64   int x_out;  in RescalerExportRowExpand_NEON()  local
77     for (x_out = 0; x_out < max_span; x_out += 8) {  in RescalerExportRowExpand_NEON()
78       LOAD_32x4(frow + x_out + 0, A0);  in RescalerExportRowExpand_NEON()
79       LOAD_32x4(frow + x_out + 4, A1);  in RescalerExportRowExpand_NEON()
85       vst1_u8(dst + x_out, D);  in RescalerExportRowExpand_NEON()
87     for (; x_out < x_out_max; ++x_out) {  in RescalerExportRowExpand_NEON()
88       const uint32_t J = frow[x_out];  in RescalerExportRowExpand_NEON()
90       dst[x_out] = (v > 255) ? 255u : (uint8_t)v;  in RescalerExportRowExpand_NEON()
95     for (x_out = 0; x_out < max_span; x_out += 8) {  in RescalerExportRowExpand_NEON()
97           Interpolate_NEON(frow + x_out + 0, irow + x_out + 0, A, B);  in RescalerExportRowExpand_NEON()
99           Interpolate_NEON(frow + x_out + 4, irow + x_out + 4, A, B);  in RescalerExportRowExpand_NEON()
105       vst1_u8(dst + x_out, F);  in RescalerExportRowExpand_NEON()
107     for (; x_out < x_out_max; ++x_out) {  in RescalerExportRowExpand_NEON()
108       const uint64_t I = (uint64_t)A * frow[x_out]  in RescalerExportRowExpand_NEON()
109                        + (uint64_t)B * irow[x_out];  in RescalerExportRowExpand_NEON()
112       dst[x_out] = (v > 255) ? 255u : (uint8_t)v;  in RescalerExportRowExpand_NEON()
118   int x_out;  in RescalerExportRowShrink_NEON()  local
133     for (x_out = 0; x_out < max_span; x_out += 8) {  in RescalerExportRowShrink_NEON()
134       LOAD_32x8(frow + x_out, in0, in1);  in RescalerExportRowShrink_NEON()
135       LOAD_32x8(irow + x_out, in2, in3);  in RescalerExportRowShrink_NEON()
145       vst1_u8(dst + x_out, E);  in RescalerExportRowShrink_NEON()
146       STORE_32x8(A0, A1, irow + x_out);  in RescalerExportRowShrink_NEON()
148     for (; x_out < x_out_max; ++x_out) {  in RescalerExportRowShrink_NEON()
149       const uint32_t frac = (uint32_t)MULT_FIX_FLOOR_C(frow[x_out], yscale);  in RescalerExportRowShrink_NEON()
150       const int v = (int)MULT_FIX_C(irow[x_out] - frac, fxy_scale);  in RescalerExportRowShrink_NEON()
151       dst[x_out] = (v > 255) ? 255u : (uint8_t)v;  in RescalerExportRowShrink_NEON()
152       irow[x_out] = frac;   // new fractional start  in RescalerExportRowShrink_NEON()
155     for (x_out = 0; x_out < max_span; x_out += 8) {  in RescalerExportRowShrink_NEON()
156       LOAD_32x8(irow + x_out, in0, in1);  in RescalerExportRowShrink_NEON()
162       vst1_u8(dst + x_out, C);  in RescalerExportRowShrink_NEON()
163       STORE_32x8(zero, zero, irow + x_out);  in RescalerExportRowShrink_NEON()
165     for (; x_out < x_out_max; ++x_out) {  in RescalerExportRowShrink_NEON()
166       const int v = (int)MULT_FIX_C(irow[x_out], fxy_scale);  in RescalerExportRowShrink_NEON()
167       dst[x_out] = (v > 255) ? 255u : (uint8_t)v;  in RescalerExportRowShrink_NEON()
168       irow[x_out] = 0;  in RescalerExportRowShrink_NEON()