25     const struct transpose_context* context,  in xnn_compute_transposec_2d()  argument
31   const size_t log2_element_size = context->log2_element_size;  in xnn_compute_transposec_2d()
33   context->const_size_ukernel(  in xnn_compute_transposec_2d()
34 …  (const void*) ((uintptr_t) context->x + (i << log2_element_size) + j * context->input_stride[1]),  in xnn_compute_transposec_2d()
35       (void*) ((uintptr_t) context->y + (j << log2_element_size) + i * context->output_stride[0]),  in xnn_compute_transposec_2d()
36       context->input_stride[1],  in xnn_compute_transposec_2d()
37       context->output_stride[0],  in xnn_compute_transposec_2d()
43     const struct transpose_context* context,  in xnn_compute_transposec_3d()  argument
50   const size_t log2_element_size = context->log2_element_size;  in xnn_compute_transposec_3d()
51   const size_t ld_input = context->input_stride[2];  in xnn_compute_transposec_3d()
52   const size_t ld_output = context->output_stride[1];  in xnn_compute_transposec_3d()
53   const void* x = (const void*) ((uintptr_t) context->x +  in xnn_compute_transposec_3d()
54 …                     (i * context->input_stride[0] + j * context->input_stride[1]) + k * ld_input);  in xnn_compute_transposec_3d()
55 …void* y = (void*) ((uintptr_t)context->y + i * context->output_stride[0] + j * context->output_str…  in xnn_compute_transposec_3d()
58   context->const_size_ukernel(  in xnn_compute_transposec_3d()
68     const struct transpose_context* context,  in xnn_compute_transposec_4d()  argument
76   const size_t log2_element_size = context->log2_element_size;  in xnn_compute_transposec_4d()
77   const size_t ld_input = context->input_stride[3];  in xnn_compute_transposec_4d()
78   const size_t ld_output = context->output_stride[2];  in xnn_compute_transposec_4d()
79 …const void* x = (const void*) ((uintptr_t)context->x + i * context->input_stride[0] + j * context-…  in xnn_compute_transposec_4d()
80                                  k * context->input_stride[2] + l * ld_input);  in xnn_compute_transposec_4d()
81 …void* y = (void*) ((uintptr_t)context->y + i * context->output_stride[0] + j * context->output_str…  in xnn_compute_transposec_4d()
82                      k * context->output_stride[2] + (l << log2_element_size));  in xnn_compute_transposec_4d()
84   context->const_size_ukernel(  in xnn_compute_transposec_4d()
94     const struct transpose_context* context,  in xnn_compute_transposec_5d()  argument
103   const size_t log2_element_size = context->log2_element_size;  in xnn_compute_transposec_5d()
104   const size_t ld_input = context->input_stride[4];  in xnn_compute_transposec_5d()
105   const size_t ld_output = context->output_stride[3];  in xnn_compute_transposec_5d()
106 …const void* x = (const void*)((uintptr_t)context->x + i * context->input_stride[0] + j * context->…  in xnn_compute_transposec_5d()
107 …                       k * context->input_stride[2] + l * context->input_stride[3] + m * ld_input);  in xnn_compute_transposec_5d()
108 …void* y = (void*)((uintptr_t)context->y + i * context->output_stride[0] + j * context->output_stri…  in xnn_compute_transposec_5d()
109 …         k * context->output_stride[2] + l * context->output_stride[3] + (m << log2_element_size));  in xnn_compute_transposec_5d()
111   context->const_size_ukernel(  in xnn_compute_transposec_5d()
121     const struct transpose_context* context,  in xnn_compute_transposec_6d()  argument
131   const size_t log2_element_size = context->log2_element_size;  in xnn_compute_transposec_6d()
132   const size_t ld_input = context->input_stride[5];  in xnn_compute_transposec_6d()
133   const size_t ld_output = context->output_stride[4];  in xnn_compute_transposec_6d()
134 …const void* x = (const void*)((uintptr_t)context->x + i * context->input_stride[0] + j * context->…  in xnn_compute_transposec_6d()
135                                  k * context->input_stride[2] + l * context->input_stride[3] +  in xnn_compute_transposec_6d()
136                                  m * context->input_stride[4] + n * ld_input);  in xnn_compute_transposec_6d()
137 …void* y = (void*)((uintptr_t)context->y + i * context->output_stride[0] + j * context->output_stri…  in xnn_compute_transposec_6d()
138 …    k * context->output_stride[2] + l * context->output_stride[3] + m * context->output_stride[4] +  in xnn_compute_transposec_6d()
141   context->const_size_ukernel(  in xnn_compute_transposec_6d()
151     const struct transpose_context* context,  in xnn_compute_transposev_2d()  argument
157   const size_t element_size = context->element_size;  in xnn_compute_transposev_2d()
158   const size_t ld_input = context->input_stride[1];  in xnn_compute_transposev_2d()
159   const size_t ld_output = context->output_stride[0];  in xnn_compute_transposev_2d()
160   const void* x = (const void*) ((uintptr_t) context->x +  in xnn_compute_transposev_2d()
161                                  i * context->input_stride[0] + j * ld_input);  in xnn_compute_transposev_2d()
162 …void* y = (void*) ((uintptr_t) context->y + context->output_stride[1] * j + i * context->output_st…  in xnn_compute_transposev_2d()
164   context->variable_size_ukernel(  in xnn_compute_transposev_2d()
169       context->input_stride[0],  in xnn_compute_transposev_2d()
170       context->output_stride[1],  in xnn_compute_transposev_2d()
177     const struct transpose_context* context,  in xnn_compute_transposev_3d()  argument
184   const size_t element_size = context->element_size;  in xnn_compute_transposev_3d()
185   const size_t ld_input = context->input_stride[2];  in xnn_compute_transposev_3d()
186   const size_t ld_output = context->output_stride[1];  in xnn_compute_transposev_3d()
187 …const void* x = (const void*)((uintptr_t)context->x + i * context->input_stride[0] + j * context->…  in xnn_compute_transposev_3d()
189 …void* y = (void*)((uintptr_t)context->y + i * context->output_stride[0] + j * context->output_stri…  in xnn_compute_transposev_3d()
190                      k * context->output_stride[2]);  in xnn_compute_transposev_3d()
192   context->variable_size_ukernel(  in xnn_compute_transposev_3d()
197       context->input_stride[1],  in xnn_compute_transposev_3d()
198       context->output_stride[2],  in xnn_compute_transposev_3d()
205     const struct transpose_context* context,  in xnn_compute_transposev_4d()  argument
213   const size_t element_size = context->element_size;  in xnn_compute_transposev_4d()
214   const size_t ld_input = context->input_stride[3];  in xnn_compute_transposev_4d()
215   const size_t ld_output = context->output_stride[2];  in xnn_compute_transposev_4d()
216 …const void* x = (const void*)((uintptr_t)context->x + i * context->input_stride[0] + j * context->…  in xnn_compute_transposev_4d()
217                                  k * context->input_stride[2] + l * ld_input);  in xnn_compute_transposev_4d()
218 …void* y = (void*)((uintptr_t)context->y + context->output_stride[3] * l + i * context->output_stri…  in xnn_compute_transposev_4d()
219                      j * context->output_stride[1] + k * context->output_stride[2]);  in xnn_compute_transposev_4d()
221   context->variable_size_ukernel(  in xnn_compute_transposev_4d()
226       context->input_stride[2],  in xnn_compute_transposev_4d()
227       context->output_stride[3],  in xnn_compute_transposev_4d()
234     const struct transpose_context* context,  in xnn_compute_transposev_5d()  argument
243   const size_t element_size = context->element_size;  in xnn_compute_transposev_5d()
244   const size_t ld_input = context->input_stride[4];  in xnn_compute_transposev_5d()
245   const size_t ld_output = context->output_stride[3];  in xnn_compute_transposev_5d()
246 …const void* x = (const void*)((uintptr_t)context->x + i * context->input_stride[0] + j * context->…  in xnn_compute_transposev_5d()
247 …                       k * context->input_stride[2] + l * context->input_stride[3] + m * ld_input);  in xnn_compute_transposev_5d()
248 …void* y = (void*)((uintptr_t)context->y + context->output_stride[4] * m + i * context->output_stri…  in xnn_compute_transposev_5d()
249 …    j * context->output_stride[1] + k * context->output_stride[2] + l * context->output_stride[3]);  in xnn_compute_transposev_5d()
251   context->variable_size_ukernel(  in xnn_compute_transposev_5d()
256       context->input_stride[3],  in xnn_compute_transposev_5d()
257       context->output_stride[4],  in xnn_compute_transposev_5d()
264     const struct transpose_context* context,  in xnn_compute_transposev_6d()  argument
274   const size_t element_size = context->element_size;  in xnn_compute_transposev_6d()
275   const size_t ld_input = context->input_stride[5];  in xnn_compute_transposev_6d()
276   const size_t ld_output = context->output_stride[4];  in xnn_compute_transposev_6d()
277 …const void* x = (const void*)((uintptr_t)context->x + i * context->input_stride[0] + j * context->…  in xnn_compute_transposev_6d()
278                                  k * context->input_stride[2] + l * context->input_stride[3] +  in xnn_compute_transposev_6d()
279                                  m * context->input_stride[4] + n * ld_input);  in xnn_compute_transposev_6d()
280 …void* y = (void*)((uintptr_t)context->y + context->output_stride[5] * n + i * context->output_stri…  in xnn_compute_transposev_6d()
281 …    j * context->output_stride[1] + k * context->output_stride[2] + l * context->output_stride[3] +  in xnn_compute_transposev_6d()
282                      m * context->output_stride[4]);  in xnn_compute_transposev_6d()
284   context->variable_size_ukernel(  in xnn_compute_transposev_6d()
289       context->input_stride[4],  in xnn_compute_transposev_6d()
290       context->output_stride[5],  in xnn_compute_transposev_6d()
297     const struct gemm_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_grouped_gemm()
304   const size_t k_scaled  = context->k_scaled;  in xnn_compute_grouped_gemm()
305   const size_t a_stride  = context->a_stride;  in xnn_compute_grouped_gemm()
306   const size_t cm_stride = context->cm_stride;  in xnn_compute_grouped_gemm()
308   context->ukernel.function[XNN_UARCH_DEFAULT](  in xnn_compute_grouped_gemm()
312       (const void*) ((uintptr_t) context->a + mr_block_start * a_stride + group_index * k_scaled),  in xnn_compute_grouped_gemm()
314 …(const void*) ((uintptr_t) context->packed_w + nr_block_start * context->w_stride + group_index * …  in xnn_compute_grouped_gemm()
315 …(void*) ((uintptr_t) context->c + mr_block_start * cm_stride + (nr_block_start << context->log2_cs…  in xnn_compute_grouped_gemm()
317       context->cn_stride,  in xnn_compute_grouped_gemm()
318       &context->params);  in xnn_compute_grouped_gemm()
322     const struct gemm_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_gemm()
328   const size_t a_stride  = context->a_stride;  in xnn_compute_gemm()
329   const size_t cm_stride = context->cm_stride;  in xnn_compute_gemm()
331   context->ukernel.function[XNN_UARCH_DEFAULT](  in xnn_compute_gemm()
334       context->k_scaled,  in xnn_compute_gemm()
335       (const void*) ((uintptr_t) context->a + mr_block_start * a_stride),  in xnn_compute_gemm()
337       (const void*) ((uintptr_t) context->packed_w + nr_block_start * context->w_stride),  in xnn_compute_gemm()
338 …(void*) ((uintptr_t) context->c + mr_block_start * cm_stride + (nr_block_start << context->log2_cs…  in xnn_compute_gemm()
340       context->cn_stride,  in xnn_compute_gemm()
341       context->fused_params);  in xnn_compute_gemm()
345     const struct spmm_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_spmm()
350   context->ukernel(  in xnn_compute_spmm()
352       context->n,  in xnn_compute_spmm()
353 …(const void*) ((uintptr_t) context->input + batch_index * context->batched_input_stride + mr_block…  in xnn_compute_spmm()
354       context->nonzero_weights,  in xnn_compute_spmm()
355       context->input_increments,  in xnn_compute_spmm()
356       context->output_channel_nonzeros,  in xnn_compute_spmm()
357 …(void*) ((uintptr_t) context->output + batch_index * context->batched_output_stride + mr_block_sta…  in xnn_compute_spmm()
358       context->scaled_m,  in xnn_compute_spmm()
359       &context->params);  in xnn_compute_spmm()
363     const struct igemm_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_grouped_batch_igemm()
371   const size_t ks        = context->ks;  in xnn_compute_grouped_batch_igemm()
372   const size_t cm_stride = context->cm_stride;  in xnn_compute_grouped_batch_igemm()
374   context->ukernel.function[XNN_UARCH_DEFAULT](  in xnn_compute_grouped_batch_igemm()
377       context->kc,  in xnn_compute_grouped_batch_igemm()
378       context->ks_scaled,  in xnn_compute_grouped_batch_igemm()
379       (const void**) ((uintptr_t) context->indirect_a + mr_block_start * ks * sizeof(void*)),  in xnn_compute_grouped_batch_igemm()
380 …(const void*) ((uintptr_t) context->packed_w + nr_block_start * context->w_stride + group_index * …  in xnn_compute_grouped_batch_igemm()
381 …ptr_t) context->c + group_index * context->gc_stride + batch_index * context->bc_stride + mr_block…  in xnn_compute_grouped_batch_igemm()
383       context->cn_stride,  in xnn_compute_grouped_batch_igemm()
384       context->a_offset + group_index * context->ga_stride + batch_index * context->ba_stride,  in xnn_compute_grouped_batch_igemm()
385       context->zero,  in xnn_compute_grouped_batch_igemm()
386       &context->params);  in xnn_compute_grouped_batch_igemm()
390     const struct igemm_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_grouped_igemm()
397   const size_t ks        = context->ks;  in xnn_compute_grouped_igemm()
398   const size_t cm_stride = context->cm_stride;  in xnn_compute_grouped_igemm()
400   context->ukernel.function[XNN_UARCH_DEFAULT](  in xnn_compute_grouped_igemm()
403       context->kc,  in xnn_compute_grouped_igemm()
404       context->ks_scaled,  in xnn_compute_grouped_igemm()
405       (const void**) ((uintptr_t) context->indirect_a + mr_block_start * ks * sizeof(void*)),  in xnn_compute_grouped_igemm()
406 …(const void*) ((uintptr_t) context->packed_w + nr_block_start * context->w_stride + group_index * …  in xnn_compute_grouped_igemm()
407 …(void*) ((uintptr_t) context->c + group_index * context->gc_stride + mr_block_start * cm_stride + …  in xnn_compute_grouped_igemm()
409       context->cn_stride,  in xnn_compute_grouped_igemm()
410       context->a_offset + group_index * context->ga_stride,  in xnn_compute_grouped_igemm()
411       context->zero,  in xnn_compute_grouped_igemm()
412       &context->params);  in xnn_compute_grouped_igemm()
416     const struct igemm_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_batch_igemm()
423   const size_t ks        = context->ks;  in xnn_compute_batch_igemm()
424   const size_t cm_stride = context->cm_stride;  in xnn_compute_batch_igemm()
426   context->ukernel.function[XNN_UARCH_DEFAULT](  in xnn_compute_batch_igemm()
429       context->kc,  in xnn_compute_batch_igemm()
430       context->ks_scaled,  in xnn_compute_batch_igemm()
431       (const void**) ((uintptr_t) context->indirect_a + mr_block_start * ks * sizeof(void*)),  in xnn_compute_batch_igemm()
432       (const void*) ((uintptr_t) context->packed_w + nr_block_start * context->w_stride),  in xnn_compute_batch_igemm()
433 …(void*) ((uintptr_t) context->c + batch_index * context->bc_stride + mr_block_start * cm_stride + …  in xnn_compute_batch_igemm()
435       context->cn_stride,  in xnn_compute_batch_igemm()
436       context->a_offset + batch_index * context->ba_stride,  in xnn_compute_batch_igemm()
437       context->zero,  in xnn_compute_batch_igemm()
438       &context->params);  in xnn_compute_batch_igemm()
442     const struct igemm_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_igemm()
448   const size_t ks        = context->ks;  in xnn_compute_igemm()
449   const size_t cm_stride = context->cm_stride;  in xnn_compute_igemm()
451   context->ukernel.function[XNN_UARCH_DEFAULT](  in xnn_compute_igemm()
454       context->kc,  in xnn_compute_igemm()
455       context->ks_scaled,  in xnn_compute_igemm()
456       (const void**) ((uintptr_t) context->indirect_a + mr_block_start * ks * sizeof(void*)),  in xnn_compute_igemm()
457       (const void*) ((uintptr_t) context->packed_w + nr_block_start * context->w_stride),  in xnn_compute_igemm()
458 …(void*) ((uintptr_t) context->c + mr_block_start * cm_stride + (nr_block_start << context->log2_cs…  in xnn_compute_igemm()
460       context->cn_stride,  in xnn_compute_igemm()
461       context->a_offset,  in xnn_compute_igemm()
462       context->zero,  in xnn_compute_igemm()
463       &context->params);  in xnn_compute_igemm()
467       const struct subgemm_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_grouped_subgemm2d()
477 …const struct subconvolution_params* subconvolution_params = &context->subconvolution_params[subker…  in xnn_compute_grouped_subgemm2d()
489   const size_t ax_stride = context->ax_stride;  in xnn_compute_grouped_subgemm2d()
490   const size_t cx_stride = context->cx_stride;  in xnn_compute_grouped_subgemm2d()
491   context->ukernel.function[XNN_UARCH_DEFAULT](  in xnn_compute_grouped_subgemm2d()
494       context->kc,  in xnn_compute_grouped_subgemm2d()
495 …(uintptr_t) context->a + group_index * context->ga_stride + slice_y * context->ay_stride + slice_x…  in xnn_compute_grouped_subgemm2d()
497 …ms->weights + nc_block_start * subconvolution_params->w_stride + group_index * context->gw_stride),  in xnn_compute_grouped_subgemm2d()
498 …dex * context->gc_stride + slice_y * context->cy_stride + slice_x_start * cx_stride + batch_index …  in xnn_compute_grouped_subgemm2d()
500       context->cn_stride,  in xnn_compute_grouped_subgemm2d()
501       &context->params);  in xnn_compute_grouped_subgemm2d()
505       const struct subgemm_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_subgemm2d()
514 …const struct subconvolution_params* subconvolution_params = &context->subconvolution_params[subker…  in xnn_compute_subgemm2d()
526   const size_t ax_stride = context->ax_stride;  in xnn_compute_subgemm2d()
527   const size_t cx_stride = context->cx_stride;  in xnn_compute_subgemm2d()
528   context->ukernel.function[XNN_UARCH_DEFAULT](  in xnn_compute_subgemm2d()
531       context->kc,  in xnn_compute_subgemm2d()
532 …(const void*) ((uintptr_t) context->a + slice_y * context->ay_stride + slice_x_start * ax_stride +…  in xnn_compute_subgemm2d()
535 …->output + slice_y * context->cy_stride + slice_x_start * cx_stride + batch_index * context->bc_st…  in xnn_compute_subgemm2d()
537       context->cn_stride,  in xnn_compute_subgemm2d()
538       &context->params);  in xnn_compute_subgemm2d()
542       const struct subconv_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_grouped_subconv2d()
552 …const struct subconvolution_params* subconvolution_params = &context->subconvolution_params[subker…  in xnn_compute_grouped_subconv2d()
564   const size_t cx_stride = context->cx_stride;  in xnn_compute_grouped_subconv2d()
565   context->ukernel.function[XNN_UARCH_DEFAULT](  in xnn_compute_grouped_subconv2d()
568       context->kc,  in xnn_compute_grouped_subconv2d()
571 …ms->weights + nc_block_start * subconvolution_params->w_stride + group_index * context->gw_stride),  in xnn_compute_grouped_subconv2d()
572 …dex * context->gc_stride + slice_y * context->cy_stride + slice_x_start * cx_stride + batch_index …  in xnn_compute_grouped_subconv2d()
574       context->cn_stride,  in xnn_compute_grouped_subconv2d()
575       context->a_offset + group_index * context->ga_stride + batch_index * context->ba_stride,  in xnn_compute_grouped_subconv2d()
576       context->zero,  in xnn_compute_grouped_subconv2d()
577       &context->params);  in xnn_compute_grouped_subconv2d()
581       const struct subconv_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_subconv2d()
590 …const struct subconvolution_params* subconvolution_params = &context->subconvolution_params[subker…  in xnn_compute_subconv2d()
602   const size_t cx_stride = context->cx_stride;  in xnn_compute_subconv2d()
603   context->ukernel.function[XNN_UARCH_DEFAULT](  in xnn_compute_subconv2d()
606       context->kc,  in xnn_compute_subconv2d()
610 …->output + slice_y * context->cy_stride + slice_x_start * cx_stride + batch_index * context->bc_st…  in xnn_compute_subconv2d()
612       context->cn_stride,  in xnn_compute_subconv2d()
613       context->a_offset + batch_index * context->ba_stride,  in xnn_compute_subconv2d()
614       context->zero,  in xnn_compute_subconv2d()
615       &context->params);  in xnn_compute_subconv2d()
619       const struct conv2d_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_conv2d_hwc2chw()
624   context->hwc2chw_ukernel(  in xnn_compute_conv2d_hwc2chw()
625       context->input_height,  in xnn_compute_conv2d_hwc2chw()
626       context->input_width,  in xnn_compute_conv2d_hwc2chw()
629       (const void*) ((uintptr_t) context->input + batch_index * context->input_batch_stride),  in xnn_compute_conv2d_hwc2chw()
630       context->zero,  in xnn_compute_conv2d_hwc2chw()
631       context->packed_weights,  in xnn_compute_conv2d_hwc2chw()
632       (void*) ((uintptr_t) context->output + batch_index * context->output_batch_stride),  in xnn_compute_conv2d_hwc2chw()
633       context->input_padding_top,  in xnn_compute_conv2d_hwc2chw()
634       context->output_channels,  in xnn_compute_conv2d_hwc2chw()
635       context->output_height_stride,  in xnn_compute_conv2d_hwc2chw()
636       context->output_channel_stride,  in xnn_compute_conv2d_hwc2chw()
637       &context->params);  in xnn_compute_conv2d_hwc2chw()
641     const struct dwconv_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_dwconv_unipass()
646 …(const void**) ((uintptr_t) context->indirect_input + output_y * context->indirect_input_height_st…  in xnn_compute_dwconv_unipass()
647   const size_t input_offset = context->input_offset + batch_index * context->input_batch_stride;  in xnn_compute_dwconv_unipass()
648   void* output = (void*) ((uintptr_t) context->output +  in xnn_compute_dwconv_unipass()
649     batch_index * context->output_batch_stride + output_y * context->output_height_stride);  in xnn_compute_dwconv_unipass()
651   context->unipass_ukernel(  in xnn_compute_dwconv_unipass()
652     context->groups, context->output_width,  in xnn_compute_dwconv_unipass()
653     indirect_input, context->packed_weights, output,  in xnn_compute_dwconv_unipass()
654     context->indirect_input_width_stride, context->output_increment,  in xnn_compute_dwconv_unipass()
655     input_offset, context->zero,  in xnn_compute_dwconv_unipass()
656     &context->params);  in xnn_compute_dwconv_unipass()
660     const struct dwconv2d_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_dwconv2d_chw()
664   context->chw_ukernel(  in xnn_compute_dwconv2d_chw()
665     context->input_height,  in xnn_compute_dwconv2d_chw()
666     context->input_width,  in xnn_compute_dwconv2d_chw()
667 …(const void*) ((uintptr_t) context->input + channel * context->input_channel_stride + batch_index …  in xnn_compute_dwconv2d_chw()
668     (const void*) ((uintptr_t) context->packed_weights + channel * context->weights_channel_stride),  in xnn_compute_dwconv2d_chw()
669     context->zero,  in xnn_compute_dwconv2d_chw()
670 …(void*) ((uintptr_t) context->output + channel * context->output_channel_stride + batch_index * co…  in xnn_compute_dwconv2d_chw()
671     context->input_padding_top,  in xnn_compute_dwconv2d_chw()
672     &context->params);  in xnn_compute_dwconv2d_chw()
676     const struct argmax_pooling_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_argmax_pooling_unipass()
680   const void** indirect_input = (const void**) ((uintptr_t) context->indirect_input +  in xnn_compute_argmax_pooling_unipass()
681     output_y * context->indirect_input_height_stride);  in xnn_compute_argmax_pooling_unipass()
682   const size_t input_offset = context->input_offset + batch_index * context->input_batch_stride;  in xnn_compute_argmax_pooling_unipass()
683   void* output = (void*) ((uintptr_t) context->output +  in xnn_compute_argmax_pooling_unipass()
684     batch_index * context->output_batch_stride + output_y * context->output_height_stride);  in xnn_compute_argmax_pooling_unipass()
685   uint32_t* index = (uint32_t*) ((uintptr_t) context->index +  in xnn_compute_argmax_pooling_unipass()
686     batch_index * context->index_batch_stride + output_y * context->index_height_stride);  in xnn_compute_argmax_pooling_unipass()
688   context->unipass_ukernel(  in xnn_compute_argmax_pooling_unipass()
689     context->output_width, context->pooling_size, context->channels,  in xnn_compute_argmax_pooling_unipass()
691     context->input_increment, context->output_increment);  in xnn_compute_argmax_pooling_unipass()
695     const struct argmax_pooling_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_argmax_pooling_multipass()
699   const void** indirect_input = (const void**) ((uintptr_t) context->indirect_input +  in xnn_compute_argmax_pooling_multipass()
700     output_y * context->indirect_input_height_stride);  in xnn_compute_argmax_pooling_multipass()
701   const size_t input_offset = context->input_offset + batch_index * context->input_batch_stride;  in xnn_compute_argmax_pooling_multipass()
702   void* output = (void*) ((uintptr_t) context->output +  in xnn_compute_argmax_pooling_multipass()
703     batch_index * context->output_batch_stride + output_y * context->output_height_stride);  in xnn_compute_argmax_pooling_multipass()
704   uint32_t* index = (uint32_t*) ((uintptr_t) context->index +  in xnn_compute_argmax_pooling_multipass()
705     batch_index * context->index_batch_stride + output_y * context->index_height_stride);  in xnn_compute_argmax_pooling_multipass()
707 …void* multipass_accumulation_buffer = XNN_SIMD_ALLOCA(context->channels * sizeof(float) + XNN_EXTR…  in xnn_compute_argmax_pooling_multipass()
708 …void* multipass_index_buffer = XNN_SIMD_ALLOCA(context->channels * sizeof(uint32_t) + XNN_EXTRA_BY…  in xnn_compute_argmax_pooling_multipass()
710   context->multipass_ukernel(  in xnn_compute_argmax_pooling_multipass()
711     context->output_width, context->pooling_size, context->channels,  in xnn_compute_argmax_pooling_multipass()
713     context->input_increment, context->output_increment);  in xnn_compute_argmax_pooling_multipass()
717     const struct max_pooling_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_max_pooling()
721   const void** indirect_input = (const void**) ((uintptr_t) context->indirect_input +  in xnn_compute_max_pooling()
722     output_y * context->indirect_input_height_stride);  in xnn_compute_max_pooling()
723   const size_t input_offset = context->input_offset + batch_index * context->input_batch_stride;  in xnn_compute_max_pooling()
724   void* output = (void*) ((uintptr_t) context->output +  in xnn_compute_max_pooling()
725     batch_index * context->output_batch_stride + output_y * context->output_height_stride);  in xnn_compute_max_pooling()
727   context->ukernel(  in xnn_compute_max_pooling()
728     context->output_width, context->pooling_size, context->channels,  in xnn_compute_max_pooling()
730     context->input_increment, context->output_increment,  in xnn_compute_max_pooling()
731     &context->params);  in xnn_compute_max_pooling()
735     const struct unpooling_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_unpooling()
739   const void* input = (const void*) ((uintptr_t) context->input +  in xnn_compute_unpooling()
740       input_y * context->input_height_stride + input_x * context->input_width_stride);  in xnn_compute_unpooling()
741   const uint32_t* index = (const uint32_t*) ((uintptr_t) context->index +  in xnn_compute_unpooling()
742       input_y * context->index_height_stride + input_x * context->index_width_stride);  in xnn_compute_unpooling()
744     (void**) ((uintptr_t) context->indirect_output +  in xnn_compute_unpooling()
745 …input_y * context->indirect_output_height_stride + input_x * context->indirect_output_width_stride…  in xnn_compute_unpooling()
747   context->ukernel(  in xnn_compute_unpooling()
748     context->pooling_size,  in xnn_compute_unpooling()
749     context->channels,  in xnn_compute_unpooling()
750     context->fill_value,  in xnn_compute_unpooling()
755     const struct average_pooling_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_average_pooling_unipass()
760 …(const void**) ((uintptr_t) context->indirect_input + output_y * context->indirect_input_height_st…  in xnn_compute_average_pooling_unipass()
761   const size_t input_offset = context->input_offset + batch_index * context->input_batch_stride;  in xnn_compute_average_pooling_unipass()
762   void* output = (void*) ((uintptr_t) context->output +  in xnn_compute_average_pooling_unipass()
763     batch_index * context->output_batch_stride + output_y * context->output_height_stride);  in xnn_compute_average_pooling_unipass()
765   context->unipass_ukernel(  in xnn_compute_average_pooling_unipass()
766     context->output_width, context->pooling_size, context->channels,  in xnn_compute_average_pooling_unipass()
767     indirect_input, input_offset, context->zero, output,  in xnn_compute_average_pooling_unipass()
768     context->input_increment, context->output_increment,  in xnn_compute_average_pooling_unipass()
769     &context->params);  in xnn_compute_average_pooling_unipass()
773     const struct average_pooling_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_average_pooling_multipass()
778 …(const void**) ((uintptr_t) context->indirect_input + output_y * context->indirect_input_height_st…  in xnn_compute_average_pooling_multipass()
779   const size_t input_offset = context->input_offset + batch_index * context->input_batch_stride;  in xnn_compute_average_pooling_multipass()
780   void* output = (void*) ((uintptr_t) context->output +  in xnn_compute_average_pooling_multipass()
781     batch_index * context->output_batch_stride + output_y * context->output_height_stride);  in xnn_compute_average_pooling_multipass()
784 …XNN_SIMD_ALLOCA(context->channels * sizeof(int32_t) + XNN_EXTRA_BYTES * sizeof(int32_t) / sizeof(u…  in xnn_compute_average_pooling_multipass()
786   context->multipass_ukernel(  in xnn_compute_average_pooling_multipass()
787     context->output_width, context->pooling_size, context->channels,  in xnn_compute_average_pooling_multipass()
788     indirect_input, input_offset, context->zero, multipass_buffer, output,  in xnn_compute_average_pooling_multipass()
789     context->input_increment, context->output_increment,  in xnn_compute_average_pooling_multipass()
790     &context->params);  in xnn_compute_average_pooling_multipass()
794     const struct pixelwise_average_pooling_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_pixelwise_average_pooling_unipass()
799 …(const void**) ((uintptr_t) context->indirect_input + output_y * context->indirect_input_height_st…  in xnn_compute_pixelwise_average_pooling_unipass()
800   const size_t input_offset = context->input_offset + batch_index * context->input_batch_stride;  in xnn_compute_pixelwise_average_pooling_unipass()
802 …(const void*) ((uintptr_t) context->pixelwise_buffer + output_y * context->pixelwise_buffer_height…  in xnn_compute_pixelwise_average_pooling_unipass()
803   void* output = (void*) ((uintptr_t) context->output +  in xnn_compute_pixelwise_average_pooling_unipass()
804     batch_index * context->output_batch_stride + output_y * context->output_height_stride);  in xnn_compute_pixelwise_average_pooling_unipass()
806   context->unipass_ukernel(  in xnn_compute_pixelwise_average_pooling_unipass()
807     context->output_width, context->pooling_size, context->channels,  in xnn_compute_pixelwise_average_pooling_unipass()
808     indirect_input, input_offset, context->zero, pixelwise_buffer, output,  in xnn_compute_pixelwise_average_pooling_unipass()
809     context->input_increment, context->output_increment,  in xnn_compute_pixelwise_average_pooling_unipass()
810     &context->params);  in xnn_compute_pixelwise_average_pooling_unipass()
814     const struct pixelwise_average_pooling_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_pixelwise_average_pooling_multipass()
819 …(const void**) ((uintptr_t) context->indirect_input + output_y * context->indirect_input_height_st…  in xnn_compute_pixelwise_average_pooling_multipass()
820   const size_t input_offset = context->input_offset + batch_index * context->input_batch_stride;  in xnn_compute_pixelwise_average_pooling_multipass()
822 …(const void*) ((uintptr_t) context->pixelwise_buffer + output_y * context->pixelwise_buffer_height…  in xnn_compute_pixelwise_average_pooling_multipass()
823   void* output = (void*) ((uintptr_t) context->output +  in xnn_compute_pixelwise_average_pooling_multipass()
824     batch_index * context->output_batch_stride + output_y * context->output_height_stride);  in xnn_compute_pixelwise_average_pooling_multipass()
826 …void* multipass_buffer = XNN_SIMD_ALLOCA(context->channels * sizeof(int32_t) + XNN_EXTRA_BYTES * s…  in xnn_compute_pixelwise_average_pooling_multipass()
828   context->multipass_ukernel(  in xnn_compute_pixelwise_average_pooling_multipass()
829     context->output_width, context->pooling_size, context->channels,  in xnn_compute_pixelwise_average_pooling_multipass()
830     indirect_input, input_offset, context->zero, pixelwise_buffer, multipass_buffer, output,  in xnn_compute_pixelwise_average_pooling_multipass()
831     context->input_increment, context->output_increment,  in xnn_compute_pixelwise_average_pooling_multipass()
832     &context->params);  in xnn_compute_pixelwise_average_pooling_multipass()
836     const struct global_average_pooling_nwc_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_global_average_pooling_nwc_unipass()
840     (const void*) ((uintptr_t) context->input + batch_index * context->input_batch_stride);  in xnn_compute_global_average_pooling_nwc_unipass()
842     (void*) ((uintptr_t) context->output + batch_index * context->output_batch_stride);  in xnn_compute_global_average_pooling_nwc_unipass()
844   context->unipass_ukernel(  in xnn_compute_global_average_pooling_nwc_unipass()
845     context->input_elements,  in xnn_compute_global_average_pooling_nwc_unipass()
846     context->channels,  in xnn_compute_global_average_pooling_nwc_unipass()
848     context->input_pixel_stride,  in xnn_compute_global_average_pooling_nwc_unipass()
849     context->zero,  in xnn_compute_global_average_pooling_nwc_unipass()
851     &context->params);  in xnn_compute_global_average_pooling_nwc_unipass()
855     const struct global_average_pooling_nwc_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_global_average_pooling_nwc_multipass()
859     (const void*) ((uintptr_t) context->input + batch_index * context->input_batch_stride);  in xnn_compute_global_average_pooling_nwc_multipass()
861     (void*) ((uintptr_t) context->output + batch_index * context->output_batch_stride);  in xnn_compute_global_average_pooling_nwc_multipass()
864 …XNN_SIMD_ALLOCA(context->channels * sizeof(int32_t) + XNN_EXTRA_BYTES * sizeof(int32_t) / sizeof(u…  in xnn_compute_global_average_pooling_nwc_multipass()
866   context->multipass_ukernel(  in xnn_compute_global_average_pooling_nwc_multipass()
867     context->input_elements,  in xnn_compute_global_average_pooling_nwc_multipass()
868     context->channels,  in xnn_compute_global_average_pooling_nwc_multipass()
870     context->input_pixel_stride,  in xnn_compute_global_average_pooling_nwc_multipass()
871     context->zero,  in xnn_compute_global_average_pooling_nwc_multipass()
874     &context->params);  in xnn_compute_global_average_pooling_nwc_multipass()
878     const struct global_average_pooling_ncw_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_global_average_pooling_ncw()
883   const void* input = (const void*) ((uintptr_t) context->input +  in xnn_compute_global_average_pooling_ncw()
884     channels_start * context->input_channel_stride + batch_index * context->input_batch_stride);  in xnn_compute_global_average_pooling_ncw()
885   void* output = (void*) ((uintptr_t) context->output +  in xnn_compute_global_average_pooling_ncw()
886     channels_start * context->output_channel_stride + batch_index * context->output_batch_stride);  in xnn_compute_global_average_pooling_ncw()
888   context->ukernel(  in xnn_compute_global_average_pooling_ncw()
889     context->input_elements,  in xnn_compute_global_average_pooling_ncw()
893     &context->params);  in xnn_compute_global_average_pooling_ncw()
897     const struct resize_bilinear_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_resize_bilinear()
903 …(void*) ((uintptr_t) context->output + pixel_start * context->output_pixel_stride + batch_index * …  in xnn_compute_resize_bilinear()
905   context->ukernel(  in xnn_compute_resize_bilinear()
907     context->scaled_channels,  in xnn_compute_resize_bilinear()
908     context->indirect_input + pixel_start * 4,  in xnn_compute_resize_bilinear()
909     context->input_offset + batch_index * context->input_batch_stride,  in xnn_compute_resize_bilinear()
910     (const void*) ((uintptr_t) context->packed_weights + (pixel_start << context->log2_wsize)),  in xnn_compute_resize_bilinear()
912     context->output_pixel_stride - context->scaled_channels);  in xnn_compute_resize_bilinear()
916     const struct resize_bilinear_chw_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_resize_bilinear_chw()
922 …(void*) ((uintptr_t) context->output + channel_start * context->output_channel_stride + batch_inde…  in xnn_compute_resize_bilinear_chw()
923 …const size_t input_offset = context->input_offset + batch_index * context->input_batch_stride + ch…  in xnn_compute_resize_bilinear_chw()
925   context->ukernel(  in xnn_compute_resize_bilinear_chw()
926     context->output_pixels,  in xnn_compute_resize_bilinear_chw()
928     context->indirect_input,  in xnn_compute_resize_bilinear_chw()
930     context->packed_weights,  in xnn_compute_resize_bilinear_chw()
932     context->input_channel_stride);  in xnn_compute_resize_bilinear_chw()
936     const struct prelu_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_prelu()
940   const size_t x_stride = context->x_stride;  in xnn_compute_prelu()
941   const size_t y_stride = context->y_stride;  in xnn_compute_prelu()
942   const void* x = (const void*) ((uintptr_t) context->x + x_stride * batch_start);  in xnn_compute_prelu()
943   void* y = (void*) ((uintptr_t) context->y + y_stride * batch_start);  in xnn_compute_prelu()
945   context->ukernel(batch_range, context->n, x, x_stride, context->w, y, y_stride);  in xnn_compute_prelu()
949     const struct pad_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_pad_5d()
952   const void* input = (const void*) ((uintptr_t) context->input +  in xnn_compute_pad_5d()
953 …i * context->input_stride[4] + j * context->input_stride[3] + k * context->input_stride[2] + l * c…  in xnn_compute_pad_5d()
954   void* output = (void*) ((uintptr_t) context->output +  in xnn_compute_pad_5d()
955 …i * context->output_stride[4] + j * context->output_stride[3] + k * context->output_stride[2] + l …  in xnn_compute_pad_5d()
957   const size_t i_padding = context->pre_paddings[5];  in xnn_compute_pad_5d()
958   const size_t j_padding = context->pre_paddings[4];  in xnn_compute_pad_5d()
959   const size_t k_padding = context->pre_paddings[3];  in xnn_compute_pad_5d()
960   const size_t l_padding = context->pre_paddings[2];  in xnn_compute_pad_5d()
961   const size_t m_padding = context->pre_paddings[1];  in xnn_compute_pad_5d()
963   const size_t i_size = context->input_size[5];  in xnn_compute_pad_5d()
964   const size_t j_size = context->input_size[4];  in xnn_compute_pad_5d()
965   const size_t k_size = context->input_size[3];  in xnn_compute_pad_5d()
966   const size_t l_size = context->input_size[2];  in xnn_compute_pad_5d()
967   const size_t m_size = context->input_size[1];  in xnn_compute_pad_5d()
972     context->pad_ukernel(  in xnn_compute_pad_5d()
974       context->input_size[0], context->pre_paddings[0], context->post_paddings[0],  in xnn_compute_pad_5d()
976       context->padding_value);  in xnn_compute_pad_5d()
978 …context->fill_ukernel(1 /* rows */, context->output_size[0], output, 0 /* output stride */, contex…  in xnn_compute_pad_5d()
983     const struct elementwise_binary_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_elementwise_binary_1d()
986   const void* a = (const void*) ((uintptr_t) context->a + i * context->a_stride[4]);  in xnn_compute_elementwise_binary_1d()
987   const void* b = (const void*) ((uintptr_t) context->b + i * context->b_stride[4]);  in xnn_compute_elementwise_binary_1d()
988   void* y = (void*) ((uintptr_t) context->y + i * context->y_stride[4]);  in xnn_compute_elementwise_binary_1d()
989   context->ukernel(context->elements, a, b, y, &context->params);  in xnn_compute_elementwise_binary_1d()
993     const struct elementwise_binary_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_elementwise_binary_2d()
996 …const void* a = (const void*) ((uintptr_t) context->a + i * context->a_stride[3] + j * context->a_…  in xnn_compute_elementwise_binary_2d()
997 …const void* b = (const void*) ((uintptr_t) context->b + i * context->b_stride[3] + j * context->b_…  in xnn_compute_elementwise_binary_2d()
998   void* y = (void*) ((uintptr_t) context->y + i * context->y_stride[3] + j * context->y_stride[4]);  in xnn_compute_elementwise_binary_2d()
999   context->ukernel(context->elements, a, b, y, &context->params);  in xnn_compute_elementwise_binary_2d()
1003     const struct elementwise_binary_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_elementwise_binary_3d()
1006   const void* a = (const void*) ((uintptr_t) context->a +  in xnn_compute_elementwise_binary_3d()
1007     i * context->a_stride[2] + j * context->a_stride[3] + k * context->a_stride[4]);  in xnn_compute_elementwise_binary_3d()
1008   const void* b = (const void*) ((uintptr_t) context->b +  in xnn_compute_elementwise_binary_3d()
1009     i * context->b_stride[2] + j * context->b_stride[3] + k * context->b_stride[4]);  in xnn_compute_elementwise_binary_3d()
1010   void* y = (void*) ((uintptr_t) context->y +  in xnn_compute_elementwise_binary_3d()
1011     i * context->y_stride[2] + j * context->y_stride[3] + k * context->y_stride[4]);  in xnn_compute_elementwise_binary_3d()
1012   context->ukernel(context->elements, a, b, y, &context->params);  in xnn_compute_elementwise_binary_3d()
1016     const struct elementwise_binary_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_elementwise_binary_4d()
1019   const void* a = (const void*) ((uintptr_t) context->a +  in xnn_compute_elementwise_binary_4d()
1020 …i * context->a_stride[1] + j * context->a_stride[2] + k * context->a_stride[3] + l * context->a_st…  in xnn_compute_elementwise_binary_4d()
1021   const void* b = (const void*) ((uintptr_t) context->b +  in xnn_compute_elementwise_binary_4d()
1022 …i * context->b_stride[1] + j * context->b_stride[2] + k * context->b_stride[3] + l * context->b_st…  in xnn_compute_elementwise_binary_4d()
1023   void* y = (void*) ((uintptr_t) context->y +  in xnn_compute_elementwise_binary_4d()
1024 …i * context->y_stride[1] + j * context->y_stride[2] + k * context->y_stride[3] + l * context->y_st…  in xnn_compute_elementwise_binary_4d()
1025   context->ukernel(context->elements, a, b, y, &context->params);  in xnn_compute_elementwise_binary_4d()
1029     const struct elementwise_binary_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_elementwise_binary_5d()
1032   const void* a = (const void*) ((uintptr_t) context->a +  in xnn_compute_elementwise_binary_5d()
1033 …i * context->a_stride[0] + j * context->a_stride[1] + k * context->a_stride[2] + l * context->a_st…  in xnn_compute_elementwise_binary_5d()
1034   const void* b = (const void*) ((uintptr_t) context->b +  in xnn_compute_elementwise_binary_5d()
1035 …i * context->b_stride[0] + j * context->b_stride[1] + k * context->b_stride[2] + l * context->b_st…  in xnn_compute_elementwise_binary_5d()
1036   void* y = (void*) ((uintptr_t) context->y +  in xnn_compute_elementwise_binary_5d()
1037 …i * context->y_stride[0] + j * context->y_stride[1] + k * context->y_stride[2] + l * context->y_st…  in xnn_compute_elementwise_binary_5d()
1038   context->ukernel(context->elements, a, b, y, &context->params);  in xnn_compute_elementwise_binary_5d()
1042     const struct channel_shuffle_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_channel_shuffle_fixed()
1045   const void* x = (const void*) ((uintptr_t) context->x + index * context->x_stride);  in xnn_compute_channel_shuffle_fixed()
1046   void* y = (void*) ((uintptr_t) context->y + index * context->y_stride);  in xnn_compute_channel_shuffle_fixed()
1048   context->fixed_ukernel(context->n, x, y);  in xnn_compute_channel_shuffle_fixed()
1052     const struct channel_shuffle_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_channel_shuffle_variable()
1055   const void* x = (const void*) ((uintptr_t) context->x + index * context->x_stride);  in xnn_compute_channel_shuffle_variable()
1056   void* y = (void*) ((uintptr_t) context->y + index * context->y_stride);  in xnn_compute_channel_shuffle_variable()
1058   context->variable_ukernel(context->n, context->m, x, y);  in xnn_compute_channel_shuffle_variable()
1062     const struct lut_strided_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_lut_strided()
1065   const void* x = (const void*) ((uintptr_t) context->x + context->x_stride * batch_index);  in xnn_compute_lut_strided()
1066   void* y = (void*) ((uintptr_t) context->y + context->y_stride * batch_index);  in xnn_compute_lut_strided()
1068   context->ukernel(context->n, x, y, context->t);  in xnn_compute_lut_strided()
1072     const struct lut_contiguous_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_lut_contiguous()
1076   const void* x = (const void*) ((uintptr_t) context->x + offset);  in xnn_compute_lut_contiguous()
1077   void* y = (void*) ((uintptr_t) context->y + offset);  in xnn_compute_lut_contiguous()
1079   context->ukernel(size, x, y, context->t);  in xnn_compute_lut_contiguous()
1083     const struct univector_strided_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_univector_strided()
1087   const size_t x_stride = context->x_stride;  in xnn_compute_univector_strided()
1088   const size_t y_stride = context->y_stride;  in xnn_compute_univector_strided()
1090   const void* x = (const void*) ((uintptr_t) context->x + x_stride * batch_index);  in xnn_compute_univector_strided()
1091   void* y = (void*) ((uintptr_t) context->y + y_stride * batch_index);  in xnn_compute_univector_strided()
1093     context->ukernel(context->n, x, y, &context->params);  in xnn_compute_univector_strided()
1100     const struct univector_contiguous_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_univector_contiguous()
1104   const uint32_t log2_xsize = context->log2_xsize;  in xnn_compute_univector_contiguous()
1105   const uint32_t log2_ysize = context->log2_ysize;  in xnn_compute_univector_contiguous()
1106   const void* x = (const void*) ((uintptr_t) context->x + offset);  in xnn_compute_univector_contiguous()
1107   void* y = (void*) ((uintptr_t) context->y + ((offset >> log2_xsize) << log2_ysize));  in xnn_compute_univector_contiguous()
1108   context->ukernel(size, x, y, &context->params);  in xnn_compute_univector_contiguous()
1112     const struct u8_softmax_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_u8_softmax()
1115   const uint8_t* x = (const uint8_t*) ((uintptr_t) context->x + context->x_stride * batch_index);  in xnn_compute_u8_softmax()
1116   uint8_t* y = (uint8_t*) ((uintptr_t) context->y + context->y_stride * batch_index);  in xnn_compute_u8_softmax()
1117   const size_t n = context->n;  in xnn_compute_u8_softmax()
1120   context->rmax_ukernel(n, x, &x_max);  in xnn_compute_u8_softmax()
1122   const uint32_t* t = (const uint32_t*) context->t + adjustment;  in xnn_compute_u8_softmax()
1123   context->lut_norm_ukernel(n, x, t, y);  in xnn_compute_u8_softmax()
1127     const struct floating_point_softmax_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_floating_point_softmax()
1130   const void* x = (const void*) ((uintptr_t) context->x + context->x_stride * batch_index);  in xnn_compute_floating_point_softmax()
1131   void* y = (void*) ((uintptr_t) context->y + context->y_stride * batch_index);  in xnn_compute_floating_point_softmax()
1132   const size_t n = context->n;  in xnn_compute_floating_point_softmax()
1139   context->rmax_ukernel(n, x, &x_max);  in xnn_compute_floating_point_softmax()
1146   context->raddstoreexpminusmax_ukernel(n, x, &x_max, y, &y_sum, &context->expminus_params);  in xnn_compute_floating_point_softmax()
1153   context->compute_reciprocal(&y_sum, &y_scale);  in xnn_compute_floating_point_softmax()
1154   context->vmulc_ukernel(n, y, &y_scale, y, &context->minmax_params);  in xnn_compute_floating_point_softmax()
1158     const struct vmulcaddc_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_vmulcaddc()
1162   const size_t x_stride = context->x_stride;  in xnn_compute_vmulcaddc()
1163   const size_t y_stride = context->y_stride;  in xnn_compute_vmulcaddc()
1165   const void* x = (const void*) ((uintptr_t) context->x + x_stride * batch_start);  in xnn_compute_vmulcaddc()
1166   void* y = (void*) ((uintptr_t) context->y + y_stride * batch_start);  in xnn_compute_vmulcaddc()
1168   context->ukernel(  in xnn_compute_vmulcaddc()
1170     context->n,  in xnn_compute_vmulcaddc()
1172     context->w,  in xnn_compute_vmulcaddc()
1174     &context->params);  in xnn_compute_vmulcaddc()
1179       const struct gemm_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_hmp_grouped_gemm()
1187     const size_t k_scaled  = context->k_scaled;  in xnn_compute_hmp_grouped_gemm()
1188     const size_t a_stride  = context->a_stride;  in xnn_compute_hmp_grouped_gemm()
1189     const size_t cm_stride = context->cm_stride;  in xnn_compute_hmp_grouped_gemm()
1191     context->ukernel.function[uarch_index](  in xnn_compute_hmp_grouped_gemm()
1195         (const void*) ((uintptr_t) context->a + mr_block_start * a_stride + group_index * k_scaled),  in xnn_compute_hmp_grouped_gemm()
1197 …(const void*) ((uintptr_t) context->packed_w + nr_block_start * context->w_stride + group_index * …  in xnn_compute_hmp_grouped_gemm()
1198 …(void*) ((uintptr_t) context->c + mr_block_start * cm_stride + (nr_block_start << context->log2_cs…  in xnn_compute_hmp_grouped_gemm()
1200         context->cn_stride,  in xnn_compute_hmp_grouped_gemm()
1201         &context->params);  in xnn_compute_hmp_grouped_gemm()
1205       const struct gemm_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_hmp_gemm()
1212     const size_t a_stride  = context->a_stride;  in xnn_compute_hmp_gemm()
1213     const size_t cm_stride = context->cm_stride;  in xnn_compute_hmp_gemm()
1215     context->ukernel.function[uarch_index](  in xnn_compute_hmp_gemm()
1218         context->k_scaled,  in xnn_compute_hmp_gemm()
1219         (const void*) ((uintptr_t) context->a + mr_block_start * a_stride),  in xnn_compute_hmp_gemm()
1221         (const void*) ((uintptr_t) context->packed_w + nr_block_start * context->w_stride),  in xnn_compute_hmp_gemm()
1222 …(void*) ((uintptr_t) context->c + mr_block_start * cm_stride + (nr_block_start << context->log2_cs…  in xnn_compute_hmp_gemm()
1224         context->cn_stride,  in xnn_compute_hmp_gemm()
1225         context->fused_params);  in xnn_compute_hmp_gemm()
1229       const struct igemm_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_hmp_grouped_batch_igemm()
1238     const size_t ks        = context->ks;  in xnn_compute_hmp_grouped_batch_igemm()
1239     const size_t cm_stride = context->cm_stride;  in xnn_compute_hmp_grouped_batch_igemm()
1241     context->ukernel.function[uarch_index](  in xnn_compute_hmp_grouped_batch_igemm()
1244         context->kc,  in xnn_compute_hmp_grouped_batch_igemm()
1245         context->ks_scaled,  in xnn_compute_hmp_grouped_batch_igemm()
1246         (const void**) ((uintptr_t) context->indirect_a + mr_block_start * ks * sizeof(void*)),  in xnn_compute_hmp_grouped_batch_igemm()
1247 …(const void*) ((uintptr_t) context->packed_w + nr_block_start * context->w_stride + group_index * …  in xnn_compute_hmp_grouped_batch_igemm()
1248 …ptr_t) context->c + group_index * context->gc_stride + batch_index * context->bc_stride + mr_block…  in xnn_compute_hmp_grouped_batch_igemm()
1250         context->cn_stride,  in xnn_compute_hmp_grouped_batch_igemm()
1251         context->a_offset + group_index * context->ga_stride + batch_index * context->ba_stride,  in xnn_compute_hmp_grouped_batch_igemm()
1252         context->zero,  in xnn_compute_hmp_grouped_batch_igemm()
1253         &context->params);  in xnn_compute_hmp_grouped_batch_igemm()
1257       const struct igemm_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_hmp_grouped_igemm()
1265     const size_t ks        = context->ks;  in xnn_compute_hmp_grouped_igemm()
1266     const size_t cm_stride = context->cm_stride;  in xnn_compute_hmp_grouped_igemm()
1268     context->ukernel.function[uarch_index](  in xnn_compute_hmp_grouped_igemm()
1271         context->kc,  in xnn_compute_hmp_grouped_igemm()
1272         context->ks_scaled,  in xnn_compute_hmp_grouped_igemm()
1273         (const void**) ((uintptr_t) context->indirect_a + mr_block_start * ks * sizeof(void*)),  in xnn_compute_hmp_grouped_igemm()
1274 …(const void*) ((uintptr_t) context->packed_w + nr_block_start * context->w_stride + group_index * …  in xnn_compute_hmp_grouped_igemm()
1275 …(void*) ((uintptr_t) context->c + group_index * context->gc_stride + mr_block_start * cm_stride + …  in xnn_compute_hmp_grouped_igemm()
1277         context->cn_stride,  in xnn_compute_hmp_grouped_igemm()
1278         context->a_offset + group_index * context->ga_stride,  in xnn_compute_hmp_grouped_igemm()
1279         context->zero,  in xnn_compute_hmp_grouped_igemm()
1280         &context->params);  in xnn_compute_hmp_grouped_igemm()
1284       const struct igemm_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_batch_hmp_igemm()
1292     const size_t ks        = context->ks;  in xnn_compute_batch_hmp_igemm()
1293     const size_t cm_stride = context->cm_stride;  in xnn_compute_batch_hmp_igemm()
1295     context->ukernel.function[uarch_index](  in xnn_compute_batch_hmp_igemm()
1298         context->kc,  in xnn_compute_batch_hmp_igemm()
1299         context->ks_scaled,  in xnn_compute_batch_hmp_igemm()
1300         (const void**) ((uintptr_t) context->indirect_a + mr_block_start * ks * sizeof(void*)),  in xnn_compute_batch_hmp_igemm()
1301         (const void*) ((uintptr_t) context->packed_w + nr_block_start * context->w_stride),  in xnn_compute_batch_hmp_igemm()
1302 …(void*) ((uintptr_t) context->c + batch_index * context->bc_stride + mr_block_start * cm_stride + …  in xnn_compute_batch_hmp_igemm()
1304         context->cn_stride,  in xnn_compute_batch_hmp_igemm()
1305         context->a_offset + batch_index * context->ba_stride,  in xnn_compute_batch_hmp_igemm()
1306         context->zero,  in xnn_compute_batch_hmp_igemm()
1307         &context->params);  in xnn_compute_batch_hmp_igemm()
1311       const struct igemm_context context[restrict XNN_MIN_ELEMENTS(1)],  in xnn_compute_hmp_igemm()
1318     const size_t ks        = context->ks;  in xnn_compute_hmp_igemm()
1319     const size_t cm_stride = context->cm_stride;  in xnn_compute_hmp_igemm()
1321     context->ukernel.function[uarch_index](  in xnn_compute_hmp_igemm()
1324         context->kc,  in xnn_compute_hmp_igemm()
1325         context->ks_scaled,  in xnn_compute_hmp_igemm()
1326         (const void**) ((uintptr_t) context->indirect_a + mr_block_start * ks * sizeof(void*)),  in xnn_compute_hmp_igemm()
1327         (const void*) ((uintptr_t) context->packed_w + nr_block_start * context->w_stride),  in xnn_compute_hmp_igemm()
1328 …(void*) ((uintptr_t) context->c + mr_block_start * cm_stride + (nr_block_start << context->log2_cs…  in xnn_compute_hmp_igemm()
1330         context->cn_stride,  in xnn_compute_hmp_igemm()
1331         context->a_offset,  in xnn_compute_hmp_igemm()
1332         context->zero,  in xnn_compute_hmp_igemm()
1333         &context->params);  in xnn_compute_hmp_igemm()
1365           &op->context,  in xnn_run_operator()
1375           &op->context,  in xnn_run_operator()
1386           &op->context,  in xnn_run_operator()
1397           &op->context,  in xnn_run_operator()
1410           &op->context,  in xnn_run_operator()
1422           &op->context,  in xnn_run_operator()
1435           &op->context,  in xnn_run_operator()
1448           &op->context,  in xnn_run_operator()
1462           &op->context,  in xnn_run_operator()
1476           &op->context,  in xnn_run_operator()
1491           &op->context,  in xnn_run_operator()
1508           &op->context,  in xnn_run_operator()
1522           &op->context,  in xnn_run_operator()
1537           &op->context,  in xnn_run_operator()
1553           &op->context,  in xnn_run_operator()