Richard Henderson <richard.hender...@linaro.org> writes:
> With a canonical representation of NaNs, we can silence an SNaN
> immediately rather than delay until the final format is known.
>
> Reviewed-by: Peter Maydell <peter.mayd...@linaro.org>
> Signed-off-by: Richard Henderson <richard.hender...@linaro.org>
Reviewed-by: Alex Bennée <alex.ben...@linaro.org>
> ---
> fpu/softfloat-specialize.h | 23 ++++++++++++++++++++++
> fpu/softfloat.c | 40 ++++++++++----------------------------
> 2 files changed, 33 insertions(+), 30 deletions(-)
>
> diff --git a/fpu/softfloat-specialize.h b/fpu/softfloat-specialize.h
> index 0d3d81a52b..571d1df378 100644
> --- a/fpu/softfloat-specialize.h
> +++ b/fpu/softfloat-specialize.h
> @@ -138,6 +138,29 @@ static FloatParts parts_default_nan(float_status *status)
> };
> }
>
> +/*----------------------------------------------------------------------------
> +| Returns a quiet NaN from a signalling NaN for the deconstructed
> +| floating-point parts.
> +*----------------------------------------------------------------------------*/
> +
> +static FloatParts parts_silence_nan(FloatParts a, float_status *status)
> +{
> +#ifdef NO_SIGNALING_NANS
> + g_assert_not_reached();
> +#elif defined(TARGET_HPPA)
> + a.frac &= ~(1ULL << (DECOMPOSED_BINARY_POINT - 1));
> + a.frac |= 1ULL << (DECOMPOSED_BINARY_POINT - 2);
> +#else
> + if (status->snan_bit_is_one) {
> + return parts_default_nan(status);
> + } else {
> + a.frac |= 1ULL << (DECOMPOSED_BINARY_POINT - 1);
> + }
> +#endif
> + a.cls = float_class_qnan;
> + return a;
> +}
> +
>
> /*----------------------------------------------------------------------------
> | The pattern for a default generated half-precision NaN.
>
> *----------------------------------------------------------------------------*/
> diff --git a/fpu/softfloat.c b/fpu/softfloat.c
> index 51780b718f..41253c6749 100644
> --- a/fpu/softfloat.c
> +++ b/fpu/softfloat.c
> @@ -188,7 +188,6 @@ typedef enum __attribute__ ((__packed__)) {
> float_class_inf,
> float_class_qnan, /* all NaNs from here */
> float_class_snan,
> - float_class_msnan, /* maybe silenced */
> } FloatClass;
>
> /*
> @@ -492,14 +491,7 @@ static FloatParts float16_unpack_canonical(float16 f,
> float_status *s)
>
> static float16 float16_round_pack_canonical(FloatParts p, float_status *s)
> {
> - switch (p.cls) {
> - case float_class_msnan:
> - p.frac >>= float16_params.frac_shift;
> - return float16_maybe_silence_nan(float16_pack_raw(p), s);
> - default:
> - p = round_canonical(p, s, &float16_params);
> - return float16_pack_raw(p);
> - }
> + return float16_pack_raw(round_canonical(p, s, &float16_params));
> }
>
> static FloatParts float32_unpack_canonical(float32 f, float_status *s)
> @@ -509,14 +501,7 @@ static FloatParts float32_unpack_canonical(float32 f,
> float_status *s)
>
> static float32 float32_round_pack_canonical(FloatParts p, float_status *s)
> {
> - switch (p.cls) {
> - case float_class_msnan:
> - p.frac >>= float32_params.frac_shift;
> - return float32_maybe_silence_nan(float32_pack_raw(p), s);
> - default:
> - p = round_canonical(p, s, &float32_params);
> - return float32_pack_raw(p);
> - }
> + return float32_pack_raw(round_canonical(p, s, &float32_params));
> }
>
> static FloatParts float64_unpack_canonical(float64 f, float_status *s)
> @@ -526,14 +511,7 @@ static FloatParts float64_unpack_canonical(float64 f,
> float_status *s)
>
> static float64 float64_round_pack_canonical(FloatParts p, float_status *s)
> {
> - switch (p.cls) {
> - case float_class_msnan:
> - p.frac >>= float64_params.frac_shift;
> - return float64_maybe_silence_nan(float64_pack_raw(p), s);
> - default:
> - p = round_canonical(p, s, &float64_params);
> - return float64_pack_raw(p);
> - }
> + return float64_pack_raw(round_canonical(p, s, &float64_params));
> }
>
> /* Simple helpers for checking if what NaN we have */
> @@ -555,7 +533,7 @@ static FloatParts return_nan(FloatParts a, float_status
> *s)
> switch (a.cls) {
> case float_class_snan:
> s->float_exception_flags |= float_flag_invalid;
> - a.cls = float_class_msnan;
> + a = parts_silence_nan(a, s);
> /* fall through */
> case float_class_qnan:
> if (s->default_nan_mode) {
> @@ -584,7 +562,9 @@ static FloatParts pick_nan(FloatParts a, FloatParts b,
> float_status *s)
> (a.frac == b.frac && a.sign < b.sign))) {
> a = b;
> }
> - a.cls = float_class_msnan;
> + if (is_snan(a.cls)) {
> + return parts_silence_nan(a, s);
> + }
> }
> return a;
> }
> @@ -624,8 +604,10 @@ static FloatParts pick_nan_muladd(FloatParts a,
> FloatParts b, FloatParts c,
> default:
> g_assert_not_reached();
> }
> - a.cls = float_class_msnan;
>
> + if (is_snan(a.cls)) {
> + return parts_silence_nan(a, s);
> + }
> return a;
> }
>
> @@ -1334,7 +1316,6 @@ static int64_t round_to_int_and_pack(FloatParts in, int
> rmode,
> switch (p.cls) {
> case float_class_snan:
> case float_class_qnan:
> - case float_class_msnan:
> s->float_exception_flags = orig_flags | float_flag_invalid;
> return max;
> case float_class_inf:
> @@ -1425,7 +1406,6 @@ static uint64_t round_to_uint_and_pack(FloatParts in,
> int rmode, uint64_t max,
> switch (p.cls) {
> case float_class_snan:
> case float_class_qnan:
> - case float_class_msnan:
> s->float_exception_flags = orig_flags | float_flag_invalid;
> return max;
> case float_class_inf:
--
Alex Bennée
