wok diff syslinux/stuff/iso2exe/unlzma.S @ rev 13713
syslinux/iso2exe: add loram support
| author | Pascal Bellard <pascal.bellard@slitaz.org> |
|---|---|
| date | Tue Dec 18 16:09:07 2012 +0100 (2012-12-18) |
| parents | |
| children |
line diff
1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 1.2 +++ b/syslinux/stuff/iso2exe/unlzma.S Tue Dec 18 16:09:07 2012 +0100 1.3 @@ -0,0 +1,826 @@ 1.4 +// #define RC_NORMALIZE if (Range < kTopValue) { Range <<= 8; Code = (Code << 8) | RC_READ_BYTE; } 1.5 +// 1.6 +// #define IfBit0(p) RC_NORMALIZE; bound = (Range >> kNumBitModelTotalBits) * *(p); if (Code < bound) 1.7 +// #define UpdateBit0(p) Range = bound; *(p) += (kBitModelTotal - *(p)) >> kNumMoveBits; 1.8 +// #define UpdateBit1(p) Range -= bound; Code -= bound; *(p) -= (*(p)) >> kNumMoveBits; 1.9 +// 1.10 +//#define RC_GET_BIT2(p, mi, A0, A1) IfBit0(p) \ 1.11 +// { UpdateBit0(p); mi <<= 1; A0; } else \ 1.12 +// { UpdateBit1(p); mi = (mi + mi) + 1; A1; } 1.13 +// 1.14 +// #define RC_GET_BIT(p, mi) RC_GET_BIT2(p, mi, ; , ;) 1.15 +// 1.16 +// #define RangeDecoderBitTreeDecode(probs, numLevels, res) \ 1.17 +// { int i = numLevels; res = 1; \ 1.18 +// do { CProb *p = probs + res; RC_GET_BIT(p, res) } while(--i != 0); \ 1.19 +// res -= (1 << numLevels); } 1.20 +/* 1.21 + * Compression with : lzma e src dst -eos -pb2 -lp0 -lc3 1.22 + */ 1.23 + 1.24 +#define PROP_PB 2 1.25 +#define PROP_LP 0 1.26 +#define PROP_LC 3 1.27 +#define PROPS (PROP_LC+(PROP_LP*9)+(PROP_PB*45)) 1.28 + 1.29 +// static const Byte *Buffer; 1.30 +// static UInt32 bound, Code, Range; 1.31 + 1.32 +/* 1.33 + * Buffer register DS:SI 1.34 + * all var based ws=ss:bp 1.35 + */ 1.36 + 1.37 +rep0 = -4 // long 1.38 +rep1 = rep0-4 // long 1.39 +rep2 = rep0-8 // long 1.40 +rep3 = rep0-12 // long 1.41 +state = -17 // byte, 0..11 1.42 +posState = state-1 // byte, 0..15 1.43 +posState2 = posState-1 // byte, 0..15 1.44 +scratched = rep0-16 // byte = 1 1.45 +Code = -24 // long 1.46 +outStream = -28 // long 1.47 +nowPos = outStream // long 1.48 +Range = Code-8 // long 1.49 +#define LOCALS 32 1.50 + 1.51 +// int LzmaDecode(CLzmaDecoderState *vs, 1.52 +// const unsigned char *inStream, 1.53 +// unsigned char *outStream) 1.54 +// { 1.55 +// CProb *p = vs->Probs; 1.56 +// SizeT nowPos = 0; 1.57 +// #define posStateMask = (1 << (vs->Properties.pb)) - 1; 1.58 +// #define literalPosMask = (1 << (vs->Properties.lp)) - 1; 1.59 +// int lc = vs->Properties.lc, state = 0, len = 0; 1.60 +// UInt32 rep0 = 1, rep1 = 1, rep2 = 1, rep3 = 1; 1.61 +// 1.62 +// { 1.63 +// UInt32 i, numProbs = Literal /*1846*/ 1.64 +// + ((UInt32)LZMA_LIT_SIZE /*768*/ << (lc + vs->Properties.lp)); 1.65 +// for (i = 0; i < numProbs; i++) p[i] = kBitModelTotal /*2048*/ >> 1; 1.66 + 1.67 +#define WS (1846+(768<<(PROP_LC+PROP_LP))) 1.68 +#if (WS+WS+LOCALS) >= 65000 1.69 +/* MAX WS = (1846+(768<<(8+4))) > 3MB! */ 1.70 +#error invalid (lc,lp,pb) : out of memory 1.71 +#endif 1.72 + 1.73 +ws1 = WS 1.74 +ws2 = ws1*2 1.75 +ws = ws2+LOCALS+15 1.76 + 1.77 +#ifndef FLAT16 1.78 +#define INC incl 1.79 +#else 1.80 +#define INC incw 1.81 +#endif 1.82 +#ifndef FLAT32 1.83 +#define AX %ax 1.84 +#define BX %bx 1.85 +#define CX %cx 1.86 +#define DX %dx 1.87 +#define SI %si 1.88 +#define DI %di 1.89 +#define BP %bp 1.90 +#define SP %sp 1.91 +#define CWD cwd 1.92 +#else 1.93 +#define AX %eax 1.94 +#define BX %ebx 1.95 +#define CX %ecx 1.96 +#define DX %edx 1.97 +#define SI %esi 1.98 +#define DI %edi 1.99 +#define BP %ebp 1.100 +#define SP %esp 1.101 +#define CWD cdq 1.102 +#endif 1.103 +/* 1.104 + * LzmaDecode: 1.105 +#ifndef FLAT32 1.106 + * input ds:si=inStream, es:di=outStream 1.107 + * output outStream[], ds:si, es:di 1.108 + .code 16 1.109 +#else 1.110 + * input esi=inStream, edi=outStream 1.111 + * output outStream[], esi, edi 1.112 + .code 32 1.113 +#endif 1.114 + */ 1.115 + 1.116 + mov $ws1, CX 1.117 +lzd1: 1.118 + pushw $2048/2 1.119 + loop lzd1 1.120 + mov SP, BP 1.121 + movb $((LOCALS+3)/4)*2, %cl 1.122 +initlocals: 1.123 + pushl $1 1.124 + loop initlocals 1.125 + 1.126 +#if !defined(FLAT32) && !defined(FLAT16) 1.127 + movb $4, %cl 1.128 + movw %es, %bx 1.129 + shrw %cl, %bx 1.130 + movw %es, %dx 1.131 + shlw %cl, %dx 1.132 + addw %dx, %di 1.133 + movw %di, outStream(%bp) 1.134 + adcb %bh, outStream+2(%bp) 1.135 + incw %cx 1.136 +#else 1.137 + movb $5, %cl 1.138 + mov DI, outStream(BP) 1.139 +#endif 1.140 + 1.141 +// Byte previousByte = 0; 1.142 + xor BX, BX 1.143 + 1.144 +// #define RC_INIT(buffer) 1.145 +// Buffer = buffer; Code = 0; Range = 0xFFFFFFFF; 1.146 +// { int i; for(i=0; i<5; i++) { Code = (Code<<8) | RC_READ_BYTE; }} 1.147 +// } 1.148 +// RC_INIT(inStream); 1.149 + 1.150 + add $13, SI // skip header 1.151 +setrep: 1.152 + call RC_LOAD_BYTE 1.153 + decb Range(BP) 1.154 + loop setrep 1.155 + 1.156 +lzdmainlp: 1.157 +// while(1) { 1.158 +// CProb *prob; 1.159 +// int posState = (int)((nowPos) & posStateMask); 1.160 +// 1.161 +// prob = p + IsMatch /*0*/ + (state << kNumPosBitsMax /*4*/) + posState; 1.162 +// if (Bit0(prob)) { /* char */ 1.163 + 1.164 + xor DX, DX 1.165 + call Bit1state // Bit1(dx + (state << kNumPosBitsMax /*4*/) + posState) 1.166 + mov $state, DI 1.167 + jc lzdstring 1.168 + 1.169 +// prob = p + Literal /*1846*/ + (LZMA_LIT_SIZE /*768*/ * 1.170 +// ((((nowPos) & literalPosMask) << lc) + (previousByte >> (8 - lc)))); 1.171 + 1.172 +#if PROP_LC != 0 1.173 + shrb $8-PROP_LC, %bl 1.174 +#endif 1.175 + 1.176 +#if PROP_LP != 0 1.177 + movb posState2(BP), %dl 1.178 + shl $PROP_LC, DX 1.179 + movb $0, %bh 1.180 + add BX, DX 1.181 +#endif 1.182 + 1.183 + movb $3, %ah 1.184 + mul BX // dx = 3*bh 1.185 + add $1846, AX 1.186 + 1.187 +// int symbol = 1; 1.188 + 1.189 + CWD 1.190 + inc DX // symbol = 1 1.191 + xchg AX, CX // save prob 1.192 + 1.193 +// if (state >= kNumLitStates /*7*/) { /* previous was string */ 1.194 +// if (state < 4) state = 0; 1.195 + 1.196 +lzd6z: 1.197 + subb $3, (BP, DI) 1.198 + 1.199 +// if (state < 4) state = 0; 1.200 + 1.201 + jnc lzd6 1.202 + movb %dh, (BP, DI) // %dh = 0 1.203 + 1.204 +lzd6: 1.205 +// else if (state < 10) state -= 3; 1.206 + 1.207 + cmpb $10-3, (BP, DI) 1.208 + 1.209 +// else state -= 6; 1.210 + 1.211 + jnb lzd6z 1.212 + cmpb $7-3-1, (BP, DI) 1.213 + jbe lzd3 1.214 + 1.215 +// int matchByte = outStream[nowPos - rep0]; 1.216 + 1.217 + call DicoRep02ESDI // %bl = outStream[nowPos - rep0]; 1.218 + 1.219 +// do { 1.220 +// int bit; 1.221 +// CProb *probLit; 1.222 +// matchByte <<= 1; bit = (matchByte & 0x100); 1.223 + 1.224 + movb $1, %bh 1.225 +lzd4: 1.226 + shlb $1, %bl // matchByte <<= 1 1.227 + sbb DI, DI // save bit=C 1.228 + 1.229 +// probLit = prob + 0x100 + bit + symbol; 1.230 + 1.231 + mov CX, AX // restore prob 1.232 + adcb %bh, %ah // + bit + 0x100 1.233 + 1.234 +// RC_GET_BIT2(probLit, symbol, if (bit) break, if (!bit) break) 1.235 + 1.236 + call Bit1axdx // C,%ax = Bit1(prob+%ax) 1.237 + rclb $1, %dl // symbol <<= 1; symbol |= C 1.238 + jc lzd5 // if symbol >= 0x100 1.239 + cmp DI, AX 1.240 + jz lzd4 // if bit == Bit1(prob+%ax) 1.241 + 1.242 +// } while (symbol < 0x100); 1.243 +// } 1.244 +lzd3: 1.245 +// while (symbol < 0x100) { 1.246 +// CProb *probLit = prob + symbol; 1.247 +// RC_GET_BIT(probLit, symbol) 1.248 +// } 1.249 + 1.250 + xor BX, BX 1.251 + jmp lzd4 1.252 +lzd5: 1.253 + 1.254 +// outStream[nowPos++] = previousByte = (Byte)symbol; 1.255 + 1.256 + xchg AX, DX 1.257 + call outchar // %bl = outStream[nowPos++] = %al; 1.258 + jmp lzdmainlp 1.259 + 1.260 +// } 1.261 + 1.262 +lzdstring: 1.263 + mov $1, CX 1.264 + 1.265 +// else { /* string */ 1.266 +// prob = p + IsRep /*192*/ + state; 1.267 + 1.268 + movb $192, %dl 1.269 + addb (BP, DI), %dl 1.270 + mov $rep0, DI 1.271 + 1.272 +// if (Bit0(prob)) { 1.273 + 1.274 + call Bit1dx // Bit1(prob) 1.275 + jc lzd8 1.276 + 1.277 +// rep3 = rep2; rep2 = rep1; rep1 = rep0; 1.278 +// state = (state < kNumLitStates /*7*/) ? 0 : 3; 1.279 + 1.280 + stc 1.281 + 1.282 +// prob = p + LenCoder /*818*/; 1.283 + 1.284 + mov $818, DX 1.285 + 1.286 +// } 1.287 + 1.288 + jmp lzd11a 1.289 + 1.290 +// else { 1.291 +lzd8: 1.292 +// prob += kNumStates /*12*/; 1.293 +// if (Bit0(prob)) { 1.294 + call Bit1dx12 // prob += 12; Bit1(prob) 1.295 + jc lzd11 1.296 +// prob = p + IsRep0Long /*240*/ + (state << kNumPosBitsMax /*4*/) 1.297 +// + posState; 1.298 + movb $240, %dl // dh=0 1.299 + 1.300 +// if (Bit0(prob)) { 1.301 + 1.302 + call Bit1state // Bit1(dx + (state << kNumPosBitsMax /*4*/) + posState) 1.303 + jc lzd12 1.304 + 1.305 +// // if (nowPos == 0) return LZMA_RESULT_DATA_ERROR; 1.306 +// state = (state < kNumLitStates /*7*/) ? 9 : 11; 1.307 + 1.308 + movb $9, %dl 1.309 + 1.310 +// len++; goto string; 1.311 + jmp lzd13string // ax = 0 1.312 +// } 1.313 +// } 1.314 +// else { 1.315 +lzd11: 1.316 +// UInt32 distance = rep1; 1.317 +// prob += kNumStates /*12*/; 1.318 +// if (!Bit0(prob)) { 1.319 + 1.320 + call Bit1dx12 // prob += 12; Bit1(prob) 1.321 + jnc lzd11z 1.322 + 1.323 +// prob += kNumStates /*12*/; 1.324 +// if (Bit0(prob)) distance = rep2; 1.325 + 1.326 + call Bit1dx12 // prob += 12; Bit1(prob) 1.327 +lzd11a: 1.328 + adcb %cl, %cl 1.329 + 1.330 +// else { distance = rep3; rep3 = rep2; } 1.331 +// rep2 = rep1; 1.332 +// } 1.333 +// rep1 = rep0; rep0 = distance; 1.334 + 1.335 +lzd11z: 1.336 + shl $2, CX // 8->32 bits 1.337 + sub CX, DI // &rep[cx] 1.338 + movl (BP, DI), %eax 1.339 +rotreplp: 1.340 + movb 4(BP, DI), %bl 1.341 + movb %bl, (BP, DI) 1.342 + inc DI 1.343 + loop rotreplp 1.344 + testb %dh, %dh 1.345 + jnz lzd10 1.346 + movl %eax, (BP, DI) 1.347 + 1.348 +// } 1.349 +lzd12: 1.350 +// state = (state < kNumLitStates /*7*/) ? 8 : 11; 1.351 + 1.352 + movb $0x08, %cl 1.353 + 1.354 +// prob = p + RepLenCoder /*1332*/; 1.355 + 1.356 + mov $1332, DX 1.357 + 1.358 +// } 1.359 +lzd10: 1.360 + push CX // CX = 0 1.361 + 1.362 +// { /* get len */ 1.363 +// int numBits, offset; 1.364 +// CProb *probLen = prob + LenChoice /*0*/; 1.365 +// numBits = kLenNumLowBits /*3*/; 1.366 + 1.367 + movb $8, %cl // numBits : 3,3,8 1.368 + 1.369 +// if (Bit0(probLen)) { 1.370 + 1.371 + call Bit1dx // Bit1(prob) 1.372 + xchg AX, BX 1.373 + inc DX 1.374 + jnc lzd15 // bx=0 1.375 + 1.376 +// probLen = prob + LenLow/*2*/ + (posState << kLenNumLowBits/*3*/); 1.377 +// offset = 0; 1.378 +// } 1.379 +// else { 1.380 +// probLen = prob + LenChoice2 /*1*/; 1.381 + 1.382 + call Bit1dx // Bit1(prob) 1.383 + add AX, BX 1.384 + 1.385 +#if PROP_PB != 0 1.386 + inc AX // ah=0 1.387 +#endif 1.388 + jc lzd16 // %ax=0, %bx=-2 1.389 +lzd15: 1.390 +#if PROP_PB != 0 1.391 + movb $8, %al 1.392 + mulb posState(BP) 1.393 +#endif 1.394 + 1.395 +// if (Bit0(probLen)) { 1.396 +// probLen = prob + LenMid/*130*/ + (posState << kLenNumMidBits/*3*/); 1.397 + 1.398 + movb $3, %cl // numBits : 3,3,8 1.399 +lzd16: 1.400 +#if PROP_PB != 0 1.401 + add $2-128-1, AX // probLen : 2,130,258 1.402 +#else 1.403 + mov $2-128-1, AX // probLen : 2,130,258 1.404 +#endif 1.405 + add DX, AX 1.406 + mov $-8+1, DX // offset : 0,8,16 1.407 +lzdargslp: 1.408 + add $8, DX 1.409 + add $128, AX 1.410 + inc BX 1.411 + jle lzdargslp // leave with bx=1 1.412 + 1.413 +// offset = kLenNumLowSymbols /*8*/; 1.414 +// //numBits = kLenNumMidBits /*3*/; 1.415 +// } 1.416 +// else { 1.417 +// probLen = prob + LenHigh /*258*/; 1.418 +// offset = kLenNumLowSymbols /*8*/ + kLenNumMidSymbols /*8*/; 1.419 +// numBits = kLenNumHighBits /*8*/; 1.420 +// } 1.421 +// } 1.422 +// RangeDecoderBitTreeDecode(probLen, numBits, len); len += offset; 1.423 + 1.424 + push DX 1.425 + call RangeDecoder // %ax=probs, %cx=numLevels, %ax=res 1.426 + pop DX 1.427 + add DX, AX // offset 1.428 + pop DX // 0 1.429 +lzd13string: 1.430 + push AX 1.431 + 1.432 +// state = (state < kNumLitStates /*7*/) ? dl : dl|3; 1.433 + 1.434 + movb $7, %cl 1.435 + cmpb %cl, state(BP) 1.436 + jb new_state 1.437 + orb $3, %dl 1.438 +new_state: 1.439 + movb %dl, state(BP) 1.440 + 1.441 +// } /* get len */ 1.442 +// if (state < 4) { 1.443 + 1.444 + cmpb $4-1, %dl 1.445 + ja lzd19 1.446 + 1.447 +// int posSlot; 1.448 +// state += kNumLitStates /*7*/; 1.449 + 1.450 + addb %cl, state(BP) 1.451 + 1.452 +// prob = p + PosSlot /*432*/ + (((len < kNumLenToPosStates /*4*/) ? 1.453 +// len : kNumLenToPosStates - 1) << kNumPosSlotBits /*6*/); 1.454 + 1.455 + cmp $4+1, AX 1.456 + jb lzd21 1.457 + mov $3+1, AX 1.458 + 1.459 +lzd21: 1.460 + 1.461 + dec CX // cx = 6 1.462 + shl %cl, AX 1.463 + add $432-64, AX 1.464 + 1.465 +// RangeDecoderBitTreeDecode(prob, kNumPosSlotBits /*6*/, posSlot); 1.466 + 1.467 + call RangeDecoder // %ax=probs, %cx=numLevels, %ax=res 1.468 + 1.469 +// if (posSlot >= kStartPosModelIndex /*4*/) { 1.470 +// int numDirectBits = ((posSlot >> 1) - 1); 1.471 + 1.472 +#ifndef FLAT32 1.473 + movw %cx, 2(%bp, %di) // %cx = 0 1.474 +#endif 1.475 + mov AX, (BP, DI) 1.476 + mov AX, CX 1.477 + shrw $1, CX 1.478 + dec CX 1.479 + cmpb $4, %al 1.480 + jb lzd22 1.481 + 1.482 +// rep0 = (2 | ((UInt32)posSlot & 1)); 1.483 + 1.484 + andb %bl, (BP, DI) // %bx=1 1.485 + orb $2, (BP, DI) 1.486 + 1.487 +// if (posSlot < kEndPosModelIndex /*14*/) { 1.488 + 1.489 + cmpb $14, %al 1.490 + jnb lzd23 1.491 + 1.492 +// rep0 <<= numDirectBits; 1.493 + 1.494 + neg AX 1.495 + shll %cl, (BP, DI) 1.496 + add (BP, DI), AX 1.497 + 1.498 +// prob = p + SpecPos /*688*/ + rep0 - posSlot - 1; 1.499 + 1.500 + add $687, AX 1.501 + jmp lzd24 1.502 + 1.503 +// } 1.504 +// else { 1.505 +lzd23: 1.506 +// numDirectBits -= kNumAlignBits /*4*/; 1.507 +// do { 1.508 +// RC_NORMALIZE; Range >>= 1; rep0 <<= 1; 1.509 +// if (Code >= Range) { Code -= Range; rep0 |= 1; } 1.510 + 1.511 +lzd23z: 1.512 + call RC_NORMALIZE 1.513 + shrl $1, Range(BP) 1.514 + movl Range(BP), %eax 1.515 + cmpl Code(BP), %eax 1.516 + ja lzd25 1.517 + subl %eax, Code(BP) 1.518 + stc 1.519 +lzd25: 1.520 + rcll $1, (BP, DI) 1.521 + 1.522 +// } while (--numDirectBits != 0); 1.523 + 1.524 + cmpb $4+1, %cl 1.525 + loopne lzd23z 1.526 + 1.527 +// prob = p + Align /* 802 */; numDirectBits = kNumAlignBits /*4*/; 1.528 +// rep0 <<= numDirectBits; 1.529 + 1.530 + shll %cl, (BP, DI) 1.531 + mov $802, AX 1.532 +// } 1.533 + 1.534 +lzd24: 1.535 + call RangeDecoder // %ax=probs, %cx=numLevels, %ax=res 1.536 + 1.537 +// { 1.538 +// int i = 1, mi = 1; 1.539 +// do { 1.540 +// CProb *prob3 = prob + mi; 1.541 +// RC_GET_BIT2(prob3, mi, ; , rep0 |= i); 1.542 + 1.543 + orb %dh, (BP, DI) // update rep0 with DirectBits 1.544 + 1.545 +// i <<= 1; 1.546 +// } while(--numDirectBits != 0); 1.547 +// } 1.548 +// } else rep0 = posSlot; 1.549 +lzd22: 1.550 +// if (++rep0 == (UInt32)(0)) break; /* EOF */ 1.551 + 1.552 + incl (BP, DI) 1.553 + 1.554 +lzd19: 1.555 + pop CX 1.556 + jz lzdone 1.557 + 1.558 +// } 1.559 +// len += kMatchMinLen;/*2*/ 1.560 + 1.561 + inc CX 1.562 + 1.563 +// string: // if (rep0 > nowPos) return LZMA_RESULT_DATA_ERROR; 1.564 +// do { 1.565 +lzd13z: 1.566 +// previousByte = outStream[nowPos - rep0]; 1.567 +// outStream[nowPos++] = previousByte; 1.568 + 1.569 + call outcharDico // %bl = outStream[nowPos++] = outStream[nowPos - rep0] 1.570 + 1.571 +// } while(--len != 0); 1.572 + 1.573 + loop lzd13z 1.574 + 1.575 +// } /* char/string */ 1.576 +// } 1.577 + 1.578 + jmp lzdmainlp 1.579 + 1.580 +lzdone: 1.581 +// //RC_NORMALIZE; 1.582 +// //*inSizeProcessed = (SizeT)(Buffer - inStream); *outSizeProcessed = nowPos; 1.583 +// return LZMA_RESULT_OK; 1.584 + call Dico2ESDI // set es & di (rep0 = 0) 1.585 + lea ws2(BP), SP // dealloc 1.586 + ret 1.587 +// } 1.588 + 1.589 +// al = outStream[nowPos - rep0]; 1.590 + 1.591 +/* 1.592 + * output es:di, al 1.593 + * scratch bh, cl, flags 1.594 + */ 1.595 + 1.596 +DicoRep02ESDI: 1.597 + stc 1.598 + 1.599 +// bl = outStream[nowPos]; 1.600 + 1.601 +/* 1.602 + * output es:di, bl 1.603 + * scratch bh, cl, flags 1.604 + */ 1.605 + 1.606 +Dico2ESDI: 1.607 +#if !defined(FLAT32) && !defined(FLAT16) 1.608 + movl nowPos(%bp), %ebx 1.609 + jnc Dico2ESDIz 1.610 + subl rep0(%bp), %ebx 1.611 +Dico2ESDIz: 1.612 + movw %bx, %di 1.613 + xorw %bx, %bx 1.614 + shrl $4, %ebx 1.615 + movw %bx, %es 1.616 + movb %es:(%di), %bl 1.617 +#else 1.618 + mov nowPos(BP), DI 1.619 + jnc Dico2ESDIz 1.620 + sub rep0(BP), DI 1.621 +Dico2ESDIz: 1.622 + movb (DI), %bl 1.623 +#endif 1.624 + ret 1.625 + 1.626 +outcharDico: 1.627 + 1.628 +// bl = outStream[nowPos++] = outStream[nowPos - rep0] 1.629 + 1.630 +/* 1.631 + * output es:di, bl 1.632 + * update nowPos 1.633 + * scratch ax, dx, bh, cl, flags 1.634 + */ 1.635 + 1.636 + call DicoRep02ESDI // %bl = outStream[nowPos - rep0] 1.637 + xchg AX, BX 1.638 +outchar: 1.639 + 1.640 +// bl = outStream[nowPos++] = previousByte = al; 1.641 + 1.642 +/* 1.643 + * output bl 1.644 + * update nowPos 1.645 + * scratch ax, dx, bh, di, cl, flags 1.646 + */ 1.647 + 1.648 + clc 1.649 + call Dico2ESDI 1.650 + stosb 1.651 + xchg AX, BX // previous byte 1.652 + 1.653 +// int posState = (int)((nowPos) & posStateMask); 1.654 + 1.655 +#if PROP_PB != 0 && PROP_LP != 0 1.656 + addw $0x0101, posState2(BP) 1.657 + andb $(((1 << PROP_PB) -1)<<8)+((1 << PROP_LP) -1), posState2(BP) 1.658 +#else 1.659 +# if PROP_PB != 0 1.660 + incb posState(BP) 1.661 + andb $((1 << PROP_PB) -1), posState(BP) 1.662 +# endif 1.663 +# if PROP_LP != 0 1.664 + incb posState2(BP) 1.665 + andb $((1 << PROP_LP) -1), posState2(BP) 1.666 +# endif 1.667 +#endif 1.668 + INC nowPos(BP) 1.669 + ret 1.670 + 1.671 +// 1.672 +// #define RC_NORMALIZE if (Range < kTopValue) 1.673 +// { Range <<= 8; Code = (Code << 8) | RC_READ_BYTE; } 1.674 + 1.675 +/* 1.676 + * update Range, Code, ds:si 1.677 + * scratch flags 1.678 + */ 1.679 + 1.680 +RC_NORMALIZE: 1.681 + cmpb $0, Range+3(BP) 1.682 + jne RC_NORMALIZE_1 1.683 +RC_LOAD_BYTE: 1.684 + push AX 1.685 + shll $8, Range(BP) 1.686 + shll $8, Code(BP) 1.687 +#if !defined(FLAT32) && !defined(FLAT16) 1.688 + testw %si, %si 1.689 + jns RC_READ_BYTE 1.690 + movw %ds, %ax 1.691 + incw %ax 1.692 + movw %ax, %ds 1.693 + addw $-16, %si 1.694 +RC_READ_BYTE: 1.695 +#endif 1.696 + lodsb 1.697 + movb %al, Code(BP) 1.698 + pop AX 1.699 +RC_NORMALIZE_1: 1.700 + ret 1.701 + 1.702 +// Bit1(dx + (state << kNumPosBitsMax /*4*/) + posState) 1.703 + 1.704 +Bit1state: 1.705 + movb $16, %al 1.706 + mulb state(BP) 1.707 +# if PROP_PB != 0 1.708 + addb posState(BP), %al 1.709 +# endif 1.710 +Bit1axdx: 1.711 + add DX, AX 1.712 + jmp Bit1 1.713 + 1.714 +// prob += 12; Bit1(prob) 1.715 + 1.716 +Bit1dx12: 1.717 + add $12, DX 1.718 +Bit1dx: 1.719 + mov DX, AX 1.720 + 1.721 +// static int Bit1(CProb *p) 1.722 + 1.723 +Bit1: 1.724 +/* 1.725 + * input ax=p 1.726 + * output C, ax 1.727 + * update bound, Range, Code, ds:si 1.728 + * scratch flags 1.729 + */ 1.730 + 1.731 +// { 1.732 +// RC_NORMALIZE; 1.733 + 1.734 + call RC_NORMALIZE // kill %ax, update %si 1.735 + 1.736 + pushal 1.737 + 1.738 + xchg AX, DI 1.739 + add DI, DI // short * 1.740 + 1.741 + 1.742 +// bound = (Range>>kNumBitModelTotalBits /*11*/) * *(p); 1.743 + 1.744 + movl Range(BP), %eax 1.745 + shrl $11, %eax 1.746 + movzwl (BP, DI), %edx 1.747 + mull %edx 1.748 + 1.749 +// if (Code < bound) { 1.750 + 1.751 + cmpl Code(BP), %eax 1.752 + jbe Bit1_1 1.753 + 1.754 +// Range = bound; 1.755 + 1.756 + movl %eax, Range(BP) 1.757 + 1.758 +// *(p) += (kBitModelTotal /*2048*/ - *(p)) >> kNumMoveBits /*5*/; 1.759 + 1.760 + movw $2048, %ax 1.761 + 1.762 +// return 0; 1.763 + 1.764 + jmp Bit1_2 1.765 + 1.766 +// } 1.767 +// else { 1.768 + 1.769 +Bit1_1: 1.770 + 1.771 +// Range -= bound; Code -= bound; 1.772 + 1.773 + subl %eax, Range(BP) 1.774 + subl %eax, Code(BP) 1.775 + 1.776 +// *(p) -= (*(p)) >> kNumMoveBits /*5*/; 1.777 + 1.778 + movw $31, %ax 1.779 + 1.780 +// return 1; 1.781 + 1.782 + stc 1.783 +Bit1_2: 1.784 + pushf 1.785 + subw (BP, DI), %ax 1.786 + sarw $5, %ax 1.787 + addw %ax, (BP, DI) 1.788 + popf 1.789 + popal 1.790 + sbb AX, AX 1.791 + 1.792 +// } 1.793 +// } 1.794 + 1.795 + ret 1.796 + 1.797 +RangeDecoder: 1.798 + 1.799 +/* 1.800 + * input ax=probs cx=numLevels (< 8) bx=1 1.801 + * output ax=res (backward), dh (forward) 1.802 + * update bound, Range, Code, ds:si 1.803 + * scratch flags, cx=0, dl 1.804 + */ 1.805 + 1.806 + push BX 1.807 + 1.808 +// { int i = numLevels; res = 1; 1.809 + mov BX, DX // res = 1 1.810 + 1.811 +// do { CProb *p = probs + res; RC_GET_BIT(p, res) } while(--i != 0); 1.812 + 1.813 +RangeDecoder_1: 1.814 + push AX 1.815 + call Bit1axdx // C,%ax = Bit1(prob+%ax) 1.816 + rclb $1, %dl // res <<= 1; res |= C 1.817 + andb %bl, %al // current bit 1.818 + orb %al, %bh // store in bh 1.819 + shlb $1, %bl // update max 1.820 + pop AX 1.821 + loop RangeDecoder_1 1.822 + 1.823 +// res -= (1 << numLevels); } 1.824 + 1.825 + xchg AX, BX // move bh to dh 1.826 + xchg AX, DX // and dl to al 1.827 + sub %dl, %al // sub max 1.828 + pop BX 1.829 + ret