diff options
Diffstat (limited to 'media/libstagefright/codecs/mp3dec/src/pvmp3_mdct_18.cpp')
-rw-r--r-- | media/libstagefright/codecs/mp3dec/src/pvmp3_mdct_18.cpp | 289 |
1 files changed, 289 insertions, 0 deletions
diff --git a/media/libstagefright/codecs/mp3dec/src/pvmp3_mdct_18.cpp b/media/libstagefright/codecs/mp3dec/src/pvmp3_mdct_18.cpp new file mode 100644 index 0000000..09a735b --- /dev/null +++ b/media/libstagefright/codecs/mp3dec/src/pvmp3_mdct_18.cpp @@ -0,0 +1,289 @@ +/* ------------------------------------------------------------------ + * Copyright (C) 1998-2009 PacketVideo + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either + * express or implied. + * See the License for the specific language governing permissions + * and limitations under the License. + * ------------------------------------------------------------------- + */ +/* +------------------------------------------------------------------------------ + + PacketVideo Corp. + MP3 Decoder Library + + Filename: mdct_18.cpp + + Date: 09/21/2007 + +------------------------------------------------------------------------------ + REVISION HISTORY + + + Description: + +------------------------------------------------------------------------------ + INPUT AND OUTPUT DEFINITIONS + +Input + int32 vec[], input vector of length 18 + int32 *history input for overlap and add, vector updated with + next overlap and add values + const int32 *window sine window used in the mdct, three types are allowed + noraml, start and stop +Returns + none mdct computation in-place + + +------------------------------------------------------------------------------ + FUNCTION DESCRIPTION + + Returns the mdct of length 18 of the input vector, as well as the overlap + vector for next iteration ( on history[]) + +------------------------------------------------------------------------------ + REQUIREMENTS + + +------------------------------------------------------------------------------ + REFERENCES + +------------------------------------------------------------------------------ + PSEUDO-CODE + +------------------------------------------------------------------------------ +*/ + +#if ( !defined(PV_ARM_GCC_V5) && !defined(PV_ARM_GCC_V4) && !defined(PV_ARM_V5) && !defined(PV_ARM_V4) ) +/*---------------------------------------------------------------------------- +; INCLUDES +----------------------------------------------------------------------------*/ + +#include "pv_mp3dec_fxd_op.h" +#include "pvmp3_mdct_18.h" + + +/*---------------------------------------------------------------------------- +; MACROS +; Define module specific macros here +----------------------------------------------------------------------------*/ + + +/*---------------------------------------------------------------------------- +; DEFINES +; Include all pre-processor statements here. Include conditional +; compile variables also. +----------------------------------------------------------------------------*/ + +/*---------------------------------------------------------------------------- +; LOCAL FUNCTION DEFINITIONS +; Function Prototype declaration +----------------------------------------------------------------------------*/ + +/*---------------------------------------------------------------------------- +; LOCAL STORE/BUFFER/POINTER DEFINITIONS +; Variable declaration - defined here and used outside this module +----------------------------------------------------------------------------*/ +const int32 cosTerms_dct18[9] = +{ + Qfmt(0.50190991877167f), Qfmt(0.51763809020504f), Qfmt(0.55168895948125f), + Qfmt(0.61038729438073f), Qfmt(0.70710678118655f), Qfmt(0.87172339781055f), + Qfmt(1.18310079157625f), Qfmt(1.93185165257814f), Qfmt(5.73685662283493f) +}; + + +const int32 cosTerms_1_ov_cos_phi[18] = +{ + + Qfmt1(0.50047634258166f), Qfmt1(0.50431448029008f), Qfmt1(0.51213975715725f), + Qfmt1(0.52426456257041f), Qfmt1(0.54119610014620f), Qfmt1(0.56369097343317f), + Qfmt1(0.59284452371708f), Qfmt1(0.63023620700513f), Qfmt1(0.67817085245463f), + + Qfmt2(0.74009361646113f), Qfmt2(0.82133981585229f), Qfmt2(0.93057949835179f), + Qfmt2(1.08284028510010f), Qfmt2(1.30656296487638f), Qfmt2(1.66275476171152f), + Qfmt2(2.31011315767265f), Qfmt2(3.83064878777019f), Qfmt2(11.46279281302667f) +}; + +/*---------------------------------------------------------------------------- +; EXTERNAL FUNCTION REFERENCES +; Declare functions defined elsewhere and referenced in this module +----------------------------------------------------------------------------*/ + +/*---------------------------------------------------------------------------- +; EXTERNAL GLOBAL STORE/BUFFER/POINTER REFERENCES +; Declare variables used in this module but defined elsewhere +----------------------------------------------------------------------------*/ + +/*---------------------------------------------------------------------------- +; FUNCTION CODE +----------------------------------------------------------------------------*/ + + + +void pvmp3_mdct_18(int32 vec[], int32 *history, const int32 *window) +{ + int32 i; + int32 tmp; + int32 tmp1; + int32 tmp2; + int32 tmp3; + int32 tmp4; + + + + const int32 *pt_cos_split = cosTerms_dct18; + const int32 *pt_cos = cosTerms_1_ov_cos_phi; + const int32 *pt_cos_x = &cosTerms_1_ov_cos_phi[17]; + int32 *pt_vec = vec; + int32 *pt_vec_o = &vec[17]; + + + for (i = 9; i != 0; i--) + { + tmp = *(pt_vec); + tmp1 = *(pt_vec_o); + tmp = fxp_mul32_Q32(tmp << 1, *(pt_cos++)); + tmp1 = fxp_mul32_Q27(tmp1, *(pt_cos_x--)); + *(pt_vec++) = tmp + tmp1 ; + *(pt_vec_o--) = fxp_mul32_Q28((tmp - tmp1), *(pt_cos_split++)); + } + + + pvmp3_dct_9(vec); // Even terms + pvmp3_dct_9(&vec[9]); // Odd terms + + + tmp3 = vec[16]; // + vec[16] = vec[ 8]; + tmp4 = vec[14]; // + vec[14] = vec[ 7]; + tmp = vec[12]; + vec[12] = vec[ 6]; + tmp2 = vec[10]; // vec[10] + vec[10] = vec[ 5]; + vec[ 8] = vec[ 4]; + vec[ 6] = vec[ 3]; + vec[ 4] = vec[ 2]; + vec[ 2] = vec[ 1]; + vec[ 1] = vec[ 9] - tmp2; // vec[9] + vec[10] + vec[ 3] = vec[11] - tmp2; + vec[ 5] = vec[11] - tmp; + vec[ 7] = vec[13] - tmp; + vec[ 9] = vec[13] - tmp4; + vec[11] = vec[15] - tmp4; + vec[13] = vec[15] - tmp3; + vec[15] = vec[17] - tmp3; + + + /* overlap and add */ + + tmp2 = vec[0]; + tmp3 = vec[9]; + + for (i = 0; i < 6; i++) + { + tmp = history[ i]; + tmp4 = vec[i+10]; + vec[i+10] = tmp3 + tmp4; + tmp1 = vec[i+1]; + vec[ i] = fxp_mac32_Q32(tmp, (vec[i+10]), window[ i]); + tmp3 = tmp4; + history[i ] = -(tmp2 + tmp1); + tmp2 = tmp1; + } + + tmp = history[ 6]; + tmp4 = vec[16]; + vec[16] = tmp3 + tmp4; + tmp1 = vec[7]; + vec[ 6] = fxp_mac32_Q32(tmp, vec[16] << 1, window[ i]); + tmp = history[ 7]; + history[6] = -(tmp2 + tmp1); + history[7] = -(tmp1 + vec[8]); + + tmp1 = history[ 8]; + tmp4 = vec[17] + tmp4; + vec[ 7] = fxp_mac32_Q32(tmp, tmp4 << 1, window[ 7]); + history[8] = -(vec[8] + vec[9]); + vec[ 8] = fxp_mac32_Q32(tmp1, vec[17] << 1, window[ 8]); + + tmp = history[9]; + tmp1 = history[17]; + tmp2 = history[16]; + vec[ 9] = fxp_mac32_Q32(tmp, vec[17] << 1, window[ 9]); + + vec[17] = fxp_mac32_Q32(tmp1, vec[10] << 1, window[17]); + vec[10] = -vec[ 16]; + vec[16] = fxp_mac32_Q32(tmp2, vec[11] << 1, window[16]); + tmp1 = history[15]; + tmp2 = history[14]; + vec[11] = -vec[ 15]; + vec[15] = fxp_mac32_Q32(tmp1, vec[12] << 1, window[15]); + vec[12] = -vec[ 14]; + vec[14] = fxp_mac32_Q32(tmp2, vec[13] << 1, window[14]); + + tmp = history[13]; + tmp1 = history[12]; + tmp2 = history[11]; + tmp3 = history[10]; + vec[13] = fxp_mac32_Q32(tmp, vec[12] << 1, window[13]); + vec[12] = fxp_mac32_Q32(tmp1, vec[11] << 1, window[12]); + vec[11] = fxp_mac32_Q32(tmp2, vec[10] << 1, window[11]); + vec[10] = fxp_mac32_Q32(tmp3, tmp4 << 1, window[10]); + + + /* next iteration overlap */ + + tmp1 = history[ 8]; + tmp3 = history[ 7]; + tmp2 = history[ 1]; + tmp = history[ 0]; + tmp1 <<= 1; + tmp3 <<= 1; + + history[ 0] = fxp_mul32_Q32(tmp1, window[18]); + history[17] = fxp_mul32_Q32(tmp1, window[35]); + history[ 1] = fxp_mul32_Q32(tmp3, window[19]); + history[16] = fxp_mul32_Q32(tmp3, window[34]); + + tmp2 <<= 1; + tmp <<= 1; + history[ 7] = fxp_mul32_Q32(tmp2, window[25]); + history[10] = fxp_mul32_Q32(tmp2, window[28]); + history[ 8] = fxp_mul32_Q32(tmp, window[26]); + history[ 9] = fxp_mul32_Q32(tmp, window[27]); + + tmp1 = history[ 6]; + tmp3 = history[ 5]; + tmp4 = history[ 4]; + tmp2 = history[ 3]; + tmp = history[ 2]; + + tmp1 <<= 1; + tmp3 <<= 1; + tmp4 <<= 1; + + history[ 2] = fxp_mul32_Q32(tmp1, window[20]); + history[15] = fxp_mul32_Q32(tmp1, window[33]); + history[ 3] = fxp_mul32_Q32(tmp3, window[21]); + history[14] = fxp_mul32_Q32(tmp3, window[32]); + history[ 4] = fxp_mul32_Q32(tmp4, window[22]); + history[13] = fxp_mul32_Q32(tmp4, window[31]); + tmp2 <<= 1; + tmp <<= 1; + history[ 5] = fxp_mul32_Q32(tmp2, window[23]); + history[12] = fxp_mul32_Q32(tmp2, window[30]); + history[ 6] = fxp_mul32_Q32(tmp, window[24]); + history[11] = fxp_mul32_Q32(tmp, window[29]); +} + +#endif // If not assembly |