230   int x_out;  in RescalerExportRowExpandSSE2()  local
241     for (x_out = 0; x_out + 8 <= x_out_max; x_out += 8) {  in RescalerExportRowExpandSSE2()
243       LoadDispatchAndMult(frow + x_out, NULL, &A0, &A1, &A2, &A3);  in RescalerExportRowExpandSSE2()
244       ProcessRow(&A0, &A1, &A2, &A3, &mult, dst + x_out);  in RescalerExportRowExpandSSE2()
246     for (; x_out < x_out_max; ++x_out) {  in RescalerExportRowExpandSSE2()
247       const uint32_t J = frow[x_out];  in RescalerExportRowExpandSSE2()
250       dst[x_out] = v;  in RescalerExportRowExpandSSE2()
258     for (x_out = 0; x_out + 8 <= x_out_max; x_out += 8) {  in RescalerExportRowExpandSSE2()
260       LoadDispatchAndMult(frow + x_out, &mA, &A0, &A1, &A2, &A3);  in RescalerExportRowExpandSSE2()
261       LoadDispatchAndMult(irow + x_out, &mB, &B0, &B1, &B2, &B3);  in RescalerExportRowExpandSSE2()
275         ProcessRow(&E0, &E1, &E2, &E3, &mult, dst + x_out);  in RescalerExportRowExpandSSE2()
278     for (; x_out < x_out_max; ++x_out) {  in RescalerExportRowExpandSSE2()
279       const uint64_t I = (uint64_t)A * frow[x_out]  in RescalerExportRowExpandSSE2()
280                        + (uint64_t)B * irow[x_out];  in RescalerExportRowExpandSSE2()
284       dst[x_out] = v;  in RescalerExportRowExpandSSE2()
290   int x_out;  in RescalerExportRowShrinkSSE2()  local
304     for (x_out = 0; x_out + 8 <= x_out_max; x_out += 8) {  in RescalerExportRowShrinkSSE2()
306       LoadDispatchAndMult(irow + x_out, NULL, &A0, &A1, &A2, &A3);  in RescalerExportRowShrinkSSE2()
307       LoadDispatchAndMult(frow + x_out, &mult_y, &B0, &B1, &B2, &B3);  in RescalerExportRowShrinkSSE2()
325         _mm_storeu_si128((__m128i*)(irow + x_out + 0), G0);  in RescalerExportRowShrinkSSE2()
326         _mm_storeu_si128((__m128i*)(irow + x_out + 4), G1);  in RescalerExportRowShrinkSSE2()
327         ProcessRow(&E0, &E1, &E2, &E3, &mult_xy, dst + x_out);  in RescalerExportRowShrinkSSE2()
330     for (; x_out < x_out_max; ++x_out) {  in RescalerExportRowShrinkSSE2()
331       const uint32_t frac = (int)MULT_FIX(frow[x_out], yscale);  in RescalerExportRowShrinkSSE2()
332       const int v = (int)MULT_FIX(irow[x_out] - frac, wrk->fxy_scale);  in RescalerExportRowShrinkSSE2()
334       dst[x_out] = v;  in RescalerExportRowShrinkSSE2()
335       irow[x_out] = frac;   // new fractional start  in RescalerExportRowShrinkSSE2()
341     for (x_out = 0; x_out + 8 <= x_out_max; x_out += 8) {  in RescalerExportRowShrinkSSE2()
343       LoadDispatchAndMult(irow + x_out, NULL, &A0, &A1, &A2, &A3);  in RescalerExportRowShrinkSSE2()
344       _mm_storeu_si128((__m128i*)(irow + x_out + 0), zero);  in RescalerExportRowShrinkSSE2()
345       _mm_storeu_si128((__m128i*)(irow + x_out + 4), zero);  in RescalerExportRowShrinkSSE2()
346       ProcessRow(&A0, &A1, &A2, &A3, &mult, dst + x_out);  in RescalerExportRowShrinkSSE2()
348     for (; x_out < x_out_max; ++x_out) {  in RescalerExportRowShrinkSSE2()
349       const int v = (int)MULT_FIX(irow[x_out], scale);  in RescalerExportRowShrinkSSE2()
351       dst[x_out] = v;  in RescalerExportRowShrinkSSE2()
352       irow[x_out] = 0;  in RescalerExportRowShrinkSSE2()