Folkert de Vries 73cddef788
optimize is_finite assembly (#169402)
Fixes https://github.com/llvm/llvm-project/issues/169270

Changes the implementation of `is_finite` to emit fewer instructions,
e.g.

X86_64

```asm
old: # 18 bytes
        movd    %xmm0, %eax
        andl    $2147483647, %eax
        cmpl    $2139095040, %eax
        setl    %al
        retq
new: # 15 bytes
        movd    %xmm0, %eax
        addl    %eax, %eax
        cmpl    $-16777216, %eax
        setb    %al
        retq
```

Aarch64

```asm
old:
        fmov    w9, s0
        mov     w8, #2139095040
        and     w9, w9, #0x7fffffff
        cmp     w9, w8
        cset    w0, lt
        ret
new:
        fmov    w8, s0
        ubfx    w8, w8, #23, #8
        cmp     w8, #255
        cset    w0, lo
        ret
```

See the issue for more information.
2026-03-29 14:28:07 +00:00
..
2025-05-22 15:42:22 +01:00