block(rb_execution_context_t *ec, const rb_iseq_t *iseq, VALUE self, cons
 static VALUE
 invoke_bmethod(rb_execution_context_t *ec, const rb_iseq_t *iseq, VALUE self, const struct rb_captured_block *captured, const rb_callable_method_entry_t *me, VALUE type, int opt_pc)
 {
-    /* bmethod */
+    /* bmethod call from outside the VM */
     int arg_size = ISEQ_BODY(iseq)->param.size;
     VALUE ret;
 
@@ -1408,7 +1408,7 @@ invoke_bmethod(rb_execution_context_t *ec, const rb_iseq_t *iseq, VALUE self, co
                   VM_GUARDED_PREV_EP(captured->ep),
                   (VALUE)me,
                   ISEQ_BODY(iseq)->iseq_encoded + opt_pc,
-                  ec->cfp->sp + arg_size,
+                  ec->cfp->sp + 1 /* self */ + arg_size,
                   ISEQ_BODY(iseq)->local_table_size - arg_size,
                   ISEQ_BODY(iseq)->stack_max);
 
@@ -1429,7 +1429,7 @@ invoke_iseq_block_from_c(rb_execution_context_t *ec, const struct rb_captured_bl
                          const rb_cref_t *cref, int is_lambda, const rb_callable_method_entry_t *me)
 {
     const rb_iseq_t *iseq = rb_iseq_check(captured->code.iseq);
-    int i, opt_pc;
+    int opt_pc;
     VALUE type = VM_FRAME_MAGIC_BLOCK | (is_lambda ? VM_FRAME_FLAG_LAMBDA : 0);
     rb_control_frame_t *cfp = ec->cfp;
     VALUE *sp = cfp->sp;
@@ -1448,14 +1448,18 @@ invoke_iseq_block_from_c(rb_execution_context_t *ec, const struct rb_captured_bl
         use_argv = vm_argv_ruby_array(av, argv, &flags, &argc, kw_splat);
     }
 
-    CHECK_VM_STACK_OVERFLOW(cfp, argc);
+    CHECK_VM_STACK_OVERFLOW(cfp, argc + 1);
     vm_check_canary(ec, sp);
-    cfp->sp = sp + argc;
-    for (i=0; i<argc; i++) {
-        sp[i] = use_argv[i];
+
+    VALUE *stack_argv = sp;
+    if (me) {
+        *sp = self; // bemthods need `self` on the VM stack
+        stack_argv++;
     }
+    cfp->sp = stack_argv + argc;
+    MEMCPY(stack_argv, use_argv, VALUE, argc); // restrict: new stack space
 
-    opt_pc = vm_yield_setup_args(ec, iseq, argc, sp, flags, passed_block_handler,
+    opt_pc = vm_yield_setup_args(ec, iseq, argc, stack_argv, flags, passed_block_handler,
                                  (is_lambda ? arg_setup_method : arg_setup_block));
     cfp->sp = sp;