diff options
| author | Emilio G. Cota <cota@braap.org> | 2018-03-17 00:30:40 -0400 |
|---|---|---|
| committer | Alex Bennée <alex.bennee@linaro.org> | 2018-12-17 08:25:25 +0000 |
| commit | f131bae8a7b7ed1928cc94c69df291db609c316a (patch) | |
| tree | 10c20d7f1abec552934fd03c05450f62dc9232da /fpu | |
| parent | ccf770ba7396c240ca8a1564740083742dd04c08 (diff) | |
hardfloat: implement float32/64 square root
Performance results for fp-bench:
Host: Intel(R) Core(TM) i7-6700K CPU @ 4.00GHz
- before:
sqrt-single: 42.30 MFlops
sqrt-double: 22.97 MFlops
- after:
sqrt-single: 311.42 MFlops
sqrt-double: 311.08 MFlops
Here USE_FP makes a huge difference for f64's, with throughput
going from ~200 MFlops to ~300 MFlops.
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Diffstat (limited to 'fpu')
| -rw-r--r-- | fpu/softfloat.c | 60 |
1 files changed, 58 insertions, 2 deletions
diff --git a/fpu/softfloat.c b/fpu/softfloat.c index 7554d63495..fbd66fd8dc 100644 --- a/fpu/softfloat.c +++ b/fpu/softfloat.c @@ -3044,20 +3044,76 @@ float16 QEMU_FLATTEN float16_sqrt(float16 a, float_status *status) return float16_round_pack_canonical(pr, status); } -float32 QEMU_FLATTEN float32_sqrt(float32 a, float_status *status) +static float32 QEMU_SOFTFLOAT_ATTR +soft_f32_sqrt(float32 a, float_status *status) { FloatParts pa = float32_unpack_canonical(a, status); FloatParts pr = sqrt_float(pa, status, &float32_params); return float32_round_pack_canonical(pr, status); } -float64 QEMU_FLATTEN float64_sqrt(float64 a, float_status *status) +static float64 QEMU_SOFTFLOAT_ATTR +soft_f64_sqrt(float64 a, float_status *status) { FloatParts pa = float64_unpack_canonical(a, status); FloatParts pr = sqrt_float(pa, status, &float64_params); return float64_round_pack_canonical(pr, status); } +float32 QEMU_FLATTEN float32_sqrt(float32 xa, float_status *s) +{ + union_float32 ua, ur; + + ua.s = xa; + if (unlikely(!can_use_fpu(s))) { + goto soft; + } + + float32_input_flush1(&ua.s, s); + if (QEMU_HARDFLOAT_1F32_USE_FP) { + if (unlikely(!(fpclassify(ua.h) == FP_NORMAL || + fpclassify(ua.h) == FP_ZERO) || + signbit(ua.h))) { + goto soft; + } + } else if (unlikely(!float32_is_zero_or_normal(ua.s) || + float32_is_neg(ua.s))) { + goto soft; + } + ur.h = sqrtf(ua.h); + return ur.s; + + soft: + return soft_f32_sqrt(ua.s, s); +} + +float64 QEMU_FLATTEN float64_sqrt(float64 xa, float_status *s) +{ + union_float64 ua, ur; + + ua.s = xa; + if (unlikely(!can_use_fpu(s))) { + goto soft; + } + + float64_input_flush1(&ua.s, s); + if (QEMU_HARDFLOAT_1F64_USE_FP) { + if (unlikely(!(fpclassify(ua.h) == FP_NORMAL || + fpclassify(ua.h) == FP_ZERO) || + signbit(ua.h))) { + goto soft; + } + } else if (unlikely(!float64_is_zero_or_normal(ua.s) || + float64_is_neg(ua.s))) { + goto soft; + } + ur.h = sqrt(ua.h); + return ur.s; + + soft: + return soft_f64_sqrt(ua.s, s); +} + /*---------------------------------------------------------------------------- | The pattern for a default generated NaN. *----------------------------------------------------------------------------*/ |