229   int x_out;  in RescalerExportRowExpandSSE2()  local
240     for (x_out = 0; x_out + 8 <= x_out_max; x_out += 8) {  in RescalerExportRowExpandSSE2()
242       LoadDispatchAndMult(frow + x_out, NULL, &A0, &A1, &A2, &A3);  in RescalerExportRowExpandSSE2()
243       ProcessRow(&A0, &A1, &A2, &A3, &mult, dst + x_out);  in RescalerExportRowExpandSSE2()
245     for (; x_out < x_out_max; ++x_out) {  in RescalerExportRowExpandSSE2()
246       const uint32_t J = frow[x_out];  in RescalerExportRowExpandSSE2()
249       dst[x_out] = v;  in RescalerExportRowExpandSSE2()
257     for (x_out = 0; x_out + 8 <= x_out_max; x_out += 8) {  in RescalerExportRowExpandSSE2()
259       LoadDispatchAndMult(frow + x_out, &mA, &A0, &A1, &A2, &A3);  in RescalerExportRowExpandSSE2()
260       LoadDispatchAndMult(irow + x_out, &mB, &B0, &B1, &B2, &B3);  in RescalerExportRowExpandSSE2()
274         ProcessRow(&E0, &E1, &E2, &E3, &mult, dst + x_out);  in RescalerExportRowExpandSSE2()
277     for (; x_out < x_out_max; ++x_out) {  in RescalerExportRowExpandSSE2()
278       const uint64_t I = (uint64_t)A * frow[x_out]  in RescalerExportRowExpandSSE2()
279                        + (uint64_t)B * irow[x_out];  in RescalerExportRowExpandSSE2()
283       dst[x_out] = v;  in RescalerExportRowExpandSSE2()
289   int x_out;  in RescalerExportRowShrinkSSE2()  local
303     for (x_out = 0; x_out + 8 <= x_out_max; x_out += 8) {  in RescalerExportRowShrinkSSE2()
305       LoadDispatchAndMult(irow + x_out, NULL, &A0, &A1, &A2, &A3);  in RescalerExportRowShrinkSSE2()
306       LoadDispatchAndMult(frow + x_out, &mult_y, &B0, &B1, &B2, &B3);  in RescalerExportRowShrinkSSE2()
324         _mm_storeu_si128((__m128i*)(irow + x_out + 0), G0);  in RescalerExportRowShrinkSSE2()
325         _mm_storeu_si128((__m128i*)(irow + x_out + 4), G1);  in RescalerExportRowShrinkSSE2()
326         ProcessRow(&E0, &E1, &E2, &E3, &mult_xy, dst + x_out);  in RescalerExportRowShrinkSSE2()
329     for (; x_out < x_out_max; ++x_out) {  in RescalerExportRowShrinkSSE2()
330       const uint32_t frac = (int)MULT_FIX(frow[x_out], yscale);  in RescalerExportRowShrinkSSE2()
331       const int v = (int)MULT_FIX(irow[x_out] - frac, wrk->fxy_scale);  in RescalerExportRowShrinkSSE2()
333       dst[x_out] = v;  in RescalerExportRowShrinkSSE2()
334       irow[x_out] = frac;   // new fractional start  in RescalerExportRowShrinkSSE2()
340     for (x_out = 0; x_out + 8 <= x_out_max; x_out += 8) {  in RescalerExportRowShrinkSSE2()
342       LoadDispatchAndMult(irow + x_out, NULL, &A0, &A1, &A2, &A3);  in RescalerExportRowShrinkSSE2()
343       _mm_storeu_si128((__m128i*)(irow + x_out + 0), zero);  in RescalerExportRowShrinkSSE2()
344       _mm_storeu_si128((__m128i*)(irow + x_out + 4), zero);  in RescalerExportRowShrinkSSE2()
345       ProcessRow(&A0, &A1, &A2, &A3, &mult, dst + x_out);  in RescalerExportRowShrinkSSE2()
347     for (; x_out < x_out_max; ++x_out) {  in RescalerExportRowShrinkSSE2()
348       const int v = (int)MULT_FIX(irow[x_out], scale);  in RescalerExportRowShrinkSSE2()
350       dst[x_out] = v;  in RescalerExportRowShrinkSSE2()
351       irow[x_out] = 0;  in RescalerExportRowShrinkSSE2()