| /external/XNNPACK/src/f32-vsigmoid/gen/ |
| D | vsigmoid-scalar-rr2-lut2048-p1-div-x4.c | 48 const float vz3 = fabsf(vx3); in xnn_f32_vsigmoid_ukernel__scalar_rr2_lut2048_p1_div_x4() local
53 float vn3 = vz3 * vminus_log2e + vmagic_bias; in xnn_f32_vsigmoid_ukernel__scalar_rr2_lut2048_p1_div_x4()
77 float vt3 = vn3 * vln2_hi + vz3; in xnn_f32_vsigmoid_ukernel__scalar_rr2_lut2048_p1_div_x4()
113 if XNN_UNPREDICTABLE(vz3 > vdenorm_cutoff) { in xnn_f32_vsigmoid_ukernel__scalar_rr2_lut2048_p1_div_x4()
|
| D | vsigmoid-scalar-rr2-p5-div-x4.c | 48 const float vz3 = fabsf(vx3); in xnn_f32_vsigmoid_ukernel__scalar_rr2_p5_div_x4() local
53 float vn3 = vz3 * vminus_log2e + vmagic_bias; in xnn_f32_vsigmoid_ukernel__scalar_rr2_p5_div_x4()
68 float vt3 = vn3 * vln2_hi + vz3; in xnn_f32_vsigmoid_ukernel__scalar_rr2_p5_div_x4()
124 if XNN_UNPREDICTABLE(vz3 > vdenorm_cutoff) { in xnn_f32_vsigmoid_ukernel__scalar_rr2_p5_div_x4()
|
| D | vsigmoid-scalar-rr2-lut64-p2-div-x4.c | 48 const float vz3 = fabsf(vx3); in xnn_f32_vsigmoid_ukernel__scalar_rr2_lut64_p2_div_x4() local
53 float vn3 = vz3 * vminus_log2e + vmagic_bias; in xnn_f32_vsigmoid_ukernel__scalar_rr2_lut64_p2_div_x4()
77 float vt3 = vn3 * vln2_hi + vz3; in xnn_f32_vsigmoid_ukernel__scalar_rr2_lut64_p2_div_x4()
118 if XNN_UNPREDICTABLE(vz3 > vdenorm_cutoff) { in xnn_f32_vsigmoid_ukernel__scalar_rr2_lut64_p2_div_x4()
|
| D | vsigmoid-avx2-rr1-p5-div-x32.c | 48 const __m256 vz3 = _mm256_or_ps(vx3, vsign_mask); in xnn_f32_vsigmoid_ukernel__avx2_rr1_p5_div_x32() local
53 __m256 vn3 = _mm256_fmadd_ps(vz3, vlog2e, vmagic_bias); in xnn_f32_vsigmoid_ukernel__avx2_rr1_p5_div_x32()
68 __m256 vt3 = _mm256_fmadd_ps(vn3, vminus_ln2, vz3); in xnn_f32_vsigmoid_ukernel__avx2_rr1_p5_div_x32()
113 vf3 = _mm256_andnot_ps(_mm256_cmp_ps(vz3, vdenorm_cutoff, _CMP_LT_OS), vf3); in xnn_f32_vsigmoid_ukernel__avx2_rr1_p5_div_x32()
|
| D | vsigmoid-avx2-rr1-p5-nr2fma-x32.c | 48 const __m256 vz3 = _mm256_or_ps(vx3, vsign_mask); in xnn_f32_vsigmoid_ukernel__avx2_rr1_p5_nr2fma_x32() local
53 __m256 vn3 = _mm256_fmadd_ps(vz3, vlog2e, vmagic_bias); in xnn_f32_vsigmoid_ukernel__avx2_rr1_p5_nr2fma_x32()
68 __m256 vt3 = _mm256_fmadd_ps(vn3, vminus_ln2, vz3); in xnn_f32_vsigmoid_ukernel__avx2_rr1_p5_nr2fma_x32()
128 vf3 = _mm256_andnot_ps(_mm256_cmp_ps(vz3, vdenorm_cutoff, _CMP_LT_OS), vf3); in xnn_f32_vsigmoid_ukernel__avx2_rr1_p5_nr2fma_x32()
|
| D | vsigmoid-avx2-rr1-p5-div-x40.c | 49 const __m256 vz3 = _mm256_or_ps(vx3, vsign_mask); in xnn_f32_vsigmoid_ukernel__avx2_rr1_p5_div_x40() local
55 __m256 vn3 = _mm256_fmadd_ps(vz3, vlog2e, vmagic_bias); in xnn_f32_vsigmoid_ukernel__avx2_rr1_p5_div_x40()
73 __m256 vt3 = _mm256_fmadd_ps(vn3, vminus_ln2, vz3); in xnn_f32_vsigmoid_ukernel__avx2_rr1_p5_div_x40()
127 vf3 = _mm256_andnot_ps(_mm256_cmp_ps(vz3, vdenorm_cutoff, _CMP_LT_OS), vf3); in xnn_f32_vsigmoid_ukernel__avx2_rr1_p5_div_x40()
|
| D | vsigmoid-avx2-rr1-p5-nr1fma-x32.c | 48 const __m256 vz3 = _mm256_or_ps(vx3, vsign_mask); in xnn_f32_vsigmoid_ukernel__avx2_rr1_p5_nr1fma_x32() local
53 __m256 vn3 = _mm256_fmadd_ps(vz3, vlog2e, vmagic_bias); in xnn_f32_vsigmoid_ukernel__avx2_rr1_p5_nr1fma_x32()
68 __m256 vt3 = _mm256_fmadd_ps(vn3, vminus_ln2, vz3); in xnn_f32_vsigmoid_ukernel__avx2_rr1_p5_nr1fma_x32()
124 vf3 = _mm256_andnot_ps(_mm256_cmp_ps(vz3, vdenorm_cutoff, _CMP_LT_OS), vf3); in xnn_f32_vsigmoid_ukernel__avx2_rr1_p5_nr1fma_x32()
|
| /external/XNNPACK/src/f32-velu/gen/ |
| D | velu-scalar-rr2-lut16-p3-x4.c | 51 const float vz3 = vx3 * vprescale; in xnn_f32_velu_ukernel__scalar_rr2_lut16_p3_x4() local
56 float vn3 = vz3 * vlog2e + vmagic_bias; in xnn_f32_velu_ukernel__scalar_rr2_lut16_p3_x4()
77 float vt3 = vn3 * vminus_ln2_hi + vz3; in xnn_f32_velu_ukernel__scalar_rr2_lut16_p3_x4()
96 if XNN_UNPREDICTABLE(vz3 <= vsat_cutoff) { in xnn_f32_velu_ukernel__scalar_rr2_lut16_p3_x4()
|
| D | velu-scalar-rr2-p6-x4.c | 51 const float vz3 = vx3 * vprescale; in xnn_f32_velu_ukernel__scalar_rr2_p6_x4() local
56 float vn3 = vz3 * vlog2e + vmagic_bias; in xnn_f32_velu_ukernel__scalar_rr2_p6_x4()
70 float vt3 = vn3 * vminus_ln2_hi + vz3; in xnn_f32_velu_ukernel__scalar_rr2_p6_x4()
89 if XNN_UNPREDICTABLE(vz3 <= vsat_cutoff) { in xnn_f32_velu_ukernel__scalar_rr2_p6_x4()
|
| D | velu-scalar-rr2-lut16-p3-x5.c | 52 const float vz3 = vx3 * vprescale; in xnn_f32_velu_ukernel__scalar_rr2_lut16_p3_x5() local
58 float vn3 = vz3 * vlog2e + vmagic_bias; in xnn_f32_velu_ukernel__scalar_rr2_lut16_p3_x5()
83 float vt3 = vn3 * vminus_ln2_hi + vz3; in xnn_f32_velu_ukernel__scalar_rr2_lut16_p3_x5()
104 if XNN_UNPREDICTABLE(vz3 <= vsat_cutoff) { in xnn_f32_velu_ukernel__scalar_rr2_lut16_p3_x5()
|
| D | velu-scalar-rr2-p6-x5.c | 52 const float vz3 = vx3 * vprescale; in xnn_f32_velu_ukernel__scalar_rr2_p6_x5() local
58 float vn3 = vz3 * vlog2e + vmagic_bias; in xnn_f32_velu_ukernel__scalar_rr2_p6_x5()
75 float vt3 = vn3 * vminus_ln2_hi + vz3; in xnn_f32_velu_ukernel__scalar_rr2_p6_x5()
96 if XNN_UNPREDICTABLE(vz3 <= vsat_cutoff) { in xnn_f32_velu_ukernel__scalar_rr2_p6_x5()
|
| D | velu-scalar-rr2-lut16-p3-x6.c | 53 const float vz3 = vx3 * vprescale; in xnn_f32_velu_ukernel__scalar_rr2_lut16_p3_x6() local
60 float vn3 = vz3 * vlog2e + vmagic_bias; in xnn_f32_velu_ukernel__scalar_rr2_lut16_p3_x6()
89 float vt3 = vn3 * vminus_ln2_hi + vz3; in xnn_f32_velu_ukernel__scalar_rr2_lut16_p3_x6()
112 if XNN_UNPREDICTABLE(vz3 <= vsat_cutoff) { in xnn_f32_velu_ukernel__scalar_rr2_lut16_p3_x6()
|
| D | velu-scalar-rr2-p6-x6.c | 53 const float vz3 = vx3 * vprescale; in xnn_f32_velu_ukernel__scalar_rr2_p6_x6() local
60 float vn3 = vz3 * vlog2e + vmagic_bias; in xnn_f32_velu_ukernel__scalar_rr2_p6_x6()
80 float vt3 = vn3 * vminus_ln2_hi + vz3; in xnn_f32_velu_ukernel__scalar_rr2_p6_x6()
103 if XNN_UNPREDICTABLE(vz3 <= vsat_cutoff) { in xnn_f32_velu_ukernel__scalar_rr2_p6_x6()
|
| D | velu-wasm-rr2-lut16-p3-x4.c | 51 …const float vz3 = __builtin_wasm_min_f32(__builtin_wasm_max_f32(vx3 * vprescale, vsat_cutoff), 0.0… in xnn_f32_velu_ukernel__wasm_rr2_lut16_p3_x4() local
56 float vn3 = vz3 * vlog2e + vmagic_bias; in xnn_f32_velu_ukernel__wasm_rr2_lut16_p3_x4()
77 float vt3 = vn3 * vminus_ln2_hi + vz3; in xnn_f32_velu_ukernel__wasm_rr2_lut16_p3_x4()
|
| D | velu-wasm-rr2-p6-x4.c | 51 …const float vz3 = __builtin_wasm_min_f32(__builtin_wasm_max_f32(vx3 * vprescale, vsat_cutoff), 0.0… in xnn_f32_velu_ukernel__wasm_rr2_p6_x4() local
56 float vn3 = vz3 * vlog2e + vmagic_bias; in xnn_f32_velu_ukernel__wasm_rr2_p6_x4()
70 float vt3 = vn3 * vminus_ln2_hi + vz3; in xnn_f32_velu_ukernel__wasm_rr2_p6_x4()
|
| D | velu-wasm-rr2-lut16-p3-x5.c | 52 …const float vz3 = __builtin_wasm_min_f32(__builtin_wasm_max_f32(vx3 * vprescale, vsat_cutoff), 0.0… in xnn_f32_velu_ukernel__wasm_rr2_lut16_p3_x5() local
58 float vn3 = vz3 * vlog2e + vmagic_bias; in xnn_f32_velu_ukernel__wasm_rr2_lut16_p3_x5()
83 float vt3 = vn3 * vminus_ln2_hi + vz3; in xnn_f32_velu_ukernel__wasm_rr2_lut16_p3_x5()
|
| D | velu-wasm-rr2-p6-x5.c | 52 …const float vz3 = __builtin_wasm_min_f32(__builtin_wasm_max_f32(vx3 * vprescale, vsat_cutoff), 0.0… in xnn_f32_velu_ukernel__wasm_rr2_p6_x5() local
58 float vn3 = vz3 * vlog2e + vmagic_bias; in xnn_f32_velu_ukernel__wasm_rr2_p6_x5()
75 float vt3 = vn3 * vminus_ln2_hi + vz3; in xnn_f32_velu_ukernel__wasm_rr2_p6_x5()
|
| D | velu-wasm-rr2-lut16-p3-x6.c | 53 …const float vz3 = __builtin_wasm_min_f32(__builtin_wasm_max_f32(vx3 * vprescale, vsat_cutoff), 0.0… in xnn_f32_velu_ukernel__wasm_rr2_lut16_p3_x6() local
60 float vn3 = vz3 * vlog2e + vmagic_bias; in xnn_f32_velu_ukernel__wasm_rr2_lut16_p3_x6()
89 float vt3 = vn3 * vminus_ln2_hi + vz3; in xnn_f32_velu_ukernel__wasm_rr2_lut16_p3_x6()
|
| D | velu-wasm-rr2-p6-x6.c | 53 …const float vz3 = __builtin_wasm_min_f32(__builtin_wasm_max_f32(vx3 * vprescale, vsat_cutoff), 0.0… in xnn_f32_velu_ukernel__wasm_rr2_p6_x6() local
60 float vn3 = vz3 * vlog2e + vmagic_bias; in xnn_f32_velu_ukernel__wasm_rr2_p6_x6()
80 float vt3 = vn3 * vminus_ln2_hi + vz3; in xnn_f32_velu_ukernel__wasm_rr2_p6_x6()
|
| /external/XNNPACK/src/f16-vsigmoid/gen/ |
| D | vsigmoid-avx2-rr1-p2-div-x32.c | 48 const __m256 vz3 = _mm256_or_ps(vx3, vsign_mask); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_div_x32() local
53 __m256 vn3 = _mm256_fmadd_ps(vz3, vlog2e, vmagic_bias); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_div_x32()
68 __m256 vt3 = _mm256_fmadd_ps(vn3, vminus_ln2, vz3); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_div_x32()
98 vf3 = _mm256_andnot_ps(_mm256_cmp_ps(vz3, vdenorm_cutoff, _CMP_LT_OS), vf3); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_div_x32()
|
| D | vsigmoid-avx2-rr1-p2-rcp-x32.c | 48 const __m256 vz3 = _mm256_or_ps(vx3, vsign_mask); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_rcp_x32() local
53 __m256 vn3 = _mm256_fmadd_ps(vz3, vlog2e, vmagic_bias); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_rcp_x32()
68 __m256 vt3 = _mm256_fmadd_ps(vn3, vminus_ln2, vz3); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_rcp_x32()
103 vf3 = _mm256_andnot_ps(_mm256_cmp_ps(vz3, vdenorm_cutoff, _CMP_LT_OS), vf3); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_rcp_x32()
|
| D | vsigmoid-avx2-rr1-p2-div-x40.c | 49 const __m256 vz3 = _mm256_or_ps(vx3, vsign_mask); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_div_x40() local
55 __m256 vn3 = _mm256_fmadd_ps(vz3, vlog2e, vmagic_bias); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_div_x40()
73 __m256 vt3 = _mm256_fmadd_ps(vn3, vminus_ln2, vz3); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_div_x40()
109 vf3 = _mm256_andnot_ps(_mm256_cmp_ps(vz3, vdenorm_cutoff, _CMP_LT_OS), vf3); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_div_x40()
|
| D | vsigmoid-avx2-rr1-p2-rcp-x40.c | 49 const __m256 vz3 = _mm256_or_ps(vx3, vsign_mask); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_rcp_x40() local
55 __m256 vn3 = _mm256_fmadd_ps(vz3, vlog2e, vmagic_bias); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_rcp_x40()
73 __m256 vt3 = _mm256_fmadd_ps(vn3, vminus_ln2, vz3); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_rcp_x40()
115 vf3 = _mm256_andnot_ps(_mm256_cmp_ps(vz3, vdenorm_cutoff, _CMP_LT_OS), vf3); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_rcp_x40()
|
| D | vsigmoid-avx2-rr1-p2-rcp-x48.c | 50 const __m256 vz3 = _mm256_or_ps(vx3, vsign_mask); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_rcp_x48() local
57 __m256 vn3 = _mm256_fmadd_ps(vz3, vlog2e, vmagic_bias); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_rcp_x48()
78 __m256 vt3 = _mm256_fmadd_ps(vn3, vminus_ln2, vz3); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_rcp_x48()
127 vf3 = _mm256_andnot_ps(_mm256_cmp_ps(vz3, vdenorm_cutoff, _CMP_LT_OS), vf3); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_rcp_x48()
|
| D | vsigmoid-avx2-rr1-p2-div-x48.c | 50 const __m256 vz3 = _mm256_or_ps(vx3, vsign_mask); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_div_x48() local
57 __m256 vn3 = _mm256_fmadd_ps(vz3, vlog2e, vmagic_bias); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_div_x48()
78 __m256 vt3 = _mm256_fmadd_ps(vn3, vminus_ln2, vz3); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_div_x48()
120 vf3 = _mm256_andnot_ps(_mm256_cmp_ps(vz3, vdenorm_cutoff, _CMP_LT_OS), vf3); in xnn_f16_vsigmoid_ukernel__avx2_rr1_p2_div_x48()
|