| /external/XNNPACK/src/f32-vsigmoid/gen/ |
| D | vsigmoid-avx512f-rr2-lut32-p2-perm2-scalef-div-x112.c | 62 __m512 vn6 = _mm512_fmadd_ps(vz6, vlog2e, vmagic_bias); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_div_x112() local
70 const __m512 vl6 = _mm512_permutex2var_ps(vtable_lo, _mm512_castps_si512(vn6), vtable_hi); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_div_x112()
78 vn6 = _mm512_sub_ps(vn6, vmagic_bias); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_div_x112()
86 __m512 vt6 = _mm512_fmadd_ps(vn6, vminus_ln2_hi, vz6); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_div_x112()
94 vt6 = _mm512_fmadd_ps(vn6, vminus_ln2_lo, vt6); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_div_x112()
126 const __m512 ve6 = _mm512_scalef_ps(vp6, vn6); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_div_x112()
|
| D | vsigmoid-avx512f-rr2-lut32-p2-perm2-scalef-div-x128.c | 64 __m512 vn6 = _mm512_fmadd_ps(vz6, vlog2e, vmagic_bias); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_div_x128() local
73 const __m512 vl6 = _mm512_permutex2var_ps(vtable_lo, _mm512_castps_si512(vn6), vtable_hi); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_div_x128()
82 vn6 = _mm512_sub_ps(vn6, vmagic_bias); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_div_x128()
91 __m512 vt6 = _mm512_fmadd_ps(vn6, vminus_ln2_hi, vz6); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_div_x128()
100 vt6 = _mm512_fmadd_ps(vn6, vminus_ln2_lo, vt6); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_div_x128()
136 const __m512 ve6 = _mm512_scalef_ps(vp6, vn6); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_div_x128()
|
| D | vsigmoid-avx512f-rr2-lut32-p2-perm2-scalef-nr1fma-x112.c | 62 __m512 vn6 = _mm512_fmadd_ps(vz6, vlog2e, vmagic_bias); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_nr1fma_x112() local
70 const __m512 vl6 = _mm512_permutex2var_ps(vtable_lo, _mm512_castps_si512(vn6), vtable_hi); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_nr1fma_x112()
78 vn6 = _mm512_sub_ps(vn6, vmagic_bias); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_nr1fma_x112()
86 __m512 vt6 = _mm512_fmadd_ps(vn6, vminus_ln2_hi, vz6); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_nr1fma_x112()
94 vt6 = _mm512_fmadd_ps(vn6, vminus_ln2_lo, vt6); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_nr1fma_x112()
126 const __m512 ve6 = _mm512_scalef_ps(vp6, vn6); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_nr1fma_x112()
|
| D | vsigmoid-avx512f-rr1-lut16-p3-perm-scalef-div-x112.c | 60 __m512 vn6 = _mm512_fmadd_ps(vz6, vlog2e, vmagic_bias); in xnn_f32_vsigmoid_ukernel__avx512f_rr1_lut16_p3_perm_scalef_div_x112() local
68 const __m512 vl6 = _mm512_permutexvar_ps(_mm512_castps_si512(vn6), vtable); in xnn_f32_vsigmoid_ukernel__avx512f_rr1_lut16_p3_perm_scalef_div_x112()
76 vn6 = _mm512_sub_ps(vn6, vmagic_bias); in xnn_f32_vsigmoid_ukernel__avx512f_rr1_lut16_p3_perm_scalef_div_x112()
84 __m512 vt6 = _mm512_fmadd_ps(vn6, vminus_ln2, vz6); in xnn_f32_vsigmoid_ukernel__avx512f_rr1_lut16_p3_perm_scalef_div_x112()
124 const __m512 ve6 = _mm512_scalef_ps(vp6, vn6); in xnn_f32_vsigmoid_ukernel__avx512f_rr1_lut16_p3_perm_scalef_div_x112()
|
| D | vsigmoid-avx512f-rr2-lut32-p2-perm2-scalef-nr1fma-x128.c | 64 __m512 vn6 = _mm512_fmadd_ps(vz6, vlog2e, vmagic_bias); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_nr1fma_x128() local
73 const __m512 vl6 = _mm512_permutex2var_ps(vtable_lo, _mm512_castps_si512(vn6), vtable_hi); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_nr1fma_x128()
82 vn6 = _mm512_sub_ps(vn6, vmagic_bias); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_nr1fma_x128()
91 __m512 vt6 = _mm512_fmadd_ps(vn6, vminus_ln2_hi, vz6); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_nr1fma_x128()
100 vt6 = _mm512_fmadd_ps(vn6, vminus_ln2_lo, vt6); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_nr1fma_x128()
136 const __m512 ve6 = _mm512_scalef_ps(vp6, vn6); in xnn_f32_vsigmoid_ukernel__avx512f_rr2_lut32_p2_perm2_scalef_nr1fma_x128()
|
| D | vsigmoid-avx512f-rr1-lut16-p3-perm-scalef-nr1fma-x112.c | 60 __m512 vn6 = _mm512_fmadd_ps(vz6, vlog2e, vmagic_bias); in xnn_f32_vsigmoid_ukernel__avx512f_rr1_lut16_p3_perm_scalef_nr1fma_x112() local
68 const __m512 vl6 = _mm512_permutexvar_ps(_mm512_castps_si512(vn6), vtable); in xnn_f32_vsigmoid_ukernel__avx512f_rr1_lut16_p3_perm_scalef_nr1fma_x112()
76 vn6 = _mm512_sub_ps(vn6, vmagic_bias); in xnn_f32_vsigmoid_ukernel__avx512f_rr1_lut16_p3_perm_scalef_nr1fma_x112()
84 __m512 vt6 = _mm512_fmadd_ps(vn6, vminus_ln2, vz6); in xnn_f32_vsigmoid_ukernel__avx512f_rr1_lut16_p3_perm_scalef_nr1fma_x112()
124 const __m512 ve6 = _mm512_scalef_ps(vp6, vn6); in xnn_f32_vsigmoid_ukernel__avx512f_rr1_lut16_p3_perm_scalef_nr1fma_x112()
|
| D | vsigmoid-avx512f-rr1-lut16-p3-perm-scalef-div-x128.c | 62 __m512 vn6 = _mm512_fmadd_ps(vz6, vlog2e, vmagic_bias); in xnn_f32_vsigmoid_ukernel__avx512f_rr1_lut16_p3_perm_scalef_div_x128() local
71 const __m512 vl6 = _mm512_permutexvar_ps(_mm512_castps_si512(vn6), vtable); in xnn_f32_vsigmoid_ukernel__avx512f_rr1_lut16_p3_perm_scalef_div_x128()
80 vn6 = _mm512_sub_ps(vn6, vmagic_bias); in xnn_f32_vsigmoid_ukernel__avx512f_rr1_lut16_p3_perm_scalef_div_x128()
89 __m512 vt6 = _mm512_fmadd_ps(vn6, vminus_ln2, vz6); in xnn_f32_vsigmoid_ukernel__avx512f_rr1_lut16_p3_perm_scalef_div_x128()
134 const __m512 ve6 = _mm512_scalef_ps(vp6, vn6); in xnn_f32_vsigmoid_ukernel__avx512f_rr1_lut16_p3_perm_scalef_div_x128()
|
| D | vsigmoid-avx-rr2-p5-div-x56.c | 63 __m256 vn6 = _mm256_add_ps(_mm256_mul_ps(vz6, vlog2e), vmagic_bias); in xnn_f32_vsigmoid_ukernel__avx_rr2_p5_div_x56() local
83 …_m128 vs_lo6 = _mm_castsi128_ps(_mm_slli_epi32(_mm_castps_si128(_mm256_castps256_ps128(vn6)), 23)); in xnn_f32_vsigmoid_ukernel__avx_rr2_p5_div_x56()
84 …128 vs_hi6 = _mm_castsi128_ps(_mm_slli_epi32(_mm_castps_si128(_mm256_extractf128_ps(vn6, 1)), 23)); in xnn_f32_vsigmoid_ukernel__avx_rr2_p5_div_x56()
93 vn6 = _mm256_sub_ps(vn6, vmagic_bias); in xnn_f32_vsigmoid_ukernel__avx_rr2_p5_div_x56()
101 __m256 vt6 = _mm256_add_ps(_mm256_mul_ps(vn6, vminus_ln2_hi), vz6); in xnn_f32_vsigmoid_ukernel__avx_rr2_p5_div_x56()
109 vt6 = _mm256_add_ps(_mm256_mul_ps(vn6, vminus_ln2_lo), vt6); in xnn_f32_vsigmoid_ukernel__avx_rr2_p5_div_x56()
|
| /external/XNNPACK/src/f32-vscaleexpminusmax/gen/ |
| D | avx2-p5-x56.c | 72 __m256 vn6 = _mm256_fmadd_ps(vx6, vlog2e, vmagic_bias); in xnn_f32_vscaleexpminusmax_ukernel__avx2_p5_x56() local
82 const __m256 vs6 = _mm256_castsi256_ps(_mm256_slli_epi32(_mm256_castps_si256(vn6), 23)); in xnn_f32_vscaleexpminusmax_ukernel__avx2_p5_x56()
91 vn6 = _mm256_sub_ps(vn6, vmagic_bias); in xnn_f32_vscaleexpminusmax_ukernel__avx2_p5_x56()
101 __m256 vt6 = _mm256_fmadd_ps(vn6, vminus_ln2_hi, vx6); in xnn_f32_vscaleexpminusmax_ukernel__avx2_p5_x56()
109 vt6 = _mm256_fmadd_ps(vn6, vminus_ln2_lo, vt6); in xnn_f32_vscaleexpminusmax_ukernel__avx2_p5_x56()
|
| D | avx2-p5-x64.c | 74 __m256 vn6 = _mm256_fmadd_ps(vx6, vlog2e, vmagic_bias); in xnn_f32_vscaleexpminusmax_ukernel__avx2_p5_x64() local
85 const __m256 vs6 = _mm256_castsi256_ps(_mm256_slli_epi32(_mm256_castps_si256(vn6), 23)); in xnn_f32_vscaleexpminusmax_ukernel__avx2_p5_x64()
95 vn6 = _mm256_sub_ps(vn6, vmagic_bias); in xnn_f32_vscaleexpminusmax_ukernel__avx2_p5_x64()
106 __m256 vt6 = _mm256_fmadd_ps(vn6, vminus_ln2_hi, vx6); in xnn_f32_vscaleexpminusmax_ukernel__avx2_p5_x64()
115 vt6 = _mm256_fmadd_ps(vn6, vminus_ln2_lo, vt6); in xnn_f32_vscaleexpminusmax_ukernel__avx2_p5_x64()
|
| D | avx2-p5-x72.c | 76 __m256 vn6 = _mm256_fmadd_ps(vx6, vlog2e, vmagic_bias); in xnn_f32_vscaleexpminusmax_ukernel__avx2_p5_x72() local
88 const __m256 vs6 = _mm256_castsi256_ps(_mm256_slli_epi32(_mm256_castps_si256(vn6), 23)); in xnn_f32_vscaleexpminusmax_ukernel__avx2_p5_x72()
99 vn6 = _mm256_sub_ps(vn6, vmagic_bias); in xnn_f32_vscaleexpminusmax_ukernel__avx2_p5_x72()
111 __m256 vt6 = _mm256_fmadd_ps(vn6, vminus_ln2_hi, vx6); in xnn_f32_vscaleexpminusmax_ukernel__avx2_p5_x72()
121 vt6 = _mm256_fmadd_ps(vn6, vminus_ln2_lo, vt6); in xnn_f32_vscaleexpminusmax_ukernel__avx2_p5_x72()
|
| /external/XNNPACK/src/f32-raddexpminusmax/gen/ |
| D | avx2-p5-x64-acc2.c | 73 __m256 vn6 = _mm256_fmadd_ps(vx6, vlog2e, vmagic_bias); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x64_acc2() local
84 const __m256 vs6 = _mm256_castsi256_ps(_mm256_slli_epi32(_mm256_castps_si256(vn6), 23)); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x64_acc2()
94 vn6 = _mm256_sub_ps(vn6, vmagic_bias); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x64_acc2()
105 __m256 vt6 = _mm256_fmadd_ps(vn6, vminus_ln2_hi, vx6); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x64_acc2()
114 vt6 = _mm256_fmadd_ps(vn6, vminus_ln2_lo, vt6); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x64_acc2()
|
| D | avx2-p5-x64.c | 72 __m256 vn6 = _mm256_fmadd_ps(vx6, vlog2e, vmagic_bias); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x64() local
83 const __m256 vs6 = _mm256_castsi256_ps(_mm256_slli_epi32(_mm256_castps_si256(vn6), 23)); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x64()
93 vn6 = _mm256_sub_ps(vn6, vmagic_bias); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x64()
104 __m256 vt6 = _mm256_fmadd_ps(vn6, vminus_ln2_hi, vx6); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x64()
113 vt6 = _mm256_fmadd_ps(vn6, vminus_ln2_lo, vt6); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x64()
|
| D | avx2-p5-x64-acc4.c | 75 __m256 vn6 = _mm256_fmadd_ps(vx6, vlog2e, vmagic_bias); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x64_acc4() local
86 const __m256 vs6 = _mm256_castsi256_ps(_mm256_slli_epi32(_mm256_castps_si256(vn6), 23)); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x64_acc4()
96 vn6 = _mm256_sub_ps(vn6, vmagic_bias); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x64_acc4()
107 __m256 vt6 = _mm256_fmadd_ps(vn6, vminus_ln2_hi, vx6); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x64_acc4()
116 vt6 = _mm256_fmadd_ps(vn6, vminus_ln2_lo, vt6); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x64_acc4()
|
| D | avx2-p5-x72-acc3.c | 76 __m256 vn6 = _mm256_fmadd_ps(vx6, vlog2e, vmagic_bias); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x72_acc3() local
88 const __m256 vs6 = _mm256_castsi256_ps(_mm256_slli_epi32(_mm256_castps_si256(vn6), 23)); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x72_acc3()
99 vn6 = _mm256_sub_ps(vn6, vmagic_bias); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x72_acc3()
111 __m256 vt6 = _mm256_fmadd_ps(vn6, vminus_ln2_hi, vx6); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x72_acc3()
121 vt6 = _mm256_fmadd_ps(vn6, vminus_ln2_lo, vt6); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x72_acc3()
|
| D | avx2-p5-x72.c | 74 __m256 vn6 = _mm256_fmadd_ps(vx6, vlog2e, vmagic_bias); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x72() local
86 const __m256 vs6 = _mm256_castsi256_ps(_mm256_slli_epi32(_mm256_castps_si256(vn6), 23)); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x72()
97 vn6 = _mm256_sub_ps(vn6, vmagic_bias); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x72()
109 __m256 vt6 = _mm256_fmadd_ps(vn6, vminus_ln2_hi, vx6); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x72()
119 vt6 = _mm256_fmadd_ps(vn6, vminus_ln2_lo, vt6); in xnn_f32_raddexpminusmax_ukernel__avx2_p5_x72()
|
| /external/XNNPACK/src/f16-vsigmoid/gen/ |
| D | vsigmoid-neonfp16arith-rr2-p2-div-x56.c | 60 float16x8_t vn6 = vfmaq_f16(vmagic_bias, vz6, vminus_log2e); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x56() local
68 const float16x8_t vs6 = vreinterpretq_f16_s16(vshlq_n_s16(vreinterpretq_s16_f16(vn6), 10)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x56()
76 vn6 = vsubq_f16(vn6, vmagic_bias); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x56()
84 float16x8_t vt6 = vfmaq_f16(vz6, vn6, vln2_hi); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x56()
92 vt6 = vfmaq_f16(vt6, vn6, vln2_lo); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x56()
|
| D | vsigmoid-neonfp16arith-rr2-p2-div-x64.c | 62 float16x8_t vn6 = vfmaq_f16(vmagic_bias, vz6, vminus_log2e); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x64() local
71 const float16x8_t vs6 = vreinterpretq_f16_s16(vshlq_n_s16(vreinterpretq_s16_f16(vn6), 10)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x64()
80 vn6 = vsubq_f16(vn6, vmagic_bias); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x64()
89 float16x8_t vt6 = vfmaq_f16(vz6, vn6, vln2_hi); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x64()
98 vt6 = vfmaq_f16(vt6, vn6, vln2_lo); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x64()
|
| /external/XNNPACK/src/f16-raddstoreexpminusmax/gen/ |
| D | neonfp16arith-rr2-p2-x64-acc4.c | 68 float16x8_t vn6 = vfmaq_f16(vmagic_bias, vx6, vlog2e); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc4() local
77 const float16x8_t vs6 = vreinterpretq_f16_s16(vshlq_n_s16(vreinterpretq_s16_f16(vn6), 10)); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc4()
86 vn6 = vsubq_f16(vn6, vmagic_bias); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc4()
95 float16x8_t vt6 = vfmaq_f16(vx6, vn6, vminus_ln2_hi); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc4()
104 vt6 = vfmaq_f16(vt6, vn6, vminus_ln2_lo); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc4()
|
| D | neonfp16arith-rr2-p2-x64-acc2.c | 66 float16x8_t vn6 = vfmaq_f16(vmagic_bias, vx6, vlog2e); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc2() local
75 const float16x8_t vs6 = vreinterpretq_f16_s16(vshlq_n_s16(vreinterpretq_s16_f16(vn6), 10)); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc2()
84 vn6 = vsubq_f16(vn6, vmagic_bias); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc2()
93 float16x8_t vt6 = vfmaq_f16(vx6, vn6, vminus_ln2_hi); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc2()
102 vt6 = vfmaq_f16(vt6, vn6, vminus_ln2_lo); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc2()
|
| D | neonfp16arith-rr2-p2-x64.c | 65 float16x8_t vn6 = vfmaq_f16(vmagic_bias, vx6, vlog2e); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64() local
74 const float16x8_t vs6 = vreinterpretq_f16_s16(vshlq_n_s16(vreinterpretq_s16_f16(vn6), 10)); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64()
83 vn6 = vsubq_f16(vn6, vmagic_bias); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64()
92 float16x8_t vt6 = vfmaq_f16(vx6, vn6, vminus_ln2_hi); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64()
101 vt6 = vfmaq_f16(vt6, vn6, vminus_ln2_lo); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64()
|
| D | neonfp16arith-rr2-p2-x72.c | 67 float16x8_t vn6 = vfmaq_f16(vmagic_bias, vx6, vlog2e); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72() local
77 const float16x8_t vs6 = vreinterpretq_f16_s16(vshlq_n_s16(vreinterpretq_s16_f16(vn6), 10)); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72()
87 vn6 = vsubq_f16(vn6, vmagic_bias); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72()
97 float16x8_t vt6 = vfmaq_f16(vx6, vn6, vminus_ln2_hi); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72()
107 vt6 = vfmaq_f16(vt6, vn6, vminus_ln2_lo); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72()
|
| /external/XNNPACK/src/f32-velu/gen/ |
| D | velu-avx512f-rr1-lut16-p3-perm-x112.c | 63 __m512 vn6 = _mm512_fmadd_ps(vz6, vlog2e, vmagic_bias); in xnn_f32_velu_ukernel__avx512f_rr1_lut16_p3_perm_x112() local
77 const __m512i ven6 = _mm512_slli_epi32(_mm512_castps_si512(vn6), 19); in xnn_f32_velu_ukernel__avx512f_rr1_lut16_p3_perm_x112()
78 const __m512i vl6 = _mm512_permutexvar_epi32(_mm512_castps_si512(vn6), vtable); in xnn_f32_velu_ukernel__avx512f_rr1_lut16_p3_perm_x112()
93 vn6 = _mm512_sub_ps(vn6, vmagic_bias); in xnn_f32_velu_ukernel__avx512f_rr1_lut16_p3_perm_x112()
101 __m512 vt6 = _mm512_fmadd_ps(vn6, vminus_ln2, vz6); in xnn_f32_velu_ukernel__avx512f_rr1_lut16_p3_perm_x112()
|
| D | velu-avx2-rr1-lut16-p3-gather-x56.c | 63 __m256 vn6 = _mm256_fmadd_ps(vz6, vlog2e, vmagic_bias); in xnn_f32_velu_ukernel__avx2_rr1_lut16_p3_gather_x56() local
77 const __m256i vidx6 = _mm256_and_si256(_mm256_castps_si256(vn6), vindex_mask); in xnn_f32_velu_ukernel__avx2_rr1_lut16_p3_gather_x56()
92 const __m256i ven6 = _mm256_slli_epi32(_mm256_castps_si256(vn6), 19); in xnn_f32_velu_ukernel__avx2_rr1_lut16_p3_gather_x56()
93 vn6 = _mm256_sub_ps(vn6, vmagic_bias); in xnn_f32_velu_ukernel__avx2_rr1_lut16_p3_gather_x56()
108 __m256 vt6 = _mm256_fmadd_ps(vn6, vminus_ln2, vz6); in xnn_f32_velu_ukernel__avx2_rr1_lut16_p3_gather_x56()
|
| D | velu-avx2-rr1-lut8-p4-perm-x56.c | 62 __m256 vn6 = _mm256_fmadd_ps(vz6, vlog2e, vmagic_bias); in xnn_f32_velu_ukernel__avx2_rr1_lut8_p4_perm_x56() local
82 const __m256i ven6 = _mm256_slli_epi32(_mm256_castps_si256(vn6), 20); in xnn_f32_velu_ukernel__avx2_rr1_lut8_p4_perm_x56()
83 const __m256i vl6 = _mm256_permutevar8x32_epi32(vtable, _mm256_castps_si256(vn6)); in xnn_f32_velu_ukernel__avx2_rr1_lut8_p4_perm_x56()
84 vn6 = _mm256_sub_ps(vn6, vmagic_bias); in xnn_f32_velu_ukernel__avx2_rr1_lut8_p4_perm_x56()
99 __m256 vt6 = _mm256_fmadd_ps(vn6, vminus_ln2, vz6); in xnn_f32_velu_ukernel__avx2_rr1_lut8_p4_perm_x56()
|