| /external/XNNPACK/src/f16-velu/gen/ |
| D | velu-neonfp16arith-rr1-p3-x8.c | 27 …const float16x8_t vprescale = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.pr… in xnn_f16_velu_ukernel__neonfp16arith_rr1_p3_x8()
28 …const float16x8_t vsat_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.… in xnn_f16_velu_ukernel__neonfp16arith_rr1_p3_x8()
29 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.… in xnn_f16_velu_ukernel__neonfp16arith_rr1_p3_x8()
30 …const float16x8_t vlog2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.log2e… in xnn_f16_velu_ukernel__neonfp16arith_rr1_p3_x8()
31 …const float16x8_t vminus_ln2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.m… in xnn_f16_velu_ukernel__neonfp16arith_rr1_p3_x8()
32 const float16x8_t vc3 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.c3)); in xnn_f16_velu_ukernel__neonfp16arith_rr1_p3_x8()
33 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.c2)); in xnn_f16_velu_ukernel__neonfp16arith_rr1_p3_x8()
34 …const float16x8_t vminus_alpha = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3… in xnn_f16_velu_ukernel__neonfp16arith_rr1_p3_x8()
35 …const float16x8_t vbeta = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.beta)); in xnn_f16_velu_ukernel__neonfp16arith_rr1_p3_x8()
|
| D | velu-neonfp16arith-rr1-p3-x16.c | 27 …const float16x8_t vprescale = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.pr… in xnn_f16_velu_ukernel__neonfp16arith_rr1_p3_x16()
28 …const float16x8_t vsat_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.… in xnn_f16_velu_ukernel__neonfp16arith_rr1_p3_x16()
29 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.… in xnn_f16_velu_ukernel__neonfp16arith_rr1_p3_x16()
30 …const float16x8_t vlog2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.log2e… in xnn_f16_velu_ukernel__neonfp16arith_rr1_p3_x16()
31 …const float16x8_t vminus_ln2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.m… in xnn_f16_velu_ukernel__neonfp16arith_rr1_p3_x16()
32 const float16x8_t vc3 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.c3)); in xnn_f16_velu_ukernel__neonfp16arith_rr1_p3_x16()
33 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.c2)); in xnn_f16_velu_ukernel__neonfp16arith_rr1_p3_x16()
34 …const float16x8_t vminus_alpha = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3… in xnn_f16_velu_ukernel__neonfp16arith_rr1_p3_x16()
35 …const float16x8_t vbeta = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.beta)); in xnn_f16_velu_ukernel__neonfp16arith_rr1_p3_x16()
|
| /external/XNNPACK/src/f16-vsigmoid/gen/ |
| D | vsigmoid-neonfp16arith-rr2-p2-div-x8.c | 26 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x8()
27 …const float16x8_t vminus_log2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x8()
28 …const float16x8_t vln2_hi = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x8()
29 …const float16x8_t vln2_lo = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x8()
30 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c2)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x8()
31 const float16x8_t vc1 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c1)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x8()
33 …const float16x8_t vdenorm_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x8()
|
| D | vsigmoid-neonfp16arith-rr2-p2-nr1fma-x8.c | 26 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x8()
27 …const float16x8_t vminus_log2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x8()
28 …const float16x8_t vln2_hi = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x8()
29 …const float16x8_t vln2_lo = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x8()
30 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c2)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x8()
31 const float16x8_t vc1 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c1)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x8()
33 …const float16x8_t vdenorm_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x8()
|
| D | vsigmoid-neonfp16arith-rr2-p2-nr1recps-x8.c | 26 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x8()
27 …const float16x8_t vminus_log2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x8()
28 …const float16x8_t vln2_hi = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x8()
29 …const float16x8_t vln2_lo = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x8()
30 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c2)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x8()
31 const float16x8_t vc1 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c1)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x8()
33 …const float16x8_t vdenorm_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x8()
|
| D | vsigmoid-neonfp16arith-rr2-p2-div-x16.c | 26 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x16()
27 …const float16x8_t vminus_log2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x16()
28 …const float16x8_t vln2_hi = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x16()
29 …const float16x8_t vln2_lo = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x16()
30 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c2)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x16()
31 const float16x8_t vc1 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c1)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x16()
33 …const float16x8_t vdenorm_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x16()
|
| D | vsigmoid-neonfp16arith-rr2-p2-nr1recps-x16.c | 26 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x16()
27 …const float16x8_t vminus_log2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x16()
28 …const float16x8_t vln2_hi = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x16()
29 …const float16x8_t vln2_lo = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x16()
30 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c2)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x16()
31 const float16x8_t vc1 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c1)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x16()
33 …const float16x8_t vdenorm_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x16()
|
| D | vsigmoid-neonfp16arith-rr2-p2-nr1fma-x16.c | 26 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x16()
27 …const float16x8_t vminus_log2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x16()
28 …const float16x8_t vln2_hi = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x16()
29 …const float16x8_t vln2_lo = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x16()
30 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c2)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x16()
31 const float16x8_t vc1 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c1)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x16()
33 …const float16x8_t vdenorm_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x16()
|
| D | vsigmoid-neonfp16arith-rr2-p2-div-x24.c | 26 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x24()
27 …const float16x8_t vminus_log2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x24()
28 …const float16x8_t vln2_hi = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x24()
29 …const float16x8_t vln2_lo = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x24()
30 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c2)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x24()
31 const float16x8_t vc1 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c1)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x24()
33 …const float16x8_t vdenorm_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x24()
|
| D | vsigmoid-neonfp16arith-rr2-p2-nr1recps-x24.c | 26 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x24()
27 …const float16x8_t vminus_log2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x24()
28 …const float16x8_t vln2_hi = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x24()
29 …const float16x8_t vln2_lo = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x24()
30 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c2)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x24()
31 const float16x8_t vc1 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c1)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x24()
33 …const float16x8_t vdenorm_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x24()
|
| D | vsigmoid-neonfp16arith-rr2-p2-nr1fma-x24.c | 26 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x24()
27 …const float16x8_t vminus_log2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x24()
28 …const float16x8_t vln2_hi = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x24()
29 …const float16x8_t vln2_lo = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x24()
30 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c2)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x24()
31 const float16x8_t vc1 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c1)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x24()
33 …const float16x8_t vdenorm_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x24()
|
| D | vsigmoid-neonfp16arith-rr2-p2-div-x32.c | 26 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x32()
27 …const float16x8_t vminus_log2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x32()
28 …const float16x8_t vln2_hi = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x32()
29 …const float16x8_t vln2_lo = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x32()
30 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c2)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x32()
31 const float16x8_t vc1 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c1)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x32()
33 …const float16x8_t vdenorm_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x32()
|
| D | vsigmoid-neonfp16arith-rr2-p2-nr1fma-x32.c | 26 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x32()
27 …const float16x8_t vminus_log2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x32()
28 …const float16x8_t vln2_hi = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x32()
29 …const float16x8_t vln2_lo = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x32()
30 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c2)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x32()
31 const float16x8_t vc1 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c1)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x32()
33 …const float16x8_t vdenorm_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x32()
|
| D | vsigmoid-neonfp16arith-rr2-p2-div-x40.c | 26 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x40()
27 …const float16x8_t vminus_log2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x40()
28 …const float16x8_t vln2_hi = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x40()
29 …const float16x8_t vln2_lo = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x40()
30 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c2)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x40()
31 const float16x8_t vc1 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c1)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x40()
33 …const float16x8_t vdenorm_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x40()
|
| D | vsigmoid-neonfp16arith-rr2-p2-nr1recps-x32.c | 26 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x32()
27 …const float16x8_t vminus_log2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x32()
28 …const float16x8_t vln2_hi = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x32()
29 …const float16x8_t vln2_lo = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.ln2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x32()
30 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c2)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x32()
31 const float16x8_t vc1 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c1)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x32()
33 …const float16x8_t vdenorm_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_… in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x32()
|
| /external/XNNPACK/src/f16-velu/ |
| D | neonfp16arith-rr1-p3.c.in | 26 …const float16x8_t vprescale = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.pr…
27 …const float16x8_t vsat_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.…
28 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.…
29 …const float16x8_t vlog2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.log2e…
30 …const float16x8_t vminus_ln2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.m…
31 const float16x8_t vc3 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.c3));
32 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.c2));
33 …const float16x8_t vminus_alpha = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3…
34 …const float16x8_t vbeta = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr1_p3.beta));
|
| /external/XNNPACK/src/f16-raddstoreexpminusmax/gen/ |
| D | neonfp16arith-rr2-p2-x32-acc2.c | 29 …const float16x8_t vlog2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.log2e… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32_acc2()
30 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32_acc2()
31 …const float16x8_t vminus_ln2_hi = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32_acc2()
32 …const float16x8_t vminus_ln2_lo = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32_acc2()
33 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c2)); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32_acc2()
34 const float16x8_t vc1 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c1)); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32_acc2()
35 …const float16x8_t vdenorm_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32_acc2()
|
| D | neonfp16arith-rr2-p2-x32.c | 29 …const float16x8_t vlog2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.log2e… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32()
30 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32()
31 …const float16x8_t vminus_ln2_hi = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32()
32 …const float16x8_t vminus_ln2_lo = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32()
33 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c2)); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32()
34 const float16x8_t vc1 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c1)); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32()
35 …const float16x8_t vdenorm_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32()
|
| D | neonfp16arith-rr2-p2-x40-acc2.c | 29 …const float16x8_t vlog2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.log2e… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc2()
30 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc2()
31 …const float16x8_t vminus_ln2_hi = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc2()
32 …const float16x8_t vminus_ln2_lo = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc2()
33 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c2)); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc2()
34 const float16x8_t vc1 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c1)); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc2()
35 …const float16x8_t vdenorm_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc2()
|
| D | neonfp16arith-rr2-p2-x32-acc4.c | 29 …const float16x8_t vlog2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.log2e… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32_acc4()
30 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32_acc4()
31 …const float16x8_t vminus_ln2_hi = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32_acc4()
32 …const float16x8_t vminus_ln2_lo = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32_acc4()
33 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c2)); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32_acc4()
34 const float16x8_t vc1 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c1)); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32_acc4()
35 …const float16x8_t vdenorm_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32_acc4()
|
| D | neonfp16arith-rr2-p2-x40.c | 29 …const float16x8_t vlog2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.log2e… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40()
30 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40()
31 …const float16x8_t vminus_ln2_hi = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40()
32 …const float16x8_t vminus_ln2_lo = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40()
33 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c2)); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40()
34 const float16x8_t vc1 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c1)); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40()
35 …const float16x8_t vdenorm_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40()
|
| D | neonfp16arith-rr2-p2-x48.c | 29 …const float16x8_t vlog2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.log2e… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48()
30 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48()
31 …const float16x8_t vminus_ln2_hi = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48()
32 …const float16x8_t vminus_ln2_lo = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48()
33 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c2)); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48()
34 const float16x8_t vc1 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c1)); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48()
35 …const float16x8_t vdenorm_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48()
|
| D | neonfp16arith-rr2-p2-x40-acc5.c | 29 …const float16x8_t vlog2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.log2e… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc5()
30 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc5()
31 …const float16x8_t vminus_ln2_hi = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc5()
32 …const float16x8_t vminus_ln2_lo = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc5()
33 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c2)); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc5()
34 const float16x8_t vc1 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c1)); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc5()
35 …const float16x8_t vdenorm_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_… in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc5()
|
| /external/XNNPACK/src/f16-raddstoreexpminusmax/ |
| D | neonfp16arith-rr2-p2.c.in | 29 …const float16x8_t vlog2e = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.log2e…
30 …const float16x8_t vmagic_bias = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.…
31 …const float16x8_t vminus_ln2_hi = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p…
32 …const float16x8_t vminus_ln2_lo = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p…
33 const float16x8_t vc2 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c2));
34 const float16x8_t vc1 = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_p2.c1));
35 …const float16x8_t vdenorm_cutoff = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neonfp16arith_rr2_…
|
| /external/XNNPACK/src/f16-vhswish/gen/ |
| D | vhswish-neonfp16arith-x8.c | 30 const float16x8_t vsixth = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neon.sixth)); in xnn_f16_vhswish_ukernel__neonfp16arith_x8()
31 const float16x8_t vthree = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->neon.three)); in xnn_f16_vhswish_ukernel__neonfp16arith_x8()
32 const int16x8_t vsix = vreinterpretq_s16_u16(vld1q_dup_u16(¶ms->neon.six)); in xnn_f16_vhswish_ukernel__neonfp16arith_x8()
|