static void Effect_DrawColor_Software(source *Source, layer_bitmap_info *BitmapInfo, memory *Memory, property_channel Property[]) { v4 FloatColor = Property[0].CurrentValue.col; blend_mode BlendMode = Property[1].CurrentValue.blendmode; __m256 ZeroReal = _mm256_set1_ps(0); __m256 ZeroPointFive = _mm256_set1_ps(0.5); __m256 One = _mm256_set1_ps(1); __m256 Two = _mm256_set1_ps(2); __m256 Fraction255 = _mm256_set1_ps(1/255.0f); __m256 Real255 = _mm256_set1_ps(255); __m256i FF = _mm256_set1_epi32(0xFF); __m256 Alpha = _mm256_set1_ps(FloatColor.a); __m256 AlphaInv = _mm256_set1_ps(1.0f - FloatColor.a); __m256 R_Col = _mm256_set1_ps(FloatColor.E[0]); __m256 R_Colx2 = _mm256_mul_ps(R_Col, Two); __m256 R_ColInv = _mm256_set1_ps(1.0f - FloatColor.E[0]); __m256 G_Col = _mm256_set1_ps(FloatColor.E[1]); __m256 G_Colx2 = _mm256_mul_ps(G_Col, Two); __m256 G_ColInv = _mm256_set1_ps(1.0f - FloatColor.E[1]); __m256 B_Col = _mm256_set1_ps(FloatColor.E[2]); __m256 B_Colx2 = _mm256_mul_ps(B_Col, Two); __m256 B_ColInv = _mm256_set1_ps(1.0f - FloatColor.E[2]); for (int16 Y = 0; Y < Source->Info.Height; Y += 2) { for (int16 X = 0; X < Source->Info.Width; X += 4) { uint32 XLookup = (X >> 2)*16 + (X % 4); uint32 YLookup = (Y >> 2)*(Source->Info.Width*4) + (Y % 4)*4; uint32 PixelToSeek = XLookup + YLookup; uint8 *Pixel = (uint8 *)BitmapInfo->BitmapBuffer + PixelToSeek*Source->Info.BytesPerPixel; __m256i DestPixel = _mm256_loadu_si256((const __m256i *)Pixel); // normalized values __m256 R_Dest = _mm256_mul_ps(_mm256_cvtepi32_ps(_mm256_and_si256( DestPixel, FF)), Fraction255); __m256 G_Dest = _mm256_mul_ps(_mm256_cvtepi32_ps(_mm256_and_si256(_mm256_srli_epi32(DestPixel, 8), FF)), Fraction255); __m256 B_Dest = _mm256_mul_ps(_mm256_cvtepi32_ps(_mm256_and_si256(_mm256_srli_epi32(DestPixel, 16), FF)), Fraction255); __m256i A_Out = _mm256_and_si256(_mm256_srli_epi32(DestPixel, 24), FF); __m256 R_Blend = R_Col; __m256 G_Blend = G_Col; __m256 B_Blend = B_Col; switch (BlendMode) { case blend_normal: { } break; case blend_multiply: { R_Blend = _mm256_mul_ps(R_Dest, R_Col); G_Blend = _mm256_mul_ps(G_Dest, G_Col); B_Blend = _mm256_mul_ps(B_Dest, B_Col); } break; case blend_colorburn: { R_Blend = _mm256_sub_ps(One, _mm256_div_ps(_mm256_sub_ps(One, R_Dest), R_Col)); G_Blend = _mm256_sub_ps(One, _mm256_div_ps(_mm256_sub_ps(One, G_Dest), G_Col)); B_Blend = _mm256_sub_ps(One, _mm256_div_ps(_mm256_sub_ps(One, B_Dest), B_Col)); } break; case blend_linearburn: { R_Blend = _mm256_sub_ps(_mm256_add_ps(R_Dest, R_Col), One); G_Blend = _mm256_sub_ps(_mm256_add_ps(G_Dest, G_Col), One); B_Blend = _mm256_sub_ps(_mm256_add_ps(B_Dest, B_Col), One); } break; case blend_add: { R_Blend = _mm256_add_ps(R_Dest, R_Col); G_Blend = _mm256_add_ps(G_Dest, G_Col); B_Blend = _mm256_add_ps(B_Dest, B_Col); } break; case blend_screen: { R_Blend = _mm256_sub_ps(One, _mm256_mul_ps(_mm256_sub_ps(One, R_Dest), R_ColInv)); G_Blend = _mm256_sub_ps(One, _mm256_mul_ps(_mm256_sub_ps(One, G_Dest), G_ColInv)); B_Blend = _mm256_sub_ps(One, _mm256_mul_ps(_mm256_sub_ps(One, B_Dest), B_ColInv)); } break; case blend_overlay: { __m256 R_Mask = _mm256_cmp_ps(R_Dest, ZeroPointFive, 1); __m256 G_Mask = _mm256_cmp_ps(G_Dest, ZeroPointFive, 1); __m256 B_Mask = _mm256_cmp_ps(B_Dest, ZeroPointFive, 1); __m256 R_Lower = _mm256_mul_ps(Two, _mm256_mul_ps(R_Dest, R_Col)); __m256 G_Lower = _mm256_mul_ps(Two, _mm256_mul_ps(G_Dest, G_Col)); __m256 B_Lower = _mm256_mul_ps(Two, _mm256_mul_ps(B_Dest, B_Col)); __m256 R_Upper = _mm256_sub_ps(One, _mm256_mul_ps(Two, _mm256_mul_ps(_mm256_sub_ps(One, R_Dest), R_ColInv))); __m256 G_Upper = _mm256_sub_ps(One, _mm256_mul_ps(Two, _mm256_mul_ps(_mm256_sub_ps(One, G_Dest), G_ColInv))); __m256 B_Upper = _mm256_sub_ps(One, _mm256_mul_ps(Two, _mm256_mul_ps(_mm256_sub_ps(One, B_Dest), B_ColInv))); R_Blend = _mm256_blendv_ps(R_Upper, R_Lower, R_Mask); G_Blend = _mm256_blendv_ps(G_Upper, G_Lower, G_Mask); B_Blend = _mm256_blendv_ps(B_Upper, B_Lower, B_Mask); } break; case blend_softlight: { // using Pegtop's equation R_Blend = _mm256_add_ps(_mm256_mul_ps(_mm256_sub_ps(One, R_Colx2), _mm256_mul_ps(R_Dest, R_Dest)), _mm256_mul_ps(R_Colx2, R_Dest)); G_Blend = _mm256_add_ps(_mm256_mul_ps(_mm256_sub_ps(One, G_Colx2), _mm256_mul_ps(G_Dest, G_Dest)), _mm256_mul_ps(G_Colx2, G_Dest)); B_Blend = _mm256_add_ps(_mm256_mul_ps(_mm256_sub_ps(One, B_Colx2), _mm256_mul_ps(B_Dest, B_Dest)), _mm256_mul_ps(B_Colx2, B_Dest)); } break; case blend_hardlight: { __m256 R_Mask = _mm256_cmp_ps(R_Dest, ZeroPointFive, 13); __m256 G_Mask = _mm256_cmp_ps(G_Dest, ZeroPointFive, 13); __m256 B_Mask = _mm256_cmp_ps(B_Dest, ZeroPointFive, 13); __m256 R_Lower = _mm256_mul_ps(Two, _mm256_mul_ps(R_Dest, R_Col)); __m256 G_Lower = _mm256_mul_ps(Two, _mm256_mul_ps(G_Dest, G_Col)); __m256 B_Lower = _mm256_mul_ps(Two, _mm256_mul_ps(B_Dest, B_Col)); __m256 R_Upper = _mm256_sub_ps(One, _mm256_mul_ps(Two, _mm256_mul_ps(_mm256_sub_ps(One, R_Dest), R_ColInv))); __m256 G_Upper = _mm256_sub_ps(One, _mm256_mul_ps(Two, _mm256_mul_ps(_mm256_sub_ps(One, G_Dest), G_ColInv))); __m256 B_Upper = _mm256_sub_ps(One, _mm256_mul_ps(Two, _mm256_mul_ps(_mm256_sub_ps(One, B_Dest), B_ColInv))); R_Blend = _mm256_blendv_ps(R_Upper, R_Lower, R_Mask); G_Blend = _mm256_blendv_ps(G_Upper, G_Lower, G_Mask); B_Blend = _mm256_blendv_ps(B_Upper, B_Lower, B_Mask); } break; case blend_subtract: { R_Blend = _mm256_sub_ps(R_Dest, R_Col); G_Blend = _mm256_sub_ps(G_Dest, G_Col); B_Blend = _mm256_sub_ps(B_Dest, B_Col); } break; case blend_divide: { R_Blend = _mm256_div_ps(R_Dest, R_Col); G_Blend = _mm256_div_ps(G_Dest, G_Col); B_Blend = _mm256_div_ps(B_Dest, B_Col); } break; case blend_difference: { __m256 R_Lower = _mm256_sub_ps(R_Col, R_Dest); __m256 G_Lower = _mm256_sub_ps(G_Col, G_Dest); __m256 B_Lower = _mm256_sub_ps(B_Col, B_Dest); __m256 R_Upper = _mm256_sub_ps(R_Dest, R_Col); __m256 G_Upper = _mm256_sub_ps(G_Dest, G_Col); __m256 B_Upper = _mm256_sub_ps(B_Dest, B_Col); __m256 R_Mask = _mm256_cmp_ps(R_Lower, ZeroReal, 14); __m256 G_Mask = _mm256_cmp_ps(G_Lower, ZeroReal, 14); __m256 B_Mask = _mm256_cmp_ps(B_Lower, ZeroReal, 14); R_Blend = _mm256_blendv_ps(R_Upper, R_Lower, R_Mask); G_Blend = _mm256_blendv_ps(G_Upper, G_Lower, G_Mask); B_Blend = _mm256_blendv_ps(B_Upper, B_Lower, B_Mask); } break; } R_Blend = _mm256_add_ps(_mm256_mul_ps(R_Dest, AlphaInv), _mm256_mul_ps(R_Blend, Alpha)); G_Blend = _mm256_add_ps(_mm256_mul_ps(G_Dest, AlphaInv), _mm256_mul_ps(G_Blend, Alpha)); B_Blend = _mm256_add_ps(_mm256_mul_ps(B_Dest, AlphaInv), _mm256_mul_ps(B_Blend, Alpha)); R_Blend = _mm256_max_ps(_mm256_min_ps(One, R_Blend), ZeroReal); G_Blend = _mm256_max_ps(_mm256_min_ps(One, G_Blend), ZeroReal); B_Blend = _mm256_max_ps(_mm256_min_ps(One, B_Blend), ZeroReal); __m256i R_Out = _mm256_cvttps_epi32(_mm256_mul_ps(R_Blend, Real255)); __m256i G_Out = _mm256_cvttps_epi32(_mm256_mul_ps(G_Blend, Real255)); __m256i B_Out = _mm256_cvttps_epi32(_mm256_mul_ps(B_Blend, Real255)); __m256i OutputPixel = _mm256_or_si256( _mm256_or_si256(R_Out, _mm256_slli_epi32(G_Out, 8)), _mm256_or_si256(_mm256_slli_epi32(B_Out, 16), _mm256_slli_epi32(A_Out, 24))); _mm256_storeu_si256((__m256i *)Pixel, OutputPixel); } } }