URL
https://opencores.org/ocsvn/bluespec-h264/bluespec-h264/trunk
Subversion Repositories bluespec-h264
[/] [bluespec-h264/] [trunk/] [src/] [mkInterpolator.bsv] - Rev 100
Compare with Previous | Blame | View Log
// The MIT License
// Copyright (c) 2006-2007 Massachusetts Institute of Technology
// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files (the "Software"), to deal
// in the Software without restriction, including without limitation the rights
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
// copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
// The above copyright notice and this permission notice shall be included in
// all copies or substantial portions of the Software.
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
// THE SOFTWARE.
//**********************************************************************
// interpolator implementation
//----------------------------------------------------------------------
//
//
import H264Types::*;
import IInterpolator::*;
import FIFO::*;
import Vector::*;
import Connectable::*;
import GetPut::*;
import ClientServer::*;
//-----------------------------------------------------------
// Local Datatypes
//-----------------------------------------------------------
typedef union tagged
{
struct { Bit#(2) xFracL; Bit#(2) yFracL; Bit#(2) offset; IPBlockType bt; } IPWLuma;
struct { Bit#(3) xFracC; Bit#(3) yFracC; Bit#(2) offset; IPBlockType bt; } IPWChroma;
}
InterpolatorWT deriving(Eq,Bits);
//-----------------------------------------------------------
// Helper functions
function Bit#(8) clip1y10to8( Bit#(10) innum );
if(innum[9] == 1)
return 0;
else if(innum[8] == 1)
return 255;
else
return truncate(innum);
endfunction
interface Interpolate8to15;
method Bit#(15) interpolate(Bit#(8) in0, Bit#(8) in1, Bit#(8) in2, Bit#(8) in3, Bit#(8) in4, Bit#(8) in5);
endinterface
module mkInterpolate8to15 (Interpolate8to15);
method Bit#(15) interpolate( Bit#(8) in0, Bit#(8) in1, Bit#(8) in2, Bit#(8) in3, Bit#(8) in4, Bit#(8) in5 );
return zeroExtend(in0) - 5*zeroExtend(in1) + 20*zeroExtend(in2) + 20*zeroExtend(in3) - 5*zeroExtend(in4) + zeroExtend(in5);
endmethod
endmodule
interface Interpolate15to8;
method Bit#(8) interpolate( Bit#(15) in0, Bit#(15) in1, Bit#(15) in2, Bit#(15) in3, Bit#(15) in4, Bit#(15) in5 );
endinterface
module mkInterpolate15to8 (Interpolate15to8);
method Bit#(8) interpolate( Bit#(15) in0, Bit#(15) in1, Bit#(15) in2, Bit#(15) in3, Bit#(15) in4, Bit#(15) in5 );
Bit#(20) temp = signExtend(in0) - 5*signExtend(in1) + 20*signExtend(in2) + 20*signExtend(in3) - 5*signExtend(in4) + signExtend(in5) + 512;
return clip1y10to8(truncate(temp>>10));
endmethod
endmodule
//-----------------------------------------------------------
// Interpolation Module
//-----------------------------------------------------------
(* synthesize *)
module mkInterpolator( Interpolator );
FIFO#(InterpolatorIT) reqfifoLoad <- mkSizedFIFO(interpolator_reqfifoLoad_size); // This fifo takes in motion vector
// pixel requests.
FIFO#(InterpolatorWT) reqfifoWork1 <- mkSizedFIFO(interpolator_reqfifoWork_size); // This is where the memory responses
// come from
Interpolate15to8 interpolate15to8 <- mkInterpolate15to8();
Interpolate8to15 interpolate8to15 <- mkInterpolate8to15();
Reg#(Maybe#(InterpolatorWT)) reqregWork2 <- mkReg(Invalid);
FIFO#(Vector#(4,Bit#(8))) outfifo <- mkFIFO;
Reg#(Bool) endOfFrameFlag <- mkReg(False);
FIFO#(InterpolatorLoadReq) memReqQ <- mkFIFO;
FIFO#(InterpolatorLoadResp) memRespQ <- mkSizedFIFO(interpolator_memRespQ_size);
Reg#(Bit#(PicWidthSz)) picWidth <- mkReg(maxPicWidthInMB);
Reg#(Bit#(PicHeightSz)) picHeight <- mkReg(0);
RFile1#(Bit#(6),Vector#(4,Bit#(15))) workFile <- mkRFile1Full();
RFile1#(Bit#(6),Vector#(4,Bit#(8))) storeFile <- mkRFile1Full();
Reg#(Bit#(1)) workFileFlag <- mkReg(0);
RFile1#(Bit#(4),Vector#(4,Bit#(8))) resultFile <- mkRFile1Full();
Reg#(Bit#(1)) loadStage <- mkReg(0);
Reg#(Bit#(2)) loadHorNum <- mkReg(0);
Reg#(Bit#(4)) loadVerNum <- mkReg(0);
Reg#(Bit#(2)) work1MbPart <- mkReg(0);//only for Chroma
Reg#(Bit#(2)) work1SubMbPart <- mkReg(0);//only for Chroma
Reg#(Bit#(1)) work1Stage <- mkReg(0);
Reg#(Bit#(2)) work1HorNum <- mkReg(0);
Reg#(Bit#(4)) work1VerNum <- mkReg(0);
Reg#(Vector#(20,Bit#(8))) work1Vector8 <- mkRegU;
Reg#(Bool) work1Done <- mkReg(False);
Reg#(Bit#(2)) work2SubMbPart <- mkReg(0);
Reg#(Bit#(2)) work2HorNum <- mkReg(0);
Reg#(Bit#(4)) work2VerNum <- mkReg(0);
Reg#(Vector#(20,Bit#(8))) work2Vector8 <- mkRegU;
Reg#(Vector#(20,Bit#(15))) work2Vector15 <- mkRegU;
Reg#(Vector#(16,Bit#(1))) resultReady <- mkReg(replicate(0));
Reg#(Bool) work2Done <- mkReg(False);
Reg#(Bool) work8x8Done <- mkReg(False);
Reg#(Bit#(2)) outBlockNum <- mkReg(0);
Reg#(Bit#(2)) outPixelNum <- mkReg(0);
Reg#(Bool) outDone <- mkReg(False);
rule sendEndOfFrameReq( endOfFrameFlag );
endOfFrameFlag <= False;
memReqQ.enq(IPLoadEndFrame);
endrule
rule loadLuma( reqfifoLoad.first() matches tagged IPLuma .reqdata &&& !endOfFrameFlag );
Bit#(2) xfracl = reqdata.mvhor[1:0];
Bit#(2) yfracl = reqdata.mvver[1:0];
Bit#(2) offset = reqdata.mvhor[3:2];
Bool twoStage = (xfracl==1||xfracl==3) && (yfracl==1||yfracl==3);
Bool horInter = (twoStage ? loadStage==1 : xfracl!=0);
Bool verInter = (twoStage ? loadStage==0 : yfracl!=0);
Bit#(2) offset2 = reqdata.mvhor[3:2] + ((twoStage&&verInter&&xfracl==3) ? 1 : 0);
Bit#(1) horOut = 0;
Bit#(TAdd#(PicWidthSz,2)) horAddr;
Bit#(TAdd#(PicHeightSz,4)) verAddr;
Bit#(TAdd#(PicWidthSz,12)) horTemp = zeroExtend({reqdata.hor,2'b00}) + zeroExtend({loadHorNum,2'b00}) + (xfracl==3&&(yfracl==1||yfracl==3)&&loadStage==0 ? 1 : 0);
Bit#(TAdd#(PicHeightSz,10)) verTemp = zeroExtend(reqdata.ver) + zeroExtend(loadVerNum) + (yfracl==3&&(xfracl==1||xfracl==3)&&loadStage==1 ? 1 : 0);
Bit#(13) mvhortemp = signExtend(reqdata.mvhor[13:2])-(horInter?2:0);
Bit#(11) mvvertemp = signExtend(reqdata.mvver[11:2])-(verInter?2:0);
if(mvhortemp[12]==1 && zeroExtend(0-mvhortemp)>horTemp)
begin
horAddr = 0;
horOut = 1;
end
else
begin
horTemp = horTemp + signExtend(mvhortemp);
if(horTemp>=zeroExtend({picWidth,4'b0000}))
begin
horAddr = {picWidth-1,2'b11};
horOut = 1;
end
else
horAddr = truncate(horTemp>>2);
end
if(mvvertemp[10]==1 && zeroExtend(0-mvvertemp)>verTemp)
verAddr = 0;
else
begin
verTemp = verTemp + signExtend(mvvertemp);
if(verTemp>=zeroExtend({picHeight,4'b0000}))
verAddr = {picHeight-1,4'b1111};
else
verAddr = truncate(verTemp);
end
memReqQ.enq(IPLoadLuma {refIdx:reqdata.refIdx,horOutOfBounds:horOut,hor:horAddr,ver:verAddr});
Bool verFirst = twoStage || (yfracl==2&&(xfracl==1||xfracl==3));
Bit#(2) loadHorNumMax = (reqdata.bt==IP8x8||reqdata.bt==IP8x4 ? 1 : 0) + (horInter ? 2 : (offset2==0 ? 0 : 1));
Bit#(4) loadVerNumMax = (reqdata.bt==IP8x8||reqdata.bt==IP4x8 ? 7 : 3) + (verInter ? 5 : 0);
// It would appear that we are collecting memory requests here, or at least we're adjusting
// the memory addresses.
if(verFirst)
begin
if(loadVerNum < loadVerNumMax)
loadVerNum <= loadVerNum+1;
else
begin
loadVerNum <= 0;
if(loadHorNum < loadHorNumMax)
begin
if(loadStage == 1)
begin
offset = offset + (xfracl==3 ? 1 : 0);
if(!(offset==1 || (xfracl==3 && offset==2)))
loadHorNum <= loadHorNumMax;
else
begin
loadHorNum <= 0;
loadStage <= 0;
reqfifoLoad.deq();
end
end
else
loadHorNum <= loadHorNum+1;
end
else
begin
if(twoStage && loadStage==0)
begin
offset = offset + (xfracl==3 ? 1 : 0);
if((xfracl==3 ? offset<3 : offset<2))
loadHorNum <= 0;
else
loadHorNum <= loadHorNumMax+1;
loadStage <= 1;
end
else
begin
loadHorNum <= 0;
loadStage <= 0;
reqfifoLoad.deq();
end
end
end
end
else
begin
if(loadHorNum < loadHorNumMax)
loadHorNum <= loadHorNum+1;
else
begin
loadHorNum <= 0;
if(loadVerNum < loadVerNumMax)
loadVerNum <= loadVerNum+1;
else
begin
loadVerNum <= 0;
reqfifoLoad.deq();
end
end
end
if(reqdata.bt==IP16x16 || reqdata.bt==IP16x8 || reqdata.bt==IP8x16)
$display( "ERROR Interpolation: loadLuma block sizes > 8x8 not supported");
$display( "Trace interpolator: loadLuma xfrac: %h yfrac: %h Hor: %h Ver: %h refIdx: %h HorAddr:%h VerAddr%h", xfracl, yfracl, loadHorNum, loadVerNum, reqdata.refIdx, horAddr, verAddr);
endrule
rule loadChroma( reqfifoLoad.first() matches tagged IPChroma .reqdata &&& !endOfFrameFlag );
Bit#(3) xfracc = reqdata.mvhor[2:0];
Bit#(3) yfracc = reqdata.mvver[2:0];
Bit#(2) offset = reqdata.mvhor[4:3]+{reqdata.hor[0],1'b0};
Bit#(1) horOut = 0;
Bit#(TAdd#(PicWidthSz,1)) horAddr;
Bit#(TAdd#(PicHeightSz,3)) verAddr;
Bit#(TAdd#(PicWidthSz,11)) horTemp = zeroExtend({reqdata.hor,1'b0}) + zeroExtend({loadHorNum,2'b00});
Bit#(TAdd#(PicHeightSz,9)) verTemp = zeroExtend(reqdata.ver) + zeroExtend(loadVerNum);
if(reqdata.mvhor[13]==1 && zeroExtend(0-reqdata.mvhor[13:3])>horTemp)
begin
horAddr = 0;
horOut = 1;
end
else
begin
horTemp = horTemp + signExtend(reqdata.mvhor[13:3]);
if(horTemp>=zeroExtend({picWidth,3'b000}))
begin
horAddr = {picWidth-1,1'b1};
horOut = 1;
end
else
horAddr = truncate(horTemp>>2);
end
if(reqdata.mvver[11]==1 && zeroExtend(0-reqdata.mvver[11:3])>verTemp)
verAddr = 0;
else
begin
verTemp = verTemp + signExtend(reqdata.mvver[11:3]);
if(verTemp>=zeroExtend({picHeight,3'b000}))
verAddr = {picHeight-1,3'b111};
else
verAddr = truncate(verTemp);
end
memReqQ.enq(IPLoadChroma {refIdx:reqdata.refIdx,uv:reqdata.uv,horOutOfBounds:horOut,hor:horAddr,ver:verAddr});
Bit#(2) loadHorNumMax = (reqdata.bt==IP4x8||reqdata.bt==IP4x4 ? (offset[1]==0||(xfracc==0&&offset!=3) ? 0 : 1) : ((reqdata.bt==IP16x16||reqdata.bt==IP16x8 ? 1 : 0) + (xfracc==0&&offset==0 ? 0 : 1)));
Bit#(4) loadVerNumMax = (reqdata.bt==IP16x16||reqdata.bt==IP8x16 ? 7 : (reqdata.bt==IP16x8||reqdata.bt==IP8x8||reqdata.bt==IP4x8 ? 3 : 1)) + (yfracc==0 ? 0 : 1);
if(loadHorNum < loadHorNumMax)
loadHorNum <= loadHorNum+1;
else
begin
loadHorNum <= 0;
if(loadVerNum < loadVerNumMax)
loadVerNum <= loadVerNum+1;
else
begin
loadVerNum <= 0;
reqfifoLoad.deq();
end
end
$display( "Trace interpolator: loadChroma xfrac: %h yfrac: %h Hor: %h Ver: %h refIdx: %h horAddr: %h verAddr: %h", xfracc, yfracc, loadHorNum, loadVerNum, reqdata.refIdx, horAddr, verAddr);
endrule
rule work1Luma ( reqfifoWork1.first() matches tagged IPWLuma .reqdata &&& !work1Done );
let xfracl = reqdata.xFracL;
let yfracl = reqdata.yFracL;
let offset = reqdata.offset;
let blockT = reqdata.bt;
Bool twoStage = (xfracl==1||xfracl==3) && (yfracl==1||yfracl==3); // are we dealing with a quarter sample
Vector#(20,Bit#(8)) work1Vector8Next = work1Vector8; // This must die.
if(memRespQ.first() matches tagged IPLoadResp .tempreaddata)
begin
memRespQ.deq();
Vector#(4,Bit#(8)) readdata = replicate(0);
readdata[0] = tempreaddata[7:0];
readdata[1] = tempreaddata[15:8];
readdata[2] = tempreaddata[23:16];
readdata[3] = tempreaddata[31:24];
//$display( "Trace interpolator: workLuma stage 0 readdata %h %h %h %h %h %h", workHorNum, workVerNum, readdata[3], readdata[2], readdata[1], readdata[0] );
Vector#(4,Bit#(8)) tempResult8 = replicate(0);
Vector#(4,Bit#(15)) tempResult15 = replicate(0);
if(xfracl==0 || yfracl==0 || xfracl==2)
begin
if(xfracl==0)//reorder
begin
for(Integer ii=0; ii<4; ii=ii+1)
begin
Bit#(2) offsetplusii = offset+fromInteger(ii);
if(offset <= 3-fromInteger(ii) && offset!=0)
tempResult8[ii] = work1Vector8[offsetplusii];
else
tempResult8[ii] = readdata[offsetplusii];
work1Vector8Next[ii] = readdata[ii];
end
for(Integer ii=0; ii<4; ii=ii+1)
tempResult15[ii] = zeroExtend({tempResult8[ii],5'b00000});
end
else//horizontal interpolation
begin
offset = offset-2;
for(Integer ii=0; ii<8; ii=ii+1)
work1Vector8Next[ii] = work1Vector8[ii+4];
for(Integer ii=0; ii<4; ii=ii+1)
begin
Bit#(4) tempIndex = fromInteger(ii) + 8 - zeroExtend(offset);
work1Vector8Next[tempIndex] = readdata[ii];
end
for(Integer ii=0; ii<4; ii=ii+1) // horizontal filtration step.
begin
tempResult15[ii] = interpolate8to15.interpolate(work1Vector8Next[ii],work1Vector8Next[ii+1],work1Vector8Next[ii+2],work1Vector8Next[ii+3],work1Vector8Next[ii+4],work1Vector8Next[ii+5]);
tempResult8[ii] = clip1y10to8(truncate((tempResult15[ii]+16)>>5));
if(xfracl == 1) // Seems to be averaging the quarter samples.
tempResult8[ii] = truncate(({1'b0,tempResult8[ii]} + {1'b0,work1Vector8Next[ii+2]} + 1) >> 1);
else if(xfracl == 3)
tempResult8[ii] = truncate(({1'b0,tempResult8[ii]} + {1'b0,work1Vector8Next[ii+3]} + 1) >> 1);
end
end
Bit#(2) workHorNumOffset = (xfracl!=0 ? 2 : (reqdata.offset==0 ? 0 : 1));
if(work1HorNum >= workHorNumOffset)
begin
Bit#(1) horAddr = truncate(work1HorNum-workHorNumOffset);
if(yfracl == 0)
begin
for(Integer ii=0; ii<4; ii=ii+1)
tempResult15[ii] = zeroExtend({tempResult8[ii],5'b00000});
end
workFile.upd({workFileFlag,work1VerNum,horAddr},tempResult15);
end
Bit#(2) workHorNumMax = (blockT==IP8x8||blockT==IP8x4 ? 1 : 0) + workHorNumOffset;
Bit#(4) workVerNumMax = (blockT==IP8x8||blockT==IP4x8 ? 7 : 3) + (yfracl!=0 ? 5 : 0);
if(work1HorNum < workHorNumMax)
work1HorNum <= work1HorNum+1;
else
begin
work1HorNum <= 0;
if(work1VerNum < workVerNumMax)
work1VerNum <= work1VerNum+1;
else
begin
work1VerNum <= 0;
work1Done <= True;
end
end
end
else if(work1Stage == 0)//vertical interpolation
begin
offset = offset + (xfracl==3&&(yfracl==1||yfracl==3) ? 1 : 0);
for(Integer ii=0; ii<4; ii=ii+1) // apply the horizontal filtration step.
tempResult15[ii] = interpolate8to15.interpolate(work1Vector8[ii],work1Vector8[ii+4],work1Vector8[ii+8],work1Vector8[ii+12],work1Vector8[ii+16],readdata[ii]);
for(Integer ii=0; ii<16; ii=ii+1) // advances the work vector
work1Vector8Next[ii] = work1Vector8[ii+4];
for(Integer ii=0; ii<4; ii=ii+1) // assigns the new work vector value
work1Vector8Next[ii+16] = readdata[ii];
Bit#(2) workHorNumMax = (blockT==IP8x8||blockT==IP8x4 ? 1 : 0) + (yfracl==2 ? 2 : (offset==0 ? 0 : 1));
Bit#(4) workVerNumMax = (blockT==IP8x8||blockT==IP4x8 ? 7 : 3) + 5;
Bit#(2) horAddr = work1HorNum;
Bit#(3) verAddr = truncate(work1VerNum-5);
if(work1VerNum > 4)
begin
workFile.upd({workFileFlag,verAddr,horAddr},tempResult15);
//$display( "Trace interpolator: workLuma stage 0 result %h %h %h %h %h %h %h", workHorNum, workVerNum, {verAddr,horAddr}, tempResult15[3], tempResult15[2], tempResult15[1], tempResult15[0]);
end
if(twoStage)
begin
Bit#(2) storeHorAddr = work1HorNum;
Bit#(4) storeVerAddr = work1VerNum;
if((xfracl==3 ? offset<3 : offset<2))
storeHorAddr = storeHorAddr+1;
if(yfracl==3)
storeVerAddr = storeVerAddr-3;
else
storeVerAddr = storeVerAddr-2;
if(storeVerAddr < 8)
storeFile.upd({workFileFlag,storeVerAddr[2:0],storeHorAddr},readdata);
end
if(work1VerNum < workVerNumMax)
work1VerNum <= work1VerNum+1;
else
begin
work1VerNum <= 0;
if(work1HorNum < workHorNumMax)
work1HorNum <= work1HorNum+1;
else
begin
if(twoStage)
begin
work1Stage <= 1;
if((xfracl==3 ? offset<3 : offset<2))
work1HorNum <= 0;
else
work1HorNum <= workHorNumMax+1;
end
else
begin
work1HorNum <= 0;
work1Done <= True;
end
end
end
end
else//second stage of twoStage
begin
storeFile.upd({workFileFlag,work1VerNum[2:0],work1HorNum},readdata);
Bit#(2) workHorNumMax = (blockT==IP8x8||blockT==IP8x4 ? 1 : 0) + 2;
Bit#(4) workVerNumMax = (blockT==IP8x8||blockT==IP4x8 ? 7 : 3);
if(work1VerNum < workVerNumMax)
work1VerNum <= work1VerNum+1;
else
begin
work1VerNum <= 0;
offset = offset + (xfracl==3 ? 1 : 0);
if(work1HorNum<workHorNumMax && !(offset==1 || (xfracl==3 && offset==2)))
work1HorNum <= workHorNumMax;
else
begin
work1HorNum <= 0;
work1Stage <= 0;
work1Done <= True;
end
end
end
end
work1Vector8 <= work1Vector8Next;
$display( "Trace interpolator: work1Luma xfrac: %h yfrac: %h horNum: %h verNum: %h offset: %h workStage: %h", xfracl, yfracl, work1HorNum, work1VerNum, offset, work1Stage);
endrule
rule work2Luma ( reqregWork2 matches tagged Valid .vdata &&& vdata matches tagged IPWLuma .reqdata &&& !work2Done &&& !work8x8Done );
let xfracl = reqdata.xFracL;
let yfracl = reqdata.yFracL;
let offset = reqdata.offset;
let blockT = reqdata.bt;
Vector#(20,Bit#(8)) work2Vector8Next = work2Vector8;
Vector#(20,Bit#(15)) work2Vector15Next = work2Vector15;
Vector#(16,Bit#(1)) resultReadyNext = resultReady;
Vector#(4,Bit#(8)) tempResult8 = replicate(0);
Vector#(4,Bit#(15)) readdata = replicate(0);
if(yfracl==0)
begin
readdata = workFile.sub({(1-workFileFlag),1'b0,work2VerNum[1],work2HorNum,work2VerNum[0]});
for(Integer ii=0; ii<4; ii=ii+1)
tempResult8[ii] = (readdata[ii])[12:5];
resultFile.upd({work2VerNum[1],work2HorNum,work2VerNum[0]},tempResult8);
resultReadyNext[{work2VerNum[1],work2HorNum,work2VerNum[0]}] = 1;
work2HorNum <= work2HorNum+1;
if(work2HorNum == 3)
begin
if(work2VerNum == 3)
begin
work2VerNum <= 0;
work2Done <= True;
if(((blockT==IP4x8 || blockT==IP8x4) && work2SubMbPart==0) || (blockT==IP4x4 && work2SubMbPart<3))
work2SubMbPart <= work2SubMbPart+1;
else
begin
work2SubMbPart <= 0;
work8x8Done <= True;
end
end
else
work2VerNum <= work2VerNum+1;
end
end
else if(xfracl==0 || xfracl==2)//vertical interpolation
begin
readdata = workFile.sub({(1-workFileFlag),work2VerNum,work2HorNum[0]});
for(Integer ii=0; ii<4; ii=ii+1)
begin
tempResult8[ii] = interpolate15to8.interpolate(work2Vector15[ii],work2Vector15[ii+4],work2Vector15[ii+8],work2Vector15[ii+12],work2Vector15[ii+16],readdata[ii]);
if(yfracl == 1)
tempResult8[ii] = truncate(({1'b0,tempResult8[ii]} + {1'b0,clip1y10to8(truncate((work2Vector15[ii+8]+16)>>5))} + 1) >> 1);
else if(yfracl == 3)
tempResult8[ii] = truncate(({1'b0,tempResult8[ii]} + {1'b0,clip1y10to8(truncate((work2Vector15[ii+12]+16)>>5))} + 1) >> 1);
end
for(Integer ii=0; ii<16; ii=ii+1)
work2Vector15Next[ii] = work2Vector15[ii+4];
for(Integer ii=0; ii<4; ii=ii+1)
work2Vector15Next[ii+16] = readdata[ii];
Bit#(2) workHorNumMax = 1;
Bit#(4) workVerNumMax = (blockT==IP8x8||blockT==IP4x8 ? 7 : 3) + 5;
if(work2VerNum > 4)
begin
Bit#(1) horAddr = truncate(work2HorNum);
Bit#(3) verAddr = truncate(work2VerNum-5);
horAddr = horAddr + ((blockT==IP4x8&&work2SubMbPart==1)||(blockT==IP4x4&&work2SubMbPart[0]==1) ? 1 : 0);
verAddr = verAddr + ((blockT==IP8x4&&work2SubMbPart==1)||(blockT==IP4x4&&work2SubMbPart[1]==1) ? 4 : 0);
resultFile.upd({verAddr,horAddr},tempResult8);
resultReadyNext[{verAddr,horAddr}] = 1;
end
if(work2VerNum < workVerNumMax)
work2VerNum <= work2VerNum+1;
else
begin
work2VerNum <= 0;
if(work2HorNum < workHorNumMax)
work2HorNum <= work2HorNum+1;
else
begin
work2HorNum <= 0;
work2Done <= True;
if(((blockT==IP4x8 || blockT==IP8x4) && work2SubMbPart==0) || (blockT==IP4x4 && work2SubMbPart<3))
work2SubMbPart <= work2SubMbPart+1;
else
begin
work2SubMbPart <= 0;
work8x8Done <= True;
end
end
end
end
else//horizontal interpolation
begin
offset = offset-2;
if(yfracl == 2)
begin
readdata = workFile.sub({(1-workFileFlag),work2VerNum[2:0],work2HorNum});
for(Integer ii=0; ii<8; ii=ii+1)
work2Vector15Next[ii] = work2Vector15[ii+4];
for(Integer ii=0; ii<4; ii=ii+1)
begin
Bit#(4) tempIndex = fromInteger(ii) + 8 - zeroExtend(offset);
work2Vector15Next[tempIndex] = readdata[ii];
end
for(Integer ii=0; ii<4; ii=ii+1)
begin
tempResult8[ii] = interpolate15to8.interpolate(work2Vector15Next[ii],work2Vector15Next[ii+1],work2Vector15Next[ii+2],work2Vector15Next[ii+3],work2Vector15Next[ii+4],work2Vector15Next[ii+5]);
if(xfracl == 1)
tempResult8[ii] = truncate(({1'b0,tempResult8[ii]} + {1'b0,clip1y10to8(truncate((work2Vector15Next[ii+2]+16)>>5))} + 1) >> 1);
else if(xfracl == 3)
tempResult8[ii] = truncate(({1'b0,tempResult8[ii]} + {1'b0,clip1y10to8(truncate((work2Vector15Next[ii+3]+16)>>5))} + 1) >> 1);
end
end
else
begin
Vector#(4,Bit#(8)) readdata8 = storeFile.sub({(1-workFileFlag),work2VerNum[2:0],work2HorNum});
for(Integer ii=0; ii<8; ii=ii+1)
work2Vector8Next[ii] = work2Vector8[ii+4];
for(Integer ii=0; ii<4; ii=ii+1)
begin
Bit#(4) tempIndex = fromInteger(ii) + 8 - zeroExtend(offset);
work2Vector8Next[tempIndex] = readdata8[ii];
end
Vector#(4,Bit#(15)) tempResult15 = replicate(0);
for(Integer ii=0; ii<4; ii=ii+1)
begin
tempResult15[ii] = interpolate8to15.interpolate(work2Vector8Next[ii],work2Vector8Next[ii+1],work2Vector8Next[ii+2],work2Vector8Next[ii+3],work2Vector8Next[ii+4],work2Vector8Next[ii+5]);
tempResult8[ii] = clip1y10to8(truncate((tempResult15[ii]+16)>>5));
end
Bit#(2) verOffset;
Vector#(4,Bit#(15)) verResult15 = replicate(0);
if(xfracl == 1)
verOffset = reqdata.offset;
else
verOffset = reqdata.offset+1;
readdata = workFile.sub({(1-workFileFlag),work2VerNum[2:0],(work2HorNum-2+(verOffset==0?0:1))});
for(Integer ii=0; ii<4; ii=ii+1)
begin
Bit#(2) offsetplusii = verOffset+fromInteger(ii);
if(verOffset <= 3-fromInteger(ii) && verOffset!=0)
verResult15[ii] = work2Vector15[offsetplusii];
else
verResult15[ii] = readdata[offsetplusii];
work2Vector15Next[ii] = readdata[ii];
end
for(Integer ii=0; ii<4; ii=ii+1)
begin
Bit#(9) tempVal = zeroExtend(clip1y10to8(truncate((verResult15[ii]+16)>>5)));
tempResult8[ii] = truncate((tempVal+zeroExtend(tempResult8[ii])+1)>>1);
end
end
if(work2HorNum >= 2)
begin
Bit#(1) horAddr = truncate(work2HorNum-2);
Bit#(3) verAddr = truncate(work2VerNum);
horAddr = horAddr + ((blockT==IP4x8&&work2SubMbPart==1)||(blockT==IP4x4&&work2SubMbPart[0]==1) ? 1 : 0);
verAddr = verAddr + ((blockT==IP8x4&&work2SubMbPart==1)||(blockT==IP4x4&&work2SubMbPart[1]==1) ? 4 : 0);
resultFile.upd({verAddr,horAddr},tempResult8);
resultReadyNext[{verAddr,horAddr}] = 1;
//$display( "Trace interpolator: workLuma stage 1 result %h %h %h %h %h %h %h %h", workHorNum, workVerNum, {verAddr,horAddr}, tempResult8[3], tempResult8[2], tempResult8[1], tempResult8[0], pack(resultReadyNext));
end
Bit#(2) workHorNumMax = (blockT==IP8x8||blockT==IP8x4 ? 1 : 0) + 2;
Bit#(4) workVerNumMax = (blockT==IP8x8||blockT==IP4x8 ? 7 : 3);
if(work2HorNum < workHorNumMax)
work2HorNum <= work2HorNum+1;
else
begin
work2HorNum <= 0;
if(work2VerNum < workVerNumMax)
work2VerNum <= work2VerNum+1;
else
begin
work2VerNum <= 0;
work2Done <= True;
if(((blockT==IP4x8 || blockT==IP8x4) && work2SubMbPart==0) || (blockT==IP4x4 && work2SubMbPart<3))
work2SubMbPart <= work2SubMbPart+1;
else
begin
work2SubMbPart <= 0;
work8x8Done <= True;
end
end
end
end
work2Vector8 <= work2Vector8Next;
work2Vector15 <= work2Vector15Next;
resultReady <= resultReadyNext;
$display( "Trace interpolator: work2Luma xfrac: %h yfrac: %h horNum: %h verNum: %h offset: %h", xfracl, yfracl, work2HorNum, work2VerNum, offset);
endrule
rule work1Chroma ( reqfifoWork1.first() matches tagged IPWChroma .reqdata &&& !work1Done );
Bit#(4) xfracc = zeroExtend(reqdata.xFracC);
Bit#(4) yfracc = zeroExtend(reqdata.yFracC);
let offset = reqdata.offset;
let blockT = reqdata.bt;
Vector#(20,Bit#(8)) work1Vector8Next = work1Vector8;
if(memRespQ.first() matches tagged IPLoadResp .tempreaddata)
begin
memRespQ.deq();
Vector#(4,Bit#(8)) readdata = replicate(0);
readdata[0] = tempreaddata[7:0];
readdata[1] = tempreaddata[15:8];
readdata[2] = tempreaddata[23:16];
readdata[3] = tempreaddata[31:24];
Vector#(5,Bit#(8)) tempWork8 = replicate(0);
Vector#(5,Bit#(8)) tempPrev8 = replicate(0);
Vector#(4,Bit#(8)) tempResult8 = replicate(0);
Bool resultReadyFlag = False;
for(Integer ii=0; ii<4; ii=ii+1)
begin
Bit#(2) offsetplusii = offset+fromInteger(ii);
if(offset <= 3-fromInteger(ii) && !((blockT==IP4x8||blockT==IP4x4)&&(offset[1]==0||(xfracc==0&&offset!=3))) && !(xfracc==0&&offset==0))
tempWork8[ii] = work1Vector8[offsetplusii];
else
tempWork8[ii] = readdata[offsetplusii];
work1Vector8Next[ii] = readdata[ii];
end
tempWork8[4] = readdata[offset];
// deals with the row major offsets
if((blockT==IP16x8 || blockT==IP16x16) && work1HorNum==(xfracc==0&&offset==0 ? 1 : 2))
begin
for(Integer ii=0; ii<5; ii=ii+1)
begin
tempPrev8[ii] = work1Vector8[ii+9];
work1Vector8Next[ii+9] = tempWork8[ii];
end
end
else
begin
for(Integer ii=0; ii<5; ii=ii+1)
tempPrev8[ii] = work1Vector8[ii+4];
if(work1HorNum==(xfracc==0&&offset==0 ? 0 : 1) || ((blockT==IP4x8||blockT==IP4x4)&&(offset[1]==0||(xfracc==0&&offset!=3))))
begin
for(Integer ii=0; ii<5; ii=ii+1)
work1Vector8Next[ii+4] = tempWork8[ii];
end
end
if(yfracc==0)
begin
for(Integer ii=0; ii<5; ii=ii+1)
tempPrev8[ii] = tempWork8[ii];
end
// Apply filter?
for(Integer ii=0; ii<4; ii=ii+1)
begin
$display("Trace interpolator: Applying filter");
Bit#(14) tempVal = zeroExtend((8-xfracc))*zeroExtend((8-yfracc))*zeroExtend(tempPrev8[ii]);
tempVal = tempVal + zeroExtend(xfracc)*zeroExtend((8-yfracc))*zeroExtend(tempPrev8[ii+1]);
tempVal = tempVal + zeroExtend((8-xfracc))*zeroExtend(yfracc)*zeroExtend(tempWork8[ii]);
tempVal = tempVal + zeroExtend(xfracc)*zeroExtend(yfracc)*zeroExtend(tempWork8[ii+1]);
tempResult8[ii] = truncate((tempVal+32)>>6);
end
if(work1VerNum > 0 || yfracc==0)
begin
if(blockT==IP4x8 || blockT==IP4x4)
begin
Bit#(5) tempIndex = 10 + zeroExtend(work1VerNum<<1);
work1Vector8Next[tempIndex] = tempResult8[0];
work1Vector8Next[tempIndex+1] = tempResult8[1];
tempResult8[2] = tempResult8[0];
tempResult8[3] = tempResult8[1];
tempResult8[0] = work1Vector8[tempIndex];
tempResult8[1] = work1Vector8[tempIndex+1];
if((work1HorNum>0 || offset[1]==0) && work1SubMbPart[0]==1)
resultReadyFlag = True;
end
else
begin
if(work1HorNum>0 || (xfracc==0 && offset==0))
resultReadyFlag = True;
end
end
if(resultReadyFlag)
begin
Bit#(1) horAddr = ((blockT==IP4x8 || blockT==IP4x4) ? 0 : truncate(((xfracc==0 && offset==0) ? work1HorNum : work1HorNum-1)));
Bit#(3) verAddr = truncate((yfracc==0 ? work1VerNum : work1VerNum-1));
horAddr = horAddr + ((blockT==IP16x8||blockT==IP16x16) ? 0 : work1MbPart[0]);
verAddr = verAddr + ((blockT==IP8x16||blockT==IP16x16) ? 0 : ((blockT==IP16x8) ? {work1MbPart[0],2'b00} : {work1MbPart[1],2'b00}));
verAddr = verAddr + ((blockT==IP8x4&&work1SubMbPart==1)||(blockT==IP4x4&&work1SubMbPart[1]==1) ? 2 : 0);
storeFile.upd({workFileFlag,1'b0,verAddr,horAddr},tempResult8);
end
Bit#(2) workHorNumMax = (blockT==IP4x8||blockT==IP4x4 ? (offset[1]==0||(xfracc==0&&offset!=3) ? 0 : 1) : ((blockT==IP16x16||blockT==IP16x8 ? 1 : 0) + (xfracc==0&&offset==0 ? 0 : 1)));
Bit#(4) workVerNumMax = (blockT==IP16x16||blockT==IP8x16 ? 7 : (blockT==IP16x8||blockT==IP8x8||blockT==IP4x8 ? 3 : 1)) + (yfracc==0 ? 0 : 1);
if(work1HorNum < workHorNumMax)
work1HorNum <= work1HorNum+1;
else
begin
work1HorNum <= 0;
if(work1VerNum < workVerNumMax)
work1VerNum <= work1VerNum+1;
else
begin
Bool allDone = False;
work1VerNum <= 0;
if(((blockT==IP4x8 || blockT==IP8x4) && work1SubMbPart==0) || (blockT==IP4x4 && work1SubMbPart<3))
work1SubMbPart <= work1SubMbPart+1;
else
begin
work1SubMbPart <= 0;
if(((blockT==IP16x8 || blockT==IP8x16) && work1MbPart==0) || (!(blockT==IP16x8 || blockT==IP8x16 || blockT==IP16x16) && work1MbPart<3))
work1MbPart <= work1MbPart+1;
else
begin
work1MbPart <= 0;
work1Done <= True;
allDone = True;
end
end
if(!allDone)
begin
reqfifoWork1.deq();
$display("Trace Interpolator: work1Chroma finished");
end
end
end
end
work1Vector8 <= work1Vector8Next;
case (blockT)
IP16x16: $display("Trace Interpolator: chroma 16x16");
IP16x8: $display("Trace Interpolator: chroma 16x8");
IP8x16: $display("Trace Interpolator: chroma 8x16");
IP8x8: $display("Trace Interpolator: chroma 8x8");
IP8x4: $display("Trace Interpolator: chroma 8x4");
IP4x8: $display("Trace Interpolator: chroma 4x8");
IP4x4: $display("Trace Interpolator: chroma 4x4");
endcase
$display( "Trace interpolator: work1Chroma xfracc: %h yfracc: %h Hor: %h Ver: %h offset: %h",
xfracc, yfracc, work1HorNum, work1VerNum, offset);
endrule
rule work2Chroma ( reqregWork2 matches tagged Valid .vdata &&& vdata matches tagged IPWChroma .reqdata &&& !work2Done &&& !work8x8Done );
Vector#(16,Bit#(1)) resultReadyNext = resultReady;
resultFile.upd({work2VerNum[1],work2HorNum,work2VerNum[0]},storeFile.sub({(1-workFileFlag),1'b0,work2VerNum[1],work2HorNum,work2VerNum[0]}));
resultReadyNext[{work2VerNum[1],work2HorNum,work2VerNum[0]}] = 1;
work2HorNum <= work2HorNum+1;
if(work2HorNum == 3)
begin
if(work2VerNum == 3)
begin
work2VerNum <= 0;
work2Done <= True;
work8x8Done <= True;
end
else
work2VerNum <= work2VerNum+1;
end
resultReady <= resultReadyNext;
$display( "Trace interpolator: work2Chroma Hor: %h Ver: %h", work2HorNum, work2VerNum);
endrule
rule outputing( !outDone && resultReady[{outBlockNum[1],outPixelNum,outBlockNum[0]}]==1 );
outfifo.enq(resultFile.sub({outBlockNum[1],outPixelNum,outBlockNum[0]}));
outPixelNum <= outPixelNum+1;
if(outPixelNum == 3)
begin
outBlockNum <= outBlockNum+1;
if(outBlockNum == 3)
outDone <= True;
end
$display( "Trace interpolator: outputing BlockNum: %h pixelNum: %h", outBlockNum, outPixelNum);
endrule
// These two rules complete the processing step, and
rule switching( work1Done && (work2Done || reqregWork2==Invalid) && !work8x8Done);
work1Done <= False;
work2Done <= False;
reqregWork2 <= (tagged Valid reqfifoWork1.first());
workFileFlag <= 1-workFileFlag;
$display("Trace Interpolator: work1 finished");
reqfifoWork1.deq();
$display( "Trace interpolator: switching blockNum: %h pixelNum: %h", outBlockNum, outPixelNum);
endrule
// this rule is kind of one of the last to run
rule switching8x8( work1Done && (work2Done || reqregWork2==Invalid) && work8x8Done && outDone);
outDone <= False;
work8x8Done <= False;
resultReady <= replicate(0);
work1Done <= False;
work2Done <= False;
reqregWork2 <= (tagged Valid reqfifoWork1.first());
workFileFlag <= 1-workFileFlag;
reqfifoWork1.deq();
$display("Trace Interpolator: work1 finished");
$display( "Trace interpolator: switching8x8 blockNum: %h pixelNum: %h", outBlockNum, outPixelNum);
endrule
method Action setPicWidth( Bit#(PicWidthSz) newPicWidth );
picWidth <= newPicWidth;
endmethod
method Action setPicHeight( Bit#(PicHeightSz) newPicHeight );
picHeight <= newPicHeight;
endmethod
method Action request( InterpolatorIT inputdata );
reqfifoLoad.enq(inputdata);
if(inputdata matches tagged IPLuma .indata)
reqfifoWork1.enq(IPWLuma {xFracL:indata.mvhor[1:0],yFracL:indata.mvver[1:0],offset:indata.mvhor[3:2],bt:indata.bt});
else if(inputdata matches tagged IPChroma .indata)
reqfifoWork1.enq(IPWChroma {xFracC:indata.mvhor[2:0],yFracC:indata.mvver[2:0],offset:indata.mvhor[4:3]+{indata.hor[0],1'b0},bt:indata.bt});
endmethod
method Vector#(4,Bit#(8)) first();
return outfifo.first();
endmethod
method Action deq();
outfifo.deq();
endmethod
method Action endOfFrame();
endOfFrameFlag <= True;
endmethod
interface Client mem_client;
interface Get request = fifoToGet(memReqQ);
interface Put response = fifoToPut(memRespQ);
endinterface
endmodule