blob: 99c7a8cda7d374f33eb2d897a11083e67469c14a [file] [log] [blame]
diff --git a/third_party/libopenjpeg20/dwt.c b/third_party/libopenjpeg20/dwt.c
index 4164ba090..a36b7ed10 100644
--- a/third_party/libopenjpeg20/dwt.c
+++ b/third_party/libopenjpeg20/dwt.c
@@ -63,9 +63,6 @@
/** @defgroup DWT DWT - Implementation of a discrete wavelet transform */
/*@{*/
-#define OPJ_WS(i) v->mem[(i)*2]
-#define OPJ_WD(i) v->mem[(1+(i)*2)]
-
#ifdef __AVX2__
/** Number of int32 values in a AVX2 register */
#define VREG_INT_COUNT 8
@@ -82,6 +79,7 @@
typedef struct dwt_local {
OPJ_INT32* mem;
+ OPJ_SIZE_T mem_count;
OPJ_INT32 dn; /* number of elements in high pass band */
OPJ_INT32 sn; /* number of elements in low pass band */
OPJ_INT32 cas; /* 0 = start on even coord, 1 = start on odd coord */
@@ -140,7 +138,7 @@ static void opj_dwt_encode_stepsize(OPJ_INT32 stepsize, OPJ_INT32 numbps,
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);
+ const opj_tcd_tilecomp_t* tilec, OPJ_UINT32 i);
static OPJ_BOOL opj_dwt_decode_partial_tile(
opj_tcd_tilecomp_t* tilec,
@@ -181,13 +179,20 @@ static OPJ_UINT32 opj_dwt_max_resolution(opj_tcd_resolution_t* OPJ_RESTRICT r,
/*@}*/
-#define OPJ_S(i) a[(i)*2]
-#define OPJ_D(i) a[(1+(i)*2)]
-#define OPJ_S_(i) ((i)<0?OPJ_S(0):((i)>=sn?OPJ_S(sn-1):OPJ_S(i)))
-#define OPJ_D_(i) ((i)<0?OPJ_D(0):((i)>=dn?OPJ_D(dn-1):OPJ_D(i)))
+#define IDX_S(i) (i)*2
+#define IDX_D(i) 1 + (i)* 2
+#define UNDERFLOW_SN(i) ((i) >= sn&&sn>0)
+#define UNDERFLOW_DN(i) ((i) >= dn&&dn>0)
+#define OVERFLOW_S(i) (IDX_S(i) >= a_count)
+#define OVERFLOW_D(i) (IDX_D(i) >= a_count)
+
+#define OPJ_S(i) a[IDX_S(i)]
+#define OPJ_D(i) a[IDX_D(i)]
+#define OPJ_S_(i) ((i)<0 ? OPJ_S(0) : (UNDERFLOW_SN(i) ? OPJ_S(sn - 1) : OVERFLOW_S(i) ? OPJ_S(i - 1) : OPJ_S(i)))
+#define OPJ_D_(i) ((i)<0 ? OPJ_D(0) : (UNDERFLOW_DN(i) ? OPJ_D(dn - 1) : OVERFLOW_D(i) ? OPJ_D(i - 1) : OPJ_D(i)))
/* new */
-#define OPJ_SS_(i) ((i)<0?OPJ_S(0):((i)>=dn?OPJ_S(dn-1):OPJ_S(i)))
-#define OPJ_DD_(i) ((i)<0?OPJ_D(0):((i)>=sn?OPJ_D(sn-1):OPJ_D(i)))
+#define OPJ_SS_(i) ((i)<0 ? OPJ_S(0) : (UNDERFLOW_DN(i) ? OPJ_S(dn - 1) : OVERFLOW_S(i) ? OPJ_S(i - 1) : OPJ_S(i)))
+#define OPJ_DD_(i) ((i)<0 ? OPJ_D(0) : (UNDERFLOW_SN(i) ? OPJ_D(sn - 1) : OVERFLOW_D(i) ? OPJ_D(i - 1) : OPJ_D(i)))
/* <summary> */
/* This table contains the norms of the 5-3 wavelets for different bands. */
@@ -296,8 +301,8 @@ static void opj_dwt_interleave_v(const opj_dwt_t* v, OPJ_INT32 *a, OPJ_INT32 x)
/* <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)
+static void opj_dwt_decode_1_(OPJ_INT32 *a, OPJ_SIZE_T a_count, OPJ_INT32 dn,
+ OPJ_INT32 sn, OPJ_INT32 cas)
{
OPJ_INT32 i;
@@ -326,7 +331,7 @@ static void opj_dwt_decode_1_(OPJ_INT32 *a, OPJ_INT32 dn, OPJ_INT32 sn,
static void opj_dwt_decode_1(const opj_dwt_t *v)
{
- opj_dwt_decode_1_(v->mem, v->dn, v->sn, v->cas);
+ opj_dwt_decode_1_(v->mem, v->mem_count, v->dn, v->sn, v->cas);
}
#endif /* STANDARD_SLOW_VERSION */
@@ -2062,7 +2067,7 @@ static void opj_dwt_decode_v_func(void* user_data, opj_tls_t* tls)
/* 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)
+ const opj_tcd_tilecomp_t* tilec, OPJ_UINT32 numres)
{
opj_dwt_t h;
opj_dwt_t v;
@@ -2084,22 +2089,23 @@ static OPJ_BOOL opj_dwt_decode_tile(opj_thread_pool_t* tp,
return OPJ_TRUE;
}
num_threads = opj_thread_pool_get_thread_count(tp);
- h_mem_size = opj_dwt_max_resolution(tr, numres);
+ h.mem_count = opj_dwt_max_resolution(tr, numres);
/* overflow check */
- if (h_mem_size > (SIZE_MAX / PARALLEL_COLS_53 / sizeof(OPJ_INT32))) {
+ if (h.mem_count > (SIZE_MAX / PARALLEL_COLS_53 / sizeof(OPJ_INT32))) {
/* FIXME event manager error callback */
return OPJ_FALSE;
}
/* We need PARALLEL_COLS_53 times the height of the array, */
/* since for the vertical pass */
/* we process PARALLEL_COLS_53 columns at a time */
- h_mem_size *= PARALLEL_COLS_53 * sizeof(OPJ_INT32);
+ h_mem_size = h.mem_count * PARALLEL_COLS_53 * sizeof(OPJ_INT32);
h.mem = (OPJ_INT32*)opj_aligned_32_malloc(h_mem_size);
if (! h.mem) {
/* FIXME event manager error callback */
return OPJ_FALSE;
}
+ v.mem_count = h.mem_count;
v.mem = h.mem;
while (--numres) {
@@ -2277,7 +2283,8 @@ static void opj_dwt_interleave_partial_v(OPJ_INT32 *dest,
OPJ_UNUSED(ret);
}
-static void opj_dwt_decode_partial_1(OPJ_INT32 *a, OPJ_INT32 dn, OPJ_INT32 sn,
+static void opj_dwt_decode_partial_1(OPJ_INT32 *a, OPJ_SIZE_T a_count,
+ OPJ_INT32 dn, OPJ_INT32 sn,
OPJ_INT32 cas,
OPJ_INT32 win_l_x0,
OPJ_INT32 win_l_x1,
@@ -2657,16 +2664,16 @@ static OPJ_BOOL opj_dwt_decode_partial_tile(
opj_sparse_array_int32_free(sa);
return OPJ_TRUE;
}
- h_mem_size = opj_dwt_max_resolution(tr, numres);
+ h.mem_count = opj_dwt_max_resolution(tr, numres);
/* overflow check */
/* in vertical pass, we process 4 columns at a time */
- if (h_mem_size > (SIZE_MAX / (4 * sizeof(OPJ_INT32)))) {
+ if (h.mem_count > (SIZE_MAX / (4 * sizeof(OPJ_INT32)))) {
/* FIXME event manager error callback */
opj_sparse_array_int32_free(sa);
return OPJ_FALSE;
}
- h_mem_size *= 4 * sizeof(OPJ_INT32);
+ h_mem_size = h.mem_count * 4 * sizeof(OPJ_INT32);
h.mem = (OPJ_INT32*)opj_aligned_32_malloc(h_mem_size);
if (! h.mem) {
/* FIXME event manager error callback */
@@ -2674,6 +2681,7 @@ static OPJ_BOOL opj_dwt_decode_partial_tile(
return OPJ_FALSE;
}
+ v.mem_count = h.mem_count;
v.mem = h.mem;
for (resno = 1; resno < numres; resno ++) {
@@ -2784,7 +2792,7 @@ static OPJ_BOOL opj_dwt_decode_partial_tile(
win_ll_x1,
win_hl_x0,
win_hl_x1);
- opj_dwt_decode_partial_1(h.mem, h.dn, h.sn, h.cas,
+ opj_dwt_decode_partial_1(h.mem, h.mem_count, h.dn, h.sn, h.cas,
(OPJ_INT32)win_ll_x0,
(OPJ_INT32)win_ll_x1,
(OPJ_INT32)win_hl_x0,