File size: 1,391 Bytes
8b7c501 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 |
// Auto-generated file. Do not edit!
// Template: src/f16-vbinary/vopc-fp16arith.c.in
// Generator: tools/xngen
//
// Copyright 2022 Google LLC
//
// This source code is licensed under the BSD-style license found in the
// LICENSE file in the root directory of this source tree.
#include <assert.h>
#include <string.h>
#include <arm_fp16.h>
#include <xnnpack/common.h>
#include <xnnpack/intrinsics-polyfill.h>
#include <xnnpack/math.h>
#include <xnnpack/vbinary.h>
void xnn_f16_vaddc_minmax_ukernel__fp16arith_x1(
size_t batch,
const void* restrict input_a,
const void* restrict input_b,
void* restrict output,
const union xnn_f16_minmax_params params[restrict XNN_MIN_ELEMENTS(1)])
{
assert(batch != 0);
assert(batch % sizeof(float16_t) == 0);
assert(input_a != NULL);
assert(input_b != NULL);
assert(output != NULL);
const float16_t* a = (const float16_t*) input_a;
const float16_t* b = (const float16_t*) input_b;
float16_t* o = (float16_t*) output;
float16_t vy_min, vy_max;
memcpy(&vy_min, ¶ms->fp16arith.min, sizeof(vy_min));
memcpy(&vy_max, ¶ms->fp16arith.max, sizeof(vy_max));
const float16_t vb = *b;
do {
float16_t vacc = *a++;
vacc = vaddh_f16(vacc, vb);
vacc = vmaxnmh_f16(vacc, vy_min);
vacc = vminnmh_f16(vacc, vy_max);
*o++ = vacc;
batch -= sizeof(float16_t);
} while (batch != 0);
}
|