Re: [Qemu-devel] [PATCH v3 12/16] target-m68k: add fscale, fgetman, fgetexp and fmod

[Richard Henderson]

On 02/07/2017 11:59 AM, Laurent Vivier wrote:
> Signed-off-by: Laurent Vivier <address@hidden>
> ---
>  target/m68k/cpu.h        |  1 +
>  target/m68k/fpu_helper.c | 56 ++++++++++++++++++++++++++++++++++++++++++++++++
>  target/m68k/helper.h     |  4 ++++
>  target/m68k/translate.c  | 14 ++++++++++++
>  4 files changed, 75 insertions(+)
>
> diff --git a/target/m68k/cpu.h b/target/m68k/cpu.h
> index 7985dc3..3042ab7 100644
> --- a/target/m68k/cpu.h
> +++ b/target/m68k/cpu.h
> @@ -253,6 +253,7 @@ typedef enum {
>  /* Quotient */
>
>  #define FPSR_QT_MASK  0x00ff0000
> +#define FPSR_QT_SHIFT 16
>
>  /* Floating-Point Control Register */
>  /* Rounding mode */
> diff --git a/target/m68k/fpu_helper.c b/target/m68k/fpu_helper.c
> index d8145e0..42f5b5c 100644
> --- a/target/m68k/fpu_helper.c
> +++ b/target/m68k/fpu_helper.c
> @@ -458,3 +458,59 @@ void HELPER(const_FP0)(CPUM68KState *env, uint32_t offset)
>      env->fp0l = fpu_rom[offset].low;
>      env->fp0h = fpu_rom[offset].high;
>  }
> +
> +void HELPER(getexp_FP0)(CPUM68KState *env)
> +{
> +    int32_t exp;
> +    floatx80 res;
> +
> +    res = FP0_to_floatx80(env);
> +    if (floatx80_is_zero_or_denormal(res) || floatx80_is_any_nan(res) ||
> +        floatx80_is_infinity(res)) {
> +        return;
> +    }
> +    exp = (env->fp0h & 0x7fff) - 0x3fff;
> +
> +    res = int32_to_floatx80(exp, &env->fp_status);
> +
> +    floatx80_to_FP0(env, res);

Failure to raise OPERR for infinities?

> +void HELPER(getman_FP0)(CPUM68KState *env)
> +{
> +    floatx80 res;
> +    res = int64_to_floatx80(env->fp0l, &env->fp_status);
> +    floatx80_to_FP0(env, res);
> +}

This seems completely wrong.  (1) NaN gets returned, (2) Inf raises OPERR, (3) 
Normal values return something in the range [1.0 ... 2.0).  Which means you 
should just force the exponent rather than convert the low part.

> +
> +void HELPER(scale_FP0_FP1)(CPUM68KState *env)
> +{
> +    int32_t scale;
> +    int32_t exp;
> +
> +    scale = floatx80_to_int32(FP0_to_floatx80(env), &env->fp_status);
> +
> +    exp = (env->fp1h & 0x7fff) + scale;
> +
> +    env->fp0h = (env->fp1h & 0x8000) | (exp & 0x7fff);
> +    env->fp0l = env->fp1l;
> +}

Missing handling for NaN, Inf, 0, denormal.


r~