| /external/XNNPACK/src/f32-vsqrt/gen/ |
| D | avx512f-nr1fma1adj-x80.c | 34 const __m512 vx4 = _mm512_loadu_ps(x + 64); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x80() local
41 const __m512 vrsqrtx4 = _mm512_rsqrt14_ps(vx4); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x80()
51 __m512 vsqrtx4 = _mm512_mul_ps(vrsqrtx4, vx4); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x80()
75 const __m512 vadjustment4 = _mm512_fnmadd_ps(vsqrtx4, vsqrtx4, vx4); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x80()
|
| D | avx512f-nr1fma1adj-x96.c | 34 const __m512 vx4 = _mm512_loadu_ps(x + 64); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x96() local
42 const __m512 vrsqrtx4 = _mm512_rsqrt14_ps(vx4); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x96()
53 __m512 vsqrtx4 = _mm512_mul_ps(vrsqrtx4, vx4); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x96()
82 const __m512 vadjustment4 = _mm512_fnmadd_ps(vsqrtx4, vsqrtx4, vx4); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x96()
|
| D | fma3-nr1fma1adj-x40.c | 33 const __m256 vx4 = _mm256_loadu_ps(x + 32); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x40() local
40 const __m256 vrsqrtx4 = _mm256_rsqrt_ps(vx4); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x40()
50 __m256 vsqrtx4 = _mm256_mul_ps(vrsqrtx4, vx4); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x40()
74 const __m256 vadjustment4 = _mm256_fnmadd_ps(vsqrtx4, vsqrtx4, vx4); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x40()
|
| D | fma3-nr1fma1adj-x48.c | 33 const __m256 vx4 = _mm256_loadu_ps(x + 32); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x48() local
41 const __m256 vrsqrtx4 = _mm256_rsqrt_ps(vx4); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x48()
52 __m256 vsqrtx4 = _mm256_mul_ps(vrsqrtx4, vx4); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x48()
81 const __m256 vadjustment4 = _mm256_fnmadd_ps(vsqrtx4, vsqrtx4, vx4); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x48()
|
| D | avx512f-nr1fma1adj-x112.c | 34 const __m512 vx4 = _mm512_loadu_ps(x + 64); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x112() local
43 const __m512 vrsqrtx4 = _mm512_rsqrt14_ps(vx4); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x112()
55 __m512 vsqrtx4 = _mm512_mul_ps(vrsqrtx4, vx4); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x112()
89 const __m512 vadjustment4 = _mm512_fnmadd_ps(vsqrtx4, vsqrtx4, vx4); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x112()
|
| D | fma3-nr1fma1adj-x56.c | 33 const __m256 vx4 = _mm256_loadu_ps(x + 32); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x56() local
42 const __m256 vrsqrtx4 = _mm256_rsqrt_ps(vx4); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x56()
54 __m256 vsqrtx4 = _mm256_mul_ps(vrsqrtx4, vx4); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x56()
88 const __m256 vadjustment4 = _mm256_fnmadd_ps(vsqrtx4, vsqrtx4, vx4); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x56()
|
| D | avx512f-nr1fma1adj-x128.c | 34 const __m512 vx4 = _mm512_loadu_ps(x + 64); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x128() local
44 const __m512 vrsqrtx4 = _mm512_rsqrt14_ps(vx4); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x128()
57 __m512 vsqrtx4 = _mm512_mul_ps(vrsqrtx4, vx4); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x128()
96 const __m512 vadjustment4 = _mm512_fnmadd_ps(vsqrtx4, vsqrtx4, vx4); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x128()
|
| D | fma3-nr1fma1adj-x64.c | 33 const __m256 vx4 = _mm256_loadu_ps(x + 32); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x64() local
43 const __m256 vrsqrtx4 = _mm256_rsqrt_ps(vx4); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x64()
56 __m256 vsqrtx4 = _mm256_mul_ps(vrsqrtx4, vx4); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x64()
95 const __m256 vadjustment4 = _mm256_fnmadd_ps(vsqrtx4, vsqrtx4, vx4); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x64()
|
| /external/XNNPACK/src/f32-velu/gen/ |
| D | velu-avx2-rr1-lut8-p4-perm-x40.c | 43 __m256 vx4 = _mm256_loadu_ps(x + 32); in xnn_f32_velu_ukernel__avx2_rr1_lut8_p4_perm_x40() local
50 const __m256 vz4 = _mm256_max_ps(vsat_cutoff, _mm256_mul_ps(vx4, vprescale)); in xnn_f32_velu_ukernel__avx2_rr1_lut8_p4_perm_x40()
128 vx4 = _mm256_mul_ps(vx4, vbeta); in xnn_f32_velu_ukernel__avx2_rr1_lut8_p4_perm_x40()
134 const __m256 vy4 = _mm256_blendv_ps(vx4, ve4, vx4); in xnn_f32_velu_ukernel__avx2_rr1_lut8_p4_perm_x40()
|
| D | velu-avx2-rr1-p6-x40.c | 44 __m256 vx4 = _mm256_loadu_ps(x + 32); in xnn_f32_velu_ukernel__avx2_rr1_p6_x40() local
51 const __m256 vz4 = _mm256_max_ps(vsat_cutoff, _mm256_mul_ps(vx4, vprescale)); in xnn_f32_velu_ukernel__avx2_rr1_p6_x40()
131 vx4 = _mm256_mul_ps(vx4, vbeta); in xnn_f32_velu_ukernel__avx2_rr1_p6_x40()
137 const __m256 vy4 = _mm256_blendv_ps(vx4, ve4, vx4); in xnn_f32_velu_ukernel__avx2_rr1_p6_x40()
|
| D | velu-avx2-rr1-lut16-p3-gather-x40.c | 44 __m256 vx4 = _mm256_loadu_ps(x + 32); in xnn_f32_velu_ukernel__avx2_rr1_lut16_p3_gather_x40() local
51 const __m256 vz4 = _mm256_max_ps(vsat_cutoff, _mm256_mul_ps(vx4, vprescale)); in xnn_f32_velu_ukernel__avx2_rr1_lut16_p3_gather_x40()
129 vx4 = _mm256_mul_ps(vx4, vbeta); in xnn_f32_velu_ukernel__avx2_rr1_lut16_p3_gather_x40()
135 const __m256 vy4 = _mm256_blendv_ps(vx4, ve4, vx4); in xnn_f32_velu_ukernel__avx2_rr1_lut16_p3_gather_x40()
|
| D | velu-avx2-rr1-lut4-p4-perm-x40.c | 43 __m256 vx4 = _mm256_loadu_ps(x + 32); in xnn_f32_velu_ukernel__avx2_rr1_lut4_p4_perm_x40() local
50 const __m256 vz4 = _mm256_max_ps(vsat_cutoff, _mm256_mul_ps(vx4, vprescale)); in xnn_f32_velu_ukernel__avx2_rr1_lut4_p4_perm_x40()
128 vx4 = _mm256_mul_ps(vx4, vbeta); in xnn_f32_velu_ukernel__avx2_rr1_lut4_p4_perm_x40()
134 const __m256 vy4 = _mm256_blendv_ps(vx4, ve4, vx4); in xnn_f32_velu_ukernel__avx2_rr1_lut4_p4_perm_x40()
|
| D | velu-scalar-rr2-lut16-p3-x5.c | 46 float vx4 = x[4]; in xnn_f32_velu_ukernel__scalar_rr2_lut16_p3_x5() local
53 const float vz4 = vx4 * vprescale; in xnn_f32_velu_ukernel__scalar_rr2_lut16_p3_x5()
152 float vy4 = vx4 * vbeta; in xnn_f32_velu_ukernel__scalar_rr2_lut16_p3_x5()
166 if XNN_UNPREDICTABLE(vx4 < 0.0f) { in xnn_f32_velu_ukernel__scalar_rr2_lut16_p3_x5()
|
| D | velu-scalar-rr2-p6-x5.c | 46 float vx4 = x[4]; in xnn_f32_velu_ukernel__scalar_rr2_p6_x5() local
53 const float vz4 = vx4 * vprescale; in xnn_f32_velu_ukernel__scalar_rr2_p6_x5()
161 float vy4 = vx4 * vbeta; in xnn_f32_velu_ukernel__scalar_rr2_p6_x5()
175 if XNN_UNPREDICTABLE(vx4 < 0.0f) { in xnn_f32_velu_ukernel__scalar_rr2_p6_x5()
|
| D | velu-avx2-rr1-lut8-p4-perm-x48.c | 43 __m256 vx4 = _mm256_loadu_ps(x + 32); in xnn_f32_velu_ukernel__avx2_rr1_lut8_p4_perm_x48() local
51 const __m256 vz4 = _mm256_max_ps(vsat_cutoff, _mm256_mul_ps(vx4, vprescale)); in xnn_f32_velu_ukernel__avx2_rr1_lut8_p4_perm_x48()
142 vx4 = _mm256_mul_ps(vx4, vbeta); in xnn_f32_velu_ukernel__avx2_rr1_lut8_p4_perm_x48()
150 const __m256 vy4 = _mm256_blendv_ps(vx4, ve4, vx4); in xnn_f32_velu_ukernel__avx2_rr1_lut8_p4_perm_x48()
|
| D | velu-avx2-rr1-lut16-p3-gather-x48.c | 44 __m256 vx4 = _mm256_loadu_ps(x + 32); in xnn_f32_velu_ukernel__avx2_rr1_lut16_p3_gather_x48() local
52 const __m256 vz4 = _mm256_max_ps(vsat_cutoff, _mm256_mul_ps(vx4, vprescale)); in xnn_f32_velu_ukernel__avx2_rr1_lut16_p3_gather_x48()
143 vx4 = _mm256_mul_ps(vx4, vbeta); in xnn_f32_velu_ukernel__avx2_rr1_lut16_p3_gather_x48()
151 const __m256 vy4 = _mm256_blendv_ps(vx4, ve4, vx4); in xnn_f32_velu_ukernel__avx2_rr1_lut16_p3_gather_x48()
|
| D | velu-avx2-rr1-lut4-p4-perm-x48.c | 43 __m256 vx4 = _mm256_loadu_ps(x + 32); in xnn_f32_velu_ukernel__avx2_rr1_lut4_p4_perm_x48() local
51 const __m256 vz4 = _mm256_max_ps(vsat_cutoff, _mm256_mul_ps(vx4, vprescale)); in xnn_f32_velu_ukernel__avx2_rr1_lut4_p4_perm_x48()
142 vx4 = _mm256_mul_ps(vx4, vbeta); in xnn_f32_velu_ukernel__avx2_rr1_lut4_p4_perm_x48()
150 const __m256 vy4 = _mm256_blendv_ps(vx4, ve4, vx4); in xnn_f32_velu_ukernel__avx2_rr1_lut4_p4_perm_x48()
|
| D | velu-avx2-rr1-p6-x48.c | 44 __m256 vx4 = _mm256_loadu_ps(x + 32); in xnn_f32_velu_ukernel__avx2_rr1_p6_x48() local
52 const __m256 vz4 = _mm256_max_ps(vsat_cutoff, _mm256_mul_ps(vx4, vprescale)); in xnn_f32_velu_ukernel__avx2_rr1_p6_x48()
145 vx4 = _mm256_mul_ps(vx4, vbeta); in xnn_f32_velu_ukernel__avx2_rr1_p6_x48()
153 const __m256 vy4 = _mm256_blendv_ps(vx4, ve4, vx4); in xnn_f32_velu_ukernel__avx2_rr1_p6_x48()
|
| D | velu-scalar-rr2-lut16-p3-x6.c | 46 float vx4 = x[4]; in xnn_f32_velu_ukernel__scalar_rr2_lut16_p3_x6() local
54 const float vz4 = vx4 * vprescale; in xnn_f32_velu_ukernel__scalar_rr2_lut16_p3_x6()
170 float vy4 = vx4 * vbeta; in xnn_f32_velu_ukernel__scalar_rr2_lut16_p3_x6()
186 if XNN_UNPREDICTABLE(vx4 < 0.0f) { in xnn_f32_velu_ukernel__scalar_rr2_lut16_p3_x6()
|
| D | velu-avx2-rr1-lut16-p3-gather-x56.c | 44 __m256 vx4 = _mm256_loadu_ps(x + 32); in xnn_f32_velu_ukernel__avx2_rr1_lut16_p3_gather_x56() local
53 const __m256 vz4 = _mm256_max_ps(vsat_cutoff, _mm256_mul_ps(vx4, vprescale)); in xnn_f32_velu_ukernel__avx2_rr1_lut16_p3_gather_x56()
157 vx4 = _mm256_mul_ps(vx4, vbeta); in xnn_f32_velu_ukernel__avx2_rr1_lut16_p3_gather_x56()
167 const __m256 vy4 = _mm256_blendv_ps(vx4, ve4, vx4); in xnn_f32_velu_ukernel__avx2_rr1_lut16_p3_gather_x56()
|
| D | velu-avx2-rr1-p6-x56.c | 44 __m256 vx4 = _mm256_loadu_ps(x + 32); in xnn_f32_velu_ukernel__avx2_rr1_p6_x56() local
53 const __m256 vz4 = _mm256_max_ps(vsat_cutoff, _mm256_mul_ps(vx4, vprescale)); in xnn_f32_velu_ukernel__avx2_rr1_p6_x56()
159 vx4 = _mm256_mul_ps(vx4, vbeta); in xnn_f32_velu_ukernel__avx2_rr1_p6_x56()
169 const __m256 vy4 = _mm256_blendv_ps(vx4, ve4, vx4); in xnn_f32_velu_ukernel__avx2_rr1_p6_x56()
|
| D | velu-avx-rr2-p6-x40.c | 46 __m256 vx4 = _mm256_loadu_ps(x + 32); in xnn_f32_velu_ukernel__avx_rr2_p6_x40() local
53 const __m256 vz4 = _mm256_max_ps(vsat_cutoff, _mm256_mul_ps(vx4, vprescale)); in xnn_f32_velu_ukernel__avx_rr2_p6_x40()
150 vx4 = _mm256_mul_ps(vx4, vbeta); in xnn_f32_velu_ukernel__avx_rr2_p6_x40()
156 const __m256 vy4 = _mm256_blendv_ps(vx4, ve4, vx4); in xnn_f32_velu_ukernel__avx_rr2_p6_x40()
|
| D | velu-avx2-rr1-lut8-p4-perm-x56.c | 43 __m256 vx4 = _mm256_loadu_ps(x + 32); in xnn_f32_velu_ukernel__avx2_rr1_lut8_p4_perm_x56() local
52 const __m256 vz4 = _mm256_max_ps(vsat_cutoff, _mm256_mul_ps(vx4, vprescale)); in xnn_f32_velu_ukernel__avx2_rr1_lut8_p4_perm_x56()
156 vx4 = _mm256_mul_ps(vx4, vbeta); in xnn_f32_velu_ukernel__avx2_rr1_lut8_p4_perm_x56()
166 const __m256 vy4 = _mm256_blendv_ps(vx4, ve4, vx4); in xnn_f32_velu_ukernel__avx2_rr1_lut8_p4_perm_x56()
|
| D | velu-avx2-rr1-lut4-p4-perm-x56.c | 43 __m256 vx4 = _mm256_loadu_ps(x + 32); in xnn_f32_velu_ukernel__avx2_rr1_lut4_p4_perm_x56() local
52 const __m256 vz4 = _mm256_max_ps(vsat_cutoff, _mm256_mul_ps(vx4, vprescale)); in xnn_f32_velu_ukernel__avx2_rr1_lut4_p4_perm_x56()
156 vx4 = _mm256_mul_ps(vx4, vbeta); in xnn_f32_velu_ukernel__avx2_rr1_lut4_p4_perm_x56()
166 const __m256 vy4 = _mm256_blendv_ps(vx4, ve4, vx4); in xnn_f32_velu_ukernel__avx2_rr1_lut4_p4_perm_x56()
|
| /external/XNNPACK/src/x8-lut/gen/ |
| D | lut-scalar-x8.c | 31 const size_t vx4 = (size_t) x[4]; in xnn_x8_lut_ukernel__scalar_x8() local
41 const uint32_t vt4 = (uint32_t) t[vx4]; in xnn_x8_lut_ukernel__scalar_x8()
|