Chris@4: Chris@4: /*-------------------------------------------------------------*/ Chris@4: /*--- Compression machinery (not incl block sorting) ---*/ Chris@4: /*--- compress.c ---*/ Chris@4: /*-------------------------------------------------------------*/ Chris@4: Chris@4: /* ------------------------------------------------------------------ Chris@4: This file is part of bzip2/libbzip2, a program and library for Chris@4: lossless, block-sorting data compression. Chris@4: Chris@4: bzip2/libbzip2 version 1.0.6 of 6 September 2010 Chris@4: Copyright (C) 1996-2010 Julian Seward Chris@4: Chris@4: Please read the WARNING, DISCLAIMER and PATENTS sections in the Chris@4: README file. Chris@4: Chris@4: This program is released under the terms of the license contained Chris@4: in the file LICENSE. Chris@4: ------------------------------------------------------------------ */ Chris@4: Chris@4: Chris@4: /* CHANGES Chris@4: 0.9.0 -- original version. Chris@4: 0.9.0a/b -- no changes in this file. Chris@4: 0.9.0c -- changed setting of nGroups in sendMTFValues() Chris@4: so as to do a bit better on small files Chris@4: */ Chris@4: Chris@4: #include "bzlib_private.h" Chris@4: Chris@4: Chris@4: /*---------------------------------------------------*/ Chris@4: /*--- Bit stream I/O ---*/ Chris@4: /*---------------------------------------------------*/ Chris@4: Chris@4: /*---------------------------------------------------*/ Chris@4: void BZ2_bsInitWrite ( EState* s ) Chris@4: { Chris@4: s->bsLive = 0; Chris@4: s->bsBuff = 0; Chris@4: } Chris@4: Chris@4: Chris@4: /*---------------------------------------------------*/ Chris@4: static Chris@4: void bsFinishWrite ( EState* s ) Chris@4: { Chris@4: while (s->bsLive > 0) { Chris@4: s->zbits[s->numZ] = (UChar)(s->bsBuff >> 24); Chris@4: s->numZ++; Chris@4: s->bsBuff <<= 8; Chris@4: s->bsLive -= 8; Chris@4: } Chris@4: } Chris@4: Chris@4: Chris@4: /*---------------------------------------------------*/ Chris@4: #define bsNEEDW(nz) \ Chris@4: { \ Chris@4: while (s->bsLive >= 8) { \ Chris@4: s->zbits[s->numZ] \ Chris@4: = (UChar)(s->bsBuff >> 24); \ Chris@4: s->numZ++; \ Chris@4: s->bsBuff <<= 8; \ Chris@4: s->bsLive -= 8; \ Chris@4: } \ Chris@4: } Chris@4: Chris@4: Chris@4: /*---------------------------------------------------*/ Chris@4: static Chris@4: __inline__ Chris@4: void bsW ( EState* s, Int32 n, UInt32 v ) Chris@4: { Chris@4: bsNEEDW ( n ); Chris@4: s->bsBuff |= (v << (32 - s->bsLive - n)); Chris@4: s->bsLive += n; Chris@4: } Chris@4: Chris@4: Chris@4: /*---------------------------------------------------*/ Chris@4: static Chris@4: void bsPutUInt32 ( EState* s, UInt32 u ) Chris@4: { Chris@4: bsW ( s, 8, (u >> 24) & 0xffL ); Chris@4: bsW ( s, 8, (u >> 16) & 0xffL ); Chris@4: bsW ( s, 8, (u >> 8) & 0xffL ); Chris@4: bsW ( s, 8, u & 0xffL ); Chris@4: } Chris@4: Chris@4: Chris@4: /*---------------------------------------------------*/ Chris@4: static Chris@4: void bsPutUChar ( EState* s, UChar c ) Chris@4: { Chris@4: bsW( s, 8, (UInt32)c ); Chris@4: } Chris@4: Chris@4: Chris@4: /*---------------------------------------------------*/ Chris@4: /*--- The back end proper ---*/ Chris@4: /*---------------------------------------------------*/ Chris@4: Chris@4: /*---------------------------------------------------*/ Chris@4: static Chris@4: void makeMaps_e ( EState* s ) Chris@4: { Chris@4: Int32 i; Chris@4: s->nInUse = 0; Chris@4: for (i = 0; i < 256; i++) Chris@4: if (s->inUse[i]) { Chris@4: s->unseqToSeq[i] = s->nInUse; Chris@4: s->nInUse++; Chris@4: } Chris@4: } Chris@4: Chris@4: Chris@4: /*---------------------------------------------------*/ Chris@4: static Chris@4: void generateMTFValues ( EState* s ) Chris@4: { Chris@4: UChar yy[256]; Chris@4: Int32 i, j; Chris@4: Int32 zPend; Chris@4: Int32 wr; Chris@4: Int32 EOB; Chris@4: Chris@4: /* Chris@4: After sorting (eg, here), Chris@4: s->arr1 [ 0 .. s->nblock-1 ] holds sorted order, Chris@4: and Chris@4: ((UChar*)s->arr2) [ 0 .. s->nblock-1 ] Chris@4: holds the original block data. Chris@4: Chris@4: The first thing to do is generate the MTF values, Chris@4: and put them in Chris@4: ((UInt16*)s->arr1) [ 0 .. s->nblock-1 ]. Chris@4: Because there are strictly fewer or equal MTF values Chris@4: than block values, ptr values in this area are overwritten Chris@4: with MTF values only when they are no longer needed. Chris@4: Chris@4: The final compressed bitstream is generated into the Chris@4: area starting at Chris@4: (UChar*) (&((UChar*)s->arr2)[s->nblock]) Chris@4: Chris@4: These storage aliases are set up in bzCompressInit(), Chris@4: except for the last one, which is arranged in Chris@4: compressBlock(). Chris@4: */ Chris@4: UInt32* ptr = s->ptr; Chris@4: UChar* block = s->block; Chris@4: UInt16* mtfv = s->mtfv; Chris@4: Chris@4: makeMaps_e ( s ); Chris@4: EOB = s->nInUse+1; Chris@4: Chris@4: for (i = 0; i <= EOB; i++) s->mtfFreq[i] = 0; Chris@4: Chris@4: wr = 0; Chris@4: zPend = 0; Chris@4: for (i = 0; i < s->nInUse; i++) yy[i] = (UChar) i; Chris@4: Chris@4: for (i = 0; i < s->nblock; i++) { Chris@4: UChar ll_i; Chris@4: AssertD ( wr <= i, "generateMTFValues(1)" ); Chris@4: j = ptr[i]-1; if (j < 0) j += s->nblock; Chris@4: ll_i = s->unseqToSeq[block[j]]; Chris@4: AssertD ( ll_i < s->nInUse, "generateMTFValues(2a)" ); Chris@4: Chris@4: if (yy[0] == ll_i) { Chris@4: zPend++; Chris@4: } else { Chris@4: Chris@4: if (zPend > 0) { Chris@4: zPend--; Chris@4: while (True) { Chris@4: if (zPend & 1) { Chris@4: mtfv[wr] = BZ_RUNB; wr++; Chris@4: s->mtfFreq[BZ_RUNB]++; Chris@4: } else { Chris@4: mtfv[wr] = BZ_RUNA; wr++; Chris@4: s->mtfFreq[BZ_RUNA]++; Chris@4: } Chris@4: if (zPend < 2) break; Chris@4: zPend = (zPend - 2) / 2; Chris@4: }; Chris@4: zPend = 0; Chris@4: } Chris@4: { Chris@4: register UChar rtmp; Chris@4: register UChar* ryy_j; Chris@4: register UChar rll_i; Chris@4: rtmp = yy[1]; Chris@4: yy[1] = yy[0]; Chris@4: ryy_j = &(yy[1]); Chris@4: rll_i = ll_i; Chris@4: while ( rll_i != rtmp ) { Chris@4: register UChar rtmp2; Chris@4: ryy_j++; Chris@4: rtmp2 = rtmp; Chris@4: rtmp = *ryy_j; Chris@4: *ryy_j = rtmp2; Chris@4: }; Chris@4: yy[0] = rtmp; Chris@4: j = ryy_j - &(yy[0]); Chris@4: mtfv[wr] = j+1; wr++; s->mtfFreq[j+1]++; Chris@4: } Chris@4: Chris@4: } Chris@4: } Chris@4: Chris@4: if (zPend > 0) { Chris@4: zPend--; Chris@4: while (True) { Chris@4: if (zPend & 1) { Chris@4: mtfv[wr] = BZ_RUNB; wr++; Chris@4: s->mtfFreq[BZ_RUNB]++; Chris@4: } else { Chris@4: mtfv[wr] = BZ_RUNA; wr++; Chris@4: s->mtfFreq[BZ_RUNA]++; Chris@4: } Chris@4: if (zPend < 2) break; Chris@4: zPend = (zPend - 2) / 2; Chris@4: }; Chris@4: zPend = 0; Chris@4: } Chris@4: Chris@4: mtfv[wr] = EOB; wr++; s->mtfFreq[EOB]++; Chris@4: Chris@4: s->nMTF = wr; Chris@4: } Chris@4: Chris@4: Chris@4: /*---------------------------------------------------*/ Chris@4: #define BZ_LESSER_ICOST 0 Chris@4: #define BZ_GREATER_ICOST 15 Chris@4: Chris@4: static Chris@4: void sendMTFValues ( EState* s ) Chris@4: { Chris@4: Int32 v, t, i, j, gs, ge, totc, bt, bc, iter; Chris@4: Int32 nSelectors, alphaSize, minLen, maxLen, selCtr; Chris@4: Int32 nGroups, nBytes; Chris@4: Chris@4: /*-- Chris@4: UChar len [BZ_N_GROUPS][BZ_MAX_ALPHA_SIZE]; Chris@4: is a global since the decoder also needs it. Chris@4: Chris@4: Int32 code[BZ_N_GROUPS][BZ_MAX_ALPHA_SIZE]; Chris@4: Int32 rfreq[BZ_N_GROUPS][BZ_MAX_ALPHA_SIZE]; Chris@4: are also globals only used in this proc. Chris@4: Made global to keep stack frame size small. Chris@4: --*/ Chris@4: Chris@4: Chris@4: UInt16 cost[BZ_N_GROUPS]; Chris@4: Int32 fave[BZ_N_GROUPS]; Chris@4: Chris@4: UInt16* mtfv = s->mtfv; Chris@4: Chris@4: if (s->verbosity >= 3) Chris@4: VPrintf3( " %d in block, %d after MTF & 1-2 coding, " Chris@4: "%d+2 syms in use\n", Chris@4: s->nblock, s->nMTF, s->nInUse ); Chris@4: Chris@4: alphaSize = s->nInUse+2; Chris@4: for (t = 0; t < BZ_N_GROUPS; t++) Chris@4: for (v = 0; v < alphaSize; v++) Chris@4: s->len[t][v] = BZ_GREATER_ICOST; Chris@4: Chris@4: /*--- Decide how many coding tables to use ---*/ Chris@4: AssertH ( s->nMTF > 0, 3001 ); Chris@4: if (s->nMTF < 200) nGroups = 2; else Chris@4: if (s->nMTF < 600) nGroups = 3; else Chris@4: if (s->nMTF < 1200) nGroups = 4; else Chris@4: if (s->nMTF < 2400) nGroups = 5; else Chris@4: nGroups = 6; Chris@4: Chris@4: /*--- Generate an initial set of coding tables ---*/ Chris@4: { Chris@4: Int32 nPart, remF, tFreq, aFreq; Chris@4: Chris@4: nPart = nGroups; Chris@4: remF = s->nMTF; Chris@4: gs = 0; Chris@4: while (nPart > 0) { Chris@4: tFreq = remF / nPart; Chris@4: ge = gs-1; Chris@4: aFreq = 0; Chris@4: while (aFreq < tFreq && ge < alphaSize-1) { Chris@4: ge++; Chris@4: aFreq += s->mtfFreq[ge]; Chris@4: } Chris@4: Chris@4: if (ge > gs Chris@4: && nPart != nGroups && nPart != 1 Chris@4: && ((nGroups-nPart) % 2 == 1)) { Chris@4: aFreq -= s->mtfFreq[ge]; Chris@4: ge--; Chris@4: } Chris@4: Chris@4: if (s->verbosity >= 3) Chris@4: VPrintf5( " initial group %d, [%d .. %d], " Chris@4: "has %d syms (%4.1f%%)\n", Chris@4: nPart, gs, ge, aFreq, Chris@4: (100.0 * (float)aFreq) / (float)(s->nMTF) ); Chris@4: Chris@4: for (v = 0; v < alphaSize; v++) Chris@4: if (v >= gs && v <= ge) Chris@4: s->len[nPart-1][v] = BZ_LESSER_ICOST; else Chris@4: s->len[nPart-1][v] = BZ_GREATER_ICOST; Chris@4: Chris@4: nPart--; Chris@4: gs = ge+1; Chris@4: remF -= aFreq; Chris@4: } Chris@4: } Chris@4: Chris@4: /*--- Chris@4: Iterate up to BZ_N_ITERS times to improve the tables. Chris@4: ---*/ Chris@4: for (iter = 0; iter < BZ_N_ITERS; iter++) { Chris@4: Chris@4: for (t = 0; t < nGroups; t++) fave[t] = 0; Chris@4: Chris@4: for (t = 0; t < nGroups; t++) Chris@4: for (v = 0; v < alphaSize; v++) Chris@4: s->rfreq[t][v] = 0; Chris@4: Chris@4: /*--- Chris@4: Set up an auxiliary length table which is used to fast-track Chris@4: the common case (nGroups == 6). Chris@4: ---*/ Chris@4: if (nGroups == 6) { Chris@4: for (v = 0; v < alphaSize; v++) { Chris@4: s->len_pack[v][0] = (s->len[1][v] << 16) | s->len[0][v]; Chris@4: s->len_pack[v][1] = (s->len[3][v] << 16) | s->len[2][v]; Chris@4: s->len_pack[v][2] = (s->len[5][v] << 16) | s->len[4][v]; Chris@4: } Chris@4: } Chris@4: Chris@4: nSelectors = 0; Chris@4: totc = 0; Chris@4: gs = 0; Chris@4: while (True) { Chris@4: Chris@4: /*--- Set group start & end marks. --*/ Chris@4: if (gs >= s->nMTF) break; Chris@4: ge = gs + BZ_G_SIZE - 1; Chris@4: if (ge >= s->nMTF) ge = s->nMTF-1; Chris@4: Chris@4: /*-- Chris@4: Calculate the cost of this group as coded Chris@4: by each of the coding tables. Chris@4: --*/ Chris@4: for (t = 0; t < nGroups; t++) cost[t] = 0; Chris@4: Chris@4: if (nGroups == 6 && 50 == ge-gs+1) { Chris@4: /*--- fast track the common case ---*/ Chris@4: register UInt32 cost01, cost23, cost45; Chris@4: register UInt16 icv; Chris@4: cost01 = cost23 = cost45 = 0; Chris@4: Chris@4: # define BZ_ITER(nn) \ Chris@4: icv = mtfv[gs+(nn)]; \ Chris@4: cost01 += s->len_pack[icv][0]; \ Chris@4: cost23 += s->len_pack[icv][1]; \ Chris@4: cost45 += s->len_pack[icv][2]; \ Chris@4: Chris@4: BZ_ITER(0); BZ_ITER(1); BZ_ITER(2); BZ_ITER(3); BZ_ITER(4); Chris@4: BZ_ITER(5); BZ_ITER(6); BZ_ITER(7); BZ_ITER(8); BZ_ITER(9); Chris@4: BZ_ITER(10); BZ_ITER(11); BZ_ITER(12); BZ_ITER(13); BZ_ITER(14); Chris@4: BZ_ITER(15); BZ_ITER(16); BZ_ITER(17); BZ_ITER(18); BZ_ITER(19); Chris@4: BZ_ITER(20); BZ_ITER(21); BZ_ITER(22); BZ_ITER(23); BZ_ITER(24); Chris@4: BZ_ITER(25); BZ_ITER(26); BZ_ITER(27); BZ_ITER(28); BZ_ITER(29); Chris@4: BZ_ITER(30); BZ_ITER(31); BZ_ITER(32); BZ_ITER(33); BZ_ITER(34); Chris@4: BZ_ITER(35); BZ_ITER(36); BZ_ITER(37); BZ_ITER(38); BZ_ITER(39); Chris@4: BZ_ITER(40); BZ_ITER(41); BZ_ITER(42); BZ_ITER(43); BZ_ITER(44); Chris@4: BZ_ITER(45); BZ_ITER(46); BZ_ITER(47); BZ_ITER(48); BZ_ITER(49); Chris@4: Chris@4: # undef BZ_ITER Chris@4: Chris@4: cost[0] = cost01 & 0xffff; cost[1] = cost01 >> 16; Chris@4: cost[2] = cost23 & 0xffff; cost[3] = cost23 >> 16; Chris@4: cost[4] = cost45 & 0xffff; cost[5] = cost45 >> 16; Chris@4: Chris@4: } else { Chris@4: /*--- slow version which correctly handles all situations ---*/ Chris@4: for (i = gs; i <= ge; i++) { Chris@4: UInt16 icv = mtfv[i]; Chris@4: for (t = 0; t < nGroups; t++) cost[t] += s->len[t][icv]; Chris@4: } Chris@4: } Chris@4: Chris@4: /*-- Chris@4: Find the coding table which is best for this group, Chris@4: and record its identity in the selector table. Chris@4: --*/ Chris@4: bc = 999999999; bt = -1; Chris@4: for (t = 0; t < nGroups; t++) Chris@4: if (cost[t] < bc) { bc = cost[t]; bt = t; }; Chris@4: totc += bc; Chris@4: fave[bt]++; Chris@4: s->selector[nSelectors] = bt; Chris@4: nSelectors++; Chris@4: Chris@4: /*-- Chris@4: Increment the symbol frequencies for the selected table. Chris@4: --*/ Chris@4: if (nGroups == 6 && 50 == ge-gs+1) { Chris@4: /*--- fast track the common case ---*/ Chris@4: Chris@4: # define BZ_ITUR(nn) s->rfreq[bt][ mtfv[gs+(nn)] ]++ Chris@4: Chris@4: BZ_ITUR(0); BZ_ITUR(1); BZ_ITUR(2); BZ_ITUR(3); BZ_ITUR(4); Chris@4: BZ_ITUR(5); BZ_ITUR(6); BZ_ITUR(7); BZ_ITUR(8); BZ_ITUR(9); Chris@4: BZ_ITUR(10); BZ_ITUR(11); BZ_ITUR(12); BZ_ITUR(13); BZ_ITUR(14); Chris@4: BZ_ITUR(15); BZ_ITUR(16); BZ_ITUR(17); BZ_ITUR(18); BZ_ITUR(19); Chris@4: BZ_ITUR(20); BZ_ITUR(21); BZ_ITUR(22); BZ_ITUR(23); BZ_ITUR(24); Chris@4: BZ_ITUR(25); BZ_ITUR(26); BZ_ITUR(27); BZ_ITUR(28); BZ_ITUR(29); Chris@4: BZ_ITUR(30); BZ_ITUR(31); BZ_ITUR(32); BZ_ITUR(33); BZ_ITUR(34); Chris@4: BZ_ITUR(35); BZ_ITUR(36); BZ_ITUR(37); BZ_ITUR(38); BZ_ITUR(39); Chris@4: BZ_ITUR(40); BZ_ITUR(41); BZ_ITUR(42); BZ_ITUR(43); BZ_ITUR(44); Chris@4: BZ_ITUR(45); BZ_ITUR(46); BZ_ITUR(47); BZ_ITUR(48); BZ_ITUR(49); Chris@4: Chris@4: # undef BZ_ITUR Chris@4: Chris@4: } else { Chris@4: /*--- slow version which correctly handles all situations ---*/ Chris@4: for (i = gs; i <= ge; i++) Chris@4: s->rfreq[bt][ mtfv[i] ]++; Chris@4: } Chris@4: Chris@4: gs = ge+1; Chris@4: } Chris@4: if (s->verbosity >= 3) { Chris@4: VPrintf2 ( " pass %d: size is %d, grp uses are ", Chris@4: iter+1, totc/8 ); Chris@4: for (t = 0; t < nGroups; t++) Chris@4: VPrintf1 ( "%d ", fave[t] ); Chris@4: VPrintf0 ( "\n" ); Chris@4: } Chris@4: Chris@4: /*-- Chris@4: Recompute the tables based on the accumulated frequencies. Chris@4: --*/ Chris@4: /* maxLen was changed from 20 to 17 in bzip2-1.0.3. See Chris@4: comment in huffman.c for details. */ Chris@4: for (t = 0; t < nGroups; t++) Chris@4: BZ2_hbMakeCodeLengths ( &(s->len[t][0]), &(s->rfreq[t][0]), Chris@4: alphaSize, 17 /*20*/ ); Chris@4: } Chris@4: Chris@4: Chris@4: AssertH( nGroups < 8, 3002 ); Chris@4: AssertH( nSelectors < 32768 && Chris@4: nSelectors <= (2 + (900000 / BZ_G_SIZE)), Chris@4: 3003 ); Chris@4: Chris@4: Chris@4: /*--- Compute MTF values for the selectors. ---*/ Chris@4: { Chris@4: UChar pos[BZ_N_GROUPS], ll_i, tmp2, tmp; Chris@4: for (i = 0; i < nGroups; i++) pos[i] = i; Chris@4: for (i = 0; i < nSelectors; i++) { Chris@4: ll_i = s->selector[i]; Chris@4: j = 0; Chris@4: tmp = pos[j]; Chris@4: while ( ll_i != tmp ) { Chris@4: j++; Chris@4: tmp2 = tmp; Chris@4: tmp = pos[j]; Chris@4: pos[j] = tmp2; Chris@4: }; Chris@4: pos[0] = tmp; Chris@4: s->selectorMtf[i] = j; Chris@4: } Chris@4: }; Chris@4: Chris@4: /*--- Assign actual codes for the tables. --*/ Chris@4: for (t = 0; t < nGroups; t++) { Chris@4: minLen = 32; Chris@4: maxLen = 0; Chris@4: for (i = 0; i < alphaSize; i++) { Chris@4: if (s->len[t][i] > maxLen) maxLen = s->len[t][i]; Chris@4: if (s->len[t][i] < minLen) minLen = s->len[t][i]; Chris@4: } Chris@4: AssertH ( !(maxLen > 17 /*20*/ ), 3004 ); Chris@4: AssertH ( !(minLen < 1), 3005 ); Chris@4: BZ2_hbAssignCodes ( &(s->code[t][0]), &(s->len[t][0]), Chris@4: minLen, maxLen, alphaSize ); Chris@4: } Chris@4: Chris@4: /*--- Transmit the mapping table. ---*/ Chris@4: { Chris@4: Bool inUse16[16]; Chris@4: for (i = 0; i < 16; i++) { Chris@4: inUse16[i] = False; Chris@4: for (j = 0; j < 16; j++) Chris@4: if (s->inUse[i * 16 + j]) inUse16[i] = True; Chris@4: } Chris@4: Chris@4: nBytes = s->numZ; Chris@4: for (i = 0; i < 16; i++) Chris@4: if (inUse16[i]) bsW(s,1,1); else bsW(s,1,0); Chris@4: Chris@4: for (i = 0; i < 16; i++) Chris@4: if (inUse16[i]) Chris@4: for (j = 0; j < 16; j++) { Chris@4: if (s->inUse[i * 16 + j]) bsW(s,1,1); else bsW(s,1,0); Chris@4: } Chris@4: Chris@4: if (s->verbosity >= 3) Chris@4: VPrintf1( " bytes: mapping %d, ", s->numZ-nBytes ); Chris@4: } Chris@4: Chris@4: /*--- Now the selectors. ---*/ Chris@4: nBytes = s->numZ; Chris@4: bsW ( s, 3, nGroups ); Chris@4: bsW ( s, 15, nSelectors ); Chris@4: for (i = 0; i < nSelectors; i++) { Chris@4: for (j = 0; j < s->selectorMtf[i]; j++) bsW(s,1,1); Chris@4: bsW(s,1,0); Chris@4: } Chris@4: if (s->verbosity >= 3) Chris@4: VPrintf1( "selectors %d, ", s->numZ-nBytes ); Chris@4: Chris@4: /*--- Now the coding tables. ---*/ Chris@4: nBytes = s->numZ; Chris@4: Chris@4: for (t = 0; t < nGroups; t++) { Chris@4: Int32 curr = s->len[t][0]; Chris@4: bsW ( s, 5, curr ); Chris@4: for (i = 0; i < alphaSize; i++) { Chris@4: while (curr < s->len[t][i]) { bsW(s,2,2); curr++; /* 10 */ }; Chris@4: while (curr > s->len[t][i]) { bsW(s,2,3); curr--; /* 11 */ }; Chris@4: bsW ( s, 1, 0 ); Chris@4: } Chris@4: } Chris@4: Chris@4: if (s->verbosity >= 3) Chris@4: VPrintf1 ( "code lengths %d, ", s->numZ-nBytes ); Chris@4: Chris@4: /*--- And finally, the block data proper ---*/ Chris@4: nBytes = s->numZ; Chris@4: selCtr = 0; Chris@4: gs = 0; Chris@4: while (True) { Chris@4: if (gs >= s->nMTF) break; Chris@4: ge = gs + BZ_G_SIZE - 1; Chris@4: if (ge >= s->nMTF) ge = s->nMTF-1; Chris@4: AssertH ( s->selector[selCtr] < nGroups, 3006 ); Chris@4: Chris@4: if (nGroups == 6 && 50 == ge-gs+1) { Chris@4: /*--- fast track the common case ---*/ Chris@4: UInt16 mtfv_i; Chris@4: UChar* s_len_sel_selCtr Chris@4: = &(s->len[s->selector[selCtr]][0]); Chris@4: Int32* s_code_sel_selCtr Chris@4: = &(s->code[s->selector[selCtr]][0]); Chris@4: Chris@4: # define BZ_ITAH(nn) \ Chris@4: mtfv_i = mtfv[gs+(nn)]; \ Chris@4: bsW ( s, \ Chris@4: s_len_sel_selCtr[mtfv_i], \ Chris@4: s_code_sel_selCtr[mtfv_i] ) Chris@4: Chris@4: BZ_ITAH(0); BZ_ITAH(1); BZ_ITAH(2); BZ_ITAH(3); BZ_ITAH(4); Chris@4: BZ_ITAH(5); BZ_ITAH(6); BZ_ITAH(7); BZ_ITAH(8); BZ_ITAH(9); Chris@4: BZ_ITAH(10); BZ_ITAH(11); BZ_ITAH(12); BZ_ITAH(13); BZ_ITAH(14); Chris@4: BZ_ITAH(15); BZ_ITAH(16); BZ_ITAH(17); BZ_ITAH(18); BZ_ITAH(19); Chris@4: BZ_ITAH(20); BZ_ITAH(21); BZ_ITAH(22); BZ_ITAH(23); BZ_ITAH(24); Chris@4: BZ_ITAH(25); BZ_ITAH(26); BZ_ITAH(27); BZ_ITAH(28); BZ_ITAH(29); Chris@4: BZ_ITAH(30); BZ_ITAH(31); BZ_ITAH(32); BZ_ITAH(33); BZ_ITAH(34); Chris@4: BZ_ITAH(35); BZ_ITAH(36); BZ_ITAH(37); BZ_ITAH(38); BZ_ITAH(39); Chris@4: BZ_ITAH(40); BZ_ITAH(41); BZ_ITAH(42); BZ_ITAH(43); BZ_ITAH(44); Chris@4: BZ_ITAH(45); BZ_ITAH(46); BZ_ITAH(47); BZ_ITAH(48); BZ_ITAH(49); Chris@4: Chris@4: # undef BZ_ITAH Chris@4: Chris@4: } else { Chris@4: /*--- slow version which correctly handles all situations ---*/ Chris@4: for (i = gs; i <= ge; i++) { Chris@4: bsW ( s, Chris@4: s->len [s->selector[selCtr]] [mtfv[i]], Chris@4: s->code [s->selector[selCtr]] [mtfv[i]] ); Chris@4: } Chris@4: } Chris@4: Chris@4: Chris@4: gs = ge+1; Chris@4: selCtr++; Chris@4: } Chris@4: AssertH( selCtr == nSelectors, 3007 ); Chris@4: Chris@4: if (s->verbosity >= 3) Chris@4: VPrintf1( "codes %d\n", s->numZ-nBytes ); Chris@4: } Chris@4: Chris@4: Chris@4: /*---------------------------------------------------*/ Chris@4: void BZ2_compressBlock ( EState* s, Bool is_last_block ) Chris@4: { Chris@4: if (s->nblock > 0) { Chris@4: Chris@4: BZ_FINALISE_CRC ( s->blockCRC ); Chris@4: s->combinedCRC = (s->combinedCRC << 1) | (s->combinedCRC >> 31); Chris@4: s->combinedCRC ^= s->blockCRC; Chris@4: if (s->blockNo > 1) s->numZ = 0; Chris@4: Chris@4: if (s->verbosity >= 2) Chris@4: VPrintf4( " block %d: crc = 0x%08x, " Chris@4: "combined CRC = 0x%08x, size = %d\n", Chris@4: s->blockNo, s->blockCRC, s->combinedCRC, s->nblock ); Chris@4: Chris@4: BZ2_blockSort ( s ); Chris@4: } Chris@4: Chris@4: s->zbits = (UChar*) (&((UChar*)s->arr2)[s->nblock]); Chris@4: Chris@4: /*-- If this is the first block, create the stream header. --*/ Chris@4: if (s->blockNo == 1) { Chris@4: BZ2_bsInitWrite ( s ); Chris@4: bsPutUChar ( s, BZ_HDR_B ); Chris@4: bsPutUChar ( s, BZ_HDR_Z ); Chris@4: bsPutUChar ( s, BZ_HDR_h ); Chris@4: bsPutUChar ( s, (UChar)(BZ_HDR_0 + s->blockSize100k) ); Chris@4: } Chris@4: Chris@4: if (s->nblock > 0) { Chris@4: Chris@4: bsPutUChar ( s, 0x31 ); bsPutUChar ( s, 0x41 ); Chris@4: bsPutUChar ( s, 0x59 ); bsPutUChar ( s, 0x26 ); Chris@4: bsPutUChar ( s, 0x53 ); bsPutUChar ( s, 0x59 ); Chris@4: Chris@4: /*-- Now the block's CRC, so it is in a known place. --*/ Chris@4: bsPutUInt32 ( s, s->blockCRC ); Chris@4: Chris@4: /*-- Chris@4: Now a single bit indicating (non-)randomisation. Chris@4: As of version 0.9.5, we use a better sorting algorithm Chris@4: which makes randomisation unnecessary. So always set Chris@4: the randomised bit to 'no'. Of course, the decoder Chris@4: still needs to be able to handle randomised blocks Chris@4: so as to maintain backwards compatibility with Chris@4: older versions of bzip2. Chris@4: --*/ Chris@4: bsW(s,1,0); Chris@4: Chris@4: bsW ( s, 24, s->origPtr ); Chris@4: generateMTFValues ( s ); Chris@4: sendMTFValues ( s ); Chris@4: } Chris@4: Chris@4: Chris@4: /*-- If this is the last block, add the stream trailer. --*/ Chris@4: if (is_last_block) { Chris@4: Chris@4: bsPutUChar ( s, 0x17 ); bsPutUChar ( s, 0x72 ); Chris@4: bsPutUChar ( s, 0x45 ); bsPutUChar ( s, 0x38 ); Chris@4: bsPutUChar ( s, 0x50 ); bsPutUChar ( s, 0x90 ); Chris@4: bsPutUInt32 ( s, s->combinedCRC ); Chris@4: if (s->verbosity >= 2) Chris@4: VPrintf1( " final combined CRC = 0x%08x\n ", s->combinedCRC ); Chris@4: bsFinishWrite ( s ); Chris@4: } Chris@4: } Chris@4: Chris@4: Chris@4: /*-------------------------------------------------------------*/ Chris@4: /*--- end compress.c ---*/ Chris@4: /*-------------------------------------------------------------*/