Stride access for DWT and some bug fixes and small improvements

.gitattributes

@@ -0,0 +1,17 @@
+# Set the default behavior, in case people don't have core.autocrlf set.
+* text=auto
+
+# Explicitly declare text files you want to always be normalized and converted
+# to native line endings on checkout.
+*.c text
+*.h text
+*.cpp text
+*.hpp text
+
+# Declare files that will always have CRLF line endings on checkout.
+*.sln text eol=crlf
+
+# Denote all files that are truly binary and should not be modified.
+*.pgx binary
+*.j2k binary
+*.ppm binary

CHANGELOG

@@ -1,3 +1,9 @@
+# [0.2.8] - 2024-11-12
+
+* Fix incorrect packet parsing for RPCL, PCRL, CPRL
+* Introduce stride access into DWT
+* Change cmake configuration for MinGW environments
+
 # [0.2.7] - 2024-06-13
 
 * Refactor non-SIMD HT cleanup decoding

CMakeLists.txt

@@ -153,9 +153,7 @@ endif()
 
 if(NOT EMSCRIPTEN)
   if(CMAKE_HOST_SYSTEM_PROCESSOR MATCHES "^[xX]86_64$|^[aA][mM][dD]64$") # x86_64
-    if(NOT MINGW)
-      option(ENABLE_AVX2 "Enable the use of Intel AVX2 intrinsics" ON)
-    endif()
+    option(ENABLE_AVX2 "Enable the use of Intel AVX2 intrinsics" ON)
     if(CMAKE_CXX_COMPILER_ID MATCHES "MSVC")
       set(CMAKE_CXX_FLAGS
           "${CMAKE_CXX_FLAGS} /arch:AVX2 /EHsc /D \"_CRT_SECURE_NO_WARNINGS\"")

source/apps/imgcmp/image_class.hpp

@@ -70,7 +70,7 @@ class image {
         bitDepth(0),
         isSigned(false),
         isBigendian(false),
-        data(nullptr) {};
+        data(nullptr){};
   // // destructor
   ~image() { delete[] data; }
 
@@ -110,9 +110,9 @@ class image {
 
   // parsing PNM/PGX header
   int read_pnmpgx(const char *name) {
-    constexpr char SP = ' ';
-    constexpr char LF = '\n';
-    constexpr char CR = 13;
+    constexpr char SP                  = ' ';
+    constexpr char LF                  = '\n';
+    [[maybe_unused]] constexpr char CR = 13;
 
     FILE *fp = fopen(name, "rb");
     if (fp == nullptr) {
@@ -223,7 +223,7 @@ class image {
         }
       }
       // read numerical value
-      while (c != SP && c != LF && c != CR) {
+      while (c >= '0' && c <= '9') {
         val *= 10;
         val += c - '0';
         c = fgetc(fp);
@@ -255,8 +255,10 @@ class image {
       }
     }
     // easting trailing spaces/LF/CR or comments
-    c = fgetc(fp);
+    c         = fgetc(fp);
+    int count = 0;
     while (c == SP || c == LF) {
+      count++;
       c = fgetc(fp);
       if (c == '#') {
         char *nouse = fgets(comment, sizeof(comment), fp);
@@ -267,7 +269,7 @@ class image {
         c = fgetc(fp);
       }
     }
-    fseek(fp, -1, SEEK_CUR);
+    fseek(fp, -count - 1, SEEK_CUR);
 
     const uint_fast8_t nbytes = static_cast<uint_fast8_t>((bitDepth + 7) / 8);  // ceil bitDepth to byte
     const size_t num_samples  = this->width * this->height * num_components;

source/core/coding/coding_units.cpp

@@ -267,8 +267,8 @@ void j2k_codeblock::create_compressed_buffer(buf_chain *tile_buf, int32_t buf_li
  *******************************************************************************/
 j2k_precinct_subband::j2k_precinct_subband(uint8_t orientation, uint8_t M_b, uint8_t R_b,
                                            uint8_t transformation, float stepsize, sprec_t *ibuf,
-                                           const element_siz &bp0, const element_siz &bp1,
-                                           const element_siz &p0, const element_siz &p1,
+                                           const element_siz &bp0, const element_siz &p0,
+                                           const element_siz &p1, const uint32_t band_stride,
                                            const uint16_t &num_layers, const element_siz &codeblock_size,
                                            const uint8_t &Cmodes)
     : j2k_region(p0, p1),
@@ -290,7 +290,7 @@ j2k_precinct_subband::j2k_precinct_subband(uint8_t orientation, uint8_t M_b, uin
   }
 
   const uint32_t num_codeblocks = this->num_codeblock_x * this->num_codeblock_y;
-  const uint32_t band_stride    = bp1.x - bp0.x;
+  //  const uint32_t band_stride    = stride;
   if (num_codeblocks != 0) {
     inclusion_info = new tagtree(this->num_codeblock_x, this->num_codeblock_y);
     ZBP_info       = new tagtree(this->num_codeblock_x, this->num_codeblock_y);
@@ -551,9 +551,7 @@ void j2k_precinct_subband::parse_packet_header(buf_chain *packet_header, uint16_
             if (!(block->Cmodes & HT_MIXED)) {
               // Must be the first HT Cleanup pass
               if (segment_bytes < 2) {
-                printf(
-                    "ERROR: Length information for a HT-codeblock is "
-                    "invalid\n");
+                printf("ERROR: Length information for a HT-codeblock is invalid\n");
                 throw std::exception();
               }
               next_segment_passes = 2;
@@ -595,9 +593,7 @@ void j2k_precinct_subband::parse_packet_header(buf_chain *packet_header, uint16_
                 if (block->Cmodes & HT_MIXED) {
                   block->Cmodes &= static_cast<uint16_t>(~(HT_PHLD | HT));
                 } else {
-                  printf(
-                      "ERROR: Length information for a HT-codeblock is "
-                      "invalid\n");
+                  printf("ERROR: Length information for a HT-codeblock is invalid\n");
                   throw std::exception();
                 }
               }
@@ -1028,7 +1024,7 @@ j2k_precinct::j2k_precinct(const uint8_t &r, const uint32_t &idx, const element_
                              ceil_int(pos1.y - yob[subband[i]->orientation], sr));
     this->pband[i] = MAKE_UNIQUE<j2k_precinct_subband>(
         subband[i]->orientation, subband[i]->M_b, subband[i]->R_b, subband[i]->transformation,
-        subband[i]->delta, subband[i]->i_samples, subband[i]->pos0, subband[i]->pos1, pbpos0, pbpos1,
+        subband[i]->delta, subband[i]->i_samples, subband[i]->pos0, pbpos0, pbpos1, subband[i]->stride,
         num_layers, codeblock_size, Cmodes);
   }
 }
@@ -1059,8 +1055,9 @@ j2k_subband::j2k_subband(element_siz p0, element_siz p1, uint8_t orientation, ui
   if (num_samples) {
     if (orientation != BAND_LL) {
       // If not the lowest resolution, buffers for subbands shall be created.
-      i_samples = static_cast<sprec_t *>(aligned_mem_alloc(sizeof(sprec_t) * num_samples, 32));
-      memset(i_samples, 0, sizeof(sprec_t) * num_samples);
+      i_samples =
+          static_cast<sprec_t *>(aligned_mem_alloc(sizeof(sprec_t) * this->stride * (pos1.y - pos0.y), 32));
+      memset(i_samples, 0, sizeof(sprec_t) * this->stride * (pos1.y - pos0.y));
     } else {
       i_samples = ibuf;
     }
@@ -1096,10 +1093,12 @@ j2k_resolution::j2k_resolution(const uint8_t &r, const element_siz &p0, const el
   i_samples = nullptr;
   if (!is_empty) {
     if (index == 0) {
-      i_samples = static_cast<sprec_t *>(aligned_mem_alloc(sizeof(sprec_t) * num_samples, 32));
+      i_samples =
+          static_cast<sprec_t *>(aligned_mem_alloc(sizeof(sprec_t) * this->stride * (pos1.y - pos0.y), 32));
       memset(i_samples, 0, sizeof(sprec_t) * num_samples);
     } else {
-      i_samples = static_cast<sprec_t *>(aligned_mem_alloc(sizeof(sprec_t) * num_samples, 32));
+      i_samples =
+          static_cast<sprec_t *>(aligned_mem_alloc(sizeof(sprec_t) * this->stride * (pos1.y - pos0.y), 32));
     }
   }
 }
@@ -2504,7 +2503,7 @@ void j2k_tile::decode() {
 
     // copy samples in resolution buffer to that in tile component buffer
     uint32_t height = tc1.y - tc0.y;
-    uint32_t width  = tc1.x - tc0.x;
+    uint32_t width  = round_up(tc1.x - tc0.x, 32U);
     uint32_t stride = round_up(width, 32U);
     // size_t num_samples = static_cast<size_t>(tc1.x - tc0.x) * (tc1.y - tc0.y);
 #if defined(OPENHTJ2K_ENABLE_ARM_NEON)
@@ -2664,7 +2663,7 @@ void j2k_tile::find_gcd_of_precinct_size(element_siz &out) {
     for (uint8_t r = 0; r <= this->tcomp[c].get_dwt_levels(); r++) {
       PP  = this->tcomp[c].get_precinct_size(r);
       PPx = (PPx > PP.x) ? static_cast<uint8_t>(PP.x) : PPx;
-      PPy = (PPy > PP.y) ? static_cast<uint8_t>(PP.y) : PPx;
+      PPy = (PPy > PP.y) ? static_cast<uint8_t>(PP.y) : PPy;
     }
   }
   out.x = PPx;
@@ -3010,7 +3009,7 @@ uint8_t *j2k_tile::encode() {
 #if defined(OPENHTJ2K_TRY_AVX2) && defined(__AVX2__)
     for (uint32_t y = 0; y < height; ++y) {
       int32_t *sp             = src + y * stride;
-      sprec_t *dp             = cr->i_samples + y * (bottom_right.x - top_left.x);
+      sprec_t *dp             = cr->i_samples + y * stride;
       uint32_t num_tc_samples = bottom_right.x - top_left.x;
       for (; num_tc_samples >= 16; num_tc_samples -= 16) {
         __m256i v0 = _mm256_load_si256((__m256i *)sp);
@@ -3027,7 +3026,7 @@ uint8_t *j2k_tile::encode() {
 #elif defined(OPENHTJ2K_ENABLE_ARM_NEON)
     for (uint32_t y = 0; y < height; ++y) {
       int32_t *sp             = src + y * stride;
-      sprec_t *dp             = cr->i_samples + y * (bottom_right.x - top_left.x);
+      sprec_t *dp             = cr->i_samples + y * stride;
       uint32_t num_tc_samples = bottom_right.x - top_left.x;
       for (; num_tc_samples >= 8; num_tc_samples -= 8) {
         auto vsrc0 = vld1q_s32(sp);
@@ -3043,7 +3042,7 @@ uint8_t *j2k_tile::encode() {
 #else
       for (uint32_t y = 0; y < height; ++y) {
         int32_t *sp             = src + y * stride;
-        sprec_t *dp             = cr->i_samples + y * (bottom_right.x - top_left.x);
+        sprec_t *dp             = cr->i_samples + y * round_up(bottom_right.x - top_left.x, 32U);
         uint32_t num_tc_samples = bottom_right.x - top_left.x;
         for (; num_tc_samples > 0; --num_tc_samples) {
           *dp++ = static_cast<sprec_t>(*sp++);

source/core/coding/coding_units.hpp

@@ -41,6 +41,9 @@ class j2k_region {
   element_siz pos0;
   // bottom-right coordinate (exclusive) of a region in the reference grid
   element_siz pos1;
+  // width for line buffer
+  uint32_t stride;
+
   // return top-left coordinate (inclusive)
   [[nodiscard]] element_siz get_pos0() const { return pos0; }
   // return bottom-right coordinate (exclusive)
@@ -55,7 +58,7 @@ class j2k_region {
   // set bottom-right coordinate (exclusive)
   void set_pos1(element_siz in) { pos1 = in; }
   j2k_region() = default;
-  j2k_region(element_siz p0, element_siz p1) : pos0(p0), pos1(p1) {}
+  j2k_region(element_siz p0, element_siz p1) : pos0(p0), pos1(p1), stride(round_up(pos1.x - pos0.x, 32U)) {}
 };
 
 /********************************************************************************
@@ -176,8 +179,8 @@ class j2k_precinct_subband : public j2k_region {
   uint32_t num_codeblock_x;
   uint32_t num_codeblock_y;
   j2k_precinct_subband(uint8_t orientation, uint8_t M_b, uint8_t R_b, uint8_t transformation,
-                       float stepsize, sprec_t *ibuf, const element_siz &bp0, const element_siz &bp1,
-                       const element_siz &p0, const element_siz &p1, const uint16_t &num_layers,
+                       float stepsize, sprec_t *ibuf, const element_siz &bp0, const element_siz &p0,
+                       const element_siz &p1, const uint32_t stride, const uint16_t &num_layers,
                        const element_siz &codeblock_size, const uint8_t &Cmodes);
   ~j2k_precinct_subband() {
     delete inclusion_info;
@@ -312,7 +315,7 @@ class j2k_resolution : public j2k_region {
   void scale();
   void destroy() {
     aligned_mem_free(i_samples);
-    for (auto b = 0; b < num_bands; ++b) {
+    for (uint8_t b = 0; b < num_bands; ++b) {
       if (subbands != nullptr) {
         subbands[b]->destroy();
       }
@@ -420,7 +423,7 @@ class j2k_tile_component : public j2k_tile_base {
   void perform_dc_offset(uint8_t transformation, bool is_signed);
 
   void destroy() {
-    for (auto r = 0; r < this->NL; ++r) {
+    for (uint8_t r = 0; r < this->NL; ++r) {
       if (resolution != nullptr) {
         auto p = resolution[r].get();
         if (p != nullptr) resolution[r]->destroy();
@@ -491,7 +494,7 @@ class j2k_tile : public j2k_tile_base {
  public:
   j2k_tile();
   void destroy() {
-    for (auto c = 0; c < this->num_components; ++c) {
+    for (uint16_t c = 0; c < this->num_components; ++c) {
       tcomp[c].destroy();
     }
   }