229   int x_out;  in RescalerExportRowExpand_SSE2()  local
240     for (x_out = 0; x_out + 8 <= x_out_max; x_out += 8) {  in RescalerExportRowExpand_SSE2()
242       LoadDispatchAndMult_SSE2(frow + x_out, NULL, &A0, &A1, &A2, &A3);  in RescalerExportRowExpand_SSE2()
243       ProcessRow_SSE2(&A0, &A1, &A2, &A3, &mult, dst + x_out);  in RescalerExportRowExpand_SSE2()
245     for (; x_out < x_out_max; ++x_out) {  in RescalerExportRowExpand_SSE2()
246       const uint32_t J = frow[x_out];  in RescalerExportRowExpand_SSE2()
248       dst[x_out] = (v > 255) ? 255u : (uint8_t)v;  in RescalerExportRowExpand_SSE2()
256     for (x_out = 0; x_out + 8 <= x_out_max; x_out += 8) {  in RescalerExportRowExpand_SSE2()
258       LoadDispatchAndMult_SSE2(frow + x_out, &mA, &A0, &A1, &A2, &A3);  in RescalerExportRowExpand_SSE2()
259       LoadDispatchAndMult_SSE2(irow + x_out, &mB, &B0, &B1, &B2, &B3);  in RescalerExportRowExpand_SSE2()
273         ProcessRow_SSE2(&E0, &E1, &E2, &E3, &mult, dst + x_out);  in RescalerExportRowExpand_SSE2()
276     for (; x_out < x_out_max; ++x_out) {  in RescalerExportRowExpand_SSE2()
277       const uint64_t I = (uint64_t)A * frow[x_out]  in RescalerExportRowExpand_SSE2()
278                        + (uint64_t)B * irow[x_out];  in RescalerExportRowExpand_SSE2()
281       dst[x_out] = (v > 255) ? 255u : (uint8_t)v;  in RescalerExportRowExpand_SSE2()
287   int x_out;  in RescalerExportRowShrink_SSE2()  local
300     for (x_out = 0; x_out + 8 <= x_out_max; x_out += 8) {  in RescalerExportRowShrink_SSE2()
302       LoadDispatchAndMult_SSE2(irow + x_out, NULL, &A0, &A1, &A2, &A3);  in RescalerExportRowShrink_SSE2()
303       LoadDispatchAndMult_SSE2(frow + x_out, &mult_y, &B0, &B1, &B2, &B3);  in RescalerExportRowShrink_SSE2()
317         _mm_storeu_si128((__m128i*)(irow + x_out + 0), G0);  in RescalerExportRowShrink_SSE2()
318         _mm_storeu_si128((__m128i*)(irow + x_out + 4), G1);  in RescalerExportRowShrink_SSE2()
319         ProcessRow_SSE2(&E0, &E1, &E2, &E3, &mult_xy, dst + x_out);  in RescalerExportRowShrink_SSE2()
322     for (; x_out < x_out_max; ++x_out) {  in RescalerExportRowShrink_SSE2()
323       const uint32_t frac = (int)MULT_FIX_FLOOR(frow[x_out], yscale);  in RescalerExportRowShrink_SSE2()
324       const int v = (int)MULT_FIX(irow[x_out] - frac, wrk->fxy_scale);  in RescalerExportRowShrink_SSE2()
325       dst[x_out] = (v > 255) ? 255u : (uint8_t)v;  in RescalerExportRowShrink_SSE2()
326       irow[x_out] = frac;   // new fractional start  in RescalerExportRowShrink_SSE2()
332     for (x_out = 0; x_out + 8 <= x_out_max; x_out += 8) {  in RescalerExportRowShrink_SSE2()
334       LoadDispatchAndMult_SSE2(irow + x_out, NULL, &A0, &A1, &A2, &A3);  in RescalerExportRowShrink_SSE2()
335       _mm_storeu_si128((__m128i*)(irow + x_out + 0), zero);  in RescalerExportRowShrink_SSE2()
336       _mm_storeu_si128((__m128i*)(irow + x_out + 4), zero);  in RescalerExportRowShrink_SSE2()
337       ProcessRow_SSE2(&A0, &A1, &A2, &A3, &mult, dst + x_out);  in RescalerExportRowShrink_SSE2()
339     for (; x_out < x_out_max; ++x_out) {  in RescalerExportRowShrink_SSE2()
340       const int v = (int)MULT_FIX(irow[x_out], scale);  in RescalerExportRowShrink_SSE2()
341       dst[x_out] = (v > 255) ? 255u : (uint8_t)v;  in RescalerExportRowShrink_SSE2()
342       irow[x_out] = 0;  in RescalerExportRowShrink_SSE2()