/*
- * The copyright in this software is being made available under the 2-clauses
- * BSD License, included below. This software may be subject to other third
+ * The copyright in this software is being made available under the 2-clauses
+ * BSD License, included below. This software may be subject to other third
* party and contributor rights, including patent rights, and no such rights
* are granted under this license.
*
* Copyright (c) 2002-2014, Professor Benoit Macq
* Copyright (c) 2001-2003, David Janssens
* Copyright (c) 2002-2003, Yannick Verschueren
- * Copyright (c) 2003-2007, Francois-Olivier Devaux
+ * Copyright (c) 2003-2007, Francois-Olivier Devaux
* Copyright (c) 2003-2014, Antonin Descampe
* Copyright (c) 2005, Herve Drolon, FreeImage Team
* Copyright (c) 2007, Jonathan Ballard <dzonatas@dzonux.net>
/*@{*/
typedef struct dwt_local {
- OPJ_INT32* mem;
- OPJ_INT32 dn;
- OPJ_INT32 sn;
- OPJ_INT32 cas;
+ OPJ_INT32* mem;
+ OPJ_INT32 dn;
+ OPJ_INT32 sn;
+ OPJ_INT32 cas;
} opj_dwt_t;
typedef union {
- OPJ_FLOAT32 f[4];
+ OPJ_FLOAT32 f[4];
} opj_v4_t;
typedef struct v4dwt_local {
- opj_v4_t* wavelet ;
- OPJ_INT32 dn ;
- OPJ_INT32 sn ;
- OPJ_INT32 cas ;
+ opj_v4_t* wavelet ;
+ OPJ_INT32 dn ;
+ OPJ_INT32 sn ;
+ OPJ_INT32 cas ;
} opj_v4dwt_t ;
static const OPJ_FLOAT32 opj_dwt_alpha = 1.586134342f; /* 12994 */
/*@}*/
/**
-Virtual function type for wavelet transform in 1-D
+Virtual function type for wavelet transform in 1-D
*/
typedef void (*DWT1DFN)(opj_dwt_t* v);
/**
Forward lazy transform (horizontal)
*/
-static void opj_dwt_deinterleave_h(OPJ_INT32 *a, OPJ_INT32 *b, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 cas);
+static void opj_dwt_deinterleave_h(OPJ_INT32 *a, OPJ_INT32 *b, OPJ_INT32 dn,
+ OPJ_INT32 sn, OPJ_INT32 cas);
/**
Forward lazy transform (vertical)
*/
-static void opj_dwt_deinterleave_v(OPJ_INT32 *a, OPJ_INT32 *b, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 x, OPJ_INT32 cas);
+static void opj_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)
*/
/**
Forward 5-3 wavelet transform in 1-D
*/
-static void opj_dwt_encode_1(OPJ_INT32 *a, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 cas);
+static void opj_dwt_encode_1(OPJ_INT32 *a, OPJ_INT32 dn, OPJ_INT32 sn,
+ OPJ_INT32 cas);
/**
Inverse 5-3 wavelet transform in 1-D
*/
static void opj_dwt_decode_1(opj_dwt_t *v);
-static void opj_dwt_decode_1_(OPJ_INT32 *a, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 cas);
+static void opj_dwt_decode_1_(OPJ_INT32 *a, OPJ_INT32 dn, OPJ_INT32 sn,
+ OPJ_INT32 cas);
/**
Forward 9-7 wavelet transform in 1-D
*/
-static void opj_dwt_encode_1_real(OPJ_INT32 *a, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 cas);
+static void opj_dwt_encode_1_real(OPJ_INT32 *a, OPJ_INT32 dn, OPJ_INT32 sn,
+ OPJ_INT32 cas);
/**
-Explicit calculation of the Quantization Stepsizes
+Explicit calculation of the Quantization Stepsizes
*/
-static void opj_dwt_encode_stepsize(OPJ_INT32 stepsize, OPJ_INT32 numbps, opj_stepsize_t *bandno_stepsize);
+static void opj_dwt_encode_stepsize(OPJ_INT32 stepsize, OPJ_INT32 numbps,
+ opj_stepsize_t *bandno_stepsize);
/**
Inverse wavelet transform in 2-D.
*/
-static OPJ_BOOL opj_dwt_decode_tile(opj_thread_pool_t* tp, opj_tcd_tilecomp_t* tilec, OPJ_UINT32 i, DWT1DFN fn);
+static OPJ_BOOL opj_dwt_decode_tile(opj_thread_pool_t* tp,
+ opj_tcd_tilecomp_t* tilec, OPJ_UINT32 i, DWT1DFN fn);
-static OPJ_BOOL opj_dwt_encode_procedure( opj_tcd_tilecomp_t * tilec,
- void (*p_function)(OPJ_INT32 *, OPJ_INT32,OPJ_INT32,OPJ_INT32) );
+static OPJ_BOOL opj_dwt_encode_procedure(opj_tcd_tilecomp_t * tilec,
+ void (*p_function)(OPJ_INT32 *, OPJ_INT32, OPJ_INT32, OPJ_INT32));
-static OPJ_UINT32 opj_dwt_max_resolution(opj_tcd_resolution_t* OPJ_RESTRICT r, OPJ_UINT32 i);
+static OPJ_UINT32 opj_dwt_max_resolution(opj_tcd_resolution_t* OPJ_RESTRICT r,
+ OPJ_UINT32 i);
/* <summary> */
/* Inverse 9-7 wavelet transform in 1-D. */
/* </summary> */
static void opj_v4dwt_decode(opj_v4dwt_t* OPJ_RESTRICT dwt);
-static void opj_v4dwt_interleave_h(opj_v4dwt_t* OPJ_RESTRICT w, OPJ_FLOAT32* OPJ_RESTRICT a, OPJ_INT32 x, OPJ_INT32 size);
+static void opj_v4dwt_interleave_h(opj_v4dwt_t* OPJ_RESTRICT w,
+ OPJ_FLOAT32* OPJ_RESTRICT a, OPJ_INT32 x, OPJ_INT32 size);
-static void opj_v4dwt_interleave_v(opj_v4dwt_t* OPJ_RESTRICT v , OPJ_FLOAT32* OPJ_RESTRICT a , OPJ_INT32 x, OPJ_INT32 nb_elts_read);
+static void opj_v4dwt_interleave_v(opj_v4dwt_t* OPJ_RESTRICT v,
+ OPJ_FLOAT32* OPJ_RESTRICT a, OPJ_INT32 x, OPJ_INT32 nb_elts_read);
#ifdef __SSE__
-static void opj_v4dwt_decode_step1_sse(opj_v4_t* w, OPJ_INT32 count, const __m128 c);
+static void opj_v4dwt_decode_step1_sse(opj_v4_t* w, OPJ_INT32 count,
+ const __m128 c);
-static void opj_v4dwt_decode_step2_sse(opj_v4_t* l, opj_v4_t* w, OPJ_INT32 k, OPJ_INT32 m, __m128 c);
+static void opj_v4dwt_decode_step2_sse(opj_v4_t* l, opj_v4_t* w, OPJ_INT32 k,
+ OPJ_INT32 m, __m128 c);
#else
-static void opj_v4dwt_decode_step1(opj_v4_t* w, OPJ_INT32 count, const OPJ_FLOAT32 c);
+static void opj_v4dwt_decode_step1(opj_v4_t* w, OPJ_INT32 count,
+ const OPJ_FLOAT32 c);
-static void opj_v4dwt_decode_step2(opj_v4_t* l, opj_v4_t* w, OPJ_INT32 k, OPJ_INT32 m, OPJ_FLOAT32 c);
+static void opj_v4dwt_decode_step2(opj_v4_t* l, opj_v4_t* w, OPJ_INT32 k,
+ OPJ_INT32 m, OPJ_FLOAT32 c);
#endif
/* This table contains the norms of the 5-3 wavelets for different bands. */
/* </summary> */
static const OPJ_FLOAT64 opj_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},
- {.7186, .9218, 1.586, 3.043, 6.019, 12.01, 24.00, 47.97, 95.93}
+ {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},
+ {.7186, .9218, 1.586, 3.043, 6.019, 12.01, 24.00, 47.97, 95.93}
};
/* <summary> */
/* This table contains the norms of the 9-7 wavelets for different bands. */
/* </summary> */
static const OPJ_FLOAT64 opj_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},
- {2.080, 3.865, 8.307, 17.18, 34.71, 69.59, 139.3, 278.6, 557.2}
+ {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},
+ {2.080, 3.865, 8.307, 17.18, 34.71, 69.59, 139.3, 278.6, 557.2}
};
-/*
+/*
==========================================================
local functions
==========================================================
*/
-/* <summary> */
+/* <summary> */
/* Forward lazy transform (horizontal). */
-/* </summary> */
-static void opj_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;
- }
-
+/* </summary> */
+static void opj_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;
+ l_src = a + 1 - cas;
- for (i=0; i<dn; ++i) {
- *l_dest++=*l_src;
- l_src += 2;
- }
+ for (i = 0; i < dn; ++i) {
+ *l_dest++ = *l_src;
+ l_src += 2;
+ }
}
-/* <summary> */
+/* <summary> */
/* Forward lazy transform (vertical). */
-/* </summary> */
-static void opj_dwt_deinterleave_v(OPJ_INT32 *a, OPJ_INT32 *b, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 x, OPJ_INT32 cas) {
+/* </summary> */
+static void opj_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;
+ 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;
+ *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)];*/
+ *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 opj_dwt_interleave_h(opj_dwt_t* h, OPJ_INT32 *a) {
+static void opj_dwt_interleave_h(opj_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;
+ 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;
+ ai = a + h->sn;
+ bi = h->mem + 1 - h->cas;
+ i = h->dn ;
+ while (i--) {
+ *bi = *(ai++);
+ bi += 2;
}
}
-/* <summary> */
+/* <summary> */
/* Inverse lazy transform (vertical). */
-/* </summary> */
-static void opj_dwt_interleave_v(opj_dwt_t* v, OPJ_INT32 *a, OPJ_INT32 x) {
+/* </summary> */
+static void opj_dwt_interleave_v(opj_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;
- ai += x;
+ while (i--) {
+ *bi = *ai;
+ bi += 2;
+ ai += x;
}
ai = a + (v->sn * x);
bi = v->mem + 1 - v->cas;
i = v->dn ;
- while( i-- ) {
- *bi = *ai;
- bi += 2;
- ai += x;
+ while (i--) {
+ *bi = *ai;
+ bi += 2;
+ ai += x;
}
}
/* <summary> */
/* Forward 5-3 wavelet transform in 1-D. */
/* </summary> */
-static void opj_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 */
- for (i = 0; i < dn; i++) OPJ_D(i) -= (OPJ_S_(i) + OPJ_S_(i + 1)) >> 1;
- for (i = 0; i < sn; i++) OPJ_S(i) += (OPJ_D_(i - 1) + OPJ_D_(i) + 2) >> 2;
- }
- } else {
- if (!sn && dn == 1) /* NEW : CASE ONE ELEMENT */
- OPJ_S(0) *= 2;
- else {
- for (i = 0; i < dn; i++) OPJ_S(i) -= (OPJ_DD_(i) + OPJ_DD_(i - 1)) >> 1;
- for (i = 0; i < sn; i++) OPJ_D(i) += (OPJ_SS_(i) + OPJ_SS_(i + 1) + 2) >> 2;
- }
- }
+static void opj_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 */
+ for (i = 0; i < dn; i++) {
+ OPJ_D(i) -= (OPJ_S_(i) + OPJ_S_(i + 1)) >> 1;
+ }
+ for (i = 0; i < sn; i++) {
+ OPJ_S(i) += (OPJ_D_(i - 1) + OPJ_D_(i) + 2) >> 2;
+ }
+ }
+ } else {
+ if (!sn && dn == 1) { /* NEW : CASE ONE ELEMENT */
+ OPJ_S(0) *= 2;
+ } else {
+ for (i = 0; i < dn; i++) {
+ OPJ_S(i) -= (OPJ_DD_(i) + OPJ_DD_(i - 1)) >> 1;
+ }
+ for (i = 0; i < sn; i++) {
+ OPJ_D(i) += (OPJ_SS_(i) + OPJ_SS_(i + 1) + 2) >> 2;
+ }
+ }
+ }
}
/* <summary> */
/* Inverse 5-3 wavelet transform in 1-D. */
-/* </summary> */
-static void opj_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 */
- for (i = 0; i < sn; i++) OPJ_S(i) -= (OPJ_D_(i - 1) + OPJ_D_(i) + 2) >> 2;
- for (i = 0; i < dn; i++) OPJ_D(i) += (OPJ_S_(i) + OPJ_S_(i + 1)) >> 1;
- }
- } else {
- if (!sn && dn == 1) /* NEW : CASE ONE ELEMENT */
- OPJ_S(0) /= 2;
- else {
- for (i = 0; i < sn; i++) OPJ_D(i) -= (OPJ_SS_(i) + OPJ_SS_(i + 1) + 2) >> 2;
- for (i = 0; i < dn; i++) OPJ_S(i) += (OPJ_DD_(i) + OPJ_DD_(i - 1)) >> 1;
- }
- }
+/* </summary> */
+static void opj_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 */
+ for (i = 0; i < sn; i++) {
+ OPJ_S(i) -= (OPJ_D_(i - 1) + OPJ_D_(i) + 2) >> 2;
+ }
+ for (i = 0; i < dn; i++) {
+ OPJ_D(i) += (OPJ_S_(i) + OPJ_S_(i + 1)) >> 1;
+ }
+ }
+ } else {
+ if (!sn && dn == 1) { /* NEW : CASE ONE ELEMENT */
+ OPJ_S(0) /= 2;
+ } else {
+ for (i = 0; i < sn; i++) {
+ OPJ_D(i) -= (OPJ_SS_(i) + OPJ_SS_(i + 1) + 2) >> 2;
+ }
+ for (i = 0; i < dn; i++) {
+ OPJ_S(i) += (OPJ_DD_(i) + OPJ_DD_(i - 1)) >> 1;
+ }
+ }
+ }
}
/* <summary> */
/* Inverse 5-3 wavelet transform in 1-D. */
-/* </summary> */
-static void opj_dwt_decode_1(opj_dwt_t *v) {
- opj_dwt_decode_1_(v->mem, v->dn, v->sn, v->cas);
+/* </summary> */
+static void opj_dwt_decode_1(opj_dwt_t *v)
+{
+ opj_dwt_decode_1_(v->mem, v->dn, v->sn, v->cas);
}
/* <summary> */
/* Forward 9-7 wavelet transform in 1-D. */
/* </summary> */
-static void opj_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++)
- OPJ_D(i) -= opj_int_fix_mul(OPJ_S_(i) + OPJ_S_(i + 1), 12993);
- for (i = 0; i < sn; i++)
- OPJ_S(i) -= opj_int_fix_mul(OPJ_D_(i - 1) + OPJ_D_(i), 434);
- for (i = 0; i < dn; i++)
- OPJ_D(i) += opj_int_fix_mul(OPJ_S_(i) + OPJ_S_(i + 1), 7233);
- for (i = 0; i < sn; i++)
- OPJ_S(i) += opj_int_fix_mul(OPJ_D_(i - 1) + OPJ_D_(i), 3633);
- for (i = 0; i < dn; i++)
- OPJ_D(i) = opj_int_fix_mul(OPJ_D(i), 5038); /*5038 */
- for (i = 0; i < sn; i++)
- OPJ_S(i) = opj_int_fix_mul(OPJ_S(i), 6659); /*6660 */
- }
- } else {
- if ((sn > 0) || (dn > 1)) { /* NEW : CASE ONE ELEMENT */
- for (i = 0; i < dn; i++)
- OPJ_S(i) -= opj_int_fix_mul(OPJ_DD_(i) + OPJ_DD_(i - 1), 12993);
- for (i = 0; i < sn; i++)
- OPJ_D(i) -= opj_int_fix_mul(OPJ_SS_(i) + OPJ_SS_(i + 1), 434);
- for (i = 0; i < dn; i++)
- OPJ_S(i) += opj_int_fix_mul(OPJ_DD_(i) + OPJ_DD_(i - 1), 7233);
- for (i = 0; i < sn; i++)
- OPJ_D(i) += opj_int_fix_mul(OPJ_SS_(i) + OPJ_SS_(i + 1), 3633);
- for (i = 0; i < dn; i++)
- OPJ_S(i) = opj_int_fix_mul(OPJ_S(i), 5038); /*5038 */
- for (i = 0; i < sn; i++)
- OPJ_D(i) = opj_int_fix_mul(OPJ_D(i), 6659); /*6660 */
- }
- }
+static void opj_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++) {
+ OPJ_D(i) -= opj_int_fix_mul(OPJ_S_(i) + OPJ_S_(i + 1), 12993);
+ }
+ for (i = 0; i < sn; i++) {
+ OPJ_S(i) -= opj_int_fix_mul(OPJ_D_(i - 1) + OPJ_D_(i), 434);
+ }
+ for (i = 0; i < dn; i++) {
+ OPJ_D(i) += opj_int_fix_mul(OPJ_S_(i) + OPJ_S_(i + 1), 7233);
+ }
+ for (i = 0; i < sn; i++) {
+ OPJ_S(i) += opj_int_fix_mul(OPJ_D_(i - 1) + OPJ_D_(i), 3633);
+ }
+ for (i = 0; i < dn; i++) {
+ OPJ_D(i) = opj_int_fix_mul(OPJ_D(i), 5038); /*5038 */
+ }
+ for (i = 0; i < sn; i++) {
+ OPJ_S(i) = opj_int_fix_mul(OPJ_S(i), 6659); /*6660 */
+ }
+ }
+ } else {
+ if ((sn > 0) || (dn > 1)) { /* NEW : CASE ONE ELEMENT */
+ for (i = 0; i < dn; i++) {
+ OPJ_S(i) -= opj_int_fix_mul(OPJ_DD_(i) + OPJ_DD_(i - 1), 12993);
+ }
+ for (i = 0; i < sn; i++) {
+ OPJ_D(i) -= opj_int_fix_mul(OPJ_SS_(i) + OPJ_SS_(i + 1), 434);
+ }
+ for (i = 0; i < dn; i++) {
+ OPJ_S(i) += opj_int_fix_mul(OPJ_DD_(i) + OPJ_DD_(i - 1), 7233);
+ }
+ for (i = 0; i < sn; i++) {
+ OPJ_D(i) += opj_int_fix_mul(OPJ_SS_(i) + OPJ_SS_(i + 1), 3633);
+ }
+ for (i = 0; i < dn; i++) {
+ OPJ_S(i) = opj_int_fix_mul(OPJ_S(i), 5038); /*5038 */
+ }
+ for (i = 0; i < sn; i++) {
+ OPJ_D(i) = opj_int_fix_mul(OPJ_D(i), 6659); /*6660 */
+ }
+ }
+ }
}
-static void opj_dwt_encode_stepsize(OPJ_INT32 stepsize, OPJ_INT32 numbps, opj_stepsize_t *bandno_stepsize) {
- OPJ_INT32 p, n;
- p = opj_int_floorlog2(stepsize) - 13;
- n = 11 - opj_int_floorlog2(stepsize);
- bandno_stepsize->mant = (n < 0 ? stepsize >> -n : stepsize << n) & 0x7ff;
- bandno_stepsize->expn = numbps - p;
+static void opj_dwt_encode_stepsize(OPJ_INT32 stepsize, OPJ_INT32 numbps,
+ opj_stepsize_t *bandno_stepsize)
+{
+ OPJ_INT32 p, n;
+ p = opj_int_floorlog2(stepsize) - 13;
+ n = 11 - opj_int_floorlog2(stepsize);
+ bandno_stepsize->mant = (n < 0 ? stepsize >> -n : stepsize << n) & 0x7ff;
+ bandno_stepsize->expn = numbps - p;
}
-/*
+/*
==========================================================
DWT interface
==========================================================
/* <summary> */
/* Forward 5-3 wavelet transform in 2-D. */
/* </summary> */
-static INLINE OPJ_BOOL opj_dwt_encode_procedure(opj_tcd_tilecomp_t * tilec,void (*p_function)(OPJ_INT32 *, OPJ_INT32,OPJ_INT32,OPJ_INT32) )
+static INLINE OPJ_BOOL opj_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 */
- size_t 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 = (OPJ_INT32)tilec->numresolutions-1;
- a = tilec->data;
-
- l_cur_res = tilec->resolutions + l;
- l_last_res = l_cur_res - 1;
-
- l_data_size = opj_dwt_max_resolution( tilec->resolutions,tilec->numresolutions);
- /* overflow check */
- if (l_data_size > (SIZE_MAX / sizeof(OPJ_INT32))) {
- /* FIXME event manager error callback */
- return OPJ_FALSE;
- }
- l_data_size *= sizeof(OPJ_INT32);
- bj = (OPJ_INT32*)opj_malloc(l_data_size);
- /* l_data_size is equal to 0 when numresolutions == 1 but bj is not used */
- /* in that case, so do not error out */
- if (l_data_size != 0 && ! bj) {
- return OPJ_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 = 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 = l_cur_res->x0 & 1;
- cas_col = l_cur_res->y0 & 1;
-
- sn = rh1;
- dn = rh - rh1;
- for (j = 0; j < rw; ++j) {
- aj = a + j;
- for (k = 0; k < rh; ++k) {
- bj[k] = aj[k*w];
- }
-
- (*p_function) (bj, dn, sn, cas_col);
-
- opj_dwt_deinterleave_v(bj, aj, dn, sn, w, cas_col);
- }
-
- sn = rw1;
- dn = rw - rw1;
-
- for (j = 0; j < rh; j++) {
- aj = a + j * w;
- for (k = 0; k < rw; k++) bj[k] = aj[k];
- (*p_function) (bj, dn, sn, cas_row);
- opj_dwt_deinterleave_h(bj, aj, dn, sn, cas_row);
- }
-
- l_cur_res = l_last_res;
-
- --l_last_res;
- }
-
- opj_free(bj);
- return OPJ_TRUE;
+ 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 */
+ size_t 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 = (OPJ_INT32)tilec->numresolutions - 1;
+ a = tilec->data;
+
+ l_cur_res = tilec->resolutions + l;
+ l_last_res = l_cur_res - 1;
+
+ l_data_size = opj_dwt_max_resolution(tilec->resolutions, tilec->numresolutions);
+ /* overflow check */
+ if (l_data_size > (SIZE_MAX / sizeof(OPJ_INT32))) {
+ /* FIXME event manager error callback */
+ return OPJ_FALSE;
+ }
+ l_data_size *= sizeof(OPJ_INT32);
+ bj = (OPJ_INT32*)opj_malloc(l_data_size);
+ /* l_data_size is equal to 0 when numresolutions == 1 but bj is not used */
+ /* in that case, so do not error out */
+ if (l_data_size != 0 && ! bj) {
+ return OPJ_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 = 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 = l_cur_res->x0 & 1;
+ cas_col = l_cur_res->y0 & 1;
+
+ sn = rh1;
+ dn = rh - rh1;
+ for (j = 0; j < rw; ++j) {
+ aj = a + j;
+ for (k = 0; k < rh; ++k) {
+ bj[k] = aj[k * w];
+ }
+
+ (*p_function)(bj, dn, sn, cas_col);
+
+ opj_dwt_deinterleave_v(bj, aj, dn, sn, w, cas_col);
+ }
+
+ sn = rw1;
+ dn = rw - rw1;
+
+ for (j = 0; j < rh; j++) {
+ aj = a + j * w;
+ for (k = 0; k < rw; k++) {
+ bj[k] = aj[k];
+ }
+ (*p_function)(bj, dn, sn, cas_row);
+ opj_dwt_deinterleave_h(bj, aj, dn, sn, cas_row);
+ }
+
+ l_cur_res = l_last_res;
+
+ --l_last_res;
+ }
+
+ opj_free(bj);
+ return OPJ_TRUE;
}
/* Forward 5-3 wavelet transform in 2-D. */
/* </summary> */
OPJ_BOOL opj_dwt_encode(opj_tcd_tilecomp_t * tilec)
{
- return opj_dwt_encode_procedure(tilec,opj_dwt_encode_1);
+ return opj_dwt_encode_procedure(tilec, opj_dwt_encode_1);
}
/* <summary> */
/* Inverse 5-3 wavelet transform in 2-D. */
/* </summary> */
-OPJ_BOOL opj_dwt_decode(opj_thread_pool_t* tp, opj_tcd_tilecomp_t* tilec, OPJ_UINT32 numres) {
- return opj_dwt_decode_tile(tp, tilec, numres, &opj_dwt_decode_1);
+OPJ_BOOL opj_dwt_decode(opj_thread_pool_t* tp, opj_tcd_tilecomp_t* tilec,
+ OPJ_UINT32 numres)
+{
+ return opj_dwt_decode_tile(tp, tilec, numres, &opj_dwt_decode_1);
}
/* <summary> */
/* Get gain of 5-3 wavelet transform. */
/* </summary> */
-OPJ_UINT32 opj_dwt_getgain(OPJ_UINT32 orient) {
- if (orient == 0)
- return 0;
- if (orient == 1 || orient == 2)
- return 1;
- return 2;
+OPJ_UINT32 opj_dwt_getgain(OPJ_UINT32 orient)
+{
+ if (orient == 0) {
+ return 0;
+ }
+ if (orient == 1 || orient == 2) {
+ return 1;
+ }
+ return 2;
}
/* <summary> */
/* Get norm of 5-3 wavelet. */
/* </summary> */
-OPJ_FLOAT64 opj_dwt_getnorm(OPJ_UINT32 level, OPJ_UINT32 orient) {
- return opj_dwt_norms[orient][level];
+OPJ_FLOAT64 opj_dwt_getnorm(OPJ_UINT32 level, OPJ_UINT32 orient)
+{
+ return opj_dwt_norms[orient][level];
}
/* <summary> */
/* </summary> */
OPJ_BOOL opj_dwt_encode_real(opj_tcd_tilecomp_t * tilec)
{
- return opj_dwt_encode_procedure(tilec,opj_dwt_encode_1_real);
+ return opj_dwt_encode_procedure(tilec, opj_dwt_encode_1_real);
}
/* <summary> */
/* Get gain of 9-7 wavelet transform. */
/* </summary> */
-OPJ_UINT32 opj_dwt_getgain_real(OPJ_UINT32 orient) {
- (void)orient;
- return 0;
+OPJ_UINT32 opj_dwt_getgain_real(OPJ_UINT32 orient)
+{
+ (void)orient;
+ return 0;
}
/* <summary> */
/* Get norm of 9-7 wavelet. */
/* </summary> */
-OPJ_FLOAT64 opj_dwt_getnorm_real(OPJ_UINT32 level, OPJ_UINT32 orient) {
- return opj_dwt_norms_real[orient][level];
+OPJ_FLOAT64 opj_dwt_getnorm_real(OPJ_UINT32 level, OPJ_UINT32 orient)
+{
+ return opj_dwt_norms_real[orient][level];
}
-void opj_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++) {
- 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);
- level = tccp->numresolutions - 1 - resno;
- gain = (tccp->qmfbid == 0) ? 0 : ((orient == 0) ? 0 : (((orient == 1) || (orient == 2)) ? 1 : 2));
- if (tccp->qntsty == J2K_CCP_QNTSTY_NOQNT) {
- stepsize = 1.0;
- } else {
- OPJ_FLOAT64 norm = opj_dwt_norms_real[orient][level];
- stepsize = (1 << (gain)) / norm;
- }
- opj_dwt_encode_stepsize((OPJ_INT32) floor(stepsize * 8192.0), (OPJ_INT32)(prec + gain), &tccp->stepsizes[bandno]);
- }
+void opj_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++) {
+ 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);
+ level = tccp->numresolutions - 1 - resno;
+ gain = (tccp->qmfbid == 0) ? 0 : ((orient == 0) ? 0 : (((orient == 1) ||
+ (orient == 2)) ? 1 : 2));
+ if (tccp->qntsty == J2K_CCP_QNTSTY_NOQNT) {
+ stepsize = 1.0;
+ } else {
+ OPJ_FLOAT64 norm = opj_dwt_norms_real[orient][level];
+ stepsize = (1 << (gain)) / norm;
+ }
+ opj_dwt_encode_stepsize((OPJ_INT32) floor(stepsize * 8192.0),
+ (OPJ_INT32)(prec + gain), &tccp->stepsizes[bandno]);
+ }
}
/* <summary> */
/* Determine maximum computed resolution level for inverse wavelet transform */
/* </summary> */
-static OPJ_UINT32 opj_dwt_max_resolution(opj_tcd_resolution_t* OPJ_RESTRICT r, OPJ_UINT32 i) {
- OPJ_UINT32 mr = 0;
- OPJ_UINT32 w;
- while( --i ) {
- ++r;
- if( mr < ( w = (OPJ_UINT32)(r->x1 - r->x0) ) )
- mr = w ;
- if( mr < ( w = (OPJ_UINT32)(r->y1 - r->y0) ) )
- mr = w ;
- }
- return mr ;
+static OPJ_UINT32 opj_dwt_max_resolution(opj_tcd_resolution_t* OPJ_RESTRICT r,
+ OPJ_UINT32 i)
+{
+ OPJ_UINT32 mr = 0;
+ OPJ_UINT32 w;
+ while (--i) {
+ ++r;
+ if (mr < (w = (OPJ_UINT32)(r->x1 - r->x0))) {
+ mr = w ;
+ }
+ if (mr < (w = (OPJ_UINT32)(r->y1 - r->y0))) {
+ mr = w ;
+ }
+ }
+ return mr ;
}
-typedef struct
-{
+typedef struct {
opj_dwt_t h;
DWT1DFN dwt_1D;
OPJ_UINT32 rw;
(void)tls;
job = (opj_dwd_decode_h_job_t*)user_data;
- for( j = job->min_j; j < job->max_j; j++ )
- {
- opj_dwt_interleave_h(&job->h, &job->tiledp[j*job->w]);
- (job->dwt_1D)(&job->h);
- memcpy(&job->tiledp[j*job->w], job->h.mem, job->rw * sizeof(OPJ_INT32));
+ for (j = job->min_j; j < job->max_j; j++) {
+ opj_dwt_interleave_h(&job->h, &job->tiledp[j * job->w]);
+ (job->dwt_1D)(&job->h);
+ memcpy(&job->tiledp[j * job->w], job->h.mem, job->rw * sizeof(OPJ_INT32));
}
opj_aligned_free(job->h.mem);
opj_free(job);
}
-typedef struct
-{
+typedef struct {
opj_dwt_t v;
DWT1DFN dwt_1D;
OPJ_UINT32 rh;
(void)tls;
job = (opj_dwd_decode_v_job_t*)user_data;
- for( j = job->min_j; j < job->max_j; j++ )
- {
+ for (j = job->min_j; j < job->max_j; j++) {
OPJ_UINT32 k;
opj_dwt_interleave_v(&job->v, &job->tiledp[j], (OPJ_INT32)job->w);
(job->dwt_1D)(&job->v);
- for(k = 0; k < job->rh; ++k) {
+ for (k = 0; k < job->rh; ++k) {
job->tiledp[k * job->w + j] = job->v.mem[k];
}
}
/* <summary> */
/* Inverse wavelet transform in 2-D. */
/* </summary> */
-static OPJ_BOOL opj_dwt_decode_tile(opj_thread_pool_t* tp, opj_tcd_tilecomp_t* tilec, OPJ_UINT32 numres, DWT1DFN dwt_1D)
+static OPJ_BOOL opj_dwt_decode_tile(opj_thread_pool_t* tp,
+ opj_tcd_tilecomp_t* tilec, OPJ_UINT32 numres, DWT1DFN dwt_1D)
{
- opj_dwt_t h;
- opj_dwt_t v;
-
- opj_tcd_resolution_t* tr = tilec->resolutions;
-
- OPJ_UINT32 rw = (OPJ_UINT32)(tr->x1 - tr->x0); /* width of the resolution level computed */
- OPJ_UINT32 rh = (OPJ_UINT32)(tr->y1 - tr->y0); /* height of the resolution level computed */
-
- OPJ_UINT32 w = (OPJ_UINT32)(tilec->x1 - tilec->x0);
- size_t h_mem_size;
- int num_threads;
-
- if (numres == 1U) {
- return OPJ_TRUE;
- }
- num_threads = opj_thread_pool_get_thread_count(tp);
- h_mem_size = opj_dwt_max_resolution(tr, numres);
- /* overflow check */
- if (h_mem_size > (SIZE_MAX / sizeof(OPJ_INT32))) {
- /* FIXME event manager error callback */
- return OPJ_FALSE;
- }
- h_mem_size *= sizeof(OPJ_INT32);
- h.mem = (OPJ_INT32*)opj_aligned_malloc(h_mem_size);
- if (! h.mem){
- /* FIXME event manager error callback */
- return OPJ_FALSE;
- }
-
- v.mem = h.mem;
-
- while( --numres) {
- OPJ_INT32 * OPJ_RESTRICT tiledp = tilec->data;
- OPJ_UINT32 j;
-
- ++tr;
- h.sn = (OPJ_INT32)rw;
- v.sn = (OPJ_INT32)rh;
-
- rw = (OPJ_UINT32)(tr->x1 - tr->x0);
- rh = (OPJ_UINT32)(tr->y1 - tr->y0);
-
- h.dn = (OPJ_INT32)(rw - (OPJ_UINT32)h.sn);
- h.cas = tr->x0 % 2;
-
- if( num_threads <= 1 || rh <= 1 )
- {
- for(j = 0; j < rh; ++j) {
- opj_dwt_interleave_h(&h, &tiledp[j*w]);
- (dwt_1D)(&h);
- memcpy(&tiledp[j*w], h.mem, rw * sizeof(OPJ_INT32));
- }
- }
- else
- {
- OPJ_UINT32 num_jobs = (OPJ_UINT32)num_threads;
- OPJ_UINT32 step_j;
-
- if( rh < num_jobs ) {
- num_jobs = rh;
- }
- step_j = (rh / num_jobs);
-
- for(j = 0; j < num_jobs; j++)
- {
- opj_dwd_decode_h_job_t* job;
-
- job = (opj_dwd_decode_h_job_t*) opj_malloc(sizeof(opj_dwd_decode_h_job_t));
- if( !job )
- {
- /* It would be nice to fallback to single thread case, but */
- /* unfortunately some jobs may be launched and have modified */
- /* tiledp, so it is not practical to recover from that error */
- /* FIXME event manager error callback */
- opj_thread_pool_wait_completion(tp, 0);
- opj_aligned_free(h.mem);
- return OPJ_FALSE;
- }
- job->h = h;
- job->dwt_1D = dwt_1D;
- job->rw = rw;
- job->w = w;
- job->tiledp = tiledp;
- job->min_j = j * step_j;
- job->max_j = (j + 1U) * step_j; /* this can overflow */
- if( j == (num_jobs - 1U) ) { /* this will take care of the overflow */
- job->max_j = rh;
- }
- job->h.mem = (OPJ_INT32*)opj_aligned_malloc(h_mem_size);
- if (!job->h.mem)
- {
- /* FIXME event manager error callback */
- opj_thread_pool_wait_completion(tp, 0);
- opj_free(job);
- opj_aligned_free(h.mem);
- return OPJ_FALSE;
- }
- opj_thread_pool_submit_job( tp, opj_dwt_decode_h_func, job );
- }
- opj_thread_pool_wait_completion(tp, 0);
- }
-
- v.dn = (OPJ_INT32)(rh - (OPJ_UINT32)v.sn);
- v.cas = tr->y0 % 2;
-
- if( num_threads <= 1 || rw <= 1 )
- {
- for(j = 0; j < rw; ++j){
- OPJ_UINT32 k;
-
- opj_dwt_interleave_v(&v, &tiledp[j], (OPJ_INT32)w);
- (dwt_1D)(&v);
- for(k = 0; k < rh; ++k) {
- tiledp[k * w + j] = v.mem[k];
- }
- }
- }
- else
- {
- OPJ_UINT32 num_jobs = (OPJ_UINT32)num_threads;
- OPJ_UINT32 step_j;
-
- if( rw < num_jobs ) {
- num_jobs = rw;
- }
- step_j = (rw / num_jobs);
-
- for( j = 0; j < num_jobs; j++ )
- {
- opj_dwd_decode_v_job_t* job;
-
- job = (opj_dwd_decode_v_job_t*) opj_malloc(sizeof(opj_dwd_decode_v_job_t));
- if( !job )
- {
- /* It would be nice to fallback to single thread case, but */
- /* unfortunately some jobs may be launched and have modified */
- /* tiledp, so it is not practical to recover from that error */
- /* FIXME event manager error callback */
- opj_thread_pool_wait_completion(tp, 0);
- opj_aligned_free(v.mem);
- return OPJ_FALSE;
- }
- job->v = v;
- job->dwt_1D = dwt_1D;
- job->rh = rh;
- job->w = w;
- job->tiledp = tiledp;
- job->min_j = j * step_j;
- job->max_j = (j + 1U) * step_j; /* this can overflow */
- if( j == (num_jobs - 1U) ) { /* this will take care of the overflow */
- job->max_j = rw;
- }
- job->v.mem = (OPJ_INT32*)opj_aligned_malloc(h_mem_size);
- if (!job->v.mem)
- {
- /* FIXME event manager error callback */
- opj_thread_pool_wait_completion(tp, 0);
- opj_free(job);
- opj_aligned_free(v.mem);
- return OPJ_FALSE;
- }
- opj_thread_pool_submit_job( tp, opj_dwt_decode_v_func, job );
- }
- opj_thread_pool_wait_completion(tp, 0);
- }
- }
- opj_aligned_free(h.mem);
- return OPJ_TRUE;
+ opj_dwt_t h;
+ opj_dwt_t v;
+
+ opj_tcd_resolution_t* tr = tilec->resolutions;
+
+ OPJ_UINT32 rw = (OPJ_UINT32)(tr->x1 -
+ tr->x0); /* width of the resolution level computed */
+ OPJ_UINT32 rh = (OPJ_UINT32)(tr->y1 -
+ tr->y0); /* height of the resolution level computed */
+
+ OPJ_UINT32 w = (OPJ_UINT32)(tilec->x1 - tilec->x0);
+ size_t h_mem_size;
+ int num_threads;
+
+ if (numres == 1U) {
+ return OPJ_TRUE;
+ }
+ num_threads = opj_thread_pool_get_thread_count(tp);
+ h_mem_size = opj_dwt_max_resolution(tr, numres);
+ /* overflow check */
+ if (h_mem_size > (SIZE_MAX / sizeof(OPJ_INT32))) {
+ /* FIXME event manager error callback */
+ return OPJ_FALSE;
+ }
+ h_mem_size *= sizeof(OPJ_INT32);
+ h.mem = (OPJ_INT32*)opj_aligned_malloc(h_mem_size);
+ if (! h.mem) {
+ /* FIXME event manager error callback */
+ return OPJ_FALSE;
+ }
+
+ v.mem = h.mem;
+
+ while (--numres) {
+ OPJ_INT32 * OPJ_RESTRICT tiledp = tilec->data;
+ OPJ_UINT32 j;
+
+ ++tr;
+ h.sn = (OPJ_INT32)rw;
+ v.sn = (OPJ_INT32)rh;
+
+ rw = (OPJ_UINT32)(tr->x1 - tr->x0);
+ rh = (OPJ_UINT32)(tr->y1 - tr->y0);
+
+ h.dn = (OPJ_INT32)(rw - (OPJ_UINT32)h.sn);
+ h.cas = tr->x0 % 2;
+
+ if (num_threads <= 1 || rh <= 1) {
+ for (j = 0; j < rh; ++j) {
+ opj_dwt_interleave_h(&h, &tiledp[j * w]);
+ (dwt_1D)(&h);
+ memcpy(&tiledp[j * w], h.mem, rw * sizeof(OPJ_INT32));
+ }
+ } else {
+ OPJ_UINT32 num_jobs = (OPJ_UINT32)num_threads;
+ OPJ_UINT32 step_j;
+
+ if (rh < num_jobs) {
+ num_jobs = rh;
+ }
+ step_j = (rh / num_jobs);
+
+ for (j = 0; j < num_jobs; j++) {
+ opj_dwd_decode_h_job_t* job;
+
+ job = (opj_dwd_decode_h_job_t*) opj_malloc(sizeof(opj_dwd_decode_h_job_t));
+ if (!job) {
+ /* It would be nice to fallback to single thread case, but */
+ /* unfortunately some jobs may be launched and have modified */
+ /* tiledp, so it is not practical to recover from that error */
+ /* FIXME event manager error callback */
+ opj_thread_pool_wait_completion(tp, 0);
+ opj_aligned_free(h.mem);
+ return OPJ_FALSE;
+ }
+ job->h = h;
+ job->dwt_1D = dwt_1D;
+ job->rw = rw;
+ job->w = w;
+ job->tiledp = tiledp;
+ job->min_j = j * step_j;
+ job->max_j = (j + 1U) * step_j; /* this can overflow */
+ if (j == (num_jobs - 1U)) { /* this will take care of the overflow */
+ job->max_j = rh;
+ }
+ job->h.mem = (OPJ_INT32*)opj_aligned_malloc(h_mem_size);
+ if (!job->h.mem) {
+ /* FIXME event manager error callback */
+ opj_thread_pool_wait_completion(tp, 0);
+ opj_free(job);
+ opj_aligned_free(h.mem);
+ return OPJ_FALSE;
+ }
+ opj_thread_pool_submit_job(tp, opj_dwt_decode_h_func, job);
+ }
+ opj_thread_pool_wait_completion(tp, 0);
+ }
+
+ v.dn = (OPJ_INT32)(rh - (OPJ_UINT32)v.sn);
+ v.cas = tr->y0 % 2;
+
+ if (num_threads <= 1 || rw <= 1) {
+ for (j = 0; j < rw; ++j) {
+ OPJ_UINT32 k;
+
+ opj_dwt_interleave_v(&v, &tiledp[j], (OPJ_INT32)w);
+ (dwt_1D)(&v);
+ for (k = 0; k < rh; ++k) {
+ tiledp[k * w + j] = v.mem[k];
+ }
+ }
+ } else {
+ OPJ_UINT32 num_jobs = (OPJ_UINT32)num_threads;
+ OPJ_UINT32 step_j;
+
+ if (rw < num_jobs) {
+ num_jobs = rw;
+ }
+ step_j = (rw / num_jobs);
+
+ for (j = 0; j < num_jobs; j++) {
+ opj_dwd_decode_v_job_t* job;
+
+ job = (opj_dwd_decode_v_job_t*) opj_malloc(sizeof(opj_dwd_decode_v_job_t));
+ if (!job) {
+ /* It would be nice to fallback to single thread case, but */
+ /* unfortunately some jobs may be launched and have modified */
+ /* tiledp, so it is not practical to recover from that error */
+ /* FIXME event manager error callback */
+ opj_thread_pool_wait_completion(tp, 0);
+ opj_aligned_free(v.mem);
+ return OPJ_FALSE;
+ }
+ job->v = v;
+ job->dwt_1D = dwt_1D;
+ job->rh = rh;
+ job->w = w;
+ job->tiledp = tiledp;
+ job->min_j = j * step_j;
+ job->max_j = (j + 1U) * step_j; /* this can overflow */
+ if (j == (num_jobs - 1U)) { /* this will take care of the overflow */
+ job->max_j = rw;
+ }
+ job->v.mem = (OPJ_INT32*)opj_aligned_malloc(h_mem_size);
+ if (!job->v.mem) {
+ /* FIXME event manager error callback */
+ opj_thread_pool_wait_completion(tp, 0);
+ opj_free(job);
+ opj_aligned_free(v.mem);
+ return OPJ_FALSE;
+ }
+ opj_thread_pool_submit_job(tp, opj_dwt_decode_v_func, job);
+ }
+ opj_thread_pool_wait_completion(tp, 0);
+ }
+ }
+ opj_aligned_free(h.mem);
+ return OPJ_TRUE;
}
-static void opj_v4dwt_interleave_h(opj_v4dwt_t* OPJ_RESTRICT w, OPJ_FLOAT32* OPJ_RESTRICT a, OPJ_INT32 x, OPJ_INT32 size)
+static void opj_v4dwt_interleave_h(opj_v4dwt_t* OPJ_RESTRICT w,
+ OPJ_FLOAT32* OPJ_RESTRICT a, OPJ_INT32 x, OPJ_INT32 size)
{
- OPJ_FLOAT32* OPJ_RESTRICT bi = (OPJ_FLOAT32*) (w->wavelet + w->cas);
- OPJ_INT32 count = w->sn;
- OPJ_INT32 i, k;
-
- for(k = 0; k < 2; ++k){
- if ( count + 3 * x < size && ((size_t) a & 0x0f) == 0 && ((size_t) bi & 0x0f) == 0 && (x & 0x0f) == 0 ) {
- /* Fast code path */
- for(i = 0; i < count; ++i){
- OPJ_INT32 j = i;
- bi[i*8 ] = a[j];
- j += x;
- bi[i*8 + 1] = a[j];
- j += x;
- bi[i*8 + 2] = a[j];
- j += x;
- bi[i*8 + 3] = a[j];
- }
- }
- else {
- /* Slow code path */
- for(i = 0; i < count; ++i){
- OPJ_INT32 j = i;
- bi[i*8 ] = a[j];
- j += x;
- if(j >= size) continue;
- bi[i*8 + 1] = a[j];
- j += x;
- if(j >= size) continue;
- bi[i*8 + 2] = a[j];
- j += x;
- if(j >= size) continue;
- bi[i*8 + 3] = a[j]; /* This one*/
- }
- }
-
- bi = (OPJ_FLOAT32*) (w->wavelet + 1 - w->cas);
- a += w->sn;
- size -= w->sn;
- count = w->dn;
- }
+ OPJ_FLOAT32* OPJ_RESTRICT bi = (OPJ_FLOAT32*)(w->wavelet + w->cas);
+ OPJ_INT32 count = w->sn;
+ OPJ_INT32 i, k;
+
+ for (k = 0; k < 2; ++k) {
+ if (count + 3 * x < size && ((size_t) a & 0x0f) == 0 &&
+ ((size_t) bi & 0x0f) == 0 && (x & 0x0f) == 0) {
+ /* Fast code path */
+ for (i = 0; i < count; ++i) {
+ OPJ_INT32 j = i;
+ bi[i * 8 ] = a[j];
+ j += x;
+ bi[i * 8 + 1] = a[j];
+ j += x;
+ bi[i * 8 + 2] = a[j];
+ j += x;
+ bi[i * 8 + 3] = a[j];
+ }
+ } else {
+ /* Slow code path */
+ for (i = 0; i < count; ++i) {
+ OPJ_INT32 j = i;
+ bi[i * 8 ] = a[j];
+ j += x;
+ if (j >= size) {
+ continue;
+ }
+ bi[i * 8 + 1] = a[j];
+ j += x;
+ if (j >= size) {
+ continue;
+ }
+ bi[i * 8 + 2] = a[j];
+ j += x;
+ if (j >= size) {
+ continue;
+ }
+ bi[i * 8 + 3] = a[j]; /* This one*/
+ }
+ }
+
+ bi = (OPJ_FLOAT32*)(w->wavelet + 1 - w->cas);
+ a += w->sn;
+ size -= w->sn;
+ count = w->dn;
+ }
}
-static void opj_v4dwt_interleave_v(opj_v4dwt_t* OPJ_RESTRICT v , OPJ_FLOAT32* OPJ_RESTRICT a , OPJ_INT32 x, OPJ_INT32 nb_elts_read){
- opj_v4_t* OPJ_RESTRICT bi = v->wavelet + v->cas;
- OPJ_INT32 i;
+static void opj_v4dwt_interleave_v(opj_v4dwt_t* OPJ_RESTRICT v,
+ OPJ_FLOAT32* OPJ_RESTRICT a, OPJ_INT32 x, OPJ_INT32 nb_elts_read)
+{
+ opj_v4_t* OPJ_RESTRICT bi = v->wavelet + v->cas;
+ OPJ_INT32 i;
- for(i = 0; i < v->sn; ++i){
- memcpy(&bi[i*2], &a[i*x], (size_t)nb_elts_read * sizeof(OPJ_FLOAT32));
- }
+ for (i = 0; i < v->sn; ++i) {
+ memcpy(&bi[i * 2], &a[i * x], (size_t)nb_elts_read * sizeof(OPJ_FLOAT32));
+ }
- a += v->sn * x;
- bi = v->wavelet + 1 - v->cas;
+ a += v->sn * x;
+ bi = v->wavelet + 1 - v->cas;
- for(i = 0; i < v->dn; ++i){
- memcpy(&bi[i*2], &a[i*x], (size_t)nb_elts_read * sizeof(OPJ_FLOAT32));
- }
+ for (i = 0; i < v->dn; ++i) {
+ memcpy(&bi[i * 2], &a[i * x], (size_t)nb_elts_read * sizeof(OPJ_FLOAT32));
+ }
}
#ifdef __SSE__
-static void opj_v4dwt_decode_step1_sse(opj_v4_t* w, OPJ_INT32 count, const __m128 c){
- __m128* OPJ_RESTRICT vw = (__m128*) w;
- OPJ_INT32 i;
- /* 4x unrolled loop */
- for(i = 0; i < count >> 2; ++i){
- *vw = _mm_mul_ps(*vw, c);
- vw += 2;
- *vw = _mm_mul_ps(*vw, c);
- vw += 2;
- *vw = _mm_mul_ps(*vw, c);
- vw += 2;
- *vw = _mm_mul_ps(*vw, c);
- vw += 2;
- }
- count &= 3;
- for(i = 0; i < count; ++i){
- *vw = _mm_mul_ps(*vw, c);
- vw += 2;
- }
+static void opj_v4dwt_decode_step1_sse(opj_v4_t* w, OPJ_INT32 count,
+ const __m128 c)
+{
+ __m128* OPJ_RESTRICT vw = (__m128*) w;
+ OPJ_INT32 i;
+ /* 4x unrolled loop */
+ for (i = 0; i < count >> 2; ++i) {
+ *vw = _mm_mul_ps(*vw, c);
+ vw += 2;
+ *vw = _mm_mul_ps(*vw, c);
+ vw += 2;
+ *vw = _mm_mul_ps(*vw, c);
+ vw += 2;
+ *vw = _mm_mul_ps(*vw, c);
+ vw += 2;
+ }
+ count &= 3;
+ for (i = 0; i < count; ++i) {
+ *vw = _mm_mul_ps(*vw, c);
+ vw += 2;
+ }
}
-void opj_v4dwt_decode_step2_sse(opj_v4_t* l, opj_v4_t* w, OPJ_INT32 k, OPJ_INT32 m, __m128 c){
- __m128* OPJ_RESTRICT vl = (__m128*) l;
- __m128* OPJ_RESTRICT vw = (__m128*) w;
- OPJ_INT32 i;
- __m128 tmp1, tmp2, tmp3;
- tmp1 = vl[0];
- for(i = 0; i < m; ++i){
- tmp2 = vw[-1];
- tmp3 = vw[ 0];
- vw[-1] = _mm_add_ps(tmp2, _mm_mul_ps(_mm_add_ps(tmp1, tmp3), c));
- tmp1 = tmp3;
- vw += 2;
- }
- vl = vw - 2;
- if(m >= k){
- return;
- }
- c = _mm_add_ps(c, c);
- c = _mm_mul_ps(c, vl[0]);
- for(; m < k; ++m){
- __m128 tmp = vw[-1];
- vw[-1] = _mm_add_ps(tmp, c);
- vw += 2;
- }
+void opj_v4dwt_decode_step2_sse(opj_v4_t* l, opj_v4_t* w, OPJ_INT32 k,
+ OPJ_INT32 m, __m128 c)
+{
+ __m128* OPJ_RESTRICT vl = (__m128*) l;
+ __m128* OPJ_RESTRICT vw = (__m128*) w;
+ OPJ_INT32 i;
+ __m128 tmp1, tmp2, tmp3;
+ tmp1 = vl[0];
+ for (i = 0; i < m; ++i) {
+ tmp2 = vw[-1];
+ tmp3 = vw[ 0];
+ vw[-1] = _mm_add_ps(tmp2, _mm_mul_ps(_mm_add_ps(tmp1, tmp3), c));
+ tmp1 = tmp3;
+ vw += 2;
+ }
+ vl = vw - 2;
+ if (m >= k) {
+ return;
+ }
+ c = _mm_add_ps(c, c);
+ c = _mm_mul_ps(c, vl[0]);
+ for (; m < k; ++m) {
+ __m128 tmp = vw[-1];
+ vw[-1] = _mm_add_ps(tmp, c);
+ vw += 2;
+ }
}
#else
-static void opj_v4dwt_decode_step1(opj_v4_t* w, OPJ_INT32 count, const OPJ_FLOAT32 c)
+static void opj_v4dwt_decode_step1(opj_v4_t* w, OPJ_INT32 count,
+ const OPJ_FLOAT32 c)
{
- OPJ_FLOAT32* OPJ_RESTRICT fw = (OPJ_FLOAT32*) w;
- OPJ_INT32 i;
- for(i = 0; i < count; ++i){
- 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;
- fw[i*8 + 3] = tmp4 * c;
- }
+ OPJ_FLOAT32* OPJ_RESTRICT fw = (OPJ_FLOAT32*) w;
+ OPJ_INT32 i;
+ for (i = 0; i < count; ++i) {
+ 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;
+ fw[i * 8 + 3] = tmp4 * c;
+ }
}
-static void opj_v4dwt_decode_step2(opj_v4_t* l, opj_v4_t* w, OPJ_INT32 k, OPJ_INT32 m, OPJ_FLOAT32 c)
+static void opj_v4dwt_decode_step2(opj_v4_t* l, opj_v4_t* w, OPJ_INT32 k,
+ OPJ_INT32 m, OPJ_FLOAT32 c)
{
- OPJ_FLOAT32* fl = (OPJ_FLOAT32*) l;
- OPJ_FLOAT32* fw = (OPJ_FLOAT32*) w;
- OPJ_INT32 i;
- for(i = 0; i < m; ++i){
- 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);
- fw[-1] = tmp2_4 + ((tmp1_4 + tmp3_4) * c);
- fl = fw;
- fw += 8;
- }
- if(m < k){
- 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){
- 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;
- fw[-1] = tmp4 + c4;
- fw += 8;
- }
- }
+ OPJ_FLOAT32* fl = (OPJ_FLOAT32*) l;
+ OPJ_FLOAT32* fw = (OPJ_FLOAT32*) w;
+ OPJ_INT32 i;
+ for (i = 0; i < m; ++i) {
+ 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);
+ fw[-1] = tmp2_4 + ((tmp1_4 + tmp3_4) * c);
+ fl = fw;
+ fw += 8;
+ }
+ if (m < k) {
+ 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) {
+ 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;
+ fw[-1] = tmp4 + c4;
+ fw += 8;
+ }
+ }
}
#endif
/* </summary> */
static void opj_v4dwt_decode(opj_v4dwt_t* OPJ_RESTRICT dwt)
{
- OPJ_INT32 a, b;
- if(dwt->cas == 0) {
- if(!((dwt->dn > 0) || (dwt->sn > 1))){
- return;
- }
- a = 0;
- b = 1;
- }else{
- if(!((dwt->sn > 0) || (dwt->dn > 1))) {
- return;
- }
- a = 1;
- b = 0;
- }
+ OPJ_INT32 a, b;
+ if (dwt->cas == 0) {
+ if (!((dwt->dn > 0) || (dwt->sn > 1))) {
+ return;
+ }
+ a = 0;
+ b = 1;
+ } else {
+ if (!((dwt->sn > 0) || (dwt->dn > 1))) {
+ return;
+ }
+ a = 1;
+ b = 0;
+ }
#ifdef __SSE__
- opj_v4dwt_decode_step1_sse(dwt->wavelet+a, dwt->sn, _mm_set1_ps(opj_K));
- opj_v4dwt_decode_step1_sse(dwt->wavelet+b, dwt->dn, _mm_set1_ps(opj_c13318));
- opj_v4dwt_decode_step2_sse(dwt->wavelet+b, dwt->wavelet+a+1, dwt->sn, opj_int_min(dwt->sn, dwt->dn-a), _mm_set1_ps(opj_dwt_delta));
- opj_v4dwt_decode_step2_sse(dwt->wavelet+a, dwt->wavelet+b+1, dwt->dn, opj_int_min(dwt->dn, dwt->sn-b), _mm_set1_ps(opj_dwt_gamma));
- opj_v4dwt_decode_step2_sse(dwt->wavelet+b, dwt->wavelet+a+1, dwt->sn, opj_int_min(dwt->sn, dwt->dn-a), _mm_set1_ps(opj_dwt_beta));
- opj_v4dwt_decode_step2_sse(dwt->wavelet+a, dwt->wavelet+b+1, dwt->dn, opj_int_min(dwt->dn, dwt->sn-b), _mm_set1_ps(opj_dwt_alpha));
+ opj_v4dwt_decode_step1_sse(dwt->wavelet + a, dwt->sn, _mm_set1_ps(opj_K));
+ opj_v4dwt_decode_step1_sse(dwt->wavelet + b, dwt->dn, _mm_set1_ps(opj_c13318));
+ opj_v4dwt_decode_step2_sse(dwt->wavelet + b, dwt->wavelet + a + 1, dwt->sn,
+ opj_int_min(dwt->sn, dwt->dn - a), _mm_set1_ps(opj_dwt_delta));
+ opj_v4dwt_decode_step2_sse(dwt->wavelet + a, dwt->wavelet + b + 1, dwt->dn,
+ opj_int_min(dwt->dn, dwt->sn - b), _mm_set1_ps(opj_dwt_gamma));
+ opj_v4dwt_decode_step2_sse(dwt->wavelet + b, dwt->wavelet + a + 1, dwt->sn,
+ opj_int_min(dwt->sn, dwt->dn - a), _mm_set1_ps(opj_dwt_beta));
+ opj_v4dwt_decode_step2_sse(dwt->wavelet + a, dwt->wavelet + b + 1, dwt->dn,
+ opj_int_min(dwt->dn, dwt->sn - b), _mm_set1_ps(opj_dwt_alpha));
#else
- opj_v4dwt_decode_step1(dwt->wavelet+a, dwt->sn, opj_K);
- opj_v4dwt_decode_step1(dwt->wavelet+b, dwt->dn, opj_c13318);
- opj_v4dwt_decode_step2(dwt->wavelet+b, dwt->wavelet+a+1, dwt->sn, opj_int_min(dwt->sn, dwt->dn-a), opj_dwt_delta);
- opj_v4dwt_decode_step2(dwt->wavelet+a, dwt->wavelet+b+1, dwt->dn, opj_int_min(dwt->dn, dwt->sn-b), opj_dwt_gamma);
- opj_v4dwt_decode_step2(dwt->wavelet+b, dwt->wavelet+a+1, dwt->sn, opj_int_min(dwt->sn, dwt->dn-a), opj_dwt_beta);
- opj_v4dwt_decode_step2(dwt->wavelet+a, dwt->wavelet+b+1, dwt->dn, opj_int_min(dwt->dn, dwt->sn-b), opj_dwt_alpha);
+ opj_v4dwt_decode_step1(dwt->wavelet + a, dwt->sn, opj_K);
+ opj_v4dwt_decode_step1(dwt->wavelet + b, dwt->dn, opj_c13318);
+ opj_v4dwt_decode_step2(dwt->wavelet + b, dwt->wavelet + a + 1, dwt->sn,
+ opj_int_min(dwt->sn, dwt->dn - a), opj_dwt_delta);
+ opj_v4dwt_decode_step2(dwt->wavelet + a, dwt->wavelet + b + 1, dwt->dn,
+ opj_int_min(dwt->dn, dwt->sn - b), opj_dwt_gamma);
+ opj_v4dwt_decode_step2(dwt->wavelet + b, dwt->wavelet + a + 1, dwt->sn,
+ opj_int_min(dwt->sn, dwt->dn - a), opj_dwt_beta);
+ opj_v4dwt_decode_step2(dwt->wavelet + a, dwt->wavelet + b + 1, dwt->dn,
+ opj_int_min(dwt->dn, dwt->sn - b), opj_dwt_alpha);
#endif
}
/* <summary> */
/* Inverse 9-7 wavelet transform in 2-D. */
/* </summary> */
-OPJ_BOOL opj_dwt_decode_real(opj_tcd_tilecomp_t* OPJ_RESTRICT tilec, OPJ_UINT32 numres)
+OPJ_BOOL opj_dwt_decode_real(opj_tcd_tilecomp_t* OPJ_RESTRICT tilec,
+ OPJ_UINT32 numres)
{
- opj_v4dwt_t h;
- opj_v4dwt_t v;
-
- opj_tcd_resolution_t* res = tilec->resolutions;
-
- OPJ_UINT32 rw = (OPJ_UINT32)(res->x1 - res->x0); /* width of the resolution level computed */
- OPJ_UINT32 rh = (OPJ_UINT32)(res->y1 - res->y0); /* height of the resolution level computed */
-
- OPJ_UINT32 w = (OPJ_UINT32)(tilec->x1 - tilec->x0);
-
- size_t l_data_size;
-
- l_data_size = opj_dwt_max_resolution(res, numres);
- /* overflow check */
- if (l_data_size > (SIZE_MAX - 5U)) {
- /* FIXME event manager error callback */
- return OPJ_FALSE;
- }
- l_data_size += 5U;
- /* overflow check */
- if (l_data_size > (SIZE_MAX / sizeof(opj_v4_t))) {
- /* FIXME event manager error callback */
- return OPJ_FALSE;
- }
- h.wavelet = (opj_v4_t*) opj_aligned_malloc(l_data_size * sizeof(opj_v4_t));
- if (!h.wavelet) {
- /* FIXME event manager error callback */
- return OPJ_FALSE;
- }
- v.wavelet = h.wavelet;
-
- while( --numres) {
- OPJ_FLOAT32 * OPJ_RESTRICT aj = (OPJ_FLOAT32*) tilec->data;
- OPJ_UINT32 bufsize = (OPJ_UINT32)((tilec->x1 - tilec->x0) * (tilec->y1 - tilec->y0));
- OPJ_INT32 j;
-
- h.sn = (OPJ_INT32)rw;
- v.sn = (OPJ_INT32)rh;
-
- ++res;
-
- rw = (OPJ_UINT32)(res->x1 - res->x0); /* width of the resolution level computed */
- rh = (OPJ_UINT32)(res->y1 - res->y0); /* height of the resolution level computed */
-
- h.dn = (OPJ_INT32)(rw - (OPJ_UINT32)h.sn);
- h.cas = res->x0 % 2;
-
- for(j = (OPJ_INT32)rh; j > 3; j -= 4) {
- OPJ_INT32 k;
- opj_v4dwt_interleave_h(&h, aj, (OPJ_INT32)w, (OPJ_INT32)bufsize);
- opj_v4dwt_decode(&h);
-
- for(k = (OPJ_INT32)rw; --k >= 0;){
- aj[k ] = h.wavelet[k].f[0];
- aj[k+(OPJ_INT32)w ] = h.wavelet[k].f[1];
- aj[k+(OPJ_INT32)w*2] = h.wavelet[k].f[2];
- aj[k+(OPJ_INT32)w*3] = h.wavelet[k].f[3];
- }
-
- aj += w*4;
- bufsize -= w*4;
- }
-
- if (rh & 0x03) {
- OPJ_INT32 k;
- j = rh & 0x03;
- opj_v4dwt_interleave_h(&h, aj, (OPJ_INT32)w, (OPJ_INT32)bufsize);
- opj_v4dwt_decode(&h);
- for(k = (OPJ_INT32)rw; --k >= 0;){
- switch(j) {
- case 3: aj[k+(OPJ_INT32)w*2] = h.wavelet[k].f[2];
- case 2: aj[k+(OPJ_INT32)w ] = h.wavelet[k].f[1];
- case 1: aj[k ] = h.wavelet[k].f[0];
- }
- }
- }
-
- v.dn = (OPJ_INT32)(rh - (OPJ_UINT32)v.sn);
- v.cas = res->y0 % 2;
-
- aj = (OPJ_FLOAT32*) tilec->data;
- for(j = (OPJ_INT32)rw; j > 3; j -= 4){
- OPJ_UINT32 k;
-
- opj_v4dwt_interleave_v(&v, aj, (OPJ_INT32)w, 4);
- opj_v4dwt_decode(&v);
-
- for(k = 0; k < rh; ++k){
- memcpy(&aj[k*w], &v.wavelet[k], 4 * sizeof(OPJ_FLOAT32));
- }
- aj += 4;
- }
-
- if (rw & 0x03){
- OPJ_UINT32 k;
-
- j = rw & 0x03;
-
- opj_v4dwt_interleave_v(&v, aj, (OPJ_INT32)w, j);
- opj_v4dwt_decode(&v);
-
- for(k = 0; k < rh; ++k){
- memcpy(&aj[k*w], &v.wavelet[k], (size_t)j * sizeof(OPJ_FLOAT32));
- }
- }
- }
-
- opj_aligned_free(h.wavelet);
- return OPJ_TRUE;
+ opj_v4dwt_t h;
+ opj_v4dwt_t v;
+
+ opj_tcd_resolution_t* res = tilec->resolutions;
+
+ OPJ_UINT32 rw = (OPJ_UINT32)(res->x1 -
+ res->x0); /* width of the resolution level computed */
+ OPJ_UINT32 rh = (OPJ_UINT32)(res->y1 -
+ res->y0); /* height of the resolution level computed */
+
+ OPJ_UINT32 w = (OPJ_UINT32)(tilec->x1 - tilec->x0);
+
+ size_t l_data_size;
+
+ l_data_size = opj_dwt_max_resolution(res, numres);
+ /* overflow check */
+ if (l_data_size > (SIZE_MAX - 5U)) {
+ /* FIXME event manager error callback */
+ return OPJ_FALSE;
+ }
+ l_data_size += 5U;
+ /* overflow check */
+ if (l_data_size > (SIZE_MAX / sizeof(opj_v4_t))) {
+ /* FIXME event manager error callback */
+ return OPJ_FALSE;
+ }
+ h.wavelet = (opj_v4_t*) opj_aligned_malloc(l_data_size * sizeof(opj_v4_t));
+ if (!h.wavelet) {
+ /* FIXME event manager error callback */
+ return OPJ_FALSE;
+ }
+ v.wavelet = h.wavelet;
+
+ while (--numres) {
+ OPJ_FLOAT32 * OPJ_RESTRICT aj = (OPJ_FLOAT32*) tilec->data;
+ OPJ_UINT32 bufsize = (OPJ_UINT32)((tilec->x1 - tilec->x0) *
+ (tilec->y1 - tilec->y0));
+ OPJ_INT32 j;
+
+ h.sn = (OPJ_INT32)rw;
+ v.sn = (OPJ_INT32)rh;
+
+ ++res;
+
+ rw = (OPJ_UINT32)(res->x1 -
+ res->x0); /* width of the resolution level computed */
+ rh = (OPJ_UINT32)(res->y1 -
+ res->y0); /* height of the resolution level computed */
+
+ h.dn = (OPJ_INT32)(rw - (OPJ_UINT32)h.sn);
+ h.cas = res->x0 % 2;
+
+ for (j = (OPJ_INT32)rh; j > 3; j -= 4) {
+ OPJ_INT32 k;
+ opj_v4dwt_interleave_h(&h, aj, (OPJ_INT32)w, (OPJ_INT32)bufsize);
+ opj_v4dwt_decode(&h);
+
+ for (k = (OPJ_INT32)rw; --k >= 0;) {
+ aj[k ] = h.wavelet[k].f[0];
+ aj[k + (OPJ_INT32)w ] = h.wavelet[k].f[1];
+ aj[k + (OPJ_INT32)w * 2] = h.wavelet[k].f[2];
+ aj[k + (OPJ_INT32)w * 3] = h.wavelet[k].f[3];
+ }
+
+ aj += w * 4;
+ bufsize -= w * 4;
+ }
+
+ if (rh & 0x03) {
+ OPJ_INT32 k;
+ j = rh & 0x03;
+ opj_v4dwt_interleave_h(&h, aj, (OPJ_INT32)w, (OPJ_INT32)bufsize);
+ opj_v4dwt_decode(&h);
+ for (k = (OPJ_INT32)rw; --k >= 0;) {
+ switch (j) {
+ case 3:
+ aj[k + (OPJ_INT32)w * 2] = h.wavelet[k].f[2];
+ case 2:
+ aj[k + (OPJ_INT32)w ] = h.wavelet[k].f[1];
+ case 1:
+ aj[k ] = h.wavelet[k].f[0];
+ }
+ }
+ }
+
+ v.dn = (OPJ_INT32)(rh - (OPJ_UINT32)v.sn);
+ v.cas = res->y0 % 2;
+
+ aj = (OPJ_FLOAT32*) tilec->data;
+ for (j = (OPJ_INT32)rw; j > 3; j -= 4) {
+ OPJ_UINT32 k;
+
+ opj_v4dwt_interleave_v(&v, aj, (OPJ_INT32)w, 4);
+ opj_v4dwt_decode(&v);
+
+ for (k = 0; k < rh; ++k) {
+ memcpy(&aj[k * w], &v.wavelet[k], 4 * sizeof(OPJ_FLOAT32));
+ }
+ aj += 4;
+ }
+
+ if (rw & 0x03) {
+ OPJ_UINT32 k;
+
+ j = rw & 0x03;
+
+ opj_v4dwt_interleave_v(&v, aj, (OPJ_INT32)w, j);
+ opj_v4dwt_decode(&v);
+
+ for (k = 0; k < rh; ++k) {
+ memcpy(&aj[k * w], &v.wavelet[k], (size_t)j * sizeof(OPJ_FLOAT32));
+ }
+ }
+ }
+
+ opj_aligned_free(h.wavelet);
+ return OPJ_TRUE;
}