boxing_word.h: embed `mrb_float` in `mrb_value` if possible.

Embedding reduce memory consumption, sacrificing precision. It clips least
significant 2 bits from `mrb_float`, so if you need to keep float precision,
define `MRB_USE_FLOAT_FULL_PRECISION`.

`MRB_WORD_BOXING` and `MRB_INT64`:

`mrb_float` (`double`) is embedded in `mrb_value` clipped last 2 bits.

`MRB_WORD_BOXING` and `MRB_INT64` and `MRB_USE_FLOAT_FULL_PRECISION`:

`mrb_float` is allocated in the heaps wrapped by `struct RFloat`.

`MRB_WORD_BOXING` and `MRB_INT32` and `MRB_USE_FLOAT32`:

`mrb_float` (`float`) is embedded in `mrb_value` clipped last 2 bits.
In addition, to reserve bit space in the `mrb_value`, maximum inline
symbol length become 4 (instead of 5) in the configuration.

`MRB_WORD_BOXING` and `MRB_INT32`:

Assume `MRB_USE_FLOAT_FULL_PRECISION` and allocate Float values in heap.

1 files changed, 17 insertions, 0 deletions

diff --git a/src/etc.c b/src/etc.c
index a5c48bbe3..4d08ba400 100644
--- a/src/etc.c
+++ b/src/etc.c
@@ -158,11 +158,28 @@ mrb_word_boxing_float_value(mrb_state *mrb, mrb_float f)
 {
   union mrb_value_ v;
 
+#ifndef MRB_USE_FLOAT_FULL_PRECISION
+  v.f = f;
+  v.w = (v.w & ~3) | 2;
+#else
   v.p = mrb_obj_alloc(mrb, MRB_TT_FLOAT, mrb->float_class);
   v.fp->f = f;
   MRB_SET_FROZEN_FLAG(v.bp);
+#endif
   return v.value;
 }
+
+
+#ifndef MRB_USE_FLOAT_FULL_PRECISION
+MRB_API mrb_float
+mrb_word_boxing_value_float(mrb_value v)
+{
+  union mrb_value_ u;
+  u.value = v;
+  u.w = u.w & ~3;
+  return u.f;
+}
+#endif
 #endif  /* MRB_NO_FLOAT */
 
 MRB_API mrb_value