66 __m128i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10, xmm11;
67 __m128i *p_target, *p_src0, *p_cntl2, *p_cntl3, *p_scalars;
69 p_target = (__m128i*)target;
70 p_src0 = (__m128i*)src0;
71 p_cntl2 = (__m128i*)cntl2;
72 p_cntl3 = (__m128i*)cntl3;
73 p_scalars = (__m128i*)scalars;
75 xmm0 = _mm_load_si128(p_scalars);
77 xmm1 = _mm_shufflelo_epi16(xmm0, 0);
78 xmm2 = _mm_shufflelo_epi16(xmm0, 0x55);
79 xmm3 = _mm_shufflelo_epi16(xmm0, 0xaa);
80 xmm4 = _mm_shufflelo_epi16(xmm0, 0xff);
82 xmm1 = _mm_shuffle_epi32(xmm1, 0x00);
83 xmm2 = _mm_shuffle_epi32(xmm2, 0x00);
84 xmm3 = _mm_shuffle_epi32(xmm3, 0x00);
85 xmm4 = _mm_shuffle_epi32(xmm4, 0x00);
87 xmm0 = _mm_load_si128((__m128i*)permuters[0]);
88 xmm6 = _mm_load_si128((__m128i*)permuters[1]);
89 xmm8 = _mm_load_si128((__m128i*)permuters[2]);
90 xmm10 = _mm_load_si128((__m128i*)permuters[3]);
92 xmm5 = _mm_load_si128(p_src0);
93 xmm0 = _mm_shuffle_epi8(xmm5, xmm0);
94 xmm6 = _mm_shuffle_epi8(xmm5, xmm6);
95 xmm8 = _mm_shuffle_epi8(xmm5, xmm8);
96 xmm10 = _mm_shuffle_epi8(xmm5, xmm10);
98 xmm5 = _mm_add_epi16(xmm1, xmm2);
100 xmm6 = _mm_add_epi16(xmm2, xmm6);
101 xmm8 = _mm_add_epi16(xmm1, xmm8);
103 xmm7 = _mm_load_si128(p_cntl2);
104 xmm9 = _mm_load_si128(p_cntl3);
106 xmm0 = _mm_add_epi16(xmm5, xmm0);
108 xmm7 = _mm_and_si128(xmm7, xmm3);
109 xmm9 = _mm_and_si128(xmm9, xmm4);
111 xmm5 = _mm_load_si128(&p_cntl2[1]);
112 xmm11 = _mm_load_si128(&p_cntl3[1]);
114 xmm7 = _mm_add_epi16(xmm7, xmm9);
116 xmm5 = _mm_and_si128(xmm5, xmm3);
117 xmm11 = _mm_and_si128(xmm11, xmm4);
119 xmm0 = _mm_add_epi16(xmm0, xmm7);
122 xmm7 = _mm_load_si128(&p_cntl2[2]);
123 xmm9 = _mm_load_si128(&p_cntl3[2]);
125 xmm5 = _mm_add_epi16(xmm5, xmm11);
127 xmm7 = _mm_and_si128(xmm7, xmm3);
128 xmm9 = _mm_and_si128(xmm9, xmm4);
130 xmm6 = _mm_add_epi16(xmm6, xmm5);
133 xmm5 = _mm_load_si128(&p_cntl2[3]);
134 xmm11 = _mm_load_si128(&p_cntl3[3]);
136 xmm7 = _mm_add_epi16(xmm7, xmm9);
138 xmm5 = _mm_and_si128(xmm5, xmm3);
139 xmm11 = _mm_and_si128(xmm11, xmm4);
141 xmm8 = _mm_add_epi16(xmm8, xmm7);
143 xmm5 = _mm_add_epi16(xmm5, xmm11);
145 _mm_store_si128(p_target, xmm0);
146 _mm_store_si128(&p_target[1], xmm6);
148 xmm10 = _mm_add_epi16(xmm5, xmm10);
150 _mm_store_si128(&p_target[2], xmm8);
152 _mm_store_si128(&p_target[3], xmm10);
170 for (;
i < bound; ++
i) {
171 target[
i * 8] = src0[((char)permuters[
i][0]) / 2] + ((
i + 1) % 2 * scalars[0]) +
172 (((
i >> 1) ^ 1) * scalars[1]) + (cntl2[
i * 8] & scalars[2]) +
173 (cntl3[
i * 8] & scalars[3]);
174 target[
i * 8 + 1] = src0[((char)permuters[
i][1 * 2]) / 2] +
175 ((
i + 1) % 2 * scalars[0]) + (((
i >> 1) ^ 1) * scalars[1]) +
176 (cntl2[
i * 8 + 1] & scalars[2]) +
177 (cntl3[
i * 8 + 1] & scalars[3]);
178 target[
i * 8 + 2] = src0[((char)permuters[
i][2 * 2]) / 2] +
179 ((
i + 1) % 2 * scalars[0]) + (((
i >> 1) ^ 1) * scalars[1]) +
180 (cntl2[
i * 8 + 2] & scalars[2]) +
181 (cntl3[
i * 8 + 2] & scalars[3]);
182 target[
i * 8 + 3] = src0[((char)permuters[
i][3 * 2]) / 2] +
183 ((
i + 1) % 2 * scalars[0]) + (((
i >> 1) ^ 1) * scalars[1]) +
184 (cntl2[
i * 8 + 3] & scalars[2]) +
185 (cntl3[
i * 8 + 3] & scalars[3]);
186 target[
i * 8 + 4] = src0[((char)permuters[
i][4 * 2]) / 2] +
187 ((
i + 1) % 2 * scalars[0]) + (((
i >> 1) ^ 1) * scalars[1]) +
188 (cntl2[
i * 8 + 4] & scalars[2]) +
189 (cntl3[
i * 8 + 4] & scalars[3]);
190 target[
i * 8 + 5] = src0[((char)permuters[
i][5 * 2]) / 2] +
191 ((
i + 1) % 2 * scalars[0]) + (((
i >> 1) ^ 1) * scalars[1]) +
192 (cntl2[
i * 8 + 5] & scalars[2]) +
193 (cntl3[
i * 8 + 5] & scalars[3]);
194 target[
i * 8 + 6] = src0[((char)permuters[
i][6 * 2]) / 2] +
195 ((
i + 1) % 2 * scalars[0]) + (((
i >> 1) ^ 1) * scalars[1]) +
196 (cntl2[
i * 8 + 6] & scalars[2]) +
197 (cntl3[
i * 8 + 6] & scalars[3]);
198 target[
i * 8 + 7] = src0[((char)permuters[
i][7 * 2]) / 2] +
199 ((
i + 1) % 2 * scalars[0]) + (((
i >> 1) ^ 1) * scalars[1]) +
200 (cntl2[
i * 8 + 7] & scalars[2]) +
201 (cntl3[
i * 8 + 7] & scalars[3]);