1 /* 2 * Copyright (c) 2019, Intel Corporation 3 * 4 * Permission is hereby granted, free of charge, to any person obtaining a 5 * copy of this software and associated documentation files (the "Software"), 6 * to deal in the Software without restriction, including without limitation 7 * the rights to use, copy, modify, merge, publish, distribute, sublicense, 8 * and/or sell copies of the Software, and to permit persons to whom the 9 * Software is furnished to do so, subject to the following conditions: 10 * 11 * The above copyright notice and this permission notice shall be included 12 * in all copies or substantial portions of the Software. 13 * 14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS 15 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 17 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR 18 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, 19 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR 20 * OTHER DEALINGS IN THE SOFTWARE. 21 */ 22 { 23 /* 24 Buffer layout after shuffle 25 _________________________________________________ 26 |_______Block0__________|_______Block1__________| 27 |_______Block2__________|_______Block3__________| 28 |_______Block4__________|_______Block5__________| 29 |_______Block6__________|_______Block7__________| 30 31 Write back buffer layout correlate to the block number#, each box stands for 1 GRF 32 _______________________________________________ 33 |____R0_________R1_____|____R2_________R3_____| 34 |____G0_________G1_____|____G2_________G3_____| 35 |____B0_________B1_____|____B2_________B3_____| 36 |____A0_________A1_____|____A2_________A3_____| 37 |____R4_________R5_____|____R6_________R7_____| 38 |____G4_________G5_____|____G6_________G7_____| 39 |____B4_________B5_____|____B6_________B7_____| 40 |____A4_________A5_____|____A6_________A7_____| 41 */ 42 43 matrix_ref<uint, 8, 8> Result = DataBuffer.format<uint, 96, 8>().select<8, 1, 8, 1>(64, 0); 44 45 SurfaceIndex Dst_Surface(MDF_FC_OUTPUT_BTI_START); 46 matrix_ref<uint, 4, 8> TempResult4x8_Top = Result.select<4, 1, 8, 1>(0, 0); 47 matrix_ref<uint, 4, 8> TempResult4x8_Bottom = Result.select<4, 1, 8, 1>(4, 0); 48 #pragma unroll 49 for (uchar i = 0; i < 2; i++, DstY += 8) 50 { 51 // Rounding 52 DataBuffer.format<ushort, 96, 16>().select<6, 1, 16, 1>(32 * i, 0) = cm_add<ushort>(DataBuffer.format<ushort, 96, 16>().select<6, 1, 16, 1>(32 * i, 0), 0x80, SAT); 53 DataBuffer.format<ushort, 96, 16>().select<6, 1, 16, 1>(32 * i + 8, 0) = cm_add<ushort>(DataBuffer.format<ushort, 96, 16>().select<6, 1, 16, 1>(32 * i + 8, 0), 0x80, SAT); 54 DataBuffer.format<ushort, 96, 16>().select<6, 1, 16, 1>(32 * i + 16, 0) = cm_add<ushort>(DataBuffer.format<ushort, 96, 16>().select<6, 1, 16, 1>(32 * i + 16, 0), 0x80, SAT); 55 DataBuffer.format<ushort, 96, 16>().select<6, 1, 16, 1>(32 * i + 24, 0) = cm_add<ushort>(DataBuffer.format<ushort, 96, 16>().select<6, 1, 16, 1>(32 * i + 24, 0), 0x80, SAT); 56 57 // First 8x16 58 { 59 // Y channel 60 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 8, 2>(0, DestinationPackedYOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 8, 2>(2 + 32 * i, 1); 61 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 8, 2>(0, 16 + DestinationPackedYOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 8, 2>(10 + 32 * i, 1); 62 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 8, 2>(1, DestinationPackedYOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 8, 2>(2 + 32 * i, 17); 63 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 8, 2>(1, 16 + DestinationPackedYOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 8, 2>(10 + 32 * i, 17); 64 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 8, 2>(2, DestinationPackedYOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 8, 2>(3 + 32 * i, 1); 65 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 8, 2>(2, 16 + DestinationPackedYOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 8, 2>(11 + 32 * i, 1); 66 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 8, 2>(3, DestinationPackedYOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 8, 2>(3 + 32 * i, 17); 67 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 8, 2>(3, 16 + DestinationPackedYOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 8, 2>(11 + 32 * i, 17); 68 69 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 8, 2>(0, DestinationPackedYOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 8, 2>(18 + 32 * i, 1); 70 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 8, 2>(0, 16 + DestinationPackedYOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 8, 2>(26 + 32 * i, 1); 71 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 8, 2>(1, DestinationPackedYOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 8, 2>(18 + 32 * i, 17); 72 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 8, 2>(1, 16 + DestinationPackedYOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 8, 2>(26 + 32 * i, 17); 73 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 8, 2>(2, DestinationPackedYOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 8, 2>(19 + 32 * i, 1); 74 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 8, 2>(2, 16 + DestinationPackedYOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 8, 2>(27 + 32 * i, 1); 75 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 8, 2>(3, DestinationPackedYOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 8, 2>(19 + 32 * i, 17); 76 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 8, 2>(3, 16 + DestinationPackedYOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 8, 2>(27 + 32 * i, 17); 77 78 // UV channel 79 // V channel 80 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 4, 4>(0, DestinationPackedVOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i, 1); 81 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 4, 4>(0, 16 + DestinationPackedVOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 8, 1); 82 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 4, 4>(1, DestinationPackedVOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i, 17); 83 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 4, 4>(1, 16 + DestinationPackedVOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 8, 17); 84 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 4, 4>(2, DestinationPackedVOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 1, 1); 85 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 4, 4>(2, 16 + DestinationPackedVOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 9, 1); 86 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 4, 4>(3, DestinationPackedVOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 1, 17); 87 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 4, 4>(3, 16 + DestinationPackedVOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 9, 17); 88 89 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 4, 4>(0, DestinationPackedVOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 16, 1); 90 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 4, 4>(0, 16 + DestinationPackedVOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 24, 1); 91 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 4, 4>(1, DestinationPackedVOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 16, 17); 92 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 4, 4>(1, 16 + DestinationPackedVOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 24, 17); 93 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 4, 4>(2, DestinationPackedVOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 17, 1); 94 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 4, 4>(2, 16 + DestinationPackedVOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 25, 1); 95 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 4, 4>(3, DestinationPackedVOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 17, 17); 96 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 4, 4>(3, 16 + DestinationPackedVOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 25, 17); 97 98 // U channel 99 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 4, 4>(0, DestinationPackedUOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 4, 1); 100 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 4, 4>(0, 16 + DestinationPackedUOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 12, 1); 101 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 4, 4>(1, DestinationPackedUOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 4, 17); 102 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 4, 4>(1, 16 + DestinationPackedUOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 12, 17); 103 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 4, 4>(2, DestinationPackedUOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 5, 1); 104 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 4, 4>(2, 16 + DestinationPackedUOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 13, 1); 105 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 4, 4>(3, DestinationPackedUOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 5, 17); 106 TempResult4x8_Top.format<uchar, 4, 32>().select<1, 1, 4, 4>(3, 16 + DestinationPackedUOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 13, 17); 107 108 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 4, 4>(0, DestinationPackedUOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 20, 1); 109 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 4, 4>(0, 16 + DestinationPackedUOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 28, 1); 110 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 4, 4>(1, DestinationPackedUOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 20, 17); 111 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 4, 4>(1, 16 + DestinationPackedUOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 28, 17); 112 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 4, 4>(2, DestinationPackedUOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 21, 1); 113 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 4, 4>(2, 16 + DestinationPackedUOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 29, 1); 114 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 4, 4>(3, DestinationPackedUOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 21, 17); 115 TempResult4x8_Bottom.format<uchar, 4, 32>().select<1, 1, 4, 4>(3, 16 + DestinationPackedUOffset) = DataBuffer.format<uchar, 96, 32>().select<1, 1, 4, 4>(32 * i + 29, 17); 116 117 write(Dst_Surface, DstX * 2, DstY, Result); 118 } 119 } 120 }