Details | Last modification | View Log | RSS feed
| Rev | Author | Line No. | Line |
|---|---|---|---|
| 5361 | serge | 1 | /* |
| 2 | * Copyright © <2010>, Intel Corporation. |
||
| 3 | * |
||
| 4 | * This program is licensed under the terms and conditions of the |
||
| 5 | * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at |
||
| 6 | * http://www.opensource.org/licenses/eclipse-1.0.php. |
||
| 7 | * |
||
| 8 | */ |
||
| 9 | ////////// AVC ILDB filter horizontal UV /////////////////////////////////////////////////////// |
||
| 10 | // |
||
| 11 | // This filter code prepares the src data and control data for ILDB filtering on all horizontal edges of UV. |
||
| 12 | // |
||
| 13 | // It sssumes the data for horizontal de-blocking is already transposed. |
||
| 14 | // |
||
| 15 | // Chroma: |
||
| 16 | // |
||
| 17 | // +-------+-------+ H0 Edge |
||
| 18 | // | | | |
||
| 19 | // | | | |
||
| 20 | // | | | |
||
| 21 | // +-------+-------+ H1 Edge |
||
| 22 | // | | | |
||
| 23 | // | | | |
||
| 24 | // | | | |
||
| 25 | // +-------+-------+ |
||
| 26 | // |
||
| 27 | ///////////////////////////////////////////////////////////////////////////// |
||
| 28 | |||
| 29 | #if defined(_DEBUG) |
||
| 30 | mov (1) EntrySignatureC:w 0xBBBC:w |
||
| 31 | #endif |
||
| 32 | |||
| 33 | //=============== Chroma deblocking ================ |
||
| 34 | |||
| 35 | //---------- Deblock U external top edge ---------- |
||
| 36 | and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterTopMbEdgeFlag:w // Check for FilterTopMbEdgeFlag |
||
| 37 | // (f0.0) jmpi BYPASS_EXT_TOP_EDGE_UV |
||
| 38 | |||
| 39 | // Get horizontal border edge control data. |
||
| 40 | |||
| 41 | //***** Need to take every other bit to form U maskA and mask B |
||
| 42 | // Get Luma maskA and maskB |
||
| 43 | shr (16) TempRow0(0)<1> r[ECM_AddrReg, wEdgeCntlMapA_ExtTopHorz0]<0;1,0>:uw RRampW(0) |
||
| 44 | shr (16) TempRow1(0)<1> r[ECM_AddrReg, wEdgeCntlMapB_ExtTopHorz0]<0;1,0>:uw RRampW(0) |
||
| 45 | |||
| 46 | (f0.0) jmpi ILDB_LABEL(BYPASS_EXT_TOP_EDGE_UV) |
||
| 47 | |||
| 48 | // Extract UV MaskA and MaskB from every other bit of Y masks |
||
| 49 | and.nz.f0.0 (8) null:w TempRow0(0)<16;8,2> 1:w |
||
| 50 | and.nz.f0.1 (8) null:w TempRow1(0)<16;8,2> 1:w |
||
| 51 | |||
| 52 | //---------- Deblock U external edge ---------- |
||
| 53 | // p1 = Prev MB U row 0 |
||
| 54 | // p0 = Prev MB U row 1 |
||
| 55 | // q0 = Cur MB U row 0 |
||
| 56 | // q1 = Cur MB U row 1 |
||
| 57 | // mov (1) P_AddrReg:w PREV_MB_U_BASE:w { NoDDClr } |
||
| 58 | mov (1) P_AddrReg:w TOP_MB_U_BASE:w { NoDDClr } |
||
| 59 | mov (1) Q_AddrReg:w SRC_MB_U_BASE:w { NoDDChk } |
||
| 60 | |||
| 61 | // alpha = bAlphaTop0_Cb, beta = bBetaTop0_Cb |
||
| 62 | mov (2) alpha<1>:w r[ECM_AddrReg, bAlphaTop0_Cb]<2;2,1>:ub { NoDDClr } |
||
| 63 | // tc0 has bTc0_h03_0_Cb + bTc0_h02_0_Cb + bTc0_h01_0_Cb + bTc0_h00_0_Cb |
||
| 64 | mov (4) tc0<1>:ub r[ECM_AddrReg, bTc0_h00_0_Cb]<4;4,1>:ub { NoDDChk } |
||
| 65 | |||
| 66 | // UV MaskA and MaskB |
||
| 67 | mov (2) MaskA<1>:uw f0.0<2;2,1>:uw |
||
| 68 | |||
| 69 | CALL(FILTER_UV, 1) |
||
| 70 | |||
| 71 | //---------- Deblock V external top edge ---------- |
||
| 72 | // p1 = Prev MB V row 0 |
||
| 73 | // p0 = Prev MB V row 1 |
||
| 74 | // q0 = Cur MB V row 0 |
||
| 75 | // q1 = Cur MB V row 1 |
||
| 76 | // mov (1) P_AddrReg:w PREV_MB_V_BASE:w { NoDDClr } |
||
| 77 | mov (1) P_AddrReg:w TOP_MB_V_BASE:w { NoDDClr } |
||
| 78 | mov (1) Q_AddrReg:w SRC_MB_V_BASE:w { NoDDChk } |
||
| 79 | |||
| 80 | // alpha = bAlphaTop0_Cr, beta = bBetaTop0_Cr |
||
| 81 | mov (2) alpha<1>:w r[ECM_AddrReg, bAlphaTop0_Cr]<2;2,1>:ub { NoDDClr } |
||
| 82 | |||
| 83 | // tc0 has bTc0_h03_0_Cr + bTc0_h02_0_Cr + bTc0_h01_0_Cr + bTc0_h00_0_Cr |
||
| 84 | mov (4) tc0<1>:ub r[ECM_AddrReg, bTc0_h00_0_Cr]<4;4,1>:ub { NoDDChk } |
||
| 85 | |||
| 86 | // UV MaskA and MaskB |
||
| 87 | mov (2) f0.0<1>:uw MaskA<2;2,1>:uw |
||
| 88 | |||
| 89 | CALL(FILTER_UV, 1) |
||
| 90 | |||
| 91 | ILDB_LABEL(BYPASS_EXT_TOP_EDGE_UV): |
||
| 92 | |||
| 93 | // Set EdgeCntlMap2 = 0, so it always uses bS < 4 algorithm. |
||
| 94 | |||
| 95 | // Bypass deblocking if FilterInternal4x4EdgesFlag = 0 |
||
| 96 | // and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterInternal4x4EdgesFlag:w // Check for FilterInternal4x4EdgesFlag |
||
| 97 | // (f0.0) jmpi BYPASS_4x4_DEBLOCK_H |
||
| 98 | |||
| 99 | //---------- Deblock U internal horz middle edge ---------- |
||
| 100 | |||
| 101 | //***** Need to take every other bit to form U maskA |
||
| 102 | // Get Luma maskA and maskB |
||
| 103 | shr (16) TempRow0(0)<1> r[ECM_AddrReg, wEdgeCntlMap_IntMidHorz]<0;1,0>:uw RRampW(0) |
||
| 104 | |||
| 105 | // p1 = Cur MB U row 2 |
||
| 106 | // p0 = Cur MB U row 3 |
||
| 107 | // q0 = Cur MB U row 4 |
||
| 108 | // q1 = Cur MB U row 5 |
||
| 109 | mov (1) P_AddrReg:w 4*UV_ROW_WIDTH+SRC_MB_U_BASE:w { NoDDClr } // Skip 2 U rows and 2 V rows |
||
| 110 | mov (1) Q_AddrReg:w 8*UV_ROW_WIDTH+SRC_MB_U_BASE:w { NoDDChk } |
||
| 111 | |||
| 112 | // alpha = bAlphaInternal_Cb, beta = bBetaInternal_Cb |
||
| 113 | mov (2) alpha<1>:w r[ECM_AddrReg, bAlphaInternal_Cb]<2;2,1>:ub { NoDDClr } |
||
| 114 | // tc0 has bTc0_h23_Cb + bTc0_h22_Cb + bTc0_h21_Cb + bTc0_h20_Cb |
||
| 115 | mov (4) tc0<1>:ub r[ECM_AddrReg, bTc0_h20_Cb]<4;4,1>:ub { NoDDChk } |
||
| 116 | |||
| 117 | // Extract UV MaskA and MaskB from every other bit of Y masks |
||
| 118 | and.nz.f0.0 (8) null:w TempRow0(0)<16;8,2> 1:w |
||
| 119 | |||
| 120 | // UV MaskA and MaskB |
||
| 121 | mov (1) f0.1:uw 0:w |
||
| 122 | mov (1) MaskB:uw 0:w { NoDDClr } |
||
| 123 | mov (1) MaskA:uw f0.0:uw { NoDDChk } |
||
| 124 | |||
| 125 | CALL(FILTER_UV, 1) |
||
| 126 | |||
| 127 | //---------- Deblock V internal horz middle edge ---------- |
||
| 128 | // p1 = Cur MB V row 2 |
||
| 129 | // p0 = Cur MB V row 3 |
||
| 130 | // q0 = Cur MB V row 4 |
||
| 131 | // q1 = Cur MB V row 5 |
||
| 132 | mov (1) P_AddrReg:w 4*UV_ROW_WIDTH+SRC_MB_V_BASE:w { NoDDClr } // Skip 2 U rows and 2 V rows |
||
| 133 | mov (1) Q_AddrReg:w 8*UV_ROW_WIDTH+SRC_MB_V_BASE:w { NoDDChk } |
||
| 134 | |||
| 135 | // alpha = bAlphaInternal_Cr, beta = bBetaInternal_Cr |
||
| 136 | mov (2) alpha<1>:w r[ECM_AddrReg, bAlphaInternal_Cr]<2;2,1>:ub { NoDDClr } |
||
| 137 | // tc0 has bTc0_h23_Cr + bTc0_h22_Cr + bTc0_h21_Cr + bTc0_h20_Cr |
||
| 138 | mov (4) tc0<1>:ub r[ECM_AddrReg, bTc0_h20_Cr]<4;4,1>:ub { NoDDChk } |
||
| 139 | |||
| 140 | // UV MaskA and MaskB |
||
| 141 | mov (2) f0.0<1>:uw MaskA<2;2,1>:uw |
||
| 142 | |||
| 143 | CALL(FILTER_UV, 1) |
||
| 144 | |||
| 145 | //BYPASS_4x4_DEBLOCK_H: |