28 #if defined(__aarch64__)
33 void a64_fp32_nhwc_3x3_s1_output2x2_mla_depthfirst_indirect_impl(
34 const float *
const *
const input_ptrs,
35 float *
const *
const outptrs,
37 unsigned int n_channels,
44 float *
const *outptrs;
47 const float *inptrs[16];
50 const float *
const *
const input_ptrs,
51 float *
const *
const outptrs,
52 const void *
const params,
55 ) : outptrs(outptrs), params(params), min(min), max(max)
57 inptrs[0] = input_ptrs[5];
58 inptrs[1] = input_ptrs[0];
59 inptrs[2] = input_ptrs[3];
60 inptrs[3] = input_ptrs[6];
61 inptrs[4] = input_ptrs[9];
62 inptrs[5] = input_ptrs[12];
63 inptrs[6] = input_ptrs[15];
64 inptrs[7] = input_ptrs[1];
65 inptrs[8] = input_ptrs[2];
66 inptrs[9] = input_ptrs[10];
67 inptrs[10] = input_ptrs[4];
68 inptrs[11] = input_ptrs[7];
69 inptrs[12] = input_ptrs[8];
70 inptrs[13] = input_ptrs[11];
71 inptrs[14] = input_ptrs[13];
72 inptrs[15] = input_ptrs[14];
77 Args params_struct(input_ptrs, outptrs, params,
81 "ldr x21, [%x[params_struct], %[offsetof_args_outptrs]]\n"
83 "lsr x15, %x[n_channels], #0x2\n"
84 "ldr x14, [%x[params_struct], %[offsetof_args_params]]\n"
85 "add x20, %x[params_struct], %[offsetof_args_min]\n"
86 "ld1r { v27.4s }, [x20]\n"
87 "add x20, %x[params_struct], %[offsetof_args_max]\n"
88 "ld1r { v26.4s }, [x20]\n"
89 "add x13, %x[params_struct], %[offsetof_Args_inptrs]\n"
90 "ldp x12, x11, [x21, #0x0]\n"
91 "ldp x10, x9, [x21, #0x10]\n"
95 "ldr q25, [x14, #0x0]\n"
96 "ldr q0, [x14, #0x10]\n"
97 "cmp x16, x15, LSL #4\n"
98 "ldr q1, [x14, #0x20]\n"
99 "ldr q2, [x14, #0x30]\n"
100 "ldr q3, [x14, #0x40]\n"
101 "ldr q4, [x14, #0x50]\n"
102 "ldr q5, [x14, #0x60]\n"
103 "ldr q6, [x14, #0x70]\n"
104 "ldr q7, [x14, #0x80]\n"
105 "ldr q8, [x14, #0x90]\n"
106 "add x14, x14, #0xa0\n"
107 "ldp x21, x20, [x13, #0x0]\n"
108 "ldr q9, [x21, x28]\n"
109 "ldr q10, [x20, x28]\n"
110 "ldp x21, x20, [x13, #0x10]\n"
111 "ldr q11, [x21, x28]\n"
112 "ldr q12, [x20, x28]\n"
113 "ldr x20, [x13, #0x20]\n"
114 "ldr q13, [x20, x28]\n"
117 "mov v24.16b, v25.16b\n fmla v24.4s, v4.4s, v9.4s\n"
118 "mov v23.16b, v25.16b\n fmla v23.4s, v3.4s, v9.4s\n"
119 "ldr x21, [x13, #0x28]\n"
120 "ldr x20, [x13, #0x30]\n"
121 "mov v22.16b, v25.16b\n fmla v22.4s, v1.4s, v9.4s\n"
122 "mov v21.16b, v25.16b\n fmla v21.4s, v0.4s, v9.4s\n"
123 "ldr q18, [x21, x28]\n"
124 "ldr q25, [x14, #0x0]\n"
125 "fmla v24.4s, v0.4s, v10.4s\n"
126 "fmla v23.4s, v2.4s, v11.4s\n"
127 "ldr q17, [x20, x28]\n"
128 "ldr x21, [x13, #0x38]\n"
129 "fmla v22.4s, v2.4s, v12.4s\n"
130 "fmla v21.4s, v1.4s, v12.4s\n"
131 "ldr x20, [x13, #0x48]\n"
132 "ldr q20, [x20, x28]\n"
133 "fmla v24.4s, v5.4s, v12.4s\n"
134 "fmla v23.4s, v4.4s, v12.4s\n"
135 "ldr q16, [x21, x28]\n"
136 "ldr x20, [x13, #0x40]\n"
137 "fmla v22.4s, v6.4s, v18.4s\n"
138 "ldr q18, [x20, x28]\n"
139 "fmla v21.4s, v3.4s, v13.4s\n"
140 "ldr x20, [x13, #0x50]\n"
141 "fmla v24.4s, v7.4s, v13.4s\n"
142 "fmla v23.4s, v6.4s, v13.4s\n"
143 "ldr x22, [x13, #0x58]\n"
144 "ldr x21, [x13, #0x60]\n"
145 "fmla v22.4s, v4.4s, v13.4s\n"
146 "fmla v21.4s, v8.4s, v17.4s\n"
147 "ldr q17, [x20, x28]\n"
148 "ldr x20, [x13, #0x68]\n"
149 "fmla v24.4s, v1.4s, v16.4s\n"
150 "fmla v23.4s, v0.4s, v16.4s\n"
151 "ldr q16, [x22, x28]\n"
152 "ldr x26, [x13, #0x70]\n"
153 "fmla v22.4s, v5.4s, v20.4s\n"
154 "fmla v21.4s, v4.4s, v20.4s\n"
155 "ldr q4, [x14, #0x50]\n"
156 "ldr x25, [x13, #0x78]\n"
157 "fmla v24.4s, v2.4s, v18.4s\n"
158 "fmla v23.4s, v1.4s, v18.4s\n"
159 "ldr q19, [x21, x28]\n"
160 "ldr q1, [x14, #0x20]\n"
161 "fmla v22.4s, v0.4s, v17.4s\n"
162 "ldr q0, [x14, #0x10]\n"
163 "fmla v21.4s, v2.4s, v16.4s\n"
164 "ldr q2, [x14, #0x30]\n"
165 "fmla v24.4s, v8.4s, v20.4s\n"
166 "fmla v23.4s, v7.4s, v20.4s\n"
167 "ldr q18, [x20, x28]\n"
168 "ldp x24, x23, [x13, #0x0]\n"
169 "fmla v22.4s, v3.4s, v19.4s\n"
170 "fmla v21.4s, v5.4s, v18.4s\n"
171 "ldp x22, x21, [x13, #0x10]\n"
172 "ldr x20, [x13, #0x20]\n"
173 "ldr q13, [x20, x16]\n"
174 "fmla v24.4s, v3.4s, v17.4s\n"
175 "ldr q17, [x26, x28]\n"
176 "fmla v23.4s, v5.4s, v16.4s\n"
177 "ldr q16, [x25, x28]\n"
178 "ldr q3, [x14, #0x40]\n"
179 "fmla v22.4s, v7.4s, v17.4s\n"
180 "fmla v21.4s, v6.4s, v17.4s\n"
181 "ldr q11, [x22, x16]\n"
182 "ldr q5, [x14, #0x60]\n"
183 "fmla v24.4s, v6.4s, v19.4s\n"
184 "fmla v23.4s, v8.4s, v18.4s\n"
185 "ldr q9, [x24, x16]\n"
186 "ldr q10, [x23, x16]\n"
187 "fmla v22.4s, v8.4s, v16.4s\n"
188 "fmla v21.4s, v7.4s, v16.4s\n"
189 "ldr q12, [x21, x16]\n"
190 "ldr q6, [x14, #0x70]\n"
191 "fmax v24.4s, v24.4s, v27.4s\n"
192 "fmax v23.4s, v23.4s, v27.4s\n"
193 "ldr q7, [x14, #0x80]\n"
194 "ldr q8, [x14, #0x90]\n"
195 "fmax v22.4s, v22.4s, v27.4s\n"
196 "fmax v21.4s, v21.4s, v27.4s\n"
197 "add x16, x16, #0x10\n"
198 "add x27, x27, #0x10\n"
199 "fmin v24.4s, v24.4s, v26.4s\n"
200 "fmin v23.4s, v23.4s, v26.4s\n"
201 "cmp x16, x15, LSL #4\n"
202 "fmin v22.4s, v22.4s, v26.4s\n"
203 "fmin v21.4s, v21.4s, v26.4s\n"
204 "add x28, x28, #0x10\n"
205 "str q24, [x12, x27]\n"
206 "add x14, x14, #0xa0\n"
207 "str q23, [x11, x27]\n"
208 "str q22, [x10, x27]\n"
209 "str q21, [x9, x27]\n"
212 "mov v24.16b, v25.16b\n fmla v24.4s, v4.4s, v9.4s\n"
213 "mov v23.16b, v25.16b\n fmla v23.4s, v3.4s, v9.4s\n"
214 "ldr x21, [x13, #0x28]\n"
215 "ldr x20, [x13, #0x30]\n"
216 "mov v22.16b, v25.16b\n fmla v22.4s, v1.4s, v9.4s\n"
217 "mov v21.16b, v25.16b\n fmla v21.4s, v0.4s, v9.4s\n"
218 "ldr q18, [x21, x28]\n"
219 "ldr x21, [x13, #0x38]\n"
220 "fmla v24.4s, v0.4s, v10.4s\n"
221 "fmla v23.4s, v2.4s, v11.4s\n"
222 "ldr q17, [x20, x28]\n"
223 "ldr x20, [x13, #0x48]\n"
224 "ldr q20, [x20, x28]\n"
225 "fmla v22.4s, v2.4s, v12.4s\n"
226 "fmla v21.4s, v1.4s, v12.4s\n"
227 "ldr x20, [x13, #0x40]\n"
228 "fmla v24.4s, v5.4s, v12.4s\n"
229 "fmla v23.4s, v4.4s, v12.4s\n"
230 "ldr q16, [x21, x28]\n"
231 "ldr x21, [x13, #0x50]\n"
232 "fmla v22.4s, v6.4s, v18.4s\n"
233 "ldr q18, [x20, x28]\n"
234 "fmla v21.4s, v3.4s, v13.4s\n"
235 "ldr x20, [x13, #0x58]\n"
236 "fmla v24.4s, v7.4s, v13.4s\n"
237 "fmla v23.4s, v6.4s, v13.4s\n"
238 "ldr x23, [x13, #0x60]\n"
239 "ldr x22, [x13, #0x68]\n"
240 "fmla v22.4s, v4.4s, v13.4s\n"
241 "fmla v21.4s, v8.4s, v17.4s\n"
242 "ldr q17, [x21, x28]\n"
243 "ldr x21, [x13, #0x70]\n"
244 "fmla v24.4s, v1.4s, v16.4s\n"
245 "fmla v23.4s, v0.4s, v16.4s\n"
246 "ldr q16, [x20, x28]\n"
247 "ldr x20, [x13, #0x78]\n"
248 "fmla v22.4s, v5.4s, v20.4s\n"
249 "fmla v21.4s, v4.4s, v20.4s\n"
250 "add x27, x27, #0x10\n"
251 "fmla v24.4s, v2.4s, v18.4s\n"
252 "fmla v23.4s, v1.4s, v18.4s\n"
253 "ldr q19, [x23, x28]\n"
254 "fmla v22.4s, v0.4s, v17.4s\n"
255 "fmla v21.4s, v2.4s, v16.4s\n"
256 "fmla v24.4s, v8.4s, v20.4s\n"
257 "fmla v23.4s, v7.4s, v20.4s\n"
258 "ldr q18, [x22, x28]\n"
259 "fmla v22.4s, v3.4s, v19.4s\n"
260 "fmla v21.4s, v5.4s, v18.4s\n"
261 "fmla v24.4s, v3.4s, v17.4s\n"
262 "ldr q17, [x21, x28]\n"
263 "fmla v23.4s, v5.4s, v16.4s\n"
264 "ldr q16, [x20, x28]\n"
265 "fmla v22.4s, v7.4s, v17.4s\n"
266 "fmla v21.4s, v6.4s, v17.4s\n"
267 "add x28, x28, #0x10\n"
268 "fmla v24.4s, v6.4s, v19.4s\n"
269 "fmla v23.4s, v8.4s, v18.4s\n"
270 "fmax v24.4s, v24.4s, v27.4s\n"
271 "fmla v22.4s, v8.4s, v16.4s\n"
272 "fmla v21.4s, v7.4s, v16.4s\n"
273 "fmax v23.4s, v23.4s, v27.4s\n"
274 "fmax v22.4s, v22.4s, v27.4s\n"
275 "fmax v21.4s, v21.4s, v27.4s\n"
276 "fmin v24.4s, v24.4s, v26.4s\n"
277 "fmin v23.4s, v23.4s, v26.4s\n"
278 "str q24, [x12, x27]\n"
279 "fmin v22.4s, v22.4s, v26.4s\n"
280 "fmin v21.4s, v21.4s, v26.4s\n"
281 "str q23, [x11, x27]\n"
282 "str q22, [x10, x27]\n"
283 "str q21, [x9, x27]\n"
285 "tst %x[n_channels], #0x3\n"
287 "ldr q25, [x14, #0x0]\n"
288 "ldr q0, [x14, #0x10]\n"
290 "add x12, x12, x20\n"
291 "ldr q1, [x14, #0x20]\n"
292 "ldr q2, [x14, #0x30]\n"
293 "add x11, x11, x20\n"
294 "add x10, x10, x20\n"
295 "ldr q3, [x14, #0x40]\n"
296 "ldr q4, [x14, #0x50]\n"
298 "ldr q5, [x14, #0x60]\n"
299 "ldr q6, [x14, #0x70]\n"
300 "ldr q7, [x14, #0x80]\n"
301 "ldr q8, [x14, #0x90]\n"
302 "ldr x24, [x13, #0x0]\n"
303 "ldr x23, [x13, #0x8]\n"
304 "add x24, x24, x28\n"
305 "add x23, x23, x28\n"
306 "ldr x22, [x13, #0x10]\n"
307 "ldr x21, [x13, #0x18]\n"
308 "add x22, x22, x28\n"
309 "add x21, x21, x28\n"
310 "ldr x20, [x13, #0x20]\n"
311 "add x20, x20, x28\n"
312 "tbz %x[n_channels], #1, 4f\n"
313 "ld1 { v9.d }[0], [x24], #0x8\n"
314 "ld1 { v10.d }[0], [x23], #0x8\n"
315 "ld1 { v11.d }[0], [x22], #0x8\n"
316 "ld1 { v12.d }[0], [x21], #0x8\n"
317 "ld1 { v13.d }[0], [x20], #0x8\n"
318 "tbz %x[n_channels], #0, 5f\n"
319 "ld1 { v9.s }[2], [x24], #0x4\n"
320 "ld1 { v10.s }[2], [x23], #0x4\n"
321 "ld1 { v11.s }[2], [x22], #0x4\n"
322 "ld1 { v12.s }[2], [x21], #0x4\n"
323 "ld1 { v13.s }[2], [x20], #0x4\n"
326 "ld1 { v9.s }[0], [x24], #0x4\n"
327 "ld1 { v10.s }[0], [x23], #0x4\n"
328 "ld1 { v11.s }[0], [x22], #0x4\n"
329 "ld1 { v12.s }[0], [x21], #0x4\n"
330 "ld1 { v13.s }[0], [x20], #0x4\n"
332 "mov v28.16b, v25.16b\n fmla v28.4s, v4.4s, v9.4s\n"
333 "mov v29.16b, v25.16b\n fmla v29.4s, v3.4s, v9.4s\n"
334 "ldr x20, [x13, #0x28]\n"
335 "add x20, x20, x28\n"
336 "mov v30.16b, v25.16b\n fmla v30.4s, v1.4s, v9.4s\n"
337 "mov v31.16b, v25.16b\n fmla v31.4s, v0.4s, v9.4s\n"
338 "fmla v28.4s, v0.4s, v10.4s\n"
339 "fmla v29.4s, v2.4s, v11.4s\n"
340 "fmla v28.4s, v5.4s, v12.4s\n"
341 "fmla v29.4s, v4.4s, v12.4s\n"
342 "fmla v30.4s, v2.4s, v12.4s\n"
343 "fmla v31.4s, v1.4s, v12.4s\n"
344 "tbz %x[n_channels], #1, 6f\n"
345 "ld1 { v9.d }[0], [x20], #0x8\n"
346 "tbz %x[n_channels], #0, 7f\n"
347 "ld1 { v9.s }[2], [x20], #0x4\n"
350 "ld1 { v9.s }[0], [x20], #0x4\n"
352 "fmla v30.4s, v6.4s, v9.4s\n"
353 "ldr x20, [x13, #0x30]\n"
354 "fmla v28.4s, v7.4s, v13.4s\n"
355 "add x20, x20, x28\n"
356 "fmla v29.4s, v6.4s, v13.4s\n"
357 "fmla v30.4s, v4.4s, v13.4s\n"
358 "fmla v31.4s, v3.4s, v13.4s\n"
359 "tbz %x[n_channels], #1, 8f\n"
360 "ld1 { v11.d }[0], [x20], #0x8\n"
361 "tbz %x[n_channels], #0, 9f\n"
362 "ld1 { v11.s }[2], [x20], #0x4\n"
365 "ld1 { v11.s }[0], [x20], #0x4\n"
367 "ldr x20, [x13, #0x38]\n"
368 "fmla v31.4s, v8.4s, v11.4s\n"
369 "add x20, x20, x28\n"
370 "tbz %x[n_channels], #1, 10f\n"
371 "ld1 { v12.d }[0], [x20], #0x8\n"
372 "tbz %x[n_channels], #0, 11f\n"
373 "ld1 { v12.s }[2], [x20], #0x4\n"
376 "ld1 { v12.s }[0], [x20], #0x4\n"
378 "ldr x20, [x13, #0x40]\n"
379 "fmla v28.4s, v1.4s, v12.4s\n"
380 "fmla v29.4s, v0.4s, v12.4s\n"
381 "add x20, x20, x28\n"
382 "tbz %x[n_channels], #1, 12f\n"
383 "ld1 { v9.d }[0], [x20], #0x8\n"
384 "tbz %x[n_channels], #0, 13f\n"
385 "ld1 { v9.s }[2], [x20], #0x4\n"
388 "ld1 { v9.s }[0], [x20], #0x4\n"
390 "ldr x20, [x13, #0x48]\n"
391 "fmla v28.4s, v2.4s, v9.4s\n"
392 "fmla v29.4s, v1.4s, v9.4s\n"
393 "add x20, x20, x28\n"
394 "tbz %x[n_channels], #1, 14f\n"
395 "ld1 { v10.d }[0], [x20], #0x8\n"
396 "tbz %x[n_channels], #0, 15f\n"
397 "ld1 { v10.s }[2], [x20], #0x4\n"
400 "ld1 { v10.s }[0], [x20], #0x4\n"
402 "ldr x20, [x13, #0x50]\n"
403 "fmla v28.4s, v8.4s, v10.4s\n"
404 "fmla v29.4s, v7.4s, v10.4s\n"
405 "add x20, x20, x28\n"
406 "fmla v30.4s, v5.4s, v10.4s\n"
407 "fmla v31.4s, v4.4s, v10.4s\n"
408 "tbz %x[n_channels], #1, 16f\n"
409 "ld1 { v11.d }[0], [x20], #0x8\n"
410 "tbz %x[n_channels], #0, 17f\n"
411 "ld1 { v11.s }[2], [x20], #0x4\n"
414 "ld1 { v11.s }[0], [x20], #0x4\n"
416 "ldr x20, [x13, #0x58]\n"
417 "fmla v28.4s, v3.4s, v11.4s\n"
418 "fmla v30.4s, v0.4s, v11.4s\n"
419 "add x20, x20, x28\n"
420 "tbz %x[n_channels], #1, 18f\n"
421 "ld1 { v12.d }[0], [x20], #0x8\n"
422 "tbz %x[n_channels], #0, 19f\n"
423 "ld1 { v12.s }[2], [x20], #0x4\n"
426 "ld1 { v12.s }[0], [x20], #0x4\n"
428 "ldr x20, [x13, #0x60]\n"
429 "fmla v29.4s, v5.4s, v12.4s\n"
430 "fmla v31.4s, v2.4s, v12.4s\n"
431 "add x20, x20, x28\n"
432 "tbz %x[n_channels], #1, 20f\n"
433 "ld1 { v9.d }[0], [x20], #0x8\n"
434 "tbz %x[n_channels], #0, 21f\n"
435 "ld1 { v9.s }[2], [x20], #0x4\n"
438 "ld1 { v9.s }[0], [x20], #0x4\n"
440 "ldr x20, [x13, #0x68]\n"
441 "fmla v28.4s, v6.4s, v9.4s\n"
442 "fmla v30.4s, v3.4s, v9.4s\n"
443 "add x20, x20, x28\n"
444 "tbz %x[n_channels], #1, 22f\n"
445 "ld1 { v10.d }[0], [x20], #0x8\n"
446 "tbz %x[n_channels], #0, 23f\n"
447 "ld1 { v10.s }[2], [x20], #0x4\n"
450 "ld1 { v10.s }[0], [x20], #0x4\n"
452 "ldr x20, [x13, #0x70]\n"
453 "fmla v29.4s, v8.4s, v10.4s\n"
454 "fmla v31.4s, v5.4s, v10.4s\n"
455 "add x20, x20, x28\n"
456 "tbz %x[n_channels], #1, 24f\n"
457 "ld1 { v11.d }[0], [x20], #0x8\n"
458 "tbz %x[n_channels], #0, 25f\n"
459 "ld1 { v11.s }[2], [x20], #0x4\n"
462 "ld1 { v11.s }[0], [x20], #0x4\n"
464 "ldr x20, [x13, #0x78]\n"
465 "fmla v30.4s, v7.4s, v11.4s\n"
466 "fmla v31.4s, v6.4s, v11.4s\n"
467 "add x20, x20, x28\n"
468 "tbz %x[n_channels], #1, 26f\n"
469 "ld1 { v12.d }[0], [x20], #0x8\n"
470 "tbz %x[n_channels], #0, 27f\n"
471 "ld1 { v12.s }[2], [x20], #0x4\n"
474 "ld1 { v12.s }[0], [x20], #0x4\n"
476 "fmla v30.4s, v8.4s, v12.4s\n"
477 "fmla v31.4s, v7.4s, v12.4s\n"
478 "fmax v28.4s, v28.4s, v27.4s\n"
479 "fmax v29.4s, v29.4s, v27.4s\n"
480 "fmax v30.4s, v30.4s, v27.4s\n"
481 "fmax v31.4s, v31.4s, v27.4s\n"
482 "fmin v28.4s, v28.4s, v26.4s\n"
483 "fmin v29.4s, v29.4s, v26.4s\n"
484 "fmin v30.4s, v30.4s, v26.4s\n"
485 "fmin v31.4s, v31.4s, v26.4s\n"
486 "tbz %x[n_channels], #1, 28f\n"
487 "st1 { v28.d }[0], [x12], #0x8\n"
488 "st1 { v29.d }[0], [x11], #0x8\n"
489 "st1 { v30.d }[0], [x10], #0x8\n"
490 "st1 { v31.d }[0], [x9], #0x8\n"
491 "tbz %x[n_channels], #0, 29f\n"
492 "st1 { v28.s }[2], [x12], #0x4\n"
493 "st1 { v29.s }[2], [x11], #0x4\n"
494 "st1 { v30.s }[2], [x10], #0x4\n"
495 "st1 { v31.s }[2], [x9], #0x4\n"
498 "st1 { v28.s }[0], [x12], #0x4\n"
499 "st1 { v29.s }[0], [x11], #0x4\n"
500 "st1 { v30.s }[0], [x10], #0x4\n"
501 "st1 { v31.s }[0], [x9], #0x4\n"
505 : [n_channels]
"r" ((
unsigned long) n_channels), [offsetof_Args_inptrs]
"I" (offsetof(Args, inptrs)), [offsetof_args_max]
"I" (offsetof(Args, max)), [offsetof_args_min]
"I" (offsetof(Args, min)), [offsetof_args_outptrs]
"I" (offsetof(Args, outptrs)), [offsetof_args_params]
"I" (offsetof(Args, params)), [params_struct]
"r" (¶ms_struct)
506 :
"cc",
"memory",
"v0",
"v1",
"v2",
"v3",
"v4",
"v5",
"v6",
"v7",
"v8",
"v9",
"v10",
"v11",
"v12",
"v13",
"v16",
"v17",
"v18",
"v19",
"v20",
"v21",
"v22",
"v23",
"v24",
"v25",
"v26",
"v27",
"v28",
"v29",
"v30",
"v31",
"x9",
"x10",
"x11",
"x12",
"x13",
"x14",
"x15",
"x16",
"x20",
"x21",
"x22",
"x23",
"x24",
"x25",
"x26",
"x27",
"x28"
513 #endif // defined(__aarch64__)