|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
#include <assert.h> |
|
#include <string.h> |
|
|
|
#include <arm_fp16.h> |
|
|
|
#include <xnnpack/common.h> |
|
#include <xnnpack/intrinsics-polyfill.h> |
|
#include <xnnpack/vbinary.h> |
|
|
|
|
|
void xnn_f16_vadd_minmax_ukernel__fp16arith_x2( |
|
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)); |
|
|
|
for (; batch >= 2 * sizeof(float16_t); batch -= 2 * sizeof(float16_t)) { |
|
const float16_t va0 = *a++; |
|
const float16_t va1 = *a++; |
|
|
|
const float16_t vb0 = *b++; |
|
const float16_t vb1 = *b++; |
|
|
|
float16_t vacc0 = vaddh_f16(va0, vb0); |
|
float16_t vacc1 = vaddh_f16(va1, vb1); |
|
|
|
|
|
vacc0 = vmaxnmh_f16(vacc0, vy_min); |
|
vacc1 = vmaxnmh_f16(vacc1, vy_min); |
|
|
|
vacc0 = vminnmh_f16(vacc0, vy_max); |
|
vacc1 = vminnmh_f16(vacc1, vy_max); |
|
|
|
*o++ = vacc0; |
|
*o++ = vacc1; |
|
} |
|
if XNN_UNLIKELY(batch != 0) { |
|
const float16_t va = *a; |
|
const float16_t vb = *b; |
|
float16_t vacc = vaddh_f16(va, vb); |
|
vacc = vmaxnmh_f16(vacc, vy_min); |
|
vacc = vminnmh_f16(vacc, vy_max); |
|
*o = vacc; |
|
} |
|
} |
|
|