/* ***** BEGIN LICENSE BLOCK ***** * Version: RCSL 1.0/RPSL 1.0 * * Portions Copyright (c) 1995-2002 RealNetworks, Inc. All Rights Reserved. * * The contents of this file, and the files included with this file, are * subject to the current version of the RealNetworks Public Source License * Version 1.0 (the "RPSL") available at * http://www.helixcommunity.org/content/rpsl unless you have licensed * the file under the RealNetworks Community Source License Version 1.0 * (the "RCSL") available at http://www.helixcommunity.org/content/rcsl, * in which case the RCSL will apply. You may also obtain the license terms * directly from RealNetworks. You may not use this file except in * compliance with the RPSL or, if you have a valid RCSL with RealNetworks * applicable to this file, the RCSL. Please see the applicable RPSL or * RCSL for the rights, obligations and limitations governing use of the * contents of the file. * * This file is part of the Helix DNA Technology. RealNetworks is the * developer of the Original Code and owns the copyrights in the portions * it created. * * This file, and the files included with this file, is distributed and made * available on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER * EXPRESS OR IMPLIED, AND REALNETWORKS HEREBY DISCLAIMS ALL SUCH WARRANTIES, * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, FITNESS * FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. * * Technology Compatibility Kit Test Suite(s) Location: * http://www.helixcommunity.org/content/tck * * Contributor(s): * * ***** END LICENSE BLOCK ***** */ /************************************************************************************** * Fixed-point MP3 decoder * Jon Recker (jrecker@real.com), Ken Cooke (kenc@real.com) * August 2003 * * dqchan.c - dequantization of transform coefficients **************************************************************************************/ #include "coder.h" #include "assembly.h" typedef int ARRAY3[3]; /* for short-block reordering */ /* optional pre-emphasis for high-frequency scale factor bands */ static const char preTab[22] = { 0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,2,2,3,3,3,2,0 }; /* pow(2,-i/4) for i=0..3, Q31 format */ static const int pow14[4] = { 0x7fffffff, 0x6ba27e65, 0x5a82799a, 0x4c1bf829 }; /* pow(2,-i/4) * pow(j,4/3) for i=0..3 j=0..15, Q25 format */ static const int pow43_14[4][16] = { { 0x00000000, 0x10000000, 0x285145f3, 0x453a5cdb, /* Q28 */ 0x0cb2ff53, 0x111989d6, 0x15ce31c8, 0x1ac7f203, 0x20000000, 0x257106b9, 0x2b16b4a3, 0x30ed74b4, 0x36f23fa5, 0x3d227bd3, 0x437be656, 0x49fc823c, }, { 0x00000000, 0x0d744fcd, 0x21e71f26, 0x3a36abd9, 0x0aadc084, 0x0e610e6e, 0x12560c1d, 0x168523cf, 0x1ae89f99, 0x1f7c03a4, 0x243bae49, 0x29249c67, 0x2e34420f, 0x33686f85, 0x38bf3dff, 0x3e370182, }, { 0x00000000, 0x0b504f33, 0x1c823e07, 0x30f39a55, 0x08facd62, 0x0c176319, 0x0f6b3522, 0x12efe2ad, 0x16a09e66, 0x1a79a317, 0x1e77e301, 0x2298d5b4, 0x26da56fc, 0x2b3a902a, 0x2fb7e7e7, 0x3450f650, }, { 0x00000000, 0x09837f05, 0x17f910d7, 0x2929c7a9, 0x078d0dfa, 0x0a2ae661, 0x0cf73154, 0x0fec91cb, 0x1306fe0a, 0x16434a6c, 0x199ee595, 0x1d17ae3d, 0x20abd76a, 0x2459d551, 0x28204fbb, 0x2bfe1808, }, }; /* pow(j,4/3) for j=16..63, Q23 format */ static const int pow43[] = { 0x1428a2fa, 0x15db1bd6, 0x1796302c, 0x19598d85, 0x1b24e8bb, 0x1cf7fcfa, 0x1ed28af2, 0x20b4582a, 0x229d2e6e, 0x248cdb55, 0x26832fda, 0x28800000, 0x2a832287, 0x2c8c70a8, 0x2e9bc5d8, 0x30b0ff99, 0x32cbfd4a, 0x34eca001, 0x3712ca62, 0x393e6088, 0x3b6f47e0, 0x3da56717, 0x3fe0a5fc, 0x4220ed72, 0x44662758, 0x46b03e7c, 0x48ff1e87, 0x4b52b3f3, 0x4daaebfd, 0x5007b497, 0x5268fc62, 0x54ceb29c, 0x5738c721, 0x59a72a59, 0x5c19cd35, 0x5e90a129, 0x610b9821, 0x638aa47f, 0x660db90f, 0x6894c90b, 0x6b1fc80c, 0x6daeaa0d, 0x70416360, 0x72d7e8b0, 0x75722ef9, 0x78102b85, 0x7ab1d3ec, 0x7d571e09, }; /* sqrt(0.5) in Q31 format */ #define SQRTHALF 0x5a82799a /* * Minimax polynomial approximation to pow(x, 4/3), over the range * poly43lo: x = [0.5, 0.7071] * poly43hi: x = [0.7071, 1.0] * * Relative error < 1E-7 * Coefs are scaled by 4, 2, 1, 0.5, 0.25 */ static const int poly43lo[5] = { 0x29a0bda9, 0xb02e4828, 0x5957aa1b, 0x236c498d, 0xff581859 }; static const int poly43hi[5] = { 0x10852163, 0xd333f6a4, 0x46e9408b, 0x27c2cef0, 0xfef577b4 }; /* pow(2, i*4/3) as exp and frac */ static const int pow2exp[8] = { 14, 13, 11, 10, 9, 7, 6, 5 }; static const int pow2frac[8] = { 0x6597fa94, 0x50a28be6, 0x7fffffff, 0x6597fa94, 0x50a28be6, 0x7fffffff, 0x6597fa94, 0x50a28be6 }; /************************************************************************************** * Function: DequantBlock * * Description: Ken's highly-optimized, low memory dequantizer performing the operation * y = pow(x, 4.0/3.0) * pow(2, 25 - scale/4.0) * * Inputs: input buffer of decode Huffman codewords (signed-magnitude) * output buffer of same length (in-place (outbuf = inbuf) is allowed) * number of samples * * Outputs: dequantized samples in Q25 format * * Return: bitwise-OR of the unsigned outputs (for guard bit calculations) **************************************************************************************/ static int DequantBlock(int *inbuf, int *outbuf, int num, int scale) { int tab4[4]; int scalef, scalei, shift; int sx, x, y; int mask = 0; const int *tab16, *coef; tab16 = pow43_14[scale & 0x3]; scalef = pow14[scale & 0x3]; scalei = MIN(scale >> 2, 31); /* smallest input scale = -47, so smallest scalei = -12 */ /* cache first 4 values */ shift = MIN(scalei + 3, 31); shift = MAX(shift, 0); tab4[0] = 0; tab4[1] = tab16[1] >> shift; tab4[2] = tab16[2] >> shift; tab4[3] = tab16[3] >> shift; do { sx = *inbuf++; x = sx & 0x7fffffff; /* sx = sign|mag */ if (x < 4) { y = tab4[x]; } else if (x < 16) { y = tab16[x]; y = (scalei < 0) ? y << -scalei : y >> scalei; } else { if (x < 64) { y = pow43[x-16]; /* fractional scale */ y = MULSHIFT32(y, scalef); shift = scalei - 3; } else { /* normalize to [0x40000000, 0x7fffffff] */ x <<= 17; shift = 0; if (x < 0x08000000) x <<= 4, shift += 4; if (x < 0x20000000) x <<= 2, shift += 2; if (x < 0x40000000) x <<= 1, shift += 1; coef = (x < SQRTHALF) ? poly43lo : poly43hi; /* polynomial */ y = coef[0]; y = MULSHIFT32(y, x) + coef[1]; y = MULSHIFT32(y, x) + coef[2]; y = MULSHIFT32(y, x) + coef[3]; y = MULSHIFT32(y, x) + coef[4]; y = MULSHIFT32(y, pow2frac[shift]) << 3; /* fractional scale */ y = MULSHIFT32(y, scalef); shift = scalei - pow2exp[shift]; } /* integer scale */ if (shift < 0) { shift = -shift; if (y > (0x7fffffff >> shift)) y = 0x7fffffff; /* clip */ else y <<= shift; } else { y >>= shift; } } /* sign and store */ mask |= y; *outbuf++ = (sx < 0) ? -y : y; } while (--num); return mask; } /************************************************************************************** * Function: DequantChannel * * Description: dequantize one granule, one channel worth of decoded Huffman codewords * * Inputs: sample buffer (decoded Huffman codewords), length = MAX_NSAMP samples * work buffer for reordering short-block, length = MAX_REORDER_SAMPS * samples (3 * width of largest short-block critical band) * non-zero bound for this channel/granule * valid FrameHeader, SideInfoSub, ScaleFactorInfoSub, and CriticalBandInfo * structures for this channel/granule * * Outputs: MAX_NSAMP dequantized samples in sampleBuf * updated non-zero bound (indicating which samples are != 0 after DQ) * filled-in cbi structure indicating start and end critical bands * * Return: minimum number of guard bits in dequantized sampleBuf * * Notes: dequantized samples in Q(DQ_FRACBITS_OUT) format **************************************************************************************/ int DequantChannel(int *sampleBuf, int *workBuf, int *nonZeroBound, FrameHeader *fh, SideInfoSub *sis, ScaleFactorInfoSub *sfis, CriticalBandInfo *cbi) { int i, j, w, cb; int cbStartL, cbEndL, cbStartS, cbEndS; int nSamps, nonZero, sfactMultiplier, gbMask; int globalGain, gainI; int cbMax[3]; ARRAY3 *buf; /* short block reorder */ /* set default start/end points for short/long blocks - will update with non-zero cb info */ if (sis->blockType == 2) { cbStartL = 0; if (sis->mixedBlock) { cbEndL = (fh->ver == MPEG1 ? 8 : 6); cbStartS = 3; } else { cbEndL = 0; cbStartS = 0; } cbEndS = 13; } else { /* long block */ cbStartL = 0; cbEndL = 22; cbStartS = 13; cbEndS = 13; } cbMax[2] = cbMax[1] = cbMax[0] = 0; gbMask = 0; i = 0; /* sfactScale = 0 --> quantizer step size = 2 * sfactScale = 1 --> quantizer step size = sqrt(2) * so sfactMultiplier = 2 or 4 (jump through globalGain by powers of 2 or sqrt(2)) */ sfactMultiplier = 2 * (sis->sfactScale + 1); /* offset globalGain by -2 if midSide enabled, for 1/sqrt(2) used in MidSideProc() * (DequantBlock() does 0.25 * gainI so knocking it down by two is the same as * dividing every sample by sqrt(2) = multiplying by 2^-.5) */ globalGain = sis->globalGain; if (fh->modeExt >> 1) globalGain -= 2; globalGain += IMDCT_SCALE; /* scale everything by sqrt(2), for fast IMDCT36 */ /* long blocks */ for (cb = 0; cb < cbEndL; cb++) { nonZero = 0; nSamps = fh->sfBand->l[cb + 1] - fh->sfBand->l[cb]; gainI = 210 - globalGain + sfactMultiplier * (sfis->l[cb] + (sis->preFlag ? (int)preTab[cb] : 0)); nonZero |= DequantBlock(sampleBuf + i, sampleBuf + i, nSamps, gainI); i += nSamps; /* update highest non-zero critical band */ if (nonZero) cbMax[0] = cb; gbMask |= nonZero; if (i >= *nonZeroBound) break; } /* set cbi (Type, EndS[], EndSMax will be overwritten if we proceed to do short blocks) */ cbi->cbType = 0; /* long only */ cbi->cbEndL = cbMax[0]; cbi->cbEndS[0] = cbi->cbEndS[1] = cbi->cbEndS[2] = 0; cbi->cbEndSMax = 0; /* early exit if no short blocks */ if (cbStartS >= 12) return CLZ(gbMask) - 1; /* short blocks */ cbMax[2] = cbMax[1] = cbMax[0] = cbStartS; for (cb = cbStartS; cb < cbEndS; cb++) { nSamps = fh->sfBand->s[cb + 1] - fh->sfBand->s[cb]; for (w = 0; w < 3; w++) { nonZero = 0; gainI = 210 - globalGain + 8*sis->subBlockGain[w] + sfactMultiplier*(sfis->s[cb][w]); nonZero |= DequantBlock(sampleBuf + i + nSamps*w, workBuf + nSamps*w, nSamps, gainI); /* update highest non-zero critical band */ if (nonZero) cbMax[w] = cb; gbMask |= nonZero; } /* reorder blocks */ buf = (ARRAY3 *)(sampleBuf + i); i += 3*nSamps; for (j = 0; j < nSamps; j++) { buf[j][0] = workBuf[0*nSamps + j]; buf[j][1] = workBuf[1*nSamps + j]; buf[j][2] = workBuf[2*nSamps + j]; } ASSERT(3*nSamps <= MAX_REORDER_SAMPS); if (i >= *nonZeroBound) break; } /* i = last non-zero INPUT sample processed, which corresponds to highest possible non-zero * OUTPUT sample (after reorder) * however, the original nzb is no longer necessarily true * for each cb, buf[][] is updated with 3*nSamps samples (i increases 3*nSamps each time) * (buf[j + 1][0] = 3 (input) samples ahead of buf[j][0]) * so update nonZeroBound to i */ *nonZeroBound = i; ASSERT(*nonZeroBound <= MAX_NSAMP); cbi->cbType = (sis->mixedBlock ? 2 : 1); /* 2 = mixed short/long, 1 = short only */ cbi->cbEndS[0] = cbMax[0]; cbi->cbEndS[1] = cbMax[1]; cbi->cbEndS[2] = cbMax[2]; cbi->cbEndSMax = cbMax[0]; cbi->cbEndSMax = MAX(cbi->cbEndSMax, cbMax[1]); cbi->cbEndSMax = MAX(cbi->cbEndSMax, cbMax[2]); return CLZ(gbMask) - 1; }