49           const float32x2_t vb = vld1_f32(w); w += 2;  in xnn_f32_spmm_ukernel_8x2__neonfma()  local
51           vacc0123c0 = vfmaq_lane_f32(vacc0123c0, va0123, vb, 0);  in xnn_f32_spmm_ukernel_8x2__neonfma()
52           vacc4567c0 = vfmaq_lane_f32(vacc4567c0, va4567, vb, 0);  in xnn_f32_spmm_ukernel_8x2__neonfma()
53           vacc0123c1 = vfmaq_lane_f32(vacc0123c1, va0123, vb, 1);  in xnn_f32_spmm_ukernel_8x2__neonfma()
54           vacc4567c1 = vfmaq_lane_f32(vacc4567c1, va4567, vb, 1);  in xnn_f32_spmm_ukernel_8x2__neonfma()
87             const float32x4_t vb = vld1q_dup_f32(w); w += 1;  in xnn_f32_spmm_ukernel_8x2__neonfma()  local
88             vacc0123 = vfmaq_f32(vacc0123, va0123, vb);  in xnn_f32_spmm_ukernel_8x2__neonfma()
89             vacc4567 = vfmaq_f32(vacc4567, va4567, vb);  in xnn_f32_spmm_ukernel_8x2__neonfma()
124             const float32x2_t vb = vld1_f32(w); w += 2;  in xnn_f32_spmm_ukernel_8x2__neonfma()  local
126             vacc0123c0 = vfmaq_lane_f32(vacc0123c0, va0123, vb, 0);  in xnn_f32_spmm_ukernel_8x2__neonfma()
127             vacc0123c1 = vfmaq_lane_f32(vacc0123c1, va0123, vb, 1);  in xnn_f32_spmm_ukernel_8x2__neonfma()
152               const float32x4_t vb = vld1q_dup_f32(w); w += 1;  in xnn_f32_spmm_ukernel_8x2__neonfma()  local
153               vacc0123 = vfmaq_f32(vacc0123, va0123, vb);  in xnn_f32_spmm_ukernel_8x2__neonfma()
183             const float32x2_t vb = vld1_f32(w); w += 2;  in xnn_f32_spmm_ukernel_8x2__neonfma()  local
185             vacc01c0 = vfma_lane_f32(vacc01c0, va01, vb, 0);  in xnn_f32_spmm_ukernel_8x2__neonfma()
186             vacc01c1 = vfma_lane_f32(vacc01c1, va01, vb, 1);  in xnn_f32_spmm_ukernel_8x2__neonfma()
211               const float32x2_t vb = vld1_dup_f32(w); w += 1;  in xnn_f32_spmm_ukernel_8x2__neonfma()  local
212               vacc01 = vfma_f32(vacc01, va01, vb);  in xnn_f32_spmm_ukernel_8x2__neonfma()
241             const float32x2_t vb = vld1_f32(w); w += 2;  in xnn_f32_spmm_ukernel_8x2__neonfma()  local
243             vacc0c0 = vfma_lane_f32(vacc0c0, va0, vb, 0);  in xnn_f32_spmm_ukernel_8x2__neonfma()
244             vacc0c1 = vfma_lane_f32(vacc0c1, va0, vb, 1);  in xnn_f32_spmm_ukernel_8x2__neonfma()
269               const float32x2_t vb = vld1_dup_f32(w); w += 1;  in xnn_f32_spmm_ukernel_8x2__neonfma()  local
270               vacc0 = vfma_f32(vacc0, va0, vb);  in xnn_f32_spmm_ukernel_8x2__neonfma()