| /external/XNNPACK/src/f16-raddstoreexpminusmax/gen/ |
| D | neonfp16arith-rr2-p2-x96.c | 159 const uint16x8_t vm0 = vcltq_f16(vx0, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96()
161 const uint16x8_t vm1 = vcltq_f16(vx1, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96()
163 const uint16x8_t vm2 = vcltq_f16(vx2, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96()
165 const uint16x8_t vm3 = vcltq_f16(vx3, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96()
167 const uint16x8_t vm4 = vcltq_f16(vx4, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96()
169 const uint16x8_t vm5 = vcltq_f16(vx5, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96()
171 const uint16x8_t vm6 = vcltq_f16(vx6, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96()
173 const uint16x8_t vm7 = vcltq_f16(vx7, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96()
175 const uint16x8_t vm8 = vcltq_f16(vx8, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96()
177 const uint16x8_t vm9 = vcltq_f16(vx9, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96()
[all …]
|
| D | neonfp16arith-rr2-p2-x96-acc3.c | 161 const uint16x8_t vm0 = vcltq_f16(vx0, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc3()
163 const uint16x8_t vm1 = vcltq_f16(vx1, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc3()
165 const uint16x8_t vm2 = vcltq_f16(vx2, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc3()
167 const uint16x8_t vm3 = vcltq_f16(vx3, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc3()
169 const uint16x8_t vm4 = vcltq_f16(vx4, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc3()
171 const uint16x8_t vm5 = vcltq_f16(vx5, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc3()
173 const uint16x8_t vm6 = vcltq_f16(vx6, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc3()
175 const uint16x8_t vm7 = vcltq_f16(vx7, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc3()
177 const uint16x8_t vm8 = vcltq_f16(vx8, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc3()
179 const uint16x8_t vm9 = vcltq_f16(vx9, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc3()
[all …]
|
| D | neonfp16arith-rr2-p2-x96-acc2.c | 160 const uint16x8_t vm0 = vcltq_f16(vx0, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc2()
162 const uint16x8_t vm1 = vcltq_f16(vx1, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc2()
164 const uint16x8_t vm2 = vcltq_f16(vx2, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc2()
166 const uint16x8_t vm3 = vcltq_f16(vx3, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc2()
168 const uint16x8_t vm4 = vcltq_f16(vx4, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc2()
170 const uint16x8_t vm5 = vcltq_f16(vx5, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc2()
172 const uint16x8_t vm6 = vcltq_f16(vx6, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc2()
174 const uint16x8_t vm7 = vcltq_f16(vx7, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc2()
176 const uint16x8_t vm8 = vcltq_f16(vx8, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc2()
178 const uint16x8_t vm9 = vcltq_f16(vx9, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc2()
[all …]
|
| D | neonfp16arith-rr2-p2-x96-acc6.c | 164 const uint16x8_t vm0 = vcltq_f16(vx0, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc6()
166 const uint16x8_t vm1 = vcltq_f16(vx1, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc6()
168 const uint16x8_t vm2 = vcltq_f16(vx2, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc6()
170 const uint16x8_t vm3 = vcltq_f16(vx3, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc6()
172 const uint16x8_t vm4 = vcltq_f16(vx4, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc6()
174 const uint16x8_t vm5 = vcltq_f16(vx5, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc6()
176 const uint16x8_t vm6 = vcltq_f16(vx6, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc6()
178 const uint16x8_t vm7 = vcltq_f16(vx7, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc6()
180 const uint16x8_t vm8 = vcltq_f16(vx8, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc6()
182 const uint16x8_t vm9 = vcltq_f16(vx9, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x96_acc6()
[all …]
|
| D | neonfp16arith-rr2-p2-x80.c | 141 const uint16x8_t vm0 = vcltq_f16(vx0, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80()
143 const uint16x8_t vm1 = vcltq_f16(vx1, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80()
145 const uint16x8_t vm2 = vcltq_f16(vx2, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80()
147 const uint16x8_t vm3 = vcltq_f16(vx3, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80()
149 const uint16x8_t vm4 = vcltq_f16(vx4, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80()
151 const uint16x8_t vm5 = vcltq_f16(vx5, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80()
153 const uint16x8_t vm6 = vcltq_f16(vx6, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80()
155 const uint16x8_t vm7 = vcltq_f16(vx7, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80()
157 const uint16x8_t vm8 = vcltq_f16(vx8, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80()
159 const uint16x8_t vm9 = vcltq_f16(vx9, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80()
[all …]
|
| D | neonfp16arith-rr2-p2-x80-acc5.c | 145 const uint16x8_t vm0 = vcltq_f16(vx0, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80_acc5()
147 const uint16x8_t vm1 = vcltq_f16(vx1, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80_acc5()
149 const uint16x8_t vm2 = vcltq_f16(vx2, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80_acc5()
151 const uint16x8_t vm3 = vcltq_f16(vx3, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80_acc5()
153 const uint16x8_t vm4 = vcltq_f16(vx4, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80_acc5()
155 const uint16x8_t vm5 = vcltq_f16(vx5, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80_acc5()
157 const uint16x8_t vm6 = vcltq_f16(vx6, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80_acc5()
159 const uint16x8_t vm7 = vcltq_f16(vx7, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80_acc5()
161 const uint16x8_t vm8 = vcltq_f16(vx8, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80_acc5()
163 const uint16x8_t vm9 = vcltq_f16(vx9, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80_acc5()
[all …]
|
| D | neonfp16arith-rr2-p2-x80-acc2.c | 142 const uint16x8_t vm0 = vcltq_f16(vx0, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80_acc2()
144 const uint16x8_t vm1 = vcltq_f16(vx1, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80_acc2()
146 const uint16x8_t vm2 = vcltq_f16(vx2, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80_acc2()
148 const uint16x8_t vm3 = vcltq_f16(vx3, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80_acc2()
150 const uint16x8_t vm4 = vcltq_f16(vx4, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80_acc2()
152 const uint16x8_t vm5 = vcltq_f16(vx5, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80_acc2()
154 const uint16x8_t vm6 = vcltq_f16(vx6, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80_acc2()
156 const uint16x8_t vm7 = vcltq_f16(vx7, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80_acc2()
158 const uint16x8_t vm8 = vcltq_f16(vx8, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80_acc2()
160 const uint16x8_t vm9 = vcltq_f16(vx9, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x80_acc2()
[all …]
|
| D | neonfp16arith-rr2-p2-x72.c | 132 const uint16x8_t vm0 = vcltq_f16(vx0, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72()
134 const uint16x8_t vm1 = vcltq_f16(vx1, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72()
136 const uint16x8_t vm2 = vcltq_f16(vx2, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72()
138 const uint16x8_t vm3 = vcltq_f16(vx3, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72()
140 const uint16x8_t vm4 = vcltq_f16(vx4, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72()
142 const uint16x8_t vm5 = vcltq_f16(vx5, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72()
144 const uint16x8_t vm6 = vcltq_f16(vx6, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72()
146 const uint16x8_t vm7 = vcltq_f16(vx7, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72()
148 const uint16x8_t vm8 = vcltq_f16(vx8, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72()
198 const uint16x8_t vm = vcltq_f16(vx, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72()
[all …]
|
| D | neonfp16arith-rr2-p2-x72-acc3.c | 134 const uint16x8_t vm0 = vcltq_f16(vx0, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72_acc3()
136 const uint16x8_t vm1 = vcltq_f16(vx1, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72_acc3()
138 const uint16x8_t vm2 = vcltq_f16(vx2, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72_acc3()
140 const uint16x8_t vm3 = vcltq_f16(vx3, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72_acc3()
142 const uint16x8_t vm4 = vcltq_f16(vx4, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72_acc3()
144 const uint16x8_t vm5 = vcltq_f16(vx5, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72_acc3()
146 const uint16x8_t vm6 = vcltq_f16(vx6, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72_acc3()
148 const uint16x8_t vm7 = vcltq_f16(vx7, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72_acc3()
150 const uint16x8_t vm8 = vcltq_f16(vx8, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72_acc3()
202 const uint16x8_t vm = vcltq_f16(vx, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x72_acc3()
[all …]
|
| D | neonfp16arith-rr2-p2-x64-acc4.c | 126 const uint16x8_t vm0 = vcltq_f16(vx0, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc4()
128 const uint16x8_t vm1 = vcltq_f16(vx1, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc4()
130 const uint16x8_t vm2 = vcltq_f16(vx2, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc4()
132 const uint16x8_t vm3 = vcltq_f16(vx3, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc4()
134 const uint16x8_t vm4 = vcltq_f16(vx4, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc4()
136 const uint16x8_t vm5 = vcltq_f16(vx5, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc4()
138 const uint16x8_t vm6 = vcltq_f16(vx6, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc4()
140 const uint16x8_t vm7 = vcltq_f16(vx7, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc4()
190 const uint16x8_t vm = vcltq_f16(vx, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc4()
216 const uint16x8_t vm = vcltq_f16(vx, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc4()
|
| D | neonfp16arith-rr2-p2-x64-acc2.c | 124 const uint16x8_t vm0 = vcltq_f16(vx0, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc2()
126 const uint16x8_t vm1 = vcltq_f16(vx1, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc2()
128 const uint16x8_t vm2 = vcltq_f16(vx2, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc2()
130 const uint16x8_t vm3 = vcltq_f16(vx3, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc2()
132 const uint16x8_t vm4 = vcltq_f16(vx4, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc2()
134 const uint16x8_t vm5 = vcltq_f16(vx5, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc2()
136 const uint16x8_t vm6 = vcltq_f16(vx6, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc2()
138 const uint16x8_t vm7 = vcltq_f16(vx7, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc2()
186 const uint16x8_t vm = vcltq_f16(vx, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc2()
212 const uint16x8_t vm = vcltq_f16(vx, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64_acc2()
|
| D | neonfp16arith-rr2-p2-x64.c | 123 const uint16x8_t vm0 = vcltq_f16(vx0, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64()
125 const uint16x8_t vm1 = vcltq_f16(vx1, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64()
127 const uint16x8_t vm2 = vcltq_f16(vx2, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64()
129 const uint16x8_t vm3 = vcltq_f16(vx3, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64()
131 const uint16x8_t vm4 = vcltq_f16(vx4, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64()
133 const uint16x8_t vm5 = vcltq_f16(vx5, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64()
135 const uint16x8_t vm6 = vcltq_f16(vx6, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64()
137 const uint16x8_t vm7 = vcltq_f16(vx7, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64()
184 const uint16x8_t vm = vcltq_f16(vx, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64()
210 const uint16x8_t vm = vcltq_f16(vx, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x64()
|
| D | neonfp16arith-rr2-p2-x48.c | 105 const uint16x8_t vm0 = vcltq_f16(vx0, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48()
107 const uint16x8_t vm1 = vcltq_f16(vx1, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48()
109 const uint16x8_t vm2 = vcltq_f16(vx2, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48()
111 const uint16x8_t vm3 = vcltq_f16(vx3, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48()
113 const uint16x8_t vm4 = vcltq_f16(vx4, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48()
115 const uint16x8_t vm5 = vcltq_f16(vx5, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48()
156 const uint16x8_t vm = vcltq_f16(vx, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48()
182 const uint16x8_t vm = vcltq_f16(vx, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48()
|
| D | neonfp16arith-rr2-p2-x40.c | 96 const uint16x8_t vm0 = vcltq_f16(vx0, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40()
98 const uint16x8_t vm1 = vcltq_f16(vx1, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40()
100 const uint16x8_t vm2 = vcltq_f16(vx2, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40()
102 const uint16x8_t vm3 = vcltq_f16(vx3, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40()
104 const uint16x8_t vm4 = vcltq_f16(vx4, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40()
142 const uint16x8_t vm = vcltq_f16(vx, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40()
168 const uint16x8_t vm = vcltq_f16(vx, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40()
|
| D | neonfp16arith-rr2-p2-x40-acc2.c | 97 const uint16x8_t vm0 = vcltq_f16(vx0, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc2()
99 const uint16x8_t vm1 = vcltq_f16(vx1, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc2()
101 const uint16x8_t vm2 = vcltq_f16(vx2, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc2()
103 const uint16x8_t vm3 = vcltq_f16(vx3, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc2()
105 const uint16x8_t vm4 = vcltq_f16(vx4, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc2()
144 const uint16x8_t vm = vcltq_f16(vx, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc2()
170 const uint16x8_t vm = vcltq_f16(vx, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc2()
|
| D | neonfp16arith-rr2-p2-x48-acc3.c | 107 const uint16x8_t vm0 = vcltq_f16(vx0, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48_acc3()
109 const uint16x8_t vm1 = vcltq_f16(vx1, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48_acc3()
111 const uint16x8_t vm2 = vcltq_f16(vx2, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48_acc3()
113 const uint16x8_t vm3 = vcltq_f16(vx3, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48_acc3()
115 const uint16x8_t vm4 = vcltq_f16(vx4, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48_acc3()
117 const uint16x8_t vm5 = vcltq_f16(vx5, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48_acc3()
160 const uint16x8_t vm = vcltq_f16(vx, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48_acc3()
186 const uint16x8_t vm = vcltq_f16(vx, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48_acc3()
|
| D | neonfp16arith-rr2-p2-x48-acc2.c | 106 const uint16x8_t vm0 = vcltq_f16(vx0, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48_acc2()
108 const uint16x8_t vm1 = vcltq_f16(vx1, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48_acc2()
110 const uint16x8_t vm2 = vcltq_f16(vx2, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48_acc2()
112 const uint16x8_t vm3 = vcltq_f16(vx3, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48_acc2()
114 const uint16x8_t vm4 = vcltq_f16(vx4, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48_acc2()
116 const uint16x8_t vm5 = vcltq_f16(vx5, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48_acc2()
158 const uint16x8_t vm = vcltq_f16(vx, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48_acc2()
184 const uint16x8_t vm = vcltq_f16(vx, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x48_acc2()
|
| D | neonfp16arith-rr2-p2-x40-acc5.c | 100 const uint16x8_t vm0 = vcltq_f16(vx0, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc5()
102 const uint16x8_t vm1 = vcltq_f16(vx1, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc5()
104 const uint16x8_t vm2 = vcltq_f16(vx2, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc5()
106 const uint16x8_t vm3 = vcltq_f16(vx3, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc5()
108 const uint16x8_t vm4 = vcltq_f16(vx4, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc5()
150 const uint16x8_t vm = vcltq_f16(vx, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc5()
176 const uint16x8_t vm = vcltq_f16(vx, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x40_acc5()
|
| D | neonfp16arith-rr2-p2-x32.c | 87 const uint16x8_t vm0 = vcltq_f16(vx0, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32()
89 const uint16x8_t vm1 = vcltq_f16(vx1, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32()
91 const uint16x8_t vm2 = vcltq_f16(vx2, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32()
93 const uint16x8_t vm3 = vcltq_f16(vx3, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32()
128 const uint16x8_t vm = vcltq_f16(vx, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32()
154 const uint16x8_t vm = vcltq_f16(vx, vdenorm_cutoff); in xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_x32()
|
| /external/XNNPACK/src/f16-vsigmoid/gen/ |
| D | vsigmoid-neonfp16arith-rr2-p2-div-x64.c | 155 const uint16x8_t vm0 = vcltq_f16(vx0, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x64()
156 const uint16x8_t vm1 = vcltq_f16(vx1, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x64()
157 const uint16x8_t vm2 = vcltq_f16(vx2, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x64()
158 const uint16x8_t vm3 = vcltq_f16(vx3, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x64()
159 const uint16x8_t vm4 = vcltq_f16(vx4, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x64()
160 const uint16x8_t vm5 = vcltq_f16(vx5, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x64()
161 const uint16x8_t vm6 = vcltq_f16(vx6, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x64()
162 const uint16x8_t vm7 = vcltq_f16(vx7, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x64()
201 const uint16x8_t vm = vcltq_f16(vx, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x64()
225 const uint16x8_t vm = vcltq_f16(vx, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x64()
|
| D | vsigmoid-neonfp16arith-rr2-p2-div-x56.c | 142 const uint16x8_t vm0 = vcltq_f16(vx0, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x56()
143 const uint16x8_t vm1 = vcltq_f16(vx1, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x56()
144 const uint16x8_t vm2 = vcltq_f16(vx2, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x56()
145 const uint16x8_t vm3 = vcltq_f16(vx3, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x56()
146 const uint16x8_t vm4 = vcltq_f16(vx4, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x56()
147 const uint16x8_t vm5 = vcltq_f16(vx5, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x56()
148 const uint16x8_t vm6 = vcltq_f16(vx6, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x56()
185 const uint16x8_t vm = vcltq_f16(vx, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x56()
209 const uint16x8_t vm = vcltq_f16(vx, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x56()
|
| D | vsigmoid-neonfp16arith-rr2-p2-nr1fma-x64.c | 182 const uint16x8_t vm0 = vcltq_f16(vx0, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x64()
183 const uint16x8_t vm1 = vcltq_f16(vx1, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x64()
184 const uint16x8_t vm2 = vcltq_f16(vx2, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x64()
185 const uint16x8_t vm3 = vcltq_f16(vx3, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x64()
186 const uint16x8_t vm4 = vcltq_f16(vx4, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x64()
187 const uint16x8_t vm5 = vcltq_f16(vx5, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x64()
188 const uint16x8_t vm6 = vcltq_f16(vx6, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x64()
189 const uint16x8_t vm7 = vcltq_f16(vx7, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x64()
232 const uint16x8_t vm = vcltq_f16(vx, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x64()
260 const uint16x8_t vm = vcltq_f16(vx, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1fma_x64()
|
| D | vsigmoid-neonfp16arith-rr2-p2-nr1recps-x64.c | 182 const uint16x8_t vm0 = vcltq_f16(vx0, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x64()
183 const uint16x8_t vm1 = vcltq_f16(vx1, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x64()
184 const uint16x8_t vm2 = vcltq_f16(vx2, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x64()
185 const uint16x8_t vm3 = vcltq_f16(vx3, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x64()
186 const uint16x8_t vm4 = vcltq_f16(vx4, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x64()
187 const uint16x8_t vm5 = vcltq_f16(vx5, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x64()
188 const uint16x8_t vm6 = vcltq_f16(vx6, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x64()
189 const uint16x8_t vm7 = vcltq_f16(vx7, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x64()
232 const uint16x8_t vm = vcltq_f16(vx, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x64()
260 const uint16x8_t vm = vcltq_f16(vx, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_nr1recps_x64()
|
| D | vsigmoid-neonfp16arith-rr2-p2-div-x48.c | 129 const uint16x8_t vm0 = vcltq_f16(vx0, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x48()
130 const uint16x8_t vm1 = vcltq_f16(vx1, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x48()
131 const uint16x8_t vm2 = vcltq_f16(vx2, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x48()
132 const uint16x8_t vm3 = vcltq_f16(vx3, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x48()
133 const uint16x8_t vm4 = vcltq_f16(vx4, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x48()
134 const uint16x8_t vm5 = vcltq_f16(vx5, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x48()
169 const uint16x8_t vm = vcltq_f16(vx, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x48()
193 const uint16x8_t vm = vcltq_f16(vx, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x48()
|
| D | vsigmoid-neonfp16arith-rr2-p2-div-x40.c | 116 const uint16x8_t vm0 = vcltq_f16(vx0, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x40()
117 const uint16x8_t vm1 = vcltq_f16(vx1, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x40()
118 const uint16x8_t vm2 = vcltq_f16(vx2, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x40()
119 const uint16x8_t vm3 = vcltq_f16(vx3, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x40()
120 const uint16x8_t vm4 = vcltq_f16(vx4, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x40()
153 const uint16x8_t vm = vcltq_f16(vx, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x40()
177 const uint16x8_t vm = vcltq_f16(vx, vmovq_n_f16(0.0f)); in xnn_f16_vsigmoid_ukernel__neonfp16arith_rr2_p2_div_x40()
|