diff options
| author | Usama Arif <usama.arif@arm.com> | 2019-05-23 14:20:33 +0100 |
|---|---|---|
| committer | Usama Arif <usama.arif@arm.com> | 2019-05-24 16:11:16 +0000 |
| commit | 0a5a57a3f794de851408bae1c63b1660b4c5cbe7 (patch) | |
| tree | 5d49d85165a3d9fd257c5676691fc5ae24fda4f3 /arm_compute/core/NEON/NEMath.h | |
| parent | 43c1bf8e47098d6ab07be5b5e685bda6d8b80344 (diff) | |
| download | ComputeLibrary-0a5a57a3f794de851408bae1c63b1660b4c5cbe7.tar.gz | |
COMPMID-2160: Implement Round for NEON
Change-Id: Ie80e2ad294eaf95bc823d979842c320e8fb41f67
Signed-off-by: Usama Arif <usama.arif@arm.com>
Reviewed-on: https://review.mlplatform.org/c/1215
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'arm_compute/core/NEON/NEMath.h')
| -rw-r--r-- | arm_compute/core/NEON/NEMath.h | 20 |
1 files changed, 18 insertions, 2 deletions
diff --git a/arm_compute/core/NEON/NEMath.h b/arm_compute/core/NEON/NEMath.h index 5c60d73de4..46d97f6a0d 100644 --- a/arm_compute/core/NEON/NEMath.h +++ b/arm_compute/core/NEON/NEMath.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -36,6 +36,14 @@ namespace arm_compute */ float32x4_t vfloorq_f32(float32x4_t val); +/** Calculate round value of a vector to nearest with ties to even. + * + * @param[in] val Input vector value in F32 format. + * + * @return The calculated round vector. + */ +float32x4_t vroundq_rte_f32(float32x4_t val); + /** Calculate inverse square root. * * @param[in] x Input value. @@ -123,12 +131,20 @@ float32x4_t vpowq_f32(float32x4_t val, float32x4_t n); * * @note We clamp x to [-5,5] to avoid overflowing issues. * - * @param[in] val Input vector value in F32 format. + * @param[in] val Input vector value in F16 format. * * @return The calculated Hyperbolic Tangent. */ float16x8_t vtanhq_f16(float16x8_t val); +/** Calculate round value of a vector to nearest with ties to even. + * + * @param[in] val Input vector value in F16 format. + * + * @return The calculated round vector. + */ +float16x8_t vroundq_rte_f16(float16x8_t val); + /** Calculate reciprocal. * * @param[in] x Input value. |