The mruby C style places `else` on its own line. Reformat the
remaining `} else {` / `} else if (...)` occurrences.
Co-authored-by: Claude <noreply@anthropic.com>
The old name referred to "truncation" of float precision, which no
longer happens with rotation encoding. The new name describes the
actual behavior: disabling inline float encoding in word boxing.
The old name is kept as an obsolete alias for backward compatibility.
Co-authored-by: Claude <noreply@anthropic.com>
On 32-bit with MRB_WORDBOX_NO_FLOAT_TRUNCATE, RFloat stores a double
(8-byte alignment) but GC heap slots only guarantee 4-byte alignment.
Use char array + memcpy accessors to avoid misaligned access (SIGBUS
on MIPS, undefined behavior per C standard).
Co-authored-by: Claude <noreply@anthropic.com>
replace lossy 2-bit truncation with rotation-based encoding for
32-bit + MRB_USE_FLOAT32, matching the technique used for 64-bit
float64. rotl32(bits - ADDEND, 3) maps biased exponents [95, 158]
(actual [-32, +31]) to properly tagged inline values with zero
precision loss. special values (0, Inf, NaN) use sentinel constants;
out-of-range floats fall back to heap-allocated RFloat.
also fix a pre-existing alignment issue: RVALUE was 20 bytes on
32-bit, causing 4-byte-aligned objects to be misidentified as
immediates by word boxing (WORDBOX_IMMEDIATE_MASK=0x07 requires
8-byte alignment). pad RVALUE to 24 bytes on 32-bit + float32.
Co-authored-by: Claude <noreply@anthropic.com>
Replace lossy 2-bit truncation with rotation-based encoding for
64-bit word boxing with float64. The new scheme uses
rotl64(float_bits - ADDEND, 3) to embed floats inline with full
52-bit mantissa precision. Floats with exponents outside [-255,+256]
(0.0, NaN, Inf, very small/large values) fall back to heap-allocated
RFloat.
Co-authored-by: Claude <noreply@anthropic.com>
This reverts commit e76bebe836 (#6309).
Because it crashes limited to gcc13 -O3.
ref. #6358.
Also, benchmark tests have shown that revert tends to be preferable in this time.
By adding a fast-path where we ignore boxed types we can gain a pretty substantial speedup of mrb_iv_get, making it about 25% faster during a standard optcarrot benchmark run.
NOTE: It is just mrb_iv_get that is that much faster, the whole benchmark seems to be about 3-5% faster with word boxing.
Adding `MRB_WORDBOX_NO_FLOAT_TRUNCATE` to the build configuration in 32-bit CPU mode had a double definition.
```console
% cat myconf.rb
MRuby::Build.new do
toolchain "clang"
defines << "MRB_WORDBOX_NO_FLOAT_TRUNCATE"
cc.flags << "-m32"
linker.flags << "-m32"
enable_debug
end
% rake CONFIG=myconf.rb
CPP src/array.c -> build/host/src/array.pi
In file included from /var/tmp/mruby/src/array.c:7:
In file included from /var/tmp/mruby/include/mruby.h:115:
In file included from /var/tmp/mruby/include/mruby/value.h:203:
/var/tmp/mruby/include/mruby/boxing_word.h:11:10: warning:
'MRB_WORDBOX_NO_FLOAT_TRUNCATE' macro redefined [-Wmacro-redefined]
# define MRB_WORDBOX_NO_FLOAT_TRUNCATE
^
<command line>:3:9: note: previous definition is here
#define MRB_WORDBOX_NO_FLOAT_TRUNCATE 1
^
1 warning generated.
...SNIP...
```
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.
- Integrate `Fixnum` and `Integer`
- Remove `Integral`
- `int / int -> int`
- Replace `mrb_fixnum()` to `mrb_int()`
- Replace `mrb_fixnum_value()` to `mrb_int_value()`.
- Use `mrb_integer_p()` instead of `mrb_fixnum_p()`
- `MRB_WITHOUT_FLOAT` => `MRB_NO_FLOAT`
- `MRB_USE_FLOAT` => `MRB_USE_FLOAT32`
The former is to use `USE_XXX` naming convention. The latter is to make
sure `float` is 32bit float and not floating point number in general.
Changes:
- `pool format is completely replaced
- supported types: `STR`, `INT32`, `INT64`, `FLOAT`
- `FLOAT` may be replaced by binary representation in the future
- insert `NUL` after string literals in `mrb` files
- `irep->pool` no longer store values in `mrb_value`
- instead it stores in `mrb_pool_value`
- less allocation
- `mrb_irep` can be stored in ROM
Previously, if `MRB_NAN_BOXING` is defined, for example,
`mrb_cptr_value()` could not keep an odd address.
If it is `MRB_32BIT`, it can be embedded in `NaN` as it is.
If it is `MRB_64BIT`, some operations are shared with `MRB_WORD_BOXING`.
In this case, the MRB_API function `mrb_nan_boxing_cptr_value()` is
defined.
The default implementations of type predicate macros use `mrb_type`.
But `mrb_type` with `MRB_WORD_BOXING` isn't very efficient, so the new
implementations avoid `mrb_type`.
The changes at b2c3d88f were inappropriate because the memory layout of bit
fields are implementation defined. Therefor, I fixed it not to use bit
fields.
The default implementation of `mrb_immediate_p` uses `mrb_type`. However,
in `MRB_WORD_BOXING`, `mrb_type` has many branches and is inefficient, so
provide an implementation that does not use `mrb_type`.
Yukihiro "Matz" Matsumoto