Lines Matching refs:__m128i
28 const __m128i zero = _mm_setzero_si128();
29 const __m128i rgb_mask = _mm_set1_epi32(0xffffff00u); // to preserve RGB
30 const __m128i all_0xff = _mm_set_epi32(0, 0, ~0u, ~0u);
31 __m128i all_alphas = all_0xff;
39 __m128i* out = (__m128i*)dst;
42 const __m128i a0 = _mm_loadl_epi64((const __m128i*)&alpha[i]);
43 const __m128i a1 = _mm_unpacklo_epi8(a0, zero);
44 const __m128i a2_lo = _mm_unpacklo_epi16(a1, zero);
45 const __m128i a2_hi = _mm_unpackhi_epi16(a1, zero);
47 const __m128i b0_lo = _mm_loadu_si128(out + 0);
48 const __m128i b0_hi = _mm_loadu_si128(out + 1);
50 const __m128i b1_lo = _mm_and_si128(b0_lo, rgb_mask);
51 const __m128i b1_hi = _mm_and_si128(b0_hi, rgb_mask);
53 const __m128i b2_lo = _mm_or_si128(b1_lo, a2_lo);
54 const __m128i b2_hi = _mm_or_si128(b1_hi, a2_hi);
79 const __m128i zero = _mm_setzero_si128();
83 const __m128i a0 = _mm_loadu_si128((const __m128i*)&alpha[i]);
84 const __m128i a1 = _mm_unpacklo_epi8(zero, a0); // note the 'zero' first!
85 const __m128i b1 = _mm_unpackhi_epi8(zero, a0);
86 const __m128i a2_lo = _mm_unpacklo_epi16(a1, zero);
87 const __m128i b2_lo = _mm_unpacklo_epi16(b1, zero);
88 const __m128i a2_hi = _mm_unpackhi_epi16(a1, zero);
89 const __m128i b2_hi = _mm_unpackhi_epi16(b1, zero);
90 _mm_storeu_si128((__m128i*)&dst[i + 0], a2_lo);
91 _mm_storeu_si128((__m128i*)&dst[i + 4], a2_hi);
92 _mm_storeu_si128((__m128i*)&dst[i + 8], b2_lo);
93 _mm_storeu_si128((__m128i*)&dst[i + 12], b2_hi);
108 const __m128i a_mask = _mm_set1_epi32(0xffu); // to preserve alpha
109 const __m128i all_0xff = _mm_set_epi32(0, 0, ~0u, ~0u);
110 __m128i all_alphas = all_0xff;
118 const __m128i* src = (const __m128i*)argb;
121 const __m128i a0 = _mm_loadu_si128(src + 0);
122 const __m128i a1 = _mm_loadu_si128(src + 1);
123 const __m128i b0 = _mm_and_si128(a0, a_mask);
124 const __m128i b1 = _mm_and_si128(a1, a_mask);
125 const __m128i c0 = _mm_packs_epi32(b0, b1);
126 const __m128i d0 = _mm_packus_epi16(c0, c0);
128 _mm_storel_epi64((__m128i*)&alpha[i], d0);
157 const __m128i argb0 = _mm_loadu_si128((const __m128i*)&(RGBX)); \
158 const __m128i argb1_lo = _mm_unpacklo_epi8(argb0, zero); \
159 const __m128i argb1_hi = _mm_unpackhi_epi8(argb0, zero); \
160 const __m128i alpha0_lo = _mm_or_si128(argb1_lo, kMask); \
161 const __m128i alpha0_hi = _mm_or_si128(argb1_hi, kMask); \
162 const __m128i alpha1_lo = _mm_shufflelo_epi16(alpha0_lo, SHUFFLE); \
163 const __m128i alpha1_hi = _mm_shufflelo_epi16(alpha0_hi, SHUFFLE); \
164 const __m128i alpha2_lo = _mm_shufflehi_epi16(alpha1_lo, SHUFFLE); \
165 const __m128i alpha2_hi = _mm_shufflehi_epi16(alpha1_hi, SHUFFLE); \
167 const __m128i A0_lo = _mm_mullo_epi16(alpha2_lo, argb1_lo); \
168 const __m128i A0_hi = _mm_mullo_epi16(alpha2_hi, argb1_hi); \
169 const __m128i A1_lo = _mm_mulhi_epu16(A0_lo, kMult); \
170 const __m128i A1_hi = _mm_mulhi_epu16(A0_hi, kMult); \
171 const __m128i A2_lo = _mm_srli_epi16(A1_lo, 7); \
172 const __m128i A2_hi = _mm_srli_epi16(A1_hi, 7); \
173 const __m128i A3 = _mm_packus_epi16(A2_lo, A2_hi); \
174 _mm_storeu_si128((__m128i*)&(RGBX), A3); \
179 const __m128i zero = _mm_setzero_si128();
180 const __m128i kMult = _mm_set1_epi16(0x8081u);
181 const __m128i kMask = _mm_set_epi16(0, 0xff, 0xff, 0, 0, 0xff, 0xff, 0);
217 const __m128i all_0xff = _mm_set1_epi8(0xff);
220 const __m128i v = _mm_loadu_si128((const __m128i*)(src + i));
221 const __m128i bits = _mm_cmpeq_epi8(v, all_0xff);
230 const __m128i alpha_mask = _mm_set1_epi32(0xff);
231 const __m128i all_0xff = _mm_set1_epi8(0xff);
238 const __m128i a0 = _mm_loadu_si128((const __m128i*)(src + i + 0));
239 const __m128i a1 = _mm_loadu_si128((const __m128i*)(src + i + 16));
240 const __m128i a2 = _mm_loadu_si128((const __m128i*)(src + i + 32));
241 const __m128i a3 = _mm_loadu_si128((const __m128i*)(src + i + 48));
242 const __m128i b0 = _mm_and_si128(a0, alpha_mask);
243 const __m128i b1 = _mm_and_si128(a1, alpha_mask);
244 const __m128i b2 = _mm_and_si128(a2, alpha_mask);
245 const __m128i b3 = _mm_and_si128(a3, alpha_mask);
246 const __m128i c0 = _mm_packs_epi32(b0, b1);
247 const __m128i c1 = _mm_packs_epi32(b2, b3);
248 const __m128i d = _mm_packus_epi16(c0, c1);
249 const __m128i bits = _mm_cmpeq_epi8(d, all_0xff);
254 const __m128i a0 = _mm_loadu_si128((const __m128i*)(src + i + 0));
255 const __m128i a1 = _mm_loadu_si128((const __m128i*)(src + i + 16));
256 const __m128i b0 = _mm_and_si128(a0, alpha_mask);
257 const __m128i b1 = _mm_and_si128(a1, alpha_mask);
258 const __m128i c = _mm_packs_epi32(b0, b1);
259 const __m128i d = _mm_packus_epi16(c, c);
260 const __m128i bits = _mm_cmpeq_epi8(d, all_0xff);
275 const __m128i zero = _mm_setzero_si128();
276 const __m128i k128 = _mm_set1_epi16(128);
277 const __m128i kMult = _mm_set1_epi16(0x0101);
278 const __m128i kMask = _mm_set_epi16(0, 0xff, 0, 0, 0, 0xff, 0, 0);
282 const __m128i A0 = _mm_loadl_epi64((const __m128i*)&ptr[x]);
283 const __m128i A1 = _mm_unpacklo_epi8(A0, zero);
284 const __m128i A2 = _mm_or_si128(A1, kMask);
285 const __m128i A3 = _mm_shufflelo_epi16(A2, _MM_SHUFFLE(2, 3, 3, 3));
286 const __m128i A4 = _mm_shufflehi_epi16(A3, _MM_SHUFFLE(2, 3, 3, 3));
288 const __m128i A5 = _mm_mullo_epi16(A4, A1);
289 const __m128i A6 = _mm_add_epi16(A5, k128);
290 const __m128i A7 = _mm_mulhi_epu16(A6, kMult);
291 const __m128i A10 = _mm_packus_epi16(A7, zero);
292 _mm_storel_epi64((__m128i*)&ptr[x], A10);
303 const __m128i zero = _mm_setzero_si128();
304 const __m128i k128 = _mm_set1_epi16(128);
305 const __m128i kMult = _mm_set1_epi16(0x0101);
307 const __m128i v0 = _mm_loadl_epi64((__m128i*)&ptr[x]);
308 const __m128i a0 = _mm_loadl_epi64((const __m128i*)&alpha[x]);
309 const __m128i v1 = _mm_unpacklo_epi8(v0, zero);
310 const __m128i a1 = _mm_unpacklo_epi8(a0, zero);
311 const __m128i v2 = _mm_mullo_epi16(v1, a1);
312 const __m128i v3 = _mm_add_epi16(v2, k128);
313 const __m128i v4 = _mm_mulhi_epu16(v3, kMult);
314 const __m128i v5 = _mm_packus_epi16(v4, zero);
315 _mm_storel_epi64((__m128i*)&ptr[x], v5);
