# Changelog

## [v2.4.0](https://github.com/uclouvain/openjpeg/releases/v2.4.0) (2020-12-28)

[Full Changelog](https://github.com/uclouvain/openjpeg/compare/v2.3.1...v2.4.0)

**Closed issues:**

- OPENJPEG\_INSTALL\_DOC\_DIR does not control a destination directory where HTML docs would be installed. [\#1309](https://github.com/uclouvain/openjpeg/issues/1309)
- Heap-buffer-overflow in lib/openjp2/pi.c:312 [\#1302](https://github.com/uclouvain/openjpeg/issues/1302)
- Heap-buffer-overflow in lib/openjp2/t2.c:973 [\#1299](https://github.com/uclouvain/openjpeg/issues/1299)
- Heap-buffer-overflow in lib/openjp2/pi.c:623 [\#1293](https://github.com/uclouvain/openjpeg/issues/1293)
- Global-buffer-overflow in lib/openjp2/dwt.c:1980 [\#1286](https://github.com/uclouvain/openjpeg/issues/1286)
- Heap-buffer-overflow in lib/openjp2/tcd.c:2417 [\#1284](https://github.com/uclouvain/openjpeg/issues/1284)
- Heap-buffer-overflow in lib/openjp2/mqc.c:499 [\#1283](https://github.com/uclouvain/openjpeg/issues/1283)
- Openjpeg could not encode 32bit RGB float image [\#1281](https://github.com/uclouvain/openjpeg/issues/1281)
- Openjpeg could not encode 32bit RGB float image [\#1280](https://github.com/uclouvain/openjpeg/issues/1280)
- ISO/IEC 15444-1:2019 \(E\) compared with 'cio.h' [\#1277](https://github.com/uclouvain/openjpeg/issues/1277)
- Test-suite failure due to hash mismatch [\#1264](https://github.com/uclouvain/openjpeg/issues/1264)
- Heap use-after-free [\#1261](https://github.com/uclouvain/openjpeg/issues/1261)
- Memory leak when failing to allocate object... [\#1259](https://github.com/uclouvain/openjpeg/issues/1259)
- Memory leak of Tier 1 handle when OpenJPEG fails to set it as TLS... [\#1257](https://github.com/uclouvain/openjpeg/issues/1257)
- Any plan to build release for CVE-2020-8112/CVE-2020-6851 [\#1247](https://github.com/uclouvain/openjpeg/issues/1247)
- failing to convert 16-bit file: opj\_t2\_encode\_packet\(\): only 5251 bytes remaining in output buffer. 5621 needed. [\#1243](https://github.com/uclouvain/openjpeg/issues/1243)
- CMake+VS2017 Compile OK, thirdparty Compile OK, but thirdparty not install [\#1239](https://github.com/uclouvain/openjpeg/issues/1239)
- New release to solve CVE-2019-6988 ? [\#1238](https://github.com/uclouvain/openjpeg/issues/1238)
- Many tests fail to pass after the update of libtiff to version 4.1.0 [\#1233](https://github.com/uclouvain/openjpeg/issues/1233)
- Another heap buffer overflow in libopenjp2 [\#1231](https://github.com/uclouvain/openjpeg/issues/1231)
- Heap buffer overflow in libopenjp2 [\#1228](https://github.com/uclouvain/openjpeg/issues/1228)
- Endianness of binary volume \(JP3D\) [\#1224](https://github.com/uclouvain/openjpeg/issues/1224)
- New release to resolve CVE-2019-12973 [\#1222](https://github.com/uclouvain/openjpeg/issues/1222)
- how to set the block size,like 128,256 ? [\#1216](https://github.com/uclouvain/openjpeg/issues/1216)
- compress YUV files to motion jpeg2000 standard [\#1213](https://github.com/uclouvain/openjpeg/issues/1213)
- Repair/update Java wrapper, and include in release [\#1208](https://github.com/uclouvain/openjpeg/issues/1208)
- abc [\#1206](https://github.com/uclouvain/openjpeg/issues/1206)
- Slow decoding [\#1202](https://github.com/uclouvain/openjpeg/issues/1202)
- Installation question [\#1201](https://github.com/uclouvain/openjpeg/issues/1201)
- Typo in test\_decode\_area - \*ptilew is assigned instead of \*ptileh [\#1195](https://github.com/uclouvain/openjpeg/issues/1195)
- Creating a J2K file with one POC is broken [\#1191](https://github.com/uclouvain/openjpeg/issues/1191)
- Make fails on Arch Linux [\#1174](https://github.com/uclouvain/openjpeg/issues/1174)
- Heap buffer overflow in opj\_t1\_clbl\_decode\_processor\(\) triggered with Ghostscript [\#1158](https://github.com/uclouvain/openjpeg/issues/1158)
- opj\_stream\_get\_number\_byte\_left: Assertion `p\_stream-\>m\_byte\_offset \>= 0' failed. [\#1151](https://github.com/uclouvain/openjpeg/issues/1151)
- The fuzzer ignores too many inputs [\#1079](https://github.com/uclouvain/openjpeg/issues/1079)
- out of bounds read [\#1068](https://github.com/uclouvain/openjpeg/issues/1068)

**Merged pull requests:**

- Change defined WIN32 [\#1310](https://github.com/uclouvain/openjpeg/pull/1310) ([Jamaika1](https://github.com/Jamaika1))
- docs: fix simple typo, producted -\> produced [\#1308](https://github.com/uclouvain/openjpeg/pull/1308) ([timgates42](https://github.com/timgates42))
- Set ${OPENJPEG\_INSTALL\_DOC\_DIR} to DESTINATION of HTMLs [\#1307](https://github.com/uclouvain/openjpeg/pull/1307) ([lemniscati](https://github.com/lemniscati))
- Use INC\_DIR for OPENJPEG\_INCLUDE\_DIRS \(fixes uclouvain\#1174\) [\#1306](https://github.com/uclouvain/openjpeg/pull/1306) ([matthew-sharp](https://github.com/matthew-sharp))
- pi.c: avoid out of bounds access with POC \(fixes \#1302\) [\#1304](https://github.com/uclouvain/openjpeg/pull/1304) ([rouault](https://github.com/rouault))
- Encoder: grow again buffer size [\#1303](https://github.com/uclouvain/openjpeg/pull/1303) ([zodf0055980](https://github.com/zodf0055980))
- opj\_j2k\_write\_sod\(\): avoid potential heap buffer overflow \(fixes \#1299\) \(probably master only\) [\#1301](https://github.com/uclouvain/openjpeg/pull/1301) ([rouault](https://github.com/rouault))
- pi.c: avoid out of bounds access with POC \(refs https://github.com/uclouvain/openjpeg/issues/1293\#issuecomment-737122836\) [\#1300](https://github.com/uclouvain/openjpeg/pull/1300) ([rouault](https://github.com/rouault))
- opj\_t2\_encode\_packet\(\): avoid out of bound access of \#1297, but likely not the proper fix [\#1298](https://github.com/uclouvain/openjpeg/pull/1298) ([rouault](https://github.com/rouault))
- opj\_t2\_encode\_packet\(\): avoid out of bound access of \#1294, but likely not the proper fix [\#1296](https://github.com/uclouvain/openjpeg/pull/1296) ([rouault](https://github.com/rouault))
- opj\_j2k\_setup\_encoder\(\): validate POC compno0 and compno1 \(fixes \#1293\) [\#1295](https://github.com/uclouvain/openjpeg/pull/1295) ([rouault](https://github.com/rouault))
- Encoder: avoid global buffer overflow on irreversible conversion when… [\#1292](https://github.com/uclouvain/openjpeg/pull/1292) ([rouault](https://github.com/rouault))
- Decoding: deal with some SPOT6 images that have tiles with a single tile-part with TPsot == 0 and TNsot == 0, and with missing EOC [\#1291](https://github.com/uclouvain/openjpeg/pull/1291) ([rouault](https://github.com/rouault))
- Free p\_tcd\_marker\_info to avoid memory leak [\#1288](https://github.com/uclouvain/openjpeg/pull/1288) ([zodf0055980](https://github.com/zodf0055980))
- Encoder: grow again buffer size [\#1287](https://github.com/uclouvain/openjpeg/pull/1287) ([zodf0055980](https://github.com/zodf0055980))
- Encoder: avoid uint32 overflow when allocating memory for codestream buffer \(fixes \#1243\) [\#1276](https://github.com/uclouvain/openjpeg/pull/1276) ([rouault](https://github.com/rouault))
- Java compatibility from 1.5 to 1.6 [\#1263](https://github.com/uclouvain/openjpeg/pull/1263) ([jiapei100](https://github.com/jiapei100))
- opj\_decompress: fix double-free on input directory with mix of valid and invalid images [\#1262](https://github.com/uclouvain/openjpeg/pull/1262) ([rouault](https://github.com/rouault))
- openjp2: Plug image leak when failing to allocate codestream index. [\#1260](https://github.com/uclouvain/openjpeg/pull/1260) ([sebras](https://github.com/sebras))
- openjp2: Plug memory leak when setting data as TLS fails. [\#1258](https://github.com/uclouvain/openjpeg/pull/1258) ([sebras](https://github.com/sebras))
- openjp2: Error out if failing to create Tier 1 handle. [\#1256](https://github.com/uclouvain/openjpeg/pull/1256) ([sebras](https://github.com/sebras))
- Testing for invalid values of width, height, numcomps [\#1254](https://github.com/uclouvain/openjpeg/pull/1254) ([szukw000](https://github.com/szukw000))
- Single-threaded performance improvements in forward DWT for 5-3 and 9-7 \(and other improvements\) [\#1253](https://github.com/uclouvain/openjpeg/pull/1253) ([rouault](https://github.com/rouault))
- Add support for multithreading in encoder [\#1248](https://github.com/uclouvain/openjpeg/pull/1248) ([rouault](https://github.com/rouault))
- Add support for generation of PLT markers in encoder [\#1246](https://github.com/uclouvain/openjpeg/pull/1246) ([rouault](https://github.com/rouault))
- Fix warnings about signed/unsigned casts in pi.c [\#1244](https://github.com/uclouvain/openjpeg/pull/1244) ([rouault](https://github.com/rouault))
- opj\_decompress: add sanity checks to avoid segfault in case of decoding error [\#1240](https://github.com/uclouvain/openjpeg/pull/1240) ([rouault](https://github.com/rouault))
- ignore wrong icc [\#1236](https://github.com/uclouvain/openjpeg/pull/1236) ([szukw000](https://github.com/szukw000))
- Implement writing of IMF profiles [\#1235](https://github.com/uclouvain/openjpeg/pull/1235) ([rouault](https://github.com/rouault))
- tests: add alternate checksums for libtiff 4.1 [\#1234](https://github.com/uclouvain/openjpeg/pull/1234) ([rouault](https://github.com/rouault))
- opj\_tcd\_init\_tile\(\): avoid integer overflow [\#1232](https://github.com/uclouvain/openjpeg/pull/1232) ([rouault](https://github.com/rouault))
- tests/fuzzers: link fuzz binaries using $LIB\_FUZZING\_ENGINE. [\#1230](https://github.com/uclouvain/openjpeg/pull/1230) ([Dor1s](https://github.com/Dor1s))
- opj\_j2k\_update\_image\_dimensions\(\): reject images whose coordinates are beyond INT\_MAX \(fixes \#1228\) [\#1229](https://github.com/uclouvain/openjpeg/pull/1229) ([rouault](https://github.com/rouault))
- Fix resource leaks [\#1226](https://github.com/uclouvain/openjpeg/pull/1226) ([dodys](https://github.com/dodys))
- abi-check.sh: fix false postive ABI error, and display output error log [\#1218](https://github.com/uclouvain/openjpeg/pull/1218) ([rouault](https://github.com/rouault))
- pi.c: avoid integer overflow, resulting in later invalid access to memory in opj\_t2\_decode\_packets\(\) [\#1217](https://github.com/uclouvain/openjpeg/pull/1217) ([rouault](https://github.com/rouault))
- Add check to validate SGcod/SPcoc/SPcod parameter values. [\#1211](https://github.com/uclouvain/openjpeg/pull/1211) ([sebras](https://github.com/sebras))
- Fix buffer overflow reading an image file less than four characters [\#1196](https://github.com/uclouvain/openjpeg/pull/1196) ([robert-ancell](https://github.com/robert-ancell))
- compression: emit POC marker when only one single POC is requested \(f… [\#1192](https://github.com/uclouvain/openjpeg/pull/1192) ([rouault](https://github.com/rouault))
- Fix several potential vulnerabilities  [\#1185](https://github.com/uclouvain/openjpeg/pull/1185) ([Young-X](https://github.com/Young-X))
- openjp2/j2k: Report error if all wanted components are not decoded. [\#1164](https://github.com/uclouvain/openjpeg/pull/1164) ([sebras](https://github.com/sebras))

## [v2.3.1](https://github.com/uclouvain/openjpeg/releases/v2.3.1) (2019-04-02)
[Full Changelog](https://github.com/uclouvain/openjpeg/compare/v2.3.0...v2.3.1)

**Closed issues:**

- v2.2.0 regression for decoding images where TNsot == 0 [\#1120](https://github.com/uclouvain/openjpeg/issues/1120)
- Int overflow in jp3d [\#1162](https://github.com/uclouvain/openjpeg/issues/1162)

# Do full dependency headers.
include_regular_expression("^.*$")

#-----------------------------------------------------------------------------
# OPENJPEG version number, useful for packaging and doxygen doc:
set(OPENJPEG_VERSION_MAJOR 2)
set(OPENJPEG_VERSION_MINOR 4)
set(OPENJPEG_VERSION_BUILD 0)
set(OPENJPEG_VERSION
  "${OPENJPEG_VERSION_MAJOR}.${OPENJPEG_VERSION_MINOR}.${OPENJPEG_VERSION_BUILD}")
set(PACKAGE_VERSION
  "${OPENJPEG_VERSION_MAJOR}.${OPENJPEG_VERSION_MINOR}.${OPENJPEG_VERSION_BUILD}")

# Because autotools does not support X.Y notation for SOVERSION, we have to use
# two numbering, one for the openjpeg version and one for openjpeg soversion

#   2.0.1 |  6
#   2.1   |  7
#   2.1.1 |  7
#   2.1.2 |  7
#   2.2.0 |  7
#   2.3.0 |  7
#   2.3.1 |  7
#   2.4.0 |  7
# above is the recommendation by the OPJ team. If you really need to override this default,
# you can specify your own OPENJPEG_SOVERSION at cmake configuration time:
# cmake -DOPENJPEG_SOVERSION:STRING=42 /path/to/openjpeg
if(NOT OPENJPEG_SOVERSION)
  set(OPENJPEG_SOVERSION 7)
endif(NOT OPENJPEG_SOVERSION)
set(OPENJPEG_LIBRARY_PROPERTIES

  find_package(Threads REQUIRED)
  if(NOT CMAKE_USE_PTHREADS_INIT)
    message(FATAL_ERROR "Only pthread are supported")
  endif()
endif()
add_subdirectory(src/lib)
option(BUILD_LUTS_GENERATOR "Build utility to generate t1_luts.h" OFF)
if(UNIX)
option(BUILD_UNIT_TESTS "Build unit tests (bench_dwt, test_sparse_array, etc..)" OFF)
endif()

#-----------------------------------------------------------------------------
# Build Applications
option(BUILD_CODEC "Build the CODEC executables" ON)
option(BUILD_MJ2 "Build the MJ2 executables." OFF)
option(BUILD_JPWL "Build the JPWL library and executables" OFF)
option(BUILD_JPIP "Build the JPIP library and executables." OFF)

# OpenJPEG NEWS

More details in the [CHANGELOG](https://github.com/uclouvain/openjpeg/blob/master/CHANGELOG.md)

## OpenJPEG 2.4.0 (December 2021)

No API/ABI break compared to v2.3.1, but additional symbols for subset of components decoding (hence the MINOR version bump).

* Encoder: add support for multithreading [\#1248](https://github.com/uclouvain/openjpeg/pull/1248)
* Encoder: add support for generation of PLT markers [\#1246](https://github.com/uclouvain/openjpeg/pull/1246)
* Encoder: single-threaded performance improvements in forward DWT for 5-3 and 9-7 (and other improvements) [\#1253](https://github.com/uclouvain/openjpeg/pull/1253)
* Encoder: support IMF profiles [\#1235](https://github.com/uclouvain/openjpeg/pull/1235)
* Many bug fixes (including security fixes)

## OpenJPEG 2.3.1 (April 2019)

No API/ABI break compared to v2.3.0

* Many bug fixes (including security fixes)

## OpenJPEG 2.3.0 (October 2017)

/* create opj_config.h for CMake */
#define OPJ_HAVE_STDINT_H 		1

/*--------------------------------------------------------------------------*/
/* OpenJPEG Versioning                                                      */

/* Version number. */
#define OPJ_VERSION_MAJOR 2
#define OPJ_VERSION_MINOR 4
#define OPJ_VERSION_BUILD 0

/* create opj_config_private.h for CMake */
#define OPJ_HAVE_INTTYPES_H 	1

#define OPJ_PACKAGE_VERSION "2.4.0"

/* Not used by openjp2*/
/*#define HAVE_MEMORY_H 1*/
/*#define HAVE_STDLIB_H 1*/
/*#define HAVE_STRINGS_H 1*/
/*#define HAVE_STRING_H 1*/
/*#define HAVE_SYS_STAT_H 1*/

version: 2.4.0.{build}
branches:
  except:
  - coverity_scan
skip_tags: false
clone_depth: 50
environment:
  matrix:

set(OPENJPEG_USE_FILE "@OPENJPEG_USE_FILE_CONFIG@")

get_filename_component(SELF_DIR "${CMAKE_CURRENT_LIST_FILE}" PATH)
if(EXISTS ${SELF_DIR}/OpenJPEGTargets.cmake)
  # This is an install tree
  include(${SELF_DIR}/OpenJPEGTargets.cmake)



  set(INC_DIR "@CMAKE_INSTALL_PREFIX@/@OPENJPEG_INSTALL_INCLUDE_DIR@")


  get_filename_component(OPENJPEG_INCLUDE_DIRS "${INC_DIR}" ABSOLUTE)

else()
  if(EXISTS ${SELF_DIR}/OpenJPEGExports.cmake)
    # This is a build tree
    set( OPENJPEG_INCLUDE_DIRS @OPENJPEG_INCLUDE_PATH@)

    include(${SELF_DIR}/OpenJPEGExports.cmake)

  add_custom_target(doc ALL
    DEPENDS ${CMAKE_BINARY_DIR}/doc/html/index.html
    COMMENT "Building doxygen documentation"
  )

  # install HTML documentation (install png files too):
  install(DIRECTORY ${CMAKE_BINARY_DIR}/doc/html
    DESTINATION ${OPENJPEG_INSTALL_DOC_DIR}
    PATTERN ".svn" EXCLUDE
  )
else()
  message(STATUS "Doxygen not found, we cannot generate the documentation")
endif()

    max_w = image->comps[0].w;
    max_h = image->comps[0].h;
    prec = (int)image->comps[0].prec;

    if (out_space == cmsSigRgbData) { /* enumCS 16 */
        unsigned int i, nr_comp = image->numcomps;

        if (nr_comp < 3) { /* GRAY or GRAYA, not RGB or RGBA */
            cmsCloseProfile(in_prof);
            return;
        }
        if (nr_comp > 4) {
            nr_comp = 4;
        }
        for (i = 1; i < nr_comp; ++i) { /* AFL test */
            if (image->comps[0].dx != image->comps[i].dx) {
                break;
            }

        new_space = OPJ_CLRSPC_SRGB;
    } else if (out_space == cmsSigGrayData) { /* enumCS 17 */
        in_type = TYPE_GRAY_8;
        out_type = TYPE_RGB_8;
        out_prof = cmsCreate_sRGBProfile();
        new_space = OPJ_CLRSPC_SRGB;
    } else if (out_space == cmsSigYCbCrData) { /* enumCS 18 */
        if (image->numcomps < 3) {
            cmsCloseProfile(in_prof);
            return;
        }
        in_type = TYPE_YCbCr_16;
        out_type = TYPE_RGB_16;
        out_prof = cmsCreate_sRGBProfile();
        new_space = OPJ_CLRSPC_SRGB;
    } else {
#ifdef DEBUG_PROFILE
        fprintf(stderr, "%s:%d: color_apply_icc_profile\n\tICC Profile has unknown "

        triple = (ncomp > 2);
        wr = (int)image->comps[0].w;
        hr = (int)image->comps[0].h;
        max = (1 << prec) - 1;
        has_alpha = (ncomp == 4 || ncomp == 2);

        red = image->comps[0].data;
        if (red == NULL) {
            fprintf(stderr,
                    "imagetopnm: planes[%d] == NULL.\n", 0);
            fprintf(stderr, "\tAborting\n");
            fclose(fdest);
            return fails;
        }

        if (triple) {
            green = image->comps[1].data;
            blue = image->comps[2].data;
            for (i = 1; i <= 2; i++) {
                if (image->comps[i].data == NULL) {
                    fprintf(stderr,
                            "imagetopnm: planes[%d] == NULL.\n", i);
                    fprintf(stderr, "\tAborting\n");
                    fclose(fdest);
                    return fails;
                }
            }
        } else {
            green = blue = NULL;
        }

        if (has_alpha) {
            const char *tt = (triple ? "RGB_ALPHA" : "GRAYSCALE_ALPHA");

    return OPJ_TRUE;
}

static OPJ_BOOL bmp_read_rle4_data(FILE* IN, OPJ_UINT8* pData,
                                   OPJ_UINT32 stride, OPJ_UINT32 width, OPJ_UINT32 height)
{
    OPJ_UINT32 x, y, written;
    OPJ_UINT8 *pix;
    const OPJ_UINT8 *beyond;

    beyond = pData + stride * height;
    pix = pData;
    x = y = written = 0U;
    while (y < height) {
        int c = getc(IN);
        if (c == EOF) {
            return OPJ_FALSE;
        }

        if (c) { /* encoded mode */
            int j, c1_int;
            OPJ_UINT8 c1;

            c1_int = getc(IN);
            if (c1_int == EOF) {
                return OPJ_FALSE;
            }
            c1 = (OPJ_UINT8)c1_int;

            for (j = 0; (j < c) && (x < width) &&
                    ((OPJ_SIZE_T)pix < (OPJ_SIZE_T)beyond); j++, x++, pix++) {
                *pix = (OPJ_UINT8)((j & 1) ? (c1 & 0x0fU) : ((c1 >> 4) & 0x0fU));
                written++;
            }
        } else { /* absolute mode */
            c = getc(IN);
            if (c == EOF) {
                return OPJ_FALSE;
            }

            if (c == 0x00) { /* EOL */
                x = 0;
                y++;
                pix = pData + y * stride;
            } else if (c == 0x01) { /* EOP */
                break;
            } else if (c == 0x02) { /* MOVE by dxdy */
                c = getc(IN);
                if (c == EOF) {
                    return OPJ_FALSE;
                }
                x += (OPJ_UINT32)c;
                c = getc(IN);
                if (c == EOF) {
                    return OPJ_FALSE;
                }
                y += (OPJ_UINT32)c;
                pix = pData + y * stride + x;
            } else { /* 03 .. 255 : absolute mode */
                int j;
                OPJ_UINT8 c1 = 0U;

                for (j = 0; (j < c) && (x < width) &&
                        ((OPJ_SIZE_T)pix < (OPJ_SIZE_T)beyond); j++, x++, pix++) {
                    if ((j & 1) == 0) {
                        int c1_int;
                        c1_int = getc(IN);
                        if (c1_int == EOF) {
                            return OPJ_FALSE;
                        }
                        c1 = (OPJ_UINT8)c1_int;
                    }
                    *pix = (OPJ_UINT8)((j & 1) ? (c1 & 0x0fU) : ((c1 >> 4) & 0x0fU));
                    written++;
                }
                if (((c & 3) == 1) || ((c & 3) == 2)) { /* skip padding byte */
                    c = getc(IN);
                    if (c == EOF) {
                        return OPJ_FALSE;
                    }
                }
            }
        }
    }  /* while(y < height) */
    if (written != width * height) {
        fprintf(stderr, "warning, image's actual size does not match advertized one\n");
        return OPJ_FALSE;
    }
    return OPJ_TRUE;
}

opj_image_t* bmptoimage(const char *filename, opj_cparameters_t *parameters)
{
    opj_image_cmptparm_t cmptparm[4];   /* maximum of 4 components */
    OPJ_UINT8 lut_R[256], lut_G[256], lut_B[256];

                             (nr_comp > 2U) ? OPJ_CLRSPC_SRGB : OPJ_CLRSPC_GRAY);
    if (image == NULL) {
        goto fin;
    }
    image->x0 = (OPJ_UINT32)params->image_offset_x0;
    image->y0 = (OPJ_UINT32)params->image_offset_y0;
    image->x1 = (OPJ_UINT32)(image->x0 + (width  - 1) * (OPJ_UINT32)
                             params->subsampling_dx + 1);
    image->y1 = (OPJ_UINT32)(image->y0 + (height - 1) * (OPJ_UINT32)
                             params->subsampling_dy + 1);

    row32s = (OPJ_INT32 *)malloc((size_t)width * nr_comp * sizeof(OPJ_INT32));
    if (row32s == NULL) {
        goto fin;
    }

    /* Set alpha channel */

    OPJ_INT32* volatile buffer32s = NULL;

    volatile int fails = 1;

    memset(&sig_bit, 0, sizeof(sig_bit));
    prec = (int)image->comps[0].prec;
    planes[0] = image->comps[0].data;
    if (planes[0] == NULL) {
        fprintf(stderr,
                "imagetopng: planes[%d] == NULL.\n", 0);
        fprintf(stderr, "\tAborting\n");
        return 1;
    }
    nr_comp = (int)image->numcomps;

    if (nr_comp > 4) {
        nr_comp = 4;
    }
    for (i = 1; i < nr_comp; ++i) {
        if (image->comps[0].dx != image->comps[i].dx) {
            break;
        }
        if (image->comps[0].dy != image->comps[i].dy) {
            break;
        }
        if (image->comps[0].prec != image->comps[i].prec) {
            break;
        }
        if (image->comps[0].sgnd != image->comps[i].sgnd) {
            break;
        }
        planes[i] = image->comps[i].data;
        if (planes[i] == NULL) {
            fprintf(stderr,
                    "imagetopng: planes[%d] == NULL.\n", i);
            fprintf(stderr, "\tAborting\n");
            return 1;
        }
    }
    if (i != nr_comp) {
        fprintf(stderr,
                "imagetopng: All components shall have the same subsampling, same bit depth, same sign.\n");
        fprintf(stderr, "\tAborting\n");
        return 1;
    }

        if (image->comps[0].prec != image->comps[i].prec) {
            break;
        }
        if (image->comps[0].sgnd != image->comps[i].sgnd) {
            break;
        }
        planes[i] = image->comps[i].data;
        if (planes[i] == NULL) {
            fprintf(stderr,
                    "imagetotif: planes[%d] == NULL.\n", i);
            fprintf(stderr, "\tAborting\n");
            return 1;
        }
    }
    if (i != numcomps) {
        fprintf(stderr,
                "imagetotif: All components shall have the same subsampling, same bit depth.\n");
        fprintf(stderr, "\tAborting\n");
        return 1;
    }

    fprintf(stdout,
            "    T<tile>=<resStart>,<compStart>,<layerEnd>,<resEnd>,<compEnd>,<progOrder>\n");
    fprintf(stdout, "      Example: -POC T1=0,0,1,5,3,CPRL/T1=5,0,1,6,3,CPRL\n");
    fprintf(stdout, "-SOP\n");
    fprintf(stdout, "    Write SOP marker before each packet.\n");
    fprintf(stdout, "-EPH\n");
    fprintf(stdout, "    Write EPH marker after each header packet.\n");
    fprintf(stdout, "-PLT\n");
    fprintf(stdout, "    Write PLT marker in tile-part header.\n");
    fprintf(stdout, "-M <key value>\n");
    fprintf(stdout, "    Mode switch.\n");
    fprintf(stdout, "    [1=BYPASS(LAZY) 2=RESET 4=RESTART(TERMALL)\n");
    fprintf(stdout, "    8=VSC 16=ERTERM(SEGTERM) 32=SEGMARK(SEGSYM)]\n");
    fprintf(stdout, "    Indicate multiple modes by adding their values.\n");
    fprintf(stdout,
            "      Example: RESTART(4) + RESET(2) + SEGMARK(32) => -M 38\n");

    fprintf(stdout, "    Digital Cinema 2K profile compliant codestream.\n");
    fprintf(stdout,
            "	Need to specify the frames per second for a 2K resolution.\n");
    fprintf(stdout, "    Only 24 or 48 fps are currently allowed.\n");
    fprintf(stdout, "-cinema4K\n");
    fprintf(stdout, "    Digital Cinema 4K profile compliant codestream.\n");
    fprintf(stdout, "	Frames per second not required. Default value is 24fps.\n");
    fprintf(stdout, "-IMF <PROFILE>[,mainlevel=X][,sublevel=Y][,framerate=FPS]\n");
    fprintf(stdout, "    Interoperable Master Format compliant codestream.\n");
    fprintf(stdout, "    <PROFILE>=2K, 4K, 8K, 2K_R, 4K_R or 8K_R.\n");
    fprintf(stdout, "    X >= 0 and X <= 11.\n");
    fprintf(stdout, "    Y >= 0 and Y <= 9.\n");
    fprintf(stdout,
            "    framerate > 0 may be specified to enhance checks and set maximum bit rate when Y > 0.\n");
    fprintf(stdout, "-jpip\n");
    fprintf(stdout, "    Write jpip codestream index box in JP2 output file.\n");
    fprintf(stdout, "    Currently supports only RPCL order.\n");
    fprintf(stdout, "-C <comment>\n");
    fprintf(stdout, "    Add <comment> in the comment marker segment.\n");
    if (opj_has_thread_support()) {
        fprintf(stdout, "  -threads <num_threads|ALL_CPUS>\n"
                "    Number of threads to use for encoding or ALL_CPUS for all available cores.\n");
    }
    /* UniPG>> */
#ifdef USE_JPWL
    fprintf(stdout, "-W <params>\n");
    fprintf(stdout, "    Adoption of JPWL (Part 11) capabilities (-W params)\n");
    fprintf(stdout,
            "    The <params> field can be written and repeated in any order:\n");
    fprintf(stdout, "    [h<tilepart><=type>,s<tilepart><=method>,a=<addr>,...\n");

}

/* ------------------------------------------------------------------------------------ */

static int parse_cmdline_encoder(int argc, char **argv,
                                 opj_cparameters_t *parameters,
                                 img_fol_t *img_fol, raw_cparameters_t *raw_cp, char *indexfilename,
                                 size_t indexfilename_size,
                                 int* pOutFramerate,
                                 OPJ_BOOL* pOutPLT,
                                 int* pOutNumThreads)
{
    OPJ_UINT32 i, j;
    int totlen, c;
    opj_option_t long_option[] = {
        {"cinema2K", REQ_ARG, NULL, 'w'},
        {"cinema4K", NO_ARG, NULL, 'y'},
        {"ImgDir", REQ_ARG, NULL, 'z'},
        {"TP", REQ_ARG, NULL, 'u'},
        {"SOP", NO_ARG, NULL, 'S'},
        {"EPH", NO_ARG, NULL, 'E'},
        {"OutFor", REQ_ARG, NULL, 'O'},
        {"POC", REQ_ARG, NULL, 'P'},
        {"ROI", REQ_ARG, NULL, 'R'},
        {"jpip", NO_ARG, NULL, 'J'},
        {"mct", REQ_ARG, NULL, 'Y'},
        {"IMF", REQ_ARG, NULL, 'Z'},
        {"PLT", NO_ARG, NULL, 'A'},
        {"threads",   REQ_ARG, NULL, 'B'}
    };

    /* parse the command line */
    const char optlist[] = "i:o:r:q:n:b:c:t:p:s:SEM:x:R:d:T:If:P:C:F:u:JY:"
#ifdef USE_JPWL
                           "W:"
#endif /* USE_JPWL */

            fprintf(stdout, "CINEMA 4K profile activated\n"
                    "Other options specified could be overridden\n");
        }
        break;

        /* ------------------------------------------------------ */

        case 'Z': {         /* IMF profile*/
            int mainlevel = 0;
            int sublevel = 0;
            int profile = 0;
            int framerate = 0;
            const char* msg =
                "Wrong value for -IMF. Should be "
                "<PROFILE>[,mainlevel=X][,sublevel=Y][,framerate=FPS] where <PROFILE> is one "
                "of 2K/4K/8K/2K_R/4K_R/8K_R.\n";
            char* comma;

            comma = strstr(opj_optarg, ",mainlevel=");
            if (comma && sscanf(comma + 1, "mainlevel=%d", &mainlevel) != 1) {
                fprintf(stderr, "%s", msg);
                return 1;
            }

            comma = strstr(opj_optarg, ",sublevel=");
            if (comma && sscanf(comma + 1, "sublevel=%d", &sublevel) != 1) {
                fprintf(stderr, "%s", msg);
                return 1;
            }

            comma = strstr(opj_optarg, ",framerate=");
            if (comma && sscanf(comma + 1, "framerate=%d", &framerate) != 1) {
                fprintf(stderr, "%s", msg);
                return 1;
            }

            comma = strchr(opj_optarg, ',');
            if (comma != NULL) {
                *comma = 0;
            }

            if (strcmp(opj_optarg, "2K") == 0) {
                profile = OPJ_PROFILE_IMF_2K;
            } else if (strcmp(opj_optarg, "4K") == 0) {
                profile = OPJ_PROFILE_IMF_4K;
            } else if (strcmp(opj_optarg, "8K") == 0) {
                profile = OPJ_PROFILE_IMF_8K;
            } else if (strcmp(opj_optarg, "2K_R") == 0) {
                profile = OPJ_PROFILE_IMF_2K_R;
            } else if (strcmp(opj_optarg, "4K_R") == 0) {
                profile = OPJ_PROFILE_IMF_4K_R;
            } else if (strcmp(opj_optarg, "8K_R") == 0) {
                profile = OPJ_PROFILE_IMF_8K_R;
            } else {
                fprintf(stderr, "%s", msg);
                return 1;
            }

            if (!(mainlevel >= 0 && mainlevel <= 15)) {
                /* Voluntarily rough validation. More fine grained done in library */
                fprintf(stderr, "Invalid mainlevel value.\n");
                return 1;
            }
            if (!(sublevel >= 0 && sublevel <= 15)) {
                /* Voluntarily rough validation. More fine grained done in library */
                fprintf(stderr, "Invalid sublevel value.\n");
                return 1;
            }
            parameters->rsiz = (OPJ_UINT16)(profile | (sublevel << 4) | mainlevel);

            fprintf(stdout, "IMF profile activated\n"
                    "Other options specified could be overridden\n");

            if (pOutFramerate) {
                *pOutFramerate = framerate;
            }
            if (framerate > 0 && sublevel > 0 && sublevel <= 9) {
                const int limitMBitsSec[] = {
                    0,
                    OPJ_IMF_SUBLEVEL_1_MBITSSEC,
                    OPJ_IMF_SUBLEVEL_2_MBITSSEC,
                    OPJ_IMF_SUBLEVEL_3_MBITSSEC,
                    OPJ_IMF_SUBLEVEL_4_MBITSSEC,
                    OPJ_IMF_SUBLEVEL_5_MBITSSEC,
                    OPJ_IMF_SUBLEVEL_6_MBITSSEC,
                    OPJ_IMF_SUBLEVEL_7_MBITSSEC,
                    OPJ_IMF_SUBLEVEL_8_MBITSSEC,
                    OPJ_IMF_SUBLEVEL_9_MBITSSEC
                };
                parameters->max_cs_size = limitMBitsSec[sublevel] * (1000 * 1000 / 8) /
                                          framerate;
                fprintf(stdout, "Setting max codestream size to %d bytes.\n",
                        parameters->max_cs_size);
            }
        }
        break;

        /* ------------------------------------------------------ */

        case 'Y': {         /* Shall we do an MCT ? 0:no_mct;1:rgb->ycc;2:custom mct (-m option required)*/
            int mct_mode = 0;
            sscanf(opj_optarg, "%d", &mct_mode);
            if (mct_mode < 0 || mct_mode > 2) {
                fprintf(stderr,
                        "MCT incorrect value!! Current accepted values are 0, 1 or 2.\n");
                return 1;

        case 'J': {         /* jpip on */
            parameters->jpip_on = OPJ_TRUE;
        }
        break;
        /* ------------------------------------------------------ */

        case 'A': {         /* PLT markers */
            *pOutPLT = OPJ_TRUE;
        }
        break;

        /* ----------------------------------------------------- */
        case 'B': { /* Number of threads */
            if (strcmp(opj_optarg, "ALL_CPUS") == 0) {
                *pOutNumThreads = opj_get_num_cpus();
                if (*pOutNumThreads == 1) {
                    *pOutNumThreads = 0;
                }
            } else {
                sscanf(opj_optarg, "%d", pOutNumThreads);
            }
        }
        break;

        /* ------------------------------------------------------ */


        default:
            fprintf(stderr, "[WARNING] An invalid option has been ignored\n");
            break;
        }
    } while (c != -1);

    dircnt_t *dirptr = NULL;

    int ret = 0;

    OPJ_BOOL bSuccess;
    OPJ_BOOL bUseTiles = OPJ_FALSE; /* OPJ_TRUE */
    OPJ_UINT32 l_nb_tiles = 4;
    int framerate = 0;
    OPJ_FLOAT64 t = opj_clock();

    OPJ_BOOL PLT = OPJ_FALSE;
    int num_threads = 0;

    /* set encoding parameters to default values */
    opj_set_default_encoder_parameters(&parameters);

    /* Initialize indexfilename and img_fol */
    *indexfilename = 0;
    memset(&img_fol, 0, sizeof(img_fol_t));

    /* raw_cp initialization */
    raw_cp.rawBitDepth = 0;
    raw_cp.rawComp = 0;
    raw_cp.rawComps = 0;
    raw_cp.rawHeight = 0;
    raw_cp.rawSigned = 0;
    raw_cp.rawWidth = 0;

    /* parse input and get user encoding parameters */
    parameters.tcp_mct = (char)
                         255; /* This will be set later according to the input image or the provided option */
    if (parse_cmdline_encoder(argc, argv, &parameters, &img_fol, &raw_cp,
                              indexfilename, sizeof(indexfilename), &framerate, &PLT, &num_threads) == 1) {
        ret = 1;
        goto fin;
    }

    /* Read directory if necessary */
    if (img_fol.set_imgdir == 1) {
        num_images = get_num_images(img_fol.imgdirpath);

            if ((parameters.tcp_mct == 2) && (!parameters.mct_data)) {
                fprintf(stderr, "Custom MCT has been set but no array-based MCT\n");
                fprintf(stderr, "has been provided. Aborting.\n");
                ret = 1;
                goto fin;
            }
        }

        if (OPJ_IS_IMF(parameters.rsiz) && framerate > 0) {
            const int mainlevel = OPJ_GET_IMF_MAINLEVEL(parameters.rsiz);
            if (mainlevel > 0 && mainlevel <= OPJ_IMF_MAINLEVEL_MAX) {
                const int limitMSamplesSec[] = {
                    0,
                    OPJ_IMF_MAINLEVEL_1_MSAMPLESEC,
                    OPJ_IMF_MAINLEVEL_2_MSAMPLESEC,
                    OPJ_IMF_MAINLEVEL_3_MSAMPLESEC,
                    OPJ_IMF_MAINLEVEL_4_MSAMPLESEC,
                    OPJ_IMF_MAINLEVEL_5_MSAMPLESEC,
                    OPJ_IMF_MAINLEVEL_6_MSAMPLESEC,
                    OPJ_IMF_MAINLEVEL_7_MSAMPLESEC,
                    OPJ_IMF_MAINLEVEL_8_MSAMPLESEC,
                    OPJ_IMF_MAINLEVEL_9_MSAMPLESEC,
                    OPJ_IMF_MAINLEVEL_10_MSAMPLESEC,
                    OPJ_IMF_MAINLEVEL_11_MSAMPLESEC
                };
                OPJ_UINT32 avgcomponents = image->numcomps;
                double msamplespersec;
                if (image->numcomps == 3 &&
                        image->comps[1].dx == 2 &&
                        image->comps[1].dy == 2) {
                    avgcomponents = 2;
                }
                msamplespersec = (double)image->x1 * image->y1 * avgcomponents * framerate /
                                 1e6;
                if (msamplespersec > limitMSamplesSec[mainlevel]) {
                    fprintf(stderr,
                            "Warning: MSamples/sec is %f, whereas limit is %d.\n",
                            msamplespersec,
                            limitMSamplesSec[mainlevel]);
                }
            }
        }

        /* encode the destination image */
        /* ---------------------------- */

        switch (parameters.cod_format) {
        case J2K_CFMT: { /* JPEG-2000 codestream */
            /* Get a decoder handle */

        if (! opj_setup_encoder(l_codec, &parameters, image)) {
            fprintf(stderr, "failed to encode image: opj_setup_encoder\n");
            opj_destroy_codec(l_codec);
            opj_image_destroy(image);
            ret = 1;
            goto fin;
        }

        if (PLT) {
            const char* const options[] = { "PLT=YES", NULL };
            if (!opj_encoder_set_extra_options(l_codec, options)) {
                fprintf(stderr, "failed to encode image: opj_encoder_set_extra_options\n");
                opj_destroy_codec(l_codec);
                opj_image_destroy(image);
                ret = 1;
                goto fin;
            }
        }

        if (num_threads >= 1 &&
                !opj_codec_set_threads(l_codec, num_threads)) {
            fprintf(stderr, "failed to set number of threads\n");
            opj_destroy_codec(l_codec);
            opj_image_destroy(image);
            ret = 1;
            goto fin;
        }

        /* open a byte stream for writing and allocate memory for all tiles */
        l_stream = opj_stream_create_default_file_stream(parameters.outfile, OPJ_FALSE);
        if (! l_stream) {
            ret = 1;
            goto fin;
        }

/**
 * OPJ_DECOMPRESS MAIN
 */
/* -------------------------------------------------------------------------- */
int main(int argc, char **argv)
{
    opj_decompress_parameters parameters;           /* decompression parameters */





    OPJ_INT32 num_images, imageno;
    img_fol_t img_fol;
    dircnt_t *dirptr = NULL;
    int failed = 0;
    OPJ_FLOAT64 t, tCumulative = 0;
    OPJ_UINT32 numDecompressedImages = 0;

        }
    } else {
        num_images = 1;
    }

    /*Decoding image one by one*/
    for (imageno = 0; imageno < num_images ; imageno++)  {
        opj_image_t* image = NULL;
        opj_stream_t *l_stream = NULL;              /* Stream */
        opj_codec_t* l_codec = NULL;                /* Handle to a decompressor */
        opj_codestream_index_t* cstr_index = NULL;

        if (!parameters.quiet) {
            fprintf(stderr, "\n");
        }

        if (img_fol.set_imgdir == 1) {
            if (get_next_file(imageno, dirptr, &img_fol, &parameters)) {

                goto fin;
            }
            if (!(parameters.quiet)) {
                fprintf(stdout, "tile %d is decoded!\n\n", parameters.tile_index);
            }
        }












        tCumulative += opj_clock() - t;
        numDecompressedImages++;

        /* Close the byte stream */
        opj_stream_destroy(l_stream);

        if (image->color_space != OPJ_CLRSPC_SYCC

        }

    }/* else if pattern*.pgx */

    if (!sliceno) {
        fprintf(stdout,
                "[ERROR] No slices with this pattern founded !! Please check input volume name\n");
        closedir(dirp);
        return NULL;
    }
    /*if ( maxslice != sliceno) {
        fprintf(stdout,"[ERROR] Slices are not sequentially numbered !! Please rename them accordingly\n");
        return NULL;
    }*/

        if (endian1 == 'M' && endian2 == 'L') {
            cmptparm.bigendian = 1;
        } else if (endian2 == 'M' && endian1 == 'L') {
            cmptparm.bigendian = 0;
        } else {
            fprintf(stdout, "[ERROR] Bad pgx header, please check input file\n");
            fclose(f);
            closedir(dirp);
            return NULL;
        }

        if (s == 0) {
            /* initialize volume component */

            cmptparm.x0 = parameters->volume_offset_x0;

            }
            comp->data[i + offset] = v;

        }
        fclose(f);
    } /* for s --> sliceno*/
    comp->bpp = int_floorlog2(maxvalue) + 1;

    closedir(dirp);

    /*dump_volume(stdout, volume);*/
    return volume;
}


int volumetopgx(opj_volume_t * volume, char *outfile)
{

    volume->y1 = parameters->volume_offset_y0 + (h - 1) *   subsampling_dy + 1;
    volume->z1 = parameters->volume_offset_z0 + (l - 1) *   subsampling_dz + 1;

    /* set volume data */
    f = fopen(filename, "rb");
    if (!f) {
        fprintf(stdout, "[ERROR] Failed to open %s for reading !!\n", filename);
        opj_free(volume);
        return 0;
    }

    /* BINARY */
    for (compno = 0; compno < volume->numcomps; compno++) {
        int whl = w * h * l;
        /* set volume data */

    volume->z1 = parameters->volume_offset_z0 + (l - 1) *   subsampling_dz + 1;

    max = 0;
    /* set volume data */
    f = fopen(filename, "rb");
    if (!f) {
        fprintf(stderr, "[ERROR] Failed to open %s for reading !!\n", filename);
        opj_free(volume);
        return 0;
    }

    /* BINARY */
    for (compno = 0; compno < volume->numcomps; compno++) {
        int whl = w * h * l;
        /* set volume data */

  install(TARGETS ${exe}
    EXPORT OpenJPEGTargets
    DESTINATION ${OPENJPEG_INSTALL_BIN_DIR} COMPONENT Applications
    )
endforeach()

# Build the two java clients:
find_package(Java 1.6 COMPONENTS Development) # javac, jar

# User can override this:
if(NOT DEFINED JAVA_SOURCE_VERSION)
  set(JAVA_SOURCE_VERSION 1.6)
endif()
if(NOT DEFINED JAVA_TARGET_VERSION)
  set(JAVA_TARGET_VERSION 1.6)
endif()

# Only build the java viewer if dev is found:
if(Java_Development_FOUND AND Java_JAVAC_EXECUTABLE)
  set(jflags $ENV{JFLAGS})
  # search for package org.apache.xerces.parsers
  find_file(APACHE_XERCES_JAR

static int jpip_to_jp2(char *argv[])
{
    jpip_dec_param_t *dec;

    dec = init_jpipdecoder(OPJ_TRUE);

    if (!(fread_jpip(argv[1], dec))) {
        destroy_jpipdecoder(&dec);
        return 1;
    }

    decode_jpip(dec);

    if (!(fwrite_jp2k(argv[2], dec))) {
        destroy_jpipdecoder(&dec);
        return 1;
    }

    /* output_log( OPJ_TRUE, OPJ_FALSE, OPJ_TRUE, dec); */

    destroy_jpipdecoder(&dec);

static int jpip_to_j2k(char *argv[])
{
    jpip_dec_param_t *dec;

    dec = init_jpipdecoder(OPJ_FALSE);

    if (!(fread_jpip(argv[1], dec))) {
        destroy_jpipdecoder(&dec);
        return 1;
    }

    decode_jpip(dec);

    if (!(fwrite_jp2k(argv[2], dec))) {
        destroy_jpipdecoder(&dec);
        return 1;
    }

    /*  output_log( OPJ_TRUE, OPJ_FALSE, OPJ_FALSE, dec); */

    destroy_jpipdecoder(&dec);

    for (i = 0; i < image->numcomps - 1; i++) {
        if ((image->comps[0].dx != image->comps[i + 1].dx)
                || (image->comps[0].dy != image->comps[i + 1].dy)
                || (image->comps[0].prec != image->comps[i + 1].prec)) {
            fprintf(stderr,
                    "Unable to create a tga file with such J2K image charateristics.");
            fclose(fdest);
            return 1;
        }
    }

    width = image->comps[0].w;
    height = image->comps[0].h;

    /* Mono with alpha, or RGB with alpha. */
    write_alpha = (image->numcomps == 2) || (image->numcomps == 4);

    /* Write TGA header  */
    bpp = write_alpha ? 32 : 24;
    if (!tga_writeheader(fdest, bpp, width, height, OPJ_TRUE)) {
        fclose(fdest);
        return 1;
    }

    alpha_channel = image->numcomps - 1;

    scale = 255.0f / (float)((1 << image->comps[0].prec) - 1);

            }

            /* TGA format writes BGR ... */
            value = (unsigned char)(b * scale);
            res = fwrite(&value, 1, 1, fdest);
            if (res < 1) {
                fprintf(stderr, "failed to write 1 byte for %s\n", outfile);
                fclose(fdest);
                return 1;
            }

            value = (unsigned char)(g * scale);
            res = fwrite(&value, 1, 1, fdest);
            if (res < 1) {
                fprintf(stderr, "failed to write 1 byte for %s\n", outfile);
                fclose(fdest);
                return 1;
            }

            value = (unsigned char)(r * scale);
            res = fwrite(&value, 1, 1, fdest);
            if (res < 1) {
                fprintf(stderr, "failed to write 1 byte for %s\n", outfile);
                fclose(fdest);
                return 1;
            }

            if (write_alpha) {
                a = (float)(image->comps[alpha_channel].data[index]);
                value = (unsigned char)(a * scale);
                res = fwrite(&value, 1, 1, fdest);
                if (res < 1) {
                    fprintf(stderr, "failed to write 1 byte for %s\n", outfile);
                    fclose(fdest);
                    return 1;
                }
            }
        }
    }

    fclose(fdest);

    return 0;
}

/* -->> -->> -->> -->>

  BMP IMAGE FORMAT

                    (h - 1) * subsampling_dy + 1;

        /* set image data */

        RGB = (unsigned char *) malloc(W * H * sizeof(unsigned char));

        if (fread(RGB, sizeof(unsigned char), W * H, IN) != W * H) {
            fclose(IN);
            free(table_R);
            free(table_G);
            free(table_B);
            free(RGB);
            opj_image_destroy(image);
            fprintf(stderr,
                    "\nError: fread return a number of element different from the expected.\n");

        char bname[256]; /* buffer for name */
        char *name = bname; /* pointer */
        int nbytes = 0;
        size_t res;
        const size_t olen = strlen(outfile);
        const size_t dotpos = olen - 4;
        const size_t total = dotpos + 1 + 1 + 4; /* '-' + '[1-3]' + '.pgx' */
        if (olen < 4 || outfile[dotpos] != '.') {
            /* `pgx` was recognized but there is no dot at expected position */
            fprintf(stderr, "ERROR -> Impossible happen.");
            return 1;
        }
        if (total > 256) {
            name = (char*)malloc(total + 1);
        }
        strncpy(name, outfile, dotpos);
        /*if (image->numcomps > 1) {*/
        sprintf(name + dotpos, "_%d.pgx", compno);
        /*} else {
            strcpy(name+dotpos, ".pgx");
        }*/
        fdest = fopen(name, "wb");
        if (!fdest) {
            fprintf(stderr, "ERROR -> failed to open %s for writing\n", name);
            free(name);
            return 1;
        }
        /* don't need name anymore */
        if (total > 256) {
            free(name);
        }

    /* Read directory if necessary */
    if (img_fol.set_imgdir == 1) {
        num_images = get_num_images(img_fol.imgdirpath);
        dirptr = (dircnt_t*)malloc(sizeof(dircnt_t));
        if (dirptr) {
            dirptr->filename_buf = (char*)malloc(num_images * OPJ_PATH_LEN * sizeof(
                    char)); /* Stores at max 10 image file names*/
            if (!dirptr->filename_buf) {
                free(parameters.cp_comment);
                free(dirptr);
                return 0;
            }
            dirptr->filename = (char**) malloc(num_images * sizeof(char*));
            if (!dirptr->filename) {
                free(parameters.cp_comment);
                free(dirptr);
                return 0;
            }
            for (i = 0; i < num_images; i++) {
                dirptr->filename[i] = dirptr->filename_buf + i * OPJ_PATH_LEN;
            }
        }
        if (load_images(dirptr, img_fol.imgdirpath) == 1) {
            free(parameters.cp_comment);
            free(dirptr);
            return 0;
        }
        if (num_images == 0) {
            free(parameters.cp_comment);
            free(dirptr);
            fprintf(stdout, "Folder is empty\n");
            return 0;
        }
    } else {
        num_images = 1;
    }
    /*Encoding image one by one*/

                fprintf(stderr, "failed to open %s for writing\n", parameters.outfile);
                return 1;
            }
            res = fwrite(cio->buffer, 1, codestream_length, f);
            if (res < (size_t)codestream_length) {  /* FIXME */
                fprintf(stderr, "failed to write %d (%s)\n", codestream_length,
                        parameters.outfile);
                fclose(f);
                return 1;
            }
            fclose(f);

            fprintf(stderr, "Generated outfile %s\n", parameters.outfile);
            /* close and free the byte stream */
            opj_cio_close(cio);

                fprintf(stderr, "failed to open %s for writing\n", parameters.outfile);
                return 1;
            }
            res = fwrite(cio->buffer, 1, codestream_length, f);
            if (res < (size_t)codestream_length) {  /* FIXME */
                fprintf(stderr, "failed to write %d (%s)\n", codestream_length,
                        parameters.outfile);
                fclose(f);
                return 1;
            }
            fclose(f);
            fprintf(stderr, "Generated outfile %s\n", parameters.outfile);
            /* close and free the byte stream */
            opj_cio_close(cio);

    if (img_fol.set_imgdir == 1) {
        num_images = get_num_images(img_fol.imgdirpath);

        dirptr = (dircnt_t*)malloc(sizeof(dircnt_t));
        if (dirptr) {
            dirptr->filename_buf = (char*)malloc(num_images * OPJ_PATH_LEN * sizeof(
                    char)); /* Stores at max 10 image file names*/
            if (!dirptr->filename_buf) {
                free(dirptr);
                return 1;
            }
            dirptr->filename = (char**) malloc(num_images * sizeof(char*));
            if (!dirptr->filename) {
                free(dirptr);
                return 1;
            }

            for (i = 0; i < num_images; i++) {
                dirptr->filename[i] = dirptr->filename_buf + i * OPJ_PATH_LEN;
            }
        }
        if (load_images(dirptr, img_fol.imgdirpath) == 1) {
            return 1;
        }

        outfile++; /* There may be a leading blank if user put space after -o */
    }

    // Checking output file
    xmlout = fopen(outfile, "w"); /* was: argv[2] */
    if (!xmlout) {
        fprintf(stderr, "Failed to open %s for writing.\n", outfile); /* was: argv[2] */
        fclose(file);
        return 1;
    }
    // Leave it open

    /*
    configure the event callbacks (not required)
    setting of each callback is optional

    /* setup encoder parameters */
    mj2_setup_encoder(movie, &mj2_parameters);

    movie->tk[0].num_samples =
        yuv_num_frames(&movie->tk[0], mj2_parameters.infile);

    if (movie->tk[0].num_samples == 0) {
        fclose(mj2file);
        return 1;
    }

    /* One sample per chunk*/
    movie->tk[0].chunk = (mj2_chunk_t*)
                         malloc(movie->tk[0].num_samples * sizeof(mj2_chunk_t));
    movie->tk[0].sample = (mj2_sample_t*)

    if (!file) {
        fprintf(stderr, "failed to open %s for reading\n", argv[1]);
        return 1;
    }

    /* Checking output file */
    outfile = fopen(argv[2], "w");
    if (!outfile) {
        fprintf(stderr, "failed to open %s for writing\n", argv[2]);
        fclose(file);
        return 1;
    }
    fclose(outfile);

    /*
    configure the event callbacks (not required)
    setting of each callback is optional

        sample = &track->sample[snum];
        if (sample->sample_size - 8 > max_codstrm_size) {
            max_codstrm_size =  sample->sample_size - 8;
            if ((frame_codestream = (unsigned char*)
                                    realloc(frame_codestream, max_codstrm_size)) == NULL) {
                printf("Error reallocation memory\n");
                free(frame_codestream);
                return 1;
            };
        }
        fseek(file, sample->offset + 8, SEEK_SET);
        fread(frame_codestream, sample->sample_size - 8, 1,
              file); /* Assuming that jp and ftyp markers size do */

  message(FATAL_ERROR "No thread library found and thread/mutex support is required by OPJ_USE_THREAD option")
endif(OPJ_USE_THREAD AND NOT Threads_FOUND)

if(OPJ_USE_THREAD AND Threads_FOUND AND CMAKE_USE_PTHREADS_INIT)
   TARGET_LINK_LIBRARIES(${OPENJPEG_LIBRARY_NAME} ${CMAKE_THREAD_LIBS_INIT})
endif(OPJ_USE_THREAD AND Threads_FOUND AND CMAKE_USE_PTHREADS_INIT)

if(BUILD_UNIT_TESTS AND UNIX)
    add_executable(bench_dwt bench_dwt.c)
    if(UNIX)
        target_link_libraries(bench_dwt m ${OPENJPEG_LIBRARY_NAME})
    endif()
    if(OPJ_USE_THREAD AND Threads_FOUND AND CMAKE_USE_PTHREADS_INIT)
        target_link_libraries(bench_dwt ${CMAKE_THREAD_LIBS_INIT})
    endif(OPJ_USE_THREAD AND Threads_FOUND AND CMAKE_USE_PTHREADS_INIT)

    add_executable(test_sparse_array test_sparse_array.c)
    if(UNIX)
        target_link_libraries(test_sparse_array m ${OPENJPEG_LIBRARY_NAME})
    endif()
    if(OPJ_USE_THREAD AND Threads_FOUND AND CMAKE_USE_PTHREADS_INIT)
        target_link_libraries(test_sparse_array ${CMAKE_THREAD_LIBS_INIT})
    endif(OPJ_USE_THREAD AND Threads_FOUND AND CMAKE_USE_PTHREADS_INIT)
endif(BUILD_UNIT_TESTS AND UNIX)

}

void init_tilec(opj_tcd_tilecomp_t * l_tilec,
                OPJ_INT32 x0,
                OPJ_INT32 y0,
                OPJ_INT32 x1,
                OPJ_INT32 y1,
                OPJ_UINT32 numresolutions,
                OPJ_BOOL irreversible)
{
    opj_tcd_resolution_t* l_res;
    OPJ_UINT32 resno, l_level_no;
    size_t i, nValues;

    memset(l_tilec, 0, sizeof(*l_tilec));
    l_tilec->x0 = x0;
    l_tilec->y0 = y0;
    l_tilec->x1 = x1;
    l_tilec->y1 = y1;
    nValues = (size_t)(l_tilec->x1 - l_tilec->x0) *
              (size_t)(l_tilec->y1 - l_tilec->y0);
    l_tilec->data = (OPJ_INT32*) opj_malloc(sizeof(OPJ_INT32) * nValues);
    for (i = 0; i < nValues; i++) {
        OPJ_INT32 val = getValue((OPJ_UINT32)i);
        if (irreversible) {
            OPJ_FLOAT32 fVal = (OPJ_FLOAT32)val;
            memcpy(&l_tilec->data[i], &fVal, sizeof(OPJ_FLOAT32));
        } else {
            l_tilec->data[i] = val;
        }
    }
    l_tilec->numresolutions = numresolutions;
    l_tilec->minimum_num_resolutions = numresolutions;
    l_tilec->resolutions = (opj_tcd_resolution_t*) opj_calloc(
                               l_tilec->numresolutions,
                               sizeof(opj_tcd_resolution_t));

    l_level_no = l_tilec->numresolutions;
    l_res = l_tilec->resolutions;

    opj_free(l_tilec->data);
    opj_free(l_tilec->resolutions);
}

void usage(void)
{
    printf(
        "bench_dwt [-decode|encode] [-I] [-size value] [-check] [-display]\n");
    printf(
        "          [-num_resolutions val] [-offset x y] [-num_threads val]\n");
    exit(1);
}


OPJ_FLOAT64 opj_clock(void)
{
#ifdef _WIN32

    /* (2a) Get the seconds */
    procTime = (OPJ_FLOAT64)(t.ru_utime.tv_sec + t.ru_stime.tv_sec);
    /* (2b) More precisely! Get the microseconds part ! */
    return (procTime + (OPJ_FLOAT64)(t.ru_utime.tv_usec + t.ru_stime.tv_usec) *
            1e-6) ;
#endif
}

static OPJ_FLOAT64 opj_wallclock(void)
{
#ifdef _WIN32
    return opj_clock();
#else
    struct timeval tv;
    gettimeofday(&tv, NULL);
    return (OPJ_FLOAT64)tv.tv_sec + 1e-6 * (OPJ_FLOAT64)tv.tv_usec;
#endif
}

int main(int argc, char** argv)
{
    int num_threads = 0;
    opj_tcd_t tcd;
    opj_tcd_image_t tcd_image;
    opj_tcd_tile_t tcd_tile;
    opj_tcd_tilecomp_t tilec;
    opj_image_t image;
    opj_image_comp_t image_comp;
    opj_thread_pool_t* tp;
    OPJ_INT32 i, j, k;
    OPJ_BOOL display = OPJ_FALSE;
    OPJ_BOOL check = OPJ_FALSE;
    OPJ_INT32 size = 16384 - 1;
    OPJ_FLOAT64 start, stop;
    OPJ_FLOAT64 start_wc, stop_wc;
    OPJ_UINT32 offset_x = ((OPJ_UINT32)size + 1) / 2 - 1;
    OPJ_UINT32 offset_y = ((OPJ_UINT32)size + 1) / 2 - 1;
    OPJ_UINT32 num_resolutions = 6;
    OPJ_BOOL bench_decode = OPJ_TRUE;
    OPJ_BOOL irreversible = OPJ_FALSE;

    for (i = 1; i < argc; i++) {
        if (strcmp(argv[i], "-encode") == 0) {
            bench_decode = OPJ_FALSE;
        } else if (strcmp(argv[i], "-decode") == 0) {
            bench_decode = OPJ_TRUE;
        } else if (strcmp(argv[i], "-display") == 0) {
            display = OPJ_TRUE;
        } else if (strcmp(argv[i], "-check") == 0) {
            check = OPJ_TRUE;
        } else if (strcmp(argv[i], "-I") == 0) {
            irreversible = OPJ_TRUE;
        } else if (strcmp(argv[i], "-size") == 0 && i + 1 < argc) {
            size = atoi(argv[i + 1]);
            i ++;
        } else if (strcmp(argv[i], "-num_threads") == 0 && i + 1 < argc) {
            num_threads = atoi(argv[i + 1]);
            i ++;
        } else if (strcmp(argv[i], "-num_resolutions") == 0 && i + 1 < argc) {

            offset_x = (OPJ_UINT32)atoi(argv[i + 1]);
            offset_y = (OPJ_UINT32)atoi(argv[i + 2]);
            i += 2;
        } else {
            usage();
        }
    }

    if (irreversible && check) {
        /* Due to irreversible inverse DWT not being symetric of forward */
        /* See BUG_WEIRD_TWO_INVK in dwt.c */
        printf("-I and -check aren't compatible\n");
        exit(1);
    }

    tp = opj_thread_pool_create(num_threads);

    init_tilec(&tilec, (OPJ_INT32)offset_x, (OPJ_INT32)offset_y,
               (OPJ_INT32)offset_x + size, (OPJ_INT32)offset_y + size,
               num_resolutions, irreversible);

    if (display) {
        printf("Before\n");
        k = 0;
        for (j = 0; j < tilec.y1 - tilec.y0; j++) {
            for (i = 0; i < tilec.x1 - tilec.x0; i++) {
                if (irreversible) {
                    printf("%f ", ((OPJ_FLOAT32*)tilec.data)[k]);
                } else {
                    printf("%d ", tilec.data[k]);
                }
                k ++;
            }
            printf("\n");
        }
    }

    memset(&tcd, 0, sizeof(tcd));

    image.numcomps = 1;
    image.comps = &image_comp;
    memset(&image_comp, 0, sizeof(image_comp));
    image_comp.dx = 1;
    image_comp.dy = 1;

    start = opj_clock();
    start_wc = opj_wallclock();
    if (bench_decode) {
        if (irreversible)  {
            opj_dwt_decode_real(&tcd, &tilec, tilec.numresolutions);
        } else {
            opj_dwt_decode(&tcd, &tilec, tilec.numresolutions);
        }
    } else {
        if (irreversible)  {
            opj_dwt_encode_real(&tcd, &tilec);
        } else {
            opj_dwt_encode(&tcd, &tilec);
        }
    }
    stop = opj_clock();
    stop_wc = opj_wallclock();
    printf("time for %s: total = %.03f s, wallclock = %.03f s\n",
           bench_decode ? "dwt_decode" : "dwt_encode",
           stop - start,
           stop_wc - start_wc);

    if (display) {
        if (bench_decode) {
            printf("After IDWT\n");
        } else {
            printf("After FDWT\n");
        }
        k = 0;
        for (j = 0; j < tilec.y1 - tilec.y0; j++) {
            for (i = 0; i < tilec.x1 - tilec.x0; i++) {
                if (irreversible) {
                    printf("%f ", ((OPJ_FLOAT32*)tilec.data)[k]);
                } else {
                    printf("%d ", tilec.data[k]);
                }
                k ++;
            }
            printf("\n");
        }
    }

    if ((display || check) && !irreversible) {

        if (bench_decode) {
            opj_dwt_encode(&tcd, &tilec);
        } else {
            opj_dwt_decode(&tcd, &tilec, tilec.numresolutions);
        }


        if (display && !irreversible) {
            if (bench_decode) {
                printf("After FDWT\n");
            } else {
                printf("After IDWT\n");
            }
            k = 0;
            for (j = 0; j < tilec.y1 - tilec.y0; j++) {
                for (i = 0; i < tilec.x1 - tilec.x0; i++) {
                    if (irreversible) {
                        printf("%f ", ((OPJ_FLOAT32*)tilec.data)[k]);
                    } else {
                        printf("%d ", tilec.data[k]);
                    }
                    k ++;
                }
                printf("\n");
            }
        }

    }

    if (check) {

        size_t idx;
        size_t nValues = (size_t)(tilec.x1 - tilec.x0) *
                         (size_t)(tilec.y1 - tilec.y0);
        for (idx = 0; idx < nValues; idx++) {
            if (tilec.data[idx] != getValue((OPJ_UINT32)idx)) {
                printf("Difference found at idx = %u\n", (OPJ_UINT32)idx);
                exit(1);

            }
        }
    }

    free_tilec(&tilec);

    opj_thread_pool_destroy(tp);
    return 0;
}

typedef struct dwt_local {
    OPJ_INT32* mem;
    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 */
} opj_dwt_t;

#define NB_ELTS_V8  8

typedef union {
    OPJ_FLOAT32 f[NB_ELTS_V8];
} opj_v8_t;

typedef struct v8dwt_local {
    opj_v8_t*   wavelet ;
    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 */
    OPJ_UINT32      win_l_x0; /* start coord in low pass band */
    OPJ_UINT32      win_l_x1; /* end coord in low pass band */
    OPJ_UINT32      win_h_x0; /* start coord in high pass band */
    OPJ_UINT32      win_h_x1; /* end coord in high pass band */
} opj_v8dwt_t ;

/* From table F.4 from the standard */
static const OPJ_FLOAT32 opj_dwt_alpha =  -1.586134342f;
static const OPJ_FLOAT32 opj_dwt_beta  =  -0.052980118f;
static const OPJ_FLOAT32 opj_dwt_gamma = 0.882911075f;
static const OPJ_FLOAT32 opj_dwt_delta = 0.443506852f;

static const OPJ_FLOAT32 opj_K      = 1.230174105f;
static const OPJ_FLOAT32 opj_invK   = (OPJ_FLOAT32)(1.0 / 1.230174105);

/*@}*/






/** @name Local static functions */
/*@{*/

/**
Forward lazy transform (horizontal)
*/
static void opj_dwt_deinterleave_h(const OPJ_INT32 * OPJ_RESTRICT a,
                                   OPJ_INT32 * OPJ_RESTRICT b,
                                   OPJ_INT32 dn,




                                   OPJ_INT32 sn, OPJ_INT32 cas);





/**
Forward 9-7 wavelet transform in 1-D
*/
static void opj_dwt_encode_1_real(void *a, OPJ_INT32 dn, OPJ_INT32 sn,
                                  OPJ_INT32 cas);
/**
Explicit calculation of the Quantization Stepsizes
*/
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);

static OPJ_BOOL opj_dwt_decode_partial_tile(
    opj_tcd_tilecomp_t* tilec,
    OPJ_UINT32 numres);

/* Forward transform, for the vertical pass, processing cols columns */
/* where cols <= NB_ELTS_V8 */
/* Where void* is a OPJ_INT32* for 5x3 and OPJ_FLOAT32* for 9x7 */
typedef void (*opj_encode_and_deinterleave_v_fnptr_type)(
    void *array,
    void *tmp,
    OPJ_UINT32 height,
    OPJ_BOOL even,
    OPJ_UINT32 stride_width,
    OPJ_UINT32 cols);

/* Where void* is a OPJ_INT32* for 5x3 and OPJ_FLOAT32* for 9x7 */
typedef void (*opj_encode_and_deinterleave_h_one_row_fnptr_type)(
    void *row,
    void *tmp,
    OPJ_UINT32 width,
    OPJ_BOOL even);

static OPJ_BOOL opj_dwt_encode_procedure(opj_thread_pool_t* tp,

        opj_tcd_tilecomp_t * tilec,
        opj_encode_and_deinterleave_v_fnptr_type p_encode_and_deinterleave_v,
        opj_encode_and_deinterleave_h_one_row_fnptr_type
        p_encode_and_deinterleave_h_one_row);

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>                            */





































/*@}*/

/*@}*/

#define OPJ_S(i) a[(i)*2]
#define OPJ_D(i) a[(1+(i)*2)]

   local functions
==========================================================
*/

/* <summary>                             */
/* Forward lazy transform (horizontal).  */
/* </summary>                            */
static void opj_dwt_deinterleave_h(const OPJ_INT32 * OPJ_RESTRICT a,
                                   OPJ_INT32 * OPJ_RESTRICT b,
                                   OPJ_INT32 dn,
                                   OPJ_INT32 sn, OPJ_INT32 cas)
{
    OPJ_INT32 i;
    OPJ_INT32 * OPJ_RESTRICT l_dest = b;
    const OPJ_INT32 * OPJ_RESTRICT 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;
    }
}




























#ifdef STANDARD_SLOW_VERSION
/* <summary>                             */
/* Inverse lazy transform (horizontal).  */
/* </summary>                            */
static void opj_dwt_interleave_h(const opj_dwt_t* h, OPJ_INT32 *a)
{
    const 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;
    }
}

/* <summary>                             */
/* Inverse lazy transform (vertical).    */
/* </summary>                            */
static void opj_dwt_interleave_v(const opj_dwt_t* v, OPJ_INT32 *a, OPJ_INT32 x)
{
    const OPJ_INT32 *ai = a;
    OPJ_INT32 *bi = v->mem + v->cas;
    OPJ_INT32  i = v->sn;
    while (i--) {
        *bi = *ai;
        bi += 2;
        ai += x;
    }
    ai = a + (v->sn * (OPJ_SIZE_T)x);
    bi = v->mem + 1 - v->cas;
    i = v->dn ;
    while (i--) {
        *bi = *ai;
        bi += 2;
        ai += x;
    }
}

#endif /* STANDARD_SLOW_VERSION */
































#ifdef STANDARD_SLOW_VERSION
/* <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)
{

            }
            return;
        }
    }
#endif
}

#if 0
static void opj_dwt_encode_step1(OPJ_FLOAT32* fw,
                                 OPJ_UINT32 end,
                                 const OPJ_FLOAT32 c)
{
    OPJ_UINT32 i = 0;
    for (; i < end; ++i) {
        fw[0] *= c;
        fw += 2;
    }
}
#else
static void opj_dwt_encode_step1_combined(OPJ_FLOAT32* fw,
        OPJ_UINT32 iters_c1,
        OPJ_UINT32 iters_c2,
        const OPJ_FLOAT32 c1,
        const OPJ_FLOAT32 c2)
{
    OPJ_UINT32 i = 0;
    const OPJ_UINT32 iters_common =  opj_uint_min(iters_c1, iters_c2);
    assert((((OPJ_SIZE_T)fw) & 0xf) == 0);
    assert(opj_int_abs((OPJ_INT32)iters_c1 - (OPJ_INT32)iters_c2) <= 1);
    for (; i + 3 < iters_common; i += 4) {
#ifdef __SSE__
        const __m128 vcst = _mm_set_ps(c2, c1, c2, c1);
        *(__m128*)fw = _mm_mul_ps(*(__m128*)fw, vcst);
        *(__m128*)(fw + 4) = _mm_mul_ps(*(__m128*)(fw + 4), vcst);
#else
        fw[0] *= c1;
        fw[1] *= c2;
        fw[2] *= c1;
        fw[3] *= c2;
        fw[4] *= c1;
        fw[5] *= c2;
        fw[6] *= c1;
        fw[7] *= c2;
#endif
        fw += 8;
    }
    for (; i < iters_common; i++) {
        fw[0] *= c1;
        fw[1] *= c2;
        fw += 2;
    }
    if (i < iters_c1) {
        fw[0] *= c1;
    } else if (i < iters_c2) {
        fw[1] *= c2;
    }
}

#endif

static void opj_dwt_encode_step2(OPJ_FLOAT32* fl, OPJ_FLOAT32* fw,
                                 OPJ_UINT32 end,
                                 OPJ_UINT32 m,
                                 OPJ_FLOAT32 c)
{
    OPJ_UINT32 i;
    OPJ_UINT32 imax = opj_uint_min(end, m);
    if (imax > 0) {
        fw[-1] += (fl[0] + fw[0]) * c;
        fw += 2;
        i = 1;
        for (; i + 3 < imax; i += 4) {
            fw[-1] += (fw[-2] + fw[0]) * c;
            fw[1] += (fw[0] + fw[2]) * c;
            fw[3] += (fw[2] + fw[4]) * c;
            fw[5] += (fw[4] + fw[6]) * c;
            fw += 8;
        }
        for (; i < imax; ++i) {
            fw[-1] += (fw[-2] + fw[0]) * c;
            fw += 2;
        }
    }
    if (m < end) {
        assert(m + 1 == end);
        fw[-1] += (2 * fw[-2]) * c;
    }
}

static void opj_dwt_encode_1_real(void *aIn, OPJ_INT32 dn, OPJ_INT32 sn,
                                  OPJ_INT32 cas)
{
    OPJ_FLOAT32* w = (OPJ_FLOAT32*)aIn;
    OPJ_INT32 a, b;
    assert(dn + sn > 1);
    if (cas == 0) {

        a = 0;

        b = 1;
    } else {
        a = 1;
        b = 0;

    }
    opj_dwt_encode_step2(w + a, w + b + 1,
                         (OPJ_UINT32)dn,
                         (OPJ_UINT32)opj_int_min(dn, sn - b),

                         opj_dwt_alpha);
    opj_dwt_encode_step2(w + b, w + a + 1,
                         (OPJ_UINT32)sn,
                         (OPJ_UINT32)opj_int_min(sn, dn - a),

                         opj_dwt_beta);
    opj_dwt_encode_step2(w + a, w + b + 1,
                         (OPJ_UINT32)dn,
                         (OPJ_UINT32)opj_int_min(dn, sn - b),
                         opj_dwt_gamma);
    opj_dwt_encode_step2(w + b, w + a + 1,
                         (OPJ_UINT32)sn,
                         (OPJ_UINT32)opj_int_min(sn, dn - a),
                         opj_dwt_delta);
#if 0
    opj_dwt_encode_step1(w + b, (OPJ_UINT32)dn,
                         opj_K);
    opj_dwt_encode_step1(w + a, (OPJ_UINT32)sn,
                         opj_invK);

#else
    if (a == 0) {
        opj_dwt_encode_step1_combined(w,
                                      (OPJ_UINT32)sn,
                                      (OPJ_UINT32)dn,
                                      opj_invK,


                                      opj_K);
    } else {
        opj_dwt_encode_step1_combined(w,

                                      (OPJ_UINT32)dn,


                                      (OPJ_UINT32)sn,








                                      opj_K,


                                      opj_invK);
    }


#endif
}

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
==========================================================
*/

/** Process one line for the horizontal pass of the 5x3 forward transform */
static
void opj_dwt_encode_and_deinterleave_h_one_row(void* rowIn,
        void* tmpIn,
        OPJ_UINT32 width,
        OPJ_BOOL even)
{
    OPJ_INT32* OPJ_RESTRICT row = (OPJ_INT32*)rowIn;
    OPJ_INT32* OPJ_RESTRICT tmp = (OPJ_INT32*)tmpIn;
    const OPJ_INT32 sn = (OPJ_INT32)((width + (even ? 1 : 0)) >> 1);
    const OPJ_INT32 dn = (OPJ_INT32)(width - (OPJ_UINT32)sn);

    if (even) {
        if (width > 1) {
            OPJ_INT32 i;
            for (i = 0; i < sn - 1; i++) {
                tmp[sn + i] = row[2 * i + 1] - ((row[(i) * 2] + row[(i + 1) * 2]) >> 1);
            }
            if ((width % 2) == 0) {
                tmp[sn + i] = row[2 * i + 1] - row[(i) * 2];
            }
            row[0] += (tmp[sn] + tmp[sn] + 2) >> 2;
            for (i = 1; i < dn; i++) {
                row[i] = row[2 * i] + ((tmp[sn + (i - 1)] + tmp[sn + i] + 2) >> 2);
            }
            if ((width % 2) == 1) {
                row[i] = row[2 * i] + ((tmp[sn + (i - 1)] + tmp[sn + (i - 1)] + 2) >> 2);
            }
            memcpy(row + sn, tmp + sn, (OPJ_SIZE_T)dn * sizeof(OPJ_INT32));
        }
    } else {
        if (width == 1) {
            row[0] *= 2;
        } else {
            OPJ_INT32 i;
            tmp[sn + 0] = row[0] - row[1];
            for (i = 1; i < sn; i++) {
                tmp[sn + i] = row[2 * i] - ((row[2 * i + 1] + row[2 * (i - 1) + 1]) >> 1);
            }
            if ((width % 2) == 1) {
                tmp[sn + i] = row[2 * i] - row[2 * (i - 1) + 1];
            }

            for (i = 0; i < dn - 1; i++) {
                row[i] = row[2 * i + 1] + ((tmp[sn + i] + tmp[sn + i + 1] + 2) >> 2);
            }
            if ((width % 2) == 0) {
                row[i] = row[2 * i + 1] + ((tmp[sn + i] + tmp[sn + i] + 2) >> 2);
            }
            memcpy(row + sn, tmp + sn, (OPJ_SIZE_T)dn * sizeof(OPJ_INT32));
        }
    }
}

/** Process one line for the horizontal pass of the 9x7 forward transform */
static
void opj_dwt_encode_and_deinterleave_h_one_row_real(void* rowIn,
        void* tmpIn,
        OPJ_UINT32 width,
        OPJ_BOOL even)
{
    OPJ_FLOAT32* OPJ_RESTRICT row = (OPJ_FLOAT32*)rowIn;
    OPJ_FLOAT32* OPJ_RESTRICT tmp = (OPJ_FLOAT32*)tmpIn;
    const OPJ_INT32 sn = (OPJ_INT32)((width + (even ? 1 : 0)) >> 1);
    const OPJ_INT32 dn = (OPJ_INT32)(width - (OPJ_UINT32)sn);
    if (width == 1) {
        return;
    }
    memcpy(tmp, row, width * sizeof(OPJ_FLOAT32));
    opj_dwt_encode_1_real(tmp, dn, sn, even ? 0 : 1);
    opj_dwt_deinterleave_h((OPJ_INT32 * OPJ_RESTRICT)tmp,
                           (OPJ_INT32 * OPJ_RESTRICT)row,
                           dn, sn, even ? 0 : 1);
}

typedef struct {
    opj_dwt_t h;
    OPJ_UINT32 rw; /* Width of the resolution to process */
    OPJ_UINT32 w; /* Width of tiledp */
    OPJ_INT32 * OPJ_RESTRICT tiledp;
    OPJ_UINT32 min_j;
    OPJ_UINT32 max_j;
    opj_encode_and_deinterleave_h_one_row_fnptr_type p_function;
} opj_dwt_encode_h_job_t;

static void opj_dwt_encode_h_func(void* user_data, opj_tls_t* tls)
{
    OPJ_UINT32 j;
    opj_dwt_encode_h_job_t* job;
    (void)tls;

    job = (opj_dwt_encode_h_job_t*)user_data;
    for (j = job->min_j; j < job->max_j; j++) {
        OPJ_INT32* OPJ_RESTRICT aj = job->tiledp + j * job->w;
        (*job->p_function)(aj, job->h.mem, job->rw,
                           job->h.cas == 0 ? OPJ_TRUE : OPJ_FALSE);
    }

    opj_aligned_free(job->h.mem);
    opj_free(job);
}

typedef struct {
    opj_dwt_t v;
    OPJ_UINT32 rh;
    OPJ_UINT32 w;
    OPJ_INT32 * OPJ_RESTRICT tiledp;
    OPJ_UINT32 min_j;
    OPJ_UINT32 max_j;
    opj_encode_and_deinterleave_v_fnptr_type p_encode_and_deinterleave_v;
} opj_dwt_encode_v_job_t;

static void opj_dwt_encode_v_func(void* user_data, opj_tls_t* tls)
{
    OPJ_UINT32 j;
    opj_dwt_encode_v_job_t* job;
    (void)tls;

    job = (opj_dwt_encode_v_job_t*)user_data;
    for (j = job->min_j; j + NB_ELTS_V8 - 1 < job->max_j; j += NB_ELTS_V8) {
        (*job->p_encode_and_deinterleave_v)(job->tiledp + j,
                                            job->v.mem,
                                            job->rh,
                                            job->v.cas == 0,
                                            job->w,
                                            NB_ELTS_V8);
    }
    if (j < job->max_j) {
        (*job->p_encode_and_deinterleave_v)(job->tiledp + j,
                                            job->v.mem,
                                            job->rh,
                                            job->v.cas == 0,
                                            job->w,
                                            job->max_j - j);
    }

    opj_aligned_free(job->v.mem);
    opj_free(job);
}

/** Fetch up to cols <= NB_ELTS_V8 for each line, and put them in tmpOut */
/* that has a NB_ELTS_V8 interleave factor. */
static void opj_dwt_fetch_cols_vertical_pass(const void *arrayIn,
        void *tmpOut,
        OPJ_UINT32 height,
        OPJ_UINT32 stride_width,
        OPJ_UINT32 cols)
{
    const OPJ_INT32* OPJ_RESTRICT array = (const OPJ_INT32 * OPJ_RESTRICT)arrayIn;
    OPJ_INT32* OPJ_RESTRICT tmp = (OPJ_INT32 * OPJ_RESTRICT)tmpOut;
    if (cols == NB_ELTS_V8) {
        OPJ_UINT32 k;
        for (k = 0; k < height; ++k) {
            memcpy(tmp + NB_ELTS_V8 * k,
                   array + k * stride_width,
                   NB_ELTS_V8 * sizeof(OPJ_INT32));
        }
    } else {
        OPJ_UINT32 k;
        for (k = 0; k < height; ++k) {
            OPJ_UINT32 c;
            for (c = 0; c < cols; c++) {
                tmp[NB_ELTS_V8 * k + c] = array[c + k * stride_width];
            }
            for (; c < NB_ELTS_V8; c++) {
                tmp[NB_ELTS_V8 * k + c] = 0;
            }
        }
    }
}

/* Deinterleave result of forward transform, where cols <= NB_ELTS_V8 */
/* and src contains NB_ELTS_V8 consecutive values for up to NB_ELTS_V8 */
/* columns. */
static INLINE void opj_dwt_deinterleave_v_cols(
    const OPJ_INT32 * OPJ_RESTRICT src,
    OPJ_INT32 * OPJ_RESTRICT dst,
    OPJ_INT32 dn,
    OPJ_INT32 sn,
    OPJ_UINT32 stride_width,
    OPJ_INT32 cas,
    OPJ_UINT32 cols)
{
    OPJ_INT32 k;
    OPJ_INT32 i = sn;
    OPJ_INT32 * OPJ_RESTRICT l_dest = dst;
    const OPJ_INT32 * OPJ_RESTRICT l_src = src + cas * NB_ELTS_V8;
    OPJ_UINT32 c;

    for (k = 0; k < 2; k++) {
        while (i--) {
            if (cols == NB_ELTS_V8) {
                memcpy(l_dest, l_src, NB_ELTS_V8 * sizeof(OPJ_INT32));
            } else {
                c = 0;
                switch (cols) {
                case 7:
                    l_dest[c] = l_src[c];
                    c++; /* fallthru */
                case 6:
                    l_dest[c] = l_src[c];
                    c++; /* fallthru */
                case 5:
                    l_dest[c] = l_src[c];
                    c++; /* fallthru */
                case 4:
                    l_dest[c] = l_src[c];
                    c++; /* fallthru */
                case 3:
                    l_dest[c] = l_src[c];
                    c++; /* fallthru */
                case 2:
                    l_dest[c] = l_src[c];
                    c++; /* fallthru */
                default:
                    l_dest[c] = l_src[c];
                    break;
                }
            }
            l_dest += stride_width;
            l_src += 2 * NB_ELTS_V8;
        }

        l_dest = dst + (OPJ_SIZE_T)sn * (OPJ_SIZE_T)stride_width;
        l_src = src + (1 - cas) * NB_ELTS_V8;
        i = dn;
    }
}


/* Forward 5-3 transform, for the vertical pass, processing cols columns */
/* where cols <= NB_ELTS_V8 */
static void opj_dwt_encode_and_deinterleave_v(
    void *arrayIn,
    void *tmpIn,
    OPJ_UINT32 height,
    OPJ_BOOL even,
    OPJ_UINT32 stride_width,
    OPJ_UINT32 cols)
{
    OPJ_INT32* OPJ_RESTRICT array = (OPJ_INT32 * OPJ_RESTRICT)arrayIn;
    OPJ_INT32* OPJ_RESTRICT tmp = (OPJ_INT32 * OPJ_RESTRICT)tmpIn;
    const OPJ_UINT32 sn = (height + (even ? 1 : 0)) >> 1;
    const OPJ_UINT32 dn = height - sn;

    opj_dwt_fetch_cols_vertical_pass(arrayIn, tmpIn, height, stride_width, cols);

#define OPJ_Sc(i) tmp[(i)*2* NB_ELTS_V8 + c]
#define OPJ_Dc(i) tmp[((1+(i)*2))* NB_ELTS_V8 + c]

#ifdef __SSE2__
    if (height == 1) {
        if (!even) {
            OPJ_UINT32 c;
            for (c = 0; c < NB_ELTS_V8; c++) {
                tmp[c] *= 2;
            }
        }
    } else if (even) {
        OPJ_UINT32 c;
        OPJ_UINT32 i;
        i = 0;
        if (i + 1 < sn) {
            __m128i xmm_Si_0 = *(const __m128i*)(tmp + 4 * 0);
            __m128i xmm_Si_1 = *(const __m128i*)(tmp + 4 * 1);
            for (; i + 1 < sn; i++) {
                __m128i xmm_Sip1_0 = *(const __m128i*)(tmp +
                                                       (i + 1) * 2 * NB_ELTS_V8 + 4 * 0);
                __m128i xmm_Sip1_1 = *(const __m128i*)(tmp +
                                                       (i + 1) * 2 * NB_ELTS_V8 + 4 * 1);
                __m128i xmm_Di_0 = *(const __m128i*)(tmp +
                                                     (1 + i * 2) * NB_ELTS_V8 + 4 * 0);
                __m128i xmm_Di_1 = *(const __m128i*)(tmp +
                                                     (1 + i * 2) * NB_ELTS_V8 + 4 * 1);
                xmm_Di_0 = _mm_sub_epi32(xmm_Di_0,
                                         _mm_srai_epi32(_mm_add_epi32(xmm_Si_0, xmm_Sip1_0), 1));
                xmm_Di_1 = _mm_sub_epi32(xmm_Di_1,
                                         _mm_srai_epi32(_mm_add_epi32(xmm_Si_1, xmm_Sip1_1), 1));
                *(__m128i*)(tmp + (1 + i * 2) * NB_ELTS_V8 + 4 * 0) =  xmm_Di_0;
                *(__m128i*)(tmp + (1 + i * 2) * NB_ELTS_V8 + 4 * 1) =  xmm_Di_1;
                xmm_Si_0 = xmm_Sip1_0;
                xmm_Si_1 = xmm_Sip1_1;
            }
        }
        if (((height) % 2) == 0) {
            for (c = 0; c < NB_ELTS_V8; c++) {
                OPJ_Dc(i) -= OPJ_Sc(i);
            }
        }
        for (c = 0; c < NB_ELTS_V8; c++) {
            OPJ_Sc(0) += (OPJ_Dc(0) + OPJ_Dc(0) + 2) >> 2;
        }
        i = 1;
        if (i < dn) {
            __m128i xmm_Dim1_0 = *(const __m128i*)(tmp + (1 +
                                                   (i - 1) * 2) * NB_ELTS_V8 + 4 * 0);
            __m128i xmm_Dim1_1 = *(const __m128i*)(tmp + (1 +
                                                   (i - 1) * 2) * NB_ELTS_V8 + 4 * 1);
            const __m128i xmm_two = _mm_set1_epi32(2);
            for (; i < dn; i++) {
                __m128i xmm_Di_0 = *(const __m128i*)(tmp +
                                                     (1 + i * 2) * NB_ELTS_V8 + 4 * 0);
                __m128i xmm_Di_1 = *(const __m128i*)(tmp +
                                                     (1 + i * 2) * NB_ELTS_V8 + 4 * 1);
                __m128i xmm_Si_0 = *(const __m128i*)(tmp +
                                                     (i * 2) * NB_ELTS_V8 + 4 * 0);
                __m128i xmm_Si_1 = *(const __m128i*)(tmp +
                                                     (i * 2) * NB_ELTS_V8 + 4 * 1);
                xmm_Si_0 = _mm_add_epi32(xmm_Si_0,
                                         _mm_srai_epi32(_mm_add_epi32(_mm_add_epi32(xmm_Dim1_0, xmm_Di_0), xmm_two), 2));
                xmm_Si_1 = _mm_add_epi32(xmm_Si_1,
                                         _mm_srai_epi32(_mm_add_epi32(_mm_add_epi32(xmm_Dim1_1, xmm_Di_1), xmm_two), 2));
                *(__m128i*)(tmp + (i * 2) * NB_ELTS_V8 + 4 * 0) = xmm_Si_0;
                *(__m128i*)(tmp + (i * 2) * NB_ELTS_V8 + 4 * 1) = xmm_Si_1;
                xmm_Dim1_0 = xmm_Di_0;
                xmm_Dim1_1 = xmm_Di_1;
            }
        }
        if (((height) % 2) == 1) {
            for (c = 0; c < NB_ELTS_V8; c++) {
                OPJ_Sc(i) += (OPJ_Dc(i - 1) + OPJ_Dc(i - 1) + 2) >> 2;
            }
        }
    } else {
        OPJ_UINT32 c;
        OPJ_UINT32 i;
        for (c = 0; c < NB_ELTS_V8; c++) {
            OPJ_Sc(0) -= OPJ_Dc(0);
        }
        i = 1;
        if (i < sn) {
            __m128i xmm_Dim1_0 = *(const __m128i*)(tmp + (1 +
                                                   (i - 1) * 2) * NB_ELTS_V8 + 4 * 0);
            __m128i xmm_Dim1_1 = *(const __m128i*)(tmp + (1 +
                                                   (i - 1) * 2) * NB_ELTS_V8 + 4 * 1);
            for (; i < sn; i++) {
                __m128i xmm_Di_0 = *(const __m128i*)(tmp +
                                                     (1 + i * 2) * NB_ELTS_V8 + 4 * 0);
                __m128i xmm_Di_1 = *(const __m128i*)(tmp +
                                                     (1 + i * 2) * NB_ELTS_V8 + 4 * 1);
                __m128i xmm_Si_0 = *(const __m128i*)(tmp +
                                                     (i * 2) * NB_ELTS_V8 + 4 * 0);
                __m128i xmm_Si_1 = *(const __m128i*)(tmp +
                                                     (i * 2) * NB_ELTS_V8 + 4 * 1);
                xmm_Si_0 = _mm_sub_epi32(xmm_Si_0,
                                         _mm_srai_epi32(_mm_add_epi32(xmm_Di_0, xmm_Dim1_0), 1));
                xmm_Si_1 = _mm_sub_epi32(xmm_Si_1,
                                         _mm_srai_epi32(_mm_add_epi32(xmm_Di_1, xmm_Dim1_1), 1));
                *(__m128i*)(tmp + (i * 2) * NB_ELTS_V8 + 4 * 0) = xmm_Si_0;
                *(__m128i*)(tmp + (i * 2) * NB_ELTS_V8 + 4 * 1) = xmm_Si_1;
                xmm_Dim1_0 = xmm_Di_0;
                xmm_Dim1_1 = xmm_Di_1;
            }
        }
        if (((height) % 2) == 1) {
            for (c = 0; c < NB_ELTS_V8; c++) {
                OPJ_Sc(i) -= OPJ_Dc(i - 1);
            }
        }
        i = 0;
        if (i + 1 < dn) {
            __m128i xmm_Si_0 = *((const __m128i*)(tmp + 4 * 0));
            __m128i xmm_Si_1 = *((const __m128i*)(tmp + 4 * 1));
            const __m128i xmm_two = _mm_set1_epi32(2);
            for (; i + 1 < dn; i++) {
                __m128i xmm_Sip1_0 = *(const __m128i*)(tmp +
                                                       (i + 1) * 2 * NB_ELTS_V8 + 4 * 0);
                __m128i xmm_Sip1_1 = *(const __m128i*)(tmp +
                                                       (i + 1) * 2 * NB_ELTS_V8 + 4 * 1);
                __m128i xmm_Di_0 = *(const __m128i*)(tmp +
                                                     (1 + i * 2) * NB_ELTS_V8 + 4 * 0);
                __m128i xmm_Di_1 = *(const __m128i*)(tmp +
                                                     (1 + i * 2) * NB_ELTS_V8 + 4 * 1);
                xmm_Di_0 = _mm_add_epi32(xmm_Di_0,
                                         _mm_srai_epi32(_mm_add_epi32(_mm_add_epi32(xmm_Si_0, xmm_Sip1_0), xmm_two), 2));
                xmm_Di_1 = _mm_add_epi32(xmm_Di_1,
                                         _mm_srai_epi32(_mm_add_epi32(_mm_add_epi32(xmm_Si_1, xmm_Sip1_1), xmm_two), 2));
                *(__m128i*)(tmp + (1 + i * 2) * NB_ELTS_V8 + 4 * 0) = xmm_Di_0;
                *(__m128i*)(tmp + (1 + i * 2) * NB_ELTS_V8 + 4 * 1) = xmm_Di_1;
                xmm_Si_0 = xmm_Sip1_0;
                xmm_Si_1 = xmm_Sip1_1;
            }
        }
        if (((height) % 2) == 0) {
            for (c = 0; c < NB_ELTS_V8; c++) {
                OPJ_Dc(i) += (OPJ_Sc(i) + OPJ_Sc(i) + 2) >> 2;
            }
        }
    }
#else
    if (even) {
        OPJ_UINT32 c;
        if (height > 1) {
            OPJ_UINT32 i;
            for (i = 0; i + 1 < sn; i++) {
                for (c = 0; c < NB_ELTS_V8; c++) {
                    OPJ_Dc(i) -= (OPJ_Sc(i) + OPJ_Sc(i + 1)) >> 1;
                }
            }
            if (((height) % 2) == 0) {
                for (c = 0; c < NB_ELTS_V8; c++) {
                    OPJ_Dc(i) -= OPJ_Sc(i);
                }
            }
            for (c = 0; c < NB_ELTS_V8; c++) {
                OPJ_Sc(0) += (OPJ_Dc(0) + OPJ_Dc(0) + 2) >> 2;
            }
            for (i = 1; i < dn; i++) {
                for (c = 0; c < NB_ELTS_V8; c++) {
                    OPJ_Sc(i) += (OPJ_Dc(i - 1) + OPJ_Dc(i) + 2) >> 2;
                }
            }
            if (((height) % 2) == 1) {
                for (c = 0; c < NB_ELTS_V8; c++) {
                    OPJ_Sc(i) += (OPJ_Dc(i - 1) + OPJ_Dc(i - 1) + 2) >> 2;
                }
            }
        }
    } else {
        OPJ_UINT32 c;
        if (height == 1) {
            for (c = 0; c < NB_ELTS_V8; c++) {
                OPJ_Sc(0) *= 2;
            }
        } else {
            OPJ_UINT32 i;
            for (c = 0; c < NB_ELTS_V8; c++) {
                OPJ_Sc(0) -= OPJ_Dc(0);
            }
            for (i = 1; i < sn; i++) {
                for (c = 0; c < NB_ELTS_V8; c++) {
                    OPJ_Sc(i) -= (OPJ_Dc(i) + OPJ_Dc(i - 1)) >> 1;
                }
            }
            if (((height) % 2) == 1) {
                for (c = 0; c < NB_ELTS_V8; c++) {
                    OPJ_Sc(i) -= OPJ_Dc(i - 1);
                }
            }
            for (i = 0; i + 1 < dn; i++) {
                for (c = 0; c < NB_ELTS_V8; c++) {
                    OPJ_Dc(i) += (OPJ_Sc(i) + OPJ_Sc(i + 1) + 2) >> 2;
                }
            }
            if (((height) % 2) == 0) {
                for (c = 0; c < NB_ELTS_V8; c++) {
                    OPJ_Dc(i) += (OPJ_Sc(i) + OPJ_Sc(i) + 2) >> 2;
                }
            }
        }
    }
#endif

    if (cols == NB_ELTS_V8) {
        opj_dwt_deinterleave_v_cols(tmp, array, (OPJ_INT32)dn, (OPJ_INT32)sn,
                                    stride_width, even ? 0 : 1, NB_ELTS_V8);
    } else {
        opj_dwt_deinterleave_v_cols(tmp, array, (OPJ_INT32)dn, (OPJ_INT32)sn,
                                    stride_width, even ? 0 : 1, cols);
    }
}

static void opj_v8dwt_encode_step1(OPJ_FLOAT32* fw,
                                   OPJ_UINT32 end,
                                   const OPJ_FLOAT32 cst)
{
    OPJ_UINT32 i;
#ifdef __SSE__
    __m128* vw = (__m128*) fw;
    const __m128 vcst = _mm_set1_ps(cst);
    for (i = 0; i < end; ++i) {
        vw[0] = _mm_mul_ps(vw[0], vcst);
        vw[1] = _mm_mul_ps(vw[1], vcst);
        vw += 2 * (NB_ELTS_V8 * sizeof(OPJ_FLOAT32) / sizeof(__m128));
    }
#else
    OPJ_UINT32 c;
    for (i = 0; i < end; ++i) {
        for (c = 0; c < NB_ELTS_V8; c++) {
            fw[i * 2 * NB_ELTS_V8 + c] *= cst;
        }
    }
#endif
}

static void opj_v8dwt_encode_step2(OPJ_FLOAT32* fl, OPJ_FLOAT32* fw,
                                   OPJ_UINT32 end,
                                   OPJ_UINT32 m,
                                   OPJ_FLOAT32 cst)
{
    OPJ_UINT32 i;
    OPJ_UINT32 imax = opj_uint_min(end, m);
#ifdef __SSE__
    __m128* vw = (__m128*) fw;
    __m128 vcst = _mm_set1_ps(cst);
    if (imax > 0) {
        __m128* vl = (__m128*) fl;
        vw[-2] = _mm_add_ps(vw[-2], _mm_mul_ps(_mm_add_ps(vl[0], vw[0]), vcst));
        vw[-1] = _mm_add_ps(vw[-1], _mm_mul_ps(_mm_add_ps(vl[1], vw[1]), vcst));
        vw += 2 * (NB_ELTS_V8 * sizeof(OPJ_FLOAT32) / sizeof(__m128));
        i = 1;

        for (; i < imax; ++i) {
            vw[-2] = _mm_add_ps(vw[-2], _mm_mul_ps(_mm_add_ps(vw[-4], vw[0]), vcst));
            vw[-1] = _mm_add_ps(vw[-1], _mm_mul_ps(_mm_add_ps(vw[-3], vw[1]), vcst));
            vw += 2 * (NB_ELTS_V8 * sizeof(OPJ_FLOAT32) / sizeof(__m128));
        }
    }
    if (m < end) {
        assert(m + 1 == end);
        vcst = _mm_add_ps(vcst, vcst);
        vw[-2] = _mm_add_ps(vw[-2], _mm_mul_ps(vw[-4], vcst));
        vw[-1] = _mm_add_ps(vw[-1], _mm_mul_ps(vw[-3], vcst));
    }
#else
    OPJ_INT32 c;
    if (imax > 0) {
        for (c = 0; c < NB_ELTS_V8; c++) {
            fw[-1 * NB_ELTS_V8 + c] += (fl[0 * NB_ELTS_V8 + c] + fw[0 * NB_ELTS_V8 + c]) *
                                       cst;
        }
        fw += 2 * NB_ELTS_V8;
        i = 1;
        for (; i < imax; ++i) {
            for (c = 0; c < NB_ELTS_V8; c++) {
                fw[-1 * NB_ELTS_V8 + c] += (fw[-2 * NB_ELTS_V8 + c] + fw[0 * NB_ELTS_V8 + c]) *
                                           cst;
            }
            fw += 2 * NB_ELTS_V8;
        }
    }
    if (m < end) {
        assert(m + 1 == end);
        for (c = 0; c < NB_ELTS_V8; c++) {
            fw[-1 * NB_ELTS_V8 + c] += (2 * fw[-2 * NB_ELTS_V8 + c]) * cst;
        }
    }
#endif
}

/* Forward 9-7 transform, for the vertical pass, processing cols columns */
/* where cols <= NB_ELTS_V8 */
static void opj_dwt_encode_and_deinterleave_v_real(
    void *arrayIn,
    void *tmpIn,
    OPJ_UINT32 height,
    OPJ_BOOL even,
    OPJ_UINT32 stride_width,
    OPJ_UINT32 cols)
{
    OPJ_FLOAT32* OPJ_RESTRICT array = (OPJ_FLOAT32 * OPJ_RESTRICT)arrayIn;
    OPJ_FLOAT32* OPJ_RESTRICT tmp = (OPJ_FLOAT32 * OPJ_RESTRICT)tmpIn;
    const OPJ_INT32 sn = (OPJ_INT32)((height + (even ? 1 : 0)) >> 1);
    const OPJ_INT32 dn = (OPJ_INT32)(height - (OPJ_UINT32)sn);
    OPJ_INT32 a, b;

    if (height == 1) {
        return;
    }

    opj_dwt_fetch_cols_vertical_pass(arrayIn, tmpIn, height, stride_width, cols);

    if (even) {
        a = 0;
        b = 1;
    } else {
        a = 1;
        b = 0;
    }
    opj_v8dwt_encode_step2(tmp + a * NB_ELTS_V8,
                           tmp + (b + 1) * NB_ELTS_V8,
                           (OPJ_UINT32)dn,
                           (OPJ_UINT32)opj_int_min(dn, sn - b),
                           opj_dwt_alpha);
    opj_v8dwt_encode_step2(tmp + b * NB_ELTS_V8,
                           tmp + (a + 1) * NB_ELTS_V8,
                           (OPJ_UINT32)sn,
                           (OPJ_UINT32)opj_int_min(sn, dn - a),
                           opj_dwt_beta);
    opj_v8dwt_encode_step2(tmp + a * NB_ELTS_V8,
                           tmp + (b + 1) * NB_ELTS_V8,
                           (OPJ_UINT32)dn,
                           (OPJ_UINT32)opj_int_min(dn, sn - b),
                           opj_dwt_gamma);
    opj_v8dwt_encode_step2(tmp + b * NB_ELTS_V8,
                           tmp + (a + 1) * NB_ELTS_V8,
                           (OPJ_UINT32)sn,
                           (OPJ_UINT32)opj_int_min(sn, dn - a),
                           opj_dwt_delta);
    opj_v8dwt_encode_step1(tmp + b * NB_ELTS_V8, (OPJ_UINT32)dn,
                           opj_K);
    opj_v8dwt_encode_step1(tmp + a * NB_ELTS_V8, (OPJ_UINT32)sn,
                           opj_invK);


    if (cols == NB_ELTS_V8) {
        opj_dwt_deinterleave_v_cols((OPJ_INT32*)tmp,
                                    (OPJ_INT32*)array,
                                    (OPJ_INT32)dn, (OPJ_INT32)sn,
                                    stride_width, even ? 0 : 1, NB_ELTS_V8);
    } else {
        opj_dwt_deinterleave_v_cols((OPJ_INT32*)tmp,
                                    (OPJ_INT32*)array,
                                    (OPJ_INT32)dn, (OPJ_INT32)sn,
                                    stride_width, even ? 0 : 1, cols);
    }
}


/* <summary>                            */
/* Forward 5-3 wavelet transform in 2-D. */
/* </summary>                           */
static INLINE OPJ_BOOL opj_dwt_encode_procedure(opj_thread_pool_t* tp,
        opj_tcd_tilecomp_t * tilec,
        opj_encode_and_deinterleave_v_fnptr_type p_encode_and_deinterleave_v,
        opj_encode_and_deinterleave_h_one_row_fnptr_type
        p_encode_and_deinterleave_h_one_row)
{
    OPJ_INT32 i;
    OPJ_INT32 *bj = 00;
    OPJ_UINT32 w;
    OPJ_INT32 l;




    OPJ_SIZE_T l_data_size;

    opj_tcd_resolution_t * l_cur_res = 0;
    opj_tcd_resolution_t * l_last_res = 0;
    const int num_threads = opj_thread_pool_get_thread_count(tp);
    OPJ_INT32 * OPJ_RESTRICT tiledp = tilec->data;

    w = (OPJ_UINT32)(tilec->x1 - tilec->x0);
    l = (OPJ_INT32)tilec->numresolutions - 1;


    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 / (NB_ELTS_V8 * sizeof(OPJ_INT32)))) {
        /* FIXME event manager error callback */
        return OPJ_FALSE;
    }
    l_data_size *= NB_ELTS_V8 * sizeof(OPJ_INT32);
    bj = (OPJ_INT32*)opj_aligned_32_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_UINT32 j;
        OPJ_UINT32 rw;           /* width of the resolution level computed   */
        OPJ_UINT32 rh;           /* height of the resolution level computed  */
        OPJ_UINT32
        rw1;      /* width of the resolution level once lower than computed one                                       */
        OPJ_UINT32
        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  = (OPJ_UINT32)(l_cur_res->x1 - l_cur_res->x0);
        rh  = (OPJ_UINT32)(l_cur_res->y1 - l_cur_res->y0);
        rw1 = (OPJ_UINT32)(l_last_res->x1 - l_last_res->x0);
        rh1 = (OPJ_UINT32)(l_last_res->y1 - l_last_res->y0);

        cas_row = l_cur_res->x0 & 1;
        cas_col = l_cur_res->y0 & 1;

        sn = (OPJ_INT32)rh1;
        dn = (OPJ_INT32)(rh - rh1);

        /* Perform vertical pass */
        if (num_threads <= 1 || rw < 2 * NB_ELTS_V8) {
            for (j = 0; j + NB_ELTS_V8 - 1 < rw; j += NB_ELTS_V8) {
                p_encode_and_deinterleave_v(tiledp + j,
                                            bj,
                                            rh,
                                            cas_col == 0,
                                            w,
                                            NB_ELTS_V8);
            }
            if (j < rw) {
                p_encode_and_deinterleave_v(tiledp + j,
                                            bj,
                                            rh,
                                            cas_col == 0,
                                            w,
                                            rw - j);
            }
        }  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) / NB_ELTS_V8) * NB_ELTS_V8;

            for (j = 0; j < num_jobs; j++) {
                opj_dwt_encode_v_job_t* job;

                job = (opj_dwt_encode_v_job_t*) opj_malloc(sizeof(opj_dwt_encode_v_job_t));
                if (!job) {
                    opj_thread_pool_wait_completion(tp, 0);
                    opj_aligned_free(bj);
                    return OPJ_FALSE;
                }
                job->v.mem = (OPJ_INT32*)opj_aligned_32_malloc(l_data_size);
                if (!job->v.mem) {
                    opj_thread_pool_wait_completion(tp, 0);
                    opj_free(job);
                    opj_aligned_free(bj);
                    return OPJ_FALSE;
                }
                job->v.dn = dn;
                job->v.sn = sn;
                job->v.cas = cas_col;
                job->rh = rh;
                job->w = w;
                job->tiledp = tiledp;
                job->min_j = j * step_j;
                job->max_j = (j + 1 == num_jobs) ? rw : (j + 1) * step_j;
                job->p_encode_and_deinterleave_v = p_encode_and_deinterleave_v;
                opj_thread_pool_submit_job(tp, opj_dwt_encode_v_func, job);
            }
            opj_thread_pool_wait_completion(tp, 0);
        }

        sn = (OPJ_INT32)rw1;
        dn = (OPJ_INT32)(rw - rw1);

        /* Perform horizontal pass */
        if (num_threads <= 1 || rh <= 1) {
            for (j = 0; j < rh; j++) {
                OPJ_INT32* OPJ_RESTRICT aj = tiledp + j * w;
                (*p_encode_and_deinterleave_h_one_row)(aj, bj, rw,
                                                       cas_row == 0 ? OPJ_TRUE : OPJ_FALSE);
            }
        }  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_dwt_encode_h_job_t* job;

                job = (opj_dwt_encode_h_job_t*) opj_malloc(sizeof(opj_dwt_encode_h_job_t));
                if (!job) {
                    opj_thread_pool_wait_completion(tp, 0);
                    opj_aligned_free(bj);
                    return OPJ_FALSE;
                }
                job->h.mem = (OPJ_INT32*)opj_aligned_32_malloc(l_data_size);
                if (!job->h.mem) {
                    opj_thread_pool_wait_completion(tp, 0);
                    opj_free(job);
                    opj_aligned_free(bj);
                    return OPJ_FALSE;
                }
                job->h.dn = dn;
                job->h.sn = sn;
                job->h.cas = cas_row;
                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->p_function = p_encode_and_deinterleave_h_one_row;
                opj_thread_pool_submit_job(tp, opj_dwt_encode_h_func, job);
            }
            opj_thread_pool_wait_completion(tp, 0);
        }

        l_cur_res = l_last_res;

        --l_last_res;
    }

    opj_aligned_free(bj);
    return OPJ_TRUE;
}

/* Forward 5-3 wavelet transform in 2-D. */
/* </summary>                           */
OPJ_BOOL opj_dwt_encode(opj_tcd_t *p_tcd,
                        opj_tcd_tilecomp_t * tilec)
{
    return opj_dwt_encode_procedure(p_tcd->thread_pool, tilec,
                                    opj_dwt_encode_and_deinterleave_v,
                                    opj_dwt_encode_and_deinterleave_h_one_row);
}

/* <summary>                            */
/* Inverse 5-3 wavelet transform in 2-D. */
/* </summary>                           */
OPJ_BOOL opj_dwt_decode(opj_tcd_t *p_tcd, opj_tcd_tilecomp_t* tilec,
                        OPJ_UINT32 numres)
{
    if (p_tcd->whole_tile_decoding) {
        return opj_dwt_decode_tile(p_tcd->thread_pool, tilec, numres);
    } else {
        return opj_dwt_decode_partial_tile(tilec, numres);
    }
}
















/* <summary>                */
/* Get norm of 5-3 wavelet. */
/* </summary>               */
OPJ_FLOAT64 opj_dwt_getnorm(OPJ_UINT32 level, OPJ_UINT32 orient)
{
    /* FIXME ! This is just a band-aid to avoid a buffer overflow */
    /* but the array should really be extended up to 33 resolution levels */
    /* See https://github.com/uclouvain/openjpeg/issues/493 */
    if (orient == 0 && level >= 10) {
        level = 9;
    } else if (orient > 0 && level >= 9) {
        level = 8;
    }
    return opj_dwt_norms[orient][level];
}

/* <summary>                             */
/* Forward 9-7 wavelet transform in 2-D. */
/* </summary>                            */
OPJ_BOOL opj_dwt_encode_real(opj_tcd_t *p_tcd,
                             opj_tcd_tilecomp_t * tilec)
{
    return opj_dwt_encode_procedure(p_tcd->thread_pool, tilec,

                                    opj_dwt_encode_and_deinterleave_v_real,



                                    opj_dwt_encode_and_deinterleave_h_one_row_real);



}

/* <summary>                */
/* Get norm of 9-7 wavelet. */
/* </summary>               */
OPJ_FLOAT64 opj_dwt_getnorm_real(OPJ_UINT32 level, OPJ_UINT32 orient)
{

        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_getnorm_real(level, orient);
            stepsize = (1 << (gain)) / norm;
        }
        opj_dwt_encode_stepsize((OPJ_INT32) floor(stepsize * 8192.0),
                                (OPJ_INT32)(prec + gain), &tccp->stepsizes[bandno]);
    }
}

typedef struct {
    opj_dwt_t h;
    OPJ_UINT32 rw;
    OPJ_UINT32 w;
    OPJ_INT32 * OPJ_RESTRICT tiledp;
    OPJ_UINT32 min_j;
    OPJ_UINT32 max_j;
} opj_dwt_decode_h_job_t;

static void opj_dwt_decode_h_func(void* user_data, opj_tls_t* tls)
{
    OPJ_UINT32 j;
    opj_dwt_decode_h_job_t* job;
    (void)tls;

    job = (opj_dwt_decode_h_job_t*)user_data;
    for (j = job->min_j; j < job->max_j; j++) {
        opj_idwt53_h(&job->h, &job->tiledp[j * job->w]);
    }

    opj_aligned_free(job->h.mem);
    opj_free(job);
}

typedef struct {
    opj_dwt_t v;
    OPJ_UINT32 rh;
    OPJ_UINT32 w;
    OPJ_INT32 * OPJ_RESTRICT tiledp;
    OPJ_UINT32 min_j;
    OPJ_UINT32 max_j;
} opj_dwt_decode_v_job_t;

static void opj_dwt_decode_v_func(void* user_data, opj_tls_t* tls)
{
    OPJ_UINT32 j;
    opj_dwt_decode_v_job_t* job;
    (void)tls;

    job = (opj_dwt_decode_v_job_t*)user_data;
    for (j = job->min_j; j + PARALLEL_COLS_53 <= job->max_j;
            j += PARALLEL_COLS_53) {
        opj_idwt53_v(&job->v, &job->tiledp[j], (OPJ_SIZE_T)job->w,
                     PARALLEL_COLS_53);
    }
    if (j < job->max_j)
        opj_idwt53_v(&job->v, &job->tiledp[j], (OPJ_SIZE_T)job->w,

            if (rh < num_jobs) {
                num_jobs = rh;
            }
            step_j = (rh / num_jobs);

            for (j = 0; j < num_jobs; j++) {
                opj_dwt_decode_h_job_t* job;

                job = (opj_dwt_decode_h_job_t*) opj_malloc(sizeof(opj_dwt_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);

            if (rw < num_jobs) {
                num_jobs = rw;
            }
            step_j = (rw / num_jobs);

            for (j = 0; j < num_jobs; j++) {
                opj_dwt_decode_v_job_t* job;

                job = (opj_dwt_decode_v_job_t*) opj_malloc(sizeof(opj_dwt_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);

        assert(ret);
        OPJ_UNUSED(ret);
    }
    opj_sparse_array_int32_free(sa);
    return OPJ_TRUE;
}

static void opj_v8dwt_interleave_h(opj_v8dwt_t* OPJ_RESTRICT dwt,
                                   OPJ_FLOAT32* OPJ_RESTRICT a,
                                   OPJ_UINT32 width,
                                   OPJ_UINT32 remaining_height)
{
    OPJ_FLOAT32* OPJ_RESTRICT bi = (OPJ_FLOAT32*)(dwt->wavelet + dwt->cas);
    OPJ_UINT32 i, k;
    OPJ_UINT32 x0 = dwt->win_l_x0;
    OPJ_UINT32 x1 = dwt->win_l_x1;

    for (k = 0; k < 2; ++k) {
        if (remaining_height >= NB_ELTS_V8 && ((OPJ_SIZE_T) a & 0x0f) == 0 &&
                ((OPJ_SIZE_T) bi & 0x0f) == 0) {
            /* Fast code path */
            for (i = x0; i < x1; ++i) {
                OPJ_UINT32 j = i;
                OPJ_FLOAT32* OPJ_RESTRICT dst = bi + i * 2 * NB_ELTS_V8;
                dst[0] = a[j];
                j += width;
                dst[1] = a[j];
                j += width;
                dst[2] = a[j];
                j += width;
                dst[3] = a[j];
                j += width;
                dst[4] = a[j];
                j += width;
                dst[5] = a[j];
                j += width;
                dst[6] = a[j];
                j += width;
                dst[7] = a[j];
            }
        } else {
            /* Slow code path */
            for (i = x0; i < x1; ++i) {
                OPJ_UINT32 j = i;
                OPJ_FLOAT32* OPJ_RESTRICT dst = bi + i * 2 * NB_ELTS_V8;
                dst[0] = a[j];
                j += width;
                if (remaining_height == 1) {
                    continue;
                }
                dst[1] = a[j];
                j += width;
                if (remaining_height == 2) {
                    continue;
                }
                dst[2] = a[j];
                j += width;
                if (remaining_height == 3) {
                    continue;
                }
                dst[3] = a[j];
                j += width;
                if (remaining_height == 4) {
                    continue;
                }
                dst[4] = a[j];
                j += width;
                if (remaining_height == 5) {
                    continue;
                }
                dst[5] = a[j];
                j += width;
                if (remaining_height == 6) {
                    continue;
                }
                dst[6] = a[j];
                j += width;
                if (remaining_height == 7) {
                    continue;
                }
                dst[7] = a[j];
            }
        }

        bi = (OPJ_FLOAT32*)(dwt->wavelet + 1 - dwt->cas);
        a += dwt->sn;
        x0 = dwt->win_h_x0;
        x1 = dwt->win_h_x1;
    }
}

static void opj_v8dwt_interleave_partial_h(opj_v8dwt_t* dwt,
        opj_sparse_array_int32_t* sa,
        OPJ_UINT32 sa_line,
        OPJ_UINT32 remaining_height)
{
    OPJ_UINT32 i;
    for (i = 0; i < remaining_height; i++) {
        OPJ_BOOL ret;
        ret = opj_sparse_array_int32_read(sa,
                                          dwt->win_l_x0, sa_line + i,
                                          dwt->win_l_x1, sa_line + i + 1,
                                          /* Nasty cast from float* to int32* */
                                          (OPJ_INT32*)(dwt->wavelet + dwt->cas + 2 * dwt->win_l_x0) + i,
                                          2 * NB_ELTS_V8, 0, OPJ_TRUE);
        assert(ret);
        ret = opj_sparse_array_int32_read(sa,
                                          (OPJ_UINT32)dwt->sn + dwt->win_h_x0, sa_line + i,
                                          (OPJ_UINT32)dwt->sn + dwt->win_h_x1, sa_line + i + 1,
                                          /* Nasty cast from float* to int32* */
                                          (OPJ_INT32*)(dwt->wavelet + 1 - dwt->cas + 2 * dwt->win_h_x0) + i,
                                          2 * NB_ELTS_V8, 0, OPJ_TRUE);
        assert(ret);
        OPJ_UNUSED(ret);
    }
}

static INLINE void opj_v8dwt_interleave_v(opj_v8dwt_t* OPJ_RESTRICT dwt,
        OPJ_FLOAT32* OPJ_RESTRICT a,
        OPJ_UINT32 width,
        OPJ_UINT32 nb_elts_read)
{
    opj_v8_t* OPJ_RESTRICT bi = dwt->wavelet + dwt->cas;
    OPJ_UINT32 i;

    for (i = dwt->win_l_x0; i < dwt->win_l_x1; ++i) {
        memcpy(&bi[i * 2], &a[i * (OPJ_SIZE_T)width],
               (OPJ_SIZE_T)nb_elts_read * sizeof(OPJ_FLOAT32));
    }

    a += (OPJ_UINT32)dwt->sn * (OPJ_SIZE_T)width;
    bi = dwt->wavelet + 1 - dwt->cas;

    for (i = dwt->win_h_x0; i < dwt->win_h_x1; ++i) {
        memcpy(&bi[i * 2], &a[i * (OPJ_SIZE_T)width],
               (OPJ_SIZE_T)nb_elts_read * sizeof(OPJ_FLOAT32));
    }
}

static void opj_v8dwt_interleave_partial_v(opj_v8dwt_t* OPJ_RESTRICT dwt,
        opj_sparse_array_int32_t* sa,
        OPJ_UINT32 sa_col,
        OPJ_UINT32 nb_elts_read)
{
    OPJ_BOOL ret;
    ret = opj_sparse_array_int32_read(sa,
                                      sa_col, dwt->win_l_x0,
                                      sa_col + nb_elts_read, dwt->win_l_x1,
                                      (OPJ_INT32*)(dwt->wavelet + dwt->cas + 2 * dwt->win_l_x0),
                                      1, 2 * NB_ELTS_V8, OPJ_TRUE);
    assert(ret);
    ret = opj_sparse_array_int32_read(sa,
                                      sa_col, (OPJ_UINT32)dwt->sn + dwt->win_h_x0,
                                      sa_col + nb_elts_read, (OPJ_UINT32)dwt->sn + dwt->win_h_x1,
                                      (OPJ_INT32*)(dwt->wavelet + 1 - dwt->cas + 2 * dwt->win_h_x0),
                                      1, 2 * NB_ELTS_V8, OPJ_TRUE);
    assert(ret);
    OPJ_UNUSED(ret);
}

#ifdef __SSE__

static void opj_v8dwt_decode_step1_sse(opj_v8_t* w,
                                       OPJ_UINT32 start,
                                       OPJ_UINT32 end,
                                       const __m128 c)
{
    __m128* OPJ_RESTRICT vw = (__m128*) w;
    OPJ_UINT32 i = start;
    /* To be adapted if NB_ELTS_V8 changes */
    vw += 4 * start;
    /* Note: attempt at loop unrolling x2 doesn't help */









    for (; i < end; ++i, vw += 4) {
        vw[0] = _mm_mul_ps(vw[0], c);
        vw[1] = _mm_mul_ps(vw[1], c);
    }
}

static void opj_v8dwt_decode_step2_sse(opj_v8_t* l, opj_v8_t* w,
                                       OPJ_UINT32 start,
                                       OPJ_UINT32 end,
                                       OPJ_UINT32 m,
                                       __m128 c)
{
    __m128* OPJ_RESTRICT vl = (__m128*) l;
    __m128* OPJ_RESTRICT vw = (__m128*) w;
    /* To be adapted if NB_ELTS_V8 changes */
    OPJ_UINT32 i;
    OPJ_UINT32 imax = opj_uint_min(end, m);

    if (start == 0) {
        if (imax >= 1) {
            vw[-2] = _mm_add_ps(vw[-2], _mm_mul_ps(_mm_add_ps(vl[0], vw[0]), c));
            vw[-1] = _mm_add_ps(vw[-1], _mm_mul_ps(_mm_add_ps(vl[1], vw[1]), c));
            vw += 4;
            start = 1;
        }
    } else {
        vw += start * 4;

    }

    i = start;

    /* Note: attempt at loop unrolling x2 doesn't help */


















    for (; i < imax; ++i) {
        vw[-2] = _mm_add_ps(vw[-2], _mm_mul_ps(_mm_add_ps(vw[-4], vw[0]), c));

        vw[-1] = _mm_add_ps(vw[-1], _mm_mul_ps(_mm_add_ps(vw[-3], vw[1]), c));

        vw += 4;
    }
    if (m < end) {
        assert(m + 1 == end);
        c = _mm_add_ps(c, c);
        vw[-2] = _mm_add_ps(vw[-2], _mm_mul_ps(c, vw[-4]));
        vw[-1] = _mm_add_ps(vw[-1], _mm_mul_ps(c, vw[-3]));
    }
}

#else

static void opj_v8dwt_decode_step1(opj_v8_t* w,
                                   OPJ_UINT32 start,
                                   OPJ_UINT32 end,
                                   const OPJ_FLOAT32 c)
{
    OPJ_FLOAT32* OPJ_RESTRICT fw = (OPJ_FLOAT32*) w;
    OPJ_UINT32 i;
    /* To be adapted if NB_ELTS_V8 changes */
    for (i = start; i < end; ++i) {
        fw[i * 2 * 8    ] = fw[i * 2 * 8    ] * c;
        fw[i * 2 * 8 + 1] = fw[i * 2 * 8 + 1] * c;
        fw[i * 2 * 8 + 2] = fw[i * 2 * 8 + 2] * c;
        fw[i * 2 * 8 + 3] = fw[i * 2 * 8 + 3] * c;
        fw[i * 2 * 8 + 4] = fw[i * 2 * 8 + 4] * c;
        fw[i * 2 * 8 + 5] = fw[i * 2 * 8 + 5] * c;
        fw[i * 2 * 8 + 6] = fw[i * 2 * 8 + 6] * c;
        fw[i * 2 * 8 + 7] = fw[i * 2 * 8 + 7] * c;
    }
}

static void opj_v8dwt_decode_step2(opj_v8_t* l, opj_v8_t* w,
                                   OPJ_UINT32 start,
                                   OPJ_UINT32 end,
                                   OPJ_UINT32 m,
                                   OPJ_FLOAT32 c)
{
    OPJ_FLOAT32* fl = (OPJ_FLOAT32*) l;
    OPJ_FLOAT32* fw = (OPJ_FLOAT32*) w;
    OPJ_UINT32 i;
    OPJ_UINT32 imax = opj_uint_min(end, m);
    if (start > 0) {
        fw += 2 * NB_ELTS_V8 * start;
        fl = fw - 2 * NB_ELTS_V8;
    }
    /* To be adapted if NB_ELTS_V8 changes */
    for (i = start; i < imax; ++i) {




        fw[-8] = fw[-8] + ((fl[0] + fw[0]) * c);
        fw[-7] = fw[-7] + ((fl[1] + fw[1]) * c);
        fw[-6] = fw[-6] + ((fl[2] + fw[2]) * c);
        fw[-5] = fw[-5] + ((fl[3] + fw[3]) * c);




        fw[-4] = fw[-4] + ((fl[4] + fw[4]) * c);
        fw[-3] = fw[-3] + ((fl[5] + fw[5]) * c);
        fw[-2] = fw[-2] + ((fl[6] + fw[6]) * c);
        fw[-1] = fw[-1] + ((fl[7] + fw[7]) * c);
        fl = fw;
        fw += 2 * NB_ELTS_V8;
    }
    if (m < end) {
        assert(m + 1 == end);
        c += c;
        fw[-8] = fw[-8] + fl[0] * c;
        fw[-7] = fw[-7] + fl[1] * c;
        fw[-6] = fw[-6] + fl[2] * c;
        fw[-5] = fw[-5] + fl[3] * c;
        fw[-4] = fw[-4] + fl[4] * c;
        fw[-3] = fw[-3] + fl[5] * c;
        fw[-2] = fw[-2] + fl[6] * c;
        fw[-1] = fw[-1] + fl[7] * c;
    }
}

#endif

/* <summary>                             */
/* Inverse 9-7 wavelet transform in 1-D. */
/* </summary>                            */
static void opj_v8dwt_decode(opj_v8dwt_t* OPJ_RESTRICT dwt)
{
    OPJ_INT32 a, b;
    /* BUG_WEIRD_TWO_INVK (look for this identifier in tcd.c) */
    /* Historic value for 2 / opj_invK */
    /* Normally, we should use invK, but if we do so, we have failures in the */
    /* conformance test, due to MSE and peak errors significantly higher than */
    /* accepted value */
    /* Due to using two_invK instead of invK, we have to compensate in tcd.c */
    /* the computation of the stepsize for the non LL subbands */
    const float two_invK = 1.625732422f;
    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_v8dwt_decode_step1_sse(dwt->wavelet + a, dwt->win_l_x0, dwt->win_l_x1,
                               _mm_set1_ps(opj_K));
    opj_v8dwt_decode_step1_sse(dwt->wavelet + b, dwt->win_h_x0, dwt->win_h_x1,
                               _mm_set1_ps(two_invK));
    opj_v8dwt_decode_step2_sse(dwt->wavelet + b, dwt->wavelet + a + 1,
                               dwt->win_l_x0, dwt->win_l_x1,
                               (OPJ_UINT32)opj_int_min(dwt->sn, dwt->dn - a),
                               _mm_set1_ps(-opj_dwt_delta));
    opj_v8dwt_decode_step2_sse(dwt->wavelet + a, dwt->wavelet + b + 1,
                               dwt->win_h_x0, dwt->win_h_x1,
                               (OPJ_UINT32)opj_int_min(dwt->dn, dwt->sn - b),
                               _mm_set1_ps(-opj_dwt_gamma));
    opj_v8dwt_decode_step2_sse(dwt->wavelet + b, dwt->wavelet + a + 1,
                               dwt->win_l_x0, dwt->win_l_x1,
                               (OPJ_UINT32)opj_int_min(dwt->sn, dwt->dn - a),
                               _mm_set1_ps(-opj_dwt_beta));
    opj_v8dwt_decode_step2_sse(dwt->wavelet + a, dwt->wavelet + b + 1,
                               dwt->win_h_x0, dwt->win_h_x1,
                               (OPJ_UINT32)opj_int_min(dwt->dn, dwt->sn - b),
                               _mm_set1_ps(-opj_dwt_alpha));
#else
    opj_v8dwt_decode_step1(dwt->wavelet + a, dwt->win_l_x0, dwt->win_l_x1,
                           opj_K);
    opj_v8dwt_decode_step1(dwt->wavelet + b, dwt->win_h_x0, dwt->win_h_x1,
                           two_invK);
    opj_v8dwt_decode_step2(dwt->wavelet + b, dwt->wavelet + a + 1,
                           dwt->win_l_x0, dwt->win_l_x1,
                           (OPJ_UINT32)opj_int_min(dwt->sn, dwt->dn - a),
                           -opj_dwt_delta);
    opj_v8dwt_decode_step2(dwt->wavelet + a, dwt->wavelet + b + 1,
                           dwt->win_h_x0, dwt->win_h_x1,
                           (OPJ_UINT32)opj_int_min(dwt->dn, dwt->sn - b),
                           -opj_dwt_gamma);
    opj_v8dwt_decode_step2(dwt->wavelet + b, dwt->wavelet + a + 1,
                           dwt->win_l_x0, dwt->win_l_x1,
                           (OPJ_UINT32)opj_int_min(dwt->sn, dwt->dn - a),
                           -opj_dwt_beta);
    opj_v8dwt_decode_step2(dwt->wavelet + a, dwt->wavelet + b + 1,
                           dwt->win_h_x0, dwt->win_h_x1,
                           (OPJ_UINT32)opj_int_min(dwt->dn, dwt->sn - b),
                           -opj_dwt_alpha);
#endif
}

typedef struct {
    opj_v8dwt_t h;
    OPJ_UINT32 rw;
    OPJ_UINT32 w;
    OPJ_FLOAT32 * OPJ_RESTRICT aj;
    OPJ_UINT32 nb_rows;
} opj_dwt97_decode_h_job_t;

static void opj_dwt97_decode_h_func(void* user_data, opj_tls_t* tls)
{
    OPJ_UINT32 j;
    opj_dwt97_decode_h_job_t* job;
    OPJ_FLOAT32 * OPJ_RESTRICT aj;
    OPJ_UINT32 w;
    (void)tls;

    job = (opj_dwt97_decode_h_job_t*)user_data;
    w = job->w;

    assert((job->nb_rows % NB_ELTS_V8) == 0);

    aj = job->aj;
    for (j = 0; j + NB_ELTS_V8 <= job->nb_rows; j += NB_ELTS_V8) {
        OPJ_UINT32 k;
        opj_v8dwt_interleave_h(&job->h, aj, job->w, NB_ELTS_V8);
        opj_v8dwt_decode(&job->h);

        /* To be adapted if NB_ELTS_V8 changes */
        for (k = 0; k < job->rw; k++) {
            aj[k      ] = job->h.wavelet[k].f[0];
            aj[k + (OPJ_SIZE_T)w  ] = job->h.wavelet[k].f[1];
            aj[k + (OPJ_SIZE_T)w * 2] = job->h.wavelet[k].f[2];
            aj[k + (OPJ_SIZE_T)w * 3] = job->h.wavelet[k].f[3];
        }
        for (k = 0; k < job->rw; k++) {
            aj[k + (OPJ_SIZE_T)w * 4] = job->h.wavelet[k].f[4];
            aj[k + (OPJ_SIZE_T)w * 5] = job->h.wavelet[k].f[5];
            aj[k + (OPJ_SIZE_T)w * 6] = job->h.wavelet[k].f[6];
            aj[k + (OPJ_SIZE_T)w * 7] = job->h.wavelet[k].f[7];
        }

        aj += w * NB_ELTS_V8;
    }

    opj_aligned_free(job->h.wavelet);
    opj_free(job);
}


typedef struct {
    opj_v8dwt_t v;
    OPJ_UINT32 rh;
    OPJ_UINT32 w;
    OPJ_FLOAT32 * OPJ_RESTRICT aj;
    OPJ_UINT32 nb_columns;
} opj_dwt97_decode_v_job_t;

static void opj_dwt97_decode_v_func(void* user_data, opj_tls_t* tls)
{
    OPJ_UINT32 j;
    opj_dwt97_decode_v_job_t* job;
    OPJ_FLOAT32 * OPJ_RESTRICT aj;
    (void)tls;

    job = (opj_dwt97_decode_v_job_t*)user_data;

    assert((job->nb_columns % NB_ELTS_V8) == 0);

    aj = job->aj;
    for (j = 0; j + NB_ELTS_V8 <= job->nb_columns; j += NB_ELTS_V8) {
        OPJ_UINT32 k;

        opj_v8dwt_interleave_v(&job->v, aj, job->w, NB_ELTS_V8);
        opj_v8dwt_decode(&job->v);

        for (k = 0; k < job->rh; ++k) {
            memcpy(&aj[k * (OPJ_SIZE_T)job->w], &job->v.wavelet[k],
                   NB_ELTS_V8 * sizeof(OPJ_FLOAT32));
        }
        aj += NB_ELTS_V8;
    }

    opj_aligned_free(job->v.wavelet);
    opj_free(job);
}


/* <summary>                             */
/* Inverse 9-7 wavelet transform in 2-D. */
/* </summary>                            */
static
OPJ_BOOL opj_dwt_decode_tile_97(opj_thread_pool_t* tp,
                                opj_tcd_tilecomp_t* OPJ_RESTRICT tilec,
                                OPJ_UINT32 numres)
{
    opj_v8dwt_t h;
    opj_v8dwt_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->resolutions[tilec->minimum_num_resolutions -
                                                               1].x1 -
                                tilec->resolutions[tilec->minimum_num_resolutions - 1].x0);

    OPJ_SIZE_T l_data_size;
    const int num_threads = opj_thread_pool_get_thread_count(tp);

    if (numres == 1) {
        return OPJ_TRUE;
    }

    l_data_size = opj_dwt_max_resolution(res, numres);
    /* overflow check */






    if (l_data_size > (SIZE_MAX / sizeof(opj_v8_t))) {
        /* FIXME event manager error callback */
        return OPJ_FALSE;
    }
    h.wavelet = (opj_v8_t*) opj_aligned_malloc(l_data_size * sizeof(opj_v8_t));
    if (!h.wavelet) {
        /* FIXME event manager error callback */
        return OPJ_FALSE;
    }
    v.wavelet = h.wavelet;

    while (--numres) {

        h.dn = (OPJ_INT32)(rw - (OPJ_UINT32)h.sn);
        h.cas = res->x0 % 2;

        h.win_l_x0 = 0;
        h.win_l_x1 = (OPJ_UINT32)h.sn;
        h.win_h_x0 = 0;
        h.win_h_x1 = (OPJ_UINT32)h.dn;

        if (num_threads <= 1 || rh < 2 * NB_ELTS_V8) {
            for (j = 0; j + (NB_ELTS_V8 - 1) < rh; j += NB_ELTS_V8) {
                OPJ_UINT32 k;
                opj_v8dwt_interleave_h(&h, aj, w, NB_ELTS_V8);
                opj_v8dwt_decode(&h);

                /* To be adapted if NB_ELTS_V8 changes */
                for (k = 0; k < rw; k++) {
                    aj[k      ] = h.wavelet[k].f[0];
                    aj[k + (OPJ_SIZE_T)w  ] = h.wavelet[k].f[1];
                    aj[k + (OPJ_SIZE_T)w * 2] = h.wavelet[k].f[2];
                    aj[k + (OPJ_SIZE_T)w * 3] = h.wavelet[k].f[3];
                }
                for (k = 0; k < rw; k++) {
                    aj[k + (OPJ_SIZE_T)w * 4] = h.wavelet[k].f[4];
                    aj[k + (OPJ_SIZE_T)w * 5] = h.wavelet[k].f[5];
                    aj[k + (OPJ_SIZE_T)w * 6] = h.wavelet[k].f[6];
                    aj[k + (OPJ_SIZE_T)w * 7] = h.wavelet[k].f[7];
                }

                aj += w * NB_ELTS_V8;
            }
        } else {
            OPJ_UINT32 num_jobs = (OPJ_UINT32)num_threads;
            OPJ_UINT32 step_j;

            if ((rh / NB_ELTS_V8) < num_jobs) {
                num_jobs = rh / NB_ELTS_V8;
            }
            step_j = ((rh / num_jobs) / NB_ELTS_V8) * NB_ELTS_V8;
            for (j = 0; j < num_jobs; j++) {
                opj_dwt97_decode_h_job_t* job;

                job = (opj_dwt97_decode_h_job_t*) opj_malloc(sizeof(opj_dwt97_decode_h_job_t));
                if (!job) {
                    opj_thread_pool_wait_completion(tp, 0);
                    opj_aligned_free(h.wavelet);
                    return OPJ_FALSE;
                }
                job->h.wavelet = (opj_v8_t*)opj_aligned_malloc(l_data_size * sizeof(opj_v8_t));
                if (!job->h.wavelet) {
                    opj_thread_pool_wait_completion(tp, 0);
                    opj_free(job);
                    opj_aligned_free(h.wavelet);
                    return OPJ_FALSE;
                }
                job->h.dn = h.dn;
                job->h.sn = h.sn;
                job->h.cas = h.cas;
                job->h.win_l_x0 = h.win_l_x0;
                job->h.win_l_x1 = h.win_l_x1;
                job->h.win_h_x0 = h.win_h_x0;
                job->h.win_h_x1 = h.win_h_x1;
                job->rw = rw;
                job->w = w;
                job->aj = aj;
                job->nb_rows = (j + 1 == num_jobs) ? (rh & (OPJ_UINT32)~
                                                      (NB_ELTS_V8 - 1)) - j * step_j : step_j;
                aj += w * job->nb_rows;
                opj_thread_pool_submit_job(tp, opj_dwt97_decode_h_func, job);
            }
            opj_thread_pool_wait_completion(tp, 0);
            j = rh & (OPJ_UINT32)~(NB_ELTS_V8 - 1);
        }

        if (j < rh) {
            OPJ_UINT32 k;
            opj_v8dwt_interleave_h(&h, aj, w, rh - j);
            opj_v8dwt_decode(&h);
            for (k = 0; k < rw; k++) {
                OPJ_UINT32 l;
                for (l = 0; l < rh - j; l++) {







                    aj[k + (OPJ_SIZE_T)w  * l ] = h.wavelet[k].f[l];
                }
            }
        }

        v.dn = (OPJ_INT32)(rh - (OPJ_UINT32)v.sn);
        v.cas = res->y0 % 2;
        v.win_l_x0 = 0;
        v.win_l_x1 = (OPJ_UINT32)v.sn;
        v.win_h_x0 = 0;
        v.win_h_x1 = (OPJ_UINT32)v.dn;

        aj = (OPJ_FLOAT32*) tilec->data;
        if (num_threads <= 1 || rw < 2 * NB_ELTS_V8) {
            for (j = rw; j > (NB_ELTS_V8 - 1); j -= NB_ELTS_V8) {
                OPJ_UINT32 k;

                opj_v8dwt_interleave_v(&v, aj, w, NB_ELTS_V8);
                opj_v8dwt_decode(&v);

                for (k = 0; k < rh; ++k) {
                    memcpy(&aj[k * (OPJ_SIZE_T)w], &v.wavelet[k], NB_ELTS_V8 * sizeof(OPJ_FLOAT32));
                }
                aj += NB_ELTS_V8;
            }
        } else {
            /* "bench_dwt -I" shows that scaling is poor, likely due to RAM
                transfer being the limiting factor. So limit the number of
                threads.
             */
            OPJ_UINT32 num_jobs = opj_uint_max((OPJ_UINT32)num_threads / 2, 2U);
            OPJ_UINT32 step_j;

            if ((rw / NB_ELTS_V8) < num_jobs) {
                num_jobs = rw / NB_ELTS_V8;
            }
            step_j = ((rw / num_jobs) / NB_ELTS_V8) * NB_ELTS_V8;
            for (j = 0; j < num_jobs; j++) {
                opj_dwt97_decode_v_job_t* job;

                job = (opj_dwt97_decode_v_job_t*) opj_malloc(sizeof(opj_dwt97_decode_v_job_t));
                if (!job) {
                    opj_thread_pool_wait_completion(tp, 0);
                    opj_aligned_free(h.wavelet);
                    return OPJ_FALSE;
                }
                job->v.wavelet = (opj_v8_t*)opj_aligned_malloc(l_data_size * sizeof(opj_v8_t));
                if (!job->v.wavelet) {
                    opj_thread_pool_wait_completion(tp, 0);
                    opj_free(job);
                    opj_aligned_free(h.wavelet);
                    return OPJ_FALSE;
                }
                job->v.dn = v.dn;
                job->v.sn = v.sn;
                job->v.cas = v.cas;
                job->v.win_l_x0 = v.win_l_x0;
                job->v.win_l_x1 = v.win_l_x1;
                job->v.win_h_x0 = v.win_h_x0;
                job->v.win_h_x1 = v.win_h_x1;
                job->rh = rh;
                job->w = w;
                job->aj = aj;
                job->nb_columns = (j + 1 == num_jobs) ? (rw & (OPJ_UINT32)~
                                  (NB_ELTS_V8 - 1)) - j * step_j : step_j;
                aj += job->nb_columns;
                opj_thread_pool_submit_job(tp, opj_dwt97_decode_v_func, job);
            }
            opj_thread_pool_wait_completion(tp, 0);
        }

        if (rw & (NB_ELTS_V8 - 1)) {
            OPJ_UINT32 k;

            j = rw & (NB_ELTS_V8 - 1);

            opj_v8dwt_interleave_v(&v, aj, w, j);
            opj_v8dwt_decode(&v);

            for (k = 0; k < rh; ++k) {
                memcpy(&aj[k * (OPJ_SIZE_T)w], &v.wavelet[k],
                       (OPJ_SIZE_T)j * sizeof(OPJ_FLOAT32));
            }
        }
    }

    opj_aligned_free(h.wavelet);
    return OPJ_TRUE;
}

static
OPJ_BOOL opj_dwt_decode_partial_97(opj_tcd_tilecomp_t* OPJ_RESTRICT tilec,
                                   OPJ_UINT32 numres)
{
    opj_sparse_array_int32_t* sa;
    opj_v8dwt_t h;
    opj_v8dwt_t v;
    OPJ_UINT32 resno;
    /* This value matches the maximum left/right extension given in tables */
    /* F.2 and F.3 of the standard. Note: in opj_tcd_is_subband_area_of_interest() */
    /* we currently use 3. */
    const OPJ_UINT32 filter_width = 4U;

    opj_tcd_resolution_t* tr = tilec->resolutions;

        OPJ_UNUSED(ret);
        opj_sparse_array_int32_free(sa);
        return OPJ_TRUE;
    }

    l_data_size = opj_dwt_max_resolution(tr, numres);
    /* overflow check */
    if (l_data_size > (SIZE_MAX / sizeof(opj_v8_t))) {
        /* FIXME event manager error callback */
        opj_sparse_array_int32_free(sa);
        return OPJ_FALSE;
    }







    h.wavelet = (opj_v8_t*) opj_aligned_malloc(l_data_size * sizeof(opj_v8_t));
    if (!h.wavelet) {
        /* FIXME event manager error callback */
        opj_sparse_array_int32_free(sa);
        return OPJ_FALSE;
    }
    v.wavelet = h.wavelet;

            win_tr_y1 = opj_uint_min(opj_uint_max(2 * win_lh_y1, 2 * win_ll_y1 + 1), rh);
        }

        h.win_l_x0 = win_ll_x0;
        h.win_l_x1 = win_ll_x1;
        h.win_h_x0 = win_hl_x0;
        h.win_h_x1 = win_hl_x1;
        for (j = 0; j + (NB_ELTS_V8 - 1) < rh; j += NB_ELTS_V8) {
            if ((j + (NB_ELTS_V8 - 1) >= win_ll_y0 && j < win_ll_y1) ||
                    (j + (NB_ELTS_V8 - 1) >= win_lh_y0 + (OPJ_UINT32)v.sn &&
                     j < win_lh_y1 + (OPJ_UINT32)v.sn)) {
                opj_v8dwt_interleave_partial_h(&h, sa, j, opj_uint_min(NB_ELTS_V8, rh - j));
                opj_v8dwt_decode(&h);
                if (!opj_sparse_array_int32_write(sa,
                                                  win_tr_x0, j,
                                                  win_tr_x1, j + NB_ELTS_V8,
                                                  (OPJ_INT32*)&h.wavelet[win_tr_x0].f[0],
                                                  NB_ELTS_V8, 1, OPJ_TRUE)) {
                    /* FIXME event manager error callback */
                    opj_sparse_array_int32_free(sa);
                    opj_aligned_free(h.wavelet);
                    return OPJ_FALSE;
                }
            }
        }

        if (j < rh &&
                ((j + (NB_ELTS_V8 - 1) >= win_ll_y0 && j < win_ll_y1) ||
                 (j + (NB_ELTS_V8 - 1) >= win_lh_y0 + (OPJ_UINT32)v.sn &&
                  j < win_lh_y1 + (OPJ_UINT32)v.sn))) {
            opj_v8dwt_interleave_partial_h(&h, sa, j, rh - j);
            opj_v8dwt_decode(&h);
            if (!opj_sparse_array_int32_write(sa,
                                              win_tr_x0, j,
                                              win_tr_x1, rh,
                                              (OPJ_INT32*)&h.wavelet[win_tr_x0].f[0],
                                              NB_ELTS_V8, 1, OPJ_TRUE)) {
                /* FIXME event manager error callback */
                opj_sparse_array_int32_free(sa);
                opj_aligned_free(h.wavelet);
                return OPJ_FALSE;
            }
        }

        v.win_l_x0 = win_ll_y0;
        v.win_l_x1 = win_ll_y1;
        v.win_h_x0 = win_lh_y0;
        v.win_h_x1 = win_lh_y1;
        for (j = win_tr_x0; j < win_tr_x1; j += NB_ELTS_V8) {
            OPJ_UINT32 nb_elts = opj_uint_min(NB_ELTS_V8, win_tr_x1 - j);

            opj_v8dwt_interleave_partial_v(&v, sa, j, nb_elts);
            opj_v8dwt_decode(&v);

            if (!opj_sparse_array_int32_write(sa,
                                              j, win_tr_y0,
                                              j + nb_elts, win_tr_y1,
                                              (OPJ_INT32*)&h.wavelet[win_tr_y0].f[0],
                                              1, NB_ELTS_V8, OPJ_TRUE)) {
                /* FIXME event manager error callback */
                opj_sparse_array_int32_free(sa);
                opj_aligned_free(h.wavelet);
                return OPJ_FALSE;
            }
        }
    }

OPJ_BOOL opj_dwt_decode_real(opj_tcd_t *p_tcd,
                             opj_tcd_tilecomp_t* OPJ_RESTRICT tilec,
                             OPJ_UINT32 numres)
{
    if (p_tcd->whole_tile_decoding) {
        return opj_dwt_decode_tile_97(p_tcd->thread_pool, tilec, numres);
    } else {
        return opj_dwt_decode_partial_97(tilec, numres);
    }
}

/** @name Exported functions */
/*@{*/
/* ----------------------------------------------------------------------- */
/**
Forward 5-3 wavelet transform in 2-D.
Apply a reversible DWT transform to a component of an image.
@param p_tcd TCD handle
@param tilec Tile component information (current tile)
*/
OPJ_BOOL opj_dwt_encode(opj_tcd_t *p_tcd,
                        opj_tcd_tilecomp_t * tilec);

/**
Inverse 5-3 wavelet transform in 2-D.
Apply a reversible inverse DWT transform to a component of an image.
@param p_tcd TCD handle
@param tilec Tile component information (current tile)
@param numres Number of resolution levels to decode
*/
OPJ_BOOL opj_dwt_decode(opj_tcd_t *p_tcd,
                        opj_tcd_tilecomp_t* tilec,
                        OPJ_UINT32 numres);

/**






Get the norm of a wavelet function of a subband at a specified level for the reversible 5-3 DWT.
@param level Level of the wavelet function
@param orient Band of the wavelet function
@return Returns the norm of the wavelet function
*/
OPJ_FLOAT64 opj_dwt_getnorm(OPJ_UINT32 level, OPJ_UINT32 orient);
/**
Forward 9-7 wavelet transform in 2-D.
Apply an irreversible DWT transform to a component of an image.
@param p_tcd TCD handle
@param tilec Tile component information (current tile)
*/
OPJ_BOOL opj_dwt_encode_real(opj_tcd_t *p_tcd,
                             opj_tcd_tilecomp_t * tilec);
/**
Inverse 9-7 wavelet transform in 2-D.
Apply an irreversible inverse DWT transform to a component of an image.
@param p_tcd TCD handle
@param tilec Tile component information (current tile)
@param numres Number of resolution levels to decode
*/
OPJ_BOOL opj_dwt_decode_real(opj_tcd_t *p_tcd,
                             opj_tcd_tilecomp_t* OPJ_RESTRICT tilec,
                             OPJ_UINT32 numres);

/**






Get the norm of a wavelet function of a subband at a specified level for the irreversible 9-7 DWT
@param level Level of the wavelet function
@param orient Band of the wavelet function
@return Returns the norm of the 9-7 wavelet
*/
OPJ_FLOAT64 opj_dwt_getnorm_real(OPJ_UINT32 level, OPJ_UINT32 orient);
/**

 */
static OPJ_BOOL opj_j2k_setup_header_writing(opj_j2k_t *p_j2k,
        opj_event_mgr_t * p_manager);

static OPJ_BOOL opj_j2k_write_first_tile_part(opj_j2k_t *p_j2k,
        OPJ_BYTE * p_data,
        OPJ_UINT32 * p_data_written,
        OPJ_UINT32 total_data_size,
        opj_stream_private_t *p_stream,
        struct opj_event_mgr * p_manager);

static OPJ_BOOL opj_j2k_write_all_tile_parts(opj_j2k_t *p_j2k,
        OPJ_BYTE * p_data,
        OPJ_UINT32 * p_data_written,
        OPJ_UINT32 total_data_size,
        opj_stream_private_t *p_stream,
        struct opj_event_mgr * p_manager);

/**
 * Gets the offset of the header.
 *
 * @param       p_stream                the stream to write data to.

                                  opj_event_mgr_t * p_manager);

/**
 * Writes the SOT marker (Start of tile-part)
 *
 * @param       p_j2k            J2K codec.
 * @param       p_data           Output buffer
 * @param       total_data_size  Output buffer size
 * @param       p_data_written   Number of bytes written into stream
 * @param       p_stream         the stream to write data to.
 * @param       p_manager        the user event manager.
*/
static OPJ_BOOL opj_j2k_write_sot(opj_j2k_t *p_j2k,
                                  OPJ_BYTE * p_data,
                                  OPJ_UINT32 total_data_size,
                                  OPJ_UINT32 * p_data_written,
                                  const opj_stream_private_t *p_stream,
                                  opj_event_mgr_t * p_manager);

/**
 * Reads values from a SOT marker (Start of tile-part)
 *

static OPJ_BOOL opj_j2k_read_sot(opj_j2k_t *p_j2k,
                                 OPJ_BYTE * p_header_data,
                                 OPJ_UINT32 p_header_size,
                                 opj_event_mgr_t * p_manager);
/**
 * Writes the SOD marker (Start of data)
 *
 * This also writes optional PLT markers (before SOD)
 *
 * @param       p_j2k               J2K codec.
 * @param       p_tile_coder        FIXME DOC
 * @param       p_data              FIXME DOC
 * @param       p_data_written      FIXME DOC
 * @param       total_data_size   FIXME DOC
 * @param       p_stream            the stream to write data to.
 * @param       p_manager           the user event manager.
*/
static OPJ_BOOL opj_j2k_write_sod(opj_j2k_t *p_j2k,
                                  opj_tcd_t * p_tile_coder,
                                  OPJ_BYTE * p_data,
                                  OPJ_UINT32 * p_data_written,
                                  OPJ_UINT32 total_data_size,
                                  const opj_stream_private_t *p_stream,
                                  opj_event_mgr_t * p_manager);

/**
 * Reads a SOD marker (Start Of Data)
 *
 * @param       p_j2k                   the jpeg2000 codec.

                                  opj_event_mgr_t * p_manager);

/**
 * Checks the progression order changes values. Tells of the poc given as input are valid.
 * A nice message is outputted at errors.
 *
 * @param       p_pocs                  the progression order changes.
 * @param       tileno                  the tile number of interest
 * @param       p_nb_pocs               the number of progression order changes.
 * @param       p_nb_resolutions        the number of resolutions.
 * @param       numcomps                the number of components
 * @param       numlayers               the number of layers.
 * @param       p_manager               the user event manager.
 *
 * @return      true if the pocs are valid.
 */
static OPJ_BOOL opj_j2k_check_poc_val(const opj_poc_t *p_pocs,
                                      OPJ_UINT32 tileno,
                                      OPJ_UINT32 p_nb_pocs,
                                      OPJ_UINT32 p_nb_resolutions,
                                      OPJ_UINT32 numcomps,
                                      OPJ_UINT32 numlayers,
                                      opj_event_mgr_t * p_manager);

/**

static int opj_j2k_initialise_4K_poc(opj_poc_t *POC, int numres);

static void opj_j2k_set_cinema_parameters(opj_cparameters_t *parameters,
        opj_image_t *image, opj_event_mgr_t *p_manager);

static OPJ_BOOL opj_j2k_is_cinema_compliant(opj_image_t *image, OPJ_UINT16 rsiz,
        opj_event_mgr_t *p_manager);

static void opj_j2k_set_imf_parameters(opj_cparameters_t *parameters,
                                       opj_image_t *image, opj_event_mgr_t *p_manager);

static OPJ_BOOL opj_j2k_is_imf_compliant(opj_cparameters_t *parameters,
        opj_image_t *image,
        opj_event_mgr_t *p_manager);

/**
 * Checks for invalid number of tile-parts in SOT marker (TPsot==TNsot). See issue 254.
 *
 * @param       p_stream            the stream to read data from.
 * @param       tile_no             tile number we're looking for.
 * @param       p_correction_needed output value. if true, non conformant codestream needs TNsot correction.

            return po->str_prog;
        }
    }
    return po->str_prog;
}

static OPJ_BOOL opj_j2k_check_poc_val(const opj_poc_t *p_pocs,
                                      OPJ_UINT32 tileno,
                                      OPJ_UINT32 p_nb_pocs,
                                      OPJ_UINT32 p_nb_resolutions,
                                      OPJ_UINT32 p_num_comps,
                                      OPJ_UINT32 p_num_layers,
                                      opj_event_mgr_t * p_manager)
{
    OPJ_UINT32* packet_array;
    OPJ_UINT32 index, resno, compno, layno;
    OPJ_UINT32 i;
    OPJ_UINT32 step_c = 1;
    OPJ_UINT32 step_r = p_num_comps * step_c;
    OPJ_UINT32 step_l = p_nb_resolutions * step_r;
    OPJ_BOOL loss = OPJ_FALSE;

    assert(p_nb_pocs > 0);

    packet_array = (OPJ_UINT32*) opj_calloc(step_l * p_num_layers,
                                            sizeof(OPJ_UINT32));
    if (packet_array == 00) {
        opj_event_msg(p_manager, EVT_ERROR,
                      "Not enough memory for checking the poc values.\n");
        return OPJ_FALSE;
    }

    /* iterate through all the pocs that match our tile of interest. */
    for (i = 0; i < p_nb_pocs; ++i) {
        const opj_poc_t *poc = &p_pocs[i];


        if (tileno + 1 == poc->tile) {
            index = step_r * poc->resno0;

            /* take each resolution for each poc */
            for (resno = poc->resno0 ;
                    resno < opj_uint_min(poc->resno1, p_nb_resolutions); ++resno) {
                OPJ_UINT32 res_index = index + poc->compno0 * step_c;

                /* take each comp of each resolution for each poc */
                for (compno = poc->compno0 ;
                        compno < opj_uint_min(poc->compno1, p_num_comps); ++compno) {
                    /* The layer index always starts at zero for every progression. */
                    const OPJ_UINT32 layno0 = 0;
                    OPJ_UINT32 comp_index = res_index + layno0 * step_l;

                    /* and finally take each layer of each res of ... */
                    for (layno = layno0; layno < opj_uint_min(poc->layno1, p_num_layers);
                            ++layno) {
                        packet_array[comp_index] = 1;
                        comp_index += step_l;
                    }

                    res_index += step_c;
                }

                index += step_r;
            }

        }





























    }

    index = 0;
    for (layno = 0; layno < p_num_layers ; ++layno) {
        for (resno = 0; resno < p_nb_resolutions; ++resno) {
            for (compno = 0; compno < p_num_comps; ++compno) {
                loss |= (packet_array[index] != 1);
#ifdef DEBUG_VERBOSE
                if (packet_array[index] != 1) {
                    fprintf(stderr,
                            "Missing packet in POC: layno=%d resno=%d compno=%d\n",
                            layno, resno, compno);
                }
#endif
                index += step_c;
            }
        }
    }

    if (loss) {
        opj_event_msg(p_manager, EVT_ERROR, "Missing packets possible loss of data\n");

        l_tcp->num_layers_to_decode = l_cp->m_specific_param.m_dec.m_layer;
    } else {
        l_tcp->num_layers_to_decode = l_tcp->numlayers;
    }

    opj_read_bytes(p_header_data, &l_tcp->mct, 1);          /* SGcod (C) */
    ++p_header_data;

    if (l_tcp->mct > 1) {
#if 0
        opj_event_msg(p_manager, EVT_ERROR,
                      "Invalid multiple component transformation\n");
        return OPJ_FALSE;
#endif
    }

    p_header_size -= 5;
    for (i = 0; i < l_image->numcomps; ++i) {
        l_tcp->tccps[i].csty = l_tcp->csty & J2K_CCP_CSTY_PRT;
    }

    if (! opj_j2k_read_SPCod_SPCoc(p_j2k, 0, p_header_data, &p_header_size,

        l_nb_bytes += l_nb_comps * l_coc_bytes;

        l_qcc_bytes = opj_j2k_get_max_qcc_size(p_j2k);
        l_nb_bytes += l_nb_comps * l_qcc_bytes;
    }

    l_nb_bytes += opj_j2k_get_max_poc_size(p_j2k);

    if (p_j2k->m_specific_param.m_encoder.m_PLT) {
        /* Reserve space for PLT markers */

        OPJ_UINT32 i;
        const opj_cp_t * l_cp = &(p_j2k->m_cp);
        OPJ_UINT32 l_max_packet_count = 0;
        for (i = 0; i < l_cp->th * l_cp->tw; ++i) {
            l_max_packet_count = opj_uint_max(l_max_packet_count,
                                              opj_get_encoding_packet_count(p_j2k->m_private_image, l_cp, i));
        }
        /* Minimum 6 bytes per PLT marker, and at a minimum (taking a pessimistic */
        /* estimate of 4 bytes for a packet size), one can write */
        /* (65536-6) / 4 = 16382 paquet sizes per PLT marker */
        p_j2k->m_specific_param.m_encoder.m_reserved_bytes_for_PLT =
            6 * opj_uint_ceildiv(l_max_packet_count, 16382);
        /* Maximum 5 bytes per packet to encode a full UINT32 */
        p_j2k->m_specific_param.m_encoder.m_reserved_bytes_for_PLT +=
            l_nb_bytes += 5 * l_max_packet_count;
        p_j2k->m_specific_param.m_encoder.m_reserved_bytes_for_PLT += 1;
        l_nb_bytes += p_j2k->m_specific_param.m_encoder.m_reserved_bytes_for_PLT;
    }

    /*** DEVELOPER CORNER, Add room for your headers ***/

    return l_nb_bytes;
}

/**

    }

    return OPJ_TRUE;
}

static OPJ_BOOL opj_j2k_write_sot(opj_j2k_t *p_j2k,
                                  OPJ_BYTE * p_data,
                                  OPJ_UINT32 total_data_size,
                                  OPJ_UINT32 * p_data_written,
                                  const opj_stream_private_t *p_stream,
                                  opj_event_mgr_t * p_manager
                                 )
{
    /* preconditions */
    assert(p_j2k != 00);
    assert(p_manager != 00);
    assert(p_stream != 00);

    OPJ_UNUSED(p_stream);

    if (total_data_size < 12) {
        opj_event_msg(p_manager, EVT_ERROR,
                      "Not enough bytes in output buffer to write SOT marker\n");
        return OPJ_FALSE;
    }

    opj_write_bytes(p_data, J2K_MS_SOT,
                    2);                                 /* SOT */

       p_j2k->cstr_info->tile[tileno].tp[partno].tp_start_pos = p_stream_tell(p_stream) - 12;
       p_j2k->cstr_info->tile[tileno].tp[partno].tp_end_pos =
       p_j2k->cstr_info->tile[tileno].tp[partno].tp_start_pos + totlen - 1;
       }*/
    return OPJ_TRUE;
}

/**
 * Write one or more PLT markers in the provided buffer
 */
static OPJ_BOOL opj_j2k_write_plt_in_memory(opj_j2k_t *p_j2k,
        opj_tcd_marker_info_t* marker_info,
        OPJ_BYTE * p_data,
        OPJ_UINT32 * p_data_written,
        opj_event_mgr_t * p_manager)
{
    OPJ_BYTE Zplt = 0;
    OPJ_UINT16 Lplt;
    OPJ_BYTE* p_data_start = p_data;
    OPJ_BYTE* p_data_Lplt = p_data + 2;
    OPJ_UINT32 i;

    OPJ_UNUSED(p_j2k);

    opj_write_bytes(p_data, J2K_MS_PLT, 2);
    p_data += 2;

    /* Reserve space for Lplt */
    p_data += 2;

    opj_write_bytes(p_data, Zplt, 1);
    p_data += 1;

    Lplt = 3;

    for (i = 0; i < marker_info->packet_count; i++) {
        OPJ_BYTE var_bytes[5];
        OPJ_UINT8 var_bytes_size = 0;
        OPJ_UINT32 packet_size = marker_info->p_packet_size[i];

        /* Packet size written in variable-length way, starting with LSB */
        var_bytes[var_bytes_size] = (OPJ_BYTE)(packet_size & 0x7f);
        var_bytes_size ++;
        packet_size >>= 7;
        while (packet_size > 0) {
            var_bytes[var_bytes_size] = (OPJ_BYTE)((packet_size & 0x7f) | 0x80);
            var_bytes_size ++;
            packet_size >>= 7;
        }

        /* Check if that can fit in the current PLT marker. If not, finish */
        /* current one, and start a new one */
        if (Lplt + var_bytes_size > 65535) {
            if (Zplt == 255) {
                opj_event_msg(p_manager, EVT_ERROR,
                              "More than 255 PLT markers would be needed for current tile-part !\n");
                return OPJ_FALSE;
            }

            /* Patch Lplt */
            opj_write_bytes(p_data_Lplt, Lplt, 2);

            /* Start new segment */
            opj_write_bytes(p_data, J2K_MS_PLT, 2);
            p_data += 2;

            /* Reserve space for Lplt */
            p_data_Lplt = p_data;
            p_data += 2;

            Zplt ++;
            opj_write_bytes(p_data, Zplt, 1);
            p_data += 1;

            Lplt = 3;
        }

        Lplt = (OPJ_UINT16)(Lplt + var_bytes_size);

        /* Serialize variable-length packet size, starting with MSB */
        for (; var_bytes_size > 0; --var_bytes_size) {
            opj_write_bytes(p_data, var_bytes[var_bytes_size - 1], 1);
            p_data += 1;
        }
    }

    *p_data_written = (OPJ_UINT32)(p_data - p_data_start);

    /* Patch Lplt */
    opj_write_bytes(p_data_Lplt, Lplt, 2);

    return OPJ_TRUE;
}

static OPJ_BOOL opj_j2k_write_sod(opj_j2k_t *p_j2k,
                                  opj_tcd_t * p_tile_coder,
                                  OPJ_BYTE * p_data,
                                  OPJ_UINT32 * p_data_written,
                                  OPJ_UINT32 total_data_size,
                                  const opj_stream_private_t *p_stream,
                                  opj_event_mgr_t * p_manager
                                 )
{
    opj_codestream_info_t *l_cstr_info = 00;
    OPJ_UINT32 l_remaining_data;
    opj_tcd_marker_info_t* marker_info = NULL;

    /* preconditions */
    assert(p_j2k != 00);
    assert(p_manager != 00);
    assert(p_stream != 00);

    OPJ_UNUSED(p_stream);

    if (total_data_size < 4) {
        opj_event_msg(p_manager, EVT_ERROR,
                      "Not enough bytes in output buffer to write SOD marker\n");
        return OPJ_FALSE;
    }

    opj_write_bytes(p_data, J2K_MS_SOD,
                    2);                                 /* SOD */


    /* make room for the EOF marker */
    l_remaining_data =  total_data_size - 4;

    /* update tile coder */
    p_tile_coder->tp_num =
        p_j2k->m_specific_param.m_encoder.m_current_poc_tile_part_number ;
    p_tile_coder->cur_tp_num =
        p_j2k->m_specific_param.m_encoder.m_current_tile_part_number;

            l_cstr_info->packno = 0;
        }
#endif
    }

    *p_data_written = 0;

    if (p_j2k->m_specific_param.m_encoder.m_PLT) {
        marker_info = opj_tcd_marker_info_create(
                          p_j2k->m_specific_param.m_encoder.m_PLT);
        if (marker_info == NULL) {
            opj_event_msg(p_manager, EVT_ERROR,
                          "Cannot encode tile: opj_tcd_marker_info_create() failed\n");
            return OPJ_FALSE;
        }
    }

    if (l_remaining_data <
            p_j2k->m_specific_param.m_encoder.m_reserved_bytes_for_PLT) {
        opj_event_msg(p_manager, EVT_ERROR,
                      "Not enough bytes in output buffer to write SOD marker\n");
        opj_tcd_marker_info_destroy(marker_info);
        return OPJ_FALSE;
    }
    l_remaining_data -= p_j2k->m_specific_param.m_encoder.m_reserved_bytes_for_PLT;

    if (! opj_tcd_encode_tile(p_tile_coder, p_j2k->m_current_tile_number,
                              p_data + 2,
                              p_data_written, l_remaining_data, l_cstr_info,
                              marker_info,
                              p_manager)) {
        opj_event_msg(p_manager, EVT_ERROR, "Cannot encode tile\n");
        opj_tcd_marker_info_destroy(marker_info);
        return OPJ_FALSE;
    }

    /* For SOD */
    *p_data_written += 2;

    if (p_j2k->m_specific_param.m_encoder.m_PLT) {
        OPJ_UINT32 l_data_written_PLT = 0;
        OPJ_BYTE* p_PLT_buffer = (OPJ_BYTE*)opj_malloc(
                                     p_j2k->m_specific_param.m_encoder.m_reserved_bytes_for_PLT);
        if (!p_PLT_buffer) {
            opj_event_msg(p_manager, EVT_ERROR, "Cannot allocate memory\n");
            opj_tcd_marker_info_destroy(marker_info);
            return OPJ_FALSE;
        }
        if (!opj_j2k_write_plt_in_memory(p_j2k,
                                         marker_info,
                                         p_PLT_buffer,
                                         &l_data_written_PLT,
                                         p_manager)) {
            opj_tcd_marker_info_destroy(marker_info);
            opj_free(p_PLT_buffer);
            return OPJ_FALSE;
        }

        assert(l_data_written_PLT <=
               p_j2k->m_specific_param.m_encoder.m_reserved_bytes_for_PLT);

        /* Move PLT marker(s) before SOD */
        memmove(p_data + l_data_written_PLT, p_data, *p_data_written);
        memcpy(p_data, p_PLT_buffer, l_data_written_PLT);
        opj_free(p_PLT_buffer);
        *p_data_written += l_data_written_PLT;
    }

    opj_tcd_marker_info_destroy(marker_info);

    return OPJ_TRUE;
}

static OPJ_BOOL opj_j2k_read_sod(opj_j2k_t *p_j2k,
                                 opj_stream_private_t *p_stream,
                                 opj_event_mgr_t * p_manager

    opj_image_comp_t * l_img_comp = 00;

    OPJ_UINT32 i, j, k;
    OPJ_INT32 l_x0, l_y0, l_x1, l_y1;
    OPJ_FLOAT32 * l_rates = 0;
    OPJ_FLOAT32 l_sot_remove;
    OPJ_UINT32 l_bits_empty, l_size_pixel;
    OPJ_UINT64 l_tile_size = 0;
    OPJ_UINT32 l_last_res;
    OPJ_FLOAT32(* l_tp_stride_func)(opj_tcp_t *) = 00;

    /* preconditions */
    assert(p_j2k != 00);
    assert(p_manager != 00);
    assert(p_stream != 00);

                               (OPJ_INT32)l_image->x1);
            l_y1 = opj_int_min((OPJ_INT32)(l_cp->ty0 + (i + 1) * l_cp->tdy),
                               (OPJ_INT32)l_image->y1);

            l_rates = l_tcp->rates;

            /* Modification of the RATE >> */












            for (k = 0; k < l_tcp->numlayers; ++k) {
                if (*l_rates > 0.0f) {
                    *l_rates = (OPJ_FLOAT32)(((OPJ_FLOAT64)l_size_pixel * (OPJ_UINT32)(
                                                  l_x1 - l_x0) *
                                              (OPJ_UINT32)(l_y1 - l_y0))

                                             / ((*l_rates) * (OPJ_FLOAT32)l_bits_empty))

                               -
                               l_offset;
                }

                ++l_rates;
            }

        }
    }

    l_img_comp = l_image->comps;
    l_tile_size = 0;

    for (i = 0; i < l_image->numcomps; ++i) {
        l_tile_size += (OPJ_UINT64)opj_uint_ceildiv(l_cp->tdx, l_img_comp->dx)
                       *
                       opj_uint_ceildiv(l_cp->tdy, l_img_comp->dy)
                       *
                       l_img_comp->prec;


        ++l_img_comp;
    }

    /* TODO: where does this magic value come from ? */
    /* This used to be 1.3 / 8, but with random data and very small code */
    /* block sizes, this is not enough. For example with */
    /* bin/test_tile_encoder 1 256 256 32 32 8 0 reversible_with_precinct.j2k 4 4 3 0 0 1 16 16 */
    /* TODO revise this to take into account the overhead linked to the */
    /* number of packets and number of code blocks in packets */
    l_tile_size = (OPJ_UINT64)((double)l_tile_size * 1.4 / 8);

    /* Arbitrary amount to make the following work: */
    /* bin/test_tile_encoder 1 256 256 17 16 8 0 reversible_no_precinct.j2k 4 4 3 0 0 1 */
    l_tile_size += 500;

    l_tile_size += opj_j2k_get_specific_header_sizes(p_j2k);

    if (l_tile_size > UINT_MAX) {
        l_tile_size = UINT_MAX;
    }

    p_j2k->m_specific_param.m_encoder.m_encoded_tile_size = (OPJ_UINT32)l_tile_size;
    p_j2k->m_specific_param.m_encoder.m_encoded_tile_data =
        (OPJ_BYTE *) opj_malloc(p_j2k->m_specific_param.m_encoder.m_encoded_tile_size);
    if (p_j2k->m_specific_param.m_encoder.m_encoded_tile_data == 00) {
        opj_event_msg(p_manager, EVT_ERROR,
                      "Not enough memory to allocate m_encoded_tile_data. %u MB required\n",
                      (OPJ_UINT32)(l_tile_size / 1024 / 1024));
        return OPJ_FALSE;
    }

    if (OPJ_IS_CINEMA(l_cp->rsiz) || OPJ_IS_IMF(l_cp->rsiz)) {
        p_j2k->m_specific_param.m_encoder.m_tlm_sot_offsets_buffer =
            (OPJ_BYTE *) opj_malloc(5 *
                                    p_j2k->m_specific_param.m_encoder.m_total_tile_parts);
        if (! p_j2k->m_specific_param.m_encoder.m_tlm_sot_offsets_buffer) {
            return OPJ_FALSE;
        }

        }
        break;
    default :
        break;
    }

    /* Precincts */
    parameters->csty |= J2K_CP_CSTY_PRT;
    if (parameters->numresolution == 1) {
        parameters->res_spec = 1;
        parameters->prcw_init[0] = 128;
        parameters->prch_init[0] = 128;
    } else {
        parameters->res_spec = parameters->numresolution - 1;
        for (i = 0; i < parameters->res_spec; i++) {

        break;
    default :
        break;
    }

    return OPJ_TRUE;
}

static int opj_j2k_get_imf_max_NL(opj_cparameters_t *parameters,
                                  opj_image_t *image)
{
    /* Decomposition levels */
    const OPJ_UINT16 rsiz = parameters->rsiz;
    const OPJ_UINT16 profile = OPJ_GET_IMF_PROFILE(rsiz);
    const OPJ_UINT32 XTsiz = parameters->tile_size_on ? (OPJ_UINT32)
                             parameters->cp_tdx : image->x1;
    switch (profile) {
    case OPJ_PROFILE_IMF_2K:
        return 5;
    case OPJ_PROFILE_IMF_4K:
        return 6;
    case OPJ_PROFILE_IMF_8K:
        return 7;
    case OPJ_PROFILE_IMF_2K_R: {
        if (XTsiz >= 2048) {
            return 5;
        } else if (XTsiz >= 1024) {
            return 4;
        }
        break;
    }
    case OPJ_PROFILE_IMF_4K_R: {
        if (XTsiz >= 4096) {
            return 6;
        } else if (XTsiz >= 2048) {
            return 5;
        } else if (XTsiz >= 1024) {
            return 4;
        }
        break;
    }
    case OPJ_PROFILE_IMF_8K_R: {
        if (XTsiz >= 8192) {
            return 7;
        } else if (XTsiz >= 4096) {
            return 6;
        } else if (XTsiz >= 2048) {
            return 5;
        } else if (XTsiz >= 1024) {
            return 4;
        }
        break;
    }
    default:
        break;
    }
    return -1;
}

static void opj_j2k_set_imf_parameters(opj_cparameters_t *parameters,
                                       opj_image_t *image, opj_event_mgr_t *p_manager)
{
    const OPJ_UINT16 rsiz = parameters->rsiz;
    const OPJ_UINT16 profile = OPJ_GET_IMF_PROFILE(rsiz);

    OPJ_UNUSED(p_manager);

    /* Override defaults set by opj_set_default_encoder_parameters */
    if (parameters->cblockw_init == OPJ_COMP_PARAM_DEFAULT_CBLOCKW &&
            parameters->cblockh_init == OPJ_COMP_PARAM_DEFAULT_CBLOCKH) {
        parameters->cblockw_init = 32;
        parameters->cblockh_init = 32;
    }

    /* One tile part for each component */
    parameters->tp_flag = 'C';
    parameters->tp_on = 1;

    if (parameters->prog_order == OPJ_COMP_PARAM_DEFAULT_PROG_ORDER) {
        parameters->prog_order = OPJ_CPRL;
    }

    if (profile == OPJ_PROFILE_IMF_2K ||
            profile == OPJ_PROFILE_IMF_4K ||
            profile == OPJ_PROFILE_IMF_8K) {
        /* 9-7 transform */
        parameters->irreversible = 1;
    }

    /* Adjust the number of resolutions if set to its defaults */
    if (parameters->numresolution == OPJ_COMP_PARAM_DEFAULT_NUMRESOLUTION &&
            image->x0 == 0 &&
            image->y0 == 0) {
        const int max_NL = opj_j2k_get_imf_max_NL(parameters, image);
        if (max_NL >= 0 && parameters->numresolution > max_NL) {
            parameters->numresolution = max_NL + 1;
        }

        /* Note: below is generic logic */
        if (!parameters->tile_size_on) {
            while (parameters->numresolution > 0) {
                if (image->x1 < (1U << ((OPJ_UINT32)parameters->numresolution - 1U))) {
                    parameters->numresolution --;
                    continue;
                }
                if (image->y1 < (1U << ((OPJ_UINT32)parameters->numresolution - 1U))) {
                    parameters->numresolution --;
                    continue;
                }
                break;
            }
        }
    }

    /* Set defaults precincts */
    if (parameters->csty == 0) {
        parameters->csty |= J2K_CP_CSTY_PRT;
        if (parameters->numresolution == 1) {
            parameters->res_spec = 1;
            parameters->prcw_init[0] = 128;
            parameters->prch_init[0] = 128;
        } else {
            int i;
            parameters->res_spec = parameters->numresolution - 1;
            for (i = 0; i < parameters->res_spec; i++) {
                parameters->prcw_init[i] = 256;
                parameters->prch_init[i] = 256;
            }
        }
    }
}

/* Table A.53 from JPEG2000 standard */
static const OPJ_UINT16 tabMaxSubLevelFromMainLevel[] = {
    15, /* unspecified */
    1,
    1,
    1,
    2,
    3,
    4,
    5,
    6,
    7,
    8,
    9
};

static OPJ_BOOL opj_j2k_is_imf_compliant(opj_cparameters_t *parameters,
        opj_image_t *image,
        opj_event_mgr_t *p_manager)
{
    OPJ_UINT32 i;
    const OPJ_UINT16 rsiz = parameters->rsiz;
    const OPJ_UINT16 profile = OPJ_GET_IMF_PROFILE(rsiz);
    const OPJ_UINT16 mainlevel = OPJ_GET_IMF_MAINLEVEL(rsiz);
    const OPJ_UINT16 sublevel = OPJ_GET_IMF_SUBLEVEL(rsiz);
    const int NL = parameters->numresolution - 1;
    const OPJ_UINT32 XTsiz = parameters->tile_size_on ? (OPJ_UINT32)
                             parameters->cp_tdx : image->x1;
    OPJ_BOOL ret = OPJ_TRUE;

    /* Validate mainlevel */
    if (mainlevel > OPJ_IMF_MAINLEVEL_MAX) {
        opj_event_msg(p_manager, EVT_WARNING,
                      "IMF profile require mainlevel <= 11.\n"
                      "-> %d is thus not compliant\n"
                      "-> Non-IMF codestream will be generated\n",
                      mainlevel);
        ret = OPJ_FALSE;
    }

    /* Validate sublevel */
    assert(sizeof(tabMaxSubLevelFromMainLevel) ==
           (OPJ_IMF_MAINLEVEL_MAX + 1) * sizeof(tabMaxSubLevelFromMainLevel[0]));
    if (sublevel > tabMaxSubLevelFromMainLevel[mainlevel]) {
        opj_event_msg(p_manager, EVT_WARNING,
                      "IMF profile require sublevel <= %d for mainlevel = %d.\n"
                      "-> %d is thus not compliant\n"
                      "-> Non-IMF codestream will be generated\n",
                      tabMaxSubLevelFromMainLevel[mainlevel],
                      mainlevel,
                      sublevel);
        ret = OPJ_FALSE;
    }

    /* Number of components */
    if (image->numcomps > 3) {
        opj_event_msg(p_manager, EVT_WARNING,
                      "IMF profiles require at most 3 components.\n"
                      "-> Number of components of input image (%d) is not compliant\n"
                      "-> Non-IMF codestream will be generated\n",
                      image->numcomps);
        ret = OPJ_FALSE;
    }

    if (image->x0 != 0 || image->y0 != 0) {
        opj_event_msg(p_manager, EVT_WARNING,
                      "IMF profiles require image origin to be at 0,0.\n"
                      "-> %d,%d is not compliant\n"
                      "-> Non-IMF codestream will be generated\n",
                      image->x0, image->y0 != 0);
        ret = OPJ_FALSE;
    }

    if (parameters->cp_tx0 != 0 || parameters->cp_ty0 != 0) {
        opj_event_msg(p_manager, EVT_WARNING,
                      "IMF profiles require tile origin to be at 0,0.\n"
                      "-> %d,%d is not compliant\n"
                      "-> Non-IMF codestream will be generated\n",
                      parameters->cp_tx0, parameters->cp_ty0);
        ret = OPJ_FALSE;
    }

    if (parameters->tile_size_on) {
        if (profile == OPJ_PROFILE_IMF_2K ||
                profile == OPJ_PROFILE_IMF_4K ||
                profile == OPJ_PROFILE_IMF_8K) {
            if ((OPJ_UINT32)parameters->cp_tdx < image->x1 ||
                    (OPJ_UINT32)parameters->cp_tdy < image->y1) {
                opj_event_msg(p_manager, EVT_WARNING,
                              "IMF 2K/4K/8K single tile profiles require tile to be greater or equal to image size.\n"
                              "-> %d,%d is lesser than %d,%d\n"
                              "-> Non-IMF codestream will be generated\n",
                              parameters->cp_tdx,
                              parameters->cp_tdy,
                              image->x1,
                              image->y1);
                ret = OPJ_FALSE;
            }
        } else {
            if ((OPJ_UINT32)parameters->cp_tdx >= image->x1 &&
                    (OPJ_UINT32)parameters->cp_tdy >= image->y1) {
                /* ok */
            } else if (parameters->cp_tdx == 1024 &&
                       parameters->cp_tdy == 1024) {
                /* ok */
            } else if (parameters->cp_tdx == 2048 &&
                       parameters->cp_tdy == 2048 &&
                       (profile == OPJ_PROFILE_IMF_4K ||
                        profile == OPJ_PROFILE_IMF_8K)) {
                /* ok */
            } else if (parameters->cp_tdx == 4096 &&
                       parameters->cp_tdy == 4096 &&
                       profile == OPJ_PROFILE_IMF_8K) {
                /* ok */
            } else {
                opj_event_msg(p_manager, EVT_WARNING,
                              "IMF 2K_R/4K_R/8K_R single/multiple tile profiles "
                              "require tile to be greater or equal to image size,\n"
                              "or to be (1024,1024), or (2048,2048) for 4K_R/8K_R "
                              "or (4096,4096) for 8K_R.\n"
                              "-> %d,%d is non conformant\n"
                              "-> Non-IMF codestream will be generated\n",
                              parameters->cp_tdx,
                              parameters->cp_tdy);
                ret = OPJ_FALSE;
            }
        }
    }

    /* Bitdepth */
    for (i = 0; i < image->numcomps; i++) {
        if (!(image->comps[i].bpp >= 8 && image->comps[i].bpp <= 16) ||
                (image->comps[i].sgnd)) {
            char signed_str[] = "signed";
            char unsigned_str[] = "unsigned";
            char *tmp_str = image->comps[i].sgnd ? signed_str : unsigned_str;
            opj_event_msg(p_manager, EVT_WARNING,
                          "IMF profiles require precision of each component to b in [8-16] bits unsigned"
                          "-> At least component %d of input image (%d bits, %s) is not compliant\n"
                          "-> Non-IMF codestream will be generated\n",
                          i, image->comps[i].bpp, tmp_str);
            ret = OPJ_FALSE;
        }
    }

    /* Sub-sampling */
    for (i = 0; i < image->numcomps; i++) {
        if (i == 0 && image->comps[i].dx != 1) {
            opj_event_msg(p_manager, EVT_WARNING,
                          "IMF profiles require XRSiz1 == 1. Here it is set to %d.\n"
                          "-> Non-IMF codestream will be generated\n",
                          image->comps[i].dx);
            ret = OPJ_FALSE;
        }
        if (i == 1 && image->comps[i].dx != 1 && image->comps[i].dx != 2) {
            opj_event_msg(p_manager, EVT_WARNING,
                          "IMF profiles require XRSiz2 == 1 or 2. Here it is set to %d.\n"
                          "-> Non-IMF codestream will be generated\n",
                          image->comps[i].dx);
            ret = OPJ_FALSE;
        }
        if (i > 1 && image->comps[i].dx != image->comps[i - 1].dx) {
            opj_event_msg(p_manager, EVT_WARNING,
                          "IMF profiles require XRSiz%d to be the same as XRSiz2. "
                          "Here it is set to %d instead of %d.\n"
                          "-> Non-IMF codestream will be generated\n",
                          i + 1, image->comps[i].dx, image->comps[i - 1].dx);
            ret = OPJ_FALSE;
        }
        if (image->comps[i].dy != 1) {
            opj_event_msg(p_manager, EVT_WARNING,
                          "IMF profiles require YRsiz == 1. "
                          "Here it is set to %d for component i.\n"
                          "-> Non-IMF codestream will be generated\n",
                          image->comps[i].dy, i);
            ret = OPJ_FALSE;
        }
    }

    /* Image size */
    switch (profile) {
    case OPJ_PROFILE_IMF_2K:
    case OPJ_PROFILE_IMF_2K_R:
        if (((image->comps[0].w > 2048) | (image->comps[0].h > 1556))) {
            opj_event_msg(p_manager, EVT_WARNING,
                          "IMF 2K/2K_R profile require:\n"
                          "width <= 2048 and height <= 1556\n"
                          "-> Input image size %d x %d is not compliant\n"
                          "-> Non-IMF codestream will be generated\n",
                          image->comps[0].w, image->comps[0].h);
            ret = OPJ_FALSE;
        }
        break;
    case OPJ_PROFILE_IMF_4K:
    case OPJ_PROFILE_IMF_4K_R:
        if (((image->comps[0].w > 4096) | (image->comps[0].h > 3112))) {
            opj_event_msg(p_manager, EVT_WARNING,
                          "IMF 4K/4K_R profile require:\n"
                          "width <= 4096 and height <= 3112\n"
                          "-> Input image size %d x %d is not compliant\n"
                          "-> Non-IMF codestream will be generated\n",
                          image->comps[0].w, image->comps[0].h);
            ret = OPJ_FALSE;
        }
        break;
    case OPJ_PROFILE_IMF_8K:
    case OPJ_PROFILE_IMF_8K_R:
        if (((image->comps[0].w > 8192) | (image->comps[0].h > 6224))) {
            opj_event_msg(p_manager, EVT_WARNING,
                          "IMF 8K/8K_R profile require:\n"
                          "width <= 8192 and height <= 6224\n"
                          "-> Input image size %d x %d is not compliant\n"
                          "-> Non-IMF codestream will be generated\n",
                          image->comps[0].w, image->comps[0].h);
            ret = OPJ_FALSE;
        }
        break;
    default :
        assert(0);
        return OPJ_FALSE;
    }

    if (parameters->roi_compno != -1) {
        opj_event_msg(p_manager, EVT_WARNING,
                      "IMF profile forbid RGN / region of interest marker.\n"
                      "-> Compression parameters specify a ROI\n"
                      "-> Non-IMF codestream will be generated\n");
        ret = OPJ_FALSE;
    }

    if (parameters->cblockw_init != 32 || parameters->cblockh_init != 32) {
        opj_event_msg(p_manager, EVT_WARNING,
                      "IMF profile require code block size to be 32x32.\n"
                      "-> Compression parameters set it to %dx%d.\n"
                      "-> Non-IMF codestream will be generated\n",
                      parameters->cblockw_init,
                      parameters->cblockh_init);
        ret = OPJ_FALSE;
    }

    if (parameters->prog_order != OPJ_CPRL) {
        opj_event_msg(p_manager, EVT_WARNING,
                      "IMF profile require progression order to be CPRL.\n"
                      "-> Compression parameters set it to %d.\n"
                      "-> Non-IMF codestream will be generated\n",
                      parameters->prog_order);
        ret = OPJ_FALSE;
    }

    if (parameters->numpocs != 0) {
        opj_event_msg(p_manager, EVT_WARNING,
                      "IMF profile forbid POC markers.\n"
                      "-> Compression parameters set %d POC.\n"
                      "-> Non-IMF codestream will be generated\n",
                      parameters->numpocs);
        ret = OPJ_FALSE;
    }

    /* Codeblock style: no mode switch enabled */
    if (parameters->mode != 0) {
        opj_event_msg(p_manager, EVT_WARNING,
                      "IMF profile forbid mode switch in code block style.\n"
                      "-> Compression parameters set code block style to %d.\n"
                      "-> Non-IMF codestream will be generated\n",
                      parameters->mode);
        ret = OPJ_FALSE;
    }

    if (profile == OPJ_PROFILE_IMF_2K ||
            profile == OPJ_PROFILE_IMF_4K ||
            profile == OPJ_PROFILE_IMF_8K) {
        /* Expect 9-7 transform */
        if (parameters->irreversible != 1) {
            opj_event_msg(p_manager, EVT_WARNING,
                          "IMF 2K/4K/8K profiles require 9-7 Irreversible Transform.\n"
                          "-> Compression parameters set it to reversible.\n"
                          "-> Non-IMF codestream will be generated\n");
            ret = OPJ_FALSE;
        }
    } else {
        /* Expect 5-3 transform */
        if (parameters->irreversible != 0) {
            opj_event_msg(p_manager, EVT_WARNING,
                          "IMF 2K/4K/8K profiles require 5-3 reversible Transform.\n"
                          "-> Compression parameters set it to irreversible.\n"
                          "-> Non-IMF codestream will be generated\n");
            ret = OPJ_FALSE;
        }
    }

    /* Number of layers */
    if (parameters->tcp_numlayers != 1) {
        opj_event_msg(p_manager, EVT_WARNING,
                      "IMF 2K/4K/8K profiles require 1 single quality layer.\n"
                      "-> Number of layers is %d.\n"
                      "-> Non-IMF codestream will be generated\n",
                      parameters->tcp_numlayers);
        ret = OPJ_FALSE;
    }

    /* Decomposition levels */
    switch (profile) {
    case OPJ_PROFILE_IMF_2K:
        if (!(NL >= 1 && NL <= 5)) {
            opj_event_msg(p_manager, EVT_WARNING,
                          "IMF 2K profile requires 1 <= NL <= 5:\n"
                          "-> Number of decomposition levels is %d.\n"
                          "-> Non-IMF codestream will be generated\n",
                          NL);
            ret = OPJ_FALSE;
        }
        break;
    case OPJ_PROFILE_IMF_4K:
        if (!(NL >= 1 && NL <= 6)) {
            opj_event_msg(p_manager, EVT_WARNING,
                          "IMF 4K profile requires 1 <= NL <= 6:\n"
                          "-> Number of decomposition levels is %d.\n"
                          "-> Non-IMF codestream will be generated\n",
                          NL);
            ret = OPJ_FALSE;
        }
        break;
    case OPJ_PROFILE_IMF_8K:
        if (!(NL >= 1 && NL <= 7)) {
            opj_event_msg(p_manager, EVT_WARNING,
                          "IMF 8K profile requires 1 <= NL <= 7:\n"
                          "-> Number of decomposition levels is %d.\n"
                          "-> Non-IMF codestream will be generated\n",
                          NL);
            ret = OPJ_FALSE;
        }
        break;
    case OPJ_PROFILE_IMF_2K_R: {
        if (XTsiz >= 2048) {
            if (!(NL >= 1 && NL <= 5)) {
                opj_event_msg(p_manager, EVT_WARNING,
                              "IMF 2K_R profile requires 1 <= NL <= 5 for XTsiz >= 2048:\n"
                              "-> Number of decomposition levels is %d.\n"
                              "-> Non-IMF codestream will be generated\n",
                              NL);
                ret = OPJ_FALSE;
            }
        } else if (XTsiz >= 1024) {
            if (!(NL >= 1 && NL <= 4)) {
                opj_event_msg(p_manager, EVT_WARNING,
                              "IMF 2K_R profile requires 1 <= NL <= 4 for XTsiz in [1024,2048[:\n"
                              "-> Number of decomposition levels is %d.\n"
                              "-> Non-IMF codestream will be generated\n",
                              NL);
                ret = OPJ_FALSE;
            }
        }
        break;
    }
    case OPJ_PROFILE_IMF_4K_R: {
        if (XTsiz >= 4096) {
            if (!(NL >= 1 && NL <= 6)) {
                opj_event_msg(p_manager, EVT_WARNING,
                              "IMF 4K_R profile requires 1 <= NL <= 6 for XTsiz >= 4096:\n"
                              "-> Number of decomposition levels is %d.\n"
                              "-> Non-IMF codestream will be generated\n",
                              NL);
                ret = OPJ_FALSE;
            }
        } else if (XTsiz >= 2048) {
            if (!(NL >= 1 && NL <= 5)) {
                opj_event_msg(p_manager, EVT_WARNING,
                              "IMF 4K_R profile requires 1 <= NL <= 5 for XTsiz in [2048,4096[:\n"
                              "-> Number of decomposition levels is %d.\n"
                              "-> Non-IMF codestream will be generated\n",
                              NL);
                ret = OPJ_FALSE;
            }
        } else if (XTsiz >= 1024) {
            if (!(NL >= 1 && NL <= 4)) {
                opj_event_msg(p_manager, EVT_WARNING,
                              "IMF 4K_R profile requires 1 <= NL <= 4 for XTsiz in [1024,2048[:\n"
                              "-> Number of decomposition levels is %d.\n"
                              "-> Non-IMF codestream will be generated\n",
                              NL);
                ret = OPJ_FALSE;
            }
        }
        break;
    }
    case OPJ_PROFILE_IMF_8K_R: {
        if (XTsiz >= 8192) {
            if (!(NL >= 1 && NL <= 7)) {
                opj_event_msg(p_manager, EVT_WARNING,
                              "IMF 4K_R profile requires 1 <= NL <= 7 for XTsiz >= 8192:\n"
                              "-> Number of decomposition levels is %d.\n"
                              "-> Non-IMF codestream will be generated\n",
                              NL);
                ret = OPJ_FALSE;
            }
        } else if (XTsiz >= 4096) {
            if (!(NL >= 1 && NL <= 6)) {
                opj_event_msg(p_manager, EVT_WARNING,
                              "IMF 4K_R profile requires 1 <= NL <= 6 for XTsiz in [4096,8192[:\n"
                              "-> Number of decomposition levels is %d.\n"
                              "-> Non-IMF codestream will be generated\n",
                              NL);
                ret = OPJ_FALSE;
            }
        } else if (XTsiz >= 2048) {
            if (!(NL >= 1 && NL <= 5)) {
                opj_event_msg(p_manager, EVT_WARNING,
                              "IMF 4K_R profile requires 1 <= NL <= 5 for XTsiz in [2048,4096[:\n"
                              "-> Number of decomposition levels is %d.\n"
                              "-> Non-IMF codestream will be generated\n",
                              NL);
                ret = OPJ_FALSE;
            }
        } else if (XTsiz >= 1024) {
            if (!(NL >= 1 && NL <= 4)) {
                opj_event_msg(p_manager, EVT_WARNING,
                              "IMF 4K_R profile requires 1 <= NL <= 4 for XTsiz in [1024,2048[:\n"
                              "-> Number of decomposition levels is %d.\n"
                              "-> Non-IMF codestream will be generated\n",
                              NL);
                ret = OPJ_FALSE;
            }
        }
        break;
    }
    default:
        break;
    }

    if (parameters->numresolution == 1) {
        if (parameters->res_spec != 1 ||
                parameters->prcw_init[0] != 128 ||
                parameters->prch_init[0] != 128) {
            opj_event_msg(p_manager, EVT_WARNING,
                          "IMF profiles require PPx = PPy = 7 for NLLL band, else 8.\n"
                          "-> Supplied values are different from that.\n"
                          "-> Non-IMF codestream will be generated\n",
                          NL);
            ret = OPJ_FALSE;
        }
    } else {
        int i;
        for (i = 0; i < parameters->res_spec; i++) {
            if (parameters->prcw_init[i] != 256 ||
                    parameters->prch_init[i] != 256) {
                opj_event_msg(p_manager, EVT_WARNING,
                              "IMF profiles require PPx = PPy = 7 for NLLL band, else 8.\n"
                              "-> Supplied values are different from that.\n"
                              "-> Non-IMF codestream will be generated\n",
                              NL);
                ret = OPJ_FALSE;
            }
        }
    }

    return ret;
}


OPJ_BOOL opj_j2k_setup_encoder(opj_j2k_t *p_j2k,
                               opj_cparameters_t *parameters,
                               opj_image_t *image,
                               opj_event_mgr_t * p_manager)
{
    OPJ_UINT32 i, j, tileno, numpocs_tile;

            }
        } else {
            parameters->max_cs_size = 0;
        }
    } else {
        OPJ_FLOAT32 temp_rate;
        OPJ_BOOL cap = OPJ_FALSE;

        if (OPJ_IS_IMF(parameters->rsiz) && parameters->max_cs_size > 0 &&
                parameters->tcp_numlayers == 1 && parameters->tcp_rates[0] == 0) {
            parameters->tcp_rates[0] = (OPJ_FLOAT32)(image->numcomps * image->comps[0].w *
                                       image->comps[0].h * image->comps[0].prec) /
                                       (OPJ_FLOAT32)(((OPJ_UINT32)parameters->max_cs_size) * 8 * image->comps[0].dx *
                                               image->comps[0].dy);
        }

        temp_rate = (OPJ_FLOAT32)(((double)image->numcomps * image->comps[0].w *
                                   image->comps[0].h * image->comps[0].prec) /
                                  (((double)parameters->max_cs_size) * 8 * image->comps[0].dx *
                                   image->comps[0].dy));
        for (i = 0; i < (OPJ_UINT32) parameters->tcp_numlayers; i++) {
            if (parameters->tcp_rates[i] < temp_rate) {
                parameters->tcp_rates[i] = temp_rate;

                      "JPEG 2000 Long Term Storage profile not yet supported\n");
        parameters->rsiz = OPJ_PROFILE_NONE;
    } else if (OPJ_IS_BROADCAST(parameters->rsiz)) {
        opj_event_msg(p_manager, EVT_WARNING,
                      "JPEG 2000 Broadcast profiles not yet supported\n");
        parameters->rsiz = OPJ_PROFILE_NONE;
    } else if (OPJ_IS_IMF(parameters->rsiz)) {
        opj_j2k_set_imf_parameters(parameters, image, p_manager);
        if (!opj_j2k_is_imf_compliant(parameters, image, p_manager)) {
            parameters->rsiz = OPJ_PROFILE_NONE;
        }
    } else if (OPJ_IS_PART2(parameters->rsiz)) {
        if (parameters->rsiz == ((OPJ_PROFILE_PART2) | (OPJ_EXTENSION_NONE))) {
            opj_event_msg(p_manager, EVT_WARNING,
                          "JPEG 2000 Part-2 profile defined\n"
                          "but no Part-2 extension enabled.\n"
                          "Profile set to NONE.\n");
            parameters->rsiz = OPJ_PROFILE_NONE;

    }

    /*
    calculate other encoding parameters
    */

    if (parameters->tile_size_on) {
        if (cp->tdx == 0) {
            opj_event_msg(p_manager, EVT_ERROR, "Invalid tile width\n");
            return OPJ_FALSE;
        }
        if (cp->tdy == 0) {
            opj_event_msg(p_manager, EVT_ERROR, "Invalid tile height\n");
            return OPJ_FALSE;
        }
        cp->tw = (OPJ_UINT32)opj_int_ceildiv((OPJ_INT32)(image->x1 - cp->tx0),
                                             (OPJ_INT32)cp->tdx);
        cp->th = (OPJ_UINT32)opj_int_ceildiv((OPJ_INT32)(image->y1 - cp->ty0),
                                             (OPJ_INT32)cp->tdy);
    } else {
        cp->tdx = image->x1 - cp->tx0;
        cp->tdy = image->y1 - cp->ty0;

    /* ---------------------------- */
    cp->tcps = (opj_tcp_t*) opj_calloc(cp->tw * cp->th, sizeof(opj_tcp_t));
    if (!cp->tcps) {
        opj_event_msg(p_manager, EVT_ERROR,
                      "Not enough memory to allocate tile coding parameters\n");
        return OPJ_FALSE;
    }








    for (tileno = 0; tileno < cp->tw * cp->th; tileno++) {
        opj_tcp_t *tcp = &cp->tcps[tileno];
        tcp->numlayers = (OPJ_UINT32)parameters->tcp_numlayers;

        for (j = 0; j < tcp->numlayers; j++) {
            if (OPJ_IS_CINEMA(cp->rsiz) || OPJ_IS_IMF(cp->rsiz)) {
                if (cp->m_specific_param.m_enc.m_fixed_quality) {
                    tcp->distoratio[j] = parameters->tcp_distoratio[j];
                }
                tcp->rates[j] = parameters->tcp_rates[j];
            } else {
                if (cp->m_specific_param.m_enc.m_fixed_quality) {       /* add fixed_quality */
                    tcp->distoratio[j] = parameters->tcp_distoratio[j];

        tcp->mct = (OPJ_UINT32)parameters->tcp_mct;

        numpocs_tile = 0;
        tcp->POC = 0;

        if (parameters->numpocs) {
            /* initialisation of POC */

            for (i = 0; i < parameters->numpocs; i++) {
                if (tileno + 1 == parameters->POC[i].tile)  {
                    opj_poc_t *tcp_poc = &tcp->pocs[numpocs_tile];

                    if (parameters->POC[numpocs_tile].compno0 >= image->numcomps) {
                        opj_event_msg(p_manager, EVT_ERROR,
                                      "Invalid compno0 for POC %d\n", i);
                        return OPJ_FALSE;
                    }

                    tcp_poc->resno0         = parameters->POC[numpocs_tile].resno0;
                    tcp_poc->compno0        = parameters->POC[numpocs_tile].compno0;
                    tcp_poc->layno1         = parameters->POC[numpocs_tile].layno1;
                    tcp_poc->resno1         = parameters->POC[numpocs_tile].resno1;
                    tcp_poc->compno1        = opj_uint_min(parameters->POC[numpocs_tile].compno1,
                                                           image->numcomps);
                    tcp_poc->prg1           = parameters->POC[numpocs_tile].prg1;
                    tcp_poc->tile           = parameters->POC[numpocs_tile].tile;

                    numpocs_tile++;
                }
            }

            if (numpocs_tile) {

                /* TODO MSD use the return value*/
                opj_j2k_check_poc_val(parameters->POC, tileno, parameters->numpocs,
                                      (OPJ_UINT32)parameters->numresolution, image->numcomps,
                                      (OPJ_UINT32)parameters->tcp_numlayers, p_manager);

                tcp->POC = 1;
                tcp->numpocs = numpocs_tile - 1 ;
            }
        } else {
            tcp->numpocs = 0;
        }

        tcp->tccps = (opj_tccp_t*) opj_calloc(image->numcomps, sizeof(opj_tccp_t));
        if (!tcp->tccps) {
            opj_event_msg(p_manager, EVT_ERROR,

    }

    /* Copy codestream image information to the output image */
    opj_copy_image_header(p_j2k->m_private_image, *p_image);

    /*Allocate and initialize some elements of codestrem index*/
    if (!opj_j2k_allocate_tile_element_cstr_index(p_j2k)) {
        opj_image_destroy(*p_image);
        *p_image = NULL;
        return OPJ_FALSE;
    }

    return OPJ_TRUE;
}

static OPJ_BOOL opj_j2k_setup_header_reading(opj_j2k_t *p_j2k,

                                  opj_stream_private_t *p_stream,
                                  opj_event_mgr_t * p_manager)
{
    OPJ_UINT32 l_current_marker = J2K_MS_SOT;
    OPJ_UINT32 l_marker_size;
    const opj_dec_memory_marker_handler_t * l_marker_handler = 00;
    opj_tcp_t * l_tcp = NULL;
    const OPJ_UINT32 l_nb_tiles = p_j2k->m_cp.tw * p_j2k->m_cp.th;

    /* preconditions */
    assert(p_stream != 00);
    assert(p_j2k != 00);
    assert(p_manager != 00);

    /* Reach the End Of Codestream ?*/

                if (!opj_j2k_need_nb_tile_parts_correction(p_stream,
                        p_j2k->m_current_tile_number, &l_correction_needed, p_manager)) {
                    opj_event_msg(p_manager, EVT_ERROR,
                                  "opj_j2k_apply_nb_tile_parts_correction error\n");
                    return OPJ_FALSE;
                }
                if (l_correction_needed) {

                    OPJ_UINT32 l_tile_no;

                    p_j2k->m_specific_param.m_decoder.m_can_decode = 0;
                    p_j2k->m_specific_param.m_decoder.m_nb_tile_parts_correction = 1;
                    /* correct tiles */
                    for (l_tile_no = 0U; l_tile_no < l_nb_tiles; ++l_tile_no) {
                        if (p_j2k->m_cp.tcps[l_tile_no].m_nb_tile_parts != 0U) {
                            p_j2k->m_cp.tcps[l_tile_no].m_nb_tile_parts += 1;
                        }
                    }
                    opj_event_msg(p_manager, EVT_WARNING,
                                  "Non conformant codestream TPsot==TNsot.\n");
                }
            }












        } else {
            /* Indicate we will try to read a new tile-part header*/
            p_j2k->m_specific_param.m_decoder.m_skip_data = 0;
            p_j2k->m_specific_param.m_decoder.m_can_decode = 0;
            p_j2k->m_specific_param.m_decoder.m_state = J2K_STATE_TPHSOT;
        }

        if (! p_j2k->m_specific_param.m_decoder.m_can_decode) {
            /* Try to read 2 bytes (the next marker ID) from stream and copy them into the buffer */
            if (opj_stream_read_data(p_stream,
                                     p_j2k->m_specific_param.m_decoder.m_header_data, 2, p_manager) != 2) {

                /* Deal with likely non conformant SPOT6 files, where the last */
                /* row of tiles have TPsot == 0 and TNsot == 0, and missing EOC, */
                /* but no other tile-parts were found. */
                if (p_j2k->m_current_tile_number + 1 == l_nb_tiles) {
                    OPJ_UINT32 l_tile_no;
                    for (l_tile_no = 0U; l_tile_no < l_nb_tiles; ++l_tile_no) {
                        if (p_j2k->m_cp.tcps[l_tile_no].m_current_tile_part_number == 0 &&
                                p_j2k->m_cp.tcps[l_tile_no].m_nb_tile_parts == 0) {
                            break;
                        }
                    }
                    if (l_tile_no < l_nb_tiles) {
                        opj_event_msg(p_manager, EVT_INFO,
                                      "Tile %u has TPsot == 0 and TNsot == 0, "
                                      "but no other tile-parts were found. "
                                      "EOC is also missing.\n",
                                      l_tile_no);
                        p_j2k->m_current_tile_number = l_tile_no;
                        l_current_marker = J2K_MS_EOC;
                        p_j2k->m_specific_param.m_decoder.m_state = J2K_STATE_EOC;
                        break;
                    }
                }

                opj_event_msg(p_manager, EVT_ERROR, "Stream too short\n");
                return OPJ_FALSE;
            }

            /* Read 2 bytes from buffer as the new marker ID */
            opj_read_bytes(p_j2k->m_specific_param.m_decoder.m_header_data,
                           &l_current_marker, 2);
        }
    }

    /* Current marker is the EOC marker ?*/
    if (l_current_marker == J2K_MS_EOC) {
        if (p_j2k->m_specific_param.m_decoder.m_state != J2K_STATE_EOC) {
            p_j2k->m_current_tile_number = 0;
            p_j2k->m_specific_param.m_decoder.m_state = J2K_STATE_EOC;
        }
    }

    /* Deal with tiles that have a single tile-part with TPsot == 0 and TNsot == 0 */
    if (! p_j2k->m_specific_param.m_decoder.m_can_decode) {

        l_tcp = p_j2k->m_cp.tcps + p_j2k->m_current_tile_number;

        while ((p_j2k->m_current_tile_number < l_nb_tiles) && (l_tcp->m_data == 00)) {
            ++p_j2k->m_current_tile_number;
            ++l_tcp;
        }

    OPJ_UINT32 it_comp;
    OPJ_INT32 l_comp_x1, l_comp_y1;
    opj_image_comp_t* l_img_comp = NULL;

    l_img_comp = p_image->comps;
    for (it_comp = 0; it_comp < p_image->numcomps; ++it_comp) {
        OPJ_INT32 l_h, l_w;
        if (p_image->x0 > (OPJ_UINT32)INT_MAX ||
                p_image->y0 > (OPJ_UINT32)INT_MAX ||
                p_image->x1 > (OPJ_UINT32)INT_MAX ||
                p_image->y1 > (OPJ_UINT32)INT_MAX) {
            opj_event_msg(p_manager, EVT_ERROR,
                          "Image coordinates above INT_MAX are not supported\n");
            return OPJ_FALSE;
        }

        l_img_comp->x0 = (OPJ_UINT32)opj_int_ceildiv((OPJ_INT32)p_image->x0,
                         (OPJ_INT32)l_img_comp->dx);
        l_img_comp->y0 = (OPJ_UINT32)opj_int_ceildiv((OPJ_INT32)p_image->y0,
                         (OPJ_INT32)l_img_comp->dy);
        l_comp_x1 = opj_int_ceildiv((OPJ_INT32)p_image->x1, (OPJ_INT32)l_img_comp->dx);
        l_comp_y1 = opj_int_ceildiv((OPJ_INT32)p_image->y1, (OPJ_INT32)l_img_comp->dy);

    /* make sure room is sufficient */
    if (*p_header_size < 5) {
        opj_event_msg(p_manager, EVT_ERROR, "Error reading SPCod SPCoc element\n");
        return OPJ_FALSE;
    }

    /* SPcod (D) / SPcoc (A) */
    opj_read_bytes(l_current_ptr, &l_tccp->numresolutions, 1);

    ++l_tccp->numresolutions;  /* tccp->numresolutions = read() + 1 */
    if (l_tccp->numresolutions > OPJ_J2K_MAXRLVLS) {
        opj_event_msg(p_manager, EVT_ERROR,
                      "Invalid value for numresolutions : %d, max value is set in openjpeg.h at %d\n",
                      l_tccp->numresolutions, OPJ_J2K_MAXRLVLS);
        return OPJ_FALSE;
    }
    ++l_current_ptr;

    /* If user wants to remove more resolutions than the codestream contains, return error */
    if (l_cp->m_specific_param.m_dec.m_reduce >= l_tccp->numresolutions) {
        opj_event_msg(p_manager, EVT_ERROR,
                      "Error decoding component %d.\nThe number of resolutions "
                      "to remove (%d) is greater or equal than the number "
                      "of resolutions of this component (%d)\nModify the cp_reduce parameter.\n\n",
                      compno, l_cp->m_specific_param.m_dec.m_reduce, l_tccp->numresolutions);
        p_j2k->m_specific_param.m_decoder.m_state |=
            0x8000;/* FIXME J2K_DEC_STATE_ERR;*/
        return OPJ_FALSE;
    }

    /* SPcod (E) / SPcoc (B) */
    opj_read_bytes(l_current_ptr, &l_tccp->cblkw, 1);
    ++l_current_ptr;
    l_tccp->cblkw += 2;

    /* SPcod (F) / SPcoc (C) */
    opj_read_bytes(l_current_ptr, &l_tccp->cblkh, 1);
    ++l_current_ptr;
    l_tccp->cblkh += 2;

    if ((l_tccp->cblkw > 10) || (l_tccp->cblkh > 10) ||
            ((l_tccp->cblkw + l_tccp->cblkh) > 12)) {
        opj_event_msg(p_manager, EVT_ERROR,
                      "Error reading SPCod SPCoc element, Invalid cblkw/cblkh combination\n");
        return OPJ_FALSE;
    }

    /* SPcod (G) / SPcoc (D) */
    opj_read_bytes(l_current_ptr, &l_tccp->cblksty, 1);
    ++l_current_ptr;
    if (l_tccp->cblksty & 0xC0U) { /* 2 msb are reserved, assume we can't read */
        opj_event_msg(p_manager, EVT_ERROR,
                      "Error reading SPCod SPCoc element, Invalid code-block style found\n");
        return OPJ_FALSE;
    }

    /* SPcod (H) / SPcoc (E) */
    opj_read_bytes(l_current_ptr, &l_tccp->qmfbid, 1);
    ++l_current_ptr;

    if (l_tccp->qmfbid > 1) {
        opj_event_msg(p_manager, EVT_ERROR,
                      "Error reading SPCod SPCoc element, Invalid transformation found\n");
        return OPJ_FALSE;
    }

    *p_header_size = *p_header_size - 5;

    /* use custom precinct size ? */
    if (l_tccp->csty & J2K_CCP_CSTY_PRT) {
        if (*p_header_size < l_tccp->numresolutions) {
            opj_event_msg(p_manager, EVT_ERROR, "Error reading SPCod SPCoc element\n");
            return OPJ_FALSE;
        }

        /* SPcod (I_i) / SPcoc (F_i) */
        for (i = 0; i < l_tccp->numresolutions; ++i) {
            opj_read_bytes(l_current_ptr, &l_tmp, 1);
            ++l_current_ptr;
            /* Precinct exponent 0 is only allowed for lowest resolution level (Table A.21) */
            if ((i != 0) && (((l_tmp & 0xf) == 0) || ((l_tmp >> 4) == 0))) {
                opj_event_msg(p_manager, EVT_ERROR, "Invalid precinct size\n");
                return OPJ_FALSE;
            }
            l_tccp->prcw[i] = l_tmp & 0xf;

        if (!p_j2k->cstr_index->tile_index[it_tile].marker) {
            return OPJ_FALSE;
        }
    }

    return OPJ_TRUE;
}

static OPJ_BOOL opj_j2k_are_all_used_components_decoded(opj_j2k_t *p_j2k,
        opj_event_mgr_t * p_manager)
{
    OPJ_UINT32 compno;
    OPJ_BOOL decoded_all_used_components = OPJ_TRUE;

    if (p_j2k->m_specific_param.m_decoder.m_numcomps_to_decode) {
        for (compno = 0;
                compno < p_j2k->m_specific_param.m_decoder.m_numcomps_to_decode; compno++) {
            OPJ_UINT32 dec_compno =
                p_j2k->m_specific_param.m_decoder.m_comps_indices_to_decode[compno];
            if (p_j2k->m_output_image->comps[dec_compno].data == NULL) {
                opj_event_msg(p_manager, EVT_WARNING, "Failed to decode component %d\n",
                              dec_compno);
                decoded_all_used_components = OPJ_FALSE;
            }
        }
    } else {
        for (compno = 0; compno < p_j2k->m_output_image->numcomps; compno++) {
            if (p_j2k->m_output_image->comps[compno].data == NULL) {
                opj_event_msg(p_manager, EVT_WARNING, "Failed to decode component %d\n",
                              compno);
                decoded_all_used_components = OPJ_FALSE;
            }
        }
    }

    if (decoded_all_used_components == OPJ_FALSE) {
        opj_event_msg(p_manager, EVT_ERROR, "Failed to decode all used components\n");
        return OPJ_FALSE;
    }

    return OPJ_TRUE;
}


static OPJ_BOOL opj_j2k_decode_tiles(opj_j2k_t *p_j2k,
                                     opj_stream_private_t *p_stream,
                                     opj_event_mgr_t * p_manager)
{
    OPJ_BOOL l_go_on = OPJ_TRUE;
    OPJ_UINT32 l_current_tile_no;

                && p_j2k->m_specific_param.m_decoder.m_state == J2K_STATE_NEOC) {
            break;
        }
        if (++nr_tiles ==  p_j2k->m_cp.th * p_j2k->m_cp.tw) {
            break;
        }
    }

    if (! opj_j2k_are_all_used_components_decoded(p_j2k, p_manager)) {
        return OPJ_FALSE;
    }

    return OPJ_TRUE;
}

/**
 * Sets up the procedures to do on decoding data. Developpers wanting to extend the library can add their own reading procedures.
 */

        } else {
            opj_event_msg(p_manager, EVT_WARNING,
                          "Tile read, decoded and updated is not the desired one (%d vs %d).\n",
                          l_current_tile_no + 1, l_tile_no_to_dec + 1);
        }

    }

    if (! opj_j2k_are_all_used_components_decoded(p_j2k, p_manager)) {
        return OPJ_FALSE;
    }

    return OPJ_TRUE;
}

/**
 * Sets up the procedures to do on decoding one tile. Developpers wanting to extend the library can add their own reading procedures.
 */

            }
        }
    }

    return OPJ_FALSE;
}

/* ----------------------------------------------------------------------- */

OPJ_BOOL opj_j2k_encoder_set_extra_options(
    opj_j2k_t *p_j2k,
    const char* const* p_options,
    opj_event_mgr_t * p_manager)
{
    const char* const* p_option_iter;

    if (p_options == NULL) {
        return OPJ_TRUE;
    }

    for (p_option_iter = p_options; *p_option_iter != NULL; ++p_option_iter) {
        if (strncmp(*p_option_iter, "PLT=", 4) == 0) {
            if (strcmp(*p_option_iter, "PLT=YES") == 0) {
                p_j2k->m_specific_param.m_encoder.m_PLT = OPJ_TRUE;
            } else if (strcmp(*p_option_iter, "PLT=NO") == 0) {
                p_j2k->m_specific_param.m_encoder.m_PLT = OPJ_FALSE;
            } else {
                opj_event_msg(p_manager, EVT_ERROR,
                              "Invalid value for option: %s.\n", *p_option_iter);
                return OPJ_FALSE;
            }
        } else {
            opj_event_msg(p_manager, EVT_ERROR,
                          "Invalid option: %s.\n", *p_option_iter);
            return OPJ_FALSE;
        }
    }

    return OPJ_TRUE;
}

/* ----------------------------------------------------------------------- */

OPJ_BOOL opj_j2k_encode(opj_j2k_t * p_j2k,
                        opj_stream_private_t *p_stream,
                        opj_event_mgr_t * p_manager)
{
    OPJ_UINT32 i, j;
    OPJ_UINT32 l_nb_tiles;
    OPJ_SIZE_T l_max_tile_size = 0, l_current_tile_size;

                    if (l_current_data) {
                        opj_free(l_current_data);
                    }
                    return OPJ_FALSE;
                }
            }
        }
        l_current_tile_size = opj_tcd_get_encoder_input_buffer_size(p_j2k->m_tcd);
        if (!l_reuse_data) {
            if (l_current_tile_size > l_max_tile_size) {
                OPJ_BYTE *l_new_current_data = (OPJ_BYTE *) opj_realloc(l_current_data,
                                               l_current_tile_size);
                if (! l_new_current_data) {
                    if (l_current_data) {
                        opj_free(l_current_data);

    /* DEVELOPER CORNER, insert your custom procedures */
    if (! opj_procedure_list_add_procedure(p_j2k->m_procedure_list,
                                           (opj_procedure)opj_j2k_write_eoc, p_manager)) {
        return OPJ_FALSE;
    }

    if (OPJ_IS_CINEMA(p_j2k->m_cp.rsiz) || OPJ_IS_IMF(p_j2k->m_cp.rsiz)) {
        if (! opj_procedure_list_add_procedure(p_j2k->m_procedure_list,
                                               (opj_procedure)opj_j2k_write_updated_tlm, p_manager)) {
            return OPJ_FALSE;
        }
    }

    if (! opj_procedure_list_add_procedure(p_j2k->m_procedure_list,

        return OPJ_FALSE;
    }
    if (! opj_procedure_list_add_procedure(p_j2k->m_procedure_list,
                                           (opj_procedure)opj_j2k_write_all_qcc, p_manager)) {
        return OPJ_FALSE;
    }

    if (OPJ_IS_CINEMA(p_j2k->m_cp.rsiz) || OPJ_IS_IMF(p_j2k->m_cp.rsiz)) {
        if (! opj_procedure_list_add_procedure(p_j2k->m_procedure_list,
                                               (opj_procedure)opj_j2k_write_tlm, p_manager)) {
            return OPJ_FALSE;
        }

        if (p_j2k->m_cp.rsiz == OPJ_PROFILE_CINEMA_4K) {
            if (! opj_procedure_list_add_procedure(p_j2k->m_procedure_list,

        if (! opj_procedure_list_add_procedure(p_j2k->m_procedure_list,
                                               (opj_procedure)opj_j2k_write_com, p_manager)) {
            return OPJ_FALSE;
        }
    }

    /* DEVELOPER CORNER, insert your custom procedures */
    if ((p_j2k->m_cp.rsiz & (OPJ_PROFILE_PART2 | OPJ_EXTENSION_MCT)) ==
            (OPJ_PROFILE_PART2 | OPJ_EXTENSION_MCT)) {
        if (! opj_procedure_list_add_procedure(p_j2k->m_procedure_list,
                                               (opj_procedure)opj_j2k_write_mct_data_group, p_manager)) {
            return OPJ_FALSE;
        }
    }
    /* End of Developer Corner */

    return OPJ_TRUE;
}

static OPJ_BOOL opj_j2k_write_first_tile_part(opj_j2k_t *p_j2k,
        OPJ_BYTE * p_data,
        OPJ_UINT32 * p_data_written,
        OPJ_UINT32 total_data_size,
        opj_stream_private_t *p_stream,
        struct opj_event_mgr * p_manager)
{
    OPJ_UINT32 l_nb_bytes_written = 0;
    OPJ_UINT32 l_current_nb_bytes_written;
    OPJ_BYTE * l_begin_data = 00;

    p_j2k->m_specific_param.m_encoder.m_current_poc_tile_part_number = 0;

    /* INDEX >> */
    /* << INDEX */

    l_current_nb_bytes_written = 0;
    l_begin_data = p_data;
    if (! opj_j2k_write_sot(p_j2k, p_data, total_data_size,
                            &l_current_nb_bytes_written, p_stream,
                            p_manager)) {
        return OPJ_FALSE;
    }

    l_nb_bytes_written += l_current_nb_bytes_written;
    p_data += l_current_nb_bytes_written;
    total_data_size -= l_current_nb_bytes_written;

    if (!OPJ_IS_CINEMA(l_cp->rsiz)) {
#if 0
        for (compno = 1; compno < p_j2k->m_private_image->numcomps; compno++) {
            l_current_nb_bytes_written = 0;
            opj_j2k_write_coc_in_memory(p_j2k, compno, p_data, &l_current_nb_bytes_written,
                                        p_manager);
            l_nb_bytes_written += l_current_nb_bytes_written;
            p_data += l_current_nb_bytes_written;
            total_data_size -= l_current_nb_bytes_written;

            l_current_nb_bytes_written = 0;
            opj_j2k_write_qcc_in_memory(p_j2k, compno, p_data, &l_current_nb_bytes_written,
                                        p_manager);
            l_nb_bytes_written += l_current_nb_bytes_written;
            p_data += l_current_nb_bytes_written;
            total_data_size -= l_current_nb_bytes_written;
        }
#endif
        if (l_cp->tcps[p_j2k->m_current_tile_number].POC) {
            l_current_nb_bytes_written = 0;
            opj_j2k_write_poc_in_memory(p_j2k, p_data, &l_current_nb_bytes_written,
                                        p_manager);
            l_nb_bytes_written += l_current_nb_bytes_written;
            p_data += l_current_nb_bytes_written;
            total_data_size -= l_current_nb_bytes_written;
        }
    }

    l_current_nb_bytes_written = 0;
    if (! opj_j2k_write_sod(p_j2k, l_tcd, p_data, &l_current_nb_bytes_written,
                            total_data_size, p_stream, p_manager)) {
        return OPJ_FALSE;
    }

    l_nb_bytes_written += l_current_nb_bytes_written;
    * p_data_written = l_nb_bytes_written;

    /* Writing Psot in SOT marker */
    opj_write_bytes(l_begin_data + 6, l_nb_bytes_written,
                    4);                                 /* PSOT */

    if (OPJ_IS_CINEMA(l_cp->rsiz) || OPJ_IS_IMF(l_cp->rsiz)) {
        opj_j2k_update_tlm(p_j2k, l_nb_bytes_written);
    }

    return OPJ_TRUE;
}

static OPJ_BOOL opj_j2k_write_all_tile_parts(opj_j2k_t *p_j2k,
        OPJ_BYTE * p_data,
        OPJ_UINT32 * p_data_written,
        OPJ_UINT32 total_data_size,
        opj_stream_private_t *p_stream,
        struct opj_event_mgr * p_manager
                                            )
{
    OPJ_UINT32 tilepartno = 0;
    OPJ_UINT32 l_nb_bytes_written = 0;
    OPJ_UINT32 l_current_nb_bytes_written;

    for (tilepartno = 1; tilepartno < tot_num_tp ; ++tilepartno) {
        p_j2k->m_specific_param.m_encoder.m_current_poc_tile_part_number = tilepartno;
        l_current_nb_bytes_written = 0;
        l_part_tile_size = 0;
        l_begin_data = p_data;

        if (! opj_j2k_write_sot(p_j2k, p_data,
                                total_data_size,
                                &l_current_nb_bytes_written,
                                p_stream,
                                p_manager)) {
            return OPJ_FALSE;
        }

        l_nb_bytes_written += l_current_nb_bytes_written;
        p_data += l_current_nb_bytes_written;
        total_data_size -= l_current_nb_bytes_written;
        l_part_tile_size += l_current_nb_bytes_written;

        l_current_nb_bytes_written = 0;
        if (! opj_j2k_write_sod(p_j2k, l_tcd, p_data, &l_current_nb_bytes_written,
                                total_data_size, p_stream, p_manager)) {
            return OPJ_FALSE;
        }

        p_data += l_current_nb_bytes_written;
        l_nb_bytes_written += l_current_nb_bytes_written;
        total_data_size -= l_current_nb_bytes_written;
        l_part_tile_size += l_current_nb_bytes_written;

        /* Writing Psot in SOT marker */
        opj_write_bytes(l_begin_data + 6, l_part_tile_size,
                        4);                                   /* PSOT */

        if (OPJ_IS_CINEMA(l_cp->rsiz) || OPJ_IS_IMF(l_cp->rsiz)) {
            opj_j2k_update_tlm(p_j2k, l_part_tile_size);
        }

        ++p_j2k->m_specific_param.m_encoder.m_current_tile_part_number;
    }

    for (pino = 1; pino <= l_tcp->numpocs; ++pino) {
        l_tcd->cur_pino = pino;

        /*Get number of tile parts*/
        tot_num_tp = opj_j2k_get_num_tp(l_cp, pino, p_j2k->m_current_tile_number);
        for (tilepartno = 0; tilepartno < tot_num_tp ; ++tilepartno) {
            p_j2k->m_specific_param.m_encoder.m_current_poc_tile_part_number = tilepartno;
            l_current_nb_bytes_written = 0;
            l_part_tile_size = 0;
            l_begin_data = p_data;

            if (! opj_j2k_write_sot(p_j2k, p_data,
                                    total_data_size,
                                    &l_current_nb_bytes_written, p_stream,
                                    p_manager)) {
                return OPJ_FALSE;
            }

            l_nb_bytes_written += l_current_nb_bytes_written;
            p_data += l_current_nb_bytes_written;
            total_data_size -= l_current_nb_bytes_written;
            l_part_tile_size += l_current_nb_bytes_written;

            l_current_nb_bytes_written = 0;

            if (! opj_j2k_write_sod(p_j2k, l_tcd, p_data, &l_current_nb_bytes_written,
                                    total_data_size, p_stream, p_manager)) {
                return OPJ_FALSE;
            }

            l_nb_bytes_written += l_current_nb_bytes_written;
            p_data += l_current_nb_bytes_written;
            total_data_size -= l_current_nb_bytes_written;
            l_part_tile_size += l_current_nb_bytes_written;

            /* Writing Psot in SOT marker */
            opj_write_bytes(l_begin_data + 6, l_part_tile_size,
                            4);                                   /* PSOT */

            if (OPJ_IS_CINEMA(l_cp->rsiz) || OPJ_IS_IMF(l_cp->rsiz)) {
                opj_j2k_update_tlm(p_j2k, l_part_tile_size);
            }

            ++p_j2k->m_specific_param.m_encoder.m_current_tile_part_number;
        }
    }

    /* size of the encoded_data */
    OPJ_UINT32 m_encoded_tile_size;

    /* encoded data for a tile */
    OPJ_BYTE * m_header_tile_data;

    /* size of the encoded_data */

    OPJ_UINT32 m_header_tile_data_size;

    /* whether to generate PLT markers */
    OPJ_BOOL   m_PLT;

    /* reserved bytes in m_encoded_tile_size for PLT markers */
    OPJ_UINT32 m_reserved_bytes_for_PLT;

} opj_j2k_enc_t;



struct opj_tcd;
/**

    /** number of the tile currently concern by coding/decoding */
    OPJ_UINT32 m_current_tile_number;

    /** the current tile coder/decoder **/
    struct opj_tcd *    m_tcd;




    /** Thread pool */
    opj_thread_pool_t* m_tp;

    /** Image width coming from JP2 IHDR box. 0 from a pure codestream */
    OPJ_UINT32 ihdr_w;

    /** Image height coming from JP2 IHDR box. 0 from a pure codestream */
    OPJ_UINT32 ihdr_h;

    /** Set to 1 by the decoder initialization if OPJ_DPARAMETERS_DUMP_FLAG is set */
    unsigned int dump_state;
}
opj_j2k_t;

                          opj_event_mgr_t * p_manager,
                          OPJ_UINT32 tile_index);

OPJ_BOOL opj_j2k_set_decoded_resolution_factor(opj_j2k_t *p_j2k,
        OPJ_UINT32 res_factor,
        opj_event_mgr_t * p_manager);

/**
 * Specify extra options for the encoder.
 *
 * @param  p_j2k        the jpeg2000 codec.
 * @param  p_options    options
 * @param  p_manager    the user event manager
 *
 * @see opj_encoder_set_extra_options() for more details.
 */
OPJ_BOOL opj_j2k_encoder_set_extra_options(
    opj_j2k_t *p_j2k,
    const char* const* p_options,
    opj_event_mgr_t * p_manager);

/**
 * Writes a tile.
 * @param   p_j2k       the jpeg2000 codec.
 * @param p_tile_index FIXME DOC
 * @param p_data FIXME DOC
 * @param p_data_size FIXME DOC

    opj_read_bytes(p_image_header_data, &(jp2->h), 4);          /* HEIGHT */
    p_image_header_data += 4;
    opj_read_bytes(p_image_header_data, &(jp2->w), 4);          /* WIDTH */
    p_image_header_data += 4;
    opj_read_bytes(p_image_header_data, &(jp2->numcomps), 2);   /* NC */
    p_image_header_data += 2;

    if (jp2->h < 1 || jp2->w < 1 || jp2->numcomps < 1) {
        opj_event_msg(p_manager, EVT_ERROR,
                      "Wrong values for: w(%d) h(%d) numcomps(%d) (ihdr)\n",
                      jp2->w, jp2->h, jp2->numcomps);
        return OPJ_FALSE;
    }
    if ((jp2->numcomps - 1U) >=
            16384U) { /* unsigned underflow is well defined: 1U <= jp2->numcomps <= 16384U */
        opj_event_msg(p_manager, EVT_ERROR, "Invalid number of components (ihdr)\n");
        return OPJ_FALSE;
    }

    /* allocate memory for components */

    } else if (jp2->meth > 2) {
        /*  ISO/IEC 15444-1:2004 (E), Table I.9 Legal METH values:
        conforming JP2 reader shall ignore the entire Colour Specification box.*/
        opj_event_msg(p_manager, EVT_INFO,
                      "COLR BOX meth value is not a regular value (%d), "
                      "so we will ignore the entire Colour Specification box. \n", jp2->meth);
    }



    return OPJ_TRUE;
}

OPJ_BOOL opj_jp2_decode(opj_jp2_t *jp2,
                        opj_stream_private_t *p_stream,
                        opj_image_t* p_image,
                        opj_event_mgr_t * p_manager)

OPJ_BOOL opj_jp2_set_decoded_resolution_factor(opj_jp2_t *p_jp2,
        OPJ_UINT32 res_factor,
        opj_event_mgr_t * p_manager)
{
    return opj_j2k_set_decoded_resolution_factor(p_jp2->j2k, res_factor, p_manager);
}

/* ----------------------------------------------------------------------- */

OPJ_BOOL opj_jp2_encoder_set_extra_options(
    opj_jp2_t *p_jp2,
    const char* const* p_options,
    opj_event_mgr_t * p_manager)
{
    return opj_j2k_encoder_set_extra_options(p_jp2->j2k, p_options, p_manager);
}

/* ----------------------------------------------------------------------- */

/* JPIP specific */

#ifdef USE_JPIP
static OPJ_BOOL opj_jpip_write_iptr(opj_jp2_t *jp2,
                                    opj_stream_private_t *cio,
                                    opj_event_mgr_t * p_manager)

/**
 *
 */
OPJ_BOOL opj_jp2_set_decoded_resolution_factor(opj_jp2_t *p_jp2,
        OPJ_UINT32 res_factor,
        opj_event_mgr_t * p_manager);

/**
 * Specify extra options for the encoder.
 *
 * @param  p_jp2        the jpeg2000 codec.
 * @param  p_options    options
 * @param  p_manager    the user event manager
 *
 * @see opj_encoder_set_extra_options() for more details.
 */
OPJ_BOOL opj_jp2_encoder_set_extra_options(
    opj_jp2_t *p_jp2,
    const char* const* p_options,
    opj_event_mgr_t * p_manager);


/* TODO MSD: clean these 3 functions */
/**
 * Dump some elements from the JP2 decompression structure .
 *
 *@param p_jp2        the jp2 codec.
 *@param flag        flag to describe what elements are dump.

#else
void opj_mct_decode(
    OPJ_INT32* OPJ_RESTRICT c0,
    OPJ_INT32* OPJ_RESTRICT c1,
    OPJ_INT32* OPJ_RESTRICT c2,
    OPJ_SIZE_T n)
{
    OPJ_SIZE_T i;
    for (i = 0; i < n; ++i) {
        OPJ_INT32 y = c0[i];
        OPJ_INT32 u = c1[i];
        OPJ_INT32 v = c2[i];
        OPJ_INT32 g = y - ((u + v) >> 2);
        OPJ_INT32 r = v + g;
        OPJ_INT32 b = u + g;

{
    return opj_mct_norms[compno];
}

/* <summary> */
/* Forward irreversible MCT. */
/* </summary> */

void opj_mct_encode_real(
    OPJ_FLOAT32* OPJ_RESTRICT c0,
    OPJ_FLOAT32* OPJ_RESTRICT c1,
    OPJ_FLOAT32* OPJ_RESTRICT c2,
    OPJ_SIZE_T n)
{
    OPJ_SIZE_T i;

#ifdef __SSE__
    const __m128 YR = _mm_set1_ps(0.299f);
    const __m128 YG = _mm_set1_ps(0.587f);
    const __m128 YB = _mm_set1_ps(0.114f);
    const __m128 UR = _mm_set1_ps(-0.16875f);
    const __m128 UG = _mm_set1_ps(-0.331260f);
    const __m128 UB = _mm_set1_ps(0.5f);

    const __m128 VR = _mm_set1_ps(0.5f);
    const __m128 VG = _mm_set1_ps(-0.41869f);


    const __m128 VB = _mm_set1_ps(-0.08131f);
    for (i = 0; i < (n >> 3); i ++) {

        __m128 r, g, b, y, u, v;
















        r = _mm_load_ps(c0);
        g = _mm_load_ps(c1);
        b = _mm_load_ps(c2);
        y = _mm_add_ps(_mm_add_ps(_mm_mul_ps(r, YR), _mm_mul_ps(g, YG)),
                       _mm_mul_ps(b, YB));


        u = _mm_add_ps(_mm_add_ps(_mm_mul_ps(r, UR), _mm_mul_ps(g, UG)),


                       _mm_mul_ps(b, UB));

        v = _mm_add_ps(_mm_add_ps(_mm_mul_ps(r, VR), _mm_mul_ps(g, VG)),
                       _mm_mul_ps(b, VB));
        _mm_store_ps(c0, y);
        _mm_store_ps(c1, u);

        _mm_store_ps(c2, v);
        c0 += 4;
        c1 += 4;





        c2 += 4;









        r = _mm_load_ps(c0);
        g = _mm_load_ps(c1);


        b = _mm_load_ps(c2);


        y = _mm_add_ps(_mm_add_ps(_mm_mul_ps(r, YR), _mm_mul_ps(g, YG)),


                       _mm_mul_ps(b, YB));

        u = _mm_add_ps(_mm_add_ps(_mm_mul_ps(r, UR), _mm_mul_ps(g, UG)),

                       _mm_mul_ps(b, UB));



        v = _mm_add_ps(_mm_add_ps(_mm_mul_ps(r, VR), _mm_mul_ps(g, VG)),


                       _mm_mul_ps(b, VB));



        _mm_store_ps(c0, y);
        _mm_store_ps(c1, u);


        _mm_store_ps(c2, v);


        c0 += 4;
        c1 += 4;

        c2 += 4;






    }
    n &= 7;



#endif






    for (i = 0; i < n; ++i) {
        OPJ_FLOAT32 r = c0[i];
        OPJ_FLOAT32 g = c1[i];
        OPJ_FLOAT32 b = c2[i];
        OPJ_FLOAT32 y = 0.299f * r + 0.587f * g + 0.114f * b;

        OPJ_FLOAT32 u = -0.16875f * r - 0.331260f * g + 0.5f * b;

        OPJ_FLOAT32 v = 0.5f * r - 0.41869f * g - 0.08131f * b;

        c0[i] = y;
        c1[i] = u;
        c2[i] = v;
    }
}

























/* <summary> */
/* Inverse irreversible MCT. */
/* </summary> */
void opj_mct_decode_real(
    OPJ_FLOAT32* OPJ_RESTRICT c0,
    OPJ_FLOAT32* OPJ_RESTRICT c1,
    OPJ_FLOAT32* OPJ_RESTRICT c2,
    OPJ_SIZE_T n)
{
    OPJ_SIZE_T i;
#ifdef __SSE__
    __m128 vrv, vgu, vgv, vbu;
    vrv = _mm_set1_ps(1.402f);
    vgu = _mm_set1_ps(0.34413f);
    vgv = _mm_set1_ps(0.71414f);
    vbu = _mm_set1_ps(1.772f);
    for (i = 0; i < (n >> 3); ++i) {