diff options
| author | Antonin Descampe <antonin@gmail.com> | 2008-07-31 18:47:41 +0000 |
|---|---|---|
| committer | Antonin Descampe <antonin@gmail.com> | 2008-07-31 18:47:41 +0000 |
| commit | 7caaea18b728ead4e45226aafc09dba01e514a2d (patch) | |
| tree | 1d5d19df20b82cee63a7de94fc8e04e8426cb7c7 /libopenjpeg/dwt.c | |
| parent | 8407e057c9e10c8fc2b5f4be56e603fae3f9063d (diff) | |
Deleting obsolete files and directories, adding v2-specific files and directories, updating existing files to v2. See README.v2 for more info
Diffstat (limited to 'libopenjpeg/dwt.c')
| -rw-r--r-- | libopenjpeg/dwt.c | 510 |
1 files changed, 279 insertions, 231 deletions
diff --git a/libopenjpeg/dwt.c b/libopenjpeg/dwt.c index 78d18d17..a6ffd180 100644 --- a/libopenjpeg/dwt.c +++ b/libopenjpeg/dwt.c @@ -7,6 +7,7 @@ * Copyright (c) 2005, Herve Drolon, FreeImage Team * Copyright (c) 2007, Jonathan Ballard <dzonatas@dzonux.net> * Copyright (c) 2007, Callum Lerwick <seg@haxxed.com> + * Copyright (c) 2008, Jerome Fimes, Communications & Systemes <jerome.fimes@c-s.fr> * All rights reserved. * * Redistribution and use in source and binary forms, with or without @@ -35,7 +36,12 @@ #include <xmmintrin.h> #endif -#include "opj_includes.h" +#include "dwt.h" +#include "j2k.h" +#include "tcd.h" +#include "fix.h" +#include "opj_malloc.h" +#include "int.h" /** @defgroup DWT DWT - Implementation of a discrete wavelet transform */ /*@{*/ @@ -47,31 +53,31 @@ /*@{*/ typedef struct dwt_local { - int* mem; - int dn; - int sn; - int cas; + OPJ_INT32* mem; + OPJ_INT32 dn; + OPJ_INT32 sn; + OPJ_INT32 cas; } dwt_t; typedef union { - float f[4]; + OPJ_FLOAT32 f[4]; } v4; typedef struct v4dwt_local { v4* wavelet ; - int dn ; - int sn ; - int cas ; + OPJ_INT32 dn ; + OPJ_INT32 sn ; + OPJ_INT32 cas ; } v4dwt_t ; -static const float dwt_alpha = 1.586134342f; // 12994 -static const float dwt_beta = 0.052980118f; // 434 -static const float dwt_gamma = -0.882911075f; // -7233 -static const float dwt_delta = -0.443506852f; // -3633 +static const OPJ_FLOAT32 dwt_alpha = 1.586134342f; // 12994 +static const OPJ_FLOAT32 dwt_beta = 0.052980118f; // 434 +static const OPJ_FLOAT32 dwt_gamma = -0.882911075f; // -7233 +static const OPJ_FLOAT32 delta = -0.443506852f; // -3633 -static const float K = 1.230174105f; // 10078 +static const OPJ_FLOAT32 K = 1.230174105f; // 10078 /* FIXME: What is this constant? */ -static const float c13318 = 1.625732422f; +static const OPJ_FLOAT32 c13318 = 1.625732422f; /*@}*/ @@ -86,23 +92,23 @@ typedef void (*DWT1DFN)(dwt_t* v); /** Forward lazy transform (horizontal) */ -static void dwt_deinterleave_h(int *a, int *b, int dn, int sn, int cas); +static void dwt_deinterleave_h(OPJ_INT32 *a, OPJ_INT32 *b, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 cas); /** Forward lazy transform (vertical) */ -static void dwt_deinterleave_v(int *a, int *b, int dn, int sn, int x, int cas); +static void dwt_deinterleave_v(OPJ_INT32 *a, OPJ_INT32 *b, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 x, OPJ_INT32 cas); /** Inverse lazy transform (horizontal) */ -static void dwt_interleave_h(dwt_t* h, int *a); +static void dwt_interleave_h(dwt_t* h, OPJ_INT32 *a); /** Inverse lazy transform (vertical) */ -static void dwt_interleave_v(dwt_t* v, int *a, int x); +static void dwt_interleave_v(dwt_t* v, OPJ_INT32 *a, OPJ_INT32 x); /** Forward 5-3 wavelet transform in 1-D */ -static void dwt_encode_1(int *a, int dn, int sn, int cas); +static void dwt_encode_1(OPJ_INT32 *a, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 cas); /** Inverse 5-3 wavelet transform in 1-D */ @@ -110,16 +116,19 @@ static void dwt_decode_1(dwt_t *v); /** Forward 9-7 wavelet transform in 1-D */ -static void dwt_encode_1_real(int *a, int dn, int sn, int cas); +static void dwt_encode_1_real(OPJ_INT32 *a, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 cas); /** Explicit calculation of the Quantization Stepsizes */ -static void dwt_encode_stepsize(int stepsize, int numbps, opj_stepsize_t *bandno_stepsize); +static void dwt_encode_stepsize(OPJ_INT32 stepsize, OPJ_INT32 numbps, opj_stepsize_t *bandno_stepsize); /** Inverse wavelet transform in 2-D. */ -static void dwt_decode_tile(opj_tcd_tilecomp_t* tilec, int i, DWT1DFN fn); +static bool dwt_decode_tile(opj_tcd_tilecomp_t* tilec, OPJ_UINT32 i, DWT1DFN fn); +static OPJ_UINT32 dwt_max_resolution(opj_tcd_resolution_t* restrict r, OPJ_UINT32 i); + +static INLINE bool dwt_encode_procedure(opj_tcd_tilecomp_t * tilec,void (*p_function)(OPJ_INT32 *, OPJ_INT32,OPJ_INT32,OPJ_INT32) ); /*@}*/ /*@}*/ @@ -135,7 +144,7 @@ static void dwt_decode_tile(opj_tcd_tilecomp_t* tilec, int i, DWT1DFN fn); /* <summary> */ /* This table contains the norms of the 5-3 wavelets for different bands. */ /* </summary> */ -static const double dwt_norms[4][10] = { +static const OPJ_FLOAT64 dwt_norms[4][10] = { {1.000, 1.500, 2.750, 5.375, 10.68, 21.34, 42.67, 85.33, 170.7, 341.3}, {1.038, 1.592, 2.919, 5.703, 11.33, 22.64, 45.25, 90.48, 180.9}, {1.038, 1.592, 2.919, 5.703, 11.33, 22.64, 45.25, 90.48, 180.9}, @@ -145,7 +154,7 @@ static const double dwt_norms[4][10] = { /* <summary> */ /* This table contains the norms of the 9-7 wavelets for different bands. */ /* </summary> */ -static const double dwt_norms_real[4][10] = { +static const OPJ_FLOAT64 dwt_norms_real[4][10] = { {1.000, 1.965, 4.177, 8.403, 16.90, 33.84, 67.69, 135.3, 270.6, 540.9}, {2.022, 3.989, 8.355, 17.04, 34.27, 68.63, 137.3, 274.6, 549.0}, {2.022, 3.989, 8.355, 17.04, 34.27, 68.63, 137.3, 274.6, 549.0}, @@ -161,48 +170,88 @@ static const double dwt_norms_real[4][10] = { /* <summary> */ /* Forward lazy transform (horizontal). */ /* </summary> */ -static void dwt_deinterleave_h(int *a, int *b, int dn, int sn, int cas) { - int i; - for (i=0; i<sn; i++) b[i]=a[2*i+cas]; - for (i=0; i<dn; i++) b[sn+i]=a[(2*i+1-cas)]; +static void dwt_deinterleave_h(OPJ_INT32 *a, OPJ_INT32 *b, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 cas) { + OPJ_INT32 i; + + OPJ_INT32 * l_dest = b; + OPJ_INT32 * l_src = a+cas; + for + (i=0; i<sn; ++i) + { + *l_dest++ = *l_src; + l_src += 2; + } + l_dest = b + sn; + l_src = a + 1 - cas; + for + (i=0; i<dn; ++i) + { + *l_dest++=*l_src; + l_src += 2; + } } /* <summary> */ /* Forward lazy transform (vertical). */ /* </summary> */ -static void dwt_deinterleave_v(int *a, int *b, int dn, int sn, int x, int cas) { - int i; - for (i=0; i<sn; i++) b[i*x]=a[2*i+cas]; - for (i=0; i<dn; i++) b[(sn+i)*x]=a[(2*i+1-cas)]; +static void dwt_deinterleave_v(OPJ_INT32 *a, OPJ_INT32 *b, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 x, OPJ_INT32 cas) { + OPJ_INT32 i = sn; + OPJ_INT32 * l_dest = b; + OPJ_INT32 * l_src = a+cas; + + while + (i--) + { + *l_dest = *l_src; + l_dest += x; + l_src += 2; + /* b[i*x]=a[2*i+cas]; */ + } + l_dest = b + sn * x; + l_src = a + 1 - cas; + + i = dn; + while + (i--) + { + *l_dest = *l_src; + l_dest += x; + l_src += 2; + /*b[(sn+i)*x]=a[(2*i+1-cas)];*/ + } } /* <summary> */ /* Inverse lazy transform (horizontal). */ /* </summary> */ -static void dwt_interleave_h(dwt_t* h, int *a) { - int *ai = a; - int *bi = h->mem + h->cas; - int i = h->sn; - while( i-- ) { - *bi = *(ai++); - bi += 2; +static void dwt_interleave_h(dwt_t* h, OPJ_INT32 *a) { + OPJ_INT32 *ai = a; + OPJ_INT32 *bi = h->mem + h->cas; + OPJ_INT32 i = h->sn; + while + ( i-- ) + { + *bi = *(ai++); + bi += 2; } ai = a + h->sn; bi = h->mem + 1 - h->cas; i = h->dn ; - while( i-- ) { - *bi = *(ai++); - bi += 2; + while + ( i-- ) + { + *bi = *(ai++); + bi += 2; } } /* <summary> */ /* Inverse lazy transform (vertical). */ /* </summary> */ -static void dwt_interleave_v(dwt_t* v, int *a, int x) { - int *ai = a; - int *bi = v->mem + v->cas; - int i = v->sn; +static void dwt_interleave_v(dwt_t* v, OPJ_INT32 *a, OPJ_INT32 x) { + OPJ_INT32 *ai = a; + OPJ_INT32 *bi = v->mem + v->cas; + OPJ_INT32 i = v->sn; while( i-- ) { *bi = *ai; bi += 2; @@ -222,8 +271,8 @@ static void dwt_interleave_v(dwt_t* v, int *a, int x) { /* <summary> */ /* Forward 5-3 wavelet transform in 1-D. */ /* </summary> */ -static void dwt_encode_1(int *a, int dn, int sn, int cas) { - int i; +static void dwt_encode_1(OPJ_INT32 *a, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 cas) { + OPJ_INT32 i; if (!cas) { if ((dn > 0) || (sn > 1)) { /* NEW : CASE ONE ELEMENT */ @@ -243,8 +292,8 @@ static void dwt_encode_1(int *a, int dn, int sn, int cas) { /* <summary> */ /* Inverse 5-3 wavelet transform in 1-D. */ /* </summary> */ -static void dwt_decode_1_(int *a, int dn, int sn, int cas) { - int i; +static void dwt_decode_1_(OPJ_INT32 *a, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 cas) { + OPJ_INT32 i; if (!cas) { if ((dn > 0) || (sn > 1)) { /* NEW : CASE ONE ELEMENT */ @@ -271,8 +320,8 @@ static void dwt_decode_1(dwt_t *v) { /* <summary> */ /* Forward 9-7 wavelet transform in 1-D. */ /* </summary> */ -static void dwt_encode_1_real(int *a, int dn, int sn, int cas) { - int i; +static void dwt_encode_1_real(OPJ_INT32 *a, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 cas) { + OPJ_INT32 i; if (!cas) { if ((dn > 0) || (sn > 1)) { /* NEW : CASE ONE ELEMENT */ for (i = 0; i < dn; i++) @@ -306,8 +355,8 @@ static void dwt_encode_1_real(int *a, int dn, int sn, int cas) { } } -static void dwt_encode_stepsize(int stepsize, int numbps, opj_stepsize_t *bandno_stepsize) { - int p, n; +static void dwt_encode_stepsize(OPJ_INT32 stepsize, OPJ_INT32 numbps, opj_stepsize_t *bandno_stepsize) { + OPJ_INT32 p, n; p = int_floorlog2(stepsize) - 13; n = 11 - int_floorlog2(stepsize); bandno_stepsize->mant = (n < 0 ? stepsize >> -n : stepsize << n) & 0x7ff; @@ -323,71 +372,105 @@ static void dwt_encode_stepsize(int stepsize, int numbps, opj_stepsize_t *bandno /* <summary> */ /* Forward 5-3 wavelet transform in 2-D. */ /* </summary> */ -void dwt_encode(opj_tcd_tilecomp_t * tilec) { - int i, j, k; - int *a = NULL; - int *aj = NULL; - int *bj = NULL; - int w, l; +INLINE bool dwt_encode_procedure(opj_tcd_tilecomp_t * tilec,void (*p_function)(OPJ_INT32 *, OPJ_INT32,OPJ_INT32,OPJ_INT32) ) +{ + OPJ_INT32 i, j, k; + OPJ_INT32 *a = 00; + OPJ_INT32 *aj = 00; + OPJ_INT32 *bj = 00; + OPJ_INT32 w, l; + + OPJ_INT32 rw; /* width of the resolution level computed */ + OPJ_INT32 rh; /* height of the resolution level computed */ + OPJ_INT32 l_data_size; + opj_tcd_resolution_t * l_cur_res = 0; + opj_tcd_resolution_t * l_last_res = 0; + w = tilec->x1-tilec->x0; l = tilec->numresolutions-1; a = tilec->data; - for (i = 0; i < l; i++) { - int rw; /* width of the resolution level computed */ - int rh; /* height of the resolution level computed */ - int rw1; /* width of the resolution level once lower than computed one */ - int rh1; /* height of the resolution level once lower than computed one */ - int cas_col; /* 0 = non inversion on horizontal filtering 1 = inversion between low-pass and high-pass filtering */ - int cas_row; /* 0 = non inversion on vertical filtering 1 = inversion between low-pass and high-pass filtering */ - int dn, sn; + l_cur_res = tilec->resolutions + l; + l_last_res = l_cur_res - 1; + + rw = l_cur_res->x1 - l_cur_res->x0; + rh = l_cur_res->y1 - l_cur_res->y0; + + l_data_size = dwt_max_resolution( tilec->resolutions,tilec->numresolutions) * sizeof(OPJ_INT32); + bj = opj_malloc(l_data_size); + if + (! bj) + { + return false; + } + i = l; + + while + (i--) + { + OPJ_INT32 rw1; /* width of the resolution level once lower than computed one */ + OPJ_INT32 rh1; /* height of the resolution level once lower than computed one */ + OPJ_INT32 cas_col; /* 0 = non inversion on horizontal filtering 1 = inversion between low-pass and high-pass filtering */ + OPJ_INT32 cas_row; /* 0 = non inversion on vertical filtering 1 = inversion between low-pass and high-pass filtering */ + OPJ_INT32 dn, sn; - rw = tilec->resolutions[l - i].x1 - tilec->resolutions[l - i].x0; - rh = tilec->resolutions[l - i].y1 - tilec->resolutions[l - i].y0; - rw1= tilec->resolutions[l - i - 1].x1 - tilec->resolutions[l - i - 1].x0; - rh1= tilec->resolutions[l - i - 1].y1 - tilec->resolutions[l - i - 1].y0; + rw = l_cur_res->x1 - l_cur_res->x0; + rh = l_cur_res->y1 - l_cur_res->y0; + rw1 = l_last_res->x1 - l_last_res->x0; + rh1 = l_last_res->y1 - l_last_res->y0; - cas_row = tilec->resolutions[l - i].x0 % 2; - cas_col = tilec->resolutions[l - i].y0 % 2; + cas_row = l_cur_res->x0 & 1; + cas_col = l_cur_res->y0 & 1; sn = rh1; dn = rh - rh1; - bj = (int*)opj_malloc(rh * sizeof(int)); - for (j = 0; j < rw; j++) { + for + (j = 0; j < rw; ++j) + { aj = a + j; - for (k = 0; k < rh; k++) bj[k] = aj[k*w]; - dwt_encode_1(bj, dn, sn, cas_col); + for + (k = 0; k < rh; ++k) + { + bj[k] = aj[k*w]; + } + (*p_function) (bj, dn, sn, cas_col); dwt_deinterleave_v(bj, aj, dn, sn, w, cas_col); } - opj_free(bj); - sn = rw1; dn = rw - rw1; - bj = (int*)opj_malloc(rw * sizeof(int)); - for (j = 0; j < rh; j++) { + for (j = 0; j < rh; j++) + { aj = a + j * w; for (k = 0; k < rw; k++) bj[k] = aj[k]; - dwt_encode_1(bj, dn, sn, cas_row); + (*p_function) (bj, dn, sn, cas_row); dwt_deinterleave_h(bj, aj, dn, sn, cas_row); } - opj_free(bj); + l_cur_res = l_last_res; + --l_last_res; } + opj_free(bj); + return true; +} +/* Forward 5-3 wavelet transform in 2-D. */ +/* </summary> */ +bool dwt_encode(opj_tcd_tilecomp_t * tilec) +{ + return dwt_encode_procedure(tilec,dwt_encode_1); } - /* <summary> */ /* Inverse 5-3 wavelet transform in 2-D. */ /* </summary> */ -void dwt_decode(opj_tcd_tilecomp_t* tilec, int numres) { - dwt_decode_tile(tilec, numres, &dwt_decode_1); +bool dwt_decode(opj_tcd_tilecomp_t* tilec, OPJ_UINT32 numres) { + return dwt_decode_tile(tilec, numres, &dwt_decode_1); } /* <summary> */ /* Get gain of 5-3 wavelet transform. */ /* </summary> */ -int dwt_getgain(int orient) { +OPJ_UINT32 dwt_getgain(OPJ_UINT32 orient) { if (orient == 0) return 0; if (orient == 1 || orient == 2) @@ -398,71 +481,24 @@ int dwt_getgain(int orient) { /* <summary> */ /* Get norm of 5-3 wavelet. */ /* </summary> */ -double dwt_getnorm(int level, int orient) { +OPJ_FLOAT64 dwt_getnorm(OPJ_UINT32 level, OPJ_UINT32 orient) { return dwt_norms[orient][level]; } /* <summary> */ /* Forward 9-7 wavelet transform in 2-D. */ /* </summary> */ - -void dwt_encode_real(opj_tcd_tilecomp_t * tilec) { - int i, j, k; - int *a = NULL; - int *aj = NULL; - int *bj = NULL; - int w, l; - - w = tilec->x1-tilec->x0; - l = tilec->numresolutions-1; - a = tilec->data; - - for (i = 0; i < l; i++) { - int rw; /* width of the resolution level computed */ - int rh; /* height of the resolution level computed */ - int rw1; /* width of the resolution level once lower than computed one */ - int rh1; /* height of the resolution level once lower than computed one */ - int cas_col; /* 0 = non inversion on horizontal filtering 1 = inversion between low-pass and high-pass filtering */ - int cas_row; /* 0 = non inversion on vertical filtering 1 = inversion between low-pass and high-pass filtering */ - int dn, sn; - - rw = tilec->resolutions[l - i].x1 - tilec->resolutions[l - i].x0; - rh = tilec->resolutions[l - i].y1 - tilec->resolutions[l - i].y0; - rw1= tilec->resolutions[l - i - 1].x1 - tilec->resolutions[l - i - 1].x0; - rh1= tilec->resolutions[l - i - 1].y1 - tilec->resolutions[l - i - 1].y0; - - cas_row = tilec->resolutions[l - i].x0 % 2; - cas_col = tilec->resolutions[l - i].y0 % 2; - - sn = rh1; - dn = rh - rh1; - bj = (int*)opj_malloc(rh * sizeof(int)); - for (j = 0; j < rw; j++) { - aj = a + j; - for (k = 0; k < rh; k++) bj[k] = aj[k*w]; - dwt_encode_1_real(bj, dn, sn, cas_col); - dwt_deinterleave_v(bj, aj, dn, sn, w, cas_col); - } - opj_free(bj); - - sn = rw1; - dn = rw - rw1; - bj = (int*)opj_malloc(rw * sizeof(int)); - for (j = 0; j < rh; j++) { - aj = a + j * w; - for (k = 0; k < rw; k++) bj[k] = aj[k]; - dwt_encode_1_real(bj, dn, sn, cas_row); - dwt_deinterleave_h(bj, aj, dn, sn, cas_row); - } - opj_free(bj); - } +bool dwt_encode_real(opj_tcd_tilecomp_t * tilec) +{ + return dwt_encode_procedure(tilec,dwt_encode_1_real); } + /* <summary> */ /* Get gain of 9-7 wavelet transform. */ /* </summary> */ -int dwt_getgain_real(int orient) { +OPJ_UINT32 dwt_getgain_real(OPJ_UINT32 orient) { (void)orient; return 0; } @@ -470,16 +506,16 @@ int dwt_getgain_real(int orient) { /* <summary> */ /* Get norm of 9-7 wavelet. */ /* </summary> */ -double dwt_getnorm_real(int level, int orient) { +OPJ_FLOAT64 dwt_getnorm_real(OPJ_UINT32 level, OPJ_UINT32 orient) { return dwt_norms_real[orient][level]; } -void dwt_calc_explicit_stepsizes(opj_tccp_t * tccp, int prec) { - int numbands, bandno; +void dwt_calc_explicit_stepsizes(opj_tccp_t * tccp, OPJ_UINT32 prec) { + OPJ_UINT32 numbands, bandno; numbands = 3 * tccp->numresolutions - 2; for (bandno = 0; bandno < numbands; bandno++) { - double stepsize; - int resno, level, orient, gain; + OPJ_FLOAT64 stepsize; + OPJ_UINT32 resno, level, orient, gain; resno = (bandno == 0) ? 0 : ((bandno - 1) / 3 + 1); orient = (bandno == 0) ? 0 : ((bandno - 1) % 3 + 1); @@ -488,10 +524,10 @@ void dwt_calc_explicit_stepsizes(opj_tccp_t * tccp, int prec) { if (tccp->qntsty == J2K_CCP_QNTSTY_NOQNT) { stepsize = 1.0; } else { - double norm = dwt_norms_real[orient][level]; + OPJ_FLOAT64 norm = dwt_norms_real[orient][level]; stepsize = (1 << (gain)) / norm; } - dwt_encode_stepsize((int) floor(stepsize * 8192.0), prec + gain, &tccp->stepsizes[bandno]); + dwt_encode_stepsize((OPJ_INT32) floor(stepsize * 8192.0), prec + gain, &tccp->stepsizes[bandno]); } } @@ -499,11 +535,11 @@ void dwt_calc_explicit_stepsizes(opj_tccp_t * tccp, int prec) { /* <summary> */ /* Determine maximum computed resolution level for inverse wavelet transform */ /* </summary> */ -static int dwt_decode_max_resolution(opj_tcd_resolution_t* restrict r, int i) { - int mr = 1; - int w; +static OPJ_UINT32 dwt_max_resolution(opj_tcd_resolution_t* restrict r, OPJ_UINT32 i) { + OPJ_UINT32 mr = 0; + OPJ_UINT32 w; while( --i ) { - r++; + ++r; if( mr < ( w = r->x1 - r->x0 ) ) mr = w ; if( mr < ( w = r->y1 - r->y0 ) ) @@ -516,23 +552,29 @@ static int dwt_decode_max_resolution(opj_tcd_resolution_t* restrict r, int i) { /* <summary> */ /* Inverse wavelet transform in 2-D. */ /* </summary> */ -static void dwt_decode_tile(opj_tcd_tilecomp_t* tilec, int numres, DWT1DFN dwt_1D) { +static bool dwt_decode_tile(opj_tcd_tilecomp_t* tilec, OPJ_UINT32 numres, DWT1DFN dwt_1D) { dwt_t h; dwt_t v; opj_tcd_resolution_t* tr = tilec->resolutions; - int rw = tr->x1 - tr->x0; /* width of the resolution level computed */ - int rh = tr->y1 - tr->y0; /* height of the resolution level computed */ + OPJ_UINT32 rw = tr->x1 - tr->x0; /* width of the resolution level computed */ + OPJ_UINT32 rh = tr->y1 - tr->y0; /* height of the resolution level computed */ - int w = tilec->x1 - tilec->x0; + OPJ_UINT32 w = tilec->x1 - tilec->x0; + + h.mem = opj_aligned_malloc(dwt_max_resolution(tr, numres) * sizeof(OPJ_INT32)); + if + (! h.mem) + { + return false; + } - h.mem = opj_aligned_malloc(dwt_decode_max_resolution(tr, numres) * sizeof(int)); v.mem = h.mem; while( --numres) { - int * restrict tiledp = tilec->data; - int j; + OPJ_INT32 * restrict tiledp = tilec->data; + OPJ_UINT32 j; ++tr; h.sn = rw; @@ -547,14 +589,14 @@ static void dwt_decode_tile(opj_tcd_tilecomp_t* tilec, int numres, DWT1DFN dwt_1 for(j = 0; j < rh; ++j) { dwt_interleave_h(&h, &tiledp[j*w]); (dwt_1D)(&h); - memcpy(&tiledp[j*w], h.mem, rw * sizeof(int)); + memcpy(&tiledp[j*w], h.mem, rw * sizeof(OPJ_INT32)); } v.dn = rh - v.sn; v.cas = tr->y0 % 2; for(j = 0; j < rw; ++j){ - int k; + OPJ_UINT32 k; dwt_interleave_v(&v, &tiledp[j], w); (dwt_1D)(&v); for(k = 0; k < rh; ++k) { @@ -563,61 +605,62 @@ static void dwt_decode_tile(opj_tcd_tilecomp_t* tilec, int numres, DWT1DFN dwt_1 } } opj_aligned_free(h.mem); + return true; } -static void v4dwt_interleave_h(v4dwt_t* restrict w, float* restrict a, int x, int size){ - float* restrict bi = (float*) (w->wavelet + w->cas); - int count = w->sn; - int i, k; +static void v4dwt_interleave_h(v4dwt_t* restrict w, OPJ_FLOAT32* restrict a, OPJ_INT32 x, OPJ_INT32 size){ + OPJ_FLOAT32* restrict bi = (OPJ_FLOAT32*) (w->wavelet + w->cas); + OPJ_INT32 count = w->sn; + OPJ_INT32 i, k; for(k = 0; k < 2; ++k){ for(i = 0; i < count; ++i){ - int j = i; + OPJ_INT32 j = i; bi[i*8 ] = a[j]; j += x; - if(j > size) continue; + if(j >= size) continue; bi[i*8 + 1] = a[j]; j += x; - if(j > size) continue; + if(j >= size) continue; bi[i*8 + 2] = a[j]; j += x; - if(j > size) continue; + if(j >= size) continue; bi[i*8 + 3] = a[j]; } - bi = (float*) (w->wavelet + 1 - w->cas); + bi = (OPJ_FLOAT32*) (w->wavelet + 1 - w->cas); a += w->sn; size -= w->sn; count = w->dn; } } -static void v4dwt_interleave_v(v4dwt_t* restrict v , float* restrict a , int x){ +static void v4dwt_interleave_v(v4dwt_t* restrict v , OPJ_FLOAT32* restrict a , OPJ_INT32 x){ v4* restrict bi = v->wavelet + v->cas; - int i; + OPJ_INT32 i; for(i = 0; i < v->sn; ++i){ - memcpy(&bi[i*2], &a[i*x], 4 * sizeof(float)); + memcpy(&bi[i*2], &a[i*x], 4 * sizeof(OPJ_FLOAT32)); } a += v->sn * x; bi = v->wavelet + 1 - v->cas; for(i = 0; i < v->dn; ++i){ - memcpy(&bi[i*2], &a[i*x], 4 * sizeof(float)); + memcpy(&bi[i*2], &a[i*x], 4 * sizeof(OPJ_FLOAT32)); } } #ifdef __SSE__ -static void v4dwt_decode_step1_sse(v4* w, int count, const __m128 c){ +static void v4dwt_decode_step1_sse(v4* w, OPJ_INT32 count, const __m128 c){ __m128* restrict vw = (__m128*) w; - int i; + OPJ_INT32 i; for(i = 0; i < count; ++i){ __m128 tmp = vw[i*2]; vw[i*2] = tmp * c; } } -static void v4dwt_decode_step2_sse(v4* l, v4* w, int k, int m, __m128 c){ +static void v4dwt_decode_step2_sse(v4* l, v4* w, OPJ_INT32 k, OPJ_INT32 m, __m128 c){ __m128* restrict vl = (__m128*) l; __m128* restrict vw = (__m128*) w; - int i; + OPJ_INT32 i; for(i = 0; i < m; ++i){ __m128 tmp1 = vl[ 0]; __m128 tmp2 = vw[-1]; @@ -640,14 +683,14 @@ static void v4dwt_decode_step2_sse(v4* l, v4* w, int k, int m, __m128 c){ #else -static void v4dwt_decode_step1(v4* w, int count, const float c){ - float* restrict fw = (float*) w; - int i; +static void v4dwt_decode_step1(v4* w, OPJ_INT32 count, const OPJ_FLOAT32 c){ + OPJ_FLOAT32* restrict fw = (OPJ_FLOAT32*) w; + OPJ_INT32 i; for(i = 0; i < count; ++i){ - float tmp1 = fw[i*8 ]; - float tmp2 = fw[i*8 + 1]; - float tmp3 = fw[i*8 + 2]; - float tmp4 = fw[i*8 + 3]; + OPJ_FLOAT32 tmp1 = fw[i*8 ]; + OPJ_FLOAT32 tmp2 = fw[i*8 + 1]; + OPJ_FLOAT32 tmp3 = fw[i*8 + 2]; + OPJ_FLOAT32 tmp4 = fw[i*8 + 3]; fw[i*8 ] = tmp1 * c; fw[i*8 + 1] = tmp2 * c; fw[i*8 + 2] = tmp3 * c; @@ -655,23 +698,23 @@ static void v4dwt_decode_step1(v4* w, int count, const float c){ } } -static void v4dwt_decode_step2(v4* l, v4* w, int k, int m, float c){ - float* restrict fl = (float*) l; - float* restrict fw = (float*) w; - int i; +static void v4dwt_decode_step2(v4* l, v4* w, OPJ_INT32 k, OPJ_INT32 m, OPJ_FLOAT32 c){ + OPJ_FLOAT32* restrict fl = (OPJ_FLOAT32*) l; + OPJ_FLOAT32* restrict fw = (OPJ_FLOAT32*) w; + OPJ_INT32 i; for(i = 0; i < m; ++i){ - float tmp1_1 = fl[0]; - float tmp1_2 = fl[1]; - float tmp1_3 = fl[2]; - float tmp1_4 = fl[3]; - float tmp2_1 = fw[-4]; - float tmp2_2 = fw[-3]; - float tmp2_3 = fw[-2]; - float tmp2_4 = fw[-1]; - float tmp3_1 = fw[0]; - float tmp3_2 = fw[1]; - float tmp3_3 = fw[2]; - float tmp3_4 = fw[3]; + OPJ_FLOAT32 tmp1_1 = fl[0]; + OPJ_FLOAT32 tmp1_2 = fl[1]; + OPJ_FLOAT32 tmp1_3 = fl[2]; + OPJ_FLOAT32 tmp1_4 = fl[3]; + OPJ_FLOAT32 tmp2_1 = fw[-4]; + OPJ_FLOAT32 tmp2_2 = fw[-3]; + OPJ_FLOAT32 tmp2_3 = fw[-2]; + OPJ_FLOAT32 tmp2_4 = fw[-1]; + OPJ_FLOAT32 tmp3_1 = fw[0]; + OPJ_FLOAT32 tmp3_2 = fw[1]; + OPJ_FLOAT32 tmp3_3 = fw[2]; + OPJ_FLOAT32 tmp3_4 = fw[3]; fw[-4] = tmp2_1 + ((tmp1_1 + tmp3_1) * c); fw[-3] = tmp2_2 + ((tmp1_2 + tmp3_2) * c); fw[-2] = tmp2_3 + ((tmp1_3 + tmp3_3) * c); @@ -680,20 +723,20 @@ static void v4dwt_decode_step2(v4* l, v4* w, int k, int m, float c){ fw += 8; } if(m < k){ - float c1; - float c2; - float c3; - float c4; + OPJ_FLOAT32 c1; + OPJ_FLOAT32 c2; + OPJ_FLOAT32 c3; + OPJ_FLOAT32 c4; c += c; c1 = fl[0] * c; c2 = fl[1] * c; c3 = fl[2] * c; c4 = fl[3] * c; for(; m < k; ++m){ - float tmp1 = fw[-4]; - float tmp2 = fw[-3]; - float tmp3 = fw[-2]; - float tmp4 = fw[-1]; + OPJ_FLOAT32 tmp1 = fw[-4]; + OPJ_FLOAT32 tmp2 = fw[-3]; + OPJ_FLOAT32 tmp3 = fw[-2]; + OPJ_FLOAT32 tmp4 = fw[-1]; fw[-4] = tmp1 + c1; fw[-3] = tmp2 + c2; fw[-2] = tmp3 + c3; @@ -709,7 +752,7 @@ static void v4dwt_decode_step2(v4* l, v4* w, int k, int m, float c){ /* Inverse 9-7 wavelet transform in 1-D. */ /* </summary> */ static void v4dwt_decode(v4dwt_t* restrict dwt){ - int a, b; + OPJ_INT32 a, b; if(dwt->cas == 0) { if(!((dwt->dn > 0) || (dwt->sn > 1))){ return; @@ -726,14 +769,14 @@ static void v4dwt_decode(v4dwt_t* restrict dwt){ #ifdef __SSE__ v4dwt_decode_step1_sse(dwt->wavelet+a, dwt->sn, _mm_set1_ps(K)); v4dwt_decode_step1_sse(dwt->wavelet+b, dwt->dn, _mm_set1_ps(c13318)); - v4dwt_decode_step2_sse(dwt->wavelet+b, dwt->wavelet+a+1, dwt->sn, int_min(dwt->sn, dwt->dn-a), _mm_set1_ps(dwt_delta)); + v4dwt_decode_step2_sse(dwt->wavelet+b, dwt->wavelet+a+1, dwt->sn, int_min(dwt->sn, dwt->dn-a), _mm_set1_ps(delta)); v4dwt_decode_step2_sse(dwt->wavelet+a, dwt->wavelet+b+1, dwt->dn, int_min(dwt->dn, dwt->sn-b), _mm_set1_ps(dwt_gamma)); v4dwt_decode_step2_sse(dwt->wavelet+b, dwt->wavelet+a+1, dwt->sn, int_min(dwt->sn, dwt->dn-a), _mm_set1_ps(dwt_beta)); v4dwt_decode_step2_sse(dwt->wavelet+a, dwt->wavelet+b+1, dwt->dn, int_min(dwt->dn, dwt->sn-b), _mm_set1_ps(dwt_alpha)); #else v4dwt_decode_step1(dwt->wavelet+a, dwt->sn, K); v4dwt_decode_step1(dwt->wavelet+b, dwt->dn, c13318); - v4dwt_decode_step2(dwt->wavelet+b, dwt->wavelet+a+1, dwt->sn, int_min(dwt->sn, dwt->dn-a), dwt_delta); + v4dwt_decode_step2(dwt->wavelet+b, dwt->wavelet+a+1, dwt->sn, int_min(dwt->sn, dwt->dn-a), delta); v4dwt_decode_step2(dwt->wavelet+a, dwt->wavelet+b+1, dwt->dn, int_min(dwt->dn, dwt->sn-b), dwt_gamma); v4dwt_decode_step2(dwt->wavelet+b, dwt->wavelet+a+1, dwt->sn, int_min(dwt->sn, dwt->dn-a), dwt_beta); v4dwt_decode_step2(dwt->wavelet+a, dwt->wavelet+b+1, dwt->dn, int_min(dwt->dn, dwt->sn-b), dwt_alpha); @@ -743,24 +786,24 @@ static void v4dwt_decode(v4dwt_t* restrict dwt){ /* <summary> */ /* Inverse 9-7 wavelet transform in 2-D. */ /* </summary> */ -void dwt_decode_real(opj_tcd_tilecomp_t* restrict tilec, int numres){ +bool dwt_decode_real(opj_tcd_tilecomp_t* restrict tilec, OPJ_UINT32 numres){ v4dwt_t h; v4dwt_t v; opj_tcd_resolution_t* res = tilec->resolutions; - int rw = res->x1 - res->x0; /* width of the resolution level computed */ - int rh = res->y1 - res->y0; /* height of the resolution level computed */ + OPJ_UINT32 rw = res->x1 - res->x0; /* width of the resolution level computed */ + OPJ_UINT32 rh = res->y1 - res->y0; /* height of the resolution level computed */ - int w = tilec->x1 - tilec->x0; + OPJ_UINT32 w = tilec->x1 - tilec->x0; - h.wavelet = (v4*) opj_aligned_malloc((dwt_decode_max_resolution(res, numres)+5) * sizeof(v4)); + h.wavelet = (v4*) opj_aligned_malloc((dwt_max_resolution(res, numres)+5) * sizeof(v4)); v.wavelet = h.wavelet; while( --numres) { - float * restrict aj = (float*) tilec->data; - int bufsize = (tilec->x1 - tilec->x0) * (tilec->y1 - tilec->y0); - int j; + OPJ_FLOAT32 * restrict aj = (OPJ_FLOAT32*) tilec->data; + OPJ_UINT32 bufsize = (tilec->x1 - tilec->x0) * (tilec->y1 - tilec->y0); + OPJ_INT32 j; h.sn = rw; v.sn = rh; @@ -771,22 +814,26 @@ void dwt_decode_real(opj_tcd_tilecomp_t* restrict tilec, int numres){ rh = res->y1 - res->y0; /* height of the resolution level computed */ h.dn = rw - h.sn; - h.cas = res->x0 % 2; + h.cas = res->x0 & 1; for(j = rh; j > 0; j -= 4){ v4dwt_interleave_h(&h, aj, w, bufsize); v4dwt_decode(&h); if(j >= 4){ - int k; - for(k = rw; --k >= 0;){ + OPJ_INT32 k = rw; + while + (--k >= 0) + { aj[k ] = h.wavelet[k].f[0]; aj[k+w ] = h.wavelet[k].f[1]; aj[k+w*2] = h.wavelet[k].f[2]; aj[k+w*3] = h.wavelet[k].f[3]; } }else{ - int k; - for(k = rw; --k >= 0;){ + OPJ_INT32 k = rw; + while + (--k >= 0) + { switch(j) { case 3: aj[k+w*2] = h.wavelet[k].f[2]; case 2: aj[k+w ] = h.wavelet[k].f[1]; @@ -801,19 +848,19 @@ void dwt_decode_real(opj_tcd_tilecomp_t* restrict tilec, int numres){ v.dn = rh - v.sn; v.cas = res->y0 % 2; - aj = (float*) tilec->data; + aj = (OPJ_FLOAT32*) tilec->data; for(j = rw; j > 0; j -= 4){ v4dwt_interleave_v(&v, aj, w); v4dwt_decode(&v); if(j >= 4){ - int k; + OPJ_UINT32 k; for(k = 0; k < rh; ++k){ - memcpy(&aj[k*w], &v.wavelet[k], 4 * sizeof(float)); + memcpy(&aj[k*w], &v.wavelet[k], 4 * sizeof(OPJ_FLOAT32)); } }else{ - int k; + OPJ_UINT32 k; for(k = 0; k < rh; ++k){ - memcpy(&aj[k*w], &v.wavelet[k], j * sizeof(float)); + memcpy(&aj[k*w], &v.wavelet[k], j * sizeof(OPJ_FLOAT32)); } } aj += 4; @@ -821,5 +868,6 @@ void dwt_decode_real(opj_tcd_tilecomp_t* restrict tilec, int numres){ } opj_aligned_free(h.wavelet); + return true; } |
