* POSSIBILITY OF SUCH DAMAGE.
*/
+#define OPJ_SKIP_POISON
#include "opj_includes.h"
+
+#ifdef __SSE__
+#include <xmmintrin.h>
+#endif
+#ifdef __SSE2__
+#include <emmintrin.h>
+#endif
+
+#if defined(__GNUC__)
+#pragma GCC poison malloc calloc realloc free
+#endif
+
#include "t1_luts.h"
/** @defgroup T1 T1 - Implementation of the tier-1 coding */
if ((flags & ((T1_SIGMA_THIS | T1_PI_THIS) << (ci * 3U))) == 0U &&
(flags & (T1_SIGMA_NEIGHBOURS << (ci * 3U))) != 0U) {
OPJ_UINT32 ctxt1 = opj_t1_getctxno_zc(mqc, flags >> (ci * 3U));
- v = opj_int_abs(*datap) & one ? 1 : 0;
+ v = (opj_int_abs(*datap) & one) ? 1 : 0;
#ifdef DEBUG_ENC_SIG
fprintf(stderr, " ctxt1=%d\n", ctxt1);
#endif
OPJ_UINT32 ctxt = opj_t1_getctxno_mag(shift_flags);
*nmsedec += opj_t1_getnmsedec_ref((OPJ_UINT32)opj_int_abs(*datap),
(OPJ_UINT32)bpno);
- v = opj_int_abs(*datap) & one ? 1 : 0;
+ v = (opj_int_abs(*datap) & one) ? 1 : 0;
#ifdef DEBUG_ENC_REF
fprintf(stderr, " ctxt=%d\n", ctxt);
#endif
for (ci = runlen; ci < lim; ++ci) {
OPJ_UINT32 vsc;
opj_flag_t flags;
+ OPJ_UINT32 ctxt1;
flags = *flagsp;
}
if (!(flags & ((T1_SIGMA_THIS | T1_PI_THIS) << (ci * 3U)))) {
- OPJ_UINT32 ctxt1 = opj_t1_getctxno_zc(mqc, flags >> (ci * 3U));
+ ctxt1 = opj_t1_getctxno_zc(mqc, flags >> (ci * 3U));
#ifdef DEBUG_ENC_CLN
printf(" ctxt1=%d\n", ctxt1);
#endif
opj_mqc_setcurctx(mqc, ctxt1);
- v = opj_int_abs(*datap) & one ? 1 : 0;
+ v = (opj_int_abs(*datap) & one) ? 1 : 0;
opj_mqc_encode(mqc, v);
if (v) {
OPJ_UINT32 ctxt2, spb;
OPJ_UINT32 w,
OPJ_UINT32 h)
{
- size_t flagssize;
+ OPJ_UINT32 flagssize;
OPJ_UINT32 flags_stride;
+ /* No risk of overflow. Prior checks ensure those assert are met */
+ /* They are per the specification */
+ assert(w <= 1024);
+ assert(h <= 1024);
+ assert(w * h <= 4096);
+
/* encoder uses tile buffer, so no need to allocate */
if (!t1->encoder) {
- size_t datasize;
-
-#if (SIZE_MAX / 0xFFFFFFFFU) < 0xFFFFFFFFU /* UINT32_MAX */
- /* Overflow check */
- if ((w > 0U) && ((size_t)h > (SIZE_MAX / (size_t)w))) {
- /* FIXME event manager error callback */
- return OPJ_FALSE;
- }
-#endif
- datasize = (size_t)w * h;
+ OPJ_UINT32 datasize = w * h;
- /* Overflow check */
- if (datasize > (SIZE_MAX / sizeof(OPJ_INT32))) {
- /* FIXME event manager error callback */
- return OPJ_FALSE;
- }
-
- if (datasize > (size_t)t1->datasize) {
+ if (datasize > t1->datasize) {
opj_aligned_free(t1->data);
t1->data = (OPJ_INT32*) opj_aligned_malloc(datasize * sizeof(OPJ_INT32));
if (!t1->data) {
/* FIXME event manager error callback */
return OPJ_FALSE;
}
-#if SIZE_MAX > 0xFFFFFFFFU /* UINT32_MAX */
- /* TODO remove this if t1->datasize type changes to size_t */
- /* Overflow check */
- if (datasize > (size_t)0xFFFFFFFFU /* UINT32_MAX */) {
- /* FIXME event manager error callback */
- return OPJ_FALSE;
- }
-#endif
- t1->datasize = (OPJ_UINT32)datasize;
+ t1->datasize = datasize;
}
/* memset first arg is declared to never be null by gcc */
if (t1->data != NULL) {
}
}
- /* Overflow check */
- if (w > (0xFFFFFFFFU /* UINT32_MAX */ - 2U)) {
- /* FIXME event manager error callback */
- return OPJ_FALSE;
- }
flags_stride = w + 2U; /* can't be 0U */
-#if (SIZE_MAX - 3U) < 0xFFFFFFFFU /* UINT32_MAX */
- /* Overflow check */
- if (h > (0xFFFFFFFFU /* UINT32_MAX */ - 3U)) {
- /* FIXME event manager error callback */
- return OPJ_FALSE;
- }
-#endif
flagssize = (h + 3U) / 4U + 2U;
- /* Overflow check */
- if (flagssize > (SIZE_MAX / (size_t)flags_stride)) {
- /* FIXME event manager error callback */
- return OPJ_FALSE;
- }
- flagssize *= (size_t)flags_stride;
+ flagssize *= flags_stride;
{
- /* BIG FAT XXX */
opj_flag_t* p;
OPJ_UINT32 x;
OPJ_UINT32 flags_height = (h + 3U) / 4U;
- if (flagssize > (size_t)t1->flagssize) {
- /* Overflow check */
- if (flagssize > (SIZE_MAX / sizeof(opj_flag_t))) {
- /* FIXME event manager error callback */
- return OPJ_FALSE;
- }
+ if (flagssize > t1->flagssize) {
+
opj_aligned_free(t1->flags);
t1->flags = (opj_flag_t*) opj_aligned_malloc(flagssize * sizeof(
opj_flag_t));
/* FIXME event manager error callback */
return OPJ_FALSE;
}
-#if SIZE_MAX > 0xFFFFFFFFU /* UINT32_MAX */
- /* TODO remove this if t1->flagssize type changes to size_t */
- /* Overflow check */
- if (flagssize > (size_t)0xFFFFFFFFU /* UINT32_MAX */) {
- /* FIXME event manager error callback */
- return OPJ_FALSE;
- }
-#endif
}
- t1->flagssize = (OPJ_UINT32)flagssize;
+ t1->flagssize = flagssize;
memset(t1->flags, 0, flagssize * sizeof(opj_flag_t));
}
typedef struct {
+ OPJ_BOOL whole_tile_decoding;
OPJ_UINT32 resno;
opj_tcd_cblk_dec_t* cblk;
opj_tcd_band_t* band;
OPJ_UINT32 tile_w;
job = (opj_t1_cblk_decode_processing_job_t*) user_data;
- resno = job->resno;
+
cblk = job->cblk;
+
+ if (!job->whole_tile_decoding) {
+ cblk_w = (OPJ_UINT32)(cblk->x1 - cblk->x0);
+ cblk_h = (OPJ_UINT32)(cblk->y1 - cblk->y0);
+
+ cblk->decoded_data = (OPJ_INT32*)opj_aligned_malloc(sizeof(OPJ_INT32) *
+ cblk_w * cblk_h);
+ if (cblk->decoded_data == NULL) {
+ if (job->p_manager_mutex) {
+ opj_mutex_lock(job->p_manager_mutex);
+ }
+ opj_event_msg(job->p_manager, EVT_ERROR,
+ "Cannot allocate cblk->decoded_data\n");
+ if (job->p_manager_mutex) {
+ opj_mutex_unlock(job->p_manager_mutex);
+ }
+ *(job->pret) = OPJ_FALSE;
+ opj_free(job);
+ return;
+ }
+ /* Zero-init required */
+ memset(cblk->decoded_data, 0, sizeof(OPJ_INT32) * cblk_w * cblk_h);
+ } else if (cblk->decoded_data) {
+ /* Not sure if that code path can happen, but better be */
+ /* safe than sorry */
+ opj_aligned_free(cblk->decoded_data);
+ cblk->decoded_data = NULL;
+ }
+
+ resno = job->resno;
band = job->band;
tilec = job->tilec;
tccp = job->tccp;
- tile_w = (OPJ_UINT32)(tilec->x1 - tilec->x0);
+ tile_w = (OPJ_UINT32)(tilec->resolutions[tilec->minimum_num_resolutions - 1].x1
+ -
+ tilec->resolutions[tilec->minimum_num_resolutions - 1].x0);
if (!*(job->pret)) {
opj_free(job);
y += pres->y1 - pres->y0;
}
- datap = t1->data;
+ datap = cblk->decoded_data ? cblk->decoded_data : t1->data;
cblk_w = t1->w;
cblk_h = t1->h;
}
}
}
- if (tccp->qmfbid == 1) {
- OPJ_INT32* OPJ_RESTRICT tiledp = &tilec->data[(OPJ_UINT32)y * tile_w +
- (OPJ_UINT32)x];
+
+ /* Both can be non NULL if for example decoding a full tile and then */
+ /* partially a tile. In which case partial decoding should be the */
+ /* priority */
+ assert((cblk->decoded_data != NULL) || (tilec->data != NULL));
+
+ if (cblk->decoded_data) {
+ OPJ_UINT32 cblk_size = cblk_w * cblk_h;
+ if (tccp->qmfbid == 1) {
+ for (i = 0; i < cblk_size; ++i) {
+ datap[i] /= 2;
+ }
+ } else { /* if (tccp->qmfbid == 0) */
+ i = 0;
+#ifdef __SSE2__
+ {
+ const __m128 xmm_stepsize = _mm_set1_ps(band->stepsize);
+ for (; i < (cblk_size & ~15U); i += 16) {
+ __m128 xmm0_data = _mm_cvtepi32_ps(_mm_load_si128((__m128i * const)(
+ datap + 0)));
+ __m128 xmm1_data = _mm_cvtepi32_ps(_mm_load_si128((__m128i * const)(
+ datap + 4)));
+ __m128 xmm2_data = _mm_cvtepi32_ps(_mm_load_si128((__m128i * const)(
+ datap + 8)));
+ __m128 xmm3_data = _mm_cvtepi32_ps(_mm_load_si128((__m128i * const)(
+ datap + 12)));
+ _mm_store_ps((float*)(datap + 0), _mm_mul_ps(xmm0_data, xmm_stepsize));
+ _mm_store_ps((float*)(datap + 4), _mm_mul_ps(xmm1_data, xmm_stepsize));
+ _mm_store_ps((float*)(datap + 8), _mm_mul_ps(xmm2_data, xmm_stepsize));
+ _mm_store_ps((float*)(datap + 12), _mm_mul_ps(xmm3_data, xmm_stepsize));
+ datap += 16;
+ }
+ }
+#endif
+ for (; i < cblk_size; ++i) {
+ OPJ_FLOAT32 tmp = ((OPJ_FLOAT32)(*datap)) * band->stepsize;
+ memcpy(datap, &tmp, sizeof(tmp));
+ datap++;
+ }
+ }
+ } else if (tccp->qmfbid == 1) {
+ OPJ_INT32* OPJ_RESTRICT tiledp = &tilec->data[(OPJ_SIZE_T)y * tile_w +
+ (OPJ_SIZE_T)x];
for (j = 0; j < cblk_h; ++j) {
i = 0;
for (; i < (cblk_w & ~(OPJ_UINT32)3U); i += 4U) {
OPJ_INT32 tmp1 = datap[(j * cblk_w) + i + 1U];
OPJ_INT32 tmp2 = datap[(j * cblk_w) + i + 2U];
OPJ_INT32 tmp3 = datap[(j * cblk_w) + i + 3U];
- ((OPJ_INT32*)tiledp)[(j * tile_w) + i + 0U] = tmp0 / 2;
- ((OPJ_INT32*)tiledp)[(j * tile_w) + i + 1U] = tmp1 / 2;
- ((OPJ_INT32*)tiledp)[(j * tile_w) + i + 2U] = tmp2 / 2;
- ((OPJ_INT32*)tiledp)[(j * tile_w) + i + 3U] = tmp3 / 2;
+ ((OPJ_INT32*)tiledp)[(j * (OPJ_SIZE_T)tile_w) + i + 0U] = tmp0 / 2;
+ ((OPJ_INT32*)tiledp)[(j * (OPJ_SIZE_T)tile_w) + i + 1U] = tmp1 / 2;
+ ((OPJ_INT32*)tiledp)[(j * (OPJ_SIZE_T)tile_w) + i + 2U] = tmp2 / 2;
+ ((OPJ_INT32*)tiledp)[(j * (OPJ_SIZE_T)tile_w) + i + 3U] = tmp3 / 2;
}
for (; i < cblk_w; ++i) {
OPJ_INT32 tmp = datap[(j * cblk_w) + i];
- ((OPJ_INT32*)tiledp)[(j * tile_w) + i] = tmp / 2;
+ ((OPJ_INT32*)tiledp)[(j * (OPJ_SIZE_T)tile_w) + i] = tmp / 2;
}
}
} else { /* if (tccp->qmfbid == 0) */
- OPJ_FLOAT32* OPJ_RESTRICT tiledp = (OPJ_FLOAT32*) &tilec->data[(OPJ_UINT32)y *
- tile_w + (OPJ_UINT32)x];
+ OPJ_FLOAT32* OPJ_RESTRICT tiledp = (OPJ_FLOAT32*) &tilec->data[(OPJ_SIZE_T)y *
+ tile_w + (OPJ_SIZE_T)x];
for (j = 0; j < cblk_h; ++j) {
OPJ_FLOAT32* OPJ_RESTRICT tiledp2 = tiledp;
for (i = 0; i < cblk_w; ++i) {
}
-void opj_t1_decode_cblks(opj_thread_pool_t* tp,
+void opj_t1_decode_cblks(opj_tcd_t* tcd,
volatile OPJ_BOOL* pret,
opj_tcd_tilecomp_t* tilec,
opj_tccp_t* tccp,
OPJ_BOOL check_pterm
)
{
+ opj_thread_pool_t* tp = tcd->thread_pool;
OPJ_UINT32 resno, bandno, precno, cblkno;
+#ifdef DEBUG_VERBOSE
+ OPJ_UINT32 codeblocks_decoded = 0;
+ printf("Enter opj_t1_decode_cblks()\n");
+#endif
+
for (resno = 0; resno < tilec->minimum_num_resolutions; ++resno) {
opj_tcd_resolution_t* res = &tilec->resolutions[resno];
for (precno = 0; precno < res->pw * res->ph; ++precno) {
opj_tcd_precinct_t* precinct = &band->precincts[precno];
+ if (!opj_tcd_is_subband_area_of_interest(tcd,
+ tilec->compno,
+ resno,
+ band->bandno,
+ (OPJ_UINT32)precinct->x0,
+ (OPJ_UINT32)precinct->y0,
+ (OPJ_UINT32)precinct->x1,
+ (OPJ_UINT32)precinct->y1)) {
+ for (cblkno = 0; cblkno < precinct->cw * precinct->ch; ++cblkno) {
+ opj_tcd_cblk_dec_t* cblk = &precinct->cblks.dec[cblkno];
+ if (cblk->decoded_data) {
+#ifdef DEBUG_VERBOSE
+ printf("Discarding codeblock %d,%d at resno=%d, bandno=%d\n",
+ cblk->x0, cblk->y0, resno, bandno);
+#endif
+ opj_aligned_free(cblk->decoded_data);
+ cblk->decoded_data = NULL;
+ }
+ }
+ continue;
+ }
+
for (cblkno = 0; cblkno < precinct->cw * precinct->ch; ++cblkno) {
opj_tcd_cblk_dec_t* cblk = &precinct->cblks.dec[cblkno];
opj_t1_cblk_decode_processing_job_t* job;
+ if (!opj_tcd_is_subband_area_of_interest(tcd,
+ tilec->compno,
+ resno,
+ band->bandno,
+ (OPJ_UINT32)cblk->x0,
+ (OPJ_UINT32)cblk->y0,
+ (OPJ_UINT32)cblk->x1,
+ (OPJ_UINT32)cblk->y1)) {
+ if (cblk->decoded_data) {
+#ifdef DEBUG_VERBOSE
+ printf("Discarding codeblock %d,%d at resno=%d, bandno=%d\n",
+ cblk->x0, cblk->y0, resno, bandno);
+#endif
+ opj_aligned_free(cblk->decoded_data);
+ cblk->decoded_data = NULL;
+ }
+ continue;
+ }
+
+ if (!tcd->whole_tile_decoding) {
+ OPJ_UINT32 cblk_w = (OPJ_UINT32)(cblk->x1 - cblk->x0);
+ OPJ_UINT32 cblk_h = (OPJ_UINT32)(cblk->y1 - cblk->y0);
+ if (cblk->decoded_data != NULL) {
+#ifdef DEBUG_VERBOSE
+ printf("Reusing codeblock %d,%d at resno=%d, bandno=%d\n",
+ cblk->x0, cblk->y0, resno, bandno);
+#endif
+ continue;
+ }
+ if (cblk_w == 0 || cblk_h == 0) {
+ continue;
+ }
+#ifdef DEBUG_VERBOSE
+ printf("Decoding codeblock %d,%d at resno=%d, bandno=%d\n",
+ cblk->x0, cblk->y0, resno, bandno);
+#endif
+ }
+
job = (opj_t1_cblk_decode_processing_job_t*) opj_calloc(1,
sizeof(opj_t1_cblk_decode_processing_job_t));
if (!job) {
*pret = OPJ_FALSE;
return;
}
+ job->whole_tile_decoding = tcd->whole_tile_decoding;
job->resno = resno;
job->cblk = cblk;
job->band = band;
job->check_pterm = check_pterm;
job->mustuse_cblkdatabuffer = opj_thread_pool_get_thread_count(tp) > 1;
opj_thread_pool_submit_job(tp, opj_t1_clbl_decode_processor, job);
+#ifdef DEBUG_VERBOSE
+ codeblocks_decoded ++;
+#endif
if (!(*pret)) {
return;
}
} /* bandno */
} /* resno */
+#ifdef DEBUG_VERBOSE
+ printf("Leave opj_t1_decode_cblks(). Number decoded: %d\n", codeblocks_decoded);
+#endif
return;
}
OPJ_BYTE* cblkdata = NULL;
OPJ_UINT32 cblkdataindex = 0;
OPJ_BYTE type = T1_TYPE_MQ; /* BYPASS mode */
+ OPJ_INT32* original_t1_data = NULL;
mqc->lut_ctxno_zc_orient = lut_ctxno_zc + (orient << 9);
}
} else if (cblk->numchunks == 1) {
cblkdata = cblk->chunks[0].data;
+ } else {
+ /* Not sure if that can happen in practice, but avoid Coverity to */
+ /* think we will dereference a null cblkdta pointer */
+ return OPJ_TRUE;
+ }
+
+ /* For subtile decoding, directly decode in the decoded_data buffer of */
+ /* the code-block. Hack t1->data to point to it, and restore it later */
+ if (cblk->decoded_data) {
+ original_t1_data = t1->data;
+ t1->data = cblk->decoded_data;
}
for (segno = 0; segno < cblk->real_num_segs; ++segno) {
}
}
+ /* Restore original t1->data is needed */
+ if (cblk->decoded_data) {
+ t1->data = original_t1_data;
+ }
+
return OPJ_TRUE;
}
OPJ_INT32* OPJ_RESTRICT tiledp;
OPJ_UINT32 cblk_w;
OPJ_UINT32 cblk_h;
- OPJ_UINT32 i, j, tileIndex = 0, tileLineAdvance;
+ OPJ_UINT32 i, j, tileLineAdvance;
+ OPJ_SIZE_T tileIndex = 0;
OPJ_INT32 x = cblk->x0 - band->x0;
OPJ_INT32 y = cblk->y0 - band->y0;
cblk_h = t1->h;
tileLineAdvance = tile_w - cblk_w;
- tiledp = &tilec->data[(OPJ_UINT32)y * tile_w + (OPJ_UINT32)x];
+ tiledp = &tilec->data[(OPJ_SIZE_T)y * tile_w + (OPJ_SIZE_T)x];
t1->data = tiledp;
t1->data_stride = tile_w;
if (tccp->qmfbid == 1) {
+ /* Do multiplication on unsigned type, even if the
+ * underlying type is signed, to avoid potential
+ * int overflow on large value (the output will be
+ * incorrect in such situation, but whatever...)
+ * This assumes complement-to-2 signed integer
+ * representation
+ * Fixes https://github.com/uclouvain/openjpeg/issues/1053
+ */
+ OPJ_UINT32* OPJ_RESTRICT tiledp_u = (OPJ_UINT32*) tiledp;
for (j = 0; j < cblk_h; ++j) {
for (i = 0; i < cblk_w; ++i) {
- tiledp[tileIndex] *= (1 << T1_NMSEDEC_FRACBITS);
+ tiledp_u[tileIndex] <<= T1_NMSEDEC_FRACBITS;
tileIndex++;
}
tileIndex += tileLineAdvance;
cblk->numbps = max ? (OPJ_UINT32)((opj_int_floorlog2(max) + 1) -
T1_NMSEDEC_FRACBITS) : 0;
+ if (cblk->numbps == 0) {
+ cblk->totalpasses = 0;
+ return;
+ }
bpno = (OPJ_INT32)(cblk->numbps - 1);
passtype = 2;
projects / openjpeg.git / blobdiff