walkero
/
openjpeg
1
0
Fork 0
Code Issues Pull Requests Projects Releases 1 Wiki Activity

Change 'restrict' define to 'OPJ_RESTRICT' (#816) 

Visual Studio 2015 does not pass regression tests with `__restrict` so kept disabled for MSVC.
Need to check proper usage of OPJ_RESTRICT (if correct then there’s
probably a bug  in vc14)

Closes #661
This commit is contained in:
Matthieu Darbois 2016-09-06 00:49:53 +02:00 committed by GitHub
parent 51155950eb
commit 9f24b078c7
6 changed files with 63 additions and 54 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

34
src/lib/openjp2/dwt.c
View File

@ -129,16 +129,16 @@ static OPJ_BOOL opj_dwt_decode_tile(opj_tcd_tilecomp_t* tilec, OPJ_UINT32 i, DWT
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* 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* restrict dwt);
static void opj_v4dwt_decode(opj_v4dwt_t* OPJ_RESTRICT dwt);
static void opj_v4dwt_interleave_h(opj_v4dwt_t* restrict w, OPJ_FLOAT32* 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* restrict v , OPJ_FLOAT32* 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);
@ -543,7 +543,7 @@ void opj_dwt_calc_explicit_stepsizes(opj_tccp_t * tccp, OPJ_UINT32 prec) {
/* <summary> */
/* Determine maximum computed resolution level for inverse wavelet transform */
/* </summary> */
static OPJ_UINT32 opj_dwt_max_resolution(opj_tcd_resolution_t* restrict r, OPJ_UINT32 i) {
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 ) {
@ -582,7 +582,7 @@ static OPJ_BOOL opj_dwt_decode_tile(opj_tcd_tilecomp_t* tilec, OPJ_UINT32 numres
v.mem = h.mem;
while( --numres) {
OPJ_INT32 * restrict tiledp = tilec->data;
OPJ_INT32 * OPJ_RESTRICT tiledp = tilec->data;
OPJ_UINT32 j;
++tr;
@ -617,8 +617,8 @@ static OPJ_BOOL opj_dwt_decode_tile(opj_tcd_tilecomp_t* tilec, OPJ_UINT32 numres
return OPJ_TRUE;
}
static void opj_v4dwt_interleave_h(opj_v4dwt_t* restrict w, OPJ_FLOAT32* restrict a, OPJ_INT32 x, OPJ_INT32 size){
OPJ_FLOAT32* restrict bi = (OPJ_FLOAT32*) (w->wavelet + w->cas);
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;
@ -660,8 +660,8 @@ static void opj_v4dwt_interleave_h(opj_v4dwt_t* restrict w, OPJ_FLOAT32* restric
}
}
static void opj_v4dwt_interleave_v(opj_v4dwt_t* restrict v , OPJ_FLOAT32* restrict a , OPJ_INT32 x, OPJ_INT32 nb_elts_read){
opj_v4_t* restrict bi = v->wavelet + v->cas;
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){
@ -679,7 +679,7 @@ static void opj_v4dwt_interleave_v(opj_v4dwt_t* restrict v , OPJ_FLOAT32* restri
#ifdef __SSE__
static void opj_v4dwt_decode_step1_sse(opj_v4_t* w, OPJ_INT32 count, const __m128 c){
__m128* restrict vw = (__m128*) w;
__m128* OPJ_RESTRICT vw = (__m128*) w;
OPJ_INT32 i;
/* 4x unrolled loop */
for(i = 0; i < count >> 2; ++i){
@ -700,8 +700,8 @@ static void opj_v4dwt_decode_step1_sse(opj_v4_t* w, OPJ_INT32 count, const __m12
}
void opj_v4dwt_decode_step2_sse(opj_v4_t* l, opj_v4_t* w, OPJ_INT32 k, OPJ_INT32 m, __m128 c){
__m128* restrict vl = (__m128*) l;
__m128* restrict vw = (__m128*) w;
__m128* OPJ_RESTRICT vl = (__m128*) l;
__m128* OPJ_RESTRICT vw = (__m128*) w;
OPJ_INT32 i;
__m128 tmp1, tmp2, tmp3;
tmp1 = vl[0];
@ -729,7 +729,7 @@ void opj_v4dwt_decode_step2_sse(opj_v4_t* l, opj_v4_t* w, OPJ_INT32 k, OPJ_INT32
static void opj_v4dwt_decode_step1(opj_v4_t* w, OPJ_INT32 count, const OPJ_FLOAT32 c)
{
OPJ_FLOAT32* restrict fw = (OPJ_FLOAT32*) w;
OPJ_FLOAT32* OPJ_RESTRICT fw = (OPJ_FLOAT32*) w;
OPJ_INT32 i;
for(i = 0; i < count; ++i){
OPJ_FLOAT32 tmp1 = fw[i*8 ];
@ -797,7 +797,7 @@ static void opj_v4dwt_decode_step2(opj_v4_t* l, opj_v4_t* w, OPJ_INT32 k, OPJ_IN
/* <summary> */
/* Inverse 9-7 wavelet transform in 1-D. */
/* </summary> */
static void opj_v4dwt_decode(opj_v4dwt_t* restrict dwt)
static void opj_v4dwt_decode(opj_v4dwt_t* OPJ_RESTRICT dwt)
{
OPJ_INT32 a, b;
if(dwt->cas == 0) {
@ -834,7 +834,7 @@ static void opj_v4dwt_decode(opj_v4dwt_t* restrict dwt)
/* <summary> */
/* Inverse 9-7 wavelet transform in 2-D. */
/* </summary> */
OPJ_BOOL opj_dwt_decode_real(opj_tcd_tilecomp_t* 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;
@ -854,7 +854,7 @@ OPJ_BOOL opj_dwt_decode_real(opj_tcd_tilecomp_t* restrict tilec, OPJ_UINT32 numr
v.wavelet = h.wavelet;
while( --numres) {
OPJ_FLOAT32 * restrict aj = (OPJ_FLOAT32*) tilec->data;
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;

2
src/lib/openjp2/dwt.h
View File

@ -93,7 +93,7 @@ Apply an irreversible inverse DWT transform to a component of an image.
@param tilec Tile component information (current tile)
@param numres Number of resolution levels to decode
*/
OPJ_BOOL opj_dwt_decode_real(opj_tcd_tilecomp_t* restrict tilec, OPJ_UINT32 numres);
OPJ_BOOL opj_dwt_decode_real(opj_tcd_tilecomp_t* OPJ_RESTRICT tilec, OPJ_UINT32 numres);
/**
Get the gain of a subband for the irreversible 9-7 DWT.

42
src/lib/openjp2/mct.c
View File

@ -74,9 +74,9 @@ const OPJ_FLOAT64 * opj_mct_get_mct_norms_real ()
/* </summary> */
#ifdef __SSE2__
void opj_mct_encode(
OPJ_INT32* restrict c0,
OPJ_INT32* restrict c1,
OPJ_INT32* restrict c2,
OPJ_INT32* OPJ_RESTRICT c0,
OPJ_INT32* OPJ_RESTRICT c1,
OPJ_INT32* OPJ_RESTRICT c2,
OPJ_UINT32 n)
{
OPJ_SIZE_T i;
@ -116,9 +116,9 @@ void opj_mct_encode(
}
#else
void opj_mct_encode(
OPJ_INT32* restrict c0,
OPJ_INT32* restrict c1,
OPJ_INT32* restrict c2,
OPJ_INT32* OPJ_RESTRICT c0,
OPJ_INT32* OPJ_RESTRICT c1,
OPJ_INT32* OPJ_RESTRICT c2,
OPJ_UINT32 n)
{
OPJ_SIZE_T i;
@ -143,9 +143,9 @@ void opj_mct_encode(
/* </summary> */
#ifdef __SSE2__
void opj_mct_decode(
OPJ_INT32* restrict c0,
OPJ_INT32* restrict c1,
OPJ_INT32* restrict c2,
OPJ_INT32* OPJ_RESTRICT c0,
OPJ_INT32* OPJ_RESTRICT c1,
OPJ_INT32* OPJ_RESTRICT c2,
OPJ_UINT32 n)
{
OPJ_SIZE_T i;
@ -178,9 +178,9 @@ void opj_mct_decode(
}
#else
void opj_mct_decode(
OPJ_INT32* restrict c0,
OPJ_INT32* restrict c1,
OPJ_INT32* restrict c2,
OPJ_INT32* OPJ_RESTRICT c0,
OPJ_INT32* OPJ_RESTRICT c1,
OPJ_INT32* OPJ_RESTRICT c2,
OPJ_UINT32 n)
{
OPJ_UINT32 i;
@ -210,9 +210,9 @@ OPJ_FLOAT64 opj_mct_getnorm(OPJ_UINT32 compno) {
/* </summary> */
#ifdef __SSE4_1__
void opj_mct_encode_real(
OPJ_INT32* restrict c0,
OPJ_INT32* restrict c1,
OPJ_INT32* restrict c2,
OPJ_INT32* OPJ_RESTRICT c0,
OPJ_INT32* OPJ_RESTRICT c1,
OPJ_INT32* OPJ_RESTRICT c2,
OPJ_UINT32 n)
{
OPJ_SIZE_T i;
@ -351,9 +351,9 @@ void opj_mct_encode_real(
}
#else
void opj_mct_encode_real(
OPJ_INT32* restrict c0,
OPJ_INT32* restrict c1,
OPJ_INT32* restrict c2,
OPJ_INT32* OPJ_RESTRICT c0,
OPJ_INT32* OPJ_RESTRICT c1,
OPJ_INT32* OPJ_RESTRICT c2,
OPJ_UINT32 n)
{
OPJ_UINT32 i;
@ -375,9 +375,9 @@ void opj_mct_encode_real(
/* Inverse irreversible MCT. */
/* </summary> */
void opj_mct_decode_real(
OPJ_FLOAT32* restrict c0,
OPJ_FLOAT32* restrict c1,
OPJ_FLOAT32* restrict c2,
OPJ_FLOAT32* OPJ_RESTRICT c0,
OPJ_FLOAT32* OPJ_RESTRICT c1,
OPJ_FLOAT32* OPJ_RESTRICT c2,
OPJ_UINT32 n)
{
OPJ_UINT32 i;

8
src/lib/openjp2/mct.h
View File

@ -60,7 +60,7 @@ Apply a reversible multi-component transform to an image
@param c2 Samples blue component
@param n Number of samples for each component
*/
void opj_mct_encode(OPJ_INT32 *c0, OPJ_INT32 *c1, OPJ_INT32 *c2, OPJ_UINT32 n);
void opj_mct_encode(OPJ_INT32* OPJ_RESTRICT c0, OPJ_INT32* OPJ_RESTRICT c1, OPJ_INT32* OPJ_RESTRICT c2, OPJ_UINT32 n);
/**
Apply a reversible multi-component inverse transform to an image
@param c0 Samples for luminance component
@ -68,7 +68,7 @@ Apply a reversible multi-component inverse transform to an image
@param c2 Samples for blue chrominance component
@param n Number of samples for each component
*/
void opj_mct_decode(OPJ_INT32 *c0, OPJ_INT32 *c1, OPJ_INT32 *c2, OPJ_UINT32 n);
void opj_mct_decode(OPJ_INT32* OPJ_RESTRICT c0, OPJ_INT32* OPJ_RESTRICT c1, OPJ_INT32* OPJ_RESTRICT c2, OPJ_UINT32 n);
/**
Get norm of the basis function used for the reversible multi-component transform
@param compno Number of the component (0->Y, 1->U, 2->V)
@ -83,7 +83,7 @@ Apply an irreversible multi-component transform to an image
@param c2 Samples blue component
@param n Number of samples for each component
*/
void opj_mct_encode_real(OPJ_INT32 *c0, OPJ_INT32 *c1, OPJ_INT32 *c2, OPJ_UINT32 n);
void opj_mct_encode_real(OPJ_INT32* OPJ_RESTRICT c0, OPJ_INT32* OPJ_RESTRICT c1, OPJ_INT32* OPJ_RESTRICT c2, OPJ_UINT32 n);
/**
Apply an irreversible multi-component inverse transform to an image
@param c0 Samples for luminance component
@ -91,7 +91,7 @@ Apply an irreversible multi-component inverse transform to an image
@param c2 Samples for blue chrominance component
@param n Number of samples for each component
*/
void opj_mct_decode_real(OPJ_FLOAT32* c0, OPJ_FLOAT32* c1, OPJ_FLOAT32* c2, OPJ_UINT32 n);
void opj_mct_decode_real(OPJ_FLOAT32* OPJ_RESTRICT c0, OPJ_FLOAT32* OPJ_RESTRICT c1, OPJ_FLOAT32* OPJ_RESTRICT c2, OPJ_UINT32 n);
/**
Get norm of the basis function used for the irreversible multi-component transform
@param compno Number of the component (0->Y, 1->U, 2->V)

19
src/lib/openjp2/opj_includes.h
View File

@ -103,12 +103,21 @@
*/
/* Are restricted pointers available? (C99) */
#if (__STDC_VERSION__ != 199901L)
/* Not a C99 compiler */
#ifdef __GNUC__
#define restrict __restrict__
#if (__STDC_VERSION__ >= 199901L)
#define OPJ_RESTRICT restrict
#else
#define restrict /* restrict */
/* Not a C99 compiler */
#if defined(__GNUC__)
#define OPJ_RESTRICT __restrict__
/*
vc14 (2015) outputs wrong results.
Need to check OPJ_RESTRICT usage (or a bug in vc14)
#elif defined(_MSC_VER) && (_MSC_VER >= 1400)
#define OPJ_RESTRICT __restrict
*/
#else
#define OPJ_RESTRICT /* restrict */
#endif
#endif

14
src/lib/openjp2/t1.c
View File

@ -1283,14 +1283,14 @@ OPJ_BOOL opj_t1_decode_cblks( opj_t1_t* t1,
opj_tcd_resolution_t* res = &tilec->resolutions[resno];
for (bandno = 0; bandno < res->numbands; ++bandno) {
opj_tcd_band_t* restrict band = &res->bands[bandno];
opj_tcd_band_t* OPJ_RESTRICT band = &res->bands[bandno];
for (precno = 0; precno < res->pw * res->ph; ++precno) {
opj_tcd_precinct_t* precinct = &band->precincts[precno];
for (cblkno = 0; cblkno < precinct->cw * precinct->ch; ++cblkno) {
opj_tcd_cblk_dec_t* cblk = &precinct->cblks.dec[cblkno];
OPJ_INT32* restrict datap;
OPJ_INT32* OPJ_RESTRICT datap;
OPJ_UINT32 cblk_w, cblk_h;
OPJ_INT32 x, y;
OPJ_UINT32 i, j;
@ -1333,7 +1333,7 @@ OPJ_BOOL opj_t1_decode_cblks( opj_t1_t* t1,
}
}
if (tccp->qmfbid == 1) {
OPJ_INT32* restrict tiledp = &tilec->data[(OPJ_UINT32)y * tile_w + (OPJ_UINT32)x];
OPJ_INT32* OPJ_RESTRICT tiledp = &tilec->data[(OPJ_UINT32)y * tile_w + (OPJ_UINT32)x];
for (j = 0; j < cblk_h; ++j) {
for (i = 0; i < cblk_w; ++i) {
OPJ_INT32 tmp = datap[(j * cblk_w) + i];
@ -1341,9 +1341,9 @@ OPJ_BOOL opj_t1_decode_cblks( opj_t1_t* t1,
}
}
} else { /* if (tccp->qmfbid == 0) */
OPJ_FLOAT32* restrict tiledp = (OPJ_FLOAT32*) &tilec->data[(OPJ_UINT32)y * tile_w + (OPJ_UINT32)x];
OPJ_FLOAT32* OPJ_RESTRICT tiledp = (OPJ_FLOAT32*) &tilec->data[(OPJ_UINT32)y * tile_w + (OPJ_UINT32)x];
for (j = 0; j < cblk_h; ++j) {
OPJ_FLOAT32* restrict tiledp2 = tiledp;
OPJ_FLOAT32* OPJ_RESTRICT tiledp2 = tiledp;
for (i = 0; i < cblk_w; ++i) {
OPJ_FLOAT32 tmp = (OPJ_FLOAT32)*datap * band->stepsize;
*tiledp2 = tmp;
@ -1475,7 +1475,7 @@ OPJ_BOOL opj_t1_encode_cblks( opj_t1_t *t1,
opj_tcd_resolution_t *res = &tilec->resolutions[resno];
for (bandno = 0; bandno < res->numbands; ++bandno) {
opj_tcd_band_t* restrict band = &res->bands[bandno];
opj_tcd_band_t* OPJ_RESTRICT band = &res->bands[bandno];
OPJ_INT32 bandconst = 8192 * 8192 / ((OPJ_INT32) floor(band->stepsize * 8192));
for (precno = 0; precno < res->pw * res->ph; ++precno) {
@ -1483,7 +1483,7 @@ OPJ_BOOL opj_t1_encode_cblks( opj_t1_t *t1,
for (cblkno = 0; cblkno < prc->cw * prc->ch; ++cblkno) {
opj_tcd_cblk_enc_t* cblk = &prc->cblks.enc[cblkno];
OPJ_INT32* restrict tiledp;
OPJ_INT32* OPJ_RESTRICT tiledp;
OPJ_UINT32 cblk_w;
OPJ_UINT32 cblk_h;
OPJ_UINT32 i, j, tileIndex=0, tileLineAdvance;