Lines Matching refs:__m128i
28 const __m128i zero = _mm_setzero_si128(); in DispatchAlpha()
29 const __m128i rgb_mask = _mm_set1_epi32(0xffffff00u); // to preserve RGB in DispatchAlpha()
30 const __m128i all_0xff = _mm_set_epi32(0, 0, ~0u, ~0u); in DispatchAlpha()
31 __m128i all_alphas = all_0xff; in DispatchAlpha()
39 __m128i* out = (__m128i*)dst; in DispatchAlpha()
42 const __m128i a0 = _mm_loadl_epi64((const __m128i*)&alpha[i]); in DispatchAlpha()
43 const __m128i a1 = _mm_unpacklo_epi8(a0, zero); in DispatchAlpha()
44 const __m128i a2_lo = _mm_unpacklo_epi16(a1, zero); in DispatchAlpha()
45 const __m128i a2_hi = _mm_unpackhi_epi16(a1, zero); in DispatchAlpha()
47 const __m128i b0_lo = _mm_loadu_si128(out + 0); in DispatchAlpha()
48 const __m128i b0_hi = _mm_loadu_si128(out + 1); in DispatchAlpha()
50 const __m128i b1_lo = _mm_and_si128(b0_lo, rgb_mask); in DispatchAlpha()
51 const __m128i b1_hi = _mm_and_si128(b0_hi, rgb_mask); in DispatchAlpha()
53 const __m128i b2_lo = _mm_or_si128(b1_lo, a2_lo); in DispatchAlpha()
54 const __m128i b2_hi = _mm_or_si128(b1_hi, a2_hi); in DispatchAlpha()
79 const __m128i zero = _mm_setzero_si128(); in DispatchAlphaToGreen()
83 const __m128i a0 = _mm_loadu_si128((const __m128i*)&alpha[i]); in DispatchAlphaToGreen()
84 const __m128i a1 = _mm_unpacklo_epi8(zero, a0); // note the 'zero' first! in DispatchAlphaToGreen()
85 const __m128i b1 = _mm_unpackhi_epi8(zero, a0); in DispatchAlphaToGreen()
86 const __m128i a2_lo = _mm_unpacklo_epi16(a1, zero); in DispatchAlphaToGreen()
87 const __m128i b2_lo = _mm_unpacklo_epi16(b1, zero); in DispatchAlphaToGreen()
88 const __m128i a2_hi = _mm_unpackhi_epi16(a1, zero); in DispatchAlphaToGreen()
89 const __m128i b2_hi = _mm_unpackhi_epi16(b1, zero); in DispatchAlphaToGreen()
90 _mm_storeu_si128((__m128i*)&dst[i + 0], a2_lo); in DispatchAlphaToGreen()
91 _mm_storeu_si128((__m128i*)&dst[i + 4], a2_hi); in DispatchAlphaToGreen()
92 _mm_storeu_si128((__m128i*)&dst[i + 8], b2_lo); in DispatchAlphaToGreen()
93 _mm_storeu_si128((__m128i*)&dst[i + 12], b2_hi); in DispatchAlphaToGreen()
108 const __m128i a_mask = _mm_set1_epi32(0xffu); // to preserve alpha in ExtractAlpha()
109 const __m128i all_0xff = _mm_set_epi32(0, 0, ~0u, ~0u); in ExtractAlpha()
110 __m128i all_alphas = all_0xff; in ExtractAlpha()
118 const __m128i* src = (const __m128i*)argb; in ExtractAlpha()
121 const __m128i a0 = _mm_loadu_si128(src + 0); in ExtractAlpha()
122 const __m128i a1 = _mm_loadu_si128(src + 1); in ExtractAlpha()
123 const __m128i b0 = _mm_and_si128(a0, a_mask); in ExtractAlpha()
124 const __m128i b1 = _mm_and_si128(a1, a_mask); in ExtractAlpha()
125 const __m128i c0 = _mm_packs_epi32(b0, b1); in ExtractAlpha()
126 const __m128i d0 = _mm_packus_epi16(c0, c0); in ExtractAlpha()
128 _mm_storel_epi64((__m128i*)&alpha[i], d0); in ExtractAlpha()
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(); in ApplyAlphaMultiply_SSE2()
180 const __m128i kMult = _mm_set1_epi16(0x8081u); in ApplyAlphaMultiply_SSE2()
181 const __m128i kMask = _mm_set_epi16(0, 0xff, 0xff, 0, 0, 0xff, 0xff, 0); in ApplyAlphaMultiply_SSE2()
220 const __m128i zero = _mm_setzero_si128(); in MultARGBRow_SSE2()
221 const __m128i k128 = _mm_set1_epi16(128); in MultARGBRow_SSE2()
222 const __m128i kMult = _mm_set1_epi16(0x0101); in MultARGBRow_SSE2()
223 const __m128i kMask = _mm_set_epi16(0, 0xff, 0, 0, 0, 0xff, 0, 0); in MultARGBRow_SSE2()
227 const __m128i A0 = _mm_loadl_epi64((const __m128i*)&ptr[x]); in MultARGBRow_SSE2()
228 const __m128i A1 = _mm_unpacklo_epi8(A0, zero); in MultARGBRow_SSE2()
229 const __m128i A2 = _mm_or_si128(A1, kMask); in MultARGBRow_SSE2()
230 const __m128i A3 = _mm_shufflelo_epi16(A2, _MM_SHUFFLE(2, 3, 3, 3)); in MultARGBRow_SSE2()
231 const __m128i A4 = _mm_shufflehi_epi16(A3, _MM_SHUFFLE(2, 3, 3, 3)); in MultARGBRow_SSE2()
233 const __m128i A5 = _mm_mullo_epi16(A4, A1); in MultARGBRow_SSE2()
234 const __m128i A6 = _mm_add_epi16(A5, k128); in MultARGBRow_SSE2()
235 const __m128i A7 = _mm_mulhi_epu16(A6, kMult); in MultARGBRow_SSE2()
236 const __m128i A10 = _mm_packus_epi16(A7, zero); in MultARGBRow_SSE2()
237 _mm_storel_epi64((__m128i*)&ptr[x], A10); in MultARGBRow_SSE2()
248 const __m128i zero = _mm_setzero_si128(); in MultRow_SSE2()
249 const __m128i k128 = _mm_set1_epi16(128); in MultRow_SSE2()
250 const __m128i kMult = _mm_set1_epi16(0x0101); in MultRow_SSE2()
252 const __m128i v0 = _mm_loadl_epi64((__m128i*)&ptr[x]); in MultRow_SSE2()
253 const __m128i a0 = _mm_loadl_epi64((const __m128i*)&alpha[x]); in MultRow_SSE2()
254 const __m128i v1 = _mm_unpacklo_epi8(v0, zero); in MultRow_SSE2()
255 const __m128i a1 = _mm_unpacklo_epi8(a0, zero); in MultRow_SSE2()
256 const __m128i v2 = _mm_mullo_epi16(v1, a1); in MultRow_SSE2()
257 const __m128i v3 = _mm_add_epi16(v2, k128); in MultRow_SSE2()
258 const __m128i v4 = _mm_mulhi_epu16(v3, kMult); in MultRow_SSE2()
259 const __m128i v5 = _mm_packus_epi16(v4, zero); in MultRow_SSE2()
260 _mm_storel_epi64((__m128i*)&ptr[x], v5); in MultRow_SSE2()