From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:35788) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1f0hFI-0007P6-7b for qemu-devel@nongnu.org; Tue, 27 Mar 2018 01:34:09 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1f0hFE-0005OA-H7 for qemu-devel@nongnu.org; Tue, 27 Mar 2018 01:34:08 -0400 Received: from out5-smtp.messagingengine.com ([66.111.4.29]:34275) by eggs.gnu.org with esmtps (TLS1.0:DHE_RSA_AES_256_CBC_SHA1:32) (Exim 4.71) (envelope-from ) id 1f0hFE-0005NC-DC for qemu-devel@nongnu.org; Tue, 27 Mar 2018 01:34:04 -0400 From: "Emilio G. Cota" Date: Tue, 27 Mar 2018 01:33:58 -0400 Message-Id: <1522128840-498-13-git-send-email-cota@braap.org> In-Reply-To: <1522128840-498-1-git-send-email-cota@braap.org> References: <1522128840-498-1-git-send-email-cota@braap.org> Subject: [Qemu-devel] [PATCH v2 12/14] hardfloat: support float32/64 square root List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-devel@nongnu.org Cc: Aurelien Jarno , Peter Maydell , =?UTF-8?q?Alex=20Benn=C3=A9e?= , Laurent Vivier , Richard Henderson , Paolo Bonzini , Mark Cave-Ayland Performance results for fp-bench run under aarch64-linux-user on an Intel(R) Core(TM) i7-4790K CPU @ 4.00GHz host: - before: sqrt-single: 26.61 MFlops sqrt-double: 17.14 MFlops - after: sqrt-single: 95.06 MFlops sqrt-double: 89.05 MFlops Note that here we have a single implementation for both f32/f64. I tried the same trick we used before, but the results aren't as good: - w/ each using float32/64_is_normal or fpclassify etc.: sqrt-single: 95.50 MFlops sqrt-double: 84.55 MFlops - w/ both using fpclassify etc.: sqrt-single: 91.04 MFlops sqrt-double: 85.55 MFlops Signed-off-by: Emilio G. Cota --- fpu/softfloat.c | 26 ++++++++++++++++++++++++-- 1 file changed, 24 insertions(+), 2 deletions(-) diff --git a/fpu/softfloat.c b/fpu/softfloat.c index 2dedb13..ba7289b 100644 --- a/fpu/softfloat.c +++ b/fpu/softfloat.c @@ -2436,20 +2436,42 @@ float16 __attribute__((flatten)) float16_sqrt(float16 a, float_status *status) return float16_round_pack_canonical(pr, status); } -float32 __attribute__((flatten)) float32_sqrt(float32 a, float_status *status) +static float32 __attribute__((flatten, noinline)) +soft_float32_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 __attribute__((flatten)) float64_sqrt(float64 a, float_status *status) +static float64 __attribute__((flatten, noinline)) +soft_float64_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); } +#define GEN_FPU_SQRT(name, soft_t, host_t, host_sqrt_func) \ + soft_t name(soft_t a, float_status *s) \ + { \ + soft_t ## _input_flush1(&a, s); \ + if (likely((soft_t ## _is_normal(a) || soft_t ## _is_zero(a)) && \ + !soft_t ## _is_neg(a) && \ + s->float_exception_flags & float_flag_inexact && \ + s->float_rounding_mode == float_round_nearest_even)) { \ + host_t ha = soft_t ## _to_ ## host_t(a); \ + host_t hr = host_sqrt_func(ha); \ + \ + return host_t ## _to_ ## soft_t(hr); \ + } \ + return soft_ ## soft_t ## _sqrt(a, s); \ + } + +GEN_FPU_SQRT(float32_sqrt, float32, float, sqrtf) +GEN_FPU_SQRT(float64_sqrt, float64, double, sqrt) +#undef GEN_FPU_SQRT + /*---------------------------------------------------------------------------- | Takes a 64-bit fixed-point value `absZ' with binary point between bits 6 -- 2.7.4