| /external/ComputeLibrary/src/cpu/kernels/add/generic/neon/ |
| D | qsymm16.cpp | 97 int32x4_t rf_0{}; in add_qsymm16_neon() local
100 rf_0 = vcvtnq_s32_f32(vmulq_f32(vaddq_f32(af_0, bf_0), invvscaleo)); in add_qsymm16_neon()
103 rf_0 = vcvtq_s32_f32(vmulq_f32(vaddq_f32(af_0, bf_0), invvscaleo)); in add_qsymm16_neon()
107 const int16x8_t pa = vcombine_s16(vqmovn_s32(rf_0), vqmovn_s32(rf_1)); in add_qsymm16_neon()
148 int32x4_t rf_0{}; in add_qsymm16_neon() local
151 rf_0 = vcvtnq_s32_f32(vmulq_f32(vaddq_f32(af_0, bf_0), invvscaleo)); in add_qsymm16_neon()
154 rf_0 = vcvtq_s32_f32(vmulq_f32(vaddq_f32(af_0, bf_0), invvscaleo)); in add_qsymm16_neon()
158 const int16x8_t pa = vcombine_s16(vqmovn_s32(rf_0), vqmovn_s32(rf_1)); in add_qsymm16_neon()
|
| D | impl.cpp | 450 int32x4_t rf_0{}; in add_sub_qasymm8_neon() local
456 rf_0 = vcvtnq_s32_f32(af_0); in add_sub_qasymm8_neon()
461 rf_0 = vcvtq_s32_f32(af_0); in add_sub_qasymm8_neon()
467 const uint8x8_t pa = vqmovun_s16(vcombine_s16(vqmovn_s32(rf_0), vqmovn_s32(rf_1))); in add_sub_qasymm8_neon()
528 int32x4_t rf_0{}; in add_sub_qasymm8_neon() local
534 rf_0 = vcvtnq_s32_f32(bf_0); in add_sub_qasymm8_neon()
539 rf_0 = vcvtq_s32_f32(bf_0); in add_sub_qasymm8_neon()
545 const uint8x8_t pa = vqmovun_s16(vcombine_s16(vqmovn_s32(rf_0), vqmovn_s32(rf_1))); in add_sub_qasymm8_neon()
633 int32x4_t rf_0{}; in add_sub_qasymm8_signed_neon() local
639 rf_0 = vcvtnq_s32_f32(af_0); in add_sub_qasymm8_signed_neon()
[all …]
|
| /external/ComputeLibrary/src/core/NEON/ |
| D | SVESymm.h | 62 const auto rf_0 = svcvt_s32_f32_z(pg, svmul_f32_z(pg, svget2_f32(qv, 0), vinvscale)); in svquantize_qsymm16_z() local
64 const auto pa = svqxtnt_s32(svqxtnb_s32(rf_0), rf_1); in svquantize_qsymm16_z()
102 const auto rf_0 = svcvt_s32_f32_z(pg, svmul_f32_z(pg, svget4_f32(qv, 0), vinvscale)); in svquantize_qsymm16_z() local
107 const auto pa = svqxtnt_s32(svqxtnb_s32(rf_0), rf_1); in svquantize_qsymm16_z()
|
| D | SVEAsymm.h | 180 const auto rf_0 = svcvt_u32_f32_z(pg, svmla_f32_z(pg, voffset, svget4_f32(qv, 0), vinvscale)); in svquantize_z() local
185 const auto pa = svqxtnt_u32(svqxtnb_u32(rf_0), rf_1); in svquantize_z()
205 …const auto rf_0 = svcvt_s32_f32_z(pg, svmla_f32_z(pg, voffset, svget4_f32(qv, 0), vinvscale)… in svquantize_signed_z() local
210 const auto pa = svqxtnt_s32(svqxtnb_s32(rf_0), rf_1); in svquantize_signed_z()
231 const auto rf_0 = svcvt_u32_f32_z(pg, svmla_f32_z(pg, voffset, svget4_f32(qv, 0), vinvscale)); in svquantize_qasymm16_z() local
236 const auto pa = svqxtnt_u32(svqxtnb_u32(rf_0), rf_1); in svquantize_qasymm16_z()
|
| /external/ComputeLibrary/src/cpu/kernels/add/generic/sve2/ |
| D | qsymm16.cpp | 97 … const auto rf_0 = svcvt_s32_f32_z(pg, svmul_f32_z(pg, svadd_f32_z(pg, af_0, bf_0), invvscaleo)); in add_qsymm16_sve2() local
100 const auto res = svqxtnt_s32(svqxtnb_s32(rf_0), rf_1); in add_qsymm16_sve2()
140 … const auto rf_0 = svcvt_s32_f32_z(pg, svmul_f32_z(pg, svadd_f32_z(pg, af_0, bf_0), invvscaleo)); in add_qsymm16_sve2() local
143 const auto res = svqxtnt_s32(svqxtnb_s32(rf_0), rf_1); in add_qsymm16_sve2()
|
| D | qasymm8_signed.cpp | 104 …const auto rf_0 = svcvt_s32_f32_z(pg, svmla_f32_z(pg, voffseto, svadd_f32_z(pg, af_0, bf_0), invvs… in add_qasymm8_signed_sve2() local
109 const auto pa = svqxtnt_s32(svqxtnb_s32(rf_0), rf_1); in add_qasymm8_signed_sve2()
160 …const auto rf_0 = svcvt_s32_f32_z(pg, svmla_f32_z(pg, voffseto, svadd_f32_z(pg, af_0, bf_0), invvs… in add_qasymm8_signed_sve2() local
165 const auto pa = svqxtnt_s32(svqxtnb_s32(rf_0), rf_1); in add_qasymm8_signed_sve2()
|
| D | qasymm8.cpp | 106 …const auto rf_0 = svcvt_u32_f32_z(pg, svmla_f32_z(pg, voffseto, svadd_f32_z(pg, af_0, bf_0), invvs… in add_qasymm8_sve2() local
111 const auto pa = svqxtnt_u32(svqxtnb_u32(rf_0), rf_1); in add_qasymm8_sve2()
161 …const auto rf_0 = svcvt_u32_f32_z(pg, svmla_f32_z(pg, voffseto, svadd_f32_z(pg, af_0, bf_0), invvs… in add_qasymm8_sve2() local
166 const auto pa = svqxtnt_u32(svqxtnb_u32(rf_0), rf_1); in add_qasymm8_sve2()
|