|
|
;/* *************************************************************************;** INTEL Corporation Proprietary Information;**;** This listing is supplied under the terms of a license;** agreement with INTEL Corporation and may not be copied;** nor disclosed except in accordance with the terms of;** that agreement.;**;** Copyright (c) 1995 Intel Corporation.;** All Rights Reserved.;**;** *************************************************************************;*/
;////////////////////////////////////////////////////////////////////////////;//;// $Header: R:\h26x\h26x\src\enc\ex5fdct.asv 1.5 14 May 1996 12:18:50 BNICKERS $;// $Log: R:\h26x\h26x\src\enc\ex5fdct.asv $;// ;// Rev 1.5 14 May 1996 12:18:50 BNICKERS;// Initial debugging of MMx B-Frame ME.;// ;// Rev 1.4 11 Apr 1996 16:02:06 AKASAI;// Updated H261 encoder to new interface and macroblock action stream;// data structure in e3mbad.inc for FORWARDDCT. Files updated together;// e1enc.cpp, e1enc.h, ex5fdct.asm, e3mbad.inc.;// ;// Added IFNDEF H261 in ex5fdct so that code used only in H263 is;// not assembled for H261.;// ;// Rev 1.3 24 Jan 1996 13:21:28 BNICKERS;// Implement OBMC;// ;// Rev 1.1 27 Dec 1995 15:32:42 RMCKENZX;// Added copyright notice;//;////////////////////////////////////////////////////////////////////////////;; e35fdct -- This function performs a Forward Discrete Cosine Transform for H263, on a stream of macroblocks comprised; of 8*8 blocks of pels or pel diffs. This version is tuned for the Pentium Microprocessor.;; Arguments:;; MBlockActionStream (Input);; A stream of MacroBlock Action Descriptors. Each descriptor indicates which blocks of a macroblock are non-empty; and thus need to be transformed. There are from 0 to 12 non-empty blocks in each macroblock.;; Processing commences with the macroblock described by the first descriptor in the stream (regardless of whether; it's End-Of-Stream bit is set). Processing continues up to but not including the next descriptor that has the; End-Of-Stream bit set.;; This function requires each descripgor in the MBlockActionStream to be 16-byte aligned. Moreover, each of the; T_Blk elements in the descriptor must also be 16-byte aligned, and ordered as they are now. (Note that I am; talking about the address of these pointer variables, not the alignement of the data they point to.);; Best performance will be attained when 8*8 blocks are (or usually are) DWORD aligned. MMx implementations will; probably prefer 8-byte alignment.;; The complete format of the MacroBlock Action Descriptors is provided in e3mbad.inc.;; TargetFrameBaseAddress -- Address of upper left viewable pel in the target Y plane. When doing B frames, this; is the Target B Frame Base Address.;; PreviousFrameBaseAddress -- Address of the reconstructed previous frame. This really isn't needed for P-frame; processing, estimation since the address of each block's prediction was recorded by; MotionEstimation. It's only used by B-frame processing.;; FutureFrameBaseAddress -- Address of the reconstructed future (a.k.a. current) P-frame. Only used when processing; B frames.;; CoeffStream (Output);; A stream of storage blocks which receive the DCT output coefficient; blocks for each non-empty blocks described in the MBlockActionStream.; Each coefficient block is 128 bytes. The stream must be large enough; to hold all the output coefficient blocks.;; Best performance will be attained by assuring the storage is 32-byte; aligned. Best performance will be attained by using the output before; the data cache gets changed by other data. Consuming the coefficient; blocks in forward order is best, since they are defined in reverse; order (and thus the first blocks are most likely to be in cache).;; The complete format of the coefficient blocks is provided in encdctc.inc.;; IsBFrame (Input);; 0 (False) if doing Key or P frame. 1 (True) if doing B frame.
OPTION PROLOGUE:NoneOPTION EPILOGUE:ReturnAndRelieveEpilogueMacroOPTION M510
include e3inst.inc ; Encoder instance datainclude e3mbad.inc ; MacroBlock Action Descriptor struct layoutinclude e3dctc.inc ; DCT Coefficient block layout
.xlistinclude memmodel.inc.list.DATA
InitTbl MACRO WeightHi,WeightLo,TableLabelTableLabel LABEL DWORD CNT = -128 REPEAT 128 DWORD ((WeightHi*CNT-08000H)/010000H*010000H)+((WeightLo*CNT-08000H)/010000H) DWORD ((WeightHi*CNT-08000H)/010000H*010000H)-((WeightLo*CNT-08000H)/010000H) CNT = CNT + 1 ENDM REPEAT 128 DWORD ((WeightHi*CNT+08000H)/010000H*010000H)+((WeightLo*CNT+08000H)/010000H) DWORD ((WeightHi*CNT+08000H)/010000H*010000H)-((WeightLo*CNT+08000H)/010000H) CNT = CNT + 1 ENDMENDM
InitTbl 080000H,04545FH,P80000_P4545FP80000_N4545F = P80000_P4545F + 4
InitTbl 080000H,0A73D7H,P80000_PA73D7P80000_NA73D7 = P80000_PA73D7 + 4
BYTE 680 DUP (?) ; To assure that tables interleave nicely in cache.
InitTbl 02350BH, 06491AH,P2350B_P6491AP2350B_N6491A = P2350B_P6491A + 4
InitTbl -0B18A8H,-096831H,NB18A8_N96831NB18A8_P96831 = NB18A8_N96831 + 4
BYTE 680 DUP (?) ; To assure that tables interleave nicely in cache.
InitTbl -096831H, 02350BH,N96831_P2350BN96831_N2350B = N96831_P2350B + 4
InitTbl 06491AH, 0B18A8H,P6491A_PB18A8P6491A_NB18A8 = P6491A_PB18A8 + 4
ColsDefined DD 000000000H,000000000H,07F7F7F7FH,07F7F7F7FH DD 000000000H,07F7F7F00H,07F7F7F7FH,00000007FH DD 000000000H,07F7F0000H,07F7F7F7FH,000007F7FH DD 000000000H,07F000000H,07F7F7F7FH,0007F7F7FH,000000000H
; Right Left Chroma DB 0 ; -22.0 DB 0 ; -21.5 DB 0 ; -21.0 DB 0 ; -20.5 DB 0 ; -20.0 DB 0 ; -19.5 DB 0 ; -19.0 DB 0 ; -18.5 DB 0 ; -18.0 DB 0 ; -17.5 DB 0 ; -17.0 DB 0 ; -16.5 DB 0 ; -16.0 DB 0 ; -15.5 DB 0 ; -15.0 DB 0 ; -14.5 DB 0 ; -22.0 -14.0 DB 0 ; -21.5 -13.5 DB 0 ; -21.0 -13.0 DB 0 ; -20.5 -12.5 DB 0 ; -20.0 -12.0 DB 0 ; -19.5 -11.5 DB 0 ; -19.0 -11.0 DB 0 ; -18.5 -10.5 DB 0 ; -18.0 -10.0 DB 0 ; -17.5 -9.5 DB 0 ; -17.0 -9.0 DB 0 ; -16.5 -8.5 DB 0 ; -16.0 -8.0 DB 0 ; -15.5 -7.5 DB 48 ; -15.0 -7.0 DB 48 ; -14.5 -6.5 DB 32 ; -14.0 -6.0 DB 32 ; -13.5 -5.5 DB 16 ; -13.0 -5.0 DB 16 ; -12.5 -4.5 DB 4 ; -12.0 -4.0 DB 4 ; -11.5 -3.5 DB 52 ; -11.0 -3.0 DB 52 ; -10.5 -2.5 DB 36 ; -10.0 -2.0 DB 36 ; -9.5 -1.5 DB 20 ; -9.0 -1.0 DB 20 ; -8.5 -.5LeftYBlkColsDef DB 8 ; -8.0 0 DB 8 ; -7.5 .5 DB 8 ; -7.0 1.0 DB 8 ; -6.5 1.5 DB 8 ; -6.0 2.0 DB 8 ; -5.5 2.5 DB 8 ; -5.0 3.0 DB 8 ; -4.5 3.5 DB 8 ; -4.0 4.0 DB 8 ; -3.5 4.5 DB 8 ; -3.0 5.0 DB 8 ; -2.5 5.5 DB 8 ; -2.0 6.0 DB 8 ; -1.5 6.5 DB 8 ; -1.0 7.0 DB 8 ; -.5 7.5RightYBlkColsDef DB 8 ; 0 8.0 DB 56 ; .5 8.5 DB 56 ; 1.0 9.0 DB 40 ; 1.5 9.5 DB 40 ; 2.0 10.0 DB 24 ; 2.5 10.5 DB 24 ; 3.0 11.0 DB 12 ; 3.5 11.5 DB 12 ; 4.0 12.0 DB 60 ; 4.5 12.5 DB 60 ; 5.0 13.0 DB 44 ; 5.5 13.5 DB 44 ; 6.0 14.0 DB 28 ; 6.5 14.5 DB 28 ; 7.0 15.0 DB 0 ; 7.5 15.5 DB 0 ; 8.0 16.0 DB 0 ; 8.5 16.5 DB 0 ; 9.0 17.0 DB 0 ; 9.5 17.5 DB 0 ; 10.0 18.0 DB 0 ; 10.5 18.5 DB 0 ; 11.0 19.0 DB 0 ; 11.5 19.5 DB 0 ; 12.0 20.0 DB 0 ; 12.5 20.5 DB 0 ; 13.0 21.0 DB 0 ; 13.5 21.5 DB 0 ; 14.0 22.0 DB 0 ; 14.5 DB 0 ; 15.0 DB 0 ; 15.5 DB 0 ; 16.0 DB 0 ; 16.5 DB 0 ; 17.0 DB 0 ; 17.5 DB 0 ; 18.0 DB 0 ; 18.5 -11.0 DB 0 ; 19.0 -10.5 DB 0 ; 19.5 -10.0 DB 0 ; 20.0 -9.5 DB 0 ; 20.5 -9.0 DB 0 ; 21.0 -8.5 DB 0 ; 21.5 -8.0 DB 0 ; 22.0 -7.5 DB 48 ; -7.0 DB 48 ; -6.5 DB 32 ; -6.0 DB 32 ; -5.5 DB 16 ; -5.0 DB 16 ; -4.5 DB 4 ; -4.0 DB 4 ; -3.5 DB 52 ; -3.0 DB 52 ; -2.5 DB 36 ; -2.0 DB 36 ; -1.5 DB 20 ; -1.0 DB 20 ; -.5ChromaColsDef DB 8 ; 0 DB 56 ; .5 DB 56 ; 1.0 DB 40 ; 1.5 DB 40 ; 2.0 DB 24 ; 2.5 DB 24 ; 3.0 DB 12 ; 3.5 DB 12 ; 4.0 DB 60 ; 4.5 DB 60 ; 5.0 DB 44 ; 5.5 DB 44 ; 6.0 DB 28 ; 6.5 DB 28 ; 7.0 DB 0 ; 7.5 DB 0 ; 8.0 DB 0 ; 8.5 DB 0 ; 9.0 DB 0 ; 9.5 DB 0 ; 10.0 DB 0 ; 10.5 DB 0 ; 11.0
; Lower Upper Chroma DB 000H ; -22.0 DB 000H ; -21.5 DB 000H ; -21.0 DB 000H ; -20.5 DB 000H ; -20.0 DB 000H ; -19.5 DB 000H ; -19.0 DB 000H ; -18.5 DB 000H ; -18.0 DB 000H ; -17.5 DB 000H ; -17.0 DB 000H ; -16.5 DB 000H ; -16.0 DB 000H ; -15.5 DB 000H ; -15.0 DB 000H ; -14.5 DB 000H ; -22.0 -14.0 DB 000H ; -21.5 -13.5 DB 000H ; -21.0 -13.0 DB 000H ; -20.5 -12.5 DB 000H ; -20.0 -12.0 DB 000H ; -19.5 -11.5 DB 000H ; -19.0 -11.0 DB 000H ; -18.5 -10.5 DB 000H ; -18.0 -10.0 DB 000H ; -17.5 -9.5 DB 000H ; -17.0 -9.0 DB 000H ; -16.5 -8.5 DB 000H ; -16.0 -8.0 DB 000H ; -15.5 -7.5 DB 001H ; -15.0 -7.0 DB 001H ; -14.5 -6.5 DB 003H ; -14.0 -6.0 DB 003H ; -13.5 -5.5 DB 007H ; -13.0 -5.0 DB 007H ; -12.5 -4.5 DB 00FH ; -12.0 -4.0 DB 00FH ; -11.5 -3.5 DB 01FH ; -11.0 -3.0 DB 01FH ; -10.5 -2.5 DB 03FH ; -10.0 -2.0 DB 03FH ; -9.5 -1.5 DB 07FH ; -9.0 -1.0 DB 07FH ; -8.5 -.5UpperYBlkLinesDef DB 0FFH ; -8.0 0 DB 0FFH ; -7.5 .5 DB 0FFH ; -7.0 1.0 DB 0FFH ; -6.5 1.5 DB 0FFH ; -6.0 2.0 DB 0FFH ; -5.5 2.5 DB 0FFH ; -5.0 3.0 DB 0FFH ; -4.5 3.5 DB 0FFH ; -4.0 4.0 DB 0FFH ; -3.5 4.5 DB 0FFH ; -3.0 5.0 DB 0FFH ; -2.5 5.5 DB 0FFH ; -2.0 6.0 DB 0FFH ; -1.5 6.5 DB 0FFH ; -1.0 7.0 DB 0FFH ; -.5 7.5LowerYBlkLinesDef DB 0FFH ; 0 8.0 DB 0FEH ; .5 8.5 DB 0FEH ; 1.0 9.0 DB 0FCH ; 1.5 9.5 DB 0FCH ; 2.0 10.0 DB 0F8H ; 2.5 10.5 DB 0F8H ; 3.0 11.0 DB 0F0H ; 3.5 11.5 DB 0F0H ; 4.0 12.0 DB 0E0H ; 4.5 12.5 DB 0E0H ; 5.0 13.0 DB 0C0H ; 5.5 13.5 DB 0C0H ; 6.0 14.0 DB 080H ; 6.5 14.5 DB 080H ; 7.0 15.0 DB 000H ; 7.5 15.5 DB 000H ; 8.0 16.0 DB 000H ; 8.5 16.5 DB 000H ; 9.0 17.0 DB 000H ; 9.5 17.5 DB 000H ; 10.0 18.0 DB 000H ; 10.5 18.5 DB 000H ; 11.0 19.0 DB 000H ; 11.5 19.5 DB 000H ; 12.0 20.0 DB 000H ; 12.5 20.5 DB 000H ; 13.0 21.0 DB 000H ; 13.5 21.5 DB 000H ; 14.0 22.0 DB 000H ; 14.5 DB 000H ; 15.0 DB 000H ; 15.5 DB 000H ; 16.0 DB 000H ; 16.5 DB 000H ; 17.0 DB 000H ; 17.5 DB 000H ; 18.0 DB 000H ; 18.5 -11.0 DB 000H ; 19.0 -10.5 DB 000H ; 19.5 -10.0 DB 000H ; 20.0 -9.5 DB 000H ; 20.5 -9.0 DB 000H ; 21.0 -8.5 DB 000H ; 21.5 -8.0 DB 000H ; 22.0 -7.5 DB 001H ; -7.0 DB 001H ; -6.5 DB 003H ; -6.0 DB 003H ; -5.5 DB 007H ; -5.0 DB 007H ; -4.5 DB 00FH ; -4.0 DB 00FH ; -3.5 DB 01FH ; -3.0 DB 01FH ; -2.5 DB 03FH ; -2.0 DB 03FH ; -1.5 DB 07FH ; -1.0 DB 07FH ; -.5ChromaLinesDef DB 0FFH ; 0 DB 0FEH ; .5 DB 0FEH ; 1.0 DB 0FCH ; 1.5 DB 0FCH ; 2.0 DB 0F8H ; 2.5 DB 0F8H ; 3.0 DB 0F0H ; 3.5 DB 0F0H ; 4.0 DB 0E0H ; 4.5 DB 0E0H ; 5.0 DB 0C0H ; 5.5 DB 0C0H ; 6.0 DB 080H ; 6.5 DB 080H ; 7.0 DB 000H ; 7.5 DB 000H ; 8.0 DB 000H ; 8.5 DB 000H ; 9.0 DB 000H ; 9.5 DB 000H ; 10.0 DB 000H ; 10.5 DB 000H ; 11.0
.CODE
;ASSUME cs : FLAT;ASSUME ds : FLAT;ASSUME es : FLAT;ASSUME fs : FLAT;ASSUME gs : FLAT;ASSUME ss : FLAT
FORWARDDCT proc C AMBlockActionStream: DWORD,ATargetFrameBaseAddress: DWORD, APreviousFrameBaseAddress: DWORD, AFutureFrameBaseAddress: DWORD, ACoeffStream: DWORD, AIsBFrame: DWORD,AIsAdvancedPrediction: DWORD, AIsPOfPBPair: DWORD, AScratchBlocks: DWORD,ANumMBlksInGOB: DWORD
LocalFrameSize = 196RegisterStorageSize = 16
; Arguments:
MBlockActionStream = RegisterStorageSize + 4TargetFrameBaseAddress_arg = RegisterStorageSize + 8PreviousFrameBaseAddress_arg = RegisterStorageSize + 12FutureFrameBaseAddress_arg = RegisterStorageSize + 16CoeffStream_arg = RegisterStorageSize + 20IsBFrame = RegisterStorageSize + 24IsAdvancedPrediction = RegisterStorageSize + 28IsPOfPBPair = RegisterStorageSize + 32ScratchBlocks = RegisterStorageSize + 36NumMBlksInGOB = RegisterStorageSize + 40EndOfArgList = RegisterStorageSize + 44
; Locals (on local stack frame)
P00 EQU [esp+ 8] ; Biased Pels or Biased Pel DifferencesP01 EQU [esp+ 9]P02 EQU [esp+ 10]P03 EQU [esp+ 11]P04 EQU [esp+ 12]P05 EQU [esp+ 13]P06 EQU [esp+ 14]P07 EQU [esp+ 15]P10 EQU [esp+ 16]P11 EQU [esp+ 17]P12 EQU [esp+ 18]P13 EQU [esp+ 19]P14 EQU [esp+ 20]P15 EQU [esp+ 21]P16 EQU [esp+ 22]P17 EQU [esp+ 23]P20 EQU [esp+ 24]P21 EQU [esp+ 25]P22 EQU [esp+ 26]P23 EQU [esp+ 27]P24 EQU [esp+ 28]P25 EQU [esp+ 29]P26 EQU [esp+ 30]P27 EQU [esp+ 31]P30 EQU [esp+ 32]P31 EQU [esp+ 33]P32 EQU [esp+ 34]P33 EQU [esp+ 35]P34 EQU [esp+ 36]P35 EQU [esp+ 37]P36 EQU [esp+ 38]P37 EQU [esp+ 39]P40 EQU [esp+ 40]P41 EQU [esp+ 41]P42 EQU [esp+ 42]P43 EQU [esp+ 43]P44 EQU [esp+ 44]P45 EQU [esp+ 45]P46 EQU [esp+ 46]P47 EQU [esp+ 47]P50 EQU [esp+ 48]P51 EQU [esp+ 49]P52 EQU [esp+ 50]P53 EQU [esp+ 51]P54 EQU [esp+ 52]P55 EQU [esp+ 53]P56 EQU [esp+ 54]P57 EQU [esp+ 55]P60 EQU [esp+ 56]P61 EQU [esp+ 57]P62 EQU [esp+ 58]P63 EQU [esp+ 59]P64 EQU [esp+ 60]P65 EQU [esp+ 61]P66 EQU [esp+ 62]P67 EQU [esp+ 63]P70 EQU [esp+ 64]P71 EQU [esp+ 65]P72 EQU [esp+ 66]P73 EQU [esp+ 67]P74 EQU [esp+ 68]P75 EQU [esp+ 69]P76 EQU [esp+ 70]P77 EQU [esp+ 71]I00I02 EQU P00 ; Intermed for row 0, columns 0 and 2. I01I03 EQU P04 ; Share storage with pels.I04I06 EQU [esp+ 72]Mask00 EQU [esp+ 72]I07I05 EQU [esp+ 76]Mask04 EQU [esp+ 76]I10I12 EQU P10I11I13 EQU P14I14I16 EQU [esp+ 80]Mask10 EQU [esp+ 80]I17I15 EQU [esp+ 84]Mask14 EQU [esp+ 84]I20I22 EQU P20I21I23 EQU P24I24I26 EQU [esp+ 88]Mask20 EQU [esp+ 88]I27I25 EQU [esp+ 92]Mask24 EQU [esp+ 92]I30I32 EQU P30I31I33 EQU P34I34I36 EQU [esp+ 96]Mask30 EQU [esp+ 96]I37I35 EQU [esp+100]Mask34 EQU [esp+100]I40I42 EQU P40I41I43 EQU P44I44I46 EQU [esp+104]Mask40 EQU [esp+104]I47I45 EQU [esp+108]Mask44 EQU [esp+108]I50I52 EQU P50I51I53 EQU P54I54I56 EQU [esp+112]Mask50 EQU [esp+112]I57I55 EQU [esp+116]Mask54 EQU [esp+116]I60I62 EQU P60I61I63 EQU P64I64I66 EQU [esp+120]Mask60 EQU [esp+120]I67I65 EQU [esp+124]Mask64 EQU [esp+124]I70I72 EQU P70I71I73 EQU P74I74I76 EQU [esp+128]Mask70 EQU [esp+128]I77I75 EQU [esp+132]Mask74 EQU [esp+132]S4 EQU I10I12 ; Temp storage, shared.S7 EQU I00I02 ; Temp storage, shared.S3 EQU I30I32 ; Temp storage, shared.S0 EQU I40I42 ; Temp storage, shared.
CoeffStreamStart EQU [esp+ 0]CoeffStream EQU [esp+ 4]BlkActionDescrAddr EQU [esp+136]FutureFrameBaseAddress EQU [esp+140]DistFromTargetToPastP EQU [esp+144]TargetFrameBaseAddress EQU [esp+148]PredictionsBaseAddress EQU [esp+152]IsPlainPFrame EQU [esp+156]PreviousFrameBaseAddress EQU [esp+160]DistToBlockToLeft EQU [esp+164]DistToBlockAbove EQU [esp+168]DistToBlockToRight EQU [esp+172]DistToBlockBelow EQU [esp+176]DistFromBlk1ToBlk3Above EQU [esp+180]MBActionCursor EQU [esp+184]CentralRefAddrAndInterps EQU [esp+188]StashESP EQU [esp+192]
push esi push edi push ebp push ebx mov ebx,esp sub esp,LocalFrameSize+4 mov edi,[ebx+CoeffStream_arg] ; Get address of storage for coeffs. and esp,0FFFFFFC0H ; Get 64-byte aligned. xor ebp,ebp add esp,4 ; esp at cache line plus 4. mov esi,[ebx+MBlockActionStream] ; Get address of MB action stream. mov StashESP,ebx mov edx,[ebx+TargetFrameBaseAddress_arg] mov TargetFrameBaseAddress,edx mov eax,[ebx+PreviousFrameBaseAddress_arg] mov PreviousFrameBaseAddress,eax sub eax,edx mov ecx,[ebx+FutureFrameBaseAddress_arg] mov FutureFrameBaseAddress,ecx mov DistFromTargetToPastP,eax mov CoeffStreamStart,edi xor eax,eax xor ecx,ecx
IFNDEF H261;; H261 does not execute the OBMC code so it is included only when H261 is not defined;; cmp ebp,[ebx+IsBFrame] mov edx,PITCH jne NextBMacroBlock
cmp ebp,[ebx+IsAdvancedPrediction] je NextMacroBlock
mov eax,[ebx+ScratchBlocks] ; We must do OBMC. mov ecx,[esi].BlkY1.BlkOffset sub eax,ecx mov ebp,[ebx+IsPOfPBPair] xor ebp,1 mov PredictionsBaseAddress,eax mov IsPlainPFrame,ebp mov ebp,[ebx+NumMBlksInGOB] imul ebp,-SIZEOF T_MacroBlockActionDescr add ebp,2*SIZEOF T_Blk mov DistFromBlk1ToBlk3Above,ebp
;===============================================================================;===============================================================================; First pass builds block action stream from macroblock action stream.;===============================================================================;===============================================================================
; esi -- MacroBlockActionStream cursor; edi -- BlockActionStream cursor; edx -- Address of a block to do; bl -- BlockType;; MB edge condition: 1 off if left edge | 2: right | 4: top | 8: bottom; eax -- Coded block pattern for P block;; (Block_number - 1) * SIZEOF T_Blk
NextMacroBlock_OBMC:
mov bl,PB [esi].BlockType mov al,PB [esi].CodedBlocks ; Bits 0- 3 set for non-empty Y blks. ; Bit 4 set for non-empty U blk. ; Bit 5 set for non-empty V blk. ; Bit 6 clear except at stream end. ; Bit 7 clear. Unused. and bl,IsINTRA jne MBIsIntraCoded_OBMC
lea edx,[esi].BlkY1+12 ; Addr of block addr (plus 12). test al,1 ; Check if block 1 empty. mov [edi].BlockAddr,edx ; Store address of block address. je Block1DescrBuilt
mov al,[esi].MBEdgeType add edi,T_CoeffBlk ; Advance block descriptor ptr. shl eax,31 mov ecx,-SIZEOF T_MacroBlockActionDescr + SIZEOF T_Blk sar eax,31 mov CoeffStream,edi ; Stash block descriptor ptr. and ecx,eax ; Blk to left is blk 2 of mb to the left, or off edge. mov al,[esi].MBEdgeType shl eax,29 mov DistToBlockToLeft,ecx sar eax,31 mov ecx,DistFromBlk1ToBlk3Above and ecx,eax ; Blk above is in macroblock above, or off upper edge. mov eax,SIZEOF T_Blk ; Blk to right is blk 2 of current macroblock. mov DistToBlockAbove,ecx mov ecx,2*SIZEOF T_Blk; Blk below is blk 3 of current macroblock. mov DistToBlockToRight,eax mov DistToBlockBelow,ecx mov ebp,T_MacroBlockActionDescr.BlkY1 jmp BuildOBMCPrediction
Block1DescrBuilt:
test al,2 ; Check if block 2 empty. lea edx,[esi].BlkY2+12 ; Addr of block addr (plus 12). mov [edi].BlockAddr,edx ; Store address of block address. je Block2DescrBuilt
mov al,[esi].MBEdgeType add edi,T_CoeffBlk ; Advance block descriptor ptr. shl eax,30 mov ecx,SIZEOF T_MacroBlockActionDescr - SIZEOF T_Blk sar eax,31 mov CoeffStream,edi ; Stash block descriptor ptr. and ecx,eax ; Blk to right is blk 1 of mb to right, or off edge. mov al,[esi].MBEdgeType shl eax,29 mov DistToBlockToRight,ecx sar eax,31 mov ecx,DistFromBlk1ToBlk3Above and ecx,eax ; Blk above is in macroblock above, or off upper edge. mov eax,-SIZEOF T_Blk ; Blk to left is blk 1 of current macroblock. mov DistToBlockAbove,ecx mov ecx,2*SIZEOF T_Blk; Blk below is blk 4 of current macroblock. mov DistToBlockToLeft,eax mov DistToBlockBelow,ecx mov ebp,T_MacroBlockActionDescr.BlkY2 jmp BuildOBMCPrediction
Block1or2DescrBuilt:
mov al,PB [esi].CodedBlocks ; Bits 0- 3 set for non-empty Y blks. mov edi,CoeffStream ; Restore block descriptor ptr. jl Block1DescrBuilt
Block2DescrBuilt:
test al,4 ; Check if block 3 empty. lea edx,[esi].BlkY3+12 ; Addr of block addr (plus 12). mov [edi].BlockAddr,edx ; Store address of block address. je Block3DescrBuilt
mov al,[esi].MBEdgeType add edi,T_CoeffBlk ; Advance block descriptor ptr. shl eax,31 mov ecx,-SIZEOF T_MacroBlockActionDescr + SIZEOF T_Blk sar eax,31 mov CoeffStream,edi ; Stash block descriptor ptr. and eax,ecx ; Blk to left is blk 4 of mb to the left, or off edge. mov ecx,-2*SIZEOF T_Blk ; Blk above is blk 1 of current mb. mov DistToBlockToLeft,eax mov eax,SIZEOF T_Blk ; Blk to right is blk 4 of current macroblock. mov DistToBlockAbove,ecx xor ecx,ecx ; Blk below is current block. mov DistToBlockToRight,eax mov DistToBlockBelow,ecx mov ebp,T_MacroBlockActionDescr.BlkY3 jmp BuildOBMCPrediction
Block3DescrBuilt:
test al,8 ; Check if block 4 empty. lea edx,[esi].BlkY4+12 ; Addr of block addr (plus 12). mov [edi].BlockAddr,edx ; Store address of block address. je Block4DescrBuilt
mov al,[esi].MBEdgeType add edi,T_CoeffBlk ; Advance block descriptor ptr. shl eax,30 mov ecx,SIZEOF T_MacroBlockActionDescr - SIZEOF T_Blk sar eax,31 mov CoeffStream,edi ; Stash block descriptor ptr. and eax,ecx ; Blk to right is blk 3 of mb to right, or off edge. mov ecx,-2*SIZEOF T_Blk ; Blk above is blk 2 of current mb. mov DistToBlockToRight,eax mov eax,-SIZEOF T_Blk ; Blk to left is blk 3 of current macroblock. mov DistToBlockAbove,ecx xor ecx,ecx ; Blk below is current block. mov DistToBlockToLeft,eax mov DistToBlockBelow,ecx mov ebp,T_MacroBlockActionDescr.BlkY4
BuildOBMCPrediction:
; esi -- MacroBlockActionStream cursor; ebp -- T_MacroBlockActionDescr.BlkYN; edi -- Address at which to put prediction block
mov edi,PredictionsBaseAddress mov eax,[esi+ebp*1].T_Blk.BlkOffset; BlkOffset add edi,eax ; Compute addr at which to put OBMC pred. mov eax,[esi+ebp*1].T_Blk.MVs ; al = horz MV; ah = vert MV. test eax,1 mov edx,[esi+ebp*1].T_Blk.PastRef ; Fetch address for ref block. mov MBActionCursor,esi jne HorzInterpInCentralPred
mov [esi+ebp*1].T_Blk.PastRef,edi ; Update address for ref block. test eax,0100H mov ecx,PITCH jne VertInterpInCentralPred
; No half pel interpolation for central point required. Just copy it.
@@:
mov eax,[edx+0] mov ebx,[edx+4] mov [edi+ 0],eax mov [edi+ 4],ebx mov [edi+ 8],eax mov [edi+12],ebx mov [edi+28],eax mov [edi+32],ebx add edx,PITCH add edi,PITCH add ebp,020000000H jnc @b
sub edi,PITCH*8 sub edx,PITCH*8-080000000H ; Address of ref, xor 10 in high 2 bits. jmp CentralPredGottenForOBMC
HorzInterpInCentralPred:
mov [esi+ebp*1].T_Blk.PastRef,edi ; Update address for ref block. test eax,0100H mov ecx,1 jne BothInterpInCentralPred
VertInterpInCentralPred:
@@:
mov eax,[edx+0] mov ebx,[edx+4] add eax,[edx+ecx+0] add ebx,[edx+ecx+4] add eax,001010101H add ebx,001010101H shr eax,1 and ebx,0FEFEFEFEH shr ebx,1 and eax,07F7F7F7FH mov [edi+ 0],eax mov [edi+ 4],ebx mov [edi+ 8],eax mov [edi+12],ebx mov [edi+28],eax mov [edi+32],ebx add edx,PITCH add edi,PITCH add ebp,020000000H jnc @b
sub edi,PITCH*8 sub edx,PITCH*8 shl ecx,30 xor edx,ecx ; Address of ref, xor 00 in high 2 bits if vertically ; ; interpolated; xor 01 if horizontally interpolated. jmp CentralPredGottenForOBMC
BothInterpInCentralPred:@@: mov eax,[edx+1] ; <P04 P03 P02 P01> prediction pels. mov esi,001010101H ; Get 001010101H mask. mov ebx,[edx] ; <P03 P02 P01 P00>. add edi,4 ; Pre-increment OBMC prediction block pointer. mov ecx,[edx+PITCH+1] ; <P14 P13 P12 P11>. add eax,ebx ; <P04+P03 P03+P02 P02+P01 P01+P00>. mov ebx,[edx+PITCH] ; <P13 P12 P11 P10>. and esi,eax ; <(P04+P03)&1 ...>. shr eax,1 ; <(P04+P03)/2 ...> (dirty). add ebx,ecx ; <P14+P13 P13+P12 P12+P11 P11+P10>. and eax,07F7F7F7FH ; <(P04+P03)/2 ...> (clean). add ebx,esi ; <P14+P13+((P04+P03)&1) ...>. shr ebx,1 ; <(P14+P13+((P04+P03)&1))/2 ...> (dirty). add edx,4 ; Advance reference block pointer. and ebx,07F7F7F7FH ; <(P14+P13+((P04+P03)&1))/2 ...> (clean). add eax,001010101H ; <(P04+P03)/2+1 ...>. add ebx,eax ; <(P04+P03)/2+1+(P14+P13+((P04+P03)&1))/2 ...>. mov eax,4 shr ebx,1 ; <((P04+P03)/2+1+(P14+P13+((P04+P03)&1))/2)/2 ...>. mov esi,MBActionCursor ; Speculatively restore esi. and ebx,07F7F7F7FH ; Interpolated prediction. and eax,edi mov [edi-4],ebx mov [edi+8-4],ebx mov [edi+28-4],ebx jne @b
add edi,PITCH-8 ; Advance to next line of block. add edx,PITCH-8 ; Advance to next line of block. add ebp,020000000H ; Iterate 8 times. Quit when carry flag gets set. jnc @b
sub edx,PITCH*8 xor edx,0C0000000H ; Address of ref, xor 11 in high 2 bits. sub edi,PITCH*8
CentralPredGottenForOBMC:
; At this point, the central contribution to OBMC prediction is in its scratch; block, whose address has been written to PastRef in the block action descr.;; esi -- MacroBlockActionStream cursor; ebp -- (Block_number - 1) * SIZEOF T_Blk; edi -- Address at which to put prediction block; edx -- Address of central reference. High 2 bits xor'ed as follows:; 00 -- If central ref was interpolated vertically.; 01 -- If central ref was interpolated horizontally.; 10 -- If central ref was not interpolated.; 11 -- If central ref was interpolated both ways.; eax -- Offset to block descriptor for block to left.
mov eax,DistToBlockToLeft lea ebx,[esi+ebp] add ebx,eax ; Address of block descriptor for block to the left. mov ecx,-SIZEOF T_MacroBlockActionDescr and ecx,ebx ; Address of macroblock descr for block to the left. mov ah,IsPlainPFrame ; 0 if P of PB; 1 if run-of-the-mill P frame. mov ebx,[ebx].T_Blk.MVs mov CentralRefAddrAndInterps,edx ; Stash function of ref addr and interps. mov al,[ecx].BlockType ; Bottom bit set if left neighbor is INTRA. mov cl,bh and al,ah ; 0 if PB frame or if not INTRA jne LeftPredGottenForOBMC ; Jump if INTRA in plain P frame. (Use central) shl ebx,24 ; Get horz MV in [24:31]. mov eax,[esi+ebp*1].T_Blk.BlkOffset sar ecx,1 ; CF==1 if interp vertically. jc InterpVertForTheLeftContrib
shl ecx,25 sar ebx,25 ; Sign extend horz MV. CF==1 if interp horizontally. jc InterpHorzForTheLeftContrib
IF PITCH-384**** Magic leaks out if pitch not equal to 384ENDIF
lea ecx,[ecx+ecx*2] ; Multiply vert by 3 (to affect mult by 384) add eax,ebx ; Start accumulating left ref addr in eax. sar ecx,18 ; Sign extend vert MV. It's now linearized. mov ebx,PreviousFrameBaseAddress add eax,ebx ; Continue to accumulate left ref addr in eax. xor edx,080000000H ; Hi 2 bits of central ref same as this ref if ; ; central ref also was not interpolated. add ecx,eax ; Finish accumulating left ref addr in ecx. cmp ecx,edx ; Is central ref the same? je LeftPredGottenForOBMC
mov ebx,[ecx+PITCH*0] mov [edi+PITCH*0+8],ebx mov ebx,[ecx+PITCH*1] mov [edi+PITCH*1+8],ebx mov ebx,[ecx+PITCH*2] mov [edi+PITCH*2+8],ebx mov ebx,[ecx+PITCH*3] mov [edi+PITCH*3+8],ebx mov ebx,[ecx+PITCH*4] mov [edi+PITCH*4+8],ebx mov ebx,[ecx+PITCH*5] mov [edi+PITCH*5+8],ebx mov ebx,[ecx+PITCH*6] mov [edi+PITCH*6+8],ebx mov ebx,[ecx+PITCH*7] mov [edi+PITCH*7+8],ebx jmp LeftPredGottenForOBMC
InterpVertForTheLeftContrib:
shl ecx,25 sar ebx,25 ; Sign extend horz MV. CF==1 if interp horizontally. jc InterpBothForTheLeftContrib
IF PITCH-384**** Magic leaks out if pitch not equal to 384ENDIF
lea ecx,[ecx+ecx*2] ; Multiply vert by 3 (to affect mult by 384) add eax,ebx ; Start accumulating left ref addr in eax. sar ecx,18 ; Sign extend vert MV. It's now linearized. mov ebx,PreviousFrameBaseAddress add ebx,eax ; Continue to accumulate left ref addr in eax. ; ; Hi 2 bits of central ref same as this ref if ; ; central ref also interpolated vertically. add ecx,ebx ; Finish accumulating left ref addr in ecx. mov ebx,PITCH cmp ecx,edx ; Is central ref the same? je LeftPredGottenForOBMC
DoInterpHorzForTheLeftContrib:@@:
mov eax,[ecx+0] add edi,PITCH mov edx,[ecx+ebx+0] add eax,001010101H add eax,edx add ecx,PITCH shr eax,1 ; and eax,07F7F7F7FH add ebp,020000000H mov [edi+ 8-PITCH],eax jnc @b
sub edi,PITCH*8 jmp LeftPredGottenForOBMC
InterpBothForTheLeftContrib:
IF PITCH-384**** Magic leaks out if pitch not equal to 384ENDIF
lea ecx,[ecx+ecx*2] ; Multiply vert by 3 (to affect mult by 384) add eax,ebx ; Start accumulating left ref addr in eax. sar ecx,18 ; Sign extend vert MV. It's now linearized. mov ebx,PreviousFrameBaseAddress add eax,ebx ; Continue to accumulate left ref addr in eax. xor edx,0C0000000H ; Hi 2 bits of central ref same as this ref if ; ; central ref also interpolated both ways. add ecx,eax ; Finish accumulating left ref addr in ecx. cmp ecx,edx ; Is central ref the same? je LeftPredGottenForOBMC
@@:
mov eax,[ecx+1] ; <P04 P03 P02 P01> prediction pels. mov esi,001010101H ; Get 001010101H mask. mov ebx,[ecx] ; <P03 P02 P01 P00>. add edi,PITCH ; Pre-increment OBMC prediction block pointer. mov edx,[ecx+PITCH+1] ; <P14 P13 P12 P11>. add eax,ebx ; <P04+P03 P03+P02 P02+P01 P01+P00>. mov ebx,[ecx+PITCH] ; <P13 P12 P11 P10>. and esi,eax ; <(P04+P03)&1 ...>. shr eax,1 ; <(P04+P03)/2 ...> (dirty). add ebx,edx ; <P14+P13 P13+P12 P12+P11 P11+P10>. and eax,07F7F7F7FH ; <(P04+P03)/2 ...> (clean). add ebx,esi ; <P14+P13+((P04+P03)&1) ...>. shr ebx,1 ; <(P14+P13+((P04+P03)&1))/2 ...> (dirty). add ecx,PITCH ; Advance reference block pointer. and ebx,07F7F7F7FH ; <(P14+P13+((P04+P03)&1))/2 ...> (clean). add eax,001010101H ; <(P04+P03)/2+1 ...>. add ebx,eax ; <(P04+P03)/2+1+(P14+P13+((P04+P03)&1))/2 ...>.
shr ebx,1 ; <((P04+P03)/2+1+(P14+P13+((P04+P03)&1))/2)/2 ...>. mov esi,MBActionCursor ; Speculatively restore esi. and ebx,07F7F7F7FH ; Interpolated prediction. add ebp,020000000H ; Iterate 8 times. Quit when carry flag gets set. mov [edi+8-PITCH],ebx jnc @b
sub edi,PITCH*8 jmp LeftPredGottenForOBMC
InterpHorzForTheLeftContrib:
IF PITCH-384**** Magic leaks out if pitch not equal to 384ENDIF
lea ecx,[ecx+ecx*2] ; Multiply vert by 3 (to affect mult by 384) add eax,ebx ; Start accumulating left ref addr in eax. sar ecx,18 ; Sign extend vert MV. It's now linearized. mov ebx,PreviousFrameBaseAddress add eax,ebx ; Continue to accumulate left ref addr in eax. xor edx,040000000H ; Hi 2 bits of central ref same as this ref if ; ; central ref also interpolated horizontally. add ecx,eax ; Finish accumulating left ref addr in ecx. mov ebx,1 cmp ecx,edx ; Is central ref the same? jne DoInterpHorzForTheLeftContrib
LeftPredGottenForOBMC:
; At this point, the left contribution to OBMC prediction is in its scratch; half block. Now do the right contribution.;; esi -- MacroBlockActionStream cursor; ebp -- (Block_number - 1) * SIZEOF T_Blk; edi -- Address at which to put prediction block; edx -- Address of central reference. High 2 bits xor'ed as follows:; 00 -- If central ref was interpolated vertically.; 01 -- If central ref was interpolated horizontally.; 10 -- If central ref was not interpolated.; 11 -- If central ref was interpolated both ways.; eax -- Offset to block descriptor for block to right.
mov eax,DistToBlockToRight lea ebx,[esi+ebp] add ebx,eax mov ecx,-SIZEOF T_MacroBlockActionDescr and ecx,ebx mov ah,IsPlainPFrame mov ebx,[ebx].T_Blk.MVs mov edx,CentralRefAddrAndInterps ; Reload function of ref addr and interps. mov al,[ecx].BlockType mov cl,bh and al,ah jne RightPredGottenForOBMC shl ebx,24 mov eax,[esi+ebp*1].T_Blk.BlkOffset sar ecx,1 jc InterpVertForTheRightContrib
shl ecx,25 sar ebx,25 jc InterpHorzForTheRightContrib
IF PITCH-384**** Magic leaks out if pitch not equal to 384ENDIF
lea ecx,[ecx+ecx*2] add eax,ebx sar ecx,18 mov ebx,PreviousFrameBaseAddress add eax,ebx xor edx,080000000H add ecx,eax cmp ecx,edx je RightPredGottenForOBMC
mov ebx,[ecx+PITCH*0+4] mov [edi+PITCH*0+12],ebx mov ebx,[ecx+PITCH*1+4] mov [edi+PITCH*1+12],ebx mov ebx,[ecx+PITCH*2+4] mov [edi+PITCH*2+12],ebx mov ebx,[ecx+PITCH*3+4] mov [edi+PITCH*3+12],ebx mov ebx,[ecx+PITCH*4+4] mov [edi+PITCH*4+12],ebx mov ebx,[ecx+PITCH*5+4] mov [edi+PITCH*5+12],ebx mov ebx,[ecx+PITCH*6+4] mov [edi+PITCH*6+12],ebx mov ebx,[ecx+PITCH*7+4] mov [edi+PITCH*7+12],ebx jmp RightPredGottenForOBMC
InterpVertForTheRightContrib:
shl ecx,25 sar ebx,25 jc InterpBothForTheRightContrib
IF PITCH-384**** Magic leaks out if pitch not equal to 384ENDIF
lea ecx,[ecx+ecx*2] add eax,ebx sar ecx,18 mov ebx,PreviousFrameBaseAddress add ebx,eax add ecx,ebx mov ebx,PITCH cmp ecx,edx je RightPredGottenForOBMC
DoInterpHorzForTheRightContrib:@@:
mov eax,[ecx+4] add edi,PITCH mov edx,[ecx+ebx+4] add eax,001010101H add eax,edx add ecx,PITCH shr eax,1 ; and eax,07F7F7F7FH add ebp,020000000H mov [edi+12-PITCH],eax jnc @b
sub edi,PITCH*8 jmp RightPredGottenForOBMC
InterpBothForTheRightContrib:
IF PITCH-384**** Magic leaks out if pitch not equal to 384ENDIF
lea ecx,[ecx+ecx*2] add eax,ebx sar ecx,18 mov ebx,PreviousFrameBaseAddress add eax,ebx xor edx,0C0000000H add ecx,eax cmp ecx,edx je RightPredGottenForOBMC
@@:
mov eax,[ecx+5] mov esi,001010101H mov ebx,[ecx+4] add edi,PITCH mov edx,[ecx+PITCH+5] add eax,ebx mov ebx,[ecx+PITCH+4] and esi,eax shr eax,1 add ebx,edx and eax,07F7F7F7FH add ebx,esi shr ebx,1 add ecx,PITCH and ebx,07F7F7F7FH add eax,001010101H add ebx,eax
shr ebx,1 mov esi,MBActionCursor and ebx,07F7F7F7FH add ebp,020000000H mov [edi+12-PITCH],ebx jnc @b
sub edi,PITCH*8 jmp RightPredGottenForOBMC
InterpHorzForTheRightContrib:
IF PITCH-384**** Magic leaks out if pitch not equal to 384ENDIF
lea ecx,[ecx+ecx*2] add eax,ebx sar ecx,18 mov ebx,PreviousFrameBaseAddress add eax,ebx xor edx,040000000H add ecx,eax mov ebx,1 cmp ecx,edx jne DoInterpHorzForTheRightContrib
RightPredGottenForOBMC:
; At this point, the left and right contributions to OBMC prediction are in; their scratch half blocks. Now do the contribution for the block above.;; esi -- MacroBlockActionStream cursor; ebp -- (Block_number - 1) * SIZEOF T_Blk; edi -- Address at which to put prediction block; edx -- Address of central reference. High 2 bits xor'ed as follows:; 00 -- If central ref was interpolated vertically.; 01 -- If central ref was interpolated horizontally.; 10 -- If central ref was not interpolated.; 11 -- If central ref was interpolated both ways.; eax -- Offset to block descriptor for block above.
mov eax,DistToBlockAbove lea ebx,[esi+ebp] add ebx,eax mov ecx,-SIZEOF T_MacroBlockActionDescr and ecx,ebx mov ah,IsPlainPFrame mov ebx,[ebx].T_Blk.MVs mov edx,CentralRefAddrAndInterps mov al,[ecx].BlockType mov cl,bh and al,ah jne AbovePredGottenForOBMC shl ebx,24 mov eax,[esi+ebp*1].T_Blk.BlkOffset sar ecx,1 jc InterpVertForTheAboveContrib
shl ecx,25 sar ebx,25 jc InterpHorzForTheAboveContrib
IF PITCH-384**** Magic leaks out if pitch not equal to 384ENDIF
lea ecx,[ecx+ecx*2] add eax,ebx sar ecx,18 mov ebx,PreviousFrameBaseAddress add eax,ebx xor edx,080000000H add ecx,eax cmp ecx,edx je AbovePredGottenForOBMC
mov edx,[ecx+PITCH*0+0] mov ebx,[ecx+PITCH*0+4] mov [edi+PITCH*0+28],edx mov [edi+PITCH*0+32],ebx mov edx,[ecx+PITCH*1+0] mov ebx,[ecx+PITCH*1+4] mov [edi+PITCH*1+32],ebx mov [edi+PITCH*1+28],edx mov edx,[ecx+PITCH*2+0] mov ebx,[ecx+PITCH*2+4] mov [edi+PITCH*2+28],edx mov [edi+PITCH*2+32],ebx mov edx,[ecx+PITCH*3+0] mov ebx,[ecx+PITCH*3+4] mov [edi+PITCH*3+32],ebx mov [edi+PITCH*3+28],edx jmp AbovePredGottenForOBMC
InterpVertForTheAboveContrib:
shl ecx,25 sar ebx,25 jc InterpBothForTheAboveContrib
IF PITCH-384**** Magic leaks out if pitch not equal to 384ENDIF
lea ecx,[ecx+ecx*2] add eax,ebx sar ecx,18 mov ebx,PreviousFrameBaseAddress add ebx,eax add ecx,ebx mov ebx,PITCH cmp ecx,edx je AbovePredGottenForOBMC
DoInterpHorzForTheAboveContrib:@@:
mov eax,[ecx+0] mov edx,[ecx+4] add eax,[ecx+ebx+0] add edx,[ecx+ebx+4] add eax,001010101H add edx,001010101H shr eax,1 and edx,0FEFEFEFEH shr edx,1 and eax,07F7F7F7FH mov [edi+28],eax mov [edi+32],edx add ecx,PITCH add edi,PITCH add ebp,040000000H jnc @b
sub edi,PITCH*4 jmp AbovePredGottenForOBMC
InterpBothForTheAboveContrib:
IF PITCH-384**** Magic leaks out if pitch not equal to 384ENDIF
lea ecx,[ecx+ecx*2] add eax,ebx sar ecx,18 mov ebx,PreviousFrameBaseAddress add eax,ebx xor edx,0C0000000H add ecx,eax cmp ecx,edx je AbovePredGottenForOBMC
@@:
mov eax,[ecx+1] mov esi,001010101H mov ebx,[ecx] add edi,4 mov edx,[ecx+PITCH+1] add eax,ebx mov ebx,[ecx+PITCH] and esi,eax shr eax,1 add ebx,edx and eax,07F7F7F7FH add ebx,esi shr ebx,1 add ecx,4 and ebx,07F7F7F7FH add eax,001010101H add ebx,eax mov eax,4 shr ebx,1 mov esi,MBActionCursor and ebx,07F7F7F7FH and eax,edi mov [edi+28-4],ebx jne @b
add edi,PITCH-8 add ecx,PITCH-8 add ebp,040000000H jnc @b
sub edi,PITCH*4 jmp AbovePredGottenForOBMC
InterpHorzForTheAboveContrib:
IF PITCH-384**** Magic leaks out if pitch not equal to 384ENDIF
lea ecx,[ecx+ecx*2] add eax,ebx sar ecx,18 mov ebx,PreviousFrameBaseAddress add eax,ebx xor edx,040000000H add ecx,eax mov ebx,1 cmp ecx,edx jne DoInterpHorzForTheAboveContrib
AbovePredGottenForOBMC:
; At this point, the left, right, and above contributions to OBMC prediction; are in their scratch half blocks. Now do contribution for the block below.;; esi -- MacroBlockActionStream cursor; ebp -- (Block_number - 1) * SIZEOF T_Blk; edi -- Address at which to put prediction block; edx -- Address of central reference. High 2 bits xor'ed as follows:; 00 -- If central ref was interpolated vertically.; 01 -- If central ref was interpolated horizontally.; 10 -- If central ref was not interpolated.; 11 -- If central ref was interpolated both ways.; eax -- Offset to block descriptor for block above.
mov eax,DistToBlockBelow lea ebx,[esi+ebp] add ebx,eax mov ecx,-SIZEOF T_MacroBlockActionDescr and ecx,ebx mov ah,IsPlainPFrame mov ebx,[ebx].T_Blk.MVs mov edx,CentralRefAddrAndInterps mov al,[ecx].BlockType mov cl,bh and al,ah jne BelowPredGottenForOBMC shl ebx,24 mov eax,[esi+ebp*1].T_Blk.BlkOffset sar ecx,1 jc InterpVertForTheBelowContrib
shl ecx,25 sar ebx,25 jc InterpHorzForTheBelowContrib
IF PITCH-384**** Magic leaks out if pitch not equal to 384ENDIF
lea ecx,[ecx+ecx*2] add eax,ebx sar ecx,18 mov ebx,PreviousFrameBaseAddress add eax,ebx xor edx,080000000H add ecx,eax cmp ecx,edx je BelowPredGottenForOBMC
mov edx,[ecx+PITCH*4+0] mov ebx,[ecx+PITCH*4+4] mov [edi+PITCH*4+28],edx mov [edi+PITCH*4+32],ebx mov edx,[ecx+PITCH*5+0] mov ebx,[ecx+PITCH*5+4] mov [edi+PITCH*5+32],ebx mov [edi+PITCH*5+28],edx mov edx,[ecx+PITCH*6+0] mov ebx,[ecx+PITCH*6+4] mov [edi+PITCH*6+28],edx mov [edi+PITCH*6+32],ebx mov edx,[ecx+PITCH*7+0] mov ebx,[ecx+PITCH*7+4] mov [edi+PITCH*7+32],ebx mov [edi+PITCH*7+28],edx jmp BelowPredGottenForOBMC
InterpVertForTheBelowContrib:
shl ecx,25 sar ebx,25 jc InterpBothForTheBelowContrib
IF PITCH-384**** Magic leaks out if pitch not equal to 384ENDIF
lea ecx,[ecx+ecx*2] add eax,ebx sar ecx,18 mov ebx,PreviousFrameBaseAddress add eax,ebx add ecx,eax mov ebx,PITCH cmp ecx,edx je BelowPredGottenForOBMC
DoInterpHorzForTheBelowContrib:@@:
mov eax,[ecx+PITCH*4+0] mov edx,[ecx+PITCH*4+4] add eax,[ecx+ebx+PITCH*4+0] add edx,[ecx+ebx+PITCH*4+4] add eax,001010101H add edx,001010101H shr eax,1 and edx,0FEFEFEFEH shr edx,1 and eax,07F7F7F7FH mov [edi+PITCH*4+28],eax mov [edi+PITCH*4+32],edx add ecx,PITCH add edi,PITCH add ebp,040000000H jnc @b
sub edi,PITCH*4 jmp BelowPredGottenForOBMC
InterpBothForTheBelowContrib:
IF PITCH-384**** Magic leaks out if pitch not equal to 384ENDIF
lea ecx,[ecx+ecx*2] add eax,ebx sar ecx,18 mov ebx,PreviousFrameBaseAddress add eax,ebx xor edx,0C0000000H add ecx,eax cmp ecx,edx je BelowPredGottenForOBMC
@@:
mov eax,[ecx+PITCH*4+1] mov esi,001010101H mov ebx,[ecx+PITCH*4] add edi,4 mov edx,[ecx+PITCH*5+1] add eax,ebx mov ebx,[ecx+PITCH*5] and esi,eax shr eax,1 add ebx,edx and eax,07F7F7F7FH add ebx,esi shr ebx,1 add ecx,4 and ebx,07F7F7F7FH add eax,001010101H add ebx,eax mov eax,4 shr ebx,1 mov esi,MBActionCursor and ebx,07F7F7F7FH and eax,edi mov [edi+PITCH*4+28-4],ebx jne @b
add edi,PITCH-8 add ecx,PITCH-8 add ebp,040000000H jnc @b
sub edi,PITCH*4 jmp BelowPredGottenForOBMC
InterpHorzForTheBelowContrib:
IF PITCH-384**** Magic leaks out if pitch not equal to 384ENDIF
lea ecx,[ecx+ecx*2] add eax,ebx sar ecx,18 mov ebx,PreviousFrameBaseAddress add eax,ebx xor edx,040000000H add ecx,eax mov ebx,1 cmp ecx,edx jne DoInterpHorzForTheBelowContrib
BelowPredGottenForOBMC:
; At this point all the contributions to OBMC prediction are in their scratch; half blocks. Now combine them to get the OBMC prediction.;; ebp -- (Block_number - 1) * SIZEOF T_Blk; edi -- Address at which to put prediction block @@:
mov eax,[edi+4] ; <C07 C05 C05 C04> or <C77 C76 C75 C74> mov ebx,[edi+12] ; <R07 R06 R05 R04> or <R77 R76 R75 R74> mov ecx,[edi+32] ; <A07 A06 A05 A04> or <B77 B76 B75 B74> mov esi,[edi] ; <C03 C02 C01 C00> or <C73 C72 C71 C70> lea edx,[eax+ebx] ; <junk C6+R6 C5+R5 C4+R4> and ebx,0FF000000H ; <R7 __ __ __> shr edx,1 ; <junk (C6+R6)/2 (C5+R5)/2 (C4+R4)/2> dirty add ecx,ebx ; <A7+R7 A6 A5 A4> and edx,0007F7F7FH ; <__ (C6+R6)/2 (C5+R5)/2 (C4+R4)/2> clean mov ebx,[edi+8] ; <L03 L02 L01 L00> or <L73 L72 L71 L70> add edx,ecx ; <(2A7+2R7)/2 (2A6+C5+R5)/2 ...> add edi,PITCH*7 ; Move from line 0 to 7 (or 7 to 14) shr edx,1 ; <(2A7+2R7)/4 (2A6+C5+R5)/4 ...> dirty add ebx,esi ; <C3+L3 C2+L2 C1+L1 junk> shr ebx,1 ; <(C3+L3)/2 (C2+L2)/2 (C1+L1)/2 junk> dirty and edx,07F7F7F7FH ; <(2A7+2R7)/4 (2A6+C5+R5)/4 ...> clean and ebx,07F7F7F7FH ; <(C3+L3)/2 (C2+L2)/2 (C1+L1)/2 junk> clean mov ecx,[edi+28-PITCH*7] ; <A03 A02 A01 A00> or <B73 B72 B71 B70> lea eax,[eax+edx+001010101H]; <(2A7+4C7+2R7+4)/4 (2A6+5C5+R5+4)/4 ...> mov bl,[edi+8-PITCH*7] ; <(C3+L3)/2 (C2+L2)/2 (C1+L1)/2 L0> shr eax,1 ; <(2A7+4C7+2R7+4)/8 (2A6+5C5+R5+4)/8 ...> dirty add ebx,ecx ; <... (2A1+C1+L1)/2 (2A0+2L0)/2> shr ebx,1 ; <... (2A1+C1+L1)/4 (2A0+2L0)/4> dirty and eax,07F7F7F7FH ; <(2A7+4C7+2R7+4)/8 (2A6+5C5+R5+4)/8 ...> clean and ebx,07F7F7F7FH ; <... (2A1+C1+L1)/4 (2A0+2L0)/4> clean add esi,001010101H ; <C3+1 C2+1 C1+1 C0+1> add ebx,esi ; <... (2A1+5C1+L1+4)/4 (2A0+4C0+2L0+4)/4> mov [edi+4-PITCH*7],eax ; Store OBMC pred for pels 4-7 of line 0 or 7. shr ebx,1 ; <... (2A1+5C1+L1+4)/8 (2A0+4C0+2L0+4)/8> dirty lea esi,[edi-PITCH*13] ; Speculatively advance to line 1. and ebx,07F7F7F7FH ; <... (2A1+5C1+L1+4)/8 (2A0+4C0+2L0+4)/8> clean add ebp,080000000H mov [edi-PITCH*7],ebx ; Store OBMC pred for pels 0-3 of line 0 or 7. jnc @b
@@:
mov edx,[esi+28] ; <A13 A12 A11 A10> or <B63 B62 B61 B60> mov eax,[esi+8] ; <L13 L12 L11 L10> or <L63 L62 L61 L60> mov ecx,[esi+32] ; <A17 A16 A15 A14> or <B67 B66 B65 B64> mov ebx,[esi+12] ; <R17 R16 R15 R14> or <R67 R66 R65 R64> mov edi,[esi] ; <C13 C12 C11 C10> or <C63 C62 C61 C60> add esi,PITCH*5 ; Move from line 1 to 6 (or 6 to 11) xchg dx,ax ; edx: <A3 A2 L1 L0> eax: <L3 L2 A1 A0> xchg cx,bx ; ecx: <A7 A6 R5 R4> ebx: <R7 R6 A5 A4> add eax,edi ; <C3+L3 C2+L2 C1+A1 C0+A0> mov edi,[esi+4-PITCH*5] ; <C17 C15 C15 C14> or <C67 C66 C65 C64> shr eax,1 ; <(C3+L3)/2 (C2+L2)/2 (C1+A1)/2 (C0+A0)/2>dirty add ecx,edi ; <C7+A7 C6+A6 C5+R5 C4+R4> shr ecx,1 ; <(C7+A7)/2 (C6+A6)/2 (C5+R5)/2 (C4+R4)/2>dirty and eax,07F7F7F7FH ; <(C3+L3)/2 (C2+L2)/2 (C1+A1)/2 (C0+A0)/2>clean add eax,edx ; <(C3+L3+2A3)/2 ... (C1+2L1+A1)/2 ...> and ecx,07F7F7F7FH ; <(C7+A7)/2 (C6+A6)/2 (C5+R5)/2 (C4+R4)/2>clean shr eax,1 ; <(C3+L3+2A3)/4 ... (C1+2L1+A1)/4 ...> dirty add ecx,ebx ; <(C7+2R7+A7)/2 ... (C5+R5+2A5)/2 ...> mov ebx,[esi-PITCH*5] ; <C13 C12 C11 C10> or <C63 C62 C61 C60> and eax,07F7F7F7FH ; <(C3+L3+2A3)/4 ... (C1+2L1+A1)/4 ...> clean shr ecx,1 ; <(C7+2R7+A7)/4 ... (C5+R5+2A5)/4 ...> dirty add edi,001010101H ; <C7+1 C6+1 C5+1 C4+1> and ecx,07F7F7F7FH ; <(C7+2R7+A7)/4 ... (C5+R5+2A5)/4 ...> clean lea eax,[eax+ebx+001010101H]; <(5C3+L3+2A3+4)/4 ... (5C1+2L1+A1)/4 ...> shr eax,1 ; <(5C3+L3+2A3+4)/8 ... (5C1+2L1+A1)/8 ...>dirty add ecx,edi ; <(5C7+2R7+A7+4)/4 ... (5C5+R5+2A5)/4 ...> shr ecx,1 ; <(5C7+2R7+A7+4)/8 ... (5C5+R5+2A5)/8 ...>dirty and eax,07F7F7F7FH ; <(5C3+L3+2A3+4)/8 ... (5C1+2L1+A1)/8 ...>clean and ecx,07F7F7F7FH ; <(5C7+2R7+A7+4)/8 ... (5C5+R5+2A5)/8 ...>clean mov [esi-PITCH*5],eax ; Store OBMC pred for pels 4-7 of line 1 or 6. mov [esi+4-PITCH*5],ecx ; Store OBMC pred for pels 0-3 of line 1 or 6. lea edi,[esi-PITCH*9] ; Speculatively advance to line 2. add ebp,080000000H jnc @b
@@:
mov eax,[edi+4] ; <C27 C26 C25 C24> ... <C57 C56 C55 C54> mov ebx,[edi+12] ; <R27 R26 R25 R24> ... <R57 R56 R55 R54> add bl,al ; <R7 R6 R5 C4+R4> mov ecx,[edi] ; <C23 C22 C21 C20> ... <C53 C52 C51 C50> shr bl,1 ; <R7 R6 R5 (C4+R4)/2> mov edx,[edi+8] ; <L23 L22 L21 L20> ... <L53 L52 L51 L50> add bh,ah ; <R7 R6 C5+R5 (C4+R4)/2> add edx,ecx ; <C3+L3 C2+L2 junk junk> shr bh,1 ; <2R7/2 2R6/2 (C5+R5)/2 (C4+R4)/2> mov esi,[edi+32] ; <A27 A26 A25 A24> ... <B57 B56 B55 B54> shr edx,1 ; <(C3+L3)/2 (C2+L2)/2 junk junk> dirty add esi,eax ; <C7+A7 C6+A6 C5+A5 C4+A4> shr esi,1 ; <(C7+A7)/2 (C6+A6)/2 (C5+A5)/2 (C4+A4)/2>dirty and edx,07F7F7F7FH ; <(C3+L3)/2 (C2+L2)/2 junk junk> clean and esi,07F7F7F7FH ; <(C7+A7)/2 (C6+A6)/2 (C5+A5)/2 (C4+A4)/2>clean mov dl,[edi+8] ; <(C3+L3)/2 (C2+L2)/2 junk 2L0/2> add esi,ebx ; <(C7+2R7+A7)/2 ... (2C5+R5+A5)/2 ...> mov ebx,[edi+28] ; <A23 A22 A21 A20> ... <B53 B52 B51 B50> shr esi,1 ; <(C7+2R7+A7)/4 ... (2C5+R5+A5)/4 ...> dirty add ebx,ecx ; <C3+A3 C2+A2 C1+A1 C0+A0> shr ebx,1 ; <(C3+A3)/2 (C2+A2)/2 (C1+A1)/2 (C0+A0)/2>dirty and esi,07F7F7F7FH ; <(C7+2R7+A7)/4 ... (2C5+R5+A5)/4 ...> clean and ebx,07F7F7F7FH ; <(C3+A3)/2 (C2+A2)/2 (C1+A1)/2 (C0+A0)/2>clean mov dh,[edi+9] ; <(C3+L3)/2 (C2+L2)/2 2L1/2 2L0/2> add ebx,edx ; <(2C3+L3+A3)/2 ... (C1+2L1+A1)/2 ...> lea eax,[eax+esi+001010101H]; <(5C7+2R7+A7+4)/4 ... (6C5+R5+A5+4)/4 ...> shr ebx,1 ; <(2C3+L3+A3)/4 ... (C1+2L1+A1)/4 ...> dirty add ecx,001010101H ; <C3+1 C2+1 C1+1 C0+1> shr eax,1 ; <(5C7+2R7+A7+4)/8 ... (6C5+R5+A5+4)/8...>dirty and ebx,07F7F7F7FH ; <(2C3+L3+A3)/4 ... (C1+2L1+A1)/4 ...> clean add ebx,ecx ; <(6C3+L3+A3+4)/4 ... (5C1+2L1+A1+4)/4 ...> and eax,07F7F7F7FH ; <(5C7+2R7+A7+4)/8 ... (6C5+R5+A5+4)/8...>clean shr ebx,1 ; <(6C3+L3+A3+4)/8 ... (5C1+2L1+A1+4)/8...>dirty mov [edi+4],eax ; Store OBMC pred for pels 4-7 of line 2 thru 5. and ebx,07F7F7F7FH ; <(6C3+L3+A3+4)/8 ... (5C1+2L1+A1+4)/8...>clean mov [edi],ebx ; Store OBMC pred for pels 0-3 of line 2 thru 5. add edi,PITCH ; Advance to next line. add ebp,040000000H jnc @b
mov esi,MBActionCursor cmp ebp,T_MacroBlockActionDescr.BlkY2 jle Block1or2DescrBuilt
mov al,PB [esi].CodedBlocks mov edi,CoeffStream ; Restore block descriptor ptr. cmp ebp,T_MacroBlockActionDescr.BlkY3 je Block3DescrBuilt
Block4DescrBuilt:
shr al,5 ; Check if block 5 (U) empty. lea edx,[esi].BlkU+4 ; Addr of block addr (plus 4). sbb ebp,ebp ; -1 iff block not empty. mov [edi].BlockAddr,edx ; Store address of block address. shr al,1 ; Check if block 6 (Y) empty. lea edx,[esi].BlkV+4 ; Addr of block addr (plus 4). sbb ebx,ebx ; -1 iff block not empty. and ebp,T_CoeffBlk ; 0 iff block empty, else inc. and ebx,T_CoeffBlk ; 0 iff block empty, else inc. add esi,SIZEOF T_MacroBlockActionDescr ; Move to next macroblock descriptor. mov [edi+ebp*1].BlockAddr,edx ; Store address of block address. add edi,ebp ; Inc block descr ptr if blk non-empty. add edi,ebx ; Inc block descr ptr if blk non-empty. xor ebp,ebp and al,1 ; Are we at end-of-stream? je NextMacroBlock_OBMC
sub edi,SIZEOF T_CoeffBlk jmp BlockActionStreamBuilt
;; partial end of section only defined when H261 not defined.ENDIF BuildBlockActionDescr MACRO BlockNumber,AddrOffset shr al,1 ; Check if block empty. lea edi,[edi+ebp] ; Adjust BlockActionDescr cursor. sbb ebp,ebp ; -1 iff block not empty. lea edx,[esi].Blk[BlockNumber*SIZEOF T_Blk]+AddrOffset ; Addr of block addr. and ebp,T_CoeffBlk ; 0 iff block empty, else inc. mov [edi].BlockAddr,edx ; Store address of block address.ENDM
IFNDEF H261;; more code only used when H261 not defined
MBIsIntraCoded_OBMC:
shr al,1 ; Same as BuildBlockActionDescr macro, except don't inc edi. sbb ebp,ebp lea edx,[esi].BlkY1 and ebp,T_CoeffBlk mov [edi].BlockAddr,edx BuildBlockActionDescr 1,0 ; If blk 2 non-empty, record BAD to do as intra. BuildBlockActionDescr 2,0 ; blk 3 BuildBlockActionDescr 3,0 ; blk 4 BuildBlockActionDescr 4,0 ; blk 5 BuildBlockActionDescr 5,0 ; blk 6
add esi,SIZEOF T_MacroBlockActionDescr ; Move to next descriptor add edi,ebp test al,1 ; Are we at end-of-stream? je NextMacroBlock_OBMC
sub edi,SIZEOF T_CoeffBlk jmp BlockActionStreamBuilt
;; end of section only defined when H261 not defined.ENDIF;===============================================================================;===============================================================================; First pass builds block action stream from macroblock action stream.;===============================================================================;===============================================================================
; esi -- MacroBlockActionStream cursor; edi -- BlockActionStream cursor; ebp -- Increment for BlockActionStream cursor; edx -- Address of a block to do; al -- Coded block pattern for I or P block; bl -- BlockType
NextMacroBlock:
mov bl,PB [esi].BlockType mov al,PB [esi].CodedBlocks ; Bits 0- 3 set for non-empty Y blks. ; Bit 4 set for non-empty U blk. ; Bit 5 set for non-empty V blk. ; Bit 6 clear except at stream end. ; Bit 7 clear. Unused. and bl,IsINTRA jne MBIsIntraCoded
BuildBlockActionDescr 0,4 ; If blk 1 non-empty, record BAD to do as inter. BuildBlockActionDescr 1,4 ; blk 2 BuildBlockActionDescr 2,4 ; blk 3 BuildBlockActionDescr 3,4 ; blk 4 BuildBlockActionDescr 4,4 ; blk 5 BuildBlockActionDescr 5,4 ; blk 6 add esi,SIZEOF T_MacroBlockActionDescr ; Move to next descriptor and al,1 ; Are we at end-of-stream? je NextMacroBlock
add edi,ebp sub edi,SIZEOF T_CoeffBlk jmp BlockActionStreamBuilt MBIsIntraCoded:
BuildBlockActionDescr 0,0 ; If blk 1 non-empty, record BAD to do as intra. BuildBlockActionDescr 1,0 ; blk 2 BuildBlockActionDescr 2,0 ; blk 3 BuildBlockActionDescr 3,0 ; blk 4 BuildBlockActionDescr 4,0 ; blk 5 BuildBlockActionDescr 5,0 ; blk 6
add esi,SIZEOF T_MacroBlockActionDescr ; Move to next descriptor and al,1 ; Are we at end-of-stream? je NextMacroBlock
add edi,ebp sub edi,SIZEOF T_CoeffBlk jmp BlockActionStreamBuilt
NextBMacroBlock:
; esi -- MacroBlockActionStream cursor; edi -- BlockActionStream cursor; ebp -- Increment for BlockActionStream cursor; edx -- Address of a block to do; cl -- Used to compute defined columns mask case.; bh -- Coded block pattern for B block; bl -- Coded block pattern for I or P block; al -- Used to compute defined rows mask.
BuildBBlockActionDescr MACRO BlkNum,LinesDefFutureFrame,ColsDefFutureFrame shr bh,1 ; Check if block empty. mov cl,[esi].Blk[BlkNum*SIZEOF T_Blk].BestHMVb ; HMVb for block. lea edi,[edi+ebp] ; Adjust BlockActionDescr. mov al,[esi].Blk[BlkNum*SIZEOF T_Blk].BestVMVb ; VMVb for block. sbb ebp,ebp ; -1 iff block not empty. mov cl,ColsDefFutureFrame[ecx-96] ; Case of columns to do bidi. and ebp,T_CoeffBlk ; 0 iff block empty, else inc. mov al,LinesDefFutureFrame[eax-96] ; Mask for lines to do bidi. mov [edi].LinesDefined,al ; Stash it. mov edx,ColsDefined[ecx] mov [edi].Cols03Defined,edx ; Stash it. mov edx,ColsDefined[ecx+4] mov [edi].Cols47Defined,edx ; Stash it. lea edx,[esi].Blk[BlkNum*SIZEOF T_Blk]+8 ; Addr of block addr. mov [edi].BlockAddr,edx ; Store address of blk address.ENDM
mov ebx,PD [esi].CodedBlocks ; Bits 0- 3 set for non-empty Y blks. ; Bit 4 set for non-empty U blk. ; Bit 5 set for non-empty V blk. ; Bit 6 clear except at stream end. ; Bit 7 clear. Unused. ; Bits 8-13 like bits 0-5, but for B frame. ; Bit 14-15 clear. Unused.
BuildBBlockActionDescr 0, UpperYBlkLinesDef, LeftYBlkColsDef BuildBBlockActionDescr 1, UpperYBlkLinesDef, RightYBlkColsDef BuildBBlockActionDescr 2, LowerYBlkLinesDef, LeftYBlkColsDef BuildBBlockActionDescr 3, LowerYBlkLinesDef, RightYBlkColsDef BuildBBlockActionDescr 4, ChromaLinesDef, ChromaColsDef BuildBBlockActionDescr 5, ChromaLinesDef, ChromaColsDef add esi,SIZEOF T_MacroBlockActionDescr ; Move to next descriptor and bl,040H ; Are we at end-of-stream? je NextBMacroBlock
add edi,ebp sub edi,SIZEOF T_CoeffBlk BlockActionStreamBuilt:
mov CoeffStream,edi ; Stash address of last block of coeffs.
NextBlock:
;===============================================================================;===============================================================================; Second pass performs frame differencing of Inters and Forward DCT.;===============================================================================;===============================================================================
mov eax,[edi].BlockAddr ; Fetch address of block to do mov ebp,PITCH test eax,4 ; Is it an Inter block. jne InterOrOBMCBlock ; Jump if doing inter block.
mov edx,[eax].T_Blk.BlkOffset ; BlkOffset if INTRA; BestMVs if BiDi. mov ecx,TargetFrameBaseAddress add ecx,edx ; Target block address if INTRA mov esi,[eax-8].T_Blk.BlkOffset ; Addr of BlkOffset if BiDi
IFNDEF H261;; H261 does not execute the BiDi code so it is included only when H261 is not defined;; test eax,8 ; Is it a BiDi block? jne BiDiBlock ; Jump if doing BiDi block.ENDIF
IntraBlock:
; Register usage:; ecx,edi -- Address of block.; ebp -- Pitch.; ebx, eax -- Scratch.
mov ebx,[ecx] mov eax,[ecx+4] mov P00,ebx mov P04,eax mov eax,[ecx+ebp*1] mov edx,[ecx+ebp*1+4] lea edi,[ecx+PITCH*5] lea ecx,[ecx+ebp*2] mov P10,eax mov P14,edx mov eax,[ecx] mov edx,[ecx+4] mov P20,eax mov P24,edx mov eax,[ecx+ebp*1] mov edx,[ecx+ebp*1+4] mov P30,eax mov P34,edx mov eax,[ecx+ebp*2] mov edx,[ecx+ebp*2+4] mov P40,eax mov P44,edx mov eax,[edi] mov edx,[edi+4] mov P50,eax mov P54,edx mov eax,[edi+ebp*1] mov edx,[edi+ebp*1+4] mov P60,eax mov P64,edx mov eax,[edi+ebp*2] mov edx,[edi+ebp*2+4] mov P74,edx xor ecx,ecx and ebx,00000007FH ; Fetch P0. mov cl,P03 ; Fetch P3. mov P70,eax jmp DoForwardDCT
IFNDEF H261;; H261 does not execute the BiDi code so it is included only when H261 is not defined;;
BiDiBlock:
mov BlkActionDescrAddr,eax ; Extract VMVb. mov ebp,FutureFrameBaseAddress shr edx,25 ; CF == 1 iff VMVb is half pel. mov bl,[edi].LinesDefined lea esi,[esi+ebp-48] ; Addr 0-MV blk in Future P Frame. mov ebp,[edi].Cols47DefinedIF PITCH-384**** Magic leaks out if pitch not equal to 384ENDIF lea ecx,[edx+edx*2-48*3] ; Mult integer pel VMVb by PITCH. mov edi,[edi].Cols03Defined mov dl,[eax-8].T_Blk.BestHMVb ; Fetch HMVb. jc InterpVert_FuturePFrame
shl ecx,7 shr dl,1 ; CF == 1 iff HMVb is half pel. mov bh,bl lea esi,[esi+ecx] ; Add VMVb contrib to block addr. jc InterpHorz_FuturePFrame
add esi,edx ; Add HMVb contrib to block addr.
; esi -- Future P Frame block address.; edi -- Mask to apply to columns 0-3 of block to select columns in range.; ebp -- Mask to apply to columns 4-7 of block to select columns in range.; bl -- Mask of lines that are in range.
@@:
xor esp,4 add bl,bl ; 0A CF == 1 iff line 0 in range. sbb eax,eax ; 0B eax == -1 if line 0 in range. mov ecx,[esi] ; 0C Fetch Future P00:P03. and eax,edi ; 0D In range among P00,P01,P02,P03. add bl,bl ; 1A sbb edx,edx ; 1B mov Mask00+4,eax ; 0E Stash Mask for use with past pred. and eax,ecx ; 0F Select in-range pels. mov ecx,[esi+PITCH*1] ; 1C mov P00+4,eax ; 0G Stash in-range pels. and edx,edi ; 1D mov Mask10+4,edx ; 1E add bl,bl ; 2A sbb eax,eax ; 2B and edx,ecx ; 1F mov P10+4,edx ; 1G mov ecx,[esi+PITCH*2] ; 2C and eax,edi ; 2D add bl,bl ; 3A sbb edx,edx ; 3B mov Mask20+4,eax ; 2E and eax,ecx ; 2F mov ecx,[esi+PITCH*3] ; 3C mov P20+4,eax ; 2G and edx,edi ; 3D mov Mask30+4,edx ; 3E add bl,bl ; 4A sbb eax,eax ; 4B and edx,ecx ; 3F mov P30+4,edx ; 3G mov ecx,[esi+PITCH*4] ; 4C and eax,edi ; 4D add bl,bl ; 5A sbb edx,edx ; 5B mov Mask40+4,eax ; 4E and eax,ecx ; 4F mov ecx,[esi+PITCH*5] ; 5C mov P40+4,eax ; 4G and edx,edi ; 5D mov Mask50+4,edx ; 5E add bl,bl ; 6A sbb eax,eax ; 6B and edx,ecx ; 5F mov P50+4,edx ; 5G mov ecx,[esi+PITCH*6] ; 6C and eax,edi ; 6D add bl,bl ; 7A sbb edx,edx ; 7B mov Mask60+4,eax ; 6E and eax,ecx ; 6F mov ecx,[esi+PITCH*7] ; 7C mov P60+4,eax ; 6G and edx,edi ; 7D mov Mask70+4,edx ; 7E and edx,ecx ; 7F mov P70+4,edx ; 7G mov edi,ebp mov edx,BlkActionDescrAddr add esi,4 mov ecx,4 mov bl,bh and ecx,esp je @b
mov edi,[edx-8].T_Blk.BlkOffset xor eax,eax mov al,[edx-8].T_Blk.BestVMVf jmp BiDiFuturePredDone
InterpVert_FuturePFrame:
shl ecx,7 shr dl,1 ; CF == 1 iff HMVb is half pel. mov bh,bl lea esi,[esi+ecx] ; Add VMVb contrib to block addr. jc InterpBoth_FuturePFrame
add esi,edx ; Add HMVb contrib to block addr.
; esi -- Future P Frame block address.; edi -- Mask to apply to columns 0-3 of block to select columns in range.; ebp -- Mask to apply to columns 4-7 of block to select columns in range.; bl -- Mask of lines that are in range.
; Interpolate Future Prediction Vertically.
@@:
xor esp,4 add bl,bl ; 0A CF == 1 iff line 0 in range. sbb eax,eax ; 0B eax == -1 if line 0 in range. mov ecx,[esi] ; 0C Fetch Future P00:P03. and eax,edi ; 0D In range among P00,P01,P02,P03. mov edx,[esi+PITCH*1] ; 0E Fetch Future P10:P13. mov Mask00+4,eax ; 0F Stash Mask for use with past pred. add ecx,edx ; 0G Add P00:P03 and P10:P13. add ecx,001010101H ; 0H Add rounding. shr ecx,1 ; 0I Interpolate (divide by 2). add bl,bl ; 1A sbb edx,edx ; 1B and eax,ecx ; 0J Select in-range pels (and clean). mov P00+4,eax ; 0K Stash in-range pels. mov ecx,[esi+PITCH*1] ; 1C and edx,edi ; 1D mov eax,[esi+PITCH*2] ; 1E mov Mask10+4,edx ; 1F add ecx,eax ; 1G add ecx,001010101H ; 1H shr ecx,1 ; 1I add bl,bl ; 2A sbb eax,eax ; 2B and edx,ecx ; 1J mov P10+4,edx ; 1K mov ecx,[esi+PITCH*2] ; 2C and eax,edi ; 2D mov edx,[esi+PITCH*3] ; 2E mov Mask20+4,eax ; 2F add ecx,edx ; 2G add ecx,001010101H ; 2H shr ecx,1 ; 2I add bl,bl ; 3A sbb edx,edx ; 3B and eax,ecx ; 2J mov P20+4,eax ; 2K mov ecx,[esi+PITCH*3] ; 3C and edx,edi ; 3D mov eax,[esi+PITCH*4] ; 3E mov Mask30+4,edx ; 3F add ecx,eax ; 3G add ecx,001010101H ; 3H shr ecx,1 ; 3I add bl,bl ; 4A sbb eax,eax ; 4B and edx,ecx ; 3J mov P30+4,edx ; 3K mov ecx,[esi+PITCH*4] ; 4C and eax,edi ; 4D mov edx,[esi+PITCH*5] ; 4E mov Mask40+4,eax ; 4F add ecx,edx ; 4G add ecx,001010101H ; 4H shr ecx,1 ; 4I add bl,bl ; 5A sbb edx,edx ; 5B and eax,ecx ; 4J mov P40+4,eax ; 4K mov ecx,[esi+PITCH*5] ; 5C and edx,edi ; 5D mov eax,[esi+PITCH*6] ; 5E mov Mask50+4,edx ; 5F add ecx,eax ; 5G add ecx,001010101H ; 5H shr ecx,1 ; 5I add bl,bl ; 6A sbb eax,eax ; 6B and edx,ecx ; 5J mov P50+4,edx ; 5K mov ecx,[esi+PITCH*6] ; 6C and eax,edi ; 6D mov edx,[esi+PITCH*7] ; 6E mov Mask60+4,eax ; 6F add ecx,edx ; 6G add ecx,001010101H ; 6H add esi,4 shr ecx,1 ; 6I add bl,bl ; 7A sbb edx,edx ; 7B and eax,ecx ; 6J mov P60+4,eax ; 6K mov ecx,[esi+PITCH*7-4] ; 7C and edx,edi ; 7D mov eax,[esi+PITCH*8-4] ; 7E mov Mask70+4,edx ; 7F add ecx,eax ; 7G add ecx,001010101H ; 7H mov bl,bh shr ecx,1 ; 7I and edx,ecx ; 7J mov P70+4,edx ; 7K mov edi,ebp mov edx,BlkActionDescrAddr mov ecx,4 and ecx,esp je @b
mov edi,[edx-8].T_Blk.BlkOffset xor eax,eax mov al,[edx-8].T_Blk.BestVMVf jmp BiDiFuturePredDone
InterpHorz_FuturePFrame:
; esi -- Future P Frame block address.; edi -- Mask to apply to columns 0-3 of block to select columns in range.; ebp -- Mask to apply to columns 4-7 of block to select columns in range.; bl -- Mask of lines that are in range.
; Interpolate Future Prediction Horizontally.
add esi,edx ; Add HMVb contrib to block addr.
@@:
xor esp,4 add bl,bl ; 0A CF == 1 iff line 0 in range. sbb eax,eax ; 0B eax == -1 if line 0 in range. mov ecx,[esi] ; 0C Fetch Future P00:P03. and eax,edi ; 0D In range among P00,P01,P02,P03. mov edx,[esi+1] ; 0E Fetch Future P01:P04. mov Mask00+4,eax ; 0F Stash Mask for use with past pred. add ecx,edx ; 0G Add P00:P03 and P01:P04. add ecx,001010101H ; 0H Add rounding. shr ecx,1 ; 0I Interpolate (divide by 2). add bl,bl ; 1A sbb edx,edx ; 1B and eax,ecx ; 0J Select in-range pels (and clean). mov P00+4,eax ; 0K Stash in-range pels. mov ecx,[esi+PITCH*1] ; 1C and edx,edi ; 1D mov eax,[esi+PITCH*1+1] ; 1E mov Mask10+4,edx ; 1F add ecx,eax ; 1G add ecx,001010101H ; 1H shr ecx,1 ; 1I add bl,bl ; 2A sbb eax,eax ; 2B and edx,ecx ; 1J mov P10+4,edx ; 1K mov ecx,[esi+PITCH*2] ; 2C and eax,edi ; 2D mov edx,[esi+PITCH*2+1] ; 2E mov Mask20+4,eax ; 2F add ecx,edx ; 2G add ecx,001010101H ; 2H shr ecx,1 ; 2I add bl,bl ; 3A sbb edx,edx ; 3B and eax,ecx ; 2J mov P20+4,eax ; 2K mov ecx,[esi+PITCH*3] ; 3C and edx,edi ; 3D mov eax,[esi+PITCH*3+1] ; 3E mov Mask30+4,edx ; 3F add ecx,eax ; 3G add ecx,001010101H ; 3H shr ecx,1 ; 3I add bl,bl ; 4A sbb eax,eax ; 4B and edx,ecx ; 3J mov P30+4,edx ; 3K mov ecx,[esi+PITCH*4] ; 4C and eax,edi ; 4D mov edx,[esi+PITCH*4+1] ; 4E mov Mask40+4,eax ; 4F add ecx,edx ; 4G add ecx,001010101H ; 4H shr ecx,1 ; 4I add bl,bl ; 5A sbb edx,edx ; 5B and eax,ecx ; 4J mov P40+4,eax ; 4K mov ecx,[esi+PITCH*5] ; 5C and edx,edi ; 5D mov eax,[esi+PITCH*5+1] ; 5E mov Mask50+4,edx ; 5F add ecx,eax ; 5G add ecx,001010101H ; 5H shr ecx,1 ; 5I add bl,bl ; 6A sbb eax,eax ; 6B and edx,ecx ; 5J mov P50+4,edx ; 5K mov ecx,[esi+PITCH*6] ; 6C and eax,edi ; 6D mov edx,[esi+PITCH*6+1] ; 6E mov Mask60+4,eax ; 6F add ecx,edx ; 6G add ecx,001010101H ; 6H add esi,4 shr ecx,1 ; 6I add bl,bl ; 7A sbb edx,edx ; 7B and eax,ecx ; 6J mov P60+4,eax ; 6K mov ecx,[esi+PITCH*7-4] ; 7C and edx,edi ; 7D mov eax,[esi+PITCH*7+1-4] ; 7E mov Mask70+4,edx ; 7F add ecx,eax ; 7G add ecx,001010101H ; 7H mov bl,bh shr ecx,1 ; 7I and edx,ecx ; 7J mov P70+4,edx ; 7K mov edi,ebp mov edx,BlkActionDescrAddr mov ecx,4 and ecx,esp je @b
mov edi,[edx-8].T_Blk.BlkOffset xor eax,eax mov al,[edx-8].T_Blk.BestVMVf jmp BiDiFuturePredDone
InterpBoth_FuturePFrame:
add esi,edx ; Add HMVb contrib to block addr. sub esp,68
; esi -- Future P Frame block address.; edi -- Mask to apply to columns 0-3 of block to select columns in range.; ebp -- Mask to apply to columns 4-7 of block to select columns in range.; bl -- Mask of lines that are in range.
; Interpolate Future Prediction Vertically.
@@:
add esp,8 mov eax,[esi] ; Fetch Future P00:P03. mov ecx,001010101H ; Mask to extract halves. mov edx,[esi+1] ; Fetch Future P01:P04. add eax,edx ; <P04+P03 ...>. mov edx,[esi+PITCH+1] ; Fetch Future P11:P14. and ecx,eax ; <(P04+P03)&1 ...>. add esi,PITCH ; Advance to next line. xor eax,ecx ; <(P04+P03)/2*2 ...>. add edx,ecx ; <P14+((P04+P03)&1) ...>. shr eax,1 ; <(P04+P03)/2 ...>. mov ecx,[esi] ; Fetch Future P10:P13. add edx,ecx ; <P14+P13+((P04+P03)&1) ...>. add eax,001010101H ; <(P04+P03)/2+1 ...> shr edx,1 ; <(P14+P13+((P04+P03)&1))/2 ...> (dirty). add bl,bl ; CF == 1 iff line 0 in range. sbb ecx,ecx ; ecx == -1 if line 0 in range. and edx,07F7F7F7FH ; <(P14+P13+((P04+P03)&1))/2 ...> (clean). add eax,edx ; <(P04+P03)/2+1+(P14+P13+((P04+P03)&1))/2 ...>. and ecx,edi ; In range among P00,P01,P02,P03. shr eax,1 ; <((P04+P03)/2+1+(P14+P13+((P04+P03)&1))/2)/2 ...>. mov Mask00+60,ecx ; Stash Mask for use with past prediction. and eax,ecx ; Select in-range pels from future pred (and clean). test esp,000000038H mov P00+60,eax ; Stash in-range pels. jne @b
sub esi,PITCH*8-4 ; Move to right 4 columns. mov edx,BlkActionDescrAddr mov edi,ebp sub esp,60 mov ecx,4 mov bl,bh and ecx,esp jne @b
add esp,60 xor eax,eax mov edi,[edx-8].T_Blk.BlkOffset mov al,[edx-8].T_Blk.BestVMVf
BiDiFuturePredDone:
shr al,1 ; CF == 1 iff VMVf is half pel. mov esi,TargetFrameBaseAddress mov cl,[edx-8].T_Blk.BestHMVf mov edx,DistFromTargetToPastP lea edi,[edi+esi] jc InterpVert_PastPFrame
shr cl,1 ; CF == 1 iff HMVf is half pel. lea eax,[eax+eax*2-48*3] ; Mult integer pel VMVf by PITCH. lea esi,[edi+edx-48] ; Addr 0-MV blk in Future P Frame. jc InterpHorz_PastPFrame
shl eax,7 add esi,ecx ; Add HMVf contrib to block addr. add esi,eax ; Add VMVf contrib to block addr. sub esp,64
; esi -- Past P Frame block address.; edi -- Target block address.
@@:
mov eax,[esi] ; 0A Fetch past prediction. mov ebx,Mask00+64 ; 0B Fetch bidi-prediction mask. mov ecx,P00+64 ; 0C Fetch future pred for bidi predicted pels. and ebx,eax ; 0D Extract past for bidi predicted pels. mov edx,[esi+4] ; 4A mov ebp,Mask04+64 ; 4B lea eax,[ecx+eax*2] ; 0E (2*Past) or ((2*Past+Future) for each pel. mov ecx,P04+64 ; 4C sub eax,ebx ; 0F (2*Past) or (Past+Future) for each pel. and ebp,edx ; 4D shr eax,1 ; 0G (Past) or ((Past+Future)/2) (dirty). lea edx,[ecx+edx*2] ; 4E and eax,07F7F7F7FH ; 0H (Past) or ((Past+Future)/2) (clean). sub edx,ebp ; 4F shr edx,1 ; 4G mov ebx,[edi] ; 0I Fetch target pels. and edx,07F7F7F7FH ; 4H mov ebp,[edi+4] ; 4I sub ebx,eax ; 0J Compute correction. sub ebp,edx ; 4J add ebx,080808080H ; 0K Bias correction. add ebp,080808080H ; 4K mov P00+64,ebx ; 0K Store correction. mov P04+64,ebp ; 4K add esi,PITCH add esp,8 test esp,000000038H lea edi,[edi+PITCH] jne @b xor ebx,ebx xor ecx,ecx mov bl,P00 ; Fetch P0. mov cl,P03 ; Fetch P3. jmp DoForwardDCT
InterpVert_PastPFrame:
shr cl,1 ; CF == 1 iff HMVf is half pel. lea eax,[eax+eax*2-48*3] ; Mult integer pel VMVf by PITCH. lea esi,[edi+edx-48] ; Addr 0-MV blk in Future P Frame. jc InterpBoth_PastPFrame
shl eax,7 add esi,ecx ; Add HMVf contrib to block addr. add esi,eax ; Add VMVf contrib to block addr. sub esp,64
; esi -- Past P Frame block address.; edi -- Target block address.
@@:
mov eax,[esi] ; 0A Fetch past prediction. mov edx,[esi+4] ; 4A add eax,[esi+PITCH] ; 0B Add past prediction with which to interpolate. add edx,[esi+PITCH+4] ; 4B add eax,001010101H ; 0C Add rounding. add edx,001010101H ; 0C shr eax,1 ; 0D Divide by two (dirty). and edx,0FEFEFEFEH ; 1E shr edx,1 ; 1D Clean. and eax,07F7F7F7FH ; 0E mov ebx,Mask00+64 ; 0F Fetch bidi-prediction mask. mov ecx,P00+64 ; 0G Fetch future pred for bidi predicted pels. and ebx,eax ; 0H Extract past for bidi predicted pels. mov ebp,Mask04+64 ; 4F lea eax,[ecx+eax*2] ; 0I (2*Past) or ((2*Past+Future) for each pel. mov ecx,P04+64 ; 4G sub eax,ebx ; 0J (2*Past) or (Past+Future) for each pel. and ebp,edx ; 4H shr eax,1 ; 0K (Past) or ((Past+Future)/2) (dirty). lea edx,[ecx+edx*2] ; 4I and eax,07F7F7F7FH ; 0L (Past) or ((Past+Future)/2) (clean). sub edx,ebp ; 4J shr edx,1 ; 4K mov ebx,[edi] ; 0M Fetch target pels. and edx,07F7F7F7FH ; 4L mov ebp,[edi+4] ; 4M sub ebx,eax ; 0N Compute correction. sub ebp,edx ; 4N add ebx,080808080H ; 0O Bias correction. add ebp,080808080H ; 4O mov P00+64,ebx ; 0P Store correction. mov P04+64,ebp ; 4P add esi,PITCH add esp,8 test esp,000000038H lea edi,[edi+PITCH] jne @b xor ebx,ebx xor ecx,ecx mov bl,P00 ; Fetch P0. mov cl,P03 ; Fetch P3. jmp DoForwardDCT
InterpHorz_PastPFrame:
shl eax,7 add esi,ecx ; Add HMVf contrib to block addr. add esi,eax ; Add VMVf contrib to block addr. sub esp,64
; esi -- Past P Frame block address.; edi -- Target block address.
@@:
mov eax,[esi] ; 0A Fetch past prediction. mov edx,[esi+4] ; 4A add eax,[esi+1] ; 0B Add past prediction with which to interpolate. add edx,[esi+5] ; 4B add eax,001010101H ; 0C Add rounding. add edx,001010101H ; 0C shr eax,1 ; 0D Divide by two (dirty). and edx,0FEFEFEFEH ; 1E shr edx,1 ; 1D Clean. and eax,07F7F7F7FH ; 0E mov ebx,Mask00+64 ; 0F Fetch bidi-prediction mask. mov ecx,P00+64 ; 0G Fetch future pred for bidi predicted pels. and ebx,eax ; 0H Extract past for bidi predicted pels. mov ebp,Mask04+64 ; 4F lea eax,[ecx+eax*2] ; 0I (2*Past) or ((2*Past+Future) for each pel. mov ecx,P04+64 ; 4G sub eax,ebx ; 0J (2*Past) or (Past+Future) for each pel. and ebp,edx ; 4H shr eax,1 ; 0K (Past) or ((Past+Future)/2) (dirty). lea edx,[ecx+edx*2] ; 4I and eax,07F7F7F7FH ; 0L (Past) or ((Past+Future)/2) (clean). sub edx,ebp ; 4J shr edx,1 ; 4K mov ebx,[edi] ; 0M Fetch target pels. and edx,07F7F7F7FH ; 4L mov ebp,[edi+4] ; 4M sub ebx,eax ; 0N Compute correction. sub ebp,edx ; 4N add ebx,080808080H ; 0O Bias correction. add ebp,080808080H ; 4O mov P00+64,ebx ; 0P Store correction. mov P04+64,ebp ; 4P add esi,PITCH add esp,8 test esp,000000038H lea edi,[edi+PITCH] jne @b xor ebx,ebx xor ecx,ecx mov bl,P00 ; Fetch P0. mov cl,P03 ; Fetch P3. jmp DoForwardDCT
InterpBoth_PastPFrame:
shl eax,7 add esi,ecx ; Add HMVf contrib to block addr. add esi,eax ; Add VMVf contrib to block addr. sub esp,64
; esi -- Past P Frame block address.; edi -- Target block address.
@@:
mov eax,[esi+1] ; 0A <P04 P03 P02 P01> prediction pels. mov ebx,001010101H ; 0B Mask for extraction of halves. mov ebp,[esi+PITCH+1] ; 0C <P14 P13 P12 P11>. mov ecx,[esi] ; 0D <P03 P02 P01 P00>. add eax,ecx ; 0E <P04+P03 P03+P02 P02+P01 P01+P00>. mov ecx,[esi+PITCH] ; 0F <P13 P12 P11 P10>. and ebx,eax ; 0G <(P04+P03)&1 ...>. and eax,0FEFEFEFEH ; 0H Pre-Clean shr eax,1 ; 0I <(P04+P03)/2 ...>. add ecx,ebp ; 0J <P14+P13 P13+P12 P12+P11 P11+P10>. add eax,001010101H ; 0K <(P04+P03)/2+1 ...>. add ecx,ebx ; 0L <P14+P13+((P04+P03)&1) ...>. shr ecx,1 ; 0M <(P14+P13+((P04+P03)&1))/2 ...> (dirty). mov edx,[esi+5] ; 4A and ecx,07F7F7F7FH ; 0M <(P14+P13+((P04+P03)&1))/2 ...> (clean). mov ebx,001010101H ; 4B add eax,ecx ; 0N <(P04+P03)/2+1+(P14+P13+((P04+P03)&1))/2 ...>. mov ebp,[esi+PITCH+5] ; 4C shr eax,1 ; 0O <((P04+P03)/2+1+(P14+P13+((P04+P03)&1))/2)/2 ...> mov ecx,[esi+4] ; 4D and eax,07F7F7F7FH ; 0P Interpolated prediction. add edx,ecx ; 4E mov ecx,[esi+PITCH+4] ; 4F and ebx,edx ; 4G and edx,0FEFEFEFEH ; 4H add ecx,ebp ; 4J shr edx,1 ; 4I add ecx,ebx ; 4L shr ecx,1 ; 4M add edx,001010101H ; 4K and ecx,07F7F7F7FH ; 4M mov ebx,Mask00+64 ; 0Q Fetch bidi-prediction mask. add edx,ecx ; 4N mov ecx,P00+64 ; 0R Fetch future pred for bidi predicted pels. shr edx,1 ; 4O and ebx,eax ; 0S Extract past for bidi predicted pels. and edx,07F7F7F7FH ; 4P mov ebp,Mask04+64 ; 4Q lea eax,[ecx+eax*2] ; 0T (2*Past) or ((2*Past+Future) for each pel. mov ecx,P04+64 ; 4R sub eax,ebx ; 0U (2*Past) or (Past+Future) for each pel. and ebp,edx ; 4S shr eax,1 ; 0V (Past) or ((Past+Future)/2) (dirty). lea edx,[ecx+edx*2] ; 4T and eax,07F7F7F7FH ; 0W (Past) or ((Past+Future)/2) (clean). sub edx,ebp ; 4U shr edx,1 ; 4V mov ebx,[edi] ; 0X Fetch target pels. and edx,07F7F7F7FH ; 4W mov ebp,[edi+4] ; 4X sub ebx,eax ; 0Y Compute correction. sub ebp,edx ; 4Y add ebx,080808080H ; 0Z Bias correction. add ebp,080808080H ; 4Z mov P00+64,ebx ; 0a Store correction. mov P04+64,ebp ; 4a add esi,PITCH add esp,8 test esp,000000038H lea edi,[edi+PITCH] jne @b xor ebx,ebx xor ecx,ecx mov bl,P00 ; Fetch P0. mov cl,P03 ; Fetch P3. jmp DoForwardDCT
;; end of section of code not define when H261 definedENDIF
InterOrOBMCBlock:
mov esi,TargetFrameBaseAddress mov edi,[eax-4].T_Blk.BlkOffset ; Compute Addr of Target block.
IFNDEF H261;; H261 does not execute the OBMC code so it is included only when H261 is not defined;; test eax,8 jne OBMCBlockENDIF
add edi,esi mov esi,[eax-4].T_Blk.PastRef ; Addr of PrevRef block. mov eax,[eax-4].T_Blk.MVs ; al = Horz MV; ah = Vert MV mov ecx,080808080H
IFNDEF H261;; H261 does not execute Interp code so it is included only when H261 is not defined;; test al,1 jne InterpHorzOrBoth
ENDIF
lea edx,[ebp+ebp*2] lea ebx,[esi+ebp] test ah,1 je NoInterp
IFNDEF H261;; H261 does not execute Interp code so it is included only when H261 is not defined;;
InterpVert:InterpHorz:
; Register usage:; edi -- Address of target block.; esi -- Address of reference block.; ebx -- Address of reference plus either 1 or PITCH, for interpolation.; ebp, edx, ecx, eax -- Scratch.
sub esp,16
@@: add esp,4 mov eax,[esi] ; 0A <P03 P02 P01 P00> prediction pels. mov ecx,[ebx] ; 0B <P04 ...> or <P13 ...> prediction pels. mov edx,[edi] ; 0C <C03 C02 C01 C00> current pels. add edx,080808080H ; 0D Add bias. mov ebp,[esi+PITCH*2] ; 2A lea eax,[eax+ecx+001010101H]; 0E Sum of pred pels to interpolate. mov ecx,[ebx+PITCH*2] ; 2B shr eax,1 ; 0F Average of prediction pels (dirty). and eax,07F7F7F7FH ; 0G Average of prediction pels (clean). lea ebp,[ebp+ecx+001010101H]; 2E sub edx,eax ; 0H Current - interpolated prediction, biased. mov eax,[edi+PITCH*2] ; 2C mov P00+12,edx ; 0I Save correction. add eax,080808080H ; 2D shr ebp,1 ; 2F mov edx,[esi+PITCH*4] ; 4A and ebp,07F7F7F7FH ; 2G mov ecx,[ebx+PITCH*4] ; 4B sub eax,ebp ; 2H mov ebp,[edi+PITCH*4] ; 4C mov P20+12,eax ; 2I lea ecx,[ecx+edx+001010101H]; 4E shr ecx,1 ; 4F add ebp,080808080H ; 4D and ecx,07F7F7F7FH ; 4G mov eax,[esi+PITCH*6] ; 6A sub ebp,ecx ; 4H mov ecx,[ebx+PITCH*6] ; 6B mov P40+12,ebp ; 4I mov ebp,[edi+PITCH*6] ; 6C lea ecx,[ecx+eax+001010101H]; 6E add ebp,080808080H ; 6D shr ecx,1 ; 6F add esi,4 and ecx,07F7F7F7FH ; 6G add ebx,4 sub ebp,ecx ; 6H add edi,4 test esp,4 mov P60+12,ebp ; 6I je @b
add esi,PITCH-8 add edi,PITCH-8 test esp,8 lea ebx,[ebx+PITCH-8] jne @b
xor ebx,ebx xor ecx,ecx mov bl,P00 ; Fetch P0. mov cl,P03 ; Fetch P3. jmp DoForwardDCT
InterpHorzOrBoth:
lea ebx,[esi+1] test ah,1 je InterpHorz
InterpBoth:
; Register usage:; edi -- Address of target block.; esi -- Address of reference block.; ecx -- bias value 0x80808080, to make code size smaller.; ebp -- Pitch and scratch.; edx, ebx, eax -- Scratch.
sub esp,64
@@:
mov eax,[esi+1] ; <P04 P03 P02 P01> prediction pels. lea edx,[ecx*2+1] ; Get 001010101H mask. mov ebx,[esi] ; <P03 P02 P01 P00>. add edi,4 ; Pre-increment target block pointer. add eax,ebx ; <P04+P03 P03+P02 P02+P01 P01+P00>. mov ebx,[esi+ebp*1+1] ; <P14 P13 P12 P11>. and edx,eax ; <(P04+P03)&1 ...>. mov ebp,[esi+ebp*1] ; <P13 P12 P11 P10>. xor eax,edx ; Clear insignificant fractional bit in each byte. add ebx,ebp ; <P14+P13 P13+P12 P12+P11 P11+P10>. shr eax,1 ; <(P04+P03)/2 ...>. add ebx,edx ; <P14+P13+((P04+P03)&1) ...>. shr ebx,1 ; <(P14+P13+((P04+P03)&1))/2 ...> (dirty). add esi,4 ; Advance reference block pointer. and ebx,07F7F7F7FH ; <(P14+P13+((P04+P03)&1))/2 ...> (clean). lea eax,[eax+ecx*2+1] ; <(P04+P03)/2+1 ...>. add eax,ebx ; <(P04+P03)/2+1+(P14+P13+((P04+P03)&1))/2 ...>. mov ebx,[edi-4] ; <C03 C02 C01 C00> current pels. shr eax,1 ; <((P04+P03)/2+1+(P14+P13+((P04+P03)&1))/2)/2 ...>. add ebx,ecx ; Add bias. and eax,07F7F7F7FH ; Interpolated prediction. add esp,4 ; Advance frame difference pointer. sub ebx,eax ; Correction. mov ebp,PITCH ; Reload Pitch. test esp,4 mov P00+60,ebx ; Save correction. je @b
lea esi,[esi+ebp-8] xor ebx,ebx test esp,000000038H lea edi,[edi+ebp-8] jne @b
mov bl,P00 ; Fetch P0. xor ecx,ecx mov cl,P03 ; Fetch P3. jmp DoForwardDCT
OBMCBlock: ; Do OBMC frame differencing. OBMC prediction computed above.
mov ecx,080808080H mov edi,[eax-12].T_Blk.BlkOffset ; Compute Addr of Target block. add edi,esi mov esi,[eax-12].T_Blk.PastRef ; Addr of PrevRef block. lea edx,[ebp+ebp*2] lea ebx,[esi+ebp]
;; end of section of code not included when H261 definedENDIF
NoInterp:
; Register usage:; edi -- Address of target block.; esi -- Address of reference block.; ebp -- Pitch.; edx -- Pitch times 3.; ecx -- bias value 0x80808080, to make code size smaller.; ebx, eax -- Scratch.
@@:
xor esp,4 ; 1st time: Back off to cache line; mov eax,[edi] ; 0A <C3 C2 C1 C0> current pels. add eax,ecx ; 0C Add bias. mov ebx,[esi] ; 0B <P3 P2 P1 P0> prediction pels. sub eax,ebx ; 0D <Cn-Pn> Current - pred, biased. mov ebx,[esi+ebp*1] ; 1B mov P00+4,eax ; 0E Save <Corr3 Corr2 Corr1 Corr0> mov eax,[edi+ebp*1] ; 1A sub eax,ebx ; 1D mov ebx,[esi+ebp*2] ; 2B add eax,ecx ; 1C sub ebx,ecx ; 2C mov P10+4,eax ; 1E mov eax,[edi+ebp*2] ; 2A sub eax,ebx ; 2D mov ebx,[esi+ebp*4] ; 4B mov P20+4,eax ; 2E mov eax,[edi+ebp*4] ; 4A sub eax,ebx ; 4D mov ebx,[esi+edx*1] ; 3B add eax,ecx ; 4C sub ebx,ecx ; 3C mov P40+4,eax ; 4E mov eax,[edi+edx*1] ; 3A sub eax,ebx ; 3D mov ebx,[esi+edx*2] ; 6B mov P30+4,eax ; 3E lea esi,[esi+ebp+4] ; Advance to line 1. mov eax,[edi+edx*2] ; 6A lea edi,[edi+ebp+4] ; Advance to line 1. sub eax,ebx ; 6D mov ebx,[esi+ebp*4-4] ; 5B add eax,ecx ; 6C sub ebx,ecx ; 5C mov P60+4,eax ; 6E mov eax,[edi+ebp*4-4] ; 5A sub eax,ebx ; 5D mov ebx,[esi+edx*2-4] ; 7B mov P50+4,eax ; 5E mov eax,[edi+edx*2-4] ; 7A sub eax,ebx ; 7D sub edi,ebp ; Back off to line 0. add eax,ecx ; 7C sub esi,ebp ; Back off to line 0. test esp,4 ; Do twice. mov P70+4,eax ; 7E je @b
xor ecx,ecx xor ebx,ebx mov bl,P00 ; Fetch P0. mov cl,P03 ; Fetch P3.
DoForwardDCT:
;=============================================================================;; This section does the Forward Discrete Cosine Transform. It performs a DCT; on a 8*8 block of pels or pel differences. The row transforms are done; first using a table lookup method. Then the columns are done, using; computation.;;; Each intermediate and coefficient is a short. There are four fractional; bits. All coefficients except an intrablock's DC are biased by 08000H.
; Perform row transforms.;; Register usage:; ebp - Accumulator for contributions to intermediates I0 (hi) and I2 (lo).; edi - Accumulator for contributions to intermediates I1 (hi) and I3 (lo).; esi - Accumulator for contributions to intermediates I4 (hi) and I6 (lo).; edx - Accumulator for contributions to intermediates I7 (hi) and I5 (lo).; ecx - Pel or pel difference.; ebx - Pel or pel difference.; eax - Place in which to fetch a pel's contribution to two intermediates.
mov esi,PD P80000_P4545F [ebx*8] ; P0's contribution to I4|I6. mov eax,PD P80000_N4545F [ecx*8] ; P3's contribution to I4|I6. mov edx,PD P2350B_P6491A [ebx*8] ; P0's contribution to I7|I5. mov edi,PD NB18A8_P96831 [ecx*8] ; P3's contribution to I7|I5. lea esi,[esi+eax+40004000H] ; P0, P3 contribs to I4|I6, biased. mov eax,PD P80000_NA73D7 [ecx*8] ; P3's contribution to I0|I2. lea edx,[edx+edi+40004000H] ; P0, P3 contribs to I7|I5, biased. mov ebp,PD P80000_PA73D7 [ebx*8] ; P0's contribution to I0|I2. mov edi,PD P2350B_N6491A [ecx*8] ; P3's contribution to I1|I3. mov cl,P01 ; Fetch P1. lea ebp,[ebp+eax+40004000H] ; P0, P3 contribs to I0|I2, biased. mov eax,PD NB18A8_N96831 [ebx*8] ; P0's contribution to I1|I3. sub edi,eax ; P0, P3 contribs to I1|I3, unbiased. mov eax,PD P80000_P4545F [ecx*8] ; P1's contribution to I0|I2. add ebp,eax ; P0, P1, P3 contribs to I0|I2. mov eax,PD N96831_P2350B [ecx*8] ; P1's contribution to I1|I3. sub edi,eax ; P0, P1, P3 contribs to I1|I3, unbiased. mov eax,PD P80000_PA73D7 [ecx*8] ; P1's contribution to I4|I6. sub esi,eax ; P0, P1, P3 contribs to I4|I6. mov bl,P02 ; Fetch P2. mov eax,PD P6491A_PB18A8 [ecx*8] ; P1's contribution to I7|I5. mov cl,P04 ; Fetch P4. sub edx,eax ; P0, P1, P3 contribs to I7|I5. mov eax,PD P80000_N4545F [ebx*8] ; P2's contribution to I0|I2. add ebp,eax ; P0-P3 contribs to I0|I2. mov eax,PD P6491A_NB18A8 [ebx*8] ; P2's contribution to I1|I3. add edi,eax ; P0-P3 contribs to I1|I3, unbiased. mov eax,PD P80000_NA73D7 [ebx*8] ; P2's contribution to I4|I6. sub esi,eax ; P0-P3 contribs to I4|I6. mov eax,PD N96831_N2350B [ebx*8] ; P2's contribution to I7|I5. sub edx,eax ; P0-P3 contribs to I7|I5. mov eax,PD P80000_NA73D7 [ecx*8] ; P4's contribution to I0|I2. add ebp,eax ; P0-P4 contribs to I0|I2. mov eax,PD P2350B_N6491A [ecx*8] ; P4's contribution to I1|I3. sub edi,eax ; P0-P4 contribs to I1|I3, unbiased. mov eax,PD P80000_N4545F [ecx*8] ; P4's contribution to I4|I6. add esi,eax ; P0-P4 contribs to I4|I6. mov bl,P05 ; Fetch P5. mov eax,PD NB18A8_P96831 [ecx*8] ; P4's contribution to I7|I5. mov cl,P06 ; Fetch P6. sub edx,eax ; P0-P4 contribs to I7|I5. mov eax,PD P80000_N4545F [ebx*8] ; P5's contribution to I0|I2. add ebp,eax ; P0-P5 contribs to I0|I2. mov eax,PD P6491A_NB18A8 [ebx*8] ; P5's contribution to I1|I3. sub edi,eax ; P0-P5 contribs to I1|I3. mov eax,PD P80000_NA73D7 [ebx*8] ; P5's contribution to I4|I6. sub esi,eax ; P0-P5 contribs to I4|I6. mov eax,PD N96831_N2350B [ebx*8] ; P5's contribution to I7|I5. add edx,eax ; P0-P5 contribs to I3|I4. mov eax,PD P80000_P4545F [ecx*8] ; P6's contribution to I0|I2. add ebp,eax ; P0-P6 contribs to I0|I2. mov eax,PD N96831_P2350B [ecx*8] ; P6's contribution to I1|I3. add edi,eax ; P0-P6 contribs to I1|I3, unbiased. mov eax,PD P80000_PA73D7 [ecx*8] ; P6's contribution to I4|I6. sub esi,eax ; P0-P6 contribs to I4|I6. mov bl,P07 ; Fetch P7. mov eax,PD P6491A_PB18A8 [ecx*8] ; P6's contribution to I7|I5. mov cl,P13 ; Fetch P0. add edx,eax ; P0-P6 contribs to I7|I5. mov eax,PD P80000_PA73D7 [ebx*8] ; P7's contribution to I0|I2. add ebp,eax ; P0-P7 contribs to I0|I2. mov eax,PD P80000_P4545F [ebx*8] ; P7's contribution to I4|I6. add esi,eax ; P0-P7 contribs to I4|I6. mov eax,PD NB18A8_N96831 [ebx*8] ; P7's contribution to I1|I3. mov I00I02,ebp ; Store I0|I2 for line 0. mov I04I06,esi ; Store I4|I6 for line 0. lea edi,[edi+eax+40004000H] ; P0-P7 contribs to I1|I3, biased. mov eax,PD P2350B_P6491A [ebx*8] ; P7's contribution to I7|I5. sub edx,eax ; P0-P7 contribs to I7|I5. mov bl,P10 ; Fetch P3 of line 1. mov I01I03,edi ; Store I1|I3 for line 0. mov I07I05,edx ; Store I7|I5 for line 0.
mov esi,PD P80000_P4545F [ebx*8] mov eax,PD P80000_N4545F [ecx*8] mov edx,PD P2350B_P6491A [ebx*8] mov edi,PD NB18A8_P96831 [ecx*8] lea esi,[esi+eax+40004000H] mov eax,PD P80000_NA73D7 [ecx*8] lea edx,[edx+edi+40004000H] mov ebp,PD P80000_PA73D7 [ebx*8] mov edi,PD P2350B_N6491A [ecx*8] mov cl,P11 lea ebp,[ebp+eax+40004000H] mov eax,PD NB18A8_N96831 [ebx*8] sub edi,eax mov eax,PD P80000_P4545F [ecx*8] add ebp,eax mov eax,PD N96831_P2350B [ecx*8] sub edi,eax mov eax,PD P80000_PA73D7 [ecx*8] sub esi,eax mov bl,P12 mov eax,PD P6491A_PB18A8 [ecx*8] mov cl,P14 sub edx,eax mov eax,PD P80000_N4545F [ebx*8] add ebp,eax mov eax,PD P6491A_NB18A8 [ebx*8] add edi,eax mov eax,PD P80000_NA73D7 [ebx*8] sub esi,eax mov eax,PD N96831_N2350B [ebx*8] sub edx,eax mov eax,PD P80000_NA73D7 [ecx*8] add ebp,eax mov eax,PD P2350B_N6491A [ecx*8] sub edi,eax mov eax,PD P80000_N4545F [ecx*8] add esi,eax mov bl,P15 mov eax,PD NB18A8_P96831 [ecx*8] mov cl,P16 sub edx,eax mov eax,PD P80000_N4545F [ebx*8] add ebp,eax mov eax,PD P6491A_NB18A8 [ebx*8] sub edi,eax mov eax,PD P80000_NA73D7 [ebx*8] sub esi,eax mov eax,PD N96831_N2350B [ebx*8] add edx,eax mov eax,PD P80000_P4545F [ecx*8] add ebp,eax mov eax,PD N96831_P2350B [ecx*8] add edi,eax mov eax,PD P80000_PA73D7 [ecx*8] sub esi,eax mov bl,P17 mov eax,PD P6491A_PB18A8 [ecx*8] mov cl,P23 add edx,eax mov eax,PD P80000_PA73D7 [ebx*8] add ebp,eax mov eax,PD P80000_P4545F [ebx*8] add esi,eax mov eax,PD NB18A8_N96831 [ebx*8] mov I10I12,ebp mov I14I16,esi lea edi,[edi+eax+40004000H] mov eax,PD P2350B_P6491A [ebx*8] sub edx,eax mov bl,P20 mov I11I13,edi mov I17I15,edx
mov esi,PD P80000_P4545F [ebx*8] mov eax,PD P80000_N4545F [ecx*8] mov edx,PD P2350B_P6491A [ebx*8] mov edi,PD NB18A8_P96831 [ecx*8] lea esi,[esi+eax+40004000H] mov eax,PD P80000_NA73D7 [ecx*8] lea edx,[edx+edi+40004000H] mov ebp,PD P80000_PA73D7 [ebx*8] mov edi,PD P2350B_N6491A [ecx*8] mov cl,P21 lea ebp,[ebp+eax+40004000H] mov eax,PD NB18A8_N96831 [ebx*8] sub edi,eax mov eax,PD P80000_P4545F [ecx*8] add ebp,eax mov eax,PD N96831_P2350B [ecx*8] sub edi,eax mov eax,PD P80000_PA73D7 [ecx*8] sub esi,eax mov bl,P22 mov eax,PD P6491A_PB18A8 [ecx*8] mov cl,P24 sub edx,eax mov eax,PD P80000_N4545F [ebx*8] add ebp,eax mov eax,PD P6491A_NB18A8 [ebx*8] add edi,eax mov eax,PD P80000_NA73D7 [ebx*8] sub esi,eax mov eax,PD N96831_N2350B [ebx*8] sub edx,eax mov eax,PD P80000_NA73D7 [ecx*8] add ebp,eax mov eax,PD P2350B_N6491A [ecx*8] sub edi,eax mov eax,PD P80000_N4545F [ecx*8] add esi,eax mov bl,P25 mov eax,PD NB18A8_P96831 [ecx*8] mov cl,P26 sub edx,eax mov eax,PD P80000_N4545F [ebx*8] add ebp,eax mov eax,PD P6491A_NB18A8 [ebx*8] sub edi,eax mov eax,PD P80000_NA73D7 [ebx*8] sub esi,eax mov eax,PD N96831_N2350B [ebx*8] add edx,eax mov eax,PD P80000_P4545F [ecx*8] add ebp,eax mov eax,PD N96831_P2350B [ecx*8] add edi,eax mov eax,PD P80000_PA73D7 [ecx*8] sub esi,eax mov bl,P27 mov eax,PD P6491A_PB18A8 [ecx*8] mov cl,P33 add edx,eax mov eax,PD P80000_PA73D7 [ebx*8] add ebp,eax mov eax,PD P80000_P4545F [ebx*8] add esi,eax mov eax,PD NB18A8_N96831 [ebx*8] mov I20I22,ebp mov I24I26,esi lea edi,[edi+eax+40004000H] mov eax,PD P2350B_P6491A [ebx*8] sub edx,eax mov bl,P30 mov I21I23,edi mov I27I25,edx
mov esi,PD P80000_P4545F [ebx*8] mov eax,PD P80000_N4545F [ecx*8] mov edx,PD P2350B_P6491A [ebx*8] mov edi,PD NB18A8_P96831 [ecx*8] lea esi,[esi+eax+40004000H] mov eax,PD P80000_NA73D7 [ecx*8] lea edx,[edx+edi+40004000H] mov ebp,PD P80000_PA73D7 [ebx*8] mov edi,PD P2350B_N6491A [ecx*8] mov cl,P31 lea ebp,[ebp+eax+40004000H] mov eax,PD NB18A8_N96831 [ebx*8] sub edi,eax mov eax,PD P80000_P4545F [ecx*8] add ebp,eax mov eax,PD N96831_P2350B [ecx*8] sub edi,eax mov eax,PD P80000_PA73D7 [ecx*8] sub esi,eax mov bl,P32 mov eax,PD P6491A_PB18A8 [ecx*8] mov cl,P34 sub edx,eax mov eax,PD P80000_N4545F [ebx*8] add ebp,eax mov eax,PD P6491A_NB18A8 [ebx*8] add edi,eax mov eax,PD P80000_NA73D7 [ebx*8] sub esi,eax mov eax,PD N96831_N2350B [ebx*8] sub edx,eax mov eax,PD P80000_NA73D7 [ecx*8] add ebp,eax mov eax,PD P2350B_N6491A [ecx*8] sub edi,eax mov eax,PD P80000_N4545F [ecx*8] add esi,eax mov bl,P35 mov eax,PD NB18A8_P96831 [ecx*8] mov cl,P36 sub edx,eax mov eax,PD P80000_N4545F [ebx*8] add ebp,eax mov eax,PD P6491A_NB18A8 [ebx*8] sub edi,eax mov eax,PD P80000_NA73D7 [ebx*8] sub esi,eax mov eax,PD N96831_N2350B [ebx*8] add edx,eax mov eax,PD P80000_P4545F [ecx*8] add ebp,eax mov eax,PD N96831_P2350B [ecx*8] add edi,eax mov eax,PD P80000_PA73D7 [ecx*8] sub esi,eax mov bl,P37 mov eax,PD P6491A_PB18A8 [ecx*8] mov cl,P43 add edx,eax mov eax,PD P80000_PA73D7 [ebx*8] add ebp,eax mov eax,PD P80000_P4545F [ebx*8] add esi,eax mov eax,PD NB18A8_N96831 [ebx*8] mov I30I32,ebp mov I34I36,esi lea edi,[edi+eax+40004000H] mov eax,PD P2350B_P6491A [ebx*8] sub edx,eax mov bl,P40 mov I31I33,edi mov I37I35,edx
mov esi,PD P80000_P4545F [ebx*8] mov eax,PD P80000_N4545F [ecx*8] mov edx,PD P2350B_P6491A [ebx*8] mov edi,PD NB18A8_P96831 [ecx*8] add esi,eax mov eax,PD P80000_NA73D7 [ecx*8] add edx,edi mov ebp,PD P80000_PA73D7 [ebx*8] mov edi,PD P2350B_N6491A [ecx*8] mov cl,P41 add ebp,eax mov eax,PD NB18A8_N96831 [ebx*8] sub edi,eax mov eax,PD P80000_P4545F [ecx*8] add ebp,eax mov eax,PD N96831_P2350B [ecx*8] sub edi,eax mov eax,PD P80000_PA73D7 [ecx*8] sub esi,eax mov bl,P42 mov eax,PD P6491A_PB18A8 [ecx*8] mov cl,P44 sub edx,eax mov eax,PD P80000_N4545F [ebx*8] add ebp,eax mov eax,PD P6491A_NB18A8 [ebx*8] add edi,eax mov eax,PD P80000_NA73D7 [ebx*8] sub esi,eax mov eax,PD N96831_N2350B [ebx*8] sub edx,eax mov eax,PD P80000_NA73D7 [ecx*8] add ebp,eax mov eax,PD P2350B_N6491A [ecx*8] sub edi,eax mov eax,PD P80000_N4545F [ecx*8] add esi,eax mov bl,P45 mov eax,PD NB18A8_P96831 [ecx*8] mov cl,P46 sub edx,eax mov eax,PD P80000_N4545F [ebx*8] add ebp,eax mov eax,PD P6491A_NB18A8 [ebx*8] sub edi,eax mov eax,PD P80000_NA73D7 [ebx*8] sub esi,eax mov eax,PD N96831_N2350B [ebx*8] add edx,eax mov eax,PD P80000_P4545F [ecx*8] add ebp,eax mov eax,PD N96831_P2350B [ecx*8] add edi,eax mov eax,PD P80000_PA73D7 [ecx*8] sub esi,eax mov bl,P47 mov eax,PD P6491A_PB18A8 [ecx*8] mov cl,P53 add edx,eax mov eax,PD P80000_PA73D7 [ebx*8] add ebp,eax mov eax,PD P80000_P4545F [ebx*8] add esi,eax mov eax,PD NB18A8_N96831 [ebx*8] mov I40I42,ebp mov I44I46,esi add edi,eax mov eax,PD P2350B_P6491A [ebx*8] sub edx,eax mov bl,P50 mov I41I43,edi mov I47I45,edx
mov esi,PD P80000_P4545F [ebx*8] mov eax,PD P80000_N4545F [ecx*8] mov edx,PD P2350B_P6491A [ebx*8] mov edi,PD NB18A8_P96831 [ecx*8] add esi,eax mov eax,PD P80000_NA73D7 [ecx*8] add edx,edi mov ebp,PD P80000_PA73D7 [ebx*8] mov edi,PD P2350B_N6491A [ecx*8] mov cl,P51 add ebp,eax mov eax,PD NB18A8_N96831 [ebx*8] sub edi,eax mov eax,PD P80000_P4545F [ecx*8] add ebp,eax mov eax,PD N96831_P2350B [ecx*8] sub edi,eax mov eax,PD P80000_PA73D7 [ecx*8] sub esi,eax mov bl,P52 mov eax,PD P6491A_PB18A8 [ecx*8] mov cl,P54 sub edx,eax mov eax,PD P80000_N4545F [ebx*8] add ebp,eax mov eax,PD P6491A_NB18A8 [ebx*8] add edi,eax mov eax,PD P80000_NA73D7 [ebx*8] sub esi,eax mov eax,PD N96831_N2350B [ebx*8] sub edx,eax mov eax,PD P80000_NA73D7 [ecx*8] add ebp,eax mov eax,PD P2350B_N6491A [ecx*8] sub edi,eax mov eax,PD P80000_N4545F [ecx*8] add esi,eax mov bl,P55 mov eax,PD NB18A8_P96831 [ecx*8] mov cl,P56 sub edx,eax mov eax,PD P80000_N4545F [ebx*8] add ebp,eax mov eax,PD P6491A_NB18A8 [ebx*8] sub edi,eax mov eax,PD P80000_NA73D7 [ebx*8] sub esi,eax mov eax,PD N96831_N2350B [ebx*8] add edx,eax mov eax,PD P80000_P4545F [ecx*8] add ebp,eax mov eax,PD N96831_P2350B [ecx*8] add edi,eax mov eax,PD P80000_PA73D7 [ecx*8] sub esi,eax mov bl,P57 mov eax,PD P6491A_PB18A8 [ecx*8] mov cl,P63 add edx,eax mov eax,PD P80000_PA73D7 [ebx*8] add ebp,eax mov eax,PD P80000_P4545F [ebx*8] add esi,eax mov eax,PD NB18A8_N96831 [ebx*8] mov I50I52,ebp mov I54I56,esi add edi,eax mov eax,PD P2350B_P6491A [ebx*8] sub edx,eax mov bl,P60 mov I51I53,edi mov I57I55,edx
mov esi,PD P80000_P4545F [ebx*8] mov eax,PD P80000_N4545F [ecx*8] mov edx,PD P2350B_P6491A [ebx*8] mov edi,PD NB18A8_P96831 [ecx*8] add esi,eax mov eax,PD P80000_NA73D7 [ecx*8] add edx,edi mov ebp,PD P80000_PA73D7 [ebx*8] mov edi,PD P2350B_N6491A [ecx*8] mov cl,P61 add ebp,eax mov eax,PD NB18A8_N96831 [ebx*8] sub edi,eax mov eax,PD P80000_P4545F [ecx*8] add ebp,eax mov eax,PD N96831_P2350B [ecx*8] sub edi,eax mov eax,PD P80000_PA73D7 [ecx*8] sub esi,eax mov bl,P62 mov eax,PD P6491A_PB18A8 [ecx*8] mov cl,P64 sub edx,eax mov eax,PD P80000_N4545F [ebx*8] add ebp,eax mov eax,PD P6491A_NB18A8 [ebx*8] add edi,eax mov eax,PD P80000_NA73D7 [ebx*8] sub esi,eax mov eax,PD N96831_N2350B [ebx*8] sub edx,eax mov eax,PD P80000_NA73D7 [ecx*8] add ebp,eax mov eax,PD P2350B_N6491A [ecx*8] sub edi,eax mov eax,PD P80000_N4545F [ecx*8] add esi,eax mov bl,P65 mov eax,PD NB18A8_P96831 [ecx*8] mov cl,P66 sub edx,eax mov eax,PD P80000_N4545F [ebx*8] add ebp,eax mov eax,PD P6491A_NB18A8 [ebx*8] sub edi,eax mov eax,PD P80000_NA73D7 [ebx*8] sub esi,eax mov eax,PD N96831_N2350B [ebx*8] add edx,eax mov eax,PD P80000_P4545F [ecx*8] add ebp,eax mov eax,PD N96831_P2350B [ecx*8] add edi,eax mov eax,PD P80000_PA73D7 [ecx*8] sub esi,eax mov bl,P67 mov eax,PD P6491A_PB18A8 [ecx*8] mov cl,P73 add edx,eax mov eax,PD P80000_PA73D7 [ebx*8] add ebp,eax mov eax,PD P80000_P4545F [ebx*8] add esi,eax mov eax,PD NB18A8_N96831 [ebx*8] mov I60I62,ebp mov I64I66,esi add edi,eax mov eax,PD P2350B_P6491A [ebx*8] sub edx,eax mov bl,P70 mov I61I63,edi mov I67I65,edx
mov esi,PD P80000_P4545F [ebx*8] mov eax,PD P80000_N4545F [ecx*8] mov edx,PD P2350B_P6491A [ebx*8] mov edi,PD NB18A8_P96831 [ecx*8] add esi,eax mov eax,PD P80000_NA73D7 [ecx*8] add edx,edi mov ebp,PD P80000_PA73D7 [ebx*8] mov edi,PD P2350B_N6491A [ecx*8] mov cl,P71 add ebp,eax mov eax,PD NB18A8_N96831 [ebx*8] sub edi,eax mov eax,PD P80000_P4545F [ecx*8] add ebp,eax mov eax,PD N96831_P2350B [ecx*8] sub edi,eax mov eax,PD P80000_PA73D7 [ecx*8] sub esi,eax mov bl,P72 mov eax,PD P6491A_PB18A8 [ecx*8] mov cl,P74 sub edx,eax mov eax,PD P80000_N4545F [ebx*8] add ebp,eax mov eax,PD P6491A_NB18A8 [ebx*8] add edi,eax mov eax,PD P80000_NA73D7 [ebx*8] sub esi,eax mov eax,PD N96831_N2350B [ebx*8] sub edx,eax mov eax,PD P80000_NA73D7 [ecx*8] add ebp,eax mov eax,PD P2350B_N6491A [ecx*8] sub edi,eax mov eax,PD P80000_N4545F [ecx*8] add esi,eax mov bl,P75 mov eax,PD NB18A8_P96831 [ecx*8] mov cl,P76 sub edx,eax mov eax,PD P80000_N4545F [ebx*8] add ebp,eax mov eax,PD P6491A_NB18A8 [ebx*8] sub edi,eax mov eax,PD P80000_NA73D7 [ebx*8] sub esi,eax mov eax,PD N96831_N2350B [ebx*8] add edx,eax mov eax,PD P80000_P4545F [ecx*8] add ebp,eax mov eax,PD N96831_P2350B [ecx*8] add edi,eax mov eax,PD P80000_PA73D7 [ecx*8] sub esi,eax mov bl,P77 mov eax,PD P6491A_PB18A8 [ecx*8] mov ecx,I00I02 ; Fetch I0 (upper_lim <skew>) = 2000 4000 ; ; (lower_lim is -upper_limit) add edx,eax mov eax,PD P80000_PA73D7 [ebx*8] add ebp,eax ; I70I72, aka I7. 2000 0000 mov eax,PD P80000_P4545F [ebx*8] add esi,eax mov eax,PD NB18A8_N96831 [ebx*8] mov I74I76,esi mov esi,I30I32 ; Fetch I3 2000 4000 add edi,eax mov eax,I40I42 ; Fetch I4 2000 0000 sub esi,eax ; I3 - I4 4000 4000 sub ecx,ebp ; I0 - I7 4000 4000 shr ecx,1 ; R7 = (I0-I7)/2 (dirty) 2000 2000 and esi,0FFFEFFFFH ; pre-clean R4 shr esi,1 ; R4 = (I3-I4)/2 (dirty) 2000 2000 and ecx,0FFFF7FFFH ; R7 = (I0-I7)/2 (clean) 2000 2000 mov ebx,PD P2350B_P6491A [ebx*8] mov I71I73,edi sub edx,ebx lea ebx,[ecx+ecx*2] ; 3R7 6000 6000 mov I77I75,edx lea edi,[esi+esi*2] ; 3R4 6000 6000
; eax: I4 2000 0000 ; ebx: 3R7 6000 6000 ; ecx: R7 2000 2000 ; edx: available ; esi: R4 2000 2000 ; edi: 3R4 6000 6000 ; ebp: I7 2000 0000
lea ebp,[ebp+ecx+40004000H] ; R0 = (I0+I7)/2 2000 6000 add eax,esi ; R3 = (I3+I4)/2 2000 2000 shr ecx,1 ; R7/2 (dirty) 1000 1000 and esi,0FFFEFFFFH ; pre-clean shr esi,1 ; R4/2 (clean) 1000 1000 and ecx,0FFFF7FFFH ; clean add ebx,ecx ; 7R7/2 7000 7000 add edi,esi ; 7R4/2 7000 7000 shr ebx,6 ; 7R7/128 (dirty) 01C0 01C0 and edi,0FFC0FFFFH ; pre-clean shr edi,6 ; 7R4/128 (clean) 01C0 01C0 and ebx,0FFFF03FFH ; clean add ebx,ecx ; 71R7/128 11C0 11C0 add edi,esi ; 71R4/128 11C0 11C0 lea edx,[eax+ebp-40004000H] ; S0 = R0 + R3 4000 4000 sub ebp,eax ; S3 = R0 - R3 4000 4000 lea ecx,[ebx+ebx*2+6E406E40H] ; 213R7/128 3540 A380 lea esi,[edi+edi*2+27402740H] ; 213R4/128 3540 5C80 shr ecx,1 ; 213R7/256 (dirty) 1AA0 51C0 and esi,0FFFEFFFFH ; pre-clean shr esi,1 ; 213R4/256 (clean) 1AA0 2E40 and ecx,0FFFF7FFFH ; clean sub ecx,edi ; S7 = (213R7 - 142R4)/256 2C60 4000 mov S0,edx ; Free register for work. mov S3,ebp ; Free register for work. lea esi,[esi+ebx+80008000H] ; S4 = (142R7 + 213R3)/256 2C60 C000 mov S7,ecx ; Free register for work. mov eax,I10I12 ; Fetch I1 2000 4000 mov S4,esi ; Free register for work.
; mem: S4 2C60 C000 ; mem: S7 2C60 4000 ; mem: S0 4000 4000 ; mem: S3 4000 4000
mov ebx,I20I22 ; Fetch I2 2000 4000 mov ecx,I50I52 ; Fetch I5 2000 0000 mov edx,I60I62 ; Fetch I6 2000 0000 sub eax,edx ; I1 - I6 4000 4000 sub ebx,ecx ; I2 - I5 4000 4000 shr eax,1 ; R6 = (I1-I6)/2 (dirty) 2000 2000 and ebx,0FFFEFFFFH ; pre-clean R4 shr ebx,1 ; R5 = (I2-I5)/2 (dirty) 2000 2000 and eax,0FFFF7FFFH ; R6 = (I1-I6)/2 (clean) 2000 2000
; eax: R6 2000 2000 ; ebx: R5 2000 2000 ; ecx: I5 2000 0000 ; edx: I6 2000 0000 ; mem: S4 2C60 C000 ; mem: S7 2C60 4000 ; mem: S0 4000 4000 ; mem: S3 4000 4000
mov esi,ebx ; R5 2000 2000 mov edi,eax ; R6 2000 2000 shr esi,6 ; R5/64 0080 0080 and edi,0FFC0FFFFH ; pre-clean shr edi,6 ; R6/65 0080 0080 and esi,0FFFF03FFH ; clean lea edx,[eax+edx+20002000H] ; R1 = (I1+I6)/2 2000 4000 lea ecx,[ecx+ebx-20002000H] ; R2 = (I2+I5)/2 2000 0000 lea ebp,[ebx+ebx*2] ; 3R5 6000 6000 sub ebx,esi ; 63R5/64 1F80 1F80 shr ebp,4 ; 3R5/16 (dirty) 0600 0600 lea esi,[eax+eax*2] ; 3R6 6000 6000 sub eax,edi ; 63R6/64 1F80 1F80 mov edi,ebx ; 63R5/64 1F80 1F80 shr edi,7 ; 63R5/8192 (dirty) 003F 003F and ebp,0FFFF0FFFH ; clean shr esi,4 ; 3R6/16 (dirty) 0600 0600 and edi,0FFFF01FFH ; clean and esi,0FFFF0FFFH ; clean sub edx,ecx ; S2 = R1 - R2 4000 4000 lea edi,[edi+ebp-46BF46BFH] ; 1599R5/8192 063F -4080 mov ebp,eax ; 63R6/64 1F80 1F80 shr ebp,7 ; 63R6/8192 (dirty) 003F 003F sub eax,edi ; S6 = 8064R6/8192 - 1599R5/8192 25BF 6000 and ebp,0FFFF01FFH ; clean lea ecx,[edx+ecx*2-80008000H] ; S1 = R1 + R2 4000 -4000 add ebp,esi ; 1599R6/8192 063F 063F mov esi,S0 ; Reload S0 4000 4000 mov edi,CoeffStream ; Fetch addr at which to place blk of coeffs. sub esi,ecx ; C4 = T1 = S0 - S1 8000 8000 lea ebx,[ebx+ebp-45BF45BFH] ; S5 = 8064R5/8192 + 1599R6/8192 25BF -2000 mov ebp,S4 ; Reload S4 2C60 C000
; eax: S6 25BF 6000 ; ebx: S5 25BF -2000 ; ecx: S0 4000 4000 ; edx: S2 4000 4000 ; esi: C4 8000 8000 ; edi: Destination pointer. ; ebp: S4 2C60 C000 ; mem: S7 2C60 4000 ; mem: S3 4000 4000
sub ebp,eax ; T6 = S4 - S6 521F 6000 mov PD [edi+C40C42],esi ; Store coeffs C40 and C42. lea ecx,[esi+ecx*2+80008000H] ; C0 = T0 = S0 + S1 8000 8000 mov esi,S7 ; Reload S7 2C60 4000 sub esi,ebx ; T5 = S7 - S5 521F 6000 lea eax,[ebp+eax*2-0C000C000H] ; T4 = S4 + S6 521F 6000 mov PD [edi+C00C02],ecx ; Store coeffs C00 and C02. mov ecx,ebp ; T6 521F 6000 shr ebp,2 ; T6/4 (dirty) 1487 1800 lea ebx,[esi+ebx*2+0C000C000H] ; T7 = S7 + S5 521F E000
; eax: T4 521F 6000 ; ebx: T7 521F 6000 ; ecx: T6 521F 6000 ; edx: S2 4000 4000 ; esi: T5 521F 6000 ; edi: Destination pointer. ; ebp: T6/4 (dirty) 1487 1800 ; mem: S3 4000 4000 ; done: C0, C4
and ebp,0FFFF3FFFH ; T6/4 (clean) 1487 1800 sub ebx,eax ; C7 = T7 - T4 <7642> 8000 add ecx,ebp ; 5T6/4 66A6 7800 mov PD [edi+C70C72],ebx ; Store coeffs C70 and C72. mov ebp,ecx ; 5T6/4 66A6 7800 and ecx,0FFF8FFFFH ; pre-clean shr ecx,3 ; 5T6/32 (clean) 0CD4 0F00 lea eax,[ebx+eax*2-0C000C000H] ; C1 = T7 + T4 <7642> 8000 mov ebx,esi ; T5 521F 6000 and esi,0FFFCFFFFH ; pre-clean shr esi,2 ; T5/4 (clean) 1487 1800 lea ecx,[ecx+ebp-07000700H] ; C5 = 45T6/32 737A 8000 mov PD [edi+C50C52],ecx ; Store coeffs C50 and C52. add esi,ebx ; 5T5/4 66A6 7800 mov ebx,esi ; 5T5/4 66A6 7800 and esi,0FFF8FFFFH ; pre-clean shr esi,3 ; 5T5/32 (clean) 0CD4 0F00 mov ebp,S3 ; Reload S3 4000 4000 mov ecx,edx ; S2 4000 4000 lea esi,[esi+ebx-07000700H] ; C3 = 45T5/32 737A 8000 mov ebx,ebp ; S3 4000 4000 ;
; eax: C1 521E 8000 ; ebx: S3 4000 4000 ; ecx: S2 4000 4000 ; edx: S2 4000 4000 ; esi: C3 737A 8000 ; edi: Destination pointer. ; ebp: S3 4000 4000 ; done: C0, C4, C5, C7
shr ebp,2 ; S3/4 (dirty) 1000 1000 and ecx,0FFFCFFFFH ; pre-clean shr ecx,2 ; S2/4 (clean) 1000 1000 and ebp,0FFFF3FFFH ; S3/4 (clean) 1000 1000 mov PD [edi+C10C12],eax ; Store coeffs C10 and C12. mov PD [edi+C30C32],esi ; Store coeffs C30 and C32. lea eax,[edx+ecx] ; 5S2/4 5000 5000 lea esi,[ebx+ebp] ; 5S3/4 5000 5000 shr ebp,2 ; S3/16 (dirty) 0400 0400 and ecx,0FFFCFFFFH ; pre-clean shr ecx,2 ; S2/16 (clean) 0400 0400 and ebp,0FFFF3FFFH ; S3/16 (clean) 0400 0400 add ecx,eax ; 21S2/16 5400 5400 add ebp,esi ; 21S3/16 5400 5400 shr eax,5 ; 5S2/128 (dirty) 0280 0280 and esi,0FFE0FFFFH ; pre-clean shr esi,5 ; 5S3/128 (clean) 0280 0280 and eax,0FFFF07FFH ; 5S2/128 (clean) 0280 0280 shr edx,1 ; S2/2 (dirty) 2000 2000 and ebx,0FFFEFFFFH ; pre-clean shr ebx,1 ; S3/2 (clean) 2000 2000 and edx,0FFFF7FFFH ; S2/2 (clean) 2000 2000 sub ebx,ecx ; (64S3 - 168S2) / 128 7400 -3400 add eax,ebp ; (5S2 + 168S3) / 128 5680 5680
mov ecx,I01I03 mov ebp,I71I73 lea ebx,[ebx+esi+0B180B180H] ; C6 = (69S3 - 168S2) / 128 7680 8000 lea edx,[eax+edx+009800980H] ; C2 = (69S2 + 168S3) / 128 7680 8000 mov esi,I31I33 mov eax,I41I43 sub esi,eax sub ecx,ebp shr ecx,1 and esi,0FFFEFFFFH shr esi,1 and ecx,0FFFF7FFFH mov PD [edi+C60C62],ebx mov PD [edi+C20C22],edx lea ebx,[ecx+ecx*2] lea edi,[esi+esi*2] lea ebp,[ebp+ecx+40004000H] add eax,esi shr ecx,1 and esi,0FFFEFFFFH shr esi,1 and ecx,0FFFF7FFFH add ebx,ecx add edi,esi shr ebx,6 and edi,0FFC0FFFFH shr edi,6 and ebx,0FFFF03FFH add ebx,ecx add edi,esi lea edx,[eax+ebp-40004000H] sub ebp,eax lea ecx,[ebx+ebx*2+6E406E40H] lea esi,[edi+edi*2+27402740H] shr ecx,1 and esi,0FFFEFFFFH shr esi,1 and ecx,0FFFF7FFFH sub ecx,edi mov S0,edx mov S3,ebp lea esi,[esi+ebx+80008000H] mov S7,ecx mov eax,I11I13 mov S4,esi mov ebx,I21I23 mov ecx,I51I53 mov edx,I61I63 sub eax,edx sub ebx,ecx shr eax,1 and ebx,0FFFEFFFFH shr ebx,1 and eax,0FFFF7FFFH mov esi,ebx mov edi,eax shr esi,6 and edi,0FFC0FFFFH shr edi,6 and esi,0FFFF03FFH lea edx,[eax+edx+20002000H] lea ecx,[ecx+ebx-20002000H] lea ebp,[ebx+ebx*2] sub ebx,esi shr ebp,4 lea esi,[eax+eax*2] sub eax,edi mov edi,ebx shr edi,7 and ebp,0FFFF0FFFH shr esi,4 and edi,0FFFF01FFH and esi,0FFFF0FFFH sub edx,ecx lea edi,[edi+ebp-46BF46BFH] mov ebp,eax shr ebp,7 sub eax,edi and ebp,0FFFF01FFH lea ecx,[edx+ecx*2-80008000H] add ebp,esi mov esi,S0 mov edi,CoeffStream sub esi,ecx lea ebx,[ebx+ebp-45BF45BFH] mov ebp,S4 sub ebp,eax mov PD [edi+C41C43],esi lea ecx,[esi+ecx*2+80008000H] mov esi,S7 sub esi,ebx lea eax,[ebp+eax*2-0C000C000H] mov PD [edi+C01C03],ecx mov ecx,ebp shr ebp,2 lea ebx,[esi+ebx*2+0C000C000H] and ebp,0FFFF3FFFH sub ebx,eax add ecx,ebp mov PD [edi+C71C73],ebx mov ebp,ecx and ecx,0FFF8FFFFH shr ecx,3 lea eax,[ebx+eax*2-0C000C000H] mov ebx,esi and esi,0FFFCFFFFH shr esi,2 lea ecx,[ecx+ebp-07000700H] mov PD [edi+C51C53],ecx add esi,ebx mov ebx,esi and esi,0FFF8FFFFH shr esi,3 mov ebp,S3 mov ecx,edx lea esi,[esi+ebx-07000700H] mov ebx,ebp ; shr ebp,2 and ecx,0FFFCFFFFH shr ecx,2 and ebp,0FFFF3FFFH mov PD [edi+C11C13],eax mov PD [edi+C31C33],esi lea eax,[edx+ecx] lea esi,[ebx+ebp] shr ebp,2 and ecx,0FFFCFFFFH shr ecx,2 and ebp,0FFFF3FFFH add ecx,eax add ebp,esi shr eax,5 and esi,0FFE0FFFFH shr esi,5 and eax,0FFFF07FFH shr edx,1 and ebx,0FFFEFFFFH shr ebx,1 and edx,0FFFF7FFFH sub ebx,ecx add eax,ebp
mov ecx,I04I06 mov ebp,I74I76 lea ebx,[ebx+esi+0B180B180H] lea edx,[eax+edx+009800980H] mov esi,I34I36 mov eax,I44I46 sub esi,eax sub ecx,ebp shr ecx,1 and esi,0FFFEFFFFH shr esi,1 and ecx,0FFFF7FFFH mov PD [edi+C61C63],ebx mov PD [edi+C21C23],edx lea ebx,[ecx+ecx*2] lea edi,[esi+esi*2] lea ebp,[ebp+ecx+40004000H] add eax,esi shr ecx,1 and esi,0FFFEFFFFH shr esi,1 and ecx,0FFFF7FFFH add ebx,ecx add edi,esi shr ebx,6 and edi,0FFC0FFFFH shr edi,6 and ebx,0FFFF03FFH add ebx,ecx add edi,esi lea edx,[eax+ebp-40004000H] sub ebp,eax lea ecx,[ebx+ebx*2+6E406E40H] lea esi,[edi+edi*2+27402740H] shr ecx,1 and esi,0FFFEFFFFH shr esi,1 and ecx,0FFFF7FFFH sub ecx,edi mov S0,edx mov S3,ebp lea esi,[esi+ebx+80008000H] mov S7,ecx mov eax,I14I16 mov S4,esi mov ebx,I24I26 mov ecx,I54I56 mov edx,I64I66 sub eax,edx sub ebx,ecx shr eax,1 and ebx,0FFFEFFFFH shr ebx,1 and eax,0FFFF7FFFH mov esi,ebx mov edi,eax shr esi,6 and edi,0FFC0FFFFH shr edi,6 and esi,0FFFF03FFH lea edx,[eax+edx+20002000H] lea ecx,[ecx+ebx-20002000H] lea ebp,[ebx+ebx*2] sub ebx,esi shr ebp,4 lea esi,[eax+eax*2] sub eax,edi mov edi,ebx shr edi,7 and ebp,0FFFF0FFFH shr esi,4 and edi,0FFFF01FFH and esi,0FFFF0FFFH sub edx,ecx lea edi,[edi+ebp-46BF46BFH] mov ebp,eax shr ebp,7 sub eax,edi and ebp,0FFFF01FFH lea ecx,[edx+ecx*2-80008000H] add ebp,esi mov esi,S0 mov edi,CoeffStream sub esi,ecx lea ebx,[ebx+ebp-45BF45BFH] mov ebp,S4 sub ebp,eax mov PD [edi+C44C46],esi lea ecx,[esi+ecx*2+80008000H] mov esi,S7 sub esi,ebx lea eax,[ebp+eax*2-0C000C000H] mov PD [edi+C04C06],ecx mov ecx,ebp shr ebp,2 lea ebx,[esi+ebx*2+0C000C000H] and ebp,0FFFF3FFFH sub ebx,eax add ecx,ebp mov PD [edi+C74C76],ebx mov ebp,ecx and ecx,0FFF8FFFFH shr ecx,3 lea eax,[ebx+eax*2-0C000C000H] mov ebx,esi and esi,0FFFCFFFFH shr esi,2 lea ecx,[ecx+ebp-07000700H] mov PD [edi+C54C56],ecx add esi,ebx mov ebx,esi and esi,0FFF8FFFFH shr esi,3 mov ebp,S3 mov ecx,edx lea esi,[esi+ebx-07000700H] mov ebx,ebp ; shr ebp,2 and ecx,0FFFCFFFFH shr ecx,2 and ebp,0FFFF3FFFH mov PD [edi+C14C16],eax mov PD [edi+C34C36],esi lea eax,[edx+ecx] lea esi,[ebx+ebp] shr ebp,2 and ecx,0FFFCFFFFH shr ecx,2 and ebp,0FFFF3FFFH add ecx,eax add ebp,esi shr eax,5 and esi,0FFE0FFFFH shr esi,5 and eax,0FFFF07FFH shr edx,1 and ebx,0FFFEFFFFH shr ebx,1 and edx,0FFFF7FFFH sub ebx,ecx add eax,ebp
mov ecx,I07I05 mov ebp,I77I75 lea ebx,[ebx+esi+0B180B180H] lea edx,[eax+edx+009800980H] mov esi,I37I35 mov eax,I47I45 sub esi,eax sub ecx,ebp shr ecx,1 and esi,0FFFEFFFFH shr esi,1 and ecx,0FFFF7FFFH mov PD [edi+C64C66],ebx mov PD [edi+C24C26],edx lea ebx,[ecx+ecx*2] lea edi,[esi+esi*2] lea ebp,[ebp+ecx+40004000H] add eax,esi shr ecx,1 and esi,0FFFEFFFFH shr esi,1 and ecx,0FFFF7FFFH add ebx,ecx add edi,esi shr ebx,6 and edi,0FFC0FFFFH shr edi,6 and ebx,0FFFF03FFH add ebx,ecx add edi,esi lea edx,[eax+ebp-40004000H] sub ebp,eax lea ecx,[ebx+ebx*2+6E406E40H] lea esi,[edi+edi*2+27402740H] shr ecx,1 and esi,0FFFEFFFFH shr esi,1 and ecx,0FFFF7FFFH sub ecx,edi mov S0,edx mov S3,ebp lea esi,[esi+ebx+80008000H] mov S7,ecx mov eax,I17I15 mov S4,esi mov ebx,I27I25 mov ecx,I57I55 mov edx,I67I65 sub eax,edx sub ebx,ecx shr eax,1 and ebx,0FFFEFFFFH shr ebx,1 and eax,0FFFF7FFFH mov esi,ebx mov edi,eax shr esi,6 and edi,0FFC0FFFFH shr edi,6 and esi,0FFFF03FFH lea edx,[eax+edx+20002000H] lea ecx,[ecx+ebx-20002000H] lea ebp,[ebx+ebx*2] sub ebx,esi shr ebp,4 lea esi,[eax+eax*2] sub eax,edi mov edi,ebx shr edi,7 and ebp,0FFFF0FFFH shr esi,4 and edi,0FFFF01FFH and esi,0FFFF0FFFH sub edx,ecx lea edi,[edi+ebp-46BF46BFH] mov ebp,eax shr ebp,7 sub eax,edi and ebp,0FFFF01FFH lea ecx,[edx+ecx*2-80008000H] add ebp,esi mov esi,S0 mov edi,CoeffStream sub esi,ecx lea ebx,[ebx+ebp-45BF45BFH] mov ebp,S4 sub ebp,eax mov PD [edi+C47C45],esi lea ecx,[esi+ecx*2+80008000H] mov esi,S7 sub esi,ebx lea eax,[ebp+eax*2-0C000C000H] mov PD [edi+C07C05],ecx mov ecx,ebp shr ebp,2 lea ebx,[esi+ebx*2+0C000C000H] and ebp,0FFFF3FFFH sub ebx,eax add ecx,ebp mov PD [edi+C77C75],ebx mov ebp,ecx and ecx,0FFF8FFFFH shr ecx,3 lea eax,[ebx+eax*2-0C000C000H] mov ebx,esi and esi,0FFFCFFFFH shr esi,2 lea ecx,[ecx+ebp-07000700H] mov PD [edi+C57C55],ecx add esi,ebx mov ebx,esi and esi,0FFF8FFFFH shr esi,3 mov ebp,S3 mov ecx,edx lea esi,[esi+ebx-07000700H] mov ebx,ebp ; shr ebp,2 and ecx,0FFFCFFFFH shr ecx,2 and ebp,0FFFF3FFFH mov PD [edi+C17C15],eax mov PD [edi+C37C35],esi lea eax,[edx+ecx] lea esi,[ebx+ebp] shr ebp,2 and ecx,0FFFCFFFFH shr ecx,2 and ebp,0FFFF3FFFH add ecx,eax add ebp,esi shr eax,5 and esi,0FFE0FFFFH shr esi,5 and eax,0FFFF07FFH shr edx,1 and ebx,0FFFEFFFFH shr ebx,1 and edx,0FFFF7FFFH sub ebx,ecx add eax,ebp
mov ecx,CoeffStreamStart lea ebp,[edi-SIZEOF T_CoeffBlk] ; Advance cursor for block action stream. lea ebx,[ebx+esi+0B180B180H] lea edx,[eax+edx+009800980H] mov PD [edi+C67C65],ebx mov PD [edi+C27C25],edx
; Forward Slant Transform is done
cmp ebp,ecx mov edi,ebp mov CoeffStream,edi jae NextBlock ; Process next block.
Done:
mov esp,StashESP pop ebx pop ebp pop edi pop esi rturn
FORWARDDCT endp
END
|
