607 #define LOADU(x)    _mm256_loadu_si256((const VREG*)(x))  macro
617 #define LOADU(x)    _mm_loadu_si128((const VREG*)(x))  macro
678     s1n_0 = LOADU(in_even + 0);  in opj_idwt53_v_cas0_mcols_SSE2_OR_AVX2()
679     s1n_1 = LOADU(in_even + VREG_INT_COUNT);  in opj_idwt53_v_cas0_mcols_SSE2_OR_AVX2()
680     d1n_0 = LOADU(in_odd);  in opj_idwt53_v_cas0_mcols_SSE2_OR_AVX2()
681     d1n_1 = LOADU(in_odd + VREG_INT_COUNT);  in opj_idwt53_v_cas0_mcols_SSE2_OR_AVX2()
694         s1n_0 = LOADU(in_even + j * stride);  in opj_idwt53_v_cas0_mcols_SSE2_OR_AVX2()
695         s1n_1 = LOADU(in_even + j * stride + VREG_INT_COUNT);  in opj_idwt53_v_cas0_mcols_SSE2_OR_AVX2()
696         d1n_0 = LOADU(in_odd + j * stride);  in opj_idwt53_v_cas0_mcols_SSE2_OR_AVX2()
697         d1n_1 = LOADU(in_odd + j * stride + VREG_INT_COUNT);  in opj_idwt53_v_cas0_mcols_SSE2_OR_AVX2()
718         s1n_0 = LOADU(in_even + (OPJ_SIZE_T)((len - 1) / 2) * stride);  in opj_idwt53_v_cas0_mcols_SSE2_OR_AVX2()
726         s1n_1 = LOADU(in_even + (OPJ_SIZE_T)((len - 1) / 2) * stride + VREG_INT_COUNT);  in opj_idwt53_v_cas0_mcols_SSE2_OR_AVX2()
780     s1_0 = LOADU(in_even + stride);  in opj_idwt53_v_cas1_mcols_SSE2_OR_AVX2()
782     dc_0 = SUB(LOADU(in_odd + 0),  in opj_idwt53_v_cas1_mcols_SSE2_OR_AVX2()
783                SAR(ADD3(LOADU(in_even + 0), s1_0, two), 2));  in opj_idwt53_v_cas1_mcols_SSE2_OR_AVX2()
784     STORE(tmp + PARALLEL_COLS_53 * 0, ADD(LOADU(in_even + 0), dc_0));  in opj_idwt53_v_cas1_mcols_SSE2_OR_AVX2()
786     s1_1 = LOADU(in_even + stride + VREG_INT_COUNT);  in opj_idwt53_v_cas1_mcols_SSE2_OR_AVX2()
788     dc_1 = SUB(LOADU(in_odd + VREG_INT_COUNT),  in opj_idwt53_v_cas1_mcols_SSE2_OR_AVX2()
789                SAR(ADD3(LOADU(in_even + VREG_INT_COUNT), s1_1, two), 2));  in opj_idwt53_v_cas1_mcols_SSE2_OR_AVX2()
791           ADD(LOADU(in_even + VREG_INT_COUNT), dc_1));  in opj_idwt53_v_cas1_mcols_SSE2_OR_AVX2()
795         s2_0 = LOADU(in_even + (j + 1) * stride);  in opj_idwt53_v_cas1_mcols_SSE2_OR_AVX2()
796         s2_1 = LOADU(in_even + (j + 1) * stride + VREG_INT_COUNT);  in opj_idwt53_v_cas1_mcols_SSE2_OR_AVX2()
799         dn_0 = SUB(LOADU(in_odd + j * stride),  in opj_idwt53_v_cas1_mcols_SSE2_OR_AVX2()
801         dn_1 = SUB(LOADU(in_odd + j * stride + VREG_INT_COUNT),  in opj_idwt53_v_cas1_mcols_SSE2_OR_AVX2()
823         dn_0 = SUB(LOADU(in_odd + (OPJ_SIZE_T)(len / 2 - 1) * stride),  in opj_idwt53_v_cas1_mcols_SSE2_OR_AVX2()
825         dn_1 = SUB(LOADU(in_odd + (OPJ_SIZE_T)(len / 2 - 1) * stride + VREG_INT_COUNT),  in opj_idwt53_v_cas1_mcols_SSE2_OR_AVX2()
847 #undef LOADU