init files

loader/tools/lzsa/src/dictionary.c

@@ -0,0 +1,101 @@
+/*
+ * dictionary.c - dictionary implementation
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include "format.h"
+#include "lib.h"
+
+/**
+ * Load dictionary contents
+ *
+ * @param pszDictionaryFilename name of dictionary file, or NULL for none
+ * @param ppDictionaryData pointer to returned dictionary contents, or NULL for none
+ * @param pDictionaryDataSize pointer to returned size of dictionary contents, or 0
+ *
+ * @return LZSA_OK for success, or an error value from lzsa_status_t
+ */
+int lzsa_dictionary_load(const char *pszDictionaryFilename, void **ppDictionaryData, int *pDictionaryDataSize) {
+   unsigned char *pDictionaryData = NULL;
+   int nDictionaryDataSize = 0;
+
+   if (pszDictionaryFilename) {
+      pDictionaryData = (unsigned char *)malloc(BLOCK_SIZE);
+      if (!pDictionaryData) {
+         return LZSA_ERROR_MEMORY;
+      }
+
+      FILE *pDictionaryFile = fopen(pszDictionaryFilename, "rb");
+      if (!pDictionaryFile) {
+         free(pDictionaryData);
+         pDictionaryData = NULL;
+         return LZSA_ERROR_DICTIONARY;
+      }
+
+      fseek(pDictionaryFile, 0, SEEK_END);
+#ifdef _WIN32
+      __int64 nDictionaryFileSize = _ftelli64(pDictionaryFile);
+#else
+      off_t nDictionaryFileSize = ftello(pDictionaryFile);
+#endif
+      if (nDictionaryFileSize > BLOCK_SIZE) {
+         /* Use the last BLOCK_SIZE bytes of the dictionary */
+         fseek(pDictionaryFile, -BLOCK_SIZE, SEEK_END);
+      }
+      else {
+         fseek(pDictionaryFile, 0, SEEK_SET);
+      }
+
+      nDictionaryDataSize = (int)fread(pDictionaryData, 1, BLOCK_SIZE, pDictionaryFile);
+      if (nDictionaryDataSize < 0)
+         nDictionaryDataSize = 0;
+
+      fclose(pDictionaryFile);
+      pDictionaryFile = NULL;
+   }
+
+   *ppDictionaryData = pDictionaryData;
+   *pDictionaryDataSize = nDictionaryDataSize;
+   return LZSA_OK;
+}
+
+/**
+ * Free dictionary contents
+ *
+ * @param ppDictionaryData pointer to pointer to dictionary contents
+ */
+void lzsa_dictionary_free(void **ppDictionaryData) {
+   if (*ppDictionaryData) {
+      free(*ppDictionaryData);
+      *ppDictionaryData = NULL;
+   }
+}

loader/tools/lzsa/src/dictionary.h

@@ -0,0 +1,64 @@
+/*
+ * dictionary.h - dictionary definitions
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#ifndef _DICTIONARY_H
+#define _DICTIONARY_H
+
+#include <stdlib.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * Load dictionary contents
+ *
+ * @param pszDictionaryFilename name of dictionary file, or NULL for none
+ * @param ppDictionaryData pointer to returned dictionary contents, or NULL for none
+ * @param pDictionaryDataSize pointer to returned size of dictionary contents, or 0
+ *
+ * @return LZSA_OK for success, or an error value from lzsa_status_t
+ */
+int lzsa_dictionary_load(const char *pszDictionaryFilename, void **ppDictionaryData, int *pDictionaryDataSize);
+
+/**
+ * Free dictionary contents
+ *
+ * @param ppDictionaryData pointer to pointer to dictionary contents
+ */
+void lzsa_dictionary_free(void **ppDictionaryData);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _DICTIONARY_H */

loader/tools/lzsa/src/expand_block_v1.c

@@ -0,0 +1,224 @@
+/*
+ * expand_block_v1.c - LZSA1 block decompressor implementation
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#include <stdlib.h>
+#include <string.h>
+#include "format.h"
+#include "expand_block_v1.h"
+
+#ifdef _MSC_VER
+#define FORCE_INLINE __forceinline
+#else /* _MSC_VER */
+#define FORCE_INLINE __attribute__((always_inline))
+#endif /* _MSC_VER */
+
+static inline FORCE_INLINE int lzsa_build_literals_len_v1(const unsigned char **ppInBlock, const unsigned char *pInBlockEnd, unsigned int *nLiterals) {
+   unsigned int nByte;
+   const unsigned char *pInBlock = *ppInBlock;
+
+   if (pInBlock < pInBlockEnd) {
+      nByte = *pInBlock++;
+      (*nLiterals) += nByte;
+
+      if (nByte == 250) {
+         if (pInBlock < pInBlockEnd) {
+            (*nLiterals) = 256 + ((unsigned int)*pInBlock++);
+         }
+         else {
+            return -1;
+         }
+      }
+      else if (nByte == 249) {
+         if ((pInBlock + 1) < pInBlockEnd) {
+            (*nLiterals) = ((unsigned int)*pInBlock++);
+            (*nLiterals) |= (((unsigned int)*pInBlock++) << 8);
+         }
+         else {
+            return -1;
+         }
+      }
+
+      *ppInBlock = pInBlock;
+      return 0;
+   }
+   else {
+      return -1;
+   }
+}
+
+static inline FORCE_INLINE int lzsa_build_match_len_v1(const unsigned char **ppInBlock, const unsigned char *pInBlockEnd, unsigned int *nMatchLen) {
+   unsigned int nByte;
+   const unsigned char *pInBlock = *ppInBlock;
+
+   if (pInBlock < pInBlockEnd) {
+      nByte = *pInBlock++;
+      (*nMatchLen) += nByte;
+
+      if (nByte == 239) {
+         if (pInBlock < pInBlockEnd) {
+            (*nMatchLen) = 256 + ((unsigned int)*pInBlock++);
+         }
+         else {
+            return -1;
+         }
+      }
+      else if (nByte == 238) {
+         if ((pInBlock + 1) < pInBlockEnd) {
+            (*nMatchLen) = ((unsigned int)*pInBlock++);
+            (*nMatchLen) |= (((unsigned int)*pInBlock++) << 8);
+         }
+         else {
+            return -1;
+         }
+      }
+
+      *ppInBlock = pInBlock;
+      return 0;
+   }
+   else {
+      return -1;
+   }
+}
+
+/**
+ * Decompress one LZSA1 data block
+ *
+ * @param pInBlock pointer to compressed data
+ * @param nBlockSize size of compressed data, in bytes
+ * @param pOutData pointer to output decompression buffer (previously decompressed bytes + room for decompressing this block)
+ * @param nOutDataOffset starting index of where to store decompressed bytes in output buffer (and size of previously decompressed bytes)
+ * @param nBlockMaxSize total size of output decompression buffer, in bytes
+ *
+ * @return size of decompressed data in bytes, or -1 for error
+ */
+int lzsa_decompressor_expand_block_v1(const unsigned char *pInBlock, int nBlockSize, unsigned char *pOutData, int nOutDataOffset, int nBlockMaxSize) {
+   const unsigned char *pInBlockEnd = pInBlock + nBlockSize;
+   unsigned char *pCurOutData = pOutData + nOutDataOffset;
+   const unsigned char *pOutDataEnd = pCurOutData + nBlockMaxSize;
+   const unsigned char *pOutDataFastEnd = pOutDataEnd - 18;
+
+   while (pInBlock < pInBlockEnd) {
+      const unsigned char token = *pInBlock++;
+      unsigned int nLiterals = (unsigned int)((token & 0x70) >> 4);
+
+      if (nLiterals != LITERALS_RUN_LEN_V1 && (pInBlock + 8) <= pInBlockEnd && pCurOutData < pOutDataFastEnd) {
+         memcpy(pCurOutData, pInBlock, 8);
+         pInBlock += nLiterals;
+         pCurOutData += nLiterals;
+      }
+      else {
+         if (nLiterals == LITERALS_RUN_LEN_V1) {
+            if (lzsa_build_literals_len_v1(&pInBlock, pInBlockEnd, &nLiterals))
+               return -1;
+         }
+
+         if (nLiterals != 0) {
+            if ((pInBlock + nLiterals) <= pInBlockEnd &&
+               (pCurOutData + nLiterals) <= pOutDataEnd) {
+               memcpy(pCurOutData, pInBlock, nLiterals);
+               pInBlock += nLiterals;
+               pCurOutData += nLiterals;
+            }
+            else {
+               return -1;
+            }
+         }
+      }
+
+      if ((pInBlock + 1) < pInBlockEnd) { /* The last token in the block does not include match information */
+         unsigned int nMatchOffset;
+
+         nMatchOffset = ((unsigned int)(*pInBlock++)) ^ 0xff;
+         if (token & 0x80) {
+            nMatchOffset |= (((unsigned int)(*pInBlock++)) << 8) ^ 0xff00;
+         }
+         nMatchOffset++;
+
+         const unsigned char *pSrc = pCurOutData - nMatchOffset;
+         if (pSrc >= pOutData) {
+            unsigned int nMatchLen = (unsigned int)(token & 0x0f);
+            if (nMatchLen != MATCH_RUN_LEN_V1 && nMatchOffset >= 8 && pCurOutData < pOutDataFastEnd && (pSrc + 18) <= pOutDataEnd) {
+               memcpy(pCurOutData, pSrc, 8);
+               memcpy(pCurOutData + 8, pSrc + 8, 8);
+               memcpy(pCurOutData + 16, pSrc + 16, 2);
+               pCurOutData += (MIN_MATCH_SIZE_V1 + nMatchLen);
+            }
+            else {
+               nMatchLen += MIN_MATCH_SIZE_V1;
+               if (nMatchLen == (MATCH_RUN_LEN_V1 + MIN_MATCH_SIZE_V1)) {
+                  if (lzsa_build_match_len_v1(&pInBlock, pInBlockEnd, &nMatchLen))
+                     return -1;
+                  if (nMatchLen == 0)
+                     break;
+               }
+
+               if ((pSrc + nMatchLen) <= pOutDataEnd) {
+                  if ((pCurOutData + nMatchLen) <= pOutDataEnd) {
+                     /* Do a deterministic, left to right byte copy instead of memcpy() so as to handle overlaps */
+
+                     if (nMatchOffset >= 16 && (pCurOutData + nMatchLen) < (pOutDataFastEnd - 15)) {
+                        const unsigned char *pCopySrc = pSrc;
+                        unsigned char *pCopyDst = pCurOutData;
+                        const unsigned char *pCopyEndDst = pCurOutData + nMatchLen;
+
+                        do {
+                           memcpy(pCopyDst, pCopySrc, 16);
+                           pCopySrc += 16;
+                           pCopyDst += 16;
+                        } while (pCopyDst < pCopyEndDst);
+
+                        pCurOutData += nMatchLen;
+                     }
+                     else {
+                        while (nMatchLen) {
+                           *pCurOutData++ = *pSrc++;
+                           nMatchLen--;
+                        }
+                     }
+                  }
+                  else {
+                     return -1;
+                  }
+               }
+               else {
+                  return -1;
+               }
+            }
+         }
+         else {
+            return -1;
+         }
+      }
+   }
+
+   return (int)(pCurOutData - (pOutData + nOutDataOffset));
+}

loader/tools/lzsa/src/expand_block_v1.h

@@ -0,0 +1,49 @@
+/*
+ * expand_block_v1.h - LZSA1 block decompressor definitions
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#ifndef _EXPAND_BLOCK_V1_H
+#define _EXPAND_BLOCK_V1_H
+
+/**
+ * Decompress one LZSA1 data block
+ *
+ * @param pInBlock pointer to compressed data
+ * @param nBlockSize size of compressed data, in bytes
+ * @param pOutData pointer to output decompression buffer (previously decompressed bytes + room for decompressing this block)
+ * @param nOutDataOffset starting index of where to store decompressed bytes in output buffer (and size of previously decompressed bytes)
+ * @param nBlockMaxSize total size of output decompression buffer, in bytes
+ *
+ * @return size of decompressed data in bytes, or -1 for error
+ */
+int lzsa_decompressor_expand_block_v1(const unsigned char *pInBlock, int nBlockSize, unsigned char *pOutData, int nOutDataOffset, int nBlockMaxSize);
+
+#endif /* _EXPAND_BLOCK_V1_H */

loader/tools/lzsa/src/expand_block_v2.c

@@ -0,0 +1,249 @@
+/*
+ * expand_block_v2.c - LZSA2 block decompressor implementation
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#include <stdlib.h>
+#include <string.h>
+#include "format.h"
+#include "expand_block_v2.h"
+
+#ifdef _MSC_VER
+#define FORCE_INLINE __forceinline
+#else /* _MSC_VER */
+#define FORCE_INLINE __attribute__((always_inline))
+#endif /* _MSC_VER */
+
+static inline FORCE_INLINE unsigned int lzsa_get_nibble_v2(const unsigned char **ppInBlock, const unsigned char *pInBlockEnd, int *nCurNibbles, unsigned char *nibbles, unsigned int *nValue) {
+   if ((*nCurNibbles ^= 1) != 0) {
+      const unsigned char *pInBlock = *ppInBlock;
+      if (pInBlock < pInBlockEnd) {
+         (*nibbles) = *pInBlock++;
+         *ppInBlock = pInBlock;
+         (*nValue) = ((unsigned int)((*nibbles) & 0xf0)) >> 4;
+         return 0;
+      }
+      else {
+         return -1;
+      }
+   }
+
+   (*nValue) = (unsigned int)((*nibbles) & 0x0f);
+   return 0;
+}
+
+static inline FORCE_INLINE int lzsa_build_len_v2(const unsigned char **ppInBlock, const unsigned char *pInBlockEnd, int *nCurNibbles, unsigned char *nibbles, unsigned int *nLength) {
+   unsigned int nValue;
+
+   if (!lzsa_get_nibble_v2(ppInBlock, pInBlockEnd, nCurNibbles, nibbles, &nValue)) {
+      (*nLength) += nValue;
+
+      if (nValue == 15) {
+         const unsigned char *pInBlock = *ppInBlock;
+
+         if (pInBlock < pInBlockEnd) {
+            (*nLength) += ((unsigned int)*pInBlock++);
+
+            if ((*nLength) == 257) {
+               if ((pInBlock + 1) < pInBlockEnd) {
+                  (*nLength) = ((unsigned int)*pInBlock++) << 8;
+                  (*nLength) |= ((unsigned int)*pInBlock++);
+               }
+               else {
+                  return -1;
+               }
+            }
+            else if ((*nLength) == 256) {
+               (*nLength) = 0;
+            }
+         }
+         else {
+            return -1;
+         }
+
+         *ppInBlock = pInBlock;
+      }
+
+      return 0;
+   }
+   else {
+      return -1;
+   }
+}
+
+/**
+ * Decompress one LZSA2 data block
+ *
+ * @param pInBlock pointer to compressed data
+ * @param nBlockSize size of compressed data, in bytes
+ * @param pOutData pointer to output decompression buffer (previously decompressed bytes + room for decompressing this block)
+ * @param nOutDataOffset starting index of where to store decompressed bytes in output buffer (and size of previously decompressed bytes)
+ * @param nBlockMaxSize total size of output decompression buffer, in bytes
+ *
+ * @return size of decompressed data in bytes, or -1 for error
+ */
+int lzsa_decompressor_expand_block_v2(const unsigned char *pInBlock, int nBlockSize, unsigned char *pOutData, int nOutDataOffset, int nBlockMaxSize) {
+   const unsigned char *pInBlockEnd = pInBlock + nBlockSize;
+   unsigned char *pCurOutData = pOutData + nOutDataOffset;
+   const unsigned char *pOutDataEnd = pCurOutData + nBlockMaxSize;
+   const unsigned char *pOutDataFastEnd = pOutDataEnd - 20;
+   int nCurNibbles = 0;
+   unsigned char nibbles;
+   int nMatchOffset = 0;
+
+   while (pInBlock < pInBlockEnd) {
+      const unsigned char token = *pInBlock++;
+      unsigned int nLiterals = (unsigned int)((token >> LITERALS_LEN_POS_V2) & LITERALS_RUN_LEN_V2);
+
+      if (nLiterals != LITERALS_RUN_LEN_V2 && (pInBlock + 4) <= pInBlockEnd && pCurOutData < pOutDataFastEnd) {
+         memcpy(pCurOutData, pInBlock, 4);
+         pInBlock += nLiterals;
+         pCurOutData += nLiterals;
+      }
+      else {
+         if (nLiterals == LITERALS_RUN_LEN_V2) {
+            if (lzsa_build_len_v2(&pInBlock, pInBlockEnd, &nCurNibbles, &nibbles, &nLiterals))
+               return -1;
+         }
+
+         if (nLiterals != 0) {
+            if ((pInBlock + nLiterals) <= pInBlockEnd &&
+               (pCurOutData + nLiterals) <= pOutDataEnd) {
+               memcpy(pCurOutData, pInBlock, nLiterals);
+               pInBlock += nLiterals;
+               pCurOutData += nLiterals;
+            }
+            else {
+               return -1;
+            }
+         }
+      }
+
+      if (pInBlock < pInBlockEnd) { /* The last token in the block does not include match information */
+         unsigned char nOffsetMode = token & 0xc0;
+         unsigned int nValue;
+
+         if ((nOffsetMode & 0x80) == 0x00) {
+               if ((nOffsetMode & 0x20) == 0x00) {
+               /* 5 bit offset */
+               if (lzsa_get_nibble_v2(&pInBlock, pInBlockEnd, &nCurNibbles, &nibbles, &nValue))
+                  return -1;
+               nMatchOffset = nValue << 1;
+               nMatchOffset |= ((token & 0x40) >> 6);
+               nMatchOffset ^= 0x1e;
+               nMatchOffset++;
+            }
+            else {
+               /* 13 bit offset */
+               if (lzsa_get_nibble_v2(&pInBlock, pInBlockEnd, &nCurNibbles, &nibbles, &nValue))
+                  return -1;
+               nMatchOffset = (unsigned int)(*pInBlock++);
+               nMatchOffset |= (nValue << 9);
+               nMatchOffset |= (((unsigned int)(token & 0x40)) << 2);
+               nMatchOffset ^= 0x1eff;
+               nMatchOffset += (512 + 1);
+            }
+         }
+         else {
+            if ((nOffsetMode & 0xc0) == 0x80) {
+               /* 9 bit offset */
+               nMatchOffset = (unsigned int)(*pInBlock++);
+               nMatchOffset |= (((unsigned int)(token & 0x20)) << 3);
+               nMatchOffset ^= 0x0ff;
+               nMatchOffset++;
+            }
+            if ((nOffsetMode & 0xe0) == 0xc0) {
+               /* 16 bit offset */
+               nMatchOffset = (((unsigned int)(*pInBlock++)) << 8);
+               if (pInBlock >= pInBlockEnd) return -1;
+               nMatchOffset |= (unsigned int)(*pInBlock++);
+               nMatchOffset ^= 0xffff;
+               nMatchOffset++;
+            }
+         }
+
+         const unsigned char *pSrc = pCurOutData - nMatchOffset;
+         if (pSrc >= pOutData) {
+            unsigned int nMatchLen = (unsigned int)((token >> MATCH_LEN_POS_V2) & MATCH_RUN_LEN_V2);
+            if (nMatchLen != MATCH_RUN_LEN_V2 && nMatchOffset >= 8 && pCurOutData < pOutDataFastEnd && (pSrc + 10) <= pOutDataEnd) {
+               memcpy(pCurOutData, pSrc, 8);
+               memcpy(pCurOutData + 8, pSrc + 8, 2);
+               pCurOutData += (MIN_MATCH_SIZE_V2 + nMatchLen);
+            }
+            else {
+               nMatchLen += MIN_MATCH_SIZE_V2;
+               if (nMatchLen == (MATCH_RUN_LEN_V2 + MIN_MATCH_SIZE_V2)) {
+                  if (lzsa_build_len_v2(&pInBlock, pInBlockEnd, &nCurNibbles, &nibbles, &nMatchLen))
+                     return -1;
+                  if (nMatchLen == 0)
+                     break;
+               }
+
+               if ((pSrc + nMatchLen) <= pOutDataEnd) {
+                  if ((pCurOutData + nMatchLen) <= pOutDataEnd) {
+                     /* Do a deterministic, left to right byte copy instead of memcpy() so as to handle overlaps */
+
+                     if (nMatchOffset >= 16 && (pCurOutData + nMatchLen) < (pOutDataFastEnd - 15)) {
+                        const unsigned char *pCopySrc = pSrc;
+                        unsigned char *pCopyDst = pCurOutData;
+                        const unsigned char *pCopyEndDst = pCurOutData + nMatchLen;
+
+                        do {
+                           memcpy(pCopyDst, pCopySrc, 16);
+                           pCopySrc += 16;
+                           pCopyDst += 16;
+                        } while (pCopyDst < pCopyEndDst);
+
+                        pCurOutData += nMatchLen;
+                     }
+                     else {
+                        while (nMatchLen) {
+                           *pCurOutData++ = *pSrc++;
+                           nMatchLen--;
+                        }
+                     }
+                  }
+                  else {
+                     return -1;
+                  }
+               }
+               else {
+                  return -1;
+               }
+            }
+         }
+         else {
+            return -1;
+         }
+      }
+   }
+
+   return (int)(pCurOutData - (pOutData + nOutDataOffset));
+}

loader/tools/lzsa/src/expand_block_v2.h

@@ -0,0 +1,49 @@
+/*
+ * expand_block_v2.h - LZSA2 block decompressor definitions
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#ifndef _EXPAND_BLOCK_V2_H
+#define _EXPAND_BLOCK_V2_H
+
+/**
+ * Decompress one LZSA2 data block
+ *
+ * @param pInBlock pointer to compressed data
+ * @param nBlockSize size of compressed data, in bytes
+ * @param pOutData pointer to output decompression buffer (previously decompressed bytes + room for decompressing this block)
+ * @param nOutDataOffset starting index of where to store decompressed bytes in output buffer (and size of previously decompressed bytes)
+ * @param nBlockMaxSize total size of output decompression buffer, in bytes
+ *
+ * @return size of decompressed data in bytes, or -1 for error
+ */
+int lzsa_decompressor_expand_block_v2(const unsigned char *pInBlock, int nBlockSize, unsigned char *pOutData, int nOutDataOffset, int nBlockMaxSize);
+
+#endif /* _EXPAND_BLOCK_V2_H */

loader/tools/lzsa/src/expand_context.c

@@ -0,0 +1,76 @@
+/*
+ * expand_context.h - decompressor context definitions
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#include <stdlib.h>
+#include <string.h>
+#include "expand_context.h"
+#include "expand_block_v1.h"
+#include "expand_block_v2.h"
+#include "lib.h"
+
+/**
+ * Decompress one data block
+ *
+ * @param pInBlock pointer to compressed data
+ * @param nBlockSize size of compressed data, in bytes
+ * @param pOutData pointer to output decompression buffer (previously decompressed bytes + room for decompressing this block)
+ * @param nOutDataOffset starting index of where to store decompressed bytes in output buffer (and size of previously decompressed bytes)
+ * @param nBlockMaxSize total size of output decompression buffer, in bytes
+ * @param nFormatVersion version of format to use (1-2)
+ * @param nFlags compression flags (LZSA_FLAG_xxx)
+ *
+ * @return size of decompressed data in bytes, or -1 for error
+ */
+int lzsa_decompressor_expand_block(unsigned char *pInBlock, int nBlockSize, unsigned char *pOutData, int nOutDataOffset, int nBlockMaxSize, const int nFormatVersion, const int nFlags) {
+   int nDecompressedSize;
+
+   if (nFlags & LZSA_FLAG_RAW_BACKWARD) {
+      lzsa_reverse_buffer(pInBlock, nBlockSize);
+   }
+
+   if (nFormatVersion == 1)
+      nDecompressedSize = lzsa_decompressor_expand_block_v1(pInBlock, nBlockSize, pOutData, nOutDataOffset, nBlockMaxSize);
+   else if (nFormatVersion == 2)
+      nDecompressedSize = lzsa_decompressor_expand_block_v2(pInBlock, nBlockSize, pOutData, nOutDataOffset, nBlockMaxSize);
+   else
+      nDecompressedSize = -1;
+
+   if (nDecompressedSize != -1 && (nFlags & LZSA_FLAG_RAW_BACKWARD)) {
+      lzsa_reverse_buffer(pOutData + nOutDataOffset, nDecompressedSize);
+   }
+
+   if (nFlags & LZSA_FLAG_RAW_BACKWARD) {
+      lzsa_reverse_buffer(pInBlock, nBlockSize);
+   }
+
+   return nDecompressedSize;
+}

loader/tools/lzsa/src/expand_context.h

@@ -0,0 +1,61 @@
+/*
+ * expand_context.h - decompressor context definitions
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#ifndef _EXPAND_CONTEXT_H
+#define _EXPAND_CONTEXT_H
+
+#include <stdlib.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * Decompress one data block
+ *
+ * @param pInBlock pointer to compressed data
+ * @param nBlockSize size of compressed data, in bytes
+ * @param pOutData pointer to output decompression buffer (previously decompressed bytes + room for decompressing this block)
+ * @param nOutDataOffset starting index of where to store decompressed bytes in output buffer (and size of previously decompressed bytes)
+ * @param nBlockMaxSize total size of output decompression buffer, in bytes
+ * @param nFormatVersion version of format to use (1-2)
+ * @param nFlags compression flags (LZSA_FLAG_xxx)
+ *
+ * @return size of decompressed data in bytes, or -1 for error
+ */
+int lzsa_decompressor_expand_block(unsigned char *pInBlock, int nBlockSize, unsigned char *pOutData, int nOutDataOffset, int nBlockMaxSize, const int nFormatVersion, const int nFlags);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _EXPAND_CONTEXT_H */

loader/tools/lzsa/src/expand_inmem.c

@@ -0,0 +1,163 @@
+/*
+ * expand_inmem.c - in-memory decompression implementation
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#include <stdlib.h>
+#include <string.h>
+#include "expand_inmem.h"
+#include "lib.h"
+#include "frame.h"
+
+#define BLOCK_SIZE 65536
+
+/**
+ * Get maximum decompressed size of compressed data
+ *
+ * @param pFileData compressed data
+ * @param nFileSize compressed size in bytes
+ *
+ * @return maximum decompressed size
+ */
+size_t lzsa_get_max_decompressed_size_inmem(const unsigned char *pFileData, size_t nFileSize) {
+   const unsigned char *pCurFileData = pFileData;
+   const unsigned char *pEndFileData = pCurFileData + nFileSize;
+   int nFormatVersion = 0;
+   size_t nMaxDecompressedSize = 0;
+   const int nHeaderSize = lzsa_get_header_size();
+
+   /* Check header */
+   if ((pCurFileData + nHeaderSize) > pEndFileData ||
+       lzsa_decode_header(pCurFileData, nHeaderSize, &nFormatVersion) != 0)
+      return -1;
+
+   pCurFileData += nHeaderSize;
+
+   while (pCurFileData < pEndFileData) {
+      unsigned int nBlockDataSize = 0;
+      int nIsUncompressed = 0;
+      const int nFrameSize = lzsa_get_frame_size();
+
+      /* Decode frame header */
+      if ((pCurFileData + nFrameSize) > pEndFileData ||
+          lzsa_decode_frame(pCurFileData, nFrameSize, &nBlockDataSize, &nIsUncompressed) != 0)
+         return -1;
+      pCurFileData += nFrameSize;
+
+      if (!nBlockDataSize)
+         break;
+
+      /* Add one potentially full block to the decompressed size */
+      nMaxDecompressedSize += BLOCK_SIZE;
+
+      if ((pCurFileData + nBlockDataSize) > pEndFileData)
+         return -1;
+
+      pCurFileData += nBlockDataSize;
+   }
+
+   return nMaxDecompressedSize;
+}
+
+/**
+ * Decompress data in memory
+ *
+ * @param pFileData compressed data
+ * @param pOutBuffer buffer for decompressed data
+ * @param nFileSize compressed size in bytes
+ * @param nMaxOutBufferSize maximum capacity of decompression buffer
+ * @param nFlags compression flags (LZSA_FLAG_xxx)
+ * @param pFormatVersion pointer to format version, updated if this function is successful
+ *
+ * @return actual decompressed size, or -1 for error
+ */
+size_t lzsa_decompress_inmem(unsigned char *pFileData, unsigned char *pOutBuffer, size_t nFileSize, size_t nMaxOutBufferSize, const unsigned int nFlags, int *pFormatVersion) {
+   unsigned char *pCurFileData = pFileData;
+   const unsigned char *pEndFileData = pCurFileData + nFileSize;
+   unsigned char *pCurOutBuffer = pOutBuffer;
+   const unsigned char *pEndOutBuffer = pCurOutBuffer + nMaxOutBufferSize;
+   int nPreviousBlockSize;
+   const int nHeaderSize = lzsa_get_header_size();
+
+   if (nFlags & LZSA_FLAG_RAW_BLOCK) {
+      return (size_t)lzsa_decompressor_expand_block(pFileData, (int)nFileSize, pOutBuffer, 0, (int)nMaxOutBufferSize, *pFormatVersion, nFlags);
+   }
+
+   /* Check header */
+   if ((pCurFileData + nHeaderSize) > pEndFileData ||
+      lzsa_decode_header(pCurFileData, nHeaderSize, pFormatVersion) != 0)
+      return -1;
+
+   pCurFileData += nHeaderSize;
+   nPreviousBlockSize = 0;
+
+   while (pCurFileData < pEndFileData) {
+      unsigned int nBlockDataSize = 0;
+      int nIsUncompressed = 0;
+      const int nFrameSize = lzsa_get_frame_size();
+
+      /* Decode frame header */
+      if ((pCurFileData + nFrameSize) > pEndFileData ||
+          lzsa_decode_frame(pCurFileData, nFrameSize, &nBlockDataSize, &nIsUncompressed) != 0)
+         return -1;
+      pCurFileData += nFrameSize;
+
+      if (!nBlockDataSize)
+         break;
+
+      if (!nIsUncompressed) {
+         int nDecompressedSize;
+
+         /* Decompress block */
+         if ((pCurFileData + nBlockDataSize) > pEndFileData)
+            return -1;
+
+         nDecompressedSize = lzsa_decompressor_expand_block(pCurFileData, nBlockDataSize, pCurOutBuffer - nPreviousBlockSize, nPreviousBlockSize, (int)(pEndOutBuffer - pCurOutBuffer + nPreviousBlockSize), *pFormatVersion, nFlags);
+         if (nDecompressedSize < 0)
+            return -1;
+
+         pCurOutBuffer += nDecompressedSize;
+         nPreviousBlockSize = nDecompressedSize;
+      }
+      else {
+         /* Copy uncompressed block */
+         if ((pCurFileData + nBlockDataSize) > pEndFileData)
+            return -1;
+         if ((pCurOutBuffer + nBlockDataSize) > pEndOutBuffer)
+            return -1;
+         memcpy(pCurOutBuffer, pCurFileData, nBlockDataSize);
+         pCurOutBuffer += nBlockDataSize;
+      }
+
+      pCurFileData += nBlockDataSize;
+   }
+
+   return (int)(pCurOutBuffer - pOutBuffer);
+}

loader/tools/lzsa/src/expand_inmem.h

@@ -0,0 +1,70 @@
+/*
+ * expand_inmem.h - in-memory decompression definitions
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#ifndef _EXPAND_INMEM_H
+#define _EXPAND_INMEM_H
+
+#include <stdlib.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * Get maximum decompressed size of compressed data
+ *
+ * @param pFileData compressed data
+ * @param nFileSize compressed size in bytes
+ *
+ * @return maximum decompressed size
+ */
+size_t lzsa_get_max_decompressed_size_inmem(const unsigned char *pFileData, size_t nFileSize);
+
+/**
+ * Decompress data in memory
+ *
+ * @param pFileData compressed data
+ * @param pOutBuffer buffer for decompressed data
+ * @param nFileSize compressed size in bytes
+ * @param nMaxOutBufferSize maximum capacity of decompression buffer
+ * @param nFlags compression flags (LZSA_FLAG_xxx)
+ * @param pFormatVersion pointer to format version, updated if this function is successful
+ *
+ * @return actual decompressed size, or -1 for error
+ */
+size_t lzsa_decompress_inmem(unsigned char *pFileData, unsigned char *pOutBuffer, size_t nFileSize, size_t nMaxOutBufferSize, const unsigned int nFlags, int *pFormatVersion);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _EXPAND_INMEM_H */

loader/tools/lzsa/src/expand_streaming.c

@@ -0,0 +1,236 @@
+/*
+ * expand_streaming.c - streaming decompression definitions
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+
+#include <stdlib.h>
+#include <string.h>
+#include "expand_streaming.h"
+#include "format.h"
+#include "frame.h"
+#include "lib.h"
+
+/*-------------- File API -------------- */
+
+/**
+ * Decompress file
+ *
+ * @param pszInFilename name of input(compressed) file to decompress
+ * @param pszOutFilename name of output(decompressed) file to generate
+ * @param pszDictionaryFilename name of dictionary file, or NULL for none
+ * @param nFlags compression flags (LZSA_FLAG_RAW_BLOCK to decompress a raw block, or 0)
+ * @param nFormatVersion default version of format to use (1-2). This is used when decompressing a raw block, otherwise the version is extracted from the source file
+ * @param pOriginalSize pointer to returned output(decompressed) size, updated when this function is successful
+ * @param pCompressedSize pointer to returned input(compressed) size, updated when this function is successful
+ *
+ * @return LZSA_OK for success, or an error value from lzsa_status_t
+ */
+lzsa_status_t lzsa_decompress_file(const char *pszInFilename, const char *pszOutFilename, const char *pszDictionaryFilename, const unsigned int nFlags, int nFormatVersion,
+                                   long long *pOriginalSize, long long *pCompressedSize) {
+   lzsa_stream_t inStream, outStream;
+   void *pDictionaryData = NULL;
+   int nDictionaryDataSize = 0;
+   lzsa_status_t nStatus;
+
+   if (lzsa_filestream_open(&inStream, pszInFilename, "rb") < 0) {
+      return LZSA_ERROR_SRC;
+   }
+
+   if (lzsa_filestream_open(&outStream, pszOutFilename, "wb") < 0) {
+      inStream.close(&inStream);
+      return LZSA_ERROR_DST;
+   }
+
+   nStatus = lzsa_dictionary_load(pszDictionaryFilename, &pDictionaryData, &nDictionaryDataSize);
+   if (nStatus) {
+      outStream.close(&outStream);
+      inStream.close(&inStream);
+      return nStatus;
+   }
+
+   nStatus = lzsa_decompress_stream(&inStream, &outStream, pDictionaryData, nDictionaryDataSize, nFlags, nFormatVersion, pOriginalSize, pCompressedSize);
+
+   lzsa_dictionary_free(&pDictionaryData);
+   outStream.close(&outStream);
+   inStream.close(&inStream);
+
+   return nStatus;
+}
+
+/*-------------- Streaming API -------------- */
+
+/**
+ * Decompress stream
+ *
+ * @param pInStream input(compressed) stream to decompress
+ * @param pOutStream output(decompressed) stream to write to
+ * @param pDictionaryData dictionary contents, or NULL for none
+ * @param nDictionaryDataSize size of dictionary contents, or 0
+ * @param nFlags compression flags (LZSA_FLAG_RAW_BLOCK to decompress a raw block, or 0)
+ * @param nFormatVersion default version of format to use (1-2). This is used when decompressing a raw block, otherwise the version is extracted from the source file
+ * @param pOriginalSize pointer to returned output(decompressed) size, updated when this function is successful
+ * @param pCompressedSize pointer to returned input(compressed) size, updated when this function is successful
+ *
+ * @return LZSA_OK for success, or an error value from lzsa_status_t
+ */
+lzsa_status_t lzsa_decompress_stream(lzsa_stream_t *pInStream, lzsa_stream_t *pOutStream, const void *pDictionaryData, int nDictionaryDataSize, const unsigned int nFlags, int nFormatVersion,
+      long long *pOriginalSize, long long *pCompressedSize) {
+   long long nOriginalSize = 0LL, nCompressedSize = 0LL;
+   unsigned char cFrameData[16];
+   unsigned char *pInBlock;
+   unsigned char *pOutData;
+
+   if ((nFlags & LZSA_FLAG_RAW_BLOCK) == 0) {
+      const int nHeaderSize = lzsa_get_header_size();
+
+      memset(cFrameData, 0, 16);
+      if (pInStream->read(pInStream, cFrameData, nHeaderSize) != nHeaderSize) {
+         return LZSA_ERROR_SRC;
+      }
+
+      if (lzsa_decode_header(cFrameData, nHeaderSize, &nFormatVersion) < 0) {
+         return LZSA_ERROR_FORMAT;
+      }
+
+      nCompressedSize += (long long)nHeaderSize;
+   }
+
+   pInBlock = (unsigned char*)malloc(BLOCK_SIZE);
+   if (!pInBlock) {
+      return LZSA_ERROR_MEMORY;
+   }
+
+   pOutData = (unsigned char*)malloc(BLOCK_SIZE * 2);
+   if (!pOutData) {
+      free(pInBlock);
+      pInBlock = NULL;
+
+      return LZSA_ERROR_MEMORY;
+   }
+
+   int nDecompressionError = 0;
+   int nPrevDecompressedSize = 0;
+   int nNumBlocks = 0;
+
+   while (!pInStream->eof(pInStream) && !nDecompressionError) {
+      unsigned int nBlockSize = 0;
+      int nIsUncompressed = 0;
+
+      if (nPrevDecompressedSize != 0) {
+         memcpy(pOutData + BLOCK_SIZE - nPrevDecompressedSize, pOutData + BLOCK_SIZE, nPrevDecompressedSize);
+      }
+      else if (nDictionaryDataSize && pDictionaryData) {
+         nPrevDecompressedSize = nDictionaryDataSize;
+         memcpy(pOutData + BLOCK_SIZE - nPrevDecompressedSize, pDictionaryData, nPrevDecompressedSize);
+      }
+
+      if ((nFlags & LZSA_FLAG_RAW_BLOCK) == 0) {
+         const int nFrameSize = lzsa_get_frame_size();
+
+         memset(cFrameData, 0, 16);
+         if (pInStream->read(pInStream, cFrameData, nFrameSize) == nFrameSize) {
+            if (lzsa_decode_frame(cFrameData, nFrameSize, &nBlockSize, &nIsUncompressed) < 0) {
+               nDecompressionError = LZSA_ERROR_FORMAT;
+               nBlockSize = 0;
+            }
+
+            nCompressedSize += (long long)nFrameSize;
+         }
+         else {
+            nDecompressionError = LZSA_ERROR_SRC;
+            nBlockSize = 0;
+         }
+      }
+      else {
+         if (!nNumBlocks)
+            nBlockSize = BLOCK_SIZE;
+         else
+            nBlockSize = 0;
+      }
+
+      if (nBlockSize != 0) {
+         int nDecompressedSize = 0;
+
+         if ((int)nBlockSize > BLOCK_SIZE) {
+            nDecompressionError = LZSA_ERROR_FORMAT;
+            break;
+         }
+         size_t nReadBytes = pInStream->read(pInStream, pInBlock, nBlockSize);
+         if (nFlags & LZSA_FLAG_RAW_BLOCK) {
+            nBlockSize = (unsigned int)nReadBytes;
+         }
+
+         if (nReadBytes == nBlockSize) {
+            nCompressedSize += (long long)nReadBytes;
+
+            if (nIsUncompressed) {
+               memcpy(pOutData + BLOCK_SIZE, pInBlock, nBlockSize);
+               nDecompressedSize = nBlockSize;
+            }
+            else {
+               nDecompressedSize = lzsa_decompressor_expand_block(pInBlock, nBlockSize, pOutData, BLOCK_SIZE, BLOCK_SIZE, nFormatVersion, nFlags);
+               if (nDecompressedSize < 0) {
+                  nDecompressionError = LZSA_ERROR_DECOMPRESSION;
+                  break;
+               }
+            }
+
+            if (nDecompressedSize != 0) {
+               nOriginalSize += (long long)nDecompressedSize;
+
+               if (pOutStream->write(pOutStream, pOutData + BLOCK_SIZE, nDecompressedSize) != nDecompressedSize)
+                  nDecompressionError = LZSA_ERROR_DST;
+               nPrevDecompressedSize = nDecompressedSize;
+               nDecompressedSize = 0;
+            }
+         }
+         else {
+            break;
+         }
+
+         nNumBlocks++;
+      }
+      else {
+         break;
+      }
+   }
+
+   free(pOutData);
+   pOutData = NULL;
+
+   free(pInBlock);
+   pInBlock = NULL;
+
+   *pOriginalSize = nOriginalSize;
+   *pCompressedSize = nCompressedSize;
+   return nDecompressionError;
+}
+

loader/tools/lzsa/src/expand_streaming.h

@@ -0,0 +1,86 @@
+/*
+ * expand_streaming.h - streaming decompression definitions
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#ifndef _EXPAND_STREAMING_H
+#define _EXPAND_STREAMING_H
+
+#include "stream.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/* Forward declaration */
+typedef enum _lzsa_status_t lzsa_status_t;
+
+/*-------------- File API -------------- */
+
+/**
+ * Decompress file
+ *
+ * @param pszInFilename name of input(compressed) file to decompress
+ * @param pszOutFilename name of output(decompressed) file to generate
+ * @param pszDictionaryFilename name of dictionary file, or NULL for none
+ * @param nFlags compression flags (LZSA_FLAG_RAW_BLOCK to decompress a raw block, or 0)
+ * @param nFormatVersion default version of format to use (1-2). This is used when decompressing a raw block, otherwise the version is extracted from the source file
+ * @param pOriginalSize pointer to returned output(decompressed) size, updated when this function is successful
+ * @param pCompressedSize pointer to returned input(compressed) size, updated when this function is successful
+ *
+ * @return LZSA_OK for success, or an error value from lzsa_status_t
+ */
+lzsa_status_t lzsa_decompress_file(const char *pszInFilename, const char *pszOutFilename, const char *pszDictionaryFilename, const unsigned int nFlags, int nFormatVersion,
+   long long *pOriginalSize, long long *pCompressedSize);
+
+/*-------------- Streaming API -------------- */
+
+/**
+ * Decompress stream
+ *
+ * @param pInStream input(compressed) stream to decompress
+ * @param pOutStream output(decompressed) stream to write to
+ * @param pDictionaryData dictionary contents, or NULL for none
+ * @param nDictionaryDataSize size of dictionary contents, or 0
+ * @param nFlags compression flags (LZSA_FLAG_RAW_BLOCK to decompress a raw block, or 0)
+ * @param nFormatVersion default version of format to use (1-2). This is used when decompressing a raw block, otherwise the version is extracted from the source file
+ * @param pOriginalSize pointer to returned output(decompressed) size, updated when this function is successful
+ * @param pCompressedSize pointer to returned input(compressed) size, updated when this function is successful
+ *
+ * @return LZSA_OK for success, or an error value from lzsa_status_t
+ */
+lzsa_status_t lzsa_decompress_stream(lzsa_stream_t *pInStream, lzsa_stream_t *pOutStream, const void *pDictionaryData, int nDictionaryDataSize, const unsigned int nFlags, int nFormatVersion,
+   long long *pOriginalSize, long long *pCompressedSize);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _EXPAND_STREAMING_H */

loader/tools/lzsa/src/format.h

@@ -0,0 +1,54 @@
+/*
+ * format.h - byte stream format definitions
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#ifndef _FORMAT_H
+#define _FORMAT_H
+
+#define MIN_OFFSET 1
+#define MAX_OFFSET 0xffff
+
+#define MAX_VARLEN 0xffff
+
+#define BLOCK_SIZE 65536
+
+#define MIN_MATCH_SIZE_V1 3
+#define LITERALS_RUN_LEN_V1 7
+#define MATCH_RUN_LEN_V1 15
+
+#define MIN_MATCH_SIZE_V2 2
+#define LITERALS_RUN_LEN_V2 3
+#define MATCH_RUN_LEN_V2 7
+
+#define LITERALS_LEN_POS_V2 0
+#define MATCH_LEN_POS_V2 2
+
+#endif /* _FORMAT_H */

loader/tools/lzsa/src/frame.c

@@ -0,0 +1,236 @@
+/*
+ * frame.c - frame implementation
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+/*
+ *
+ * Changes to implement lzsa2 as depacker/packer for the bitfire demo-framework
+ * 2021 by Tobias Bindhammer aka Bitbreaker/Performers^Oxyron tobias.bindhammer@uni-ulm.de
+ *
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include "frame.h"
+
+#define LZSA_ID_0   0x7b
+#define LZSA_ID_1   0x9e
+
+/**
+ * Get compressed file header size
+ *
+ * @return file header size
+ */
+int lzsa_get_header_size(void) {
+   return 3;
+}
+
+/**
+ * Get compressed frame header size
+ *
+ * @return frame header size
+ */
+int lzsa_get_frame_size(void) {
+   return 3;
+}
+
+/**
+ * Encode file address big-endian
+ *
+ * @param pFrameData encoding buffer
+ * @param nMaxFrameDataSize max encoding buffer size, in bytes
+ * @param nAddress address to encode
+ *
+ * @return number of encoded bytes, or -1 for failure
+ */
+int lzsa_encode_addr_be(unsigned char *pFrameData, const int nMaxFrameDataSize, int nAddress) {
+   if (nMaxFrameDataSize >= 2) {
+      pFrameData[0] = (nAddress >> 8) & 0xff;
+      pFrameData[1] = nAddress & 0xff;
+      return 2;
+   }
+   else {
+      return -1;
+   }
+}
+
+/**
+ * Encode file address little-endian
+ *
+ * @param pFrameData encoding buffer
+ * @param nMaxFrameDataSize max encoding buffer size, in bytes
+ * @param nAddress address to encode
+ *
+ * @return number of encoded bytes, or -1 for failure
+ */
+int lzsa_encode_addr_le(unsigned char *pFrameData, const int nMaxFrameDataSize, int nAddress) {
+   if (nMaxFrameDataSize >= 2) {
+      pFrameData[0] = nAddress & 0xff;
+      pFrameData[1] = (nAddress >> 8) & 0xff;
+      return 2;
+   }
+   else {
+      return -1;
+   }
+}
+
+/**
+ * Encode terminal frame header
+ *
+ * @param pFrameData encoding buffer
+ * @param nMaxFrameDataSize max encoding buffer size, in bytes
+ *
+ * @return number of encoded bytes, or -1 for failure
+ */
+int lzsa_encode_footer_frame(unsigned char *pFrameData, const int nMaxFrameDataSize) {
+   if (nMaxFrameDataSize >= 0) {
+      return 0;
+   }
+   else {
+      return -1;
+   }
+}
+
+/* XXX TODO code below can be removed */
+
+
+/**
+ * Encode uncompressed block frame header
+ *
+ * @param pFrameData encoding buffer
+ * @param nMaxFrameDataSize max encoding buffer size, in bytes
+ * @param nBlockDataSize uncompressed block's data size, in bytes
+ *
+ * @return number of encoded bytes, or -1 for failure
+ */
+int lzsa_encode_uncompressed_block_frame(unsigned char *pFrameData, const int nMaxFrameDataSize, const int nBlockDataSize) {
+   if (nMaxFrameDataSize >= 3 && nBlockDataSize <= 0x7fffff) {
+      pFrameData[0] = nBlockDataSize & 0xff;
+      pFrameData[1] = (nBlockDataSize >> 8) & 0xff;
+      pFrameData[2] = ((nBlockDataSize >> 16) & 0x7f) | 0x80;   /* Uncompressed block */
+
+      return 3;
+   }
+   else {
+      return -1;
+   }
+}
+
+/**
+ * Encode file header
+ *
+ * @param pFrameData encoding buffer
+ * @param nMaxFrameDataSize max encoding buffer size, in bytes
+ *
+ * @return number of encoded bytes, or -1 for failure
+ */
+int lzsa_encode_header(unsigned char *pFrameData, const int nMaxFrameDataSize, int nFormatVersion) {
+   if (nMaxFrameDataSize >= 3 && (nFormatVersion == 1 || nFormatVersion == 2)) {
+      pFrameData[0] = LZSA_ID_0;                         /* Magic number */
+      pFrameData[1] = LZSA_ID_1;
+      pFrameData[2] = (nFormatVersion == 2) ? 0x20 : 0;  /* Format version 1 */
+
+      return 3;
+   }
+   else {
+      return -1;
+   }
+}
+
+/**
+ * Encode compressed block frame header
+ *
+ * @param pFrameData encoding buffer
+ * @param nMaxFrameDataSize max encoding buffer size, in bytes
+ * @param nBlockDataSize compressed block's data size, in bytes
+ *
+ * @return number of encoded bytes, or -1 for failure
+ */
+int lzsa_encode_compressed_block_frame(unsigned char *pFrameData, const int nMaxFrameDataSize, const int nBlockDataSize) {
+   if (nMaxFrameDataSize >= 3 && nBlockDataSize <= 0x7fffff) {
+      pFrameData[0] = nBlockDataSize & 0xff;
+      pFrameData[1] = (nBlockDataSize >> 8) & 0xff;
+      pFrameData[2] = (nBlockDataSize >> 16) & 0x7f;
+
+      return 3;
+   }
+   else {
+      return -1;
+   }
+}
+
+/**
+ * Decode file header
+ *
+ * @param pFrameData data bytes
+ * @param nFrameDataSize number of bytes to decode
+ *
+ * @return 0 for success, or -1 for failure
+ */
+int lzsa_decode_header(const unsigned char *pFrameData, const int nFrameDataSize, int *nFormatVersion) {
+   if (nFrameDataSize != 3 ||
+      pFrameData[0] != LZSA_ID_0 ||
+      pFrameData[1] != LZSA_ID_1 ||
+      (pFrameData[2] & 0x1f) != 0 ||
+      ((pFrameData[2] & 0xe0) != 0x00 && (pFrameData[2] & 0xe0) != 0x20)) {
+      return -1;
+   }
+   else {
+      *nFormatVersion = (pFrameData[2] & 0xe0) ? 2 : 1;
+      return 0;
+   }
+}
+
+/**
+ * Decode frame header
+ *
+ * @param pFrameData data bytes
+ * @param nFrameDataSize number of bytes to decode
+ * @param nBlockSize pointer to block size, updated if this function succeeds (set to 0 if this is the terminal frame)
+ * @param nIsUncompressed pointer to compressed block flag, updated if this function succeeds
+ *
+ * @return 0 for success, or -1 for failure
+ */
+int lzsa_decode_frame(const unsigned char *pFrameData, const int nFrameDataSize, unsigned int *nBlockSize, int *nIsUncompressed) {
+   if (nFrameDataSize == 3) {
+      *nBlockSize = ((unsigned int)pFrameData[0]) |
+         (((unsigned int)pFrameData[1]) << 8) |
+         (((unsigned int)pFrameData[2]) << 16);
+
+      *nIsUncompressed = ((*nBlockSize & 0x800000) != 0) ? 1 : 0;
+      *nBlockSize &= 0x7fffff;
+      return 0;
+   }
+   else {
+      return -1;
+   }
+}

loader/tools/lzsa/src/frame.h

@@ -0,0 +1,144 @@
+/*
+ * frame.h - frame definitions
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#ifndef _FRAME_H
+#define _FRAME_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * Get compressed file header size
+ *
+ * @return file header size
+ */
+int lzsa_get_header_size(void);
+
+/**
+ * Get compressed frame header size
+ *
+ * @return frame header size
+ */
+int lzsa_get_frame_size(void);
+
+/**
+ * Encode file address big-endian
+ *
+ * @param pFrameData encoding buffer
+ * @param nMaxFrameDataSize max encoding buffer size, in bytes
+ * @param nAddress address to encode
+ *
+ * @return number of encoded bytes, or -1 for failure
+ */
+int lzsa_encode_addr_be(unsigned char *pFrameData, const int nMaxFrameDataSize, int nAddress);
+
+/**
+ * Encode file address little-endian
+ *
+ * @param pFrameData encoding buffer
+ * @param nMaxFrameDataSize max encoding buffer size, in bytes
+ * @param nAddress address to encode
+ *
+ * @return number of encoded bytes, or -1 for failure
+ */
+int lzsa_encode_addr_le(unsigned char *pFrameData, const int nMaxFrameDataSize, int nAddress);
+
+/**
+ * Encode file header
+ *
+ * @param pFrameData encoding buffer
+ * @param nMaxFrameDataSize max encoding buffer size, in bytes
+ *
+ * @return number of encoded bytes, or -1 for failure
+ */
+int lzsa_encode_header(unsigned char *pFrameData, const int nMaxFrameDataSize, int nFormatVersion);
+
+/**
+ * Encode compressed block frame header
+ *
+ * @param pFrameData encoding buffer
+ * @param nMaxFrameDataSize max encoding buffer size, in bytes
+ * @param nBlockDataSize compressed block's data size, in bytes
+ *
+ * @return number of encoded bytes, or -1 for failure
+ */
+int lzsa_encode_compressed_block_frame(unsigned char *pFrameData, const int nMaxFrameDataSize, const int nBlockDataSize);
+
+/**
+ * Encode uncompressed block frame header
+ *
+ * @param pFrameData encoding buffer
+ * @param nMaxFrameDataSize max encoding buffer size, in bytes
+ * @param nBlockDataSize uncompressed block's data size, in bytes
+ *
+ * @return number of encoded bytes, or -1 for failure
+ */
+int lzsa_encode_uncompressed_block_frame(unsigned char *pFrameData, const int nMaxFrameDataSize, const int nBlockDataSize);
+
+/**
+ * Encode terminal frame header
+ *
+ * @param pFrameData encoding buffer
+ * @param nMaxFrameDataSize max encoding buffer size, in bytes
+ *
+ * @return number of encoded bytes, or -1 for failure
+ */
+int lzsa_encode_footer_frame(unsigned char *pFrameData, const int nMaxFrameDataSize);
+
+/**
+ * Decode file header
+ *
+ * @param pFrameData data bytes
+ * @param nFrameDataSize number of bytes to decode
+ *
+ * @return 0 for success, or -1 for failure
+ */
+int lzsa_decode_header(const unsigned char *pFrameData, const int nFrameDataSize, int *nFormatVersion);
+
+/**
+ * Decode frame header
+ *
+ * @param pFrameData data bytes
+ * @param nFrameDataSize number of bytes to decode
+ * @param nBlockSize pointer to block size, updated if this function succeeds (set to 0 if this is the terminal frame)
+ * @param nIsUncompressed pointer to compressed block flag, updated if this function succeeds
+ *
+ * @return 0 for success, or -1 for failure
+ */
+int lzsa_decode_frame(const unsigned char *pFrameData, const int nFrameDataSize, unsigned int *nBlockSize, int *nIsUncompressed);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _FRAME_H */

loader/tools/lzsa/src/lib.h

@@ -0,0 +1,95 @@
+/*
+ * lib.h - LZSA library definitions
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#ifndef _LIB_H
+#define _LIB_H
+
+#include "stream.h"
+#include "dictionary.h"
+#include "frame.h"
+#include "format.h"
+#include "shrink_context.h"
+#include "shrink_streaming.h"
+#include "shrink_inmem.h"
+#include "expand_context.h"
+#include "expand_streaming.h"
+#include "expand_inmem.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/** High level status for compression and decompression */
+typedef enum _lzsa_status_t {
+   LZSA_OK = 0,                           /**< Success */
+   LZSA_ERROR_SRC,                        /**< Error reading input */
+   LZSA_ERROR_DST,                        /**< Error reading output */
+   LZSA_ERROR_DICTIONARY,                 /**< Error reading dictionary */
+   LZSA_ERROR_MEMORY,                     /**< Out of memory */
+
+   /* Compression-specific status codes */
+   LZSA_ERROR_COMPRESSION,                /**< Internal compression error */
+   LZSA_ERROR_RAW_TOOLARGE,               /**< Input is too large to be compressed to a raw block */
+   LZSA_ERROR_RAW_UNCOMPRESSED,           /**< Input is incompressible and raw blocks don't support uncompressed data */
+
+   /* Decompression-specific status codes */
+   LZSA_ERROR_FORMAT,                     /**< Invalid input format or magic number when decompressing */
+   LZSA_ERROR_DECOMPRESSION               /**< Internal decompression error */
+} lzsa_status_t;
+
+/* Compression flags */
+#define LZSA_FLAG_FAVOR_RATIO    (1<<0)      /**< 1 to compress with the best ratio, 0 to trade some compression ratio for extra decompression speed */
+#define LZSA_FLAG_RAW_BLOCK      (1<<1)      /**< 1 to emit raw block */
+#define LZSA_FLAG_RAW_BACKWARD   (1<<2)      /**< 1 to compress or decompress raw block backward */
+
+/**
+ * Reverse bytes in the specified buffer
+ *
+ * @param pBuffer pointer to buffer whose contents are to be reversed
+ * @param nBufferSize size of buffer in bytes
+ */
+static inline void lzsa_reverse_buffer(unsigned char *pBuffer, const int nBufferSize) {
+   int nMidPoint = nBufferSize / 2;
+   int i, j;
+
+   for (i = 0, j = nBufferSize - 1; i < nMidPoint; i++, j--) {
+      unsigned char c = pBuffer[i];
+      pBuffer[i] = pBuffer[j];
+      pBuffer[j] = c;
+   }
+}
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _LIB_H */

loader/tools/lzsa/src/libdivsufsort/.gitignore

@@ -0,0 +1,32 @@
+# Object files
+*.o
+*.ko
+*.obj
+*.elf
+
+# Precompiled Headers
+*.gch
+*.pch
+
+# Libraries
+*.lib
+*.a
+*.la
+*.lo
+
+# Shared objects (inc. Windows DLLs)
+*.dll
+*.so
+*.so.*
+*.dylib
+
+# Executables
+*.exe
+*.out
+*.app
+*.i*86
+*.x86_64
+*.hex
+
+# CMake files/directories
+build/

loader/tools/lzsa/src/libdivsufsort/CHANGELOG.md

@@ -0,0 +1,21 @@
+# libdivsufsort Change Log
+
+See full changelog at: https://github.com/y-256/libdivsufsort/commits
+
+## [2.0.1] - 2010-11-11
+### Fixed
+* Wrong variable used in `divbwt` function
+* Enclose some string variables with double quotation marks in include/CMakeLists.txt
+* Fix typo in include/CMakeLists.txt
+
+## 2.0.0 - 2008-08-23
+### Changed
+* Switch the build system to [CMake](http://www.cmake.org/)
+* Improve the performance of the suffix-sorting algorithm
+
+### Added
+* OpenMP support
+* 64-bit version of divsufsort
+
+[Unreleased]: https://github.com/y-256/libdivsufsort/compare/2.0.1...HEAD
+[2.0.1]: https://github.com/y-256/libdivsufsort/compare/2.0.0...2.0.1

loader/tools/lzsa/src/libdivsufsort/CMakeLists.txt

@@ -0,0 +1,99 @@
+### cmake file for building libdivsufsort Package ###
+cmake_minimum_required(VERSION 2.4.4)
+set(CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/CMakeModules")
+include(AppendCompilerFlags)
+
+## Project information ##
+project(libdivsufsort C)
+set(PROJECT_VENDOR "Yuta Mori")
+set(PROJECT_CONTACT "yuta.256@gmail.com")
+set(PROJECT_URL "https://github.com/y-256/libdivsufsort")
+set(PROJECT_DESCRIPTION "A lightweight suffix sorting library")
+include(VERSION.cmake)
+
+## CPack configuration ##
+set(CPACK_GENERATOR "TGZ;TBZ2;ZIP")
+set(CPACK_SOURCE_GENERATOR "TGZ;TBZ2;ZIP")
+include(ProjectCPack)
+
+## Project options ##
+option(BUILD_SHARED_LIBS "Set to OFF to build static libraries" ON)
+option(BUILD_EXAMPLES "Build examples" ON)
+option(BUILD_DIVSUFSORT64 "Build libdivsufsort64" OFF)
+option(USE_OPENMP "Use OpenMP for parallelization" OFF)
+option(WITH_LFS "Enable Large File Support" ON)
+
+## Installation directories ##
+set(LIB_SUFFIX "" CACHE STRING "Define suffix of directory name (32 or 64)")
+
+set(CMAKE_INSTALL_RUNTIMEDIR "" CACHE PATH "Specify the output directory for dll runtimes (default is bin)")
+if(NOT CMAKE_INSTALL_RUNTIMEDIR)
+  set(CMAKE_INSTALL_RUNTIMEDIR "${CMAKE_INSTALL_PREFIX}/bin")
+endif(NOT CMAKE_INSTALL_RUNTIMEDIR)
+
+set(CMAKE_INSTALL_LIBDIR "" CACHE PATH "Specify the output directory for libraries (default is lib)")
+if(NOT CMAKE_INSTALL_LIBDIR)
+  set(CMAKE_INSTALL_LIBDIR "${CMAKE_INSTALL_PREFIX}/lib${LIB_SUFFIX}")
+endif(NOT CMAKE_INSTALL_LIBDIR)
+
+set(CMAKE_INSTALL_INCLUDEDIR "" CACHE PATH "Specify the output directory for header files (default is include)")
+if(NOT CMAKE_INSTALL_INCLUDEDIR)
+  set(CMAKE_INSTALL_INCLUDEDIR "${CMAKE_INSTALL_PREFIX}/include")
+endif(NOT CMAKE_INSTALL_INCLUDEDIR)
+
+set(CMAKE_INSTALL_PKGCONFIGDIR "" CACHE PATH "Specify the output directory for pkgconfig files (default is lib/pkgconfig)")
+if(NOT CMAKE_INSTALL_PKGCONFIGDIR)
+  set(CMAKE_INSTALL_PKGCONFIGDIR "${CMAKE_INSTALL_LIBDIR}/pkgconfig")
+endif(NOT CMAKE_INSTALL_PKGCONFIGDIR)
+
+## Build type ##
+if(NOT CMAKE_BUILD_TYPE)
+  set(CMAKE_BUILD_TYPE "Release")
+elseif(CMAKE_BUILD_TYPE STREQUAL "Debug")
+  set(CMAKE_VERBOSE_MAKEFILE ON)
+endif(NOT CMAKE_BUILD_TYPE)
+
+## Compiler options ##
+if(MSVC)
+  append_c_compiler_flags("/W4" "VC" CMAKE_C_FLAGS)
+  append_c_compiler_flags("/Oi;/Ot;/Ox;/Oy" "VC" CMAKE_C_FLAGS_RELEASE)
+  if(USE_OPENMP)
+    append_c_compiler_flags("/openmp" "VC" CMAKE_C_FLAGS)
+  endif(USE_OPENMP)
+elseif(BORLAND)
+  append_c_compiler_flags("-w" "BCC" CMAKE_C_FLAGS)
+  append_c_compiler_flags("-Oi;-Og;-Os;-Ov;-Ox" "BCC" CMAKE_C_FLAGS_RELEASE)
+else(MSVC)
+  if(CMAKE_COMPILER_IS_GNUCC)
+    append_c_compiler_flags("-Wall" "GCC" CMAKE_C_FLAGS)
+    append_c_compiler_flags("-fomit-frame-pointer" "GCC" CMAKE_C_FLAGS_RELEASE)
+    if(USE_OPENMP)
+      append_c_compiler_flags("-fopenmp" "GCC" CMAKE_C_FLAGS)
+    endif(USE_OPENMP)
+  else(CMAKE_COMPILER_IS_GNUCC)
+    append_c_compiler_flags("-Wall" "UNKNOWN" CMAKE_C_FLAGS)
+    append_c_compiler_flags("-fomit-frame-pointer" "UNKNOWN" CMAKE_C_FLAGS_RELEASE)
+    if(USE_OPENMP)
+      append_c_compiler_flags("-fopenmp;-openmp;-omp" "UNKNOWN" CMAKE_C_FLAGS)
+    endif(USE_OPENMP)
+  endif(CMAKE_COMPILER_IS_GNUCC)
+endif(MSVC)
+
+## Add definitions ##
+add_definitions(-DHAVE_CONFIG_H=1 -D__STDC_LIMIT_MACROS -D__STDC_CONSTANT_MACROS -D__STDC_FORMAT_MACROS)
+
+## Add subdirectories ##
+add_subdirectory(pkgconfig)
+add_subdirectory(include)
+add_subdirectory(lib)
+if(BUILD_EXAMPLES)
+  add_subdirectory(examples)
+endif(BUILD_EXAMPLES)
+
+## Add 'uninstall' target ##
+CONFIGURE_FILE(
+  "${CMAKE_CURRENT_SOURCE_DIR}/CMakeModules/cmake_uninstall.cmake.in"
+  "${CMAKE_CURRENT_BINARY_DIR}/CMakeModules/cmake_uninstall.cmake"
+  IMMEDIATE @ONLY)
+ADD_CUSTOM_TARGET(uninstall
+  "${CMAKE_COMMAND}" -P "${CMAKE_CURRENT_BINARY_DIR}/CMakeModules/cmake_uninstall.cmake")

loader/tools/lzsa/src/libdivsufsort/CMakeModules/AppendCompilerFlags.cmake

@@ -0,0 +1,38 @@
+include(CheckCSourceCompiles)
+include(CheckCXXSourceCompiles)
+
+macro(append_c_compiler_flags _flags _name _result)
+  set(SAFE_CMAKE_REQUIRED_FLAGS ${CMAKE_REQUIRED_FLAGS})
+  string(REGEX REPLACE "[-+/ ]" "_" cname "${_name}")
+  string(TOUPPER "${cname}" cname)
+  foreach(flag ${_flags})
+    string(REGEX REPLACE "^[-+/ ]+(.*)[-+/ ]*$" "\\1" flagname "${flag}")
+    string(REGEX REPLACE "[-+/ ]" "_" flagname "${flagname}")
+    string(TOUPPER "${flagname}" flagname)
+    set(have_flag "HAVE_${cname}_${flagname}")
+    set(CMAKE_REQUIRED_FLAGS "${flag}")
+    check_c_source_compiles("int main() { return 0; }" ${have_flag})
+    if(${have_flag})
+      set(${_result} "${${_result}} ${flag}")
+    endif(${have_flag})
+  endforeach(flag)
+  set(CMAKE_REQUIRED_FLAGS ${SAFE_CMAKE_REQUIRED_FLAGS})
+endmacro(append_c_compiler_flags)
+
+macro(append_cxx_compiler_flags _flags _name _result)
+  set(SAFE_CMAKE_REQUIRED_FLAGS ${CMAKE_REQUIRED_FLAGS})
+  string(REGEX REPLACE "[-+/ ]" "_" cname "${_name}")
+  string(TOUPPER "${cname}" cname)
+  foreach(flag ${_flags})
+    string(REGEX REPLACE "^[-+/ ]+(.*)[-+/ ]*$" "\\1" flagname "${flag}")
+    string(REGEX REPLACE "[-+/ ]" "_" flagname "${flagname}")
+    string(TOUPPER "${flagname}" flagname)
+    set(have_flag "HAVE_${cname}_${flagname}")
+    set(CMAKE_REQUIRED_FLAGS "${flag}")
+    check_cxx_source_compiles("int main() { return 0; }" ${have_flag})
+    if(${have_flag})
+      set(${_result} "${${_result}} ${flag}")
+    endif(${have_flag})
+  endforeach(flag)
+  set(CMAKE_REQUIRED_FLAGS ${SAFE_CMAKE_REQUIRED_FLAGS})
+endmacro(append_cxx_compiler_flags)

loader/tools/lzsa/src/libdivsufsort/CMakeModules/CheckFunctionKeywords.cmake

@@ -0,0 +1,15 @@
+include(CheckCSourceCompiles)
+
+macro(check_function_keywords _wordlist)
+  set(${_result} "")
+  foreach(flag ${_wordlist})
+    string(REGEX REPLACE "[-+/ ()]" "_" flagname "${flag}")
+    string(TOUPPER "${flagname}" flagname)
+    set(have_flag "HAVE_${flagname}")
+    check_c_source_compiles("${flag} void func(); void func() { } int main() { func(); return 0; }" ${have_flag})
+    if(${have_flag} AND NOT ${_result})
+      set(${_result} "${flag}")
+#      break()
+    endif(${have_flag} AND NOT ${_result})
+  endforeach(flag)
+endmacro(check_function_keywords)

loader/tools/lzsa/src/libdivsufsort/CMakeModules/CheckLFS.cmake

@@ -0,0 +1,109 @@
+## Checks for large file support ##
+include(CheckIncludeFile)
+include(CheckSymbolExists)
+include(CheckTypeSize)
+
+macro(check_lfs _isenable)
+  set(LFS_OFF_T "")
+  set(LFS_FOPEN "")
+  set(LFS_FSEEK "")
+  set(LFS_FTELL "")
+  set(LFS_PRID "")
+
+  if(${_isenable})
+    set(SAFE_CMAKE_REQUIRED_DEFINITIONS "${CMAKE_REQUIRED_DEFINITIONS}")
+    set(CMAKE_REQUIRED_DEFINITIONS ${CMAKE_REQUIRED_DEFINITIONS}
+        -D_LARGEFILE_SOURCE -D_LARGE_FILES -D_FILE_OFFSET_BITS=64
+        -D__STDC_LIMIT_MACROS -D__STDC_CONSTANT_MACROS -D__STDC_FORMAT_MACROS)
+
+    check_include_file("sys/types.h" HAVE_SYS_TYPES_H)
+    check_include_file("inttypes.h" HAVE_INTTYPES_H)
+    check_include_file("stddef.h" HAVE_STDDEF_H)
+    check_include_file("stdint.h" HAVE_STDINT_H)
+
+    # LFS type1: 8 <= sizeof(off_t), fseeko, ftello
+    check_type_size("off_t" SIZEOF_OFF_T)
+    if(SIZEOF_OFF_T GREATER 7)
+      check_symbol_exists("fseeko" "stdio.h" HAVE_FSEEKO)
+      check_symbol_exists("ftello" "stdio.h" HAVE_FTELLO)
+      if(HAVE_FSEEKO AND HAVE_FTELLO)
+        set(LFS_OFF_T "off_t")
+        set(LFS_FOPEN "fopen")
+        set(LFS_FSEEK "fseeko")
+        set(LFS_FTELL "ftello")
+        check_symbol_exists("PRIdMAX" "inttypes.h" HAVE_PRIDMAX)
+        if(HAVE_PRIDMAX)
+          set(LFS_PRID "PRIdMAX")
+        else(HAVE_PRIDMAX)
+          check_type_size("long" SIZEOF_LONG)
+          check_type_size("int" SIZEOF_INT)
+          if(SIZEOF_OFF_T GREATER SIZEOF_LONG)
+            set(LFS_PRID "\"lld\"")
+          elseif(SIZEOF_LONG GREATER SIZEOF_INT)
+            set(LFS_PRID "\"ld\"")
+          else(SIZEOF_OFF_T GREATER SIZEOF_LONG)
+            set(LFS_PRID "\"d\"")
+          endif(SIZEOF_OFF_T GREATER SIZEOF_LONG)
+        endif(HAVE_PRIDMAX)
+      endif(HAVE_FSEEKO AND HAVE_FTELLO)
+    endif(SIZEOF_OFF_T GREATER 7)
+
+    # LFS type2: 8 <= sizeof(off64_t), fopen64, fseeko64, ftello64
+    if(NOT LFS_OFF_T)
+      check_type_size("off64_t" SIZEOF_OFF64_T)
+      if(SIZEOF_OFF64_T GREATER 7)
+        check_symbol_exists("fopen64" "stdio.h" HAVE_FOPEN64)
+        check_symbol_exists("fseeko64" "stdio.h" HAVE_FSEEKO64)
+        check_symbol_exists("ftello64" "stdio.h" HAVE_FTELLO64)
+        if(HAVE_FOPEN64 AND HAVE_FSEEKO64 AND HAVE_FTELLO64)
+          set(LFS_OFF_T "off64_t")
+          set(LFS_FOPEN "fopen64")
+          set(LFS_FSEEK "fseeko64")
+          set(LFS_FTELL "ftello64")
+          check_symbol_exists("PRIdMAX" "inttypes.h" HAVE_PRIDMAX)
+          if(HAVE_PRIDMAX)
+            set(LFS_PRID "PRIdMAX")
+          else(HAVE_PRIDMAX)
+            check_type_size("long" SIZEOF_LONG)
+            check_type_size("int" SIZEOF_INT)
+            if(SIZEOF_OFF64_T GREATER SIZEOF_LONG)
+              set(LFS_PRID "\"lld\"")
+            elseif(SIZEOF_LONG GREATER SIZEOF_INT)
+              set(LFS_PRID "\"ld\"")
+            else(SIZEOF_OFF64_T GREATER SIZEOF_LONG)
+              set(LFS_PRID "\"d\"")
+            endif(SIZEOF_OFF64_T GREATER SIZEOF_LONG)
+          endif(HAVE_PRIDMAX)
+        endif(HAVE_FOPEN64 AND HAVE_FSEEKO64 AND HAVE_FTELLO64)
+      endif(SIZEOF_OFF64_T GREATER 7)
+    endif(NOT LFS_OFF_T)
+
+    # LFS type3: 8 <= sizeof(__int64), _fseeki64, _ftelli64
+    if(NOT LFS_OFF_T)
+      check_type_size("__int64" SIZEOF___INT64)
+      if(SIZEOF___INT64 GREATER 7)
+        check_symbol_exists("_fseeki64" "stdio.h" HAVE__FSEEKI64)
+        check_symbol_exists("_ftelli64" "stdio.h" HAVE__FTELLI64)
+        if(HAVE__FSEEKI64 AND HAVE__FTELLI64)
+          set(LFS_OFF_T "__int64")
+          set(LFS_FOPEN "fopen")
+          set(LFS_FSEEK "_fseeki64")
+          set(LFS_FTELL "_ftelli64")
+          set(LFS_PRID  "\"I64d\"")
+        endif(HAVE__FSEEKI64 AND HAVE__FTELLI64)
+      endif(SIZEOF___INT64 GREATER 7)
+    endif(NOT LFS_OFF_T)
+
+    set(CMAKE_REQUIRED_DEFINITIONS "${SAFE_CMAKE_REQUIRED_DEFINITIONS}")
+  endif(${_isenable})
+
+  if(NOT LFS_OFF_T)
+    ## not found
+    set(LFS_OFF_T "long")
+    set(LFS_FOPEN "fopen")
+    set(LFS_FSEEK "fseek")
+    set(LFS_FTELL "ftell")
+    set(LFS_PRID  "\"ld\"")
+  endif(NOT LFS_OFF_T)
+
+endmacro(check_lfs)

loader/tools/lzsa/src/libdivsufsort/CMakeModules/ProjectCPack.cmake

@@ -0,0 +1,38 @@
+# If the cmake version includes cpack, use it
+IF(EXISTS "${CMAKE_ROOT}/Modules/CPack.cmake")
+  SET(CPACK_PACKAGE_DESCRIPTION_SUMMARY "${PROJECT_DESCRIPTION}")
+  SET(CPACK_PACKAGE_VENDOR "${PROJECT_VENDOR}")
+  SET(CPACK_PACKAGE_DESCRIPTION_FILE "${CMAKE_CURRENT_SOURCE_DIR}/README.md")
+  SET(CPACK_RESOURCE_FILE_LICENSE "${CMAKE_CURRENT_SOURCE_DIR}/LICENSE")
+  SET(CPACK_PACKAGE_VERSION_MAJOR "${PROJECT_VERSION_MAJOR}")
+  SET(CPACK_PACKAGE_VERSION_MINOR "${PROJECT_VERSION_MINOR}")
+  SET(CPACK_PACKAGE_VERSION_PATCH "${PROJECT_VERSION_PATCH}")
+#  SET(CPACK_PACKAGE_INSTALL_DIRECTORY "${PROJECT_NAME} ${PROJECT_VERSION}")
+  SET(CPACK_SOURCE_PACKAGE_FILE_NAME "${PROJECT_NAME}-${PROJECT_VERSION_FULL}")
+
+  IF(NOT DEFINED CPACK_SYSTEM_NAME)
+    SET(CPACK_SYSTEM_NAME "${CMAKE_SYSTEM_NAME}-${CMAKE_SYSTEM_PROCESSOR}")
+  ENDIF(NOT DEFINED CPACK_SYSTEM_NAME)
+
+  IF(${CPACK_SYSTEM_NAME} MATCHES Windows)
+    IF(CMAKE_CL_64)
+      SET(CPACK_SYSTEM_NAME win64-${CMAKE_SYSTEM_PROCESSOR})
+    ELSE(CMAKE_CL_64)
+      SET(CPACK_SYSTEM_NAME win32-${CMAKE_SYSTEM_PROCESSOR})
+    ENDIF(CMAKE_CL_64)
+  ENDIF(${CPACK_SYSTEM_NAME} MATCHES Windows)
+
+  IF(NOT DEFINED CPACK_PACKAGE_FILE_NAME)
+    SET(CPACK_PACKAGE_FILE_NAME "${CPACK_SOURCE_PACKAGE_FILE_NAME}-${CPACK_SYSTEM_NAME}")
+  ENDIF(NOT DEFINED CPACK_PACKAGE_FILE_NAME)
+
+  SET(CPACK_PACKAGE_CONTACT "${PROJECT_CONTACT}")
+  IF(UNIX)
+    SET(CPACK_STRIP_FILES "")
+    SET(CPACK_SOURCE_STRIP_FILES "")
+#    SET(CPACK_PACKAGE_EXECUTABLES "ccmake" "CMake")
+  ENDIF(UNIX)
+  SET(CPACK_SOURCE_IGNORE_FILES "/CVS/" "/build/" "/\\\\.build/" "/\\\\.svn/" "~$")
+  # include CPack model once all variables are set
+  INCLUDE(CPack)
+ENDIF(EXISTS "${CMAKE_ROOT}/Modules/CPack.cmake")

loader/tools/lzsa/src/libdivsufsort/CMakeModules/cmake_uninstall.cmake.in

@@ -0,0 +1,36 @@
+IF(NOT EXISTS "@CMAKE_CURRENT_BINARY_DIR@/install_manifest.txt")
+  MESSAGE(FATAL_ERROR "Cannot find install manifest: \"@CMAKE_CURRENT_BINARY_DIR@/install_manifest.txt\"")
+ENDIF(NOT EXISTS "@CMAKE_CURRENT_BINARY_DIR@/install_manifest.txt")
+
+FILE(READ "@CMAKE_CURRENT_BINARY_DIR@/install_manifest.txt" files)
+STRING(REGEX REPLACE "\n" ";" files "${files}")
+
+SET(NUM 0)
+FOREACH(file ${files})
+  IF(EXISTS "$ENV{DESTDIR}${file}")
+    MESSAGE(STATUS "Looking for \"$ENV{DESTDIR}${file}\" - found")
+    SET(UNINSTALL_CHECK_${NUM} 1)
+  ELSE(EXISTS "$ENV{DESTDIR}${file}")
+    MESSAGE(STATUS "Looking for \"$ENV{DESTDIR}${file}\" - not found")
+    SET(UNINSTALL_CHECK_${NUM} 0)
+  ENDIF(EXISTS "$ENV{DESTDIR}${file}")
+  MATH(EXPR NUM "1 + ${NUM}")
+ENDFOREACH(file)
+
+SET(NUM 0)
+FOREACH(file ${files})
+  IF(${UNINSTALL_CHECK_${NUM}})
+    MESSAGE(STATUS "Uninstalling \"$ENV{DESTDIR}${file}\"")
+    EXEC_PROGRAM(
+      "@CMAKE_COMMAND@" ARGS "-E remove \"$ENV{DESTDIR}${file}\""
+      OUTPUT_VARIABLE rm_out
+      RETURN_VALUE rm_retval
+      )
+    IF(NOT "${rm_retval}" STREQUAL 0)
+      MESSAGE(FATAL_ERROR "Problem when removing \"$ENV{DESTDIR}${file}\"")
+    ENDIF(NOT "${rm_retval}" STREQUAL 0)
+  ENDIF(${UNINSTALL_CHECK_${NUM}})
+  MATH(EXPR NUM "1 + ${NUM}")
+ENDFOREACH(file)
+
+FILE(REMOVE "@CMAKE_CURRENT_BINARY_DIR@/install_manifest.txt")

loader/tools/lzsa/src/libdivsufsort/LICENSE

@@ -0,0 +1,21 @@
+The MIT License (MIT)
+
+Copyright (c) 2003 Yuta Mori All rights reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

loader/tools/lzsa/src/libdivsufsort/README.md

@@ -0,0 +1,140 @@
+# libdivsufsort
+
+libdivsufsort is a software library that implements a lightweight suffix array construction algorithm.
+
+## News
+* 2015-03-21: The project has moved from [Google Code](http://code.google.com/p/libdivsufsort/) to [GitHub](https://github.com/y-256/libdivsufsort)
+
+## Introduction
+This library provides a simple and an efficient C API to construct a suffix array and a Burrows-Wheeler transformed string from a given string over a constant-size alphabet.
+The algorithm runs in O(n log n) worst-case time using only 5n+O(1) bytes of memory space, where n is the length of
+the string.
+
+## Build requirements
+* An ANSI C Compiler (e.g. GNU GCC)
+* [CMake](http://www.cmake.org/ "CMake") version 2.4.2 or newer
+* CMake-supported build tool
+
+## Building on GNU/Linux
+1. Get the source code from GitHub. You can either
+    * use git to clone the repository
+    ```
+    git clone https://github.com/y-256/libdivsufsort.git
+    ```
+    * or download a [zip file](../../archive/master.zip) directly
+2. Create a `build` directory in the package source directory.
+```shell
+$ cd libdivsufsort
+$ mkdir build
+$ cd build
+```
+3. Configure the package for your system.
+If you want to install to a different location,  change the -DCMAKE_INSTALL_PREFIX option.
+```shell
+$ cmake -DCMAKE_BUILD_TYPE="Release" \
+-DCMAKE_INSTALL_PREFIX="/usr/local" ..
+```
+4. Compile the package.
+```shell
+$ make
+```
+5. (Optional) Install the library and header files.
+```shell
+$ sudo make install
+```
+
+## API
+```c
+/* Data types */
+typedef int32_t saint_t;
+typedef int32_t saidx_t;
+typedef uint8_t sauchar_t;
+
+/*
+ * Constructs the suffix array of a given string.
+ * @param T[0..n-1] The input string.
+ * @param SA[0..n-1] The output array or suffixes.
+ * @param n The length of the given string.
+ * @return 0 if no error occurred, -1 or -2 otherwise.
+ */
+saint_t
+divsufsort(const sauchar_t *T, saidx_t *SA, saidx_t n);
+
+/*
+ * Constructs the burrows-wheeler transformed string of a given string.
+ * @param T[0..n-1] The input string.
+ * @param U[0..n-1] The output string. (can be T)
+ * @param A[0..n-1] The temporary array. (can be NULL)
+ * @param n The length of the given string.
+ * @return The primary index if no error occurred, -1 or -2 otherwise.
+ */
+saidx_t
+divbwt(const sauchar_t *T, sauchar_t *U, saidx_t *A, saidx_t n);
+```
+
+## Example Usage
+```c
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include <divsufsort.h>
+
+int main() {
+    // intput data
+    char *Text = "abracadabra";
+    int n = strlen(Text);
+    int i, j;
+
+    // allocate
+    int *SA = (int *)malloc(n * sizeof(int));
+
+    // sort
+    divsufsort((unsigned char *)Text, SA, n);
+
+    // output
+    for(i = 0; i < n; ++i) {
+        printf("SA[%2d] = %2d: ", i, SA[i]);
+        for(j = SA[i]; j < n; ++j) {
+            printf("%c", Text[j]);
+        }
+        printf("$\n");
+    }
+
+    // deallocate
+    free(SA);
+
+    return 0;
+}
+```
+See the [examples](examples) directory for a few other examples.
+
+## Benchmarks
+See [Benchmarks](https://github.com/y-256/libdivsufsort/blob/wiki/SACA_Benchmarks.md) page for details.
+
+## License
+libdivsufsort is released under the [MIT license](LICENSE "MIT license").
+> The MIT License (MIT)
+>
+> Copyright (c) 2003 Yuta Mori All rights reserved.
+>
+> Permission is hereby granted, free of charge, to any person obtaining a copy
+> of this software and associated documentation files (the "Software"), to deal
+> in the Software without restriction, including without limitation the rights
+> to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+> copies of the Software, and to permit persons to whom the Software is
+> furnished to do so, subject to the following conditions:
+>
+> The above copyright notice and this permission notice shall be included in all
+> copies or substantial portions of the Software.
+>
+> THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+> IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+> FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+> AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+> LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+> OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+> SOFTWARE.
+
+## Author
+* Yuta Mori

loader/tools/lzsa/src/libdivsufsort/VERSION.cmake

@@ -0,0 +1,23 @@
+set(PROJECT_VERSION_MAJOR "2")
+set(PROJECT_VERSION_MINOR "0")
+set(PROJECT_VERSION_PATCH "2")
+set(PROJECT_VERSION_EXTRA "-1")
+set(PROJECT_VERSION "${PROJECT_VERSION_MAJOR}.${PROJECT_VERSION_MINOR}")
+set(PROJECT_VERSION_FULL "${PROJECT_VERSION_MAJOR}.${PROJECT_VERSION_MINOR}.${PROJECT_VERSION_PATCH}${PROJECT_VERSION_EXTRA}")
+
+set(LIBRARY_VERSION "3.0.1")
+set(LIBRARY_SOVERSION "3")
+
+## Git revision number ##
+if(EXISTS "${CMAKE_CURRENT_SOURCE_DIR}/.git")
+  execute_process(COMMAND git describe --tags HEAD
+    WORKING_DIRECTORY "${CMAKE_CURRENT_SOURCE_DIR}"
+    OUTPUT_VARIABLE GIT_DESCRIBE_TAGS ERROR_QUIET)
+  if(GIT_DESCRIBE_TAGS)
+    string(REGEX REPLACE "^v(.*)" "\\1" GIT_REVISION "${GIT_DESCRIBE_TAGS}")
+    string(STRIP "${GIT_REVISION}" GIT_REVISION)
+    if(GIT_REVISION)
+      set(PROJECT_VERSION_FULL "${GIT_REVISION}")
+    endif(GIT_REVISION)
+  endif(GIT_DESCRIBE_TAGS)
+endif(EXISTS "${CMAKE_CURRENT_SOURCE_DIR}/.git")

loader/tools/lzsa/src/libdivsufsort/examples/CMakeLists.txt

@@ -0,0 +1,11 @@
+## Add definitions ##
+add_definitions(-D_LARGEFILE_SOURCE -D_LARGE_FILES -D_FILE_OFFSET_BITS=64)
+
+## Targets ##
+include_directories("${CMAKE_CURRENT_SOURCE_DIR}/../include"
+                    "${CMAKE_CURRENT_BINARY_DIR}/../include")
+link_directories("${CMAKE_CURRENT_BINARY_DIR}/../lib")
+foreach(src suftest mksary sasearch bwt unbwt)
+  add_executable(${src} ${src}.c)
+  target_link_libraries(${src} divsufsort)
+endforeach(src)

loader/tools/lzsa/src/libdivsufsort/examples/bwt.c

@@ -0,0 +1,220 @@
+/*
+ * bwt.c for libdivsufsort
+ * Copyright (c) 2003-2008 Yuta Mori All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use,
+ * copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following
+ * conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+ * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#if HAVE_CONFIG_H
+# include "config.h"
+#endif
+#include <stdio.h>
+#if HAVE_STRING_H
+# include <string.h>
+#endif
+#if HAVE_STDLIB_H
+# include <stdlib.h>
+#endif
+#if HAVE_MEMORY_H
+# include <memory.h>
+#endif
+#if HAVE_STDDEF_H
+# include <stddef.h>
+#endif
+#if HAVE_STRINGS_H
+# include <strings.h>
+#endif
+#if HAVE_SYS_TYPES_H
+# include <sys/types.h>
+#endif
+#if HAVE_IO_H && HAVE_FCNTL_H
+# include <io.h>
+# include <fcntl.h>
+#endif
+#include <time.h>
+#include <divsufsort.h>
+#include "lfs.h"
+
+
+static
+size_t
+write_int(FILE *fp, saidx_t n) {
+  unsigned char c[4];
+  c[0] = (unsigned char)((n >>  0) & 0xff), c[1] = (unsigned char)((n >>  8) & 0xff),
+  c[2] = (unsigned char)((n >> 16) & 0xff), c[3] = (unsigned char)((n >> 24) & 0xff);
+  return fwrite(c, sizeof(unsigned char), 4, fp);
+}
+
+static
+void
+print_help(const char *progname, int status) {
+  fprintf(stderr,
+          "bwt, a burrows-wheeler transform program, version %s.\n",
+          divsufsort_version());
+  fprintf(stderr, "usage: %s [-b num] INFILE OUTFILE\n", progname);
+  fprintf(stderr, "  -b num    set block size to num MiB [1..512] (default: 32)\n\n");
+  exit(status);
+}
+
+int
+main(int argc, const char *argv[]) {
+  FILE *fp, *ofp;
+  const char *fname, *ofname;
+  sauchar_t *T;
+  saidx_t *SA;
+  LFS_OFF_T n;
+  size_t m;
+  saidx_t pidx;
+  clock_t start,finish;
+  saint_t i, blocksize = 32, needclose = 3;
+
+  /* Check arguments. */
+  if((argc == 1) ||
+     (strcmp(argv[1], "-h") == 0) ||
+     (strcmp(argv[1], "--help") == 0)) { print_help(argv[0], EXIT_SUCCESS); }
+  if((argc != 3) && (argc != 5)) { print_help(argv[0], EXIT_FAILURE); }
+  i = 1;
+  if(argc == 5) {
+    if(strcmp(argv[i], "-b") != 0) { print_help(argv[0], EXIT_FAILURE); }
+    blocksize = atoi(argv[i + 1]);
+    if(blocksize < 0) { blocksize = 1; }
+    else if(512 < blocksize) { blocksize = 512; }
+    i += 2;
+  }
+  blocksize <<= 20;
+
+  /* Open a file for reading. */
+  if(strcmp(argv[i], "-") != 0) {
+#if HAVE_FOPEN_S
+    if(fopen_s(&fp, fname = argv[i], "rb") != 0) {
+#else
+    if((fp = LFS_FOPEN(fname = argv[i], "rb")) == NULL) {
+#endif
+      fprintf(stderr, "%s: Cannot open file `%s': ", argv[0], fname);
+      perror(NULL);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+#if HAVE__SETMODE && HAVE__FILENO
+    if(_setmode(_fileno(stdin), _O_BINARY) == -1) {
+      fprintf(stderr, "%s: Cannot set mode: ", argv[0]);
+      perror(NULL);
+      exit(EXIT_FAILURE);
+    }
+#endif
+    fp = stdin;
+    fname = "stdin";
+    needclose ^= 1;
+  }
+  i += 1;
+
+  /* Open a file for writing. */
+  if(strcmp(argv[i], "-") != 0) {
+#if HAVE_FOPEN_S
+    if(fopen_s(&ofp, ofname = argv[i], "wb") != 0) {
+#else
+    if((ofp = LFS_FOPEN(ofname = argv[i], "wb")) == NULL) {
+#endif
+      fprintf(stderr, "%s: Cannot open file `%s': ", argv[0], ofname);
+      perror(NULL);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+#if HAVE__SETMODE && HAVE__FILENO
+    if(_setmode(_fileno(stdout), _O_BINARY) == -1) {
+      fprintf(stderr, "%s: Cannot set mode: ", argv[0]);
+      perror(NULL);
+      exit(EXIT_FAILURE);
+    }
+#endif
+    ofp = stdout;
+    ofname = "stdout";
+    needclose ^= 2;
+  }
+
+  /* Get the file size. */
+  if(LFS_FSEEK(fp, 0, SEEK_END) == 0) {
+    n = LFS_FTELL(fp);
+    rewind(fp);
+    if(n < 0) {
+      fprintf(stderr, "%s: Cannot ftell `%s': ", argv[0], fname);
+      perror(NULL);
+      exit(EXIT_FAILURE);
+    }
+    if(0x20000000L < n) { n = 0x20000000L; }
+    if((blocksize == 0) || (n < blocksize)) { blocksize = (saidx_t)n; }
+  } else if(blocksize == 0) { blocksize = 32 << 20; }
+
+  /* Allocate 5blocksize bytes of memory. */
+  T = (sauchar_t *)malloc(blocksize * sizeof(sauchar_t));
+  SA = (saidx_t *)malloc(blocksize * sizeof(saidx_t));
+  if((T == NULL) || (SA == NULL)) {
+    fprintf(stderr, "%s: Cannot allocate memory.\n", argv[0]);
+    exit(EXIT_FAILURE);
+  }
+
+  /* Write the blocksize. */
+  if(write_int(ofp, blocksize) != 4) {
+    fprintf(stderr, "%s: Cannot write to `%s': ", argv[0], ofname);
+    perror(NULL);
+    exit(EXIT_FAILURE);
+  }
+
+  fprintf(stderr, "  BWT (blocksize %" PRIdSAINT_T ") ... ", blocksize);
+  start = clock();
+  for(n = 0; 0 < (m = fread(T, sizeof(sauchar_t), blocksize, fp)); n += m) {
+    /* Burrows-Wheeler Transform. */
+    pidx = divbwt(T, T, SA, m);
+    if(pidx < 0) {
+      fprintf(stderr, "%s (bw_transform): %s.\n",
+        argv[0],
+        (pidx == -1) ? "Invalid arguments" : "Cannot allocate memory");
+      exit(EXIT_FAILURE);
+    }
+
+    /* Write the bwted data. */
+    if((write_int(ofp, pidx) != 4) ||
+       (fwrite(T, sizeof(sauchar_t), m, ofp) != m)) {
+      fprintf(stderr, "%s: Cannot write to `%s': ", argv[0], ofname);
+      perror(NULL);
+      exit(EXIT_FAILURE);
+    }
+  }
+  if(ferror(fp)) {
+    fprintf(stderr, "%s: Cannot read from `%s': ", argv[0], fname);
+    perror(NULL);
+    exit(EXIT_FAILURE);
+  }
+  finish = clock();
+  fprintf(stderr, "%" PRIdOFF_T " bytes: %.4f sec\n",
+    n, (double)(finish - start) / (double)CLOCKS_PER_SEC);
+
+  /* Close files */
+  if(needclose & 1) { fclose(fp); }
+  if(needclose & 2) { fclose(ofp); }
+
+  /* Deallocate memory. */
+  free(SA);
+  free(T);
+
+  return 0;
+}

loader/tools/lzsa/src/libdivsufsort/examples/mksary.c

@@ -0,0 +1,193 @@
+/*
+ * mksary.c for libdivsufsort
+ * Copyright (c) 2003-2008 Yuta Mori All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use,
+ * copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following
+ * conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+ * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#if HAVE_CONFIG_H
+# include "config.h"
+#endif
+#include <stdio.h>
+#if HAVE_STRING_H
+# include <string.h>
+#endif
+#if HAVE_STDLIB_H
+# include <stdlib.h>
+#endif
+#if HAVE_MEMORY_H
+# include <memory.h>
+#endif
+#if HAVE_STDDEF_H
+# include <stddef.h>
+#endif
+#if HAVE_STRINGS_H
+# include <strings.h>
+#endif
+#if HAVE_SYS_TYPES_H
+# include <sys/types.h>
+#endif
+#if HAVE_IO_H && HAVE_FCNTL_H
+# include <io.h>
+# include <fcntl.h>
+#endif
+#include <time.h>
+#include <divsufsort.h>
+#include "lfs.h"
+
+
+static
+void
+print_help(const char *progname, int status) {
+  fprintf(stderr,
+          "mksary, a simple suffix array builder, version %s.\n",
+          divsufsort_version());
+  fprintf(stderr, "usage: %s INFILE OUTFILE\n\n", progname);
+  exit(status);
+}
+
+int
+main(int argc, const char *argv[]) {
+  FILE *fp, *ofp;
+  const char *fname, *ofname;
+  sauchar_t *T;
+  saidx_t *SA;
+  LFS_OFF_T n;
+  clock_t start, finish;
+  saint_t needclose = 3;
+
+  /* Check arguments. */
+  if((argc == 1) ||
+     (strcmp(argv[1], "-h") == 0) ||
+     (strcmp(argv[1], "--help") == 0)) { print_help(argv[0], EXIT_SUCCESS); }
+  if(argc != 3) { print_help(argv[0], EXIT_FAILURE); }
+
+  /* Open a file for reading. */
+  if(strcmp(argv[1], "-") != 0) {
+#if HAVE_FOPEN_S
+    if(fopen_s(&fp, fname = argv[1], "rb") != 0) {
+#else
+    if((fp = LFS_FOPEN(fname = argv[1], "rb")) == NULL) {
+#endif
+      fprintf(stderr, "%s: Cannot open file `%s': ", argv[0], fname);
+      perror(NULL);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+#if HAVE__SETMODE && HAVE__FILENO
+    if(_setmode(_fileno(stdin), _O_BINARY) == -1) {
+      fprintf(stderr, "%s: Cannot set mode: ", argv[0]);
+      perror(NULL);
+      exit(EXIT_FAILURE);
+    }
+#endif
+    fp = stdin;
+    fname = "stdin";
+    needclose ^= 1;
+  }
+
+  /* Open a file for writing. */
+  if(strcmp(argv[2], "-") != 0) {
+#if HAVE_FOPEN_S
+    if(fopen_s(&ofp, ofname = argv[2], "wb") != 0) {
+#else
+    if((ofp = LFS_FOPEN(ofname = argv[2], "wb")) == NULL) {
+#endif
+      fprintf(stderr, "%s: Cannot open file `%s': ", argv[0], ofname);
+      perror(NULL);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+#if HAVE__SETMODE && HAVE__FILENO
+    if(_setmode(_fileno(stdout), _O_BINARY) == -1) {
+      fprintf(stderr, "%s: Cannot set mode: ", argv[0]);
+      perror(NULL);
+      exit(EXIT_FAILURE);
+    }
+#endif
+    ofp = stdout;
+    ofname = "stdout";
+    needclose ^= 2;
+  }
+
+  /* Get the file size. */
+  if(LFS_FSEEK(fp, 0, SEEK_END) == 0) {
+    n = LFS_FTELL(fp);
+    rewind(fp);
+    if(n < 0) {
+      fprintf(stderr, "%s: Cannot ftell `%s': ", argv[0], fname);
+      perror(NULL);
+      exit(EXIT_FAILURE);
+    }
+    if(0x7fffffff <= n) {
+      fprintf(stderr, "%s: Input file `%s' is too big.\n", argv[0], fname);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    fprintf(stderr, "%s: Cannot fseek `%s': ", argv[0], fname);
+    perror(NULL);
+    exit(EXIT_FAILURE);
+  }
+
+  /* Allocate 5blocksize bytes of memory. */
+  T = (sauchar_t *)malloc((size_t)n * sizeof(sauchar_t));
+  SA = (saidx_t *)malloc((size_t)n * sizeof(saidx_t));
+  if((T == NULL) || (SA == NULL)) {
+    fprintf(stderr, "%s: Cannot allocate memory.\n", argv[0]);
+    exit(EXIT_FAILURE);
+  }
+
+  /* Read n bytes of data. */
+  if(fread(T, sizeof(sauchar_t), (size_t)n, fp) != (size_t)n) {
+    fprintf(stderr, "%s: %s `%s': ",
+      argv[0],
+      (ferror(fp) || !feof(fp)) ? "Cannot read from" : "Unexpected EOF in",
+      fname);
+    perror(NULL);
+    exit(EXIT_FAILURE);
+  }
+  if(needclose & 1) { fclose(fp); }
+
+  /* Construct the suffix array. */
+  fprintf(stderr, "%s: %" PRIdOFF_T " bytes ... ", fname, n);
+  start = clock();
+  if(divsufsort(T, SA, (saidx_t)n) != 0) {
+    fprintf(stderr, "%s: Cannot allocate memory.\n", argv[0]);
+    exit(EXIT_FAILURE);
+  }
+  finish = clock();
+  fprintf(stderr, "%.4f sec\n", (double)(finish - start) / (double)CLOCKS_PER_SEC);
+
+  /* Write the suffix array. */
+  if(fwrite(SA, sizeof(saidx_t), (size_t)n, ofp) != (size_t)n) {
+    fprintf(stderr, "%s: Cannot write to `%s': ", argv[0], ofname);
+    perror(NULL);
+    exit(EXIT_FAILURE);
+  }
+  if(needclose & 2) { fclose(ofp); }
+
+  /* Deallocate memory. */
+  free(SA);
+  free(T);
+
+  return 0;
+}

loader/tools/lzsa/src/libdivsufsort/examples/sasearch.c

@@ -0,0 +1,165 @@
+/*
+ * sasearch.c for libdivsufsort
+ * Copyright (c) 2003-2008 Yuta Mori All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use,
+ * copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following
+ * conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+ * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#if HAVE_CONFIG_H
+# include "config.h"
+#endif
+#include <stdio.h>
+#if HAVE_STRING_H
+# include <string.h>
+#endif
+#if HAVE_STDLIB_H
+# include <stdlib.h>
+#endif
+#if HAVE_MEMORY_H
+# include <memory.h>
+#endif
+#if HAVE_STDDEF_H
+# include <stddef.h>
+#endif
+#if HAVE_STRINGS_H
+# include <strings.h>
+#endif
+#if HAVE_SYS_TYPES_H
+# include <sys/types.h>
+#endif
+#if HAVE_IO_H && HAVE_FCNTL_H
+# include <io.h>
+# include <fcntl.h>
+#endif
+#include <divsufsort.h>
+#include "lfs.h"
+
+
+static
+void
+print_help(const char *progname, int status) {
+  fprintf(stderr,
+          "sasearch, a simple SA-based full-text search tool, version %s\n",
+          divsufsort_version());
+  fprintf(stderr, "usage: %s PATTERN FILE SAFILE\n\n", progname);
+  exit(status);
+}
+
+int
+main(int argc, const char *argv[]) {
+  FILE *fp;
+  const char *P;
+  sauchar_t *T;
+  saidx_t *SA;
+  LFS_OFF_T n;
+  size_t Psize;
+  saidx_t i, size, left;
+
+  if((argc == 1) ||
+     (strcmp(argv[1], "-h") == 0) ||
+     (strcmp(argv[1], "--help") == 0)) { print_help(argv[0], EXIT_SUCCESS); }
+  if(argc != 4) { print_help(argv[0], EXIT_FAILURE); }
+
+  P = argv[1];
+  Psize = strlen(P);
+
+  /* Open a file for reading. */
+#if HAVE_FOPEN_S
+  if(fopen_s(&fp, argv[2], "rb") != 0) {
+#else
+  if((fp = LFS_FOPEN(argv[2], "rb")) == NULL) {
+#endif
+    fprintf(stderr, "%s: Cannot open file `%s': ", argv[0], argv[2]);
+    perror(NULL);
+    exit(EXIT_FAILURE);
+  }
+
+  /* Get the file size. */
+  if(LFS_FSEEK(fp, 0, SEEK_END) == 0) {
+    n = LFS_FTELL(fp);
+    rewind(fp);
+    if(n < 0) {
+      fprintf(stderr, "%s: Cannot ftell `%s': ", argv[0], argv[2]);
+      perror(NULL);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    fprintf(stderr, "%s: Cannot fseek `%s': ", argv[0], argv[2]);
+    perror(NULL);
+    exit(EXIT_FAILURE);
+  }
+
+  /* Allocate 5n bytes of memory. */
+  T = (sauchar_t *)malloc((size_t)n * sizeof(sauchar_t));
+  SA = (saidx_t *)malloc((size_t)n * sizeof(saidx_t));
+  if((T == NULL) || (SA == NULL)) {
+    fprintf(stderr, "%s: Cannot allocate memory.\n", argv[0]);
+    exit(EXIT_FAILURE);
+  }
+
+  /* Read n bytes of data. */
+  if(fread(T, sizeof(sauchar_t), (size_t)n, fp) != (size_t)n) {
+    fprintf(stderr, "%s: %s `%s': ",
+      argv[0],
+      (ferror(fp) || !feof(fp)) ? "Cannot read from" : "Unexpected EOF in",
+      argv[2]);
+    perror(NULL);
+    exit(EXIT_FAILURE);
+  }
+  fclose(fp);
+
+  /* Open the SA file for reading. */
+#if HAVE_FOPEN_S
+  if(fopen_s(&fp, argv[3], "rb") != 0) {
+#else
+  if((fp = LFS_FOPEN(argv[3], "rb")) == NULL) {
+#endif
+    fprintf(stderr, "%s: Cannot open file `%s': ", argv[0], argv[3]);
+    perror(NULL);
+    exit(EXIT_FAILURE);
+  }
+
+  /* Read n * sizeof(saidx_t) bytes of data. */
+  if(fread(SA, sizeof(saidx_t), (size_t)n, fp) != (size_t)n) {
+    fprintf(stderr, "%s: %s `%s': ",
+      argv[0],
+      (ferror(fp) || !feof(fp)) ? "Cannot read from" : "Unexpected EOF in",
+      argv[3]);
+    perror(NULL);
+    exit(EXIT_FAILURE);
+  }
+  fclose(fp);
+
+  /* Search and print */
+  size = sa_search(T, (saidx_t)n,
+                   (const sauchar_t *)P, (saidx_t)Psize,
+                   SA, (saidx_t)n, &left);
+  for(i = 0; i < size; ++i) {
+    fprintf(stdout, "%" PRIdSAIDX_T "\n", SA[left + i]);
+  }
+
+  /* Deallocate memory. */
+  free(SA);
+  free(T);
+
+  return 0;
+}

loader/tools/lzsa/src/libdivsufsort/examples/suftest.c

@@ -0,0 +1,164 @@
+/*
+ * suftest.c for libdivsufsort
+ * Copyright (c) 2003-2008 Yuta Mori All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use,
+ * copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following
+ * conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+ * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#if HAVE_CONFIG_H
+# include "config.h"
+#endif
+#include <stdio.h>
+#if HAVE_STRING_H
+# include <string.h>
+#endif
+#if HAVE_STDLIB_H
+# include <stdlib.h>
+#endif
+#if HAVE_MEMORY_H
+# include <memory.h>
+#endif
+#if HAVE_STDDEF_H
+# include <stddef.h>
+#endif
+#if HAVE_STRINGS_H
+# include <strings.h>
+#endif
+#if HAVE_SYS_TYPES_H
+# include <sys/types.h>
+#endif
+#if HAVE_IO_H && HAVE_FCNTL_H
+# include <io.h>
+# include <fcntl.h>
+#endif
+#include <time.h>
+#include <divsufsort.h>
+#include "lfs.h"
+
+
+static
+void
+print_help(const char *progname, int status) {
+  fprintf(stderr,
+          "suftest, a suffixsort tester, version %s.\n",
+          divsufsort_version());
+  fprintf(stderr, "usage: %s FILE\n\n", progname);
+  exit(status);
+}
+
+int
+main(int argc, const char *argv[]) {
+  FILE *fp;
+  const char *fname;
+  sauchar_t *T;
+  saidx_t *SA;
+  LFS_OFF_T n;
+  clock_t start, finish;
+  saint_t needclose = 1;
+
+  /* Check arguments. */
+  if((argc == 1) ||
+     (strcmp(argv[1], "-h") == 0) ||
+     (strcmp(argv[1], "--help") == 0)) { print_help(argv[0], EXIT_SUCCESS); }
+  if(argc != 2) { print_help(argv[0], EXIT_FAILURE); }
+
+  /* Open a file for reading. */
+  if(strcmp(argv[1], "-") != 0) {
+#if HAVE_FOPEN_S
+    if(fopen_s(&fp, fname = argv[1], "rb") != 0) {
+#else
+    if((fp = LFS_FOPEN(fname = argv[1], "rb")) == NULL) {
+#endif
+      fprintf(stderr, "%s: Cannot open file `%s': ", argv[0], fname);
+      perror(NULL);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+#if HAVE__SETMODE && HAVE__FILENO
+    if(_setmode(_fileno(stdin), _O_BINARY) == -1) {
+      fprintf(stderr, "%s: Cannot set mode: ", argv[0]);
+      perror(NULL);
+      exit(EXIT_FAILURE);
+    }
+#endif
+    fp = stdin;
+    fname = "stdin";
+    needclose = 0;
+  }
+
+  /* Get the file size. */
+  if(LFS_FSEEK(fp, 0, SEEK_END) == 0) {
+    n = LFS_FTELL(fp);
+    rewind(fp);
+    if(n < 0) {
+      fprintf(stderr, "%s: Cannot ftell `%s': ", argv[0], fname);
+      perror(NULL);
+      exit(EXIT_FAILURE);
+    }
+    if(0x7fffffff <= n) {
+      fprintf(stderr, "%s: Input file `%s' is too big.\n", argv[0], fname);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    fprintf(stderr, "%s: Cannot fseek `%s': ", argv[0], fname);
+    perror(NULL);
+    exit(EXIT_FAILURE);
+  }
+
+  /* Allocate 5n bytes of memory. */
+  T = (sauchar_t *)malloc((size_t)n * sizeof(sauchar_t));
+  SA = (saidx_t *)malloc((size_t)n * sizeof(saidx_t));
+  if((T == NULL) || (SA == NULL)) {
+    fprintf(stderr, "%s: Cannot allocate memory.\n", argv[0]);
+    exit(EXIT_FAILURE);
+  }
+
+  /* Read n bytes of data. */
+  if(fread(T, sizeof(sauchar_t), (size_t)n, fp) != (size_t)n) {
+    fprintf(stderr, "%s: %s `%s': ",
+      argv[0],
+      (ferror(fp) || !feof(fp)) ? "Cannot read from" : "Unexpected EOF in",
+      argv[1]);
+    perror(NULL);
+    exit(EXIT_FAILURE);
+  }
+  if(needclose & 1) { fclose(fp); }
+
+  /* Construct the suffix array. */
+  fprintf(stderr, "%s: %" PRIdOFF_T " bytes ... ", fname, n);
+  start = clock();
+  if(divsufsort(T, SA, (saidx_t)n) != 0) {
+    fprintf(stderr, "%s: Cannot allocate memory.\n", argv[0]);
+    exit(EXIT_FAILURE);
+  }
+  finish = clock();
+  fprintf(stderr, "%.4f sec\n", (double)(finish - start) / (double)CLOCKS_PER_SEC);
+
+  /* Check the suffix array. */
+  if(sufcheck(T, SA, (saidx_t)n, 1) != 0) { exit(EXIT_FAILURE); }
+
+  /* Deallocate memory. */
+  free(SA);
+  free(T);
+
+  return 0;
+}

loader/tools/lzsa/src/libdivsufsort/examples/unbwt.c

@@ -0,0 +1,207 @@
+/*
+ * unbwt.c for libdivsufsort
+ * Copyright (c) 2003-2008 Yuta Mori All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use,
+ * copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following
+ * conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+ * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#if HAVE_CONFIG_H
+# include "config.h"
+#endif
+#include <stdio.h>
+#if HAVE_STRING_H
+# include <string.h>
+#endif
+#if HAVE_STDLIB_H
+# include <stdlib.h>
+#endif
+#if HAVE_MEMORY_H
+# include <memory.h>
+#endif
+#if HAVE_STDDEF_H
+# include <stddef.h>
+#endif
+#if HAVE_STRINGS_H
+# include <strings.h>
+#endif
+#if HAVE_SYS_TYPES_H
+# include <sys/types.h>
+#endif
+#if HAVE_IO_H && HAVE_FCNTL_H
+# include <io.h>
+# include <fcntl.h>
+#endif
+#include <time.h>
+#include <divsufsort.h>
+#include "lfs.h"
+
+
+static
+size_t
+read_int(FILE *fp, saidx_t *n) {
+  unsigned char c[4];
+  size_t m = fread(c, sizeof(unsigned char), 4, fp);
+  if(m == 4) {
+    *n = (c[0] <<  0) | (c[1] <<  8) |
+         (c[2] << 16) | (c[3] << 24);
+  }
+  return m;
+}
+
+static
+void
+print_help(const char *progname, int status) {
+  fprintf(stderr,
+          "unbwt, an inverse burrows-wheeler transform program, version %s.\n",
+          divsufsort_version());
+  fprintf(stderr, "usage: %s INFILE OUTFILE\n\n", progname);
+  exit(status);
+}
+
+int
+main(int argc, const char *argv[]) {
+  FILE *fp, *ofp;
+  const char *fname, *ofname;
+  sauchar_t *T;
+  saidx_t *A;
+  LFS_OFF_T n;
+  size_t m;
+  saidx_t pidx;
+  clock_t start, finish;
+  saint_t err, blocksize, needclose = 3;
+
+  /* Check arguments. */
+  if((argc == 1) ||
+     (strcmp(argv[1], "-h") == 0) ||
+     (strcmp(argv[1], "--help") == 0)) { print_help(argv[0], EXIT_SUCCESS); }
+  if(argc != 3) { print_help(argv[0], EXIT_FAILURE); }
+
+  /* Open a file for reading. */
+  if(strcmp(argv[1], "-") != 0) {
+#if HAVE_FOPEN_S
+    if(fopen_s(&fp, fname = argv[1], "rb") != 0) {
+#else
+    if((fp = LFS_FOPEN(fname = argv[1], "rb")) == NULL) {
+#endif
+      fprintf(stderr, "%s: Cannot open file `%s': ", argv[0], fname);
+      perror(NULL);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+#if HAVE__SETMODE && HAVE__FILENO
+    if(_setmode(_fileno(stdin), _O_BINARY) == -1) {
+      fprintf(stderr, "%s: Cannot set mode: ", argv[0]);
+      perror(NULL);
+      exit(EXIT_FAILURE);
+    }
+#endif
+    fp = stdin;
+    fname = "stdin";
+    needclose ^= 1;
+  }
+
+  /* Open a file for writing. */
+  if(strcmp(argv[2], "-") != 0) {
+#if HAVE_FOPEN_S
+    if(fopen_s(&ofp, ofname = argv[2], "wb") != 0) {
+#else
+    if((ofp = LFS_FOPEN(ofname = argv[2], "wb")) == NULL) {
+#endif
+      fprintf(stderr, "%s: Cannot open file `%s': ", argv[0], ofname);
+      perror(NULL);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+#if HAVE__SETMODE && HAVE__FILENO
+    if(_setmode(_fileno(stdout), _O_BINARY) == -1) {
+      fprintf(stderr, "%s: Cannot set mode: ", argv[0]);
+      perror(NULL);
+      exit(EXIT_FAILURE);
+    }
+#endif
+    ofp = stdout;
+    ofname = "stdout";
+    needclose ^= 2;
+  }
+
+  /* Read the blocksize. */
+  if(read_int(fp, &blocksize) != 4) {
+    fprintf(stderr, "%s: Cannot read from `%s': ", argv[0], fname);
+    perror(NULL);
+    exit(EXIT_FAILURE);
+  }
+
+  /* Allocate 5blocksize bytes of memory. */
+  T = (sauchar_t *)malloc(blocksize * sizeof(sauchar_t));
+  A = (saidx_t *)malloc(blocksize * sizeof(saidx_t));
+  if((T == NULL) || (A == NULL)) {
+    fprintf(stderr, "%s: Cannot allocate memory.\n", argv[0]);
+    exit(EXIT_FAILURE);
+  }
+
+  fprintf(stderr, "UnBWT (blocksize %" PRIdSAINT_T ") ... ", blocksize);
+  start = clock();
+  for(n = 0; (m = read_int(fp, &pidx)) != 0; n += m) {
+    /* Read blocksize bytes of data. */
+    if((m != 4) || ((m = fread(T, sizeof(sauchar_t), blocksize, fp)) == 0)) {
+      fprintf(stderr, "%s: %s `%s': ",
+        argv[0],
+        (ferror(fp) || !feof(fp)) ? "Cannot read from" : "Unexpected EOF in",
+        fname);
+      perror(NULL);
+      exit(EXIT_FAILURE);
+    }
+
+    /* Inverse Burrows-Wheeler Transform. */
+    if((err = inverse_bw_transform(T, T, A, m, pidx)) != 0) {
+      fprintf(stderr, "%s (reverseBWT): %s.\n",
+        argv[0],
+        (err == -1) ? "Invalid data" : "Cannot allocate memory");
+      exit(EXIT_FAILURE);
+    }
+
+    /* Write m bytes of data. */
+    if(fwrite(T, sizeof(sauchar_t), m, ofp) != m) {
+      fprintf(stderr, "%s: Cannot write to `%s': ", argv[0], ofname);
+      perror(NULL);
+      exit(EXIT_FAILURE);
+    }
+  }
+  if(ferror(fp)) {
+    fprintf(stderr, "%s: Cannot read from `%s': ", argv[0], fname);
+    perror(NULL);
+    exit(EXIT_FAILURE);
+  }
+  finish = clock();
+  fprintf(stderr, "%" PRIdOFF_T " bytes: %.4f sec\n",
+    n, (double)(finish - start) / (double)CLOCKS_PER_SEC);
+
+  /* Close files */
+  if(needclose & 1) { fclose(fp); }
+  if(needclose & 2) { fclose(ofp); }
+
+  /* Deallocate memory. */
+  free(A);
+  free(T);
+
+  return 0;
+}

loader/tools/lzsa/src/libdivsufsort/include/CMakeLists.txt

@@ -0,0 +1,162 @@
+include(CheckIncludeFiles)
+include(CheckIncludeFile)
+include(CheckSymbolExists)
+include(CheckTypeSize)
+include(CheckFunctionKeywords)
+include(CheckLFS)
+
+## Checks for header files ##
+check_include_file("inttypes.h" HAVE_INTTYPES_H)
+check_include_file("memory.h" HAVE_MEMORY_H)
+check_include_file("stddef.h" HAVE_STDDEF_H)
+check_include_file("stdint.h" HAVE_STDINT_H)
+check_include_file("stdlib.h" HAVE_STDLIB_H)
+check_include_file("string.h" HAVE_STRING_H)
+check_include_file("strings.h" HAVE_STRINGS_H)
+check_include_file("sys/types.h" HAVE_SYS_TYPES_H)
+if(HAVE_INTTYPES_H)
+  set(INCFILE "#include <inttypes.h>")
+elseif(HAVE_STDINT_H)
+  set(INCFILE "#include <stdint.h>")
+else(HAVE_INTTYPES_H)
+  set(INCFILE "")
+endif(HAVE_INTTYPES_H)
+
+## create configuration files from .cmake file ##
+if(BUILD_EXAMPLES)
+  ## Checks for WinIO ##
+  if(WIN32)
+    check_include_file("io.h" HAVE_IO_H)
+    check_include_file("fcntl.h" HAVE_FCNTL_H)
+    check_symbol_exists("_setmode" "io.h;fcntl.h" HAVE__SETMODE)
+    if(NOT HAVE__SETMODE)
+      check_symbol_exists("setmode" "io.h;fcntl.h" HAVE_SETMODE)
+    endif(NOT HAVE__SETMODE)
+    check_symbol_exists("_fileno" "stdio.h" HAVE__FILENO)
+    check_symbol_exists("fopen_s" "stdio.h" HAVE_FOPEN_S)
+    check_symbol_exists("_O_BINARY" "fcntl.h" HAVE__O_BINARY)
+  endif(WIN32)
+
+  ## Checks for large file support ##
+  check_lfs(WITH_LFS)
+  configure_file("${CMAKE_CURRENT_SOURCE_DIR}/lfs.h.cmake" "${CMAKE_CURRENT_BINARY_DIR}/lfs.h" @ONLY)
+endif(BUILD_EXAMPLES)
+
+## generate config.h ##
+check_function_keywords("inline;__inline;__inline__;__declspec(dllexport);__declspec(dllimport)")
+if(HAVE_INLINE)
+  set(INLINE "inline")
+elseif(HAVE___INLINE)
+  set(INLINE "__inline")
+elseif(HAVE___INLINE__)
+  set(INLINE "__inline__")
+else(HAVE_INLINE)
+  set(INLINE "")
+endif(HAVE_INLINE)
+configure_file("${CMAKE_CURRENT_SOURCE_DIR}/config.h.cmake" "${CMAKE_CURRENT_BINARY_DIR}/config.h")
+
+## Checks for types ##
+# sauchar_t (8bit)
+check_type_size("uint8_t" UINT8_T)
+if(HAVE_UINT8_T)
+  set(SAUCHAR_TYPE "uint8_t")
+else(HAVE_UINT8_T)
+  check_type_size("unsigned char" SIZEOF_UNSIGNED_CHAR)
+  if("${SIZEOF_UNSIGNED_CHAR}" STREQUAL "1")
+    set(SAUCHAR_TYPE "unsigned char")
+  else("${SIZEOF_UNSIGNED_CHAR}" STREQUAL "1")
+    message(FATAL_ERROR "Cannot find unsigned 8-bit integer type")
+  endif("${SIZEOF_UNSIGNED_CHAR}" STREQUAL "1")
+endif(HAVE_UINT8_T)
+# saint_t (32bit)
+check_type_size("int32_t" INT32_T)
+if(HAVE_INT32_T)
+  set(SAINT32_TYPE "int32_t")
+  check_symbol_exists("PRId32" "inttypes.h" HAVE_PRID32)
+  if(HAVE_PRID32)
+    set(SAINT32_PRId "PRId32")
+  else(HAVE_PRID32)
+    set(SAINT32_PRId "\"d\"")
+  endif(HAVE_PRID32)
+else(HAVE_INT32_T)
+  check_type_size("int" SIZEOF_INT)
+  check_type_size("long" SIZEOF_LONG)
+  check_type_size("short" SIZEOF_SHORT)
+  check_type_size("__int32" SIZEOF___INT32)
+  if("${SIZEOF_INT}" STREQUAL "4")
+    set(SAINT32_TYPE "int")
+    set(SAINT32_PRId "\"d\"")
+  elseif("${SIZEOF_LONG}" STREQUAL "4")
+    set(SAINT32_TYPE "long")
+    set(SAINT32_PRId "\"ld\"")
+  elseif("${SIZEOF_SHORT}" STREQUAL "4")
+    set(SAINT32_TYPE "short")
+    set(SAINT32_PRId "\"d\"")
+  elseif("${SIZEOF___INT32}" STREQUAL "4")
+    set(SAINT32_TYPE "__int32")
+    set(SAINT32_PRId "\"d\"")
+  else("${SIZEOF_INT}" STREQUAL "4")
+    message(FATAL_ERROR "Cannot find 32-bit integer type")
+  endif("${SIZEOF_INT}" STREQUAL "4")
+endif(HAVE_INT32_T)
+# saint64_t (64bit)
+if(BUILD_DIVSUFSORT64)
+  check_type_size("int64_t" INT64_T)
+  if(HAVE_INT64_T)
+    set(SAINT64_TYPE "int64_t")
+    check_symbol_exists("PRId64" "inttypes.h" HAVE_PRID64)
+    if(HAVE_PRID64)
+      set(SAINT64_PRId "PRId64")
+    else(HAVE_PRID64)
+      set(SAINT64_PRId "\"lld\"")
+    endif(HAVE_PRID64)
+  else(HAVE_INT64_T)
+    check_type_size("int" SIZEOF_INT)
+    check_type_size("long" SIZEOF_LONG)
+    check_type_size("long long" SIZEOF_LONG_LONG)
+    check_type_size("__int64" SIZEOF___INT64)
+    if("${SIZEOF_INT}" STREQUAL "8")
+      set(SAINT64_TYPE "int")
+      set(SAINT64_PRId "\"d\"")
+    elseif("${SIZEOF_LONG}" STREQUAL "8")
+      set(SAINT64_TYPE "long")
+      set(SAINT64_PRId "\"ld\"")
+    elseif("${SIZEOF_LONG_LONG}" STREQUAL "8")
+      set(SAINT64_TYPE "long long")
+      set(SAINT64_PRId "\"lld\"")
+    elseif("${SIZEOF___INT64}" STREQUAL "8")
+      set(SAINT64_TYPE "__int64")
+      set(SAINT64_PRId "\"I64d\"")
+    else("${SIZEOF_INT}" STREQUAL "8")
+      message(SEND_ERROR "Cannot find 64-bit integer type")
+      set(BUILD_DIVSUFSORT64 OFF)
+    endif("${SIZEOF_INT}" STREQUAL "8")
+  endif(HAVE_INT64_T)
+endif(BUILD_DIVSUFSORT64)
+
+## generate divsufsort.h ##
+set(DIVSUFSORT_IMPORT "")
+set(DIVSUFSORT_EXPORT "")
+if(BUILD_SHARED_LIBS)
+  if(HAVE___DECLSPEC_DLLIMPORT_)
+    set(DIVSUFSORT_IMPORT "__declspec(dllimport)")
+  endif(HAVE___DECLSPEC_DLLIMPORT_)
+  if(HAVE___DECLSPEC_DLLEXPORT_)
+    set(DIVSUFSORT_EXPORT "__declspec(dllexport)")
+  endif(HAVE___DECLSPEC_DLLEXPORT_)
+endif(BUILD_SHARED_LIBS)
+set(W64BIT "")
+set(SAINDEX_TYPE "${SAINT32_TYPE}")
+set(SAINDEX_PRId "${SAINT32_PRId}")
+set(SAINT_PRId "${SAINT32_PRId}")
+configure_file("${CMAKE_CURRENT_SOURCE_DIR}/divsufsort.h.cmake"
+               "${CMAKE_CURRENT_BINARY_DIR}/divsufsort.h" @ONLY)
+install(FILES "${CMAKE_CURRENT_BINARY_DIR}/divsufsort.h" DESTINATION ${CMAKE_INSTALL_INCLUDEDIR})
+if(BUILD_DIVSUFSORT64)
+  set(W64BIT "64")
+  set(SAINDEX_TYPE "${SAINT64_TYPE}")
+  set(SAINDEX_PRId "${SAINT64_PRId}")
+  configure_file("${CMAKE_CURRENT_SOURCE_DIR}/divsufsort.h.cmake"
+                 "${CMAKE_CURRENT_BINARY_DIR}/divsufsort64.h" @ONLY)
+  install(FILES "${CMAKE_CURRENT_BINARY_DIR}/divsufsort64.h" DESTINATION ${CMAKE_INSTALL_INCLUDEDIR})
+endif(BUILD_DIVSUFSORT64)

loader/tools/lzsa/src/libdivsufsort/include/config.h.cmake

@@ -0,0 +1,81 @@
+/*
+ * config.h for libdivsufsort
+ * Copyright (c) 2003-2008 Yuta Mori All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use,
+ * copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following
+ * conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+ * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _CONFIG_H
+#define _CONFIG_H 1
+
+#ifdef __cplusplus
+extern "C" {
+#endif /* __cplusplus */
+
+/** Define to the version of this package. **/
+#cmakedefine PROJECT_VERSION_FULL "${PROJECT_VERSION_FULL}"
+
+/** Define to 1 if you have the header files. **/
+#cmakedefine HAVE_INTTYPES_H 1
+#cmakedefine HAVE_STDDEF_H 1
+#cmakedefine HAVE_STDINT_H 1
+#cmakedefine HAVE_STDLIB_H 1
+#cmakedefine HAVE_STRING_H 1
+#cmakedefine HAVE_STRINGS_H 1
+#cmakedefine HAVE_MEMORY_H 1
+#cmakedefine HAVE_SYS_TYPES_H 1
+
+/** for WinIO **/
+#cmakedefine HAVE_IO_H 1
+#cmakedefine HAVE_FCNTL_H 1
+#cmakedefine HAVE__SETMODE 1
+#cmakedefine HAVE_SETMODE 1
+#cmakedefine HAVE__FILENO 1
+#cmakedefine HAVE_FOPEN_S 1
+#cmakedefine HAVE__O_BINARY 1
+#ifndef HAVE__SETMODE
+# if HAVE_SETMODE
+#  define _setmode setmode
+#  define HAVE__SETMODE 1
+# endif
+# if HAVE__SETMODE && !HAVE__O_BINARY
+#  define _O_BINARY 0
+#  define HAVE__O_BINARY 1
+# endif
+#endif
+
+/** for inline **/
+#ifndef INLINE
+# define INLINE @INLINE@
+#endif
+
+/** for VC++ warning **/
+#ifdef _MSC_VER
+#pragma warning(disable: 4127)
+#endif
+
+
+#ifdef __cplusplus
+} /* extern "C" */
+#endif /* __cplusplus */
+
+#endif /* _CONFIG_H */

loader/tools/lzsa/src/libdivsufsort/include/divsufsort.h

@@ -0,0 +1,189 @@
+/*
+ * divsufsort.h for libdivsufsort
+ * Copyright (c) 2003-2008 Yuta Mori All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use,
+ * copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following
+ * conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+ * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _DIVSUFSORT_H
+#define _DIVSUFSORT_H 1
+
+#ifdef __cplusplus
+extern "C" {
+#endif /* __cplusplus */
+
+#define DIVSUFSORT_API
+
+/*- Datatypes -*/
+#ifndef SAUCHAR_T
+#define SAUCHAR_T
+typedef unsigned char sauchar_t;
+#endif /* SAUCHAR_T */
+#ifndef SAINT_T
+#define SAINT_T
+typedef int saint_t;
+#endif /* SAINT_T */
+#ifndef SAIDX_T
+#define SAIDX_T
+typedef int saidx_t;
+#endif /* SAIDX_T */
+#ifndef PRIdSAIDX_T
+#define PRIdSAIDX_T "d"
+#endif
+
+/*- divsufsort context */
+typedef struct _divsufsort_ctx_t {
+   saidx_t *bucket_A;
+   saidx_t *bucket_B;
+} divsufsort_ctx_t;
+
+/*- Prototypes -*/
+
+/**
+ * Initialize suffix array context
+ *
+ * @return 0 for success, or non-zero in case of an error
+ */
+int divsufsort_init(divsufsort_ctx_t *ctx);
+
+/**
+ * Destroy suffix array context
+ *
+ * @param ctx suffix array context to destroy
+ */
+void divsufsort_destroy(divsufsort_ctx_t *ctx);
+
+/**
+ * Constructs the suffix array of a given string.
+ * @param ctx suffix array context
+ * @param T[0..n-1] The input string.
+ * @param SA[0..n-1] The output array of suffixes.
+ * @param n The length of the given string.
+ * @return 0 if no error occurred, -1 or -2 otherwise.
+ */
+DIVSUFSORT_API
+saint_t divsufsort_build_array(divsufsort_ctx_t *ctx, const sauchar_t *T, saidx_t *SA, saidx_t n);
+
+#if 0
+/**
+ * Constructs the burrows-wheeler transformed string of a given string.
+ * @param T[0..n-1] The input string.
+ * @param U[0..n-1] The output string. (can be T)
+ * @param A[0..n-1] The temporary array. (can be NULL)
+ * @param n The length of the given string.
+ * @return The primary index if no error occurred, -1 or -2 otherwise.
+ */
+DIVSUFSORT_API
+saidx_t
+divbwt(const sauchar_t *T, sauchar_t *U, saidx_t *A, saidx_t n);
+
+/**
+ * Returns the version of the divsufsort library.
+ * @return The version number string.
+ */
+DIVSUFSORT_API
+const char *
+divsufsort_version(void);
+
+
+/**
+ * Constructs the burrows-wheeler transformed string of a given string and suffix array.
+ * @param T[0..n-1] The input string.
+ * @param U[0..n-1] The output string. (can be T)
+ * @param SA[0..n-1] The suffix array. (can be NULL)
+ * @param n The length of the given string.
+ * @param idx The output primary index.
+ * @return 0 if no error occurred, -1 or -2 otherwise.
+ */
+DIVSUFSORT_API
+saint_t
+bw_transform(const sauchar_t *T, sauchar_t *U,
+             saidx_t *SA /* can NULL */,
+             saidx_t n, saidx_t *idx);
+
+/**
+ * Inverse BW-transforms a given BWTed string.
+ * @param T[0..n-1] The input string.
+ * @param U[0..n-1] The output string. (can be T)
+ * @param A[0..n-1] The temporary array. (can be NULL)
+ * @param n The length of the given string.
+ * @param idx The primary index.
+ * @return 0 if no error occurred, -1 or -2 otherwise.
+ */
+DIVSUFSORT_API
+saint_t
+inverse_bw_transform(const sauchar_t *T, sauchar_t *U,
+                     saidx_t *A /* can NULL */,
+                     saidx_t n, saidx_t idx);
+
+/**
+ * Checks the correctness of a given suffix array.
+ * @param T[0..n-1] The input string.
+ * @param SA[0..n-1] The input suffix array.
+ * @param n The length of the given string.
+ * @param verbose The verbose mode.
+ * @return 0 if no error occurred.
+ */
+DIVSUFSORT_API
+saint_t
+sufcheck(const sauchar_t *T, const saidx_t *SA, saidx_t n, saint_t verbose);
+
+/**
+ * Search for the pattern P in the string T.
+ * @param T[0..Tsize-1] The input string.
+ * @param Tsize The length of the given string.
+ * @param P[0..Psize-1] The input pattern string.
+ * @param Psize The length of the given pattern string.
+ * @param SA[0..SAsize-1] The input suffix array.
+ * @param SAsize The length of the given suffix array.
+ * @param idx The output index.
+ * @return The count of matches if no error occurred, -1 otherwise.
+ */
+DIVSUFSORT_API
+saidx_t
+sa_search(const sauchar_t *T, saidx_t Tsize,
+          const sauchar_t *P, saidx_t Psize,
+          const saidx_t *SA, saidx_t SAsize,
+          saidx_t *left);
+
+/**
+ * Search for the character c in the string T.
+ * @param T[0..Tsize-1] The input string.
+ * @param Tsize The length of the given string.
+ * @param SA[0..SAsize-1] The input suffix array.
+ * @param SAsize The length of the given suffix array.
+ * @param c The input character.
+ * @param idx The output index.
+ * @return The count of matches if no error occurred, -1 otherwise.
+ */
+DIVSUFSORT_API
+saidx_t
+sa_simplesearch(const sauchar_t *T, saidx_t Tsize,
+                const saidx_t *SA, saidx_t SAsize,
+                saint_t c, saidx_t *left);
+#endif
+
+#ifdef __cplusplus
+} /* extern "C" */
+#endif /* __cplusplus */
+
+#endif /* _DIVSUFSORT_H */

loader/tools/lzsa/src/libdivsufsort/include/divsufsort.h.cmake

@@ -0,0 +1,180 @@
+/*
+ * divsufsort@W64BIT@.h for libdivsufsort@W64BIT@
+ * Copyright (c) 2003-2008 Yuta Mori All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use,
+ * copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following
+ * conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+ * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _DIVSUFSORT@W64BIT@_H
+#define _DIVSUFSORT@W64BIT@_H 1
+
+#ifdef __cplusplus
+extern "C" {
+#endif /* __cplusplus */
+
+@INCFILE@
+
+#ifndef DIVSUFSORT_API
+# ifdef DIVSUFSORT_BUILD_DLL
+#  define DIVSUFSORT_API @DIVSUFSORT_EXPORT@
+# else
+#  define DIVSUFSORT_API @DIVSUFSORT_IMPORT@
+# endif
+#endif
+
+/*- Datatypes -*/
+#ifndef SAUCHAR_T
+#define SAUCHAR_T
+typedef @SAUCHAR_TYPE@ sauchar_t;
+#endif /* SAUCHAR_T */
+#ifndef SAINT_T
+#define SAINT_T
+typedef @SAINT32_TYPE@ saint_t;
+#endif /* SAINT_T */
+#ifndef SAIDX@W64BIT@_T
+#define SAIDX@W64BIT@_T
+typedef @SAINDEX_TYPE@ saidx@W64BIT@_t;
+#endif /* SAIDX@W64BIT@_T */
+#ifndef PRIdSAINT_T
+#define PRIdSAINT_T @SAINT_PRId@
+#endif /* PRIdSAINT_T */
+#ifndef PRIdSAIDX@W64BIT@_T
+#define PRIdSAIDX@W64BIT@_T @SAINDEX_PRId@
+#endif /* PRIdSAIDX@W64BIT@_T */
+
+
+/*- Prototypes -*/
+
+/**
+ * Constructs the suffix array of a given string.
+ * @param T[0..n-1] The input string.
+ * @param SA[0..n-1] The output array of suffixes.
+ * @param n The length of the given string.
+ * @return 0 if no error occurred, -1 or -2 otherwise.
+ */
+DIVSUFSORT_API
+saint_t
+divsufsort@W64BIT@(const sauchar_t *T, saidx@W64BIT@_t *SA, saidx@W64BIT@_t n);
+
+/**
+ * Constructs the burrows-wheeler transformed string of a given string.
+ * @param T[0..n-1] The input string.
+ * @param U[0..n-1] The output string. (can be T)
+ * @param A[0..n-1] The temporary array. (can be NULL)
+ * @param n The length of the given string.
+ * @return The primary index if no error occurred, -1 or -2 otherwise.
+ */
+DIVSUFSORT_API
+saidx@W64BIT@_t
+divbwt@W64BIT@(const sauchar_t *T, sauchar_t *U, saidx@W64BIT@_t *A, saidx@W64BIT@_t n);
+
+/**
+ * Returns the version of the divsufsort library.
+ * @return The version number string.
+ */
+DIVSUFSORT_API
+const char *
+divsufsort@W64BIT@_version(void);
+
+
+/**
+ * Constructs the burrows-wheeler transformed string of a given string and suffix array.
+ * @param T[0..n-1] The input string.
+ * @param U[0..n-1] The output string. (can be T)
+ * @param SA[0..n-1] The suffix array. (can be NULL)
+ * @param n The length of the given string.
+ * @param idx The output primary index.
+ * @return 0 if no error occurred, -1 or -2 otherwise.
+ */
+DIVSUFSORT_API
+saint_t
+bw_transform@W64BIT@(const sauchar_t *T, sauchar_t *U,
+             saidx@W64BIT@_t *SA /* can NULL */,
+             saidx@W64BIT@_t n, saidx@W64BIT@_t *idx);
+
+/**
+ * Inverse BW-transforms a given BWTed string.
+ * @param T[0..n-1] The input string.
+ * @param U[0..n-1] The output string. (can be T)
+ * @param A[0..n-1] The temporary array. (can be NULL)
+ * @param n The length of the given string.
+ * @param idx The primary index.
+ * @return 0 if no error occurred, -1 or -2 otherwise.
+ */
+DIVSUFSORT_API
+saint_t
+inverse_bw_transform@W64BIT@(const sauchar_t *T, sauchar_t *U,
+                     saidx@W64BIT@_t *A /* can NULL */,
+                     saidx@W64BIT@_t n, saidx@W64BIT@_t idx);
+
+/**
+ * Checks the correctness of a given suffix array.
+ * @param T[0..n-1] The input string.
+ * @param SA[0..n-1] The input suffix array.
+ * @param n The length of the given string.
+ * @param verbose The verbose mode.
+ * @return 0 if no error occurred.
+ */
+DIVSUFSORT_API
+saint_t
+sufcheck@W64BIT@(const sauchar_t *T, const saidx@W64BIT@_t *SA, saidx@W64BIT@_t n, saint_t verbose);
+
+/**
+ * Search for the pattern P in the string T.
+ * @param T[0..Tsize-1] The input string.
+ * @param Tsize The length of the given string.
+ * @param P[0..Psize-1] The input pattern string.
+ * @param Psize The length of the given pattern string.
+ * @param SA[0..SAsize-1] The input suffix array.
+ * @param SAsize The length of the given suffix array.
+ * @param idx The output index.
+ * @return The count of matches if no error occurred, -1 otherwise.
+ */
+DIVSUFSORT_API
+saidx@W64BIT@_t
+sa_search@W64BIT@(const sauchar_t *T, saidx@W64BIT@_t Tsize,
+          const sauchar_t *P, saidx@W64BIT@_t Psize,
+          const saidx@W64BIT@_t *SA, saidx@W64BIT@_t SAsize,
+          saidx@W64BIT@_t *left);
+
+/**
+ * Search for the character c in the string T.
+ * @param T[0..Tsize-1] The input string.
+ * @param Tsize The length of the given string.
+ * @param SA[0..SAsize-1] The input suffix array.
+ * @param SAsize The length of the given suffix array.
+ * @param c The input character.
+ * @param idx The output index.
+ * @return The count of matches if no error occurred, -1 otherwise.
+ */
+DIVSUFSORT_API
+saidx@W64BIT@_t
+sa_simplesearch@W64BIT@(const sauchar_t *T, saidx@W64BIT@_t Tsize,
+                const saidx@W64BIT@_t *SA, saidx@W64BIT@_t SAsize,
+                saint_t c, saidx@W64BIT@_t *left);
+
+
+#ifdef __cplusplus
+} /* extern "C" */
+#endif /* __cplusplus */
+
+#endif /* _DIVSUFSORT@W64BIT@_H */

loader/tools/lzsa/src/libdivsufsort/include/divsufsort_config.h

@@ -0,0 +1,9 @@
+#define HAVE_STRING_H 1
+#define HAVE_STDLIB_H 1
+#define HAVE_MEMORY_H 1
+#define HAVE_STDINT_H 1
+#define INLINE inline
+
+#ifdef _MSC_VER
+#pragma warning( disable : 4244 )
+#endif /* _MSC_VER */

loader/tools/lzsa/src/libdivsufsort/include/divsufsort_private.h

@@ -0,0 +1,205 @@
+/*
+ * divsufsort_private.h for libdivsufsort
+ * Copyright (c) 2003-2008 Yuta Mori All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use,
+ * copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following
+ * conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+ * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _DIVSUFSORT_PRIVATE_H
+#define _DIVSUFSORT_PRIVATE_H 1
+
+#ifdef __cplusplus
+extern "C" {
+#endif /* __cplusplus */
+
+#include "divsufsort_config.h"
+#include <assert.h>
+#include <stdio.h>
+#if HAVE_STRING_H
+# include <string.h>
+#endif
+#if HAVE_STDLIB_H
+# include <stdlib.h>
+#endif
+#if HAVE_MEMORY_H
+# include <memory.h>
+#endif
+#if HAVE_STDDEF_H
+# include <stddef.h>
+#endif
+#if HAVE_STRINGS_H
+# include <strings.h>
+#endif
+#if HAVE_INTTYPES_H
+# include <inttypes.h>
+#else
+# if HAVE_STDINT_H
+#  include <stdint.h>
+# endif
+#endif
+#if defined(BUILD_DIVSUFSORT64)
+# include "divsufsort64.h"
+# ifndef SAIDX_T
+#  define SAIDX_T
+#  define saidx_t saidx64_t
+# endif /* SAIDX_T */
+# ifndef PRIdSAIDX_T
+#  define PRIdSAIDX_T PRIdSAIDX64_T
+# endif /* PRIdSAIDX_T */
+# define divsufsort divsufsort64
+# define divbwt divbwt64
+# define divsufsort_version divsufsort64_version
+# define bw_transform bw_transform64
+# define inverse_bw_transform inverse_bw_transform64
+# define sufcheck sufcheck64
+# define sa_search sa_search64
+# define sa_simplesearch sa_simplesearch64
+# define sssort sssort64
+# define trsort trsort64
+#else
+# include "divsufsort.h"
+#endif
+
+
+/*- Constants -*/
+#if !defined(UINT8_MAX)
+# define UINT8_MAX (255)
+#endif /* UINT8_MAX */
+#if defined(ALPHABET_SIZE) && (ALPHABET_SIZE < 1)
+# undef ALPHABET_SIZE
+#endif
+#if !defined(ALPHABET_SIZE)
+# define ALPHABET_SIZE (UINT8_MAX + 1)
+#endif
+/* for divsufsort.c */
+#define BUCKET_A_SIZE (ALPHABET_SIZE)
+#define BUCKET_B_SIZE (ALPHABET_SIZE * ALPHABET_SIZE)
+/* for sssort.c */
+#if defined(SS_INSERTIONSORT_THRESHOLD)
+# if SS_INSERTIONSORT_THRESHOLD < 1
+#  undef SS_INSERTIONSORT_THRESHOLD
+#  define SS_INSERTIONSORT_THRESHOLD (1)
+# endif
+#else
+# define SS_INSERTIONSORT_THRESHOLD (8)
+#endif
+#if defined(SS_BLOCKSIZE)
+# if SS_BLOCKSIZE < 0
+#  undef SS_BLOCKSIZE
+#  define SS_BLOCKSIZE (0)
+# elif 32768 <= SS_BLOCKSIZE
+#  undef SS_BLOCKSIZE
+#  define SS_BLOCKSIZE (32767)
+# endif
+#else
+# define SS_BLOCKSIZE (1024)
+#endif
+/* minstacksize = log(SS_BLOCKSIZE) / log(3) * 2 */
+#if SS_BLOCKSIZE == 0
+# if defined(BUILD_DIVSUFSORT64)
+#  define SS_MISORT_STACKSIZE (96)
+# else
+#  define SS_MISORT_STACKSIZE (64)
+# endif
+#elif SS_BLOCKSIZE <= 4096
+# define SS_MISORT_STACKSIZE (16)
+#else
+# define SS_MISORT_STACKSIZE (24)
+#endif
+#if defined(BUILD_DIVSUFSORT64)
+# define SS_SMERGE_STACKSIZE (64)
+#else
+# define SS_SMERGE_STACKSIZE (32)
+#endif
+/* for trsort.c */
+#define TR_INSERTIONSORT_THRESHOLD (8)
+#if defined(BUILD_DIVSUFSORT64)
+# define TR_STACKSIZE (96)
+#else
+# define TR_STACKSIZE (64)
+#endif
+
+
+/*- Macros -*/
+#ifndef SWAP
+# define SWAP(_a, _b) do { t = (_a); (_a) = (_b); (_b) = t; } while(0)
+#endif /* SWAP */
+#ifndef MIN
+# define MIN(_a, _b) (((_a) < (_b)) ? (_a) : (_b))
+#endif /* MIN */
+#ifndef MAX
+# define MAX(_a, _b) (((_a) > (_b)) ? (_a) : (_b))
+#endif /* MAX */
+#define STACK_PUSH(_a, _b, _c, _d)\
+  do {\
+    assert(ssize < STACK_SIZE);\
+    stack[ssize].a = (_a), stack[ssize].b = (_b),\
+    stack[ssize].c = (_c), stack[ssize++].d = (_d);\
+  } while(0)
+#define STACK_PUSH5(_a, _b, _c, _d, _e)\
+  do {\
+    assert(ssize < STACK_SIZE);\
+    stack[ssize].a = (_a), stack[ssize].b = (_b),\
+    stack[ssize].c = (_c), stack[ssize].d = (_d), stack[ssize++].e = (_e);\
+  } while(0)
+#define STACK_POP(_a, _b, _c, _d)\
+  do {\
+    assert(0 <= ssize);\
+    if(ssize == 0) { return; }\
+    (_a) = stack[--ssize].a, (_b) = stack[ssize].b,\
+    (_c) = stack[ssize].c, (_d) = stack[ssize].d;\
+  } while(0)
+#define STACK_POP5(_a, _b, _c, _d, _e)\
+  do {\
+    assert(0 <= ssize);\
+    if(ssize == 0) { return; }\
+    (_a) = stack[--ssize].a, (_b) = stack[ssize].b,\
+    (_c) = stack[ssize].c, (_d) = stack[ssize].d, (_e) = stack[ssize].e;\
+  } while(0)
+/* for divsufsort.c */
+#define BUCKET_A(_c0) bucket_A[(_c0)]
+#if ALPHABET_SIZE == 256
+#define BUCKET_B(_c0, _c1) (bucket_B[((_c1) << 8) | (_c0)])
+#define BUCKET_BSTAR(_c0, _c1) (bucket_B[((_c0) << 8) | (_c1)])
+#else
+#define BUCKET_B(_c0, _c1) (bucket_B[(_c1) * ALPHABET_SIZE + (_c0)])
+#define BUCKET_BSTAR(_c0, _c1) (bucket_B[(_c0) * ALPHABET_SIZE + (_c1)])
+#endif
+
+
+/*- Private Prototypes -*/
+/* sssort.c */
+void
+sssort(const sauchar_t *Td, const saidx_t *PA,
+       saidx_t *first, saidx_t *last,
+       saidx_t *buf, saidx_t bufsize,
+       saidx_t depth, saidx_t n, saint_t lastsuffix);
+/* trsort.c */
+void
+trsort(saidx_t *ISA, saidx_t *SA, saidx_t n, saidx_t depth);
+
+
+#ifdef __cplusplus
+} /* extern "C" */
+#endif /* __cplusplus */
+
+#endif /* _DIVSUFSORT_PRIVATE_H */

loader/tools/lzsa/src/libdivsufsort/include/lfs.h.cmake

@@ -0,0 +1,56 @@
+/*
+ * lfs.h for libdivsufsort
+ * Copyright (c) 2003-2008 Yuta Mori All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use,
+ * copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following
+ * conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+ * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _LFS_H
+#define _LFS_H 1
+
+#ifdef __cplusplus
+extern "C" {
+#endif /* __cplusplus */
+
+#ifndef __STRICT_ANSI__
+# define LFS_OFF_T @LFS_OFF_T@
+# define LFS_FOPEN @LFS_FOPEN@
+# define LFS_FTELL @LFS_FTELL@
+# define LFS_FSEEK @LFS_FSEEK@
+# define LFS_PRId  @LFS_PRID@
+#else
+# define LFS_OFF_T long
+# define LFS_FOPEN fopen
+# define LFS_FTELL ftell
+# define LFS_FSEEK fseek
+# define LFS_PRId "ld"
+#endif
+#ifndef PRIdOFF_T
+# define PRIdOFF_T LFS_PRId
+#endif
+
+
+#ifdef __cplusplus
+} /* extern "C" */
+#endif /* __cplusplus */
+
+#endif /* _LFS_H */

loader/tools/lzsa/src/libdivsufsort/lib/CMakeLists.txt

@@ -0,0 +1,31 @@
+include_directories("${CMAKE_CURRENT_SOURCE_DIR}/../include"
+                    "${CMAKE_CURRENT_BINARY_DIR}/../include")
+
+set(divsufsort_SRCS divsufsort.c sssort.c trsort.c utils.c)
+
+## libdivsufsort ##
+add_library(divsufsort ${divsufsort_SRCS})
+install(TARGETS divsufsort
+  RUNTIME DESTINATION ${CMAKE_INSTALL_RUNTIMEDIR}
+  LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR}
+  ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR})
+set_target_properties(divsufsort PROPERTIES
+  VERSION   "${LIBRARY_VERSION}"
+  SOVERSION "${LIBRARY_SOVERSION}"
+  DEFINE_SYMBOL DIVSUFSORT_BUILD_DLL
+  RUNTIME_OUTPUT_DIRECTORY "${CMAKE_CURRENT_BINARY_DIR}/../examples")
+
+## libdivsufsort64 ##
+if(BUILD_DIVSUFSORT64)
+  add_library(divsufsort64 ${divsufsort_SRCS})
+  install(TARGETS divsufsort64
+    RUNTIME DESTINATION ${CMAKE_INSTALL_RUNTIMEDIR}
+    LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR}
+    ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR})
+  set_target_properties(divsufsort64 PROPERTIES
+    VERSION   "${LIBRARY_VERSION}"
+    SOVERSION "${LIBRARY_SOVERSION}"
+    DEFINE_SYMBOL DIVSUFSORT_BUILD_DLL
+    COMPILE_FLAGS "-DBUILD_DIVSUFSORT64"
+    RUNTIME_OUTPUT_DIRECTORY "${CMAKE_CURRENT_BINARY_DIR}/../examples")
+endif(BUILD_DIVSUFSORT64)

loader/tools/lzsa/src/libdivsufsort/lib/divsufsort.c

@@ -0,0 +1,431 @@
+/*
+ * divsufsort.c for libdivsufsort
+ * Copyright (c) 2003-2008 Yuta Mori All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use,
+ * copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following
+ * conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+ * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "divsufsort_private.h"
+#ifdef _OPENMP
+# include <omp.h>
+#endif
+
+
+/*- Private Functions -*/
+
+/* Sorts suffixes of type B*. */
+static
+saidx_t
+sort_typeBstar(const sauchar_t *T, saidx_t *SA,
+               saidx_t *bucket_A, saidx_t *bucket_B,
+               saidx_t n) {
+  saidx_t *PAb, *ISAb, *buf;
+#ifdef _OPENMP
+  saidx_t *curbuf;
+  saidx_t l;
+#endif
+  saidx_t i, j, k, t, m, bufsize;
+  saint_t c0, c1;
+#ifdef _OPENMP
+  saint_t d0, d1;
+  int tmp;
+#endif
+
+  /* Initialize bucket arrays. */
+  for(i = 0; i < BUCKET_A_SIZE; ++i) { bucket_A[i] = 0; }
+  for(i = 0; i < BUCKET_B_SIZE; ++i) { bucket_B[i] = 0; }
+
+  /* Count the number of occurrences of the first one or two characters of each
+     type A, B and B* suffix. Moreover, store the beginning position of all
+     type B* suffixes into the array SA. */
+  for(i = n - 1, m = n, c0 = T[n - 1]; 0 <= i;) {
+    /* type A suffix. */
+    do { ++BUCKET_A(c1 = c0); } while((0 <= --i) && ((c0 = T[i]) >= c1));
+    if(0 <= i) {
+      /* type B* suffix. */
+      ++BUCKET_BSTAR(c0, c1);
+      SA[--m] = i;
+      /* type B suffix. */
+      for(--i, c1 = c0; (0 <= i) && ((c0 = T[i]) <= c1); --i, c1 = c0) {
+        ++BUCKET_B(c0, c1);
+      }
+    }
+  }
+  m = n - m;
+/*
+note:
+  A type B* suffix is lexicographically smaller than a type B suffix that
+  begins with the same first two characters.
+*/
+
+  /* Calculate the index of start/end point of each bucket. */
+  for(c0 = 0, i = 0, j = 0; c0 < ALPHABET_SIZE; ++c0) {
+    t = i + BUCKET_A(c0);
+    BUCKET_A(c0) = i + j; /* start point */
+    i = t + BUCKET_B(c0, c0);
+    for(c1 = c0 + 1; c1 < ALPHABET_SIZE; ++c1) {
+      j += BUCKET_BSTAR(c0, c1);
+      BUCKET_BSTAR(c0, c1) = j; /* end point */
+      i += BUCKET_B(c0, c1);
+    }
+  }
+
+  if(0 < m) {
+    /* Sort the type B* suffixes by their first two characters. */
+    PAb = SA + n - m; ISAb = SA + m;
+    for(i = m - 2; 0 <= i; --i) {
+      t = PAb[i], c0 = T[t], c1 = T[t + 1];
+      SA[--BUCKET_BSTAR(c0, c1)] = i;
+    }
+    t = PAb[m - 1], c0 = T[t], c1 = T[t + 1];
+    SA[--BUCKET_BSTAR(c0, c1)] = m - 1;
+
+    /* Sort the type B* substrings using sssort. */
+#ifdef _OPENMP
+    tmp = omp_get_max_threads();
+    buf = SA + m, bufsize = (n - (2 * m)) / tmp;
+    c0 = ALPHABET_SIZE - 2, c1 = ALPHABET_SIZE - 1, j = m;
+#pragma omp parallel default(shared) private(curbuf, k, l, d0, d1, tmp)
+    {
+      tmp = omp_get_thread_num();
+      curbuf = buf + tmp * bufsize;
+      k = 0;
+      for(;;) {
+        #pragma omp critical(sssort_lock)
+        {
+          if(0 < (l = j)) {
+            d0 = c0, d1 = c1;
+            do {
+              k = BUCKET_BSTAR(d0, d1);
+              if(--d1 <= d0) {
+                d1 = ALPHABET_SIZE - 1;
+                if(--d0 < 0) { break; }
+              }
+            } while(((l - k) <= 1) && (0 < (l = k)));
+            c0 = d0, c1 = d1, j = k;
+          }
+        }
+        if(l == 0) { break; }
+        sssort(T, PAb, SA + k, SA + l,
+               curbuf, bufsize, 2, n, *(SA + k) == (m - 1));
+      }
+    }
+#else
+    buf = SA + m, bufsize = n - (2 * m);
+    for(c0 = ALPHABET_SIZE - 2, j = m; 0 < j; --c0) {
+      for(c1 = ALPHABET_SIZE - 1; c0 < c1; j = i, --c1) {
+        i = BUCKET_BSTAR(c0, c1);
+        if(1 < (j - i)) {
+          sssort(T, PAb, SA + i, SA + j,
+                 buf, bufsize, 2, n, *(SA + i) == (m - 1));
+        }
+      }
+    }
+#endif
+
+    /* Compute ranks of type B* substrings. */
+    for(i = m - 1; 0 <= i; --i) {
+      if(0 <= SA[i]) {
+        j = i;
+        do { ISAb[SA[i]] = i; } while((0 <= --i) && (0 <= SA[i]));
+        SA[i + 1] = i - j;
+        if(i <= 0) { break; }
+      }
+      j = i;
+      do { ISAb[SA[i] = ~SA[i]] = j; } while(SA[--i] < 0);
+      ISAb[SA[i]] = j;
+    }
+
+    /* Construct the inverse suffix array of type B* suffixes using trsort. */
+    trsort(ISAb, SA, m, 1);
+
+    /* Set the sorted order of tyoe B* suffixes. */
+    for(i = n - 1, j = m, c0 = T[n - 1]; 0 <= i;) {
+      for(--i, c1 = c0; (0 <= i) && ((c0 = T[i]) >= c1); --i, c1 = c0) { }
+      if(0 <= i) {
+        t = i;
+        for(--i, c1 = c0; (0 <= i) && ((c0 = T[i]) <= c1); --i, c1 = c0) { }
+        SA[ISAb[--j]] = ((t == 0) || (1 < (t - i))) ? t : ~t;
+      }
+    }
+
+    /* Calculate the index of start/end point of each bucket. */
+    BUCKET_B(ALPHABET_SIZE - 1, ALPHABET_SIZE - 1) = n; /* end point */
+    for(c0 = ALPHABET_SIZE - 2, k = m - 1; 0 <= c0; --c0) {
+      i = BUCKET_A(c0 + 1) - 1;
+      for(c1 = ALPHABET_SIZE - 1; c0 < c1; --c1) {
+        t = i - BUCKET_B(c0, c1);
+        BUCKET_B(c0, c1) = i; /* end point */
+
+        /* Move all type B* suffixes to the correct position. */
+        for(i = t, j = BUCKET_BSTAR(c0, c1);
+            j <= k;
+            --i, --k) { SA[i] = SA[k]; }
+      }
+      BUCKET_BSTAR(c0, c0 + 1) = i - BUCKET_B(c0, c0) + 1; /* start point */
+      BUCKET_B(c0, c0) = i; /* end point */
+    }
+  }
+
+  return m;
+}
+
+/* Constructs the suffix array by using the sorted order of type B* suffixes. */
+static
+void
+construct_SA(const sauchar_t *T, saidx_t *SA,
+             saidx_t *bucket_A, saidx_t *bucket_B,
+             saidx_t n, saidx_t m) {
+  saidx_t *i, *j, *k;
+  saidx_t s;
+  saint_t c0, c1, c2;
+
+  if(0 < m) {
+    /* Construct the sorted order of type B suffixes by using
+       the sorted order of type B* suffixes. */
+    for(c1 = ALPHABET_SIZE - 2; 0 <= c1; --c1) {
+      /* Scan the suffix array from right to left. */
+      for(i = SA + BUCKET_BSTAR(c1, c1 + 1),
+          j = SA + BUCKET_A(c1 + 1) - 1, k = NULL, c2 = -1;
+          i <= j;
+          --j) {
+        if(0 < (s = *j)) {
+          assert(T[s] == c1);
+          assert(((s + 1) < n) && (T[s] <= T[s + 1]));
+          assert(T[s - 1] <= T[s]);
+          *j = ~s;
+          c0 = T[--s];
+          if((0 < s) && (T[s - 1] > c0)) { s = ~s; }
+          if(c0 != c2) {
+            if(0 <= c2) { BUCKET_B(c2, c1) = k - SA; }
+            k = SA + BUCKET_B(c2 = c0, c1);
+          }
+          assert(k < j);
+          *k-- = s;
+        } else {
+          assert(((s == 0) && (T[s] == c1)) || (s < 0));
+          *j = ~s;
+        }
+      }
+    }
+  }
+
+  /* Construct the suffix array by using
+     the sorted order of type B suffixes. */
+  k = SA + BUCKET_A(c2 = T[n - 1]);
+  *k++ = (T[n - 2] < c2) ? ~(n - 1) : (n - 1);
+  /* Scan the suffix array from left to right. */
+  for(i = SA, j = SA + n; i < j; ++i) {
+    if(0 < (s = *i)) {
+      assert(T[s - 1] >= T[s]);
+      c0 = T[--s];
+      if((s == 0) || (T[s - 1] < c0)) { s = ~s; }
+      if(c0 != c2) {
+        BUCKET_A(c2) = k - SA;
+        k = SA + BUCKET_A(c2 = c0);
+      }
+      assert(i < k);
+      *k++ = s;
+    } else {
+      assert(s < 0);
+      *i = ~s;
+    }
+  }
+}
+
+#if 0
+/* Constructs the burrows-wheeler transformed string directly
+   by using the sorted order of type B* suffixes. */
+static
+saidx_t
+construct_BWT(const sauchar_t *T, saidx_t *SA,
+              saidx_t *bucket_A, saidx_t *bucket_B,
+              saidx_t n, saidx_t m) {
+  saidx_t *i, *j, *k, *orig;
+  saidx_t s;
+  saint_t c0, c1, c2;
+
+  if(0 < m) {
+    /* Construct the sorted order of type B suffixes by using
+       the sorted order of type B* suffixes. */
+    for(c1 = ALPHABET_SIZE - 2; 0 <= c1; --c1) {
+      /* Scan the suffix array from right to left. */
+      for(i = SA + BUCKET_BSTAR(c1, c1 + 1),
+          j = SA + BUCKET_A(c1 + 1) - 1, k = NULL, c2 = -1;
+          i <= j;
+          --j) {
+        if(0 < (s = *j)) {
+          assert(T[s] == c1);
+          assert(((s + 1) < n) && (T[s] <= T[s + 1]));
+          assert(T[s - 1] <= T[s]);
+          c0 = T[--s];
+          *j = ~((saidx_t)c0);
+          if((0 < s) && (T[s - 1] > c0)) { s = ~s; }
+          if(c0 != c2) {
+            if(0 <= c2) { BUCKET_B(c2, c1) = k - SA; }
+            k = SA + BUCKET_B(c2 = c0, c1);
+          }
+          assert(k < j);
+          *k-- = s;
+        } else if(s != 0) {
+          *j = ~s;
+#ifndef NDEBUG
+        } else {
+          assert(T[s] == c1);
+#endif
+        }
+      }
+    }
+  }
+
+  /* Construct the BWTed string by using
+     the sorted order of type B suffixes. */
+  k = SA + BUCKET_A(c2 = T[n - 1]);
+  *k++ = (T[n - 2] < c2) ? ~((saidx_t)T[n - 2]) : (n - 1);
+  /* Scan the suffix array from left to right. */
+  for(i = SA, j = SA + n, orig = SA; i < j; ++i) {
+    if(0 < (s = *i)) {
+      assert(T[s - 1] >= T[s]);
+      c0 = T[--s];
+      *i = c0;
+      if((0 < s) && (T[s - 1] < c0)) { s = ~((saidx_t)T[s - 1]); }
+      if(c0 != c2) {
+        BUCKET_A(c2) = k - SA;
+        k = SA + BUCKET_A(c2 = c0);
+      }
+      assert(i < k);
+      *k++ = s;
+    } else if(s != 0) {
+      *i = ~s;
+    } else {
+      orig = i;
+    }
+  }
+
+  return orig - SA;
+}
+#endif
+
+/*---------------------------------------------------------------------------*/
+
+/**
+ * Initialize suffix array context
+ *
+ * @return 0 for success, or non-zero in case of an error
+ */
+int divsufsort_init(divsufsort_ctx_t *ctx) {
+   ctx->bucket_A = (saidx_t *)malloc(BUCKET_A_SIZE * sizeof(saidx_t));
+   ctx->bucket_B = NULL;
+
+   if (ctx->bucket_A) {
+      ctx->bucket_B = (saidx_t *)malloc(BUCKET_B_SIZE * sizeof(saidx_t));
+
+      if (ctx->bucket_B)
+         return 0;
+   }
+
+   divsufsort_destroy(ctx);
+   return -1;
+}
+
+/**
+ * Destroy suffix array context
+ *
+ * @param ctx suffix array context to destroy
+ */
+void divsufsort_destroy(divsufsort_ctx_t *ctx) {
+   if (ctx->bucket_B) {
+      free(ctx->bucket_B);
+      ctx->bucket_B = NULL;
+   }
+
+   if (ctx->bucket_A) {
+      free(ctx->bucket_A);
+      ctx->bucket_A = NULL;
+   }
+}
+
+/*- Function -*/
+
+saint_t
+divsufsort_build_array(divsufsort_ctx_t *ctx, const sauchar_t *T, saidx_t *SA, saidx_t n) {
+  saidx_t m;
+  saint_t err = 0;
+
+  /* Check arguments. */
+  if((T == NULL) || (SA == NULL) || (n < 0)) { return -1; }
+  else if(n == 0) { return 0; }
+  else if(n == 1) { SA[0] = 0; return 0; }
+  else if(n == 2) { m = (T[0] < T[1]); SA[m ^ 1] = 0, SA[m] = 1; return 0; }
+
+  /* Suffixsort. */
+  if((ctx->bucket_A != NULL) && (ctx->bucket_B != NULL)) {
+    m = sort_typeBstar(T, SA, ctx->bucket_A, ctx->bucket_B, n);
+    construct_SA(T, SA, ctx->bucket_A, ctx->bucket_B, n, m);
+  } else {
+    err = -2;
+  }
+
+  return err;
+}
+
+#if 0
+saidx_t
+divbwt(const sauchar_t *T, sauchar_t *U, saidx_t *A, saidx_t n) {
+  saidx_t *B;
+  saidx_t *bucket_A, *bucket_B;
+  saidx_t m, pidx, i;
+
+  /* Check arguments. */
+  if((T == NULL) || (U == NULL) || (n < 0)) { return -1; }
+  else if(n <= 1) { if(n == 1) { U[0] = T[0]; } return n; }
+
+  if((B = A) == NULL) { B = (saidx_t *)malloc((size_t)(n + 1) * sizeof(saidx_t)); }
+  bucket_A = (saidx_t *)malloc(BUCKET_A_SIZE * sizeof(saidx_t));
+  bucket_B = (saidx_t *)malloc(BUCKET_B_SIZE * sizeof(saidx_t));
+
+  /* Burrows-Wheeler Transform. */
+  if((B != NULL) && (bucket_A != NULL) && (bucket_B != NULL)) {
+    m = sort_typeBstar(T, B, bucket_A, bucket_B, n);
+    pidx = construct_BWT(T, B, bucket_A, bucket_B, n, m);
+
+    /* Copy to output string. */
+    U[0] = T[n - 1];
+    for(i = 0; i < pidx; ++i) { U[i + 1] = (sauchar_t)B[i]; }
+    for(i += 1; i < n; ++i) { U[i] = (sauchar_t)B[i]; }
+    pidx += 1;
+  } else {
+    pidx = -2;
+  }
+
+  free(bucket_B);
+  free(bucket_A);
+  if(A == NULL) { free(B); }
+
+  return pidx;
+}
+
+const char *
+divsufsort_version(void) {
+  return PROJECT_VERSION_FULL;
+}
+#endif

loader/tools/lzsa/src/libdivsufsort/lib/divsufsort_utils.c

@@ -0,0 +1,383 @@
+/*
+ * utils.c for libdivsufsort
+ * Copyright (c) 2003-2008 Yuta Mori All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use,
+ * copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following
+ * conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+ * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "divsufsort_private.h"
+
+
+/*- Private Function -*/
+
+#if 0
+/* Binary search for inverse bwt. */
+static
+saidx_t
+binarysearch_lower(const saidx_t *A, saidx_t size, saidx_t value) {
+  saidx_t half, i;
+  for(i = 0, half = size >> 1;
+      0 < size;
+      size = half, half >>= 1) {
+    if(A[i + half] < value) {
+      i += half + 1;
+      half -= (size & 1) ^ 1;
+    }
+  }
+  return i;
+}
+
+
+/*- Functions -*/
+
+/* Burrows-Wheeler transform. */
+saint_t
+bw_transform(const sauchar_t *T, sauchar_t *U, saidx_t *SA,
+             saidx_t n, saidx_t *idx) {
+  saidx_t *A, i, j, p, t;
+  saint_t c;
+
+  /* Check arguments. */
+  if((T == NULL) || (U == NULL) || (n < 0) || (idx == NULL)) { return -1; }
+  if(n <= 1) {
+    if(n == 1) { U[0] = T[0]; }
+    *idx = n;
+    return 0;
+  }
+
+  if((A = SA) == NULL) {
+    i = divbwt(T, U, NULL, n);
+    if(0 <= i) { *idx = i; i = 0; }
+    return (saint_t)i;
+  }
+
+  /* BW transform. */
+  if(T == U) {
+    t = n;
+    for(i = 0, j = 0; i < n; ++i) {
+      p = t - 1;
+      t = A[i];
+      if(0 <= p) {
+        c = T[j];
+        U[j] = (j <= p) ? T[p] : (sauchar_t)A[p];
+        A[j] = c;
+        j++;
+      } else {
+        *idx = i;
+      }
+    }
+    p = t - 1;
+    if(0 <= p) {
+      c = T[j];
+      U[j] = (j <= p) ? T[p] : (sauchar_t)A[p];
+      A[j] = c;
+    } else {
+      *idx = i;
+    }
+  } else {
+    U[0] = T[n - 1];
+    for(i = 0; A[i] != 0; ++i) { U[i + 1] = T[A[i] - 1]; }
+    *idx = i + 1;
+    for(++i; i < n; ++i) { U[i] = T[A[i] - 1]; }
+  }
+
+  if(SA == NULL) {
+    /* Deallocate memory. */
+    free(A);
+  }
+
+  return 0;
+}
+
+/* Inverse Burrows-Wheeler transform. */
+saint_t
+inverse_bw_transform(const sauchar_t *T, sauchar_t *U, saidx_t *A,
+                     saidx_t n, saidx_t idx) {
+  saidx_t C[ALPHABET_SIZE];
+  sauchar_t D[ALPHABET_SIZE];
+  saidx_t *B;
+  saidx_t i, p;
+  saint_t c, d;
+
+  /* Check arguments. */
+  if((T == NULL) || (U == NULL) || (n < 0) || (idx < 0) ||
+     (n < idx) || ((0 < n) && (idx == 0))) {
+    return -1;
+  }
+  if(n <= 1) { return 0; }
+
+  if((B = A) == NULL) {
+    /* Allocate n*sizeof(saidx_t) bytes of memory. */
+    if((B = (saidx_t *)malloc((size_t)n * sizeof(saidx_t))) == NULL) { return -2; }
+  }
+
+  /* Inverse BW transform. */
+  for(c = 0; c < ALPHABET_SIZE; ++c) { C[c] = 0; }
+  for(i = 0; i < n; ++i) { ++C[T[i]]; }
+  for(c = 0, d = 0, i = 0; c < ALPHABET_SIZE; ++c) {
+    p = C[c];
+    if(0 < p) {
+      C[c] = i;
+      D[d++] = (sauchar_t)c;
+      i += p;
+    }
+  }
+  for(i = 0; i < idx; ++i) { B[C[T[i]]++] = i; }
+  for( ; i < n; ++i)       { B[C[T[i]]++] = i + 1; }
+  for(c = 0; c < d; ++c) { C[c] = C[D[c]]; }
+  for(i = 0, p = idx; i < n; ++i) {
+    U[i] = D[binarysearch_lower(C, d, p)];
+    p = B[p - 1];
+  }
+
+  if(A == NULL) {
+    /* Deallocate memory. */
+    free(B);
+  }
+
+  return 0;
+}
+
+/* Checks the suffix array SA of the string T. */
+saint_t
+sufcheck(const sauchar_t *T, const saidx_t *SA,
+         saidx_t n, saint_t verbose) {
+  saidx_t C[ALPHABET_SIZE];
+  saidx_t i, p, q, t;
+  saint_t c;
+
+  if(verbose) { fprintf(stderr, "sufcheck: "); }
+
+  /* Check arguments. */
+  if((T == NULL) || (SA == NULL) || (n < 0)) {
+    if(verbose) { fprintf(stderr, "Invalid arguments.\n"); }
+    return -1;
+  }
+  if(n == 0) {
+    if(verbose) { fprintf(stderr, "Done.\n"); }
+    return 0;
+  }
+
+  /* check range: [0..n-1] */
+  for(i = 0; i < n; ++i) {
+    if((SA[i] < 0) || (n <= SA[i])) {
+      if(verbose) {
+        fprintf(stderr, "Out of the range [0,%" PRIdSAIDX_T "].\n"
+                        "  SA[%" PRIdSAIDX_T "]=%" PRIdSAIDX_T "\n",
+                        n - 1, i, SA[i]);
+      }
+      return -2;
+    }
+  }
+
+  /* check first characters. */
+  for(i = 1; i < n; ++i) {
+    if(T[SA[i - 1]] > T[SA[i]]) {
+      if(verbose) {
+        fprintf(stderr, "Suffixes in wrong order.\n"
+                        "  T[SA[%" PRIdSAIDX_T "]=%" PRIdSAIDX_T "]=%d"
+                        " > T[SA[%" PRIdSAIDX_T "]=%" PRIdSAIDX_T "]=%d\n",
+                        i - 1, SA[i - 1], T[SA[i - 1]], i, SA[i], T[SA[i]]);
+      }
+      return -3;
+    }
+  }
+
+  /* check suffixes. */
+  for(i = 0; i < ALPHABET_SIZE; ++i) { C[i] = 0; }
+  for(i = 0; i < n; ++i) { ++C[T[i]]; }
+  for(i = 0, p = 0; i < ALPHABET_SIZE; ++i) {
+    t = C[i];
+    C[i] = p;
+    p += t;
+  }
+
+  q = C[T[n - 1]];
+  C[T[n - 1]] += 1;
+  for(i = 0; i < n; ++i) {
+    p = SA[i];
+    if(0 < p) {
+      c = T[--p];
+      t = C[c];
+    } else {
+      c = T[p = n - 1];
+      t = q;
+    }
+    if((t < 0) || (p != SA[t])) {
+      if(verbose) {
+        fprintf(stderr, "Suffix in wrong position.\n"
+                        "  SA[%" PRIdSAIDX_T "]=%" PRIdSAIDX_T " or\n"
+                        "  SA[%" PRIdSAIDX_T "]=%" PRIdSAIDX_T "\n",
+                        t, (0 <= t) ? SA[t] : -1, i, SA[i]);
+      }
+      return -4;
+    }
+    if(t != q) {
+      ++C[c];
+      if((n <= C[c]) || (T[SA[C[c]]] != c)) { C[c] = -1; }
+    }
+  }
+
+  if(1 <= verbose) { fprintf(stderr, "Done.\n"); }
+  return 0;
+}
+
+
+static
+int
+_compare(const sauchar_t *T, saidx_t Tsize,
+         const sauchar_t *P, saidx_t Psize,
+         saidx_t suf, saidx_t *match) {
+  saidx_t i, j;
+  saint_t r;
+  for(i = suf + *match, j = *match, r = 0;
+      (i < Tsize) && (j < Psize) && ((r = T[i] - P[j]) == 0); ++i, ++j) { }
+  *match = j;
+  return (r == 0) ? -(j != Psize) : r;
+}
+
+/* Search for the pattern P in the string T. */
+saidx_t
+sa_search(const sauchar_t *T, saidx_t Tsize,
+          const sauchar_t *P, saidx_t Psize,
+          const saidx_t *SA, saidx_t SAsize,
+          saidx_t *idx) {
+  saidx_t size, lsize, rsize, half;
+  saidx_t match, lmatch, rmatch;
+  saidx_t llmatch, lrmatch, rlmatch, rrmatch;
+  saidx_t i, j, k;
+  saint_t r;
+
+  if(idx != NULL) { *idx = -1; }
+  if((T == NULL) || (P == NULL) || (SA == NULL) ||
+     (Tsize < 0) || (Psize < 0) || (SAsize < 0)) { return -1; }
+  if((Tsize == 0) || (SAsize == 0)) { return 0; }
+  if(Psize == 0) { if(idx != NULL) { *idx = 0; } return SAsize; }
+
+  for(i = j = k = 0, lmatch = rmatch = 0, size = SAsize, half = size >> 1;
+      0 < size;
+      size = half, half >>= 1) {
+    match = MIN(lmatch, rmatch);
+    r = _compare(T, Tsize, P, Psize, SA[i + half], &match);
+    if(r < 0) {
+      i += half + 1;
+      half -= (size & 1) ^ 1;
+      lmatch = match;
+    } else if(r > 0) {
+      rmatch = match;
+    } else {
+      lsize = half, j = i, rsize = size - half - 1, k = i + half + 1;
+
+      /* left part */
+      for(llmatch = lmatch, lrmatch = match, half = lsize >> 1;
+          0 < lsize;
+          lsize = half, half >>= 1) {
+        lmatch = MIN(llmatch, lrmatch);
+        r = _compare(T, Tsize, P, Psize, SA[j + half], &lmatch);
+        if(r < 0) {
+          j += half + 1;
+          half -= (lsize & 1) ^ 1;
+          llmatch = lmatch;
+        } else {
+          lrmatch = lmatch;
+        }
+      }
+
+      /* right part */
+      for(rlmatch = match, rrmatch = rmatch, half = rsize >> 1;
+          0 < rsize;
+          rsize = half, half >>= 1) {
+        rmatch = MIN(rlmatch, rrmatch);
+        r = _compare(T, Tsize, P, Psize, SA[k + half], &rmatch);
+        if(r <= 0) {
+          k += half + 1;
+          half -= (rsize & 1) ^ 1;
+          rlmatch = rmatch;
+        } else {
+          rrmatch = rmatch;
+        }
+      }
+
+      break;
+    }
+  }
+
+  if(idx != NULL) { *idx = (0 < (k - j)) ? j : i; }
+  return k - j;
+}
+
+/* Search for the character c in the string T. */
+saidx_t
+sa_simplesearch(const sauchar_t *T, saidx_t Tsize,
+                const saidx_t *SA, saidx_t SAsize,
+                saint_t c, saidx_t *idx) {
+  saidx_t size, lsize, rsize, half;
+  saidx_t i, j, k, p;
+  saint_t r;
+
+  if(idx != NULL) { *idx = -1; }
+  if((T == NULL) || (SA == NULL) || (Tsize < 0) || (SAsize < 0)) { return -1; }
+  if((Tsize == 0) || (SAsize == 0)) { return 0; }
+
+  for(i = j = k = 0, size = SAsize, half = size >> 1;
+      0 < size;
+      size = half, half >>= 1) {
+    p = SA[i + half];
+    r = (p < Tsize) ? T[p] - c : -1;
+    if(r < 0) {
+      i += half + 1;
+      half -= (size & 1) ^ 1;
+    } else if(r == 0) {
+      lsize = half, j = i, rsize = size - half - 1, k = i + half + 1;
+
+      /* left part */
+      for(half = lsize >> 1;
+          0 < lsize;
+          lsize = half, half >>= 1) {
+        p = SA[j + half];
+        r = (p < Tsize) ? T[p] - c : -1;
+        if(r < 0) {
+          j += half + 1;
+          half -= (lsize & 1) ^ 1;
+        }
+      }
+
+      /* right part */
+      for(half = rsize >> 1;
+          0 < rsize;
+          rsize = half, half >>= 1) {
+        p = SA[k + half];
+        r = (p < Tsize) ? T[p] - c : -1;
+        if(r <= 0) {
+          k += half + 1;
+          half -= (rsize & 1) ^ 1;
+        }
+      }
+
+      break;
+    }
+  }
+
+  if(idx != NULL) { *idx = (0 < (k - j)) ? j : i; }
+  return k - j;
+}
+#endif

loader/tools/lzsa/src/libdivsufsort/lib/sssort.c

@@ -0,0 +1,815 @@
+/*
+ * sssort.c for libdivsufsort
+ * Copyright (c) 2003-2008 Yuta Mori All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use,
+ * copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following
+ * conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+ * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "divsufsort_private.h"
+
+
+/*- Private Functions -*/
+
+static const saint_t lg_table[256]= {
+ -1,0,1,1,2,2,2,2,3,3,3,3,3,3,3,3,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,
+  5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,
+  6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,
+  6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,
+  7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,
+  7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,
+  7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,
+  7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7
+};
+
+#if (SS_BLOCKSIZE == 0) || (SS_INSERTIONSORT_THRESHOLD < SS_BLOCKSIZE)
+
+static INLINE
+saint_t
+ss_ilg(saidx_t n) {
+#if SS_BLOCKSIZE == 0
+# if defined(BUILD_DIVSUFSORT64)
+  return (n >> 32) ?
+          ((n >> 48) ?
+            ((n >> 56) ?
+              56 + lg_table[(n >> 56) & 0xff] :
+              48 + lg_table[(n >> 48) & 0xff]) :
+            ((n >> 40) ?
+              40 + lg_table[(n >> 40) & 0xff] :
+              32 + lg_table[(n >> 32) & 0xff])) :
+          ((n & 0xffff0000) ?
+            ((n & 0xff000000) ?
+              24 + lg_table[(n >> 24) & 0xff] :
+              16 + lg_table[(n >> 16) & 0xff]) :
+            ((n & 0x0000ff00) ?
+               8 + lg_table[(n >>  8) & 0xff] :
+               0 + lg_table[(n >>  0) & 0xff]));
+# else
+  return (n & 0xffff0000) ?
+          ((n & 0xff000000) ?
+            24 + lg_table[(n >> 24) & 0xff] :
+            16 + lg_table[(n >> 16) & 0xff]) :
+          ((n & 0x0000ff00) ?
+             8 + lg_table[(n >>  8) & 0xff] :
+             0 + lg_table[(n >>  0) & 0xff]);
+# endif
+#elif SS_BLOCKSIZE < 256
+  return lg_table[n];
+#else
+  return (n & 0xff00) ?
+          8 + lg_table[(n >> 8) & 0xff] :
+          0 + lg_table[(n >> 0) & 0xff];
+#endif
+}
+
+#endif /* (SS_BLOCKSIZE == 0) || (SS_INSERTIONSORT_THRESHOLD < SS_BLOCKSIZE) */
+
+#if SS_BLOCKSIZE != 0
+
+static const saint_t sqq_table[256] = {
+  0,  16,  22,  27,  32,  35,  39,  42,  45,  48,  50,  53,  55,  57,  59,  61,
+ 64,  65,  67,  69,  71,  73,  75,  76,  78,  80,  81,  83,  84,  86,  87,  89,
+ 90,  91,  93,  94,  96,  97,  98,  99, 101, 102, 103, 104, 106, 107, 108, 109,
+110, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126,
+128, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142,
+143, 144, 144, 145, 146, 147, 148, 149, 150, 150, 151, 152, 153, 154, 155, 155,
+156, 157, 158, 159, 160, 160, 161, 162, 163, 163, 164, 165, 166, 167, 167, 168,
+169, 170, 170, 171, 172, 173, 173, 174, 175, 176, 176, 177, 178, 178, 179, 180,
+181, 181, 182, 183, 183, 184, 185, 185, 186, 187, 187, 188, 189, 189, 190, 191,
+192, 192, 193, 193, 194, 195, 195, 196, 197, 197, 198, 199, 199, 200, 201, 201,
+202, 203, 203, 204, 204, 205, 206, 206, 207, 208, 208, 209, 209, 210, 211, 211,
+212, 212, 213, 214, 214, 215, 215, 216, 217, 217, 218, 218, 219, 219, 220, 221,
+221, 222, 222, 223, 224, 224, 225, 225, 226, 226, 227, 227, 228, 229, 229, 230,
+230, 231, 231, 232, 232, 233, 234, 234, 235, 235, 236, 236, 237, 237, 238, 238,
+239, 240, 240, 241, 241, 242, 242, 243, 243, 244, 244, 245, 245, 246, 246, 247,
+247, 248, 248, 249, 249, 250, 250, 251, 251, 252, 252, 253, 253, 254, 254, 255
+};
+
+static INLINE
+saidx_t
+ss_isqrt(saidx_t x) {
+  saidx_t y, e;
+
+  if(x >= (SS_BLOCKSIZE * SS_BLOCKSIZE)) { return SS_BLOCKSIZE; }
+  e = (x & 0xffff0000) ?
+        ((x & 0xff000000) ?
+          24 + lg_table[(x >> 24) & 0xff] :
+          16 + lg_table[(x >> 16) & 0xff]) :
+        ((x & 0x0000ff00) ?
+           8 + lg_table[(x >>  8) & 0xff] :
+           0 + lg_table[(x >>  0) & 0xff]);
+
+  if(e >= 16) {
+    y = sqq_table[x >> ((e - 6) - (e & 1))] << ((e >> 1) - 7);
+    if(e >= 24) { y = (y + 1 + x / y) >> 1; }
+    y = (y + 1 + x / y) >> 1;
+  } else if(e >= 8) {
+    y = (sqq_table[x >> ((e - 6) - (e & 1))] >> (7 - (e >> 1))) + 1;
+  } else {
+    return sqq_table[x] >> 4;
+  }
+
+  return (x < (y * y)) ? y - 1 : y;
+}
+
+#endif /* SS_BLOCKSIZE != 0 */
+
+
+/*---------------------------------------------------------------------------*/
+
+/* Compares two suffixes. */
+static INLINE
+saint_t
+ss_compare(const sauchar_t *T,
+           const saidx_t *p1, const saidx_t *p2,
+           saidx_t depth) {
+  const sauchar_t *U1, *U2, *U1n, *U2n;
+
+  for(U1 = T + depth + *p1,
+      U2 = T + depth + *p2,
+      U1n = T + *(p1 + 1) + 2,
+      U2n = T + *(p2 + 1) + 2;
+      (U1 < U1n) && (U2 < U2n) && (*U1 == *U2);
+      ++U1, ++U2) {
+  }
+
+  return U1 < U1n ?
+        (U2 < U2n ? *U1 - *U2 : 1) :
+        (U2 < U2n ? -1 : 0);
+}
+
+
+/*---------------------------------------------------------------------------*/
+
+#if (SS_BLOCKSIZE != 1) && (SS_INSERTIONSORT_THRESHOLD != 1)
+
+/* Insertionsort for small size groups */
+static
+void
+ss_insertionsort(const sauchar_t *T, const saidx_t *PA,
+                 saidx_t *first, saidx_t *last, saidx_t depth) {
+  saidx_t *i, *j;
+  saidx_t t;
+  saint_t r;
+
+  for(i = last - 2; first <= i; --i) {
+    for(t = *i, j = i + 1; 0 < (r = ss_compare(T, PA + t, PA + *j, depth));) {
+      do { *(j - 1) = *j; } while((++j < last) && (*j < 0));
+      if(last <= j) { break; }
+    }
+    if(r == 0) { *j = ~*j; }
+    *(j - 1) = t;
+  }
+}
+
+#endif /* (SS_BLOCKSIZE != 1) && (SS_INSERTIONSORT_THRESHOLD != 1) */
+
+
+/*---------------------------------------------------------------------------*/
+
+#if (SS_BLOCKSIZE == 0) || (SS_INSERTIONSORT_THRESHOLD < SS_BLOCKSIZE)
+
+static INLINE
+void
+ss_fixdown(const sauchar_t *Td, const saidx_t *PA,
+           saidx_t *SA, saidx_t i, saidx_t size) {
+  saidx_t j, k;
+  saidx_t v;
+  saint_t c, d, e;
+
+  for(v = SA[i], c = Td[PA[v]]; (j = 2 * i + 1) < size; SA[i] = SA[k], i = k) {
+    d = Td[PA[SA[k = j++]]];
+    if(d < (e = Td[PA[SA[j]]])) { k = j; d = e; }
+    if(d <= c) { break; }
+  }
+  SA[i] = v;
+}
+
+/* Simple top-down heapsort. */
+static
+void
+ss_heapsort(const sauchar_t *Td, const saidx_t *PA, saidx_t *SA, saidx_t size) {
+  saidx_t i, m;
+  saidx_t t;
+
+  m = size;
+  if((size % 2) == 0) {
+    m--;
+    if(Td[PA[SA[m / 2]]] < Td[PA[SA[m]]]) { SWAP(SA[m], SA[m / 2]); }
+  }
+
+  for(i = m / 2 - 1; 0 <= i; --i) { ss_fixdown(Td, PA, SA, i, m); }
+  if((size % 2) == 0) { SWAP(SA[0], SA[m]); ss_fixdown(Td, PA, SA, 0, m); }
+  for(i = m - 1; 0 < i; --i) {
+    t = SA[0], SA[0] = SA[i];
+    ss_fixdown(Td, PA, SA, 0, i);
+    SA[i] = t;
+  }
+}
+
+
+/*---------------------------------------------------------------------------*/
+
+/* Returns the median of three elements. */
+static INLINE
+saidx_t *
+ss_median3(const sauchar_t *Td, const saidx_t *PA,
+           saidx_t *v1, saidx_t *v2, saidx_t *v3) {
+  saidx_t *t;
+  if(Td[PA[*v1]] > Td[PA[*v2]]) { SWAP(v1, v2); }
+  if(Td[PA[*v2]] > Td[PA[*v3]]) {
+    if(Td[PA[*v1]] > Td[PA[*v3]]) { return v1; }
+    else { return v3; }
+  }
+  return v2;
+}
+
+/* Returns the median of five elements. */
+static INLINE
+saidx_t *
+ss_median5(const sauchar_t *Td, const saidx_t *PA,
+           saidx_t *v1, saidx_t *v2, saidx_t *v3, saidx_t *v4, saidx_t *v5) {
+  saidx_t *t;
+  if(Td[PA[*v2]] > Td[PA[*v3]]) { SWAP(v2, v3); }
+  if(Td[PA[*v4]] > Td[PA[*v5]]) { SWAP(v4, v5); }
+  if(Td[PA[*v2]] > Td[PA[*v4]]) { SWAP(v2, v4); SWAP(v3, v5); }
+  if(Td[PA[*v1]] > Td[PA[*v3]]) { SWAP(v1, v3); }
+  if(Td[PA[*v1]] > Td[PA[*v4]]) { SWAP(v1, v4); SWAP(v3, v5); }
+  if(Td[PA[*v3]] > Td[PA[*v4]]) { return v4; }
+  return v3;
+}
+
+/* Returns the pivot element. */
+static INLINE
+saidx_t *
+ss_pivot(const sauchar_t *Td, const saidx_t *PA, saidx_t *first, saidx_t *last) {
+  saidx_t *middle;
+  saidx_t t;
+
+  t = last - first;
+  middle = first + t / 2;
+
+  if(t <= 512) {
+    if(t <= 32) {
+      return ss_median3(Td, PA, first, middle, last - 1);
+    } else {
+      t >>= 2;
+      return ss_median5(Td, PA, first, first + t, middle, last - 1 - t, last - 1);
+    }
+  }
+  t >>= 3;
+  first  = ss_median3(Td, PA, first, first + t, first + (t << 1));
+  middle = ss_median3(Td, PA, middle - t, middle, middle + t);
+  last   = ss_median3(Td, PA, last - 1 - (t << 1), last - 1 - t, last - 1);
+  return ss_median3(Td, PA, first, middle, last);
+}
+
+
+/*---------------------------------------------------------------------------*/
+
+/* Binary partition for substrings. */
+static INLINE
+saidx_t *
+ss_partition(const saidx_t *PA,
+                    saidx_t *first, saidx_t *last, saidx_t depth) {
+  saidx_t *a, *b;
+  saidx_t t;
+  for(a = first - 1, b = last;;) {
+    for(; (++a < b) && ((PA[*a] + depth) >= (PA[*a + 1] + 1));) { *a = ~*a; }
+    for(; (a < --b) && ((PA[*b] + depth) <  (PA[*b + 1] + 1));) { }
+    if(b <= a) { break; }
+    t = ~*b;
+    *b = *a;
+    *a = t;
+  }
+  if(first < a) { *first = ~*first; }
+  return a;
+}
+
+/* Multikey introsort for medium size groups. */
+static
+void
+ss_mintrosort(const sauchar_t *T, const saidx_t *PA,
+              saidx_t *first, saidx_t *last,
+              saidx_t depth) {
+#define STACK_SIZE SS_MISORT_STACKSIZE
+  struct { saidx_t *a, *b, c; saint_t d; } stack[STACK_SIZE];
+  const sauchar_t *Td;
+  saidx_t *a, *b, *c, *d, *e, *f;
+  saidx_t s, t;
+  saint_t ssize;
+  saint_t limit;
+  saint_t v, x = 0;
+
+  for(ssize = 0, limit = ss_ilg(last - first);;) {
+
+    if((last - first) <= SS_INSERTIONSORT_THRESHOLD) {
+#if 1 < SS_INSERTIONSORT_THRESHOLD
+      if(1 < (last - first)) { ss_insertionsort(T, PA, first, last, depth); }
+#endif
+      STACK_POP(first, last, depth, limit);
+      continue;
+    }
+
+    Td = T + depth;
+    if(limit-- == 0) { ss_heapsort(Td, PA, first, last - first); }
+    if(limit < 0) {
+      for(a = first + 1, v = Td[PA[*first]]; a < last; ++a) {
+        if((x = Td[PA[*a]]) != v) {
+          if(1 < (a - first)) { break; }
+          v = x;
+          first = a;
+        }
+      }
+      if(Td[PA[*first] - 1] < v) {
+        first = ss_partition(PA, first, a, depth);
+      }
+      if((a - first) <= (last - a)) {
+        if(1 < (a - first)) {
+          STACK_PUSH(a, last, depth, -1);
+          last = a, depth += 1, limit = ss_ilg(a - first);
+        } else {
+          first = a, limit = -1;
+        }
+      } else {
+        if(1 < (last - a)) {
+          STACK_PUSH(first, a, depth + 1, ss_ilg(a - first));
+          first = a, limit = -1;
+        } else {
+          last = a, depth += 1, limit = ss_ilg(a - first);
+        }
+      }
+      continue;
+    }
+
+    /* choose pivot */
+    a = ss_pivot(Td, PA, first, last);
+    v = Td[PA[*a]];
+    SWAP(*first, *a);
+
+    /* partition */
+    for(b = first; (++b < last) && ((x = Td[PA[*b]]) == v);) { }
+    if(((a = b) < last) && (x < v)) {
+      for(; (++b < last) && ((x = Td[PA[*b]]) <= v);) {
+        if(x == v) { SWAP(*b, *a); ++a; }
+      }
+    }
+    for(c = last; (b < --c) && ((x = Td[PA[*c]]) == v);) { }
+    if((b < (d = c)) && (x > v)) {
+      for(; (b < --c) && ((x = Td[PA[*c]]) >= v);) {
+        if(x == v) { SWAP(*c, *d); --d; }
+      }
+    }
+    for(; b < c;) {
+      SWAP(*b, *c);
+      for(; (++b < c) && ((x = Td[PA[*b]]) <= v);) {
+        if(x == v) { SWAP(*b, *a); ++a; }
+      }
+      for(; (b < --c) && ((x = Td[PA[*c]]) >= v);) {
+        if(x == v) { SWAP(*c, *d); --d; }
+      }
+    }
+
+    if(a <= d) {
+      c = b - 1;
+
+      if((s = a - first) > (t = b - a)) { s = t; }
+      for(e = first, f = b - s; 0 < s; --s, ++e, ++f) { SWAP(*e, *f); }
+      if((s = d - c) > (t = last - d - 1)) { s = t; }
+      for(e = b, f = last - s; 0 < s; --s, ++e, ++f) { SWAP(*e, *f); }
+
+      a = first + (b - a), c = last - (d - c);
+      b = (v <= Td[PA[*a] - 1]) ? a : ss_partition(PA, a, c, depth);
+
+      if((a - first) <= (last - c)) {
+        if((last - c) <= (c - b)) {
+          STACK_PUSH(b, c, depth + 1, ss_ilg(c - b));
+          STACK_PUSH(c, last, depth, limit);
+          last = a;
+        } else if((a - first) <= (c - b)) {
+          STACK_PUSH(c, last, depth, limit);
+          STACK_PUSH(b, c, depth + 1, ss_ilg(c - b));
+          last = a;
+        } else {
+          STACK_PUSH(c, last, depth, limit);
+          STACK_PUSH(first, a, depth, limit);
+          first = b, last = c, depth += 1, limit = ss_ilg(c - b);
+        }
+      } else {
+        if((a - first) <= (c - b)) {
+          STACK_PUSH(b, c, depth + 1, ss_ilg(c - b));
+          STACK_PUSH(first, a, depth, limit);
+          first = c;
+        } else if((last - c) <= (c - b)) {
+          STACK_PUSH(first, a, depth, limit);
+          STACK_PUSH(b, c, depth + 1, ss_ilg(c - b));
+          first = c;
+        } else {
+          STACK_PUSH(first, a, depth, limit);
+          STACK_PUSH(c, last, depth, limit);
+          first = b, last = c, depth += 1, limit = ss_ilg(c - b);
+        }
+      }
+    } else {
+      limit += 1;
+      if(Td[PA[*first] - 1] < v) {
+        first = ss_partition(PA, first, last, depth);
+        limit = ss_ilg(last - first);
+      }
+      depth += 1;
+    }
+  }
+#undef STACK_SIZE
+}
+
+#endif /* (SS_BLOCKSIZE == 0) || (SS_INSERTIONSORT_THRESHOLD < SS_BLOCKSIZE) */
+
+
+/*---------------------------------------------------------------------------*/
+
+#if SS_BLOCKSIZE != 0
+
+static INLINE
+void
+ss_blockswap(saidx_t *a, saidx_t *b, saidx_t n) {
+  saidx_t t;
+  for(; 0 < n; --n, ++a, ++b) {
+    t = *a, *a = *b, *b = t;
+  }
+}
+
+static INLINE
+void
+ss_rotate(saidx_t *first, saidx_t *middle, saidx_t *last) {
+  saidx_t *a, *b, t;
+  saidx_t l, r;
+  l = middle - first, r = last - middle;
+  for(; (0 < l) && (0 < r);) {
+    if(l == r) { ss_blockswap(first, middle, l); break; }
+    if(l < r) {
+      a = last - 1, b = middle - 1;
+      t = *a;
+      do {
+        *a-- = *b, *b-- = *a;
+        if(b < first) {
+          *a = t;
+          last = a;
+          if((r -= l + 1) <= l) { break; }
+          a -= 1, b = middle - 1;
+          t = *a;
+        }
+      } while(1);
+    } else {
+      a = first, b = middle;
+      t = *a;
+      do {
+        *a++ = *b, *b++ = *a;
+        if(last <= b) {
+          *a = t;
+          first = a + 1;
+          if((l -= r + 1) <= r) { break; }
+          a += 1, b = middle;
+          t = *a;
+        }
+      } while(1);
+    }
+  }
+}
+
+
+/*---------------------------------------------------------------------------*/
+
+static
+void
+ss_inplacemerge(const sauchar_t *T, const saidx_t *PA,
+                saidx_t *first, saidx_t *middle, saidx_t *last,
+                saidx_t depth) {
+  const saidx_t *p;
+  saidx_t *a, *b;
+  saidx_t len, half;
+  saint_t q, r;
+  saint_t x;
+
+  for(;;) {
+    if(*(last - 1) < 0) { x = 1; p = PA + ~*(last - 1); }
+    else                { x = 0; p = PA +  *(last - 1); }
+    for(a = first, len = middle - first, half = len >> 1, r = -1;
+        0 < len;
+        len = half, half >>= 1) {
+      b = a + half;
+      q = ss_compare(T, PA + ((0 <= *b) ? *b : ~*b), p, depth);
+      if(q < 0) {
+        a = b + 1;
+        half -= (len & 1) ^ 1;
+      } else {
+        r = q;
+      }
+    }
+    if(a < middle) {
+      if(r == 0) { *a = ~*a; }
+      ss_rotate(a, middle, last);
+      last -= middle - a;
+      middle = a;
+      if(first == middle) { break; }
+    }
+    --last;
+    if(x != 0) { while(*--last < 0) { } }
+    if(middle == last) { break; }
+  }
+}
+
+
+/*---------------------------------------------------------------------------*/
+
+/* Merge-forward with internal buffer. */
+static
+void
+ss_mergeforward(const sauchar_t *T, const saidx_t *PA,
+                saidx_t *first, saidx_t *middle, saidx_t *last,
+                saidx_t *buf, saidx_t depth) {
+  saidx_t *a, *b, *c, *bufend;
+  saidx_t t;
+  saint_t r;
+
+  bufend = buf + (middle - first) - 1;
+  ss_blockswap(buf, first, middle - first);
+
+  for(t = *(a = first), b = buf, c = middle;;) {
+    r = ss_compare(T, PA + *b, PA + *c, depth);
+    if(r < 0) {
+      do {
+        *a++ = *b;
+        if(bufend <= b) { *bufend = t; return; }
+        *b++ = *a;
+      } while(*b < 0);
+    } else if(r > 0) {
+      do {
+        *a++ = *c, *c++ = *a;
+        if(last <= c) {
+          while(b < bufend) { *a++ = *b, *b++ = *a; }
+          *a = *b, *b = t;
+          return;
+        }
+      } while(*c < 0);
+    } else {
+      *c = ~*c;
+      do {
+        *a++ = *b;
+        if(bufend <= b) { *bufend = t; return; }
+        *b++ = *a;
+      } while(*b < 0);
+
+      do {
+        *a++ = *c, *c++ = *a;
+        if(last <= c) {
+          while(b < bufend) { *a++ = *b, *b++ = *a; }
+          *a = *b, *b = t;
+          return;
+        }
+      } while(*c < 0);
+    }
+  }
+}
+
+/* Merge-backward with internal buffer. */
+static
+void
+ss_mergebackward(const sauchar_t *T, const saidx_t *PA,
+                 saidx_t *first, saidx_t *middle, saidx_t *last,
+                 saidx_t *buf, saidx_t depth) {
+  const saidx_t *p1, *p2;
+  saidx_t *a, *b, *c, *bufend;
+  saidx_t t;
+  saint_t r;
+  saint_t x;
+
+  bufend = buf + (last - middle) - 1;
+  ss_blockswap(buf, middle, last - middle);
+
+  x = 0;
+  if(*bufend < 0)       { p1 = PA + ~*bufend; x |= 1; }
+  else                  { p1 = PA +  *bufend; }
+  if(*(middle - 1) < 0) { p2 = PA + ~*(middle - 1); x |= 2; }
+  else                  { p2 = PA +  *(middle - 1); }
+  for(t = *(a = last - 1), b = bufend, c = middle - 1;;) {
+    r = ss_compare(T, p1, p2, depth);
+    if(0 < r) {
+      if(x & 1) { do { *a-- = *b, *b-- = *a; } while(*b < 0); x ^= 1; }
+      *a-- = *b;
+      if(b <= buf) { *buf = t; break; }
+      *b-- = *a;
+      if(*b < 0) { p1 = PA + ~*b; x |= 1; }
+      else       { p1 = PA +  *b; }
+    } else if(r < 0) {
+      if(x & 2) { do { *a-- = *c, *c-- = *a; } while(*c < 0); x ^= 2; }
+      *a-- = *c, *c-- = *a;
+      if(c < first) {
+        while(buf < b) { *a-- = *b, *b-- = *a; }
+        *a = *b, *b = t;
+        break;
+      }
+      if(*c < 0) { p2 = PA + ~*c; x |= 2; }
+      else       { p2 = PA +  *c; }
+    } else {
+      if(x & 1) { do { *a-- = *b, *b-- = *a; } while(*b < 0); x ^= 1; }
+      *a-- = ~*b;
+      if(b <= buf) { *buf = t; break; }
+      *b-- = *a;
+      if(x & 2) { do { *a-- = *c, *c-- = *a; } while(*c < 0); x ^= 2; }
+      *a-- = *c, *c-- = *a;
+      if(c < first) {
+        while(buf < b) { *a-- = *b, *b-- = *a; }
+        *a = *b, *b = t;
+        break;
+      }
+      if(*b < 0) { p1 = PA + ~*b; x |= 1; }
+      else       { p1 = PA +  *b; }
+      if(*c < 0) { p2 = PA + ~*c; x |= 2; }
+      else       { p2 = PA +  *c; }
+    }
+  }
+}
+
+/* D&C based merge. */
+static
+void
+ss_swapmerge(const sauchar_t *T, const saidx_t *PA,
+             saidx_t *first, saidx_t *middle, saidx_t *last,
+             saidx_t *buf, saidx_t bufsize, saidx_t depth) {
+#define STACK_SIZE SS_SMERGE_STACKSIZE
+#define GETIDX(a) ((0 <= (a)) ? (a) : (~(a)))
+#define MERGE_CHECK(a, b, c)\
+  do {\
+    if(((c) & 1) ||\
+       (((c) & 2) && (ss_compare(T, PA + GETIDX(*((a) - 1)), PA + *(a), depth) == 0))) {\
+      *(a) = ~*(a);\
+    }\
+    if(((c) & 4) && ((ss_compare(T, PA + GETIDX(*((b) - 1)), PA + *(b), depth) == 0))) {\
+      *(b) = ~*(b);\
+    }\
+  } while(0)
+  struct { saidx_t *a, *b, *c; saint_t d; } stack[STACK_SIZE];
+  saidx_t *l, *r, *lm, *rm;
+  saidx_t m, len, half;
+  saint_t ssize;
+  saint_t check, next;
+
+  for(check = 0, ssize = 0;;) {
+    if((last - middle) <= bufsize) {
+      if((first < middle) && (middle < last)) {
+        ss_mergebackward(T, PA, first, middle, last, buf, depth);
+      }
+      MERGE_CHECK(first, last, check);
+      STACK_POP(first, middle, last, check);
+      continue;
+    }
+
+    if((middle - first) <= bufsize) {
+      if(first < middle) {
+        ss_mergeforward(T, PA, first, middle, last, buf, depth);
+      }
+      MERGE_CHECK(first, last, check);
+      STACK_POP(first, middle, last, check);
+      continue;
+    }
+
+    for(m = 0, len = MIN(middle - first, last - middle), half = len >> 1;
+        0 < len;
+        len = half, half >>= 1) {
+      if(ss_compare(T, PA + GETIDX(*(middle + m + half)),
+                       PA + GETIDX(*(middle - m - half - 1)), depth) < 0) {
+        m += half + 1;
+        half -= (len & 1) ^ 1;
+      }
+    }
+
+    if(0 < m) {
+      lm = middle - m, rm = middle + m;
+      ss_blockswap(lm, middle, m);
+      l = r = middle, next = 0;
+      if(rm < last) {
+        if(*rm < 0) {
+          *rm = ~*rm;
+          if(first < lm) { for(; *--l < 0;) { } next |= 4; }
+          next |= 1;
+        } else if(first < lm) {
+          for(; *r < 0; ++r) { }
+          next |= 2;
+        }
+      }
+
+      if((l - first) <= (last - r)) {
+        STACK_PUSH(r, rm, last, (next & 3) | (check & 4));
+        middle = lm, last = l, check = (check & 3) | (next & 4);
+      } else {
+        if((next & 2) && (r == middle)) { next ^= 6; }
+        STACK_PUSH(first, lm, l, (check & 3) | (next & 4));
+        first = r, middle = rm, check = (next & 3) | (check & 4);
+      }
+    } else {
+      if(ss_compare(T, PA + GETIDX(*(middle - 1)), PA + *middle, depth) == 0) {
+        *middle = ~*middle;
+      }
+      MERGE_CHECK(first, last, check);
+      STACK_POP(first, middle, last, check);
+    }
+  }
+#undef STACK_SIZE
+}
+
+#endif /* SS_BLOCKSIZE != 0 */
+
+
+/*---------------------------------------------------------------------------*/
+
+/*- Function -*/
+
+/* Substring sort */
+void
+sssort(const sauchar_t *T, const saidx_t *PA,
+       saidx_t *first, saidx_t *last,
+       saidx_t *buf, saidx_t bufsize,
+       saidx_t depth, saidx_t n, saint_t lastsuffix) {
+  saidx_t *a;
+#if SS_BLOCKSIZE != 0
+  saidx_t *b, *middle, *curbuf;
+  saidx_t j, k, curbufsize, limit;
+#endif
+  saidx_t i;
+
+  if(lastsuffix != 0) { ++first; }
+
+#if SS_BLOCKSIZE == 0
+  ss_mintrosort(T, PA, first, last, depth);
+#else
+  if((bufsize < SS_BLOCKSIZE) &&
+      (bufsize < (last - first)) &&
+      (bufsize < (limit = ss_isqrt(last - first)))) {
+    if(SS_BLOCKSIZE < limit) { limit = SS_BLOCKSIZE; }
+    buf = middle = last - limit, bufsize = limit;
+  } else {
+    middle = last, limit = 0;
+  }
+  for(a = first, i = 0; SS_BLOCKSIZE < (middle - a); a += SS_BLOCKSIZE, ++i) {
+#if SS_INSERTIONSORT_THRESHOLD < SS_BLOCKSIZE
+    ss_mintrosort(T, PA, a, a + SS_BLOCKSIZE, depth);
+#elif 1 < SS_BLOCKSIZE
+    ss_insertionsort(T, PA, a, a + SS_BLOCKSIZE, depth);
+#endif
+    curbufsize = last - (a + SS_BLOCKSIZE);
+    curbuf = a + SS_BLOCKSIZE;
+    if(curbufsize <= bufsize) { curbufsize = bufsize, curbuf = buf; }
+    for(b = a, k = SS_BLOCKSIZE, j = i; j & 1; b -= k, k <<= 1, j >>= 1) {
+      ss_swapmerge(T, PA, b - k, b, b + k, curbuf, curbufsize, depth);
+    }
+  }
+#if SS_INSERTIONSORT_THRESHOLD < SS_BLOCKSIZE
+  ss_mintrosort(T, PA, a, middle, depth);
+#elif 1 < SS_BLOCKSIZE
+  ss_insertionsort(T, PA, a, middle, depth);
+#endif
+  for(k = SS_BLOCKSIZE; i != 0; k <<= 1, i >>= 1) {
+    if(i & 1) {
+      ss_swapmerge(T, PA, a - k, a, middle, buf, bufsize, depth);
+      a -= k;
+    }
+  }
+  if(limit != 0) {
+#if SS_INSERTIONSORT_THRESHOLD < SS_BLOCKSIZE
+    ss_mintrosort(T, PA, middle, last, depth);
+#elif 1 < SS_BLOCKSIZE
+    ss_insertionsort(T, PA, middle, last, depth);
+#endif
+    ss_inplacemerge(T, PA, first, middle, last, depth);
+  }
+#endif
+
+  if(lastsuffix != 0) {
+    /* Insert last type B* suffix. */
+    saidx_t PAi[2]; PAi[0] = PA[*(first - 1)], PAi[1] = n - 2;
+    for(a = first, i = *(first - 1);
+        (a < last) && ((*a < 0) || (0 < ss_compare(T, &(PAi[0]), PA + *a, depth)));
+        ++a) {
+      *(a - 1) = *a;
+    }
+    *(a - 1) = i;
+  }
+}

loader/tools/lzsa/src/libdivsufsort/lib/trsort.c

@@ -0,0 +1,586 @@
+/*
+ * trsort.c for libdivsufsort
+ * Copyright (c) 2003-2008 Yuta Mori All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use,
+ * copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following
+ * conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+ * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "divsufsort_private.h"
+
+
+/*- Private Functions -*/
+
+static const saint_t lg_table[256]= {
+ -1,0,1,1,2,2,2,2,3,3,3,3,3,3,3,3,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,
+  5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,
+  6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,
+  6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,
+  7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,
+  7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,
+  7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,
+  7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7
+};
+
+static INLINE
+saint_t
+tr_ilg(saidx_t n) {
+#if defined(BUILD_DIVSUFSORT64)
+  return (n >> 32) ?
+          ((n >> 48) ?
+            ((n >> 56) ?
+              56 + lg_table[(n >> 56) & 0xff] :
+              48 + lg_table[(n >> 48) & 0xff]) :
+            ((n >> 40) ?
+              40 + lg_table[(n >> 40) & 0xff] :
+              32 + lg_table[(n >> 32) & 0xff])) :
+          ((n & 0xffff0000) ?
+            ((n & 0xff000000) ?
+              24 + lg_table[(n >> 24) & 0xff] :
+              16 + lg_table[(n >> 16) & 0xff]) :
+            ((n & 0x0000ff00) ?
+               8 + lg_table[(n >>  8) & 0xff] :
+               0 + lg_table[(n >>  0) & 0xff]));
+#else
+  return (n & 0xffff0000) ?
+          ((n & 0xff000000) ?
+            24 + lg_table[(n >> 24) & 0xff] :
+            16 + lg_table[(n >> 16) & 0xff]) :
+          ((n & 0x0000ff00) ?
+             8 + lg_table[(n >>  8) & 0xff] :
+             0 + lg_table[(n >>  0) & 0xff]);
+#endif
+}
+
+
+/*---------------------------------------------------------------------------*/
+
+/* Simple insertionsort for small size groups. */
+static
+void
+tr_insertionsort(const saidx_t *ISAd, saidx_t *first, saidx_t *last) {
+  saidx_t *a, *b;
+  saidx_t t, r;
+
+  for(a = first + 1; a < last; ++a) {
+    for(t = *a, b = a - 1; 0 > (r = ISAd[t] - ISAd[*b]);) {
+      do { *(b + 1) = *b; } while((first <= --b) && (*b < 0));
+      if(b < first) { break; }
+    }
+    if(r == 0) { *b = ~*b; }
+    *(b + 1) = t;
+  }
+}
+
+
+/*---------------------------------------------------------------------------*/
+
+static INLINE
+void
+tr_fixdown(const saidx_t *ISAd, saidx_t *SA, saidx_t i, saidx_t size) {
+  saidx_t j, k;
+  saidx_t v;
+  saidx_t c, d, e;
+
+  for(v = SA[i], c = ISAd[v]; (j = 2 * i + 1) < size; SA[i] = SA[k], i = k) {
+    d = ISAd[SA[k = j++]];
+    if(d < (e = ISAd[SA[j]])) { k = j; d = e; }
+    if(d <= c) { break; }
+  }
+  SA[i] = v;
+}
+
+/* Simple top-down heapsort. */
+static
+void
+tr_heapsort(const saidx_t *ISAd, saidx_t *SA, saidx_t size) {
+  saidx_t i, m;
+  saidx_t t;
+
+  m = size;
+  if((size % 2) == 0) {
+    m--;
+    if(ISAd[SA[m / 2]] < ISAd[SA[m]]) { SWAP(SA[m], SA[m / 2]); }
+  }
+
+  for(i = m / 2 - 1; 0 <= i; --i) { tr_fixdown(ISAd, SA, i, m); }
+  if((size % 2) == 0) { SWAP(SA[0], SA[m]); tr_fixdown(ISAd, SA, 0, m); }
+  for(i = m - 1; 0 < i; --i) {
+    t = SA[0], SA[0] = SA[i];
+    tr_fixdown(ISAd, SA, 0, i);
+    SA[i] = t;
+  }
+}
+
+
+/*---------------------------------------------------------------------------*/
+
+/* Returns the median of three elements. */
+static INLINE
+saidx_t *
+tr_median3(const saidx_t *ISAd, saidx_t *v1, saidx_t *v2, saidx_t *v3) {
+  saidx_t *t;
+  if(ISAd[*v1] > ISAd[*v2]) { SWAP(v1, v2); }
+  if(ISAd[*v2] > ISAd[*v3]) {
+    if(ISAd[*v1] > ISAd[*v3]) { return v1; }
+    else { return v3; }
+  }
+  return v2;
+}
+
+/* Returns the median of five elements. */
+static INLINE
+saidx_t *
+tr_median5(const saidx_t *ISAd,
+           saidx_t *v1, saidx_t *v2, saidx_t *v3, saidx_t *v4, saidx_t *v5) {
+  saidx_t *t;
+  if(ISAd[*v2] > ISAd[*v3]) { SWAP(v2, v3); }
+  if(ISAd[*v4] > ISAd[*v5]) { SWAP(v4, v5); }
+  if(ISAd[*v2] > ISAd[*v4]) { SWAP(v2, v4); SWAP(v3, v5); }
+  if(ISAd[*v1] > ISAd[*v3]) { SWAP(v1, v3); }
+  if(ISAd[*v1] > ISAd[*v4]) { SWAP(v1, v4); SWAP(v3, v5); }
+  if(ISAd[*v3] > ISAd[*v4]) { return v4; }
+  return v3;
+}
+
+/* Returns the pivot element. */
+static INLINE
+saidx_t *
+tr_pivot(const saidx_t *ISAd, saidx_t *first, saidx_t *last) {
+  saidx_t *middle;
+  saidx_t t;
+
+  t = last - first;
+  middle = first + t / 2;
+
+  if(t <= 512) {
+    if(t <= 32) {
+      return tr_median3(ISAd, first, middle, last - 1);
+    } else {
+      t >>= 2;
+      return tr_median5(ISAd, first, first + t, middle, last - 1 - t, last - 1);
+    }
+  }
+  t >>= 3;
+  first  = tr_median3(ISAd, first, first + t, first + (t << 1));
+  middle = tr_median3(ISAd, middle - t, middle, middle + t);
+  last   = tr_median3(ISAd, last - 1 - (t << 1), last - 1 - t, last - 1);
+  return tr_median3(ISAd, first, middle, last);
+}
+
+
+/*---------------------------------------------------------------------------*/
+
+typedef struct _trbudget_t trbudget_t;
+struct _trbudget_t {
+  saidx_t chance;
+  saidx_t remain;
+  saidx_t incval;
+  saidx_t count;
+};
+
+static INLINE
+void
+trbudget_init(trbudget_t *budget, saidx_t chance, saidx_t incval) {
+  budget->chance = chance;
+  budget->remain = budget->incval = incval;
+}
+
+static INLINE
+saint_t
+trbudget_check(trbudget_t *budget, saidx_t size) {
+  if(size <= budget->remain) { budget->remain -= size; return 1; }
+  if(budget->chance == 0) { budget->count += size; return 0; }
+  budget->remain += budget->incval - size;
+  budget->chance -= 1;
+  return 1;
+}
+
+
+/*---------------------------------------------------------------------------*/
+
+static INLINE
+void
+tr_partition(const saidx_t *ISAd,
+             saidx_t *first, saidx_t *middle, saidx_t *last,
+             saidx_t **pa, saidx_t **pb, saidx_t v) {
+  saidx_t *a, *b, *c, *d, *e, *f;
+  saidx_t t, s;
+  saidx_t x = 0;
+
+  for(b = middle - 1; (++b < last) && ((x = ISAd[*b]) == v);) { }
+  if(((a = b) < last) && (x < v)) {
+    for(; (++b < last) && ((x = ISAd[*b]) <= v);) {
+      if(x == v) { SWAP(*b, *a); ++a; }
+    }
+  }
+  for(c = last; (b < --c) && ((x = ISAd[*c]) == v);) { }
+  if((b < (d = c)) && (x > v)) {
+    for(; (b < --c) && ((x = ISAd[*c]) >= v);) {
+      if(x == v) { SWAP(*c, *d); --d; }
+    }
+  }
+  for(; b < c;) {
+    SWAP(*b, *c);
+    for(; (++b < c) && ((x = ISAd[*b]) <= v);) {
+      if(x == v) { SWAP(*b, *a); ++a; }
+    }
+    for(; (b < --c) && ((x = ISAd[*c]) >= v);) {
+      if(x == v) { SWAP(*c, *d); --d; }
+    }
+  }
+
+  if(a <= d) {
+    c = b - 1;
+    if((s = a - first) > (t = b - a)) { s = t; }
+    for(e = first, f = b - s; 0 < s; --s, ++e, ++f) { SWAP(*e, *f); }
+    if((s = d - c) > (t = last - d - 1)) { s = t; }
+    for(e = b, f = last - s; 0 < s; --s, ++e, ++f) { SWAP(*e, *f); }
+    first += (b - a), last -= (d - c);
+  }
+  *pa = first, *pb = last;
+}
+
+static
+void
+tr_copy(saidx_t *ISA, const saidx_t *SA,
+        saidx_t *first, saidx_t *a, saidx_t *b, saidx_t *last,
+        saidx_t depth) {
+  /* sort suffixes of middle partition
+     by using sorted order of suffixes of left and right partition. */
+  saidx_t *c, *d, *e;
+  saidx_t s, v;
+
+  v = b - SA - 1;
+  for(c = first, d = a - 1; c <= d; ++c) {
+    if((0 <= (s = *c - depth)) && (ISA[s] == v)) {
+      *++d = s;
+      ISA[s] = d - SA;
+    }
+  }
+  for(c = last - 1, e = d + 1, d = b; e < d; --c) {
+    if((0 <= (s = *c - depth)) && (ISA[s] == v)) {
+      *--d = s;
+      ISA[s] = d - SA;
+    }
+  }
+}
+
+static
+void
+tr_partialcopy(saidx_t *ISA, const saidx_t *SA,
+               saidx_t *first, saidx_t *a, saidx_t *b, saidx_t *last,
+               saidx_t depth) {
+  saidx_t *c, *d, *e;
+  saidx_t s, v;
+  saidx_t rank, lastrank, newrank = -1;
+
+  v = b - SA - 1;
+  lastrank = -1;
+  for(c = first, d = a - 1; c <= d; ++c) {
+    if((0 <= (s = *c - depth)) && (ISA[s] == v)) {
+      *++d = s;
+      rank = ISA[s + depth];
+      if(lastrank != rank) { lastrank = rank; newrank = d - SA; }
+      ISA[s] = newrank;
+    }
+  }
+
+  lastrank = -1;
+  for(e = d; first <= e; --e) {
+    rank = ISA[*e];
+    if(lastrank != rank) { lastrank = rank; newrank = e - SA; }
+    if(newrank != rank) { ISA[*e] = newrank; }
+  }
+
+  lastrank = -1;
+  for(c = last - 1, e = d + 1, d = b; e < d; --c) {
+    if((0 <= (s = *c - depth)) && (ISA[s] == v)) {
+      *--d = s;
+      rank = ISA[s + depth];
+      if(lastrank != rank) { lastrank = rank; newrank = d - SA; }
+      ISA[s] = newrank;
+    }
+  }
+}
+
+static
+void
+tr_introsort(saidx_t *ISA, const saidx_t *ISAd,
+             saidx_t *SA, saidx_t *first, saidx_t *last,
+             trbudget_t *budget) {
+#define STACK_SIZE TR_STACKSIZE
+  struct { const saidx_t *a; saidx_t *b, *c; saint_t d, e; }stack[STACK_SIZE];
+  saidx_t *a, *b, *c;
+  saidx_t t;
+  saidx_t v, x = 0;
+  saidx_t incr = ISAd - ISA;
+  saint_t limit, next;
+  saint_t ssize, trlink = -1;
+
+  for(ssize = 0, limit = tr_ilg(last - first);;) {
+
+    if(limit < 0) {
+      if(limit == -1) {
+        /* tandem repeat partition */
+        tr_partition(ISAd - incr, first, first, last, &a, &b, last - SA - 1);
+
+        /* update ranks */
+        if(a < last) {
+          for(c = first, v = a - SA - 1; c < a; ++c) { ISA[*c] = v; }
+        }
+        if(b < last) {
+          for(c = a, v = b - SA - 1; c < b; ++c) { ISA[*c] = v; }
+        }
+
+        /* push */
+        if(1 < (b - a)) {
+          STACK_PUSH5(NULL, a, b, 0, 0);
+          STACK_PUSH5(ISAd - incr, first, last, -2, trlink);
+          trlink = ssize - 2;
+        }
+        if((a - first) <= (last - b)) {
+          if(1 < (a - first)) {
+            STACK_PUSH5(ISAd, b, last, tr_ilg(last - b), trlink);
+            last = a, limit = tr_ilg(a - first);
+          } else if(1 < (last - b)) {
+            first = b, limit = tr_ilg(last - b);
+          } else {
+            STACK_POP5(ISAd, first, last, limit, trlink);
+          }
+        } else {
+          if(1 < (last - b)) {
+            STACK_PUSH5(ISAd, first, a, tr_ilg(a - first), trlink);
+            first = b, limit = tr_ilg(last - b);
+          } else if(1 < (a - first)) {
+            last = a, limit = tr_ilg(a - first);
+          } else {
+            STACK_POP5(ISAd, first, last, limit, trlink);
+          }
+        }
+      } else if(limit == -2) {
+        /* tandem repeat copy */
+        a = stack[--ssize].b, b = stack[ssize].c;
+        if(stack[ssize].d == 0) {
+          tr_copy(ISA, SA, first, a, b, last, ISAd - ISA);
+        } else {
+          if(0 <= trlink) { stack[trlink].d = -1; }
+          tr_partialcopy(ISA, SA, first, a, b, last, ISAd - ISA);
+        }
+        STACK_POP5(ISAd, first, last, limit, trlink);
+      } else {
+        /* sorted partition */
+        if(0 <= *first) {
+          a = first;
+          do { ISA[*a] = a - SA; } while((++a < last) && (0 <= *a));
+          first = a;
+        }
+        if(first < last) {
+          a = first; do { *a = ~*a; } while(*++a < 0);
+          next = (ISA[*a] != ISAd[*a]) ? tr_ilg(a - first + 1) : -1;
+          if(++a < last) { for(b = first, v = a - SA - 1; b < a; ++b) { ISA[*b] = v; } }
+
+          /* push */
+          if(trbudget_check(budget, a - first)) {
+            if((a - first) <= (last - a)) {
+              STACK_PUSH5(ISAd, a, last, -3, trlink);
+              ISAd += incr, last = a, limit = next;
+            } else {
+              if(1 < (last - a)) {
+                STACK_PUSH5(ISAd + incr, first, a, next, trlink);
+                first = a, limit = -3;
+              } else {
+                ISAd += incr, last = a, limit = next;
+              }
+            }
+          } else {
+            if(0 <= trlink) { stack[trlink].d = -1; }
+            if(1 < (last - a)) {
+              first = a, limit = -3;
+            } else {
+              STACK_POP5(ISAd, first, last, limit, trlink);
+            }
+          }
+        } else {
+          STACK_POP5(ISAd, first, last, limit, trlink);
+        }
+      }
+      continue;
+    }
+
+    if((last - first) <= TR_INSERTIONSORT_THRESHOLD) {
+      tr_insertionsort(ISAd, first, last);
+      limit = -3;
+      continue;
+    }
+
+    if(limit-- == 0) {
+      tr_heapsort(ISAd, first, last - first);
+      for(a = last - 1; first < a; a = b) {
+        for(x = ISAd[*a], b = a - 1; (first <= b) && (ISAd[*b] == x); --b) { *b = ~*b; }
+      }
+      limit = -3;
+      continue;
+    }
+
+    /* choose pivot */
+    a = tr_pivot(ISAd, first, last);
+    SWAP(*first, *a);
+    v = ISAd[*first];
+
+    /* partition */
+    tr_partition(ISAd, first, first + 1, last, &a, &b, v);
+    if((last - first) != (b - a)) {
+      next = (ISA[*a] != v) ? tr_ilg(b - a) : -1;
+
+      /* update ranks */
+      for(c = first, v = a - SA - 1; c < a; ++c) { ISA[*c] = v; }
+      if(b < last) { for(c = a, v = b - SA - 1; c < b; ++c) { ISA[*c] = v; } }
+
+      /* push */
+      if((1 < (b - a)) && (trbudget_check(budget, b - a))) {
+        if((a - first) <= (last - b)) {
+          if((last - b) <= (b - a)) {
+            if(1 < (a - first)) {
+              STACK_PUSH5(ISAd + incr, a, b, next, trlink);
+              STACK_PUSH5(ISAd, b, last, limit, trlink);
+              last = a;
+            } else if(1 < (last - b)) {
+              STACK_PUSH5(ISAd + incr, a, b, next, trlink);
+              first = b;
+            } else {
+              ISAd += incr, first = a, last = b, limit = next;
+            }
+          } else if((a - first) <= (b - a)) {
+            if(1 < (a - first)) {
+              STACK_PUSH5(ISAd, b, last, limit, trlink);
+              STACK_PUSH5(ISAd + incr, a, b, next, trlink);
+              last = a;
+            } else {
+              STACK_PUSH5(ISAd, b, last, limit, trlink);
+              ISAd += incr, first = a, last = b, limit = next;
+            }
+          } else {
+            STACK_PUSH5(ISAd, b, last, limit, trlink);
+            STACK_PUSH5(ISAd, first, a, limit, trlink);
+            ISAd += incr, first = a, last = b, limit = next;
+          }
+        } else {
+          if((a - first) <= (b - a)) {
+            if(1 < (last - b)) {
+              STACK_PUSH5(ISAd + incr, a, b, next, trlink);
+              STACK_PUSH5(ISAd, first, a, limit, trlink);
+              first = b;
+            } else if(1 < (a - first)) {
+              STACK_PUSH5(ISAd + incr, a, b, next, trlink);
+              last = a;
+            } else {
+              ISAd += incr, first = a, last = b, limit = next;
+            }
+          } else if((last - b) <= (b - a)) {
+            if(1 < (last - b)) {
+              STACK_PUSH5(ISAd, first, a, limit, trlink);
+              STACK_PUSH5(ISAd + incr, a, b, next, trlink);
+              first = b;
+            } else {
+              STACK_PUSH5(ISAd, first, a, limit, trlink);
+              ISAd += incr, first = a, last = b, limit = next;
+            }
+          } else {
+            STACK_PUSH5(ISAd, first, a, limit, trlink);
+            STACK_PUSH5(ISAd, b, last, limit, trlink);
+            ISAd += incr, first = a, last = b, limit = next;
+          }
+        }
+      } else {
+        if((1 < (b - a)) && (0 <= trlink)) { stack[trlink].d = -1; }
+        if((a - first) <= (last - b)) {
+          if(1 < (a - first)) {
+            STACK_PUSH5(ISAd, b, last, limit, trlink);
+            last = a;
+          } else if(1 < (last - b)) {
+            first = b;
+          } else {
+            STACK_POP5(ISAd, first, last, limit, trlink);
+          }
+        } else {
+          if(1 < (last - b)) {
+            STACK_PUSH5(ISAd, first, a, limit, trlink);
+            first = b;
+          } else if(1 < (a - first)) {
+            last = a;
+          } else {
+            STACK_POP5(ISAd, first, last, limit, trlink);
+          }
+        }
+      }
+    } else {
+      if(trbudget_check(budget, last - first)) {
+        limit = tr_ilg(last - first), ISAd += incr;
+      } else {
+        if(0 <= trlink) { stack[trlink].d = -1; }
+        STACK_POP5(ISAd, first, last, limit, trlink);
+      }
+    }
+  }
+#undef STACK_SIZE
+}
+
+
+
+/*---------------------------------------------------------------------------*/
+
+/*- Function -*/
+
+/* Tandem repeat sort */
+void
+trsort(saidx_t *ISA, saidx_t *SA, saidx_t n, saidx_t depth) {
+  saidx_t *ISAd;
+  saidx_t *first, *last;
+  trbudget_t budget;
+  saidx_t t, skip, unsorted;
+
+  trbudget_init(&budget, tr_ilg(n) * 2 / 3, n);
+/*  trbudget_init(&budget, tr_ilg(n) * 3 / 4, n); */
+  for(ISAd = ISA + depth; -n < *SA; ISAd += ISAd - ISA) {
+    first = SA;
+    skip = 0;
+    unsorted = 0;
+    do {
+      if((t = *first) < 0) { first -= t; skip += t; }
+      else {
+        if(skip != 0) { *(first + skip) = skip; skip = 0; }
+        last = SA + ISA[t] + 1;
+        if(1 < (last - first)) {
+          budget.count = 0;
+          tr_introsort(ISA, ISAd, SA, first, last, &budget);
+          if(budget.count != 0) { unsorted += budget.count; }
+          else { skip = first - last; }
+        } else if((last - first) == 1) {
+          skip = -1;
+        }
+        first = last;
+      }
+    } while(first < (SA + n));
+    if(skip != 0) { *(first + skip) = skip; }
+    if(unsorted == 0) { break; }
+  }
+}

loader/tools/lzsa/src/libdivsufsort/pkgconfig/CMakeLists.txt

@@ -0,0 +1,9 @@
+## generate libdivsufsort.pc ##
+set(W64BIT "")
+configure_file("${CMAKE_CURRENT_SOURCE_DIR}/libdivsufsort.pc.cmake" "${CMAKE_CURRENT_BINARY_DIR}/libdivsufsort.pc" @ONLY)
+install(FILES "${CMAKE_CURRENT_BINARY_DIR}/libdivsufsort.pc" DESTINATION ${CMAKE_INSTALL_PKGCONFIGDIR})
+if(BUILD_DIVSUFSORT64)
+  set(W64BIT "64")
+  configure_file("${CMAKE_CURRENT_SOURCE_DIR}/libdivsufsort.pc.cmake" "${CMAKE_CURRENT_BINARY_DIR}/libdivsufsort64.pc" @ONLY)
+  install(FILES "${CMAKE_CURRENT_BINARY_DIR}/libdivsufsort64.pc" DESTINATION ${CMAKE_INSTALL_PKGCONFIGDIR})
+endif(BUILD_DIVSUFSORT64)

loader/tools/lzsa/src/libdivsufsort/pkgconfig/libdivsufsort.pc.cmake

@@ -0,0 +1,11 @@
+prefix=@CMAKE_INSTALL_PREFIX@
+exec_prefix=${prefix}
+libdir=@CMAKE_INSTALL_LIBDIR@
+includedir=@CMAKE_INSTALL_INCLUDEDIR@
+
+Name: @PROJECT_NAME@@W64BIT@
+Description: @PROJECT_DESCRIPTION@
+Version: @PROJECT_VERSION_FULL@
+URL: @PROJECT_URL@
+Libs: -L${libdir} -ldivsufsort@W64BIT@
+Cflags: -I${includedir}

loader/tools/lzsa/src/matchfinder.c

@@ -0,0 +1,361 @@
+/*
+ * matchfinder.c - LZ match finder implementation
+ *
+ * The following copying information applies to this specific source code file:
+ *
+ * Written in 2019 by Emmanuel Marty <marty.emmanuel@gmail.com>
+ * Portions written in 2014-2015 by Eric Biggers <ebiggers3@gmail.com>
+ *
+ * To the extent possible under law, the author(s) have dedicated all copyright
+ * and related and neighboring rights to this software to the public domain
+ * worldwide via the Creative Commons Zero 1.0 Universal Public Domain
+ * Dedication (the "CC0").
+ *
+ * This software is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ * FOR A PARTICULAR PURPOSE. See the CC0 for more details.
+ *
+ * You should have received a copy of the CC0 along with this software; if not
+ * see <http://creativecommons.org/publicdomain/zero/1.0/>.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#include <stdlib.h>
+#include <string.h>
+#include "matchfinder.h"
+#include "format.h"
+#include "lib.h"
+
+/**
+ * Hash index into TAG_BITS
+ *
+ * @param nIndex index value
+ *
+ * @return hash
+ */
+static inline int lzsa_get_index_tag(unsigned int nIndex) {
+   return (int)(((unsigned long long)nIndex * 11400714819323198485ULL) >> (64ULL - TAG_BITS));
+}
+
+/**
+ * Parse input data, build suffix array and overlaid data structures to speed up match finding
+ *
+ * @param pCompressor compression context
+ * @param pInWindow pointer to input data window (previously compressed bytes + bytes to compress)
+ * @param nInWindowSize total input size in bytes (previously compressed bytes + bytes to compress)
+ *
+ * @return 0 for success, non-zero for failure
+ */
+int lzsa_build_suffix_array(lzsa_compressor *pCompressor, const unsigned char *pInWindow, const int nInWindowSize) {
+   unsigned int *intervals = pCompressor->intervals;
+
+   /* Build suffix array from input data */
+   if (divsufsort_build_array(&pCompressor->divsufsort_context, pInWindow, (saidx_t*)intervals, nInWindowSize) != 0) {
+      return 100;
+   }
+
+   int *PLCP = (int*)pCompressor->pos_data;  /* Use temporarily */
+   int *Phi = PLCP;
+   int nCurLen = 0;
+   int i, r;
+
+   /* Compute the permuted LCP first (Kärkkäinen method) */
+   Phi[intervals[0]] = -1;
+   for (i = 1; i < nInWindowSize; i++)
+      Phi[intervals[i]] = intervals[i - 1];
+   for (i = 0; i < nInWindowSize; i++) {
+      if (Phi[i] == -1) {
+         PLCP[i] = 0;
+         continue;
+      }
+      int nMaxLen = (i > Phi[i]) ? (nInWindowSize - i) : (nInWindowSize - Phi[i]);
+      while (nCurLen < nMaxLen && pInWindow[i + nCurLen] == pInWindow[Phi[i] + nCurLen]) nCurLen++;
+      PLCP[i] = nCurLen;
+      if (nCurLen > 0)
+         nCurLen--;
+   }
+
+   /* Rotate permuted LCP into the LCP. This has better cache locality than the direct Kasai LCP method. This also
+    * saves us from having to build the inverse suffix array index, as the LCP is calculated without it using this method,
+    * and the interval builder below doesn't need it either. */
+   intervals[0] &= POS_MASK;
+   int nMinMatchSize = pCompressor->min_match_size;
+
+   if (pCompressor->format_version >= 2) {
+      for (i = 1; i < nInWindowSize; i++) {
+         int nIndex = (int)(intervals[i] & POS_MASK);
+         int nLen = PLCP[nIndex];
+         if (nLen < nMinMatchSize)
+            nLen = 0;
+         if (nLen > LCP_MAX)
+            nLen = LCP_MAX;
+         int nTaggedLen = 0;
+         if (nLen)
+            nTaggedLen = (nLen << TAG_BITS) | (lzsa_get_index_tag((unsigned int)nIndex) & ((1 << TAG_BITS) - 1));
+         intervals[i] = ((unsigned int)nIndex) | (((unsigned int)nTaggedLen) << LCP_SHIFT);
+      }
+   }
+   else {
+      for (i = 1; i < nInWindowSize; i++) {
+         int nIndex = (int)(intervals[i] & POS_MASK);
+         int nLen = PLCP[nIndex];
+         if (nLen < nMinMatchSize)
+            nLen = 0;
+         if (nLen > LCP_AND_TAG_MAX)
+            nLen = LCP_AND_TAG_MAX;
+         intervals[i] = ((unsigned int)nIndex) | (((unsigned int)nLen) << LCP_SHIFT);
+      }
+   }
+
+   /**
+    * Build intervals for finding matches
+    *
+    * Methodology and code fragment taken from wimlib (CC0 license):
+    * https://wimlib.net/git/?p=wimlib;a=blob_plain;f=src/lcpit_matchfinder.c;h=a2d6a1e0cd95200d1f3a5464d8359d5736b14cbe;hb=HEAD
+    */
+   unsigned int * const SA_and_LCP = intervals;
+   unsigned int *pos_data = pCompressor->pos_data;
+   unsigned int next_interval_idx;
+   unsigned int *top = pCompressor->open_intervals;
+   unsigned int prev_pos = SA_and_LCP[0] & POS_MASK;
+
+   *top = 0;
+   intervals[0] = 0;
+   next_interval_idx = 1;
+
+   for (r = 1; r < nInWindowSize; r++) {
+      const unsigned int next_pos = SA_and_LCP[r] & POS_MASK;
+      const unsigned int next_lcp = SA_and_LCP[r] & LCP_MASK;
+      const unsigned int top_lcp = *top & LCP_MASK;
+
+      if (next_lcp == top_lcp) {
+         /* Continuing the deepest open interval  */
+         pos_data[prev_pos] = *top;
+      }
+      else if (next_lcp > top_lcp) {
+         /* Opening a new interval  */
+         *++top = next_lcp | next_interval_idx++;
+         pos_data[prev_pos] = *top;
+      }
+      else {
+         /* Closing the deepest open interval  */
+         pos_data[prev_pos] = *top;
+         for (;;) {
+            const unsigned int closed_interval_idx = *top-- & POS_MASK;
+            const unsigned int superinterval_lcp = *top & LCP_MASK;
+
+            if (next_lcp == superinterval_lcp) {
+               /* Continuing the superinterval */
+               intervals[closed_interval_idx] = *top;
+               break;
+            }
+            else if (next_lcp > superinterval_lcp) {
+               /* Creating a new interval that is a
+                * superinterval of the one being
+                * closed, but still a subinterval of
+                * its superinterval  */
+               *++top = next_lcp | next_interval_idx++;
+               intervals[closed_interval_idx] = *top;
+               break;
+            }
+            else {
+               /* Also closing the superinterval  */
+               intervals[closed_interval_idx] = *top;
+            }
+         }
+      }
+      prev_pos = next_pos;
+   }
+
+   /* Close any still-open intervals.  */
+   pos_data[prev_pos] = *top;
+   for (; top > pCompressor->open_intervals; top--)
+      intervals[*top & POS_MASK] = *(top - 1);
+
+   /* Success */
+   return 0;
+}
+
+/**
+ * Find matches at the specified offset in the input window
+ *
+ * @param pCompressor compression context
+ * @param nOffset offset to find matches at, in the input window
+ * @param pMatches pointer to returned matches
+ * @param nMaxMatches maximum number of matches to return (0 for none)
+ * @param nInWindowSize total input size in bytes (previously compressed bytes + bytes to compress)
+ *
+ * @return number of matches
+ */
+int lzsa_find_matches_at(lzsa_compressor *pCompressor, const int nOffset, lzsa_match *pMatches, const int nMaxMatches, const int nInWindowSize) {
+   unsigned int *intervals = pCompressor->intervals;
+   unsigned int *pos_data = pCompressor->pos_data;
+   unsigned int ref;
+   unsigned int super_ref;
+   unsigned int match_pos;
+   lzsa_match *matchptr;
+   int nPrevOffset = 0;
+
+   /**
+    * Find matches using intervals
+    *
+    * Taken from wimlib (CC0 license):
+    * https://wimlib.net/git/?p=wimlib;a=blob_plain;f=src/lcpit_matchfinder.c;h=a2d6a1e0cd95200d1f3a5464d8359d5736b14cbe;hb=HEAD
+    */
+
+    /* Get the deepest lcp-interval containing the current suffix. */
+   ref = pos_data[nOffset];
+
+   pos_data[nOffset] = 0;
+
+   /* Ascend until we reach a visited interval, the root, or a child of the
+    * root.  Link unvisited intervals to the current suffix as we go.  */
+   while ((super_ref = intervals[ref & POS_MASK]) & LCP_MASK) {
+      intervals[ref & POS_MASK] = nOffset | VISITED_FLAG;
+      ref = super_ref;
+   }
+
+   if (super_ref == 0) {
+      /* In this case, the current interval may be any of:
+       * (1) the root;
+       * (2) an unvisited child of the root */
+
+      if (ref != 0)  /* Not the root?  */
+         intervals[ref & POS_MASK] = nOffset | VISITED_FLAG;
+      return 0;
+   }
+
+   /* Ascend indirectly via pos_data[] links.  */
+   match_pos = super_ref & EXCL_VISITED_MASK;
+   matchptr = pMatches;
+
+   if (pCompressor->format_version >= 2 && nInWindowSize < 65536) {
+      if ((matchptr - pMatches) < nMaxMatches) {
+         int nMatchOffset = (int)(nOffset - match_pos);
+
+         if (nMatchOffset <= MAX_OFFSET) {
+            matchptr->length = (unsigned short)(ref >> (LCP_SHIFT + TAG_BITS));
+            matchptr->offset = (unsigned short)nMatchOffset;
+            matchptr++;
+
+            nPrevOffset = nMatchOffset;
+         }
+      }
+   }
+
+   for (;;) {
+      if ((super_ref = pos_data[match_pos]) > ref) {
+         match_pos = intervals[super_ref & POS_MASK] & EXCL_VISITED_MASK;
+
+         if (pCompressor->format_version >= 2 && nInWindowSize < 65536) {
+            if ((matchptr - pMatches) < nMaxMatches) {
+               int nMatchOffset = (int)(nOffset - match_pos);
+
+               if (nMatchOffset <= MAX_OFFSET && nMatchOffset != nPrevOffset) {
+                  matchptr->length = ((unsigned short)(ref >> (LCP_SHIFT + TAG_BITS))) | 0x8000;
+                  matchptr->offset = (unsigned short)nMatchOffset;
+                  matchptr++;
+
+                  nPrevOffset = nMatchOffset;
+               }
+            }
+         }
+      }
+
+      while ((super_ref = pos_data[match_pos]) > ref)
+         match_pos = intervals[super_ref & POS_MASK] & EXCL_VISITED_MASK;
+      intervals[ref & POS_MASK] = nOffset | VISITED_FLAG;
+      pos_data[match_pos] = ref;
+
+      if ((matchptr - pMatches) < nMaxMatches) {
+         int nMatchOffset = (int)(nOffset - match_pos);
+
+         if (nMatchOffset <= MAX_OFFSET && nMatchOffset != nPrevOffset) {
+            if (pCompressor->format_version >= 2) {
+               matchptr->length = (unsigned short)(ref >> (LCP_SHIFT + TAG_BITS));
+            }
+            else {
+               matchptr->length = (unsigned short)(ref >> LCP_SHIFT);
+            }
+            matchptr->offset = (unsigned short)nMatchOffset;
+            matchptr++;
+         }
+      }
+
+      if (super_ref == 0)
+         break;
+      ref = super_ref;
+      match_pos = intervals[ref & POS_MASK] & EXCL_VISITED_MASK;
+
+      if (pCompressor->format_version >= 2 && nInWindowSize < 65536) {
+         if ((matchptr - pMatches) < nMaxMatches) {
+            int nMatchOffset = (int)(nOffset - match_pos);
+
+            if (nMatchOffset <= MAX_OFFSET && nMatchOffset != nPrevOffset) {
+               matchptr->length = ((unsigned short)(ref >> (LCP_SHIFT + TAG_BITS))) | 0x8000;
+               matchptr->offset = (unsigned short)nMatchOffset;
+
+               if ((matchptr->length & 0x7fff) > 2) {
+                  matchptr++;
+
+                  nPrevOffset = nMatchOffset;
+               }
+            }
+         }
+      }
+   }
+
+   return (int)(matchptr - pMatches);
+}
+
+/**
+ * Skip previously compressed bytes
+ *
+ * @param pCompressor compression context
+ * @param nStartOffset current offset in input window (typically 0)
+ * @param nEndOffset offset to skip to in input window (typically the number of previously compressed bytes)
+ */
+void lzsa_skip_matches(lzsa_compressor *pCompressor, const int nStartOffset, const int nEndOffset) {
+   lzsa_match match;
+   int i;
+
+   /* Skipping still requires scanning for matches, as this also performs a lazy update of the intervals. However,
+    * we don't store the matches. */
+   for (i = nStartOffset; i < nEndOffset; i++) {
+      lzsa_find_matches_at(pCompressor, i, &match, 0, 0);
+   }
+}
+
+/**
+ * Find all matches for the data to be compressed
+ *
+ * @param pCompressor compression context
+ * @param nMatchesPerOffset maximum number of matches to store for each offset
+ * @param nStartOffset current offset in input window (typically the number of previously compressed bytes)
+ * @param nEndOffset offset to end finding matches at (typically the size of the total input window in bytes
+ */
+void lzsa_find_all_matches(lzsa_compressor *pCompressor, const int nMatchesPerOffset, const int nStartOffset, const int nEndOffset) {
+   lzsa_match *pMatch = pCompressor->match;
+   int i;
+
+   for (i = nStartOffset; i < nEndOffset; i++) {
+      int nMatches = lzsa_find_matches_at(pCompressor, i, pMatch, nMatchesPerOffset, nEndOffset - nStartOffset);
+
+      while (nMatches < nMatchesPerOffset) {
+         pMatch[nMatches].length = 0;
+         pMatch[nMatches].offset = 0;
+         nMatches++;
+      }
+
+      pMatch += nMatchesPerOffset;
+   }
+}

loader/tools/lzsa/src/matchfinder.h

@@ -0,0 +1,91 @@
+/*
+ * matchfinder.h - LZ match finder definitions
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#ifndef _MATCHFINDER_H
+#define _MATCHFINDER_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/* Forward declarations */
+typedef struct _lzsa_match lzsa_match;
+typedef struct _lzsa_compressor lzsa_compressor;
+
+/**
+ * Parse input data, build suffix array and overlaid data structures to speed up match finding
+ *
+ * @param pCompressor compression context
+ * @param pInWindow pointer to input data window (previously compressed bytes + bytes to compress)
+ * @param nInWindowSize total input size in bytes (previously compressed bytes + bytes to compress)
+ *
+ * @return 0 for success, non-zero for failure
+ */
+int lzsa_build_suffix_array(lzsa_compressor *pCompressor, const unsigned char *pInWindow, const int nInWindowSize);
+
+/**
+ * Find matches at the specified offset in the input window
+ *
+ * @param pCompressor compression context
+ * @param nOffset offset to find matches at, in the input window
+ * @param pMatches pointer to returned matches
+ * @param nMaxMatches maximum number of matches to return (0 for none)
+ * @param nInWindowSize total input size in bytes (previously compressed bytes + bytes to compress)
+ *
+ * @return number of matches
+ */
+int lzsa_find_matches_at(lzsa_compressor *pCompressor, const int nOffset, lzsa_match *pMatches, const int nMaxMatches, const int nInWindowSize);
+
+/**
+ * Skip previously compressed bytes
+ *
+ * @param pCompressor compression context
+ * @param nStartOffset current offset in input window (typically 0)
+ * @param nEndOffset offset to skip to in input window (typically the number of previously compressed bytes)
+ */
+void lzsa_skip_matches(lzsa_compressor *pCompressor, const int nStartOffset, const int nEndOffset);
+
+/**
+ * Find all matches for the data to be compressed
+ *
+ * @param pCompressor compression context
+ * @param nMatchesPerOffset maximum number of matches to store for each offset
+ * @param nStartOffset current offset in input window (typically the number of previously compressed bytes)
+ * @param nEndOffset offset to end finding matches at (typically the size of the total input window in bytes
+ */
+void lzsa_find_all_matches(lzsa_compressor *pCompressor, const int nMatchesPerOffset, const int nStartOffset, const int nEndOffset);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _MATCHFINDER_H */

loader/tools/lzsa/src/shrink_block_v1.c

@@ -0,0 +1,710 @@
+/*
+ * shrink_block_v1.c - LZSA1 block compressor implementation
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#include <stdlib.h>
+#include <string.h>
+#include "lib.h"
+#include "shrink_block_v1.h"
+#include "format.h"
+
+/**
+ * Get the number of extra bits required to represent a literals length
+ *
+ * @param nLength literals length
+ *
+ * @return number of extra bits required
+ */
+static inline int lzsa_get_literals_varlen_size_v1(const int nLength) {
+   if (nLength < LITERALS_RUN_LEN_V1) {
+      return 0;
+   }
+   else {
+      if (nLength < 256)
+         return 8;
+      else {
+         if (nLength < 512)
+            return 16;
+         else
+            return 24;
+      }
+   }
+}
+
+/**
+ * Write extra literals length bytes to output (compressed) buffer. The caller must first check that there is enough
+ * room to write the bytes.
+ *
+ * @param pOutData pointer to output buffer
+ * @param nOutOffset current write index into output buffer
+ * @param nLength literals length
+ */
+static inline int lzsa_write_literals_varlen_v1(unsigned char *pOutData, int nOutOffset, int nLength) {
+   if (nLength >= LITERALS_RUN_LEN_V1) {
+      if (nLength < 256)
+         pOutData[nOutOffset++] = nLength - LITERALS_RUN_LEN_V1;
+      else {
+         if (nLength < 512) {
+            pOutData[nOutOffset++] = 250;
+            pOutData[nOutOffset++] = nLength - 256;
+         }
+         else {
+            pOutData[nOutOffset++] = 249;
+            pOutData[nOutOffset++] = nLength & 0xff;
+            pOutData[nOutOffset++] = (nLength >> 8) & 0xff;
+         }
+      }
+   }
+
+   return nOutOffset;
+}
+
+/**
+ * Get the number of extra bits required to represent an encoded match length
+ *
+ * @param nLength encoded match length (actual match length - MIN_MATCH_SIZE_V1)
+ *
+ * @return number of extra bits required
+ */
+static inline int lzsa_get_match_varlen_size_v1(const int nLength) {
+   if (nLength < MATCH_RUN_LEN_V1) {
+      return 0;
+   }
+   else {
+      if ((nLength + MIN_MATCH_SIZE_V1) < 256)
+         return 8;
+      else {
+         if ((nLength + MIN_MATCH_SIZE_V1) < 512)
+            return 16;
+         else
+            return 24;
+      }
+   }
+}
+
+/**
+ * Write extra encoded match length bytes to output (compressed) buffer. The caller must first check that there is enough
+ * room to write the bytes.
+ *
+ * @param pOutData pointer to output buffer
+ * @param nOutOffset current write index into output buffer
+ * @param nLength encoded match length (actual match length - MIN_MATCH_SIZE_V1)
+ */
+static inline int lzsa_write_match_varlen_v1(unsigned char *pOutData, int nOutOffset, int nLength) {
+   if (nLength >= MATCH_RUN_LEN_V1) {
+      if ((nLength + MIN_MATCH_SIZE_V1) < 256)
+         pOutData[nOutOffset++] = nLength - MATCH_RUN_LEN_V1;
+      else {
+         if ((nLength + MIN_MATCH_SIZE_V1) < 512) {
+            pOutData[nOutOffset++] = 239;
+            pOutData[nOutOffset++] = nLength + MIN_MATCH_SIZE_V1 - 256;
+         }
+         else {
+            pOutData[nOutOffset++] = 238;
+            pOutData[nOutOffset++] = (nLength + MIN_MATCH_SIZE_V1) & 0xff;
+            pOutData[nOutOffset++] = ((nLength + MIN_MATCH_SIZE_V1) >> 8) & 0xff;
+         }
+      }
+   }
+
+   return nOutOffset;
+}
+
+/**
+ * Get offset encoding cost in bits
+ *
+ * @param nMatchOffset offset to get cost of
+ *
+ * @return cost in bits
+ */
+static inline int lzsa_get_offset_cost_v1(const unsigned int nMatchOffset) {
+   return (nMatchOffset <= 256) ? 8 : 16;
+}
+
+/**
+ * Attempt to pick optimal matches using a forward arrivals parser, so as to produce the smallest possible output that decompresses to the same input
+ *
+ * @param pCompressor compression context
+ * @param nStartOffset current offset in input window (typically the number of previously compressed bytes)
+ * @param nEndOffset offset to end finding matches at (typically the size of the total input window in bytes
+ */
+static void lzsa_optimize_forward_v1(lzsa_compressor *pCompressor, lzsa_match *pBestMatch, const int nStartOffset, const int nEndOffset, const int nReduce) {
+   lzsa_arrival *arrival = pCompressor->arrival - (nStartOffset << ARRIVALS_PER_POSITION_SHIFT);
+   const int nMinMatchSize = pCompressor->min_match_size;
+   const int nFavorRatio = (pCompressor->flags & LZSA_FLAG_FAVOR_RATIO) ? 1 : 0;
+   const int nModeSwitchPenalty = nFavorRatio ? 0 : MODESWITCH_PENALTY;
+   const int nDisableScore = nReduce ? 0 : (2 * BLOCK_SIZE);
+   int i, j, n;
+
+   if ((nEndOffset - nStartOffset) > BLOCK_SIZE) return;
+
+   memset(arrival + (nStartOffset << ARRIVALS_PER_POSITION_SHIFT), 0, sizeof(lzsa_arrival) * ((nEndOffset - nStartOffset + 1) << ARRIVALS_PER_POSITION_SHIFT));
+
+   arrival[nStartOffset << ARRIVALS_PER_POSITION_SHIFT].from_slot = -1;
+
+   for (i = nStartOffset; i != nEndOffset; i++) {
+      lzsa_arrival* cur_arrival = &arrival[i << ARRIVALS_PER_POSITION_SHIFT];
+      int m;
+
+      for (j = 0; j < NARRIVALS_PER_POSITION_V1 && cur_arrival[j].from_slot; j++) {
+         int nPrevCost = cur_arrival[j].cost;
+         int nCodingChoiceCost = nPrevCost + 8 /* literal */;
+         int nScore = cur_arrival[j].score + 1;
+         int nNumLiterals = cur_arrival[j].num_literals + 1;
+
+         if (nNumLiterals == LITERALS_RUN_LEN_V1 || nNumLiterals == 256 || nNumLiterals == 512) {
+            nCodingChoiceCost += 8;
+         }
+
+         if (nNumLiterals == 1)
+            nCodingChoiceCost += nModeSwitchPenalty;
+
+         lzsa_arrival *pDestSlots = &arrival[(i + 1) << ARRIVALS_PER_POSITION_SHIFT];
+         for (n = 0; n < NARRIVALS_PER_POSITION_V1 /* we only need the literals + short match cost + long match cost cases */; n++) {
+            lzsa_arrival *pDestArrival = &pDestSlots[n];
+
+            if (pDestArrival->from_slot == 0 ||
+               nCodingChoiceCost < pDestArrival->cost ||
+               (nCodingChoiceCost == pDestArrival->cost && nScore < (pDestArrival->score + nDisableScore))) {
+               memmove(&arrival[((i + 1) << ARRIVALS_PER_POSITION_SHIFT) + n + 1],
+                  &arrival[((i + 1) << ARRIVALS_PER_POSITION_SHIFT) + n],
+                  sizeof(lzsa_arrival) * (NARRIVALS_PER_POSITION_V1 - n - 1));
+
+               pDestArrival->cost = nCodingChoiceCost;
+               pDestArrival->from_pos = i;
+               pDestArrival->from_slot = j + 1;
+               pDestArrival->match_len = 0;
+               pDestArrival->num_literals = nNumLiterals;
+               pDestArrival->score = nScore;
+               pDestArrival->rep_offset = cur_arrival[j].rep_offset;
+               break;
+            }
+         }
+      }
+
+      const lzsa_match *match = pCompressor->match + ((i - nStartOffset) << MATCHES_PER_INDEX_SHIFT_V1);
+      int nNumArrivalsForThisPos = j;
+
+      for (m = 0; m < NMATCHES_PER_INDEX_V1 && match[m].length; m++) {
+         int nMatchLen = match[m].length;
+         int nMatchOffsetCost = lzsa_get_offset_cost_v1(match[m].offset);
+         int nStartingMatchLen, k;
+
+         if ((i + nMatchLen) > nEndOffset)
+            nMatchLen = nEndOffset - i;
+
+         if (nMatchLen >= LEAVE_ALONE_MATCH_SIZE)
+            nStartingMatchLen = nMatchLen;
+         else
+            nStartingMatchLen = nMinMatchSize;
+         for (k = nStartingMatchLen; k <= nMatchLen; k++) {
+            int nMatchLenCost = lzsa_get_match_varlen_size_v1(k - MIN_MATCH_SIZE_V1);
+
+            lzsa_arrival *pDestSlots = &arrival[(i + k) << ARRIVALS_PER_POSITION_SHIFT];
+
+            for (j = 0; j < nNumArrivalsForThisPos; j++) {
+               int nPrevCost = cur_arrival[j].cost;
+               int nCodingChoiceCost = nPrevCost + 8 /* token */ /* the actual cost of the literals themselves accumulates up the chain */ + nMatchOffsetCost + nMatchLenCost;
+               int exists = 0;
+
+               if (!cur_arrival[j].num_literals)
+                  nCodingChoiceCost += nModeSwitchPenalty;
+
+               for (n = 0;
+                  n < NARRIVALS_PER_POSITION_V1 && pDestSlots[n].from_slot && pDestSlots[n].cost <= nCodingChoiceCost;
+                  n++) {
+                  if (lzsa_get_offset_cost_v1(pDestSlots[n].rep_offset) == nMatchOffsetCost) {
+                     exists = 1;
+                     break;
+                  }
+               }
+
+               if (!exists) {
+                  int nScore = cur_arrival[j].score + 5;
+
+                  for (n = 0; n < NARRIVALS_PER_POSITION_V1 /* we only need the literals + short match cost + long match cost cases */; n++) {
+                     lzsa_arrival *pDestArrival = &pDestSlots[n];
+
+                     if (pDestArrival->from_slot == 0 ||
+                        nCodingChoiceCost < pDestArrival->cost ||
+                        (nCodingChoiceCost == pDestArrival->cost && nScore < (pDestArrival->score + nDisableScore))) {
+                        memmove(&pDestSlots[n + 1],
+                           &pDestSlots[n],
+                           sizeof(lzsa_arrival) * (NARRIVALS_PER_POSITION_V1 - n - 1));
+
+                        pDestArrival->cost = nCodingChoiceCost;
+                        pDestArrival->from_pos = i;
+                        pDestArrival->from_slot = j + 1;
+                        pDestArrival->match_len = k;
+                        pDestArrival->num_literals = 0;
+                        pDestArrival->score = nScore;
+                        pDestArrival->rep_offset = match[m].offset;
+                        j = NARRIVALS_PER_POSITION_V1;
+                        break;
+                     }
+                  }
+               }
+            }
+         }
+      }
+   }
+
+   lzsa_arrival *end_arrival = &arrival[(i << ARRIVALS_PER_POSITION_SHIFT) + 0];
+
+   while (end_arrival->from_slot > 0 && end_arrival->from_pos >= 0) {
+      if (end_arrival->from_pos >= nEndOffset) return;
+      pBestMatch[end_arrival->from_pos].length = end_arrival->match_len;
+      if (end_arrival->match_len)
+         pBestMatch[end_arrival->from_pos].offset = end_arrival->rep_offset;
+      else
+         pBestMatch[end_arrival->from_pos].offset = 0;
+
+      end_arrival = &arrival[(end_arrival->from_pos << ARRIVALS_PER_POSITION_SHIFT) + (end_arrival->from_slot - 1)];
+   }
+}
+
+/**
+ * Attempt to minimize the number of commands issued in the compressed data block, in order to speed up decompression without
+ * impacting the compression ratio
+ *
+ * @param pCompressor compression context
+ * @param pInWindow pointer to input data window (previously compressed bytes + bytes to compress)
+ * @param pBestMatch optimal matches to emit
+ * @param nStartOffset current offset in input window (typically the number of previously compressed bytes)
+ * @param nEndOffset offset to end finding matches at (typically the size of the total input window in bytes
+ *
+ * @return non-zero if the number of tokens was reduced, 0 if it wasn't
+ */
+static int lzsa_optimize_command_count_v1(lzsa_compressor *pCompressor, const unsigned char *pInWindow, lzsa_match *pBestMatch, const int nStartOffset, const int nEndOffset) {
+   int i;
+   int nNumLiterals = 0;
+   int nDidReduce = 0;
+
+   for (i = nStartOffset; i < nEndOffset; ) {
+      lzsa_match *pMatch = pBestMatch + i;
+
+      if (pMatch->length == 0 &&
+         (i + 1) < nEndOffset &&
+         pBestMatch[i + 1].length >= MIN_MATCH_SIZE_V1 &&
+         pBestMatch[i + 1].length < MAX_VARLEN &&
+         pBestMatch[i + 1].offset &&
+         i >= pBestMatch[i + 1].offset &&
+         (i + pBestMatch[i + 1].length + 1) <= nEndOffset &&
+         !memcmp(pInWindow + i - (pBestMatch[i + 1].offset), pInWindow + i, pBestMatch[i + 1].length + 1)) {
+         int nCurLenSize = lzsa_get_match_varlen_size_v1(pBestMatch[i + 1].length - MIN_MATCH_SIZE_V1);
+         int nReducedLenSize = lzsa_get_match_varlen_size_v1(pBestMatch[i + 1].length + 1 - MIN_MATCH_SIZE_V1);
+
+         if ((nReducedLenSize - nCurLenSize) <= 8) {
+            /* Merge */
+            pBestMatch[i].length = pBestMatch[i + 1].length + 1;
+            pBestMatch[i].offset = pBestMatch[i + 1].offset;
+            pBestMatch[i + 1].length = 0;
+            pBestMatch[i + 1].offset = 0;
+            nDidReduce = 1;
+            continue;
+         }
+      }
+
+      if (pMatch->length >= MIN_MATCH_SIZE_V1) {
+         if (pMatch->length <= 9 /* Don't waste time considering large matches, they will always win over literals */ &&
+            (i + pMatch->length) < nEndOffset /* Don't consider the last token in the block, we can only reduce a match inbetween other tokens */) {
+            int nNextIndex = i + pMatch->length;
+            int nNextLiterals = 0;
+
+            while (nNextIndex < nEndOffset && pBestMatch[nNextIndex].length < MIN_MATCH_SIZE_V1) {
+               nNextLiterals++;
+               nNextIndex++;
+            }
+
+            /* This command is a match, is followed by 'nNextLiterals' literals and then by another match, or the end of the input. Calculate this command's current cost (excluding 'nNumLiterals' bytes) */
+            if ((8 /* token */ + lzsa_get_literals_varlen_size_v1(nNumLiterals) + ((pMatch->offset <= 256) ? 8 : 16) /* match offset */ + lzsa_get_match_varlen_size_v1(pMatch->length - MIN_MATCH_SIZE_V1) +
+               8 /* token */ + lzsa_get_literals_varlen_size_v1(nNextLiterals)) >=
+               (8 /* token */ + (pMatch->length << 3) + lzsa_get_literals_varlen_size_v1(nNumLiterals + pMatch->length + nNextLiterals))) {
+               /* Reduce */
+               int nMatchLen = pMatch->length;
+               int j;
+
+               for (j = 0; j < nMatchLen; j++) {
+                  pBestMatch[i + j].length = 0;
+               }
+
+               nDidReduce = 1;
+               continue;
+            }
+         }
+
+         if ((i + pMatch->length) <= nEndOffset && pMatch->offset > 0 && pMatch->length >= MIN_MATCH_SIZE_V1 &&
+            pBestMatch[i + pMatch->length].offset > 0 &&
+            pBestMatch[i + pMatch->length].length >= MIN_MATCH_SIZE_V1 &&
+            (pMatch->length + pBestMatch[i + pMatch->length].length) >= LEAVE_ALONE_MATCH_SIZE &&
+            (pMatch->length + pBestMatch[i + pMatch->length].length) <= MAX_VARLEN &&
+            (i + pMatch->length) > pMatch->offset &&
+            (i + pMatch->length) > pBestMatch[i + pMatch->length].offset &&
+            (i + pMatch->length + pBestMatch[i + pMatch->length].length) <= nEndOffset &&
+            !memcmp(pInWindow + i - pMatch->offset + pMatch->length,
+               pInWindow + i + pMatch->length - pBestMatch[i + pMatch->length].offset,
+               pBestMatch[i + pMatch->length].length)) {
+
+            int nCurPartialSize = lzsa_get_match_varlen_size_v1(pMatch->length - MIN_MATCH_SIZE_V1);
+            nCurPartialSize += 8 /* token */ + lzsa_get_literals_varlen_size_v1(0) + ((pBestMatch[i + pMatch->length].offset <= 256) ? 8 : 16) /* match offset */ + lzsa_get_match_varlen_size_v1(pBestMatch[i + pMatch->length].length - MIN_MATCH_SIZE_V1);
+
+            int nReducedPartialSize = lzsa_get_match_varlen_size_v1(pMatch->length + pBestMatch[i + pMatch->length].length - MIN_MATCH_SIZE_V1);
+
+            if (nCurPartialSize >= nReducedPartialSize) {
+               int nMatchLen = pMatch->length;
+
+               /* Join */
+
+               pMatch->length += pBestMatch[i + nMatchLen].length;
+               pBestMatch[i + nMatchLen].offset = 0;
+               pBestMatch[i + nMatchLen].length = -1;
+               continue;
+            }
+         }
+
+         i += pMatch->length;
+         nNumLiterals = 0;
+      }
+      else {
+         nNumLiterals++;
+         i++;
+      }
+   }
+
+   return nDidReduce;
+}
+
+/**
+ * Get compressed data block size
+ *
+ * @param pCompressor compression context
+ * @param pBestMatch optimal matches to emit
+ * @param nStartOffset current offset in input window (typically the number of previously compressed bytes)
+ * @param nEndOffset offset to end finding matches at (typically the size of the total input window in bytes
+ *
+ * @return size of compressed data that will be written to output buffer
+ */
+static int lzsa_get_compressed_size_v1(lzsa_compressor *pCompressor, lzsa_match *pBestMatch, const int nStartOffset, const int nEndOffset) {
+   int i;
+   int nNumLiterals = 0;
+   int nCompressedSize = 0;
+
+   for (i = nStartOffset; i < nEndOffset; ) {
+      const lzsa_match *pMatch = pBestMatch + i;
+
+      if (pMatch->length >= MIN_MATCH_SIZE_V1) {
+         int nMatchOffset = pMatch->offset;
+         int nMatchLen = pMatch->length;
+         int nEncodedMatchLen = nMatchLen - MIN_MATCH_SIZE_V1;
+         int nTokenLongOffset = (nMatchOffset <= 256) ? 0x00 : 0x80;
+         int nCommandSize = 8 /* token */ + lzsa_get_literals_varlen_size_v1(nNumLiterals) + (nNumLiterals << 3) + (nTokenLongOffset ? 16 : 8) /* match offset */ + lzsa_get_match_varlen_size_v1(nEncodedMatchLen);
+
+         nCompressedSize += nCommandSize;
+         nNumLiterals = 0;
+         i += nMatchLen;
+      }
+      else {
+         nNumLiterals++;
+         i++;
+      }
+   }
+
+   {
+      int nCommandSize = 8 /* token */ + lzsa_get_literals_varlen_size_v1(nNumLiterals) + (nNumLiterals << 3);
+
+      nCompressedSize += nCommandSize;
+      nNumLiterals = 0;
+   }
+
+   if (pCompressor->flags & LZSA_FLAG_RAW_BLOCK) {
+      nCompressedSize += 8 * 4;
+   }
+
+   return nCompressedSize;
+}
+
+/**
+ * Emit block of compressed data
+ *
+ * @param pCompressor compression context
+ * @param pBestMatch optimal matches to emit
+ * @param pInWindow pointer to input data window (previously compressed bytes + bytes to compress)
+ * @param nStartOffset current offset in input window (typically the number of previously compressed bytes)
+ * @param nEndOffset offset to end finding matches at (typically the size of the total input window in bytes
+ * @param pOutData pointer to output buffer
+ * @param nMaxOutDataSize maximum size of output buffer, in bytes
+ *
+ * @return size of compressed data in output buffer, or -1 if the data is uncompressible
+ */
+static int lzsa_write_block_v1(lzsa_compressor *pCompressor, lzsa_match *pBestMatch, const unsigned char *pInWindow, const int nStartOffset, const int nEndOffset, unsigned char *pOutData, const int nMaxOutDataSize) {
+   int i;
+   int nNumLiterals = 0;
+   int nInFirstLiteralOffset = 0;
+   int nOutOffset = 0;
+
+   for (i = nStartOffset; i < nEndOffset; ) {
+      const lzsa_match *pMatch = pBestMatch + i;
+
+      if (pMatch->length >= MIN_MATCH_SIZE_V1) {
+         int nMatchOffset = pMatch->offset;
+         int nMatchLen = pMatch->length;
+         int nEncodedMatchLen = nMatchLen - MIN_MATCH_SIZE_V1;
+         int nTokenLiteralsLen = (nNumLiterals >= LITERALS_RUN_LEN_V1) ? LITERALS_RUN_LEN_V1 : nNumLiterals;
+         int nTokenMatchLen = (nEncodedMatchLen >= MATCH_RUN_LEN_V1) ? MATCH_RUN_LEN_V1 : nEncodedMatchLen;
+         int nTokenLongOffset = (nMatchOffset <= 256) ? 0x00 : 0x80;
+         int nCommandSize = 8 /* token */ + lzsa_get_literals_varlen_size_v1(nNumLiterals) + (nNumLiterals << 3) + (nTokenLongOffset ? 16 : 8) /* match offset */ + lzsa_get_match_varlen_size_v1(nEncodedMatchLen);
+
+         if ((nOutOffset + (nCommandSize >> 3)) > nMaxOutDataSize)
+            return -1;
+         if (nMatchOffset < MIN_OFFSET || nMatchOffset > MAX_OFFSET)
+            return -1;
+
+         pOutData[nOutOffset++] = nTokenLongOffset | (nTokenLiteralsLen << 4) | nTokenMatchLen;
+         nOutOffset = lzsa_write_literals_varlen_v1(pOutData, nOutOffset, nNumLiterals);
+
+         if (nNumLiterals < pCompressor->stats.min_literals || pCompressor->stats.min_literals == -1)
+            pCompressor->stats.min_literals = nNumLiterals;
+         if (nNumLiterals > pCompressor->stats.max_literals)
+            pCompressor->stats.max_literals = nNumLiterals;
+         pCompressor->stats.total_literals += nNumLiterals;
+         pCompressor->stats.literals_divisor++;
+
+         if (nNumLiterals != 0) {
+            memcpy(pOutData + nOutOffset, pInWindow + nInFirstLiteralOffset, nNumLiterals);
+            nOutOffset += nNumLiterals;
+            nNumLiterals = 0;
+         }
+
+         pOutData[nOutOffset++] = (-nMatchOffset) & 0xff;
+         if (nTokenLongOffset) {
+            pOutData[nOutOffset++] = (-nMatchOffset) >> 8;
+         }
+         nOutOffset = lzsa_write_match_varlen_v1(pOutData, nOutOffset, nEncodedMatchLen);
+
+         if (nMatchOffset < pCompressor->stats.min_offset || pCompressor->stats.min_offset == -1)
+            pCompressor->stats.min_offset = nMatchOffset;
+         if (nMatchOffset > pCompressor->stats.max_offset)
+            pCompressor->stats.max_offset = nMatchOffset;
+         pCompressor->stats.total_offsets += nMatchOffset;
+
+         if (nMatchLen < pCompressor->stats.min_match_len || pCompressor->stats.min_match_len == -1)
+            pCompressor->stats.min_match_len = nMatchLen;
+         if (nMatchLen > pCompressor->stats.max_match_len)
+            pCompressor->stats.max_match_len = nMatchLen;
+         pCompressor->stats.total_match_lens += nMatchLen;
+         pCompressor->stats.match_divisor++;
+
+         if (nMatchOffset == 1) {
+            if (nMatchLen < pCompressor->stats.min_rle1_len || pCompressor->stats.min_rle1_len == -1)
+               pCompressor->stats.min_rle1_len = nMatchLen;
+            if (nMatchLen > pCompressor->stats.max_rle1_len)
+               pCompressor->stats.max_rle1_len = nMatchLen;
+            pCompressor->stats.total_rle1_lens += nMatchLen;
+            pCompressor->stats.rle1_divisor++;
+         }
+         else if (nMatchOffset == 2) {
+            if (nMatchLen < pCompressor->stats.min_rle2_len || pCompressor->stats.min_rle2_len == -1)
+               pCompressor->stats.min_rle2_len = nMatchLen;
+            if (nMatchLen > pCompressor->stats.max_rle2_len)
+               pCompressor->stats.max_rle2_len = nMatchLen;
+            pCompressor->stats.total_rle2_lens += nMatchLen;
+            pCompressor->stats.rle2_divisor++;
+         }
+
+         i += nMatchLen;
+
+         if (pCompressor->flags & LZSA_FLAG_RAW_BLOCK) {
+            int nCurSafeDist = (i - nStartOffset) - nOutOffset;
+            if (nCurSafeDist >= 0 && pCompressor->safe_dist < nCurSafeDist)
+               pCompressor->safe_dist = nCurSafeDist;
+         }
+
+         pCompressor->num_commands++;
+      }
+      else {
+         if (nNumLiterals == 0)
+            nInFirstLiteralOffset = i;
+         nNumLiterals++;
+         i++;
+      }
+   }
+
+   {
+      int nTokenLiteralsLen = (nNumLiterals >= LITERALS_RUN_LEN_V1) ? LITERALS_RUN_LEN_V1 : nNumLiterals;
+      int nCommandSize = 8 /* token */ + lzsa_get_literals_varlen_size_v1(nNumLiterals) + (nNumLiterals << 3);
+
+      if ((nOutOffset + (nCommandSize >> 3)) > nMaxOutDataSize)
+         return -1;
+
+      if (pCompressor->flags & LZSA_FLAG_RAW_BLOCK)
+         pOutData[nOutOffset++] = (nTokenLiteralsLen << 4) | 0x0f;
+      else
+         pOutData[nOutOffset++] = (nTokenLiteralsLen << 4) | 0x00;
+      nOutOffset = lzsa_write_literals_varlen_v1(pOutData, nOutOffset, nNumLiterals);
+
+      if (nNumLiterals < pCompressor->stats.min_literals || pCompressor->stats.min_literals == -1)
+         pCompressor->stats.min_literals = nNumLiterals;
+      if (nNumLiterals > pCompressor->stats.max_literals)
+         pCompressor->stats.max_literals = nNumLiterals;
+      pCompressor->stats.total_literals += nNumLiterals;
+      pCompressor->stats.literals_divisor++;
+
+      if (nNumLiterals != 0) {
+         memcpy(pOutData + nOutOffset, pInWindow + nInFirstLiteralOffset, nNumLiterals);
+         nOutOffset += nNumLiterals;
+         nNumLiterals = 0;
+      }
+
+      if (pCompressor->flags & LZSA_FLAG_RAW_BLOCK) {
+         int nCurSafeDist = (i - nStartOffset) - nOutOffset;
+         if (nCurSafeDist >= 0 && pCompressor->safe_dist < nCurSafeDist)
+            pCompressor->safe_dist = nCurSafeDist;
+      }
+
+      pCompressor->num_commands++;
+   }
+
+   if (pCompressor->flags & LZSA_FLAG_RAW_BLOCK) {
+      /* Emit EOD marker for raw block */
+
+      if ((nOutOffset + 4) > nMaxOutDataSize)
+         return -1;
+
+      pOutData[nOutOffset++] = 0;
+      pOutData[nOutOffset++] = 238;
+      pOutData[nOutOffset++] = 0;
+      pOutData[nOutOffset++] = 0;
+   }
+
+   return nOutOffset;
+}
+
+/**
+ * Emit raw block of uncompressible data
+ *
+ * @param pCompressor compression context
+ * @param pInWindow pointer to input data window (previously compressed bytes + bytes to compress)
+ * @param nStartOffset current offset in input window (typically the number of previously compressed bytes)
+ * @param nEndOffset offset to end finding matches at (typically the size of the total input window in bytes
+ * @param pOutData pointer to output buffer
+ * @param nMaxOutDataSize maximum size of output buffer, in bytes
+ *
+ * @return size of compressed data in output buffer, or -1 if the data is uncompressible
+ */
+static int lzsa_write_raw_uncompressed_block_v1(lzsa_compressor *pCompressor, const unsigned char *pInWindow, const int nStartOffset, const int nEndOffset, unsigned char *pOutData, const int nMaxOutDataSize) {
+   int nNumLiterals = nEndOffset - nStartOffset;
+   int nTokenLiteralsLen = (nNumLiterals >= LITERALS_RUN_LEN_V1) ? LITERALS_RUN_LEN_V1 : nNumLiterals;
+   int nOutOffset = 0;
+
+   int nCommandSize = 8 /* token */ + lzsa_get_literals_varlen_size_v1(nNumLiterals) + (nNumLiterals << 3) + 4;
+   if ((nOutOffset + (nCommandSize >> 3)) > nMaxOutDataSize)
+      return -1;
+
+   pCompressor->num_commands = 0;
+   pOutData[nOutOffset++] = (nTokenLiteralsLen << 4) | 0x0f;
+   
+   nOutOffset = lzsa_write_literals_varlen_v1(pOutData, nOutOffset, nNumLiterals);
+
+   if (nNumLiterals != 0) {
+      memcpy(pOutData + nOutOffset, pInWindow + nStartOffset, nNumLiterals);
+      nOutOffset += nNumLiterals;
+      nNumLiterals = 0;
+   }
+
+   pCompressor->num_commands++;
+
+   /* Emit EOD marker for raw block */
+
+   pOutData[nOutOffset++] = 0;
+   pOutData[nOutOffset++] = 238;
+   pOutData[nOutOffset++] = 0;
+   pOutData[nOutOffset++] = 0;
+
+   return nOutOffset;
+}
+
+/**
+ * Select the most optimal matches, reduce the token count if possible, and then emit a block of compressed LZSA1 data
+ *
+ * @param pCompressor compression context
+ * @param pInWindow pointer to input data window (previously compressed bytes + bytes to compress)
+ * @param nPreviousBlockSize number of previously compressed bytes (or 0 for none)
+ * @param nInDataSize number of input bytes to compress
+ * @param pOutData pointer to output buffer
+ * @param nMaxOutDataSize maximum size of output buffer, in bytes
+ *
+ * @return size of compressed data in output buffer, or -1 if the data is uncompressible
+ */
+int lzsa_optimize_and_write_block_v1(lzsa_compressor *pCompressor, const unsigned char *pInWindow, const int nPreviousBlockSize, const int nInDataSize, unsigned char *pOutData, const int nMaxOutDataSize) {
+   int nResult, nBaseCompressedSize;
+
+   /* Compress optimally without breaking ties in favor of less tokens */
+
+   memset(pCompressor->best_match, 0, BLOCK_SIZE * sizeof(lzsa_match));
+   lzsa_optimize_forward_v1(pCompressor, pCompressor->best_match - nPreviousBlockSize, nPreviousBlockSize, nPreviousBlockSize + nInDataSize, 0 /* reduce */);
+
+   int nDidReduce;
+   int nPasses = 0;
+   do {
+      nDidReduce = lzsa_optimize_command_count_v1(pCompressor, pInWindow, pCompressor->best_match - nPreviousBlockSize, nPreviousBlockSize, nPreviousBlockSize + nInDataSize);
+      nPasses++;
+   } while (nDidReduce && nPasses < 20);
+
+   nBaseCompressedSize = lzsa_get_compressed_size_v1(pCompressor, pCompressor->best_match - nPreviousBlockSize, nPreviousBlockSize, nPreviousBlockSize + nInDataSize);
+   lzsa_match *pBestMatch = pCompressor->best_match - nPreviousBlockSize;
+
+   if (nBaseCompressedSize > 0 && nInDataSize < 65536) {
+      int nReducedCompressedSize;
+
+      /* Compress optimally and do break ties in favor of less tokens */
+      memset(pCompressor->improved_match, 0, BLOCK_SIZE * sizeof(lzsa_match));
+      lzsa_optimize_forward_v1(pCompressor, pCompressor->improved_match - nPreviousBlockSize, nPreviousBlockSize, nPreviousBlockSize + nInDataSize, 1 /* reduce */);
+      
+      nPasses = 0;
+      do {
+         nDidReduce = lzsa_optimize_command_count_v1(pCompressor, pInWindow, pCompressor->improved_match - nPreviousBlockSize, nPreviousBlockSize, nPreviousBlockSize + nInDataSize);
+         nPasses++;
+      } while (nDidReduce && nPasses < 20);
+
+      nReducedCompressedSize = lzsa_get_compressed_size_v1(pCompressor, pCompressor->improved_match - nPreviousBlockSize, nPreviousBlockSize, nPreviousBlockSize + nInDataSize);
+      if (nReducedCompressedSize > 0 && nReducedCompressedSize <= nBaseCompressedSize) {
+         /* Pick the parse with the reduced number of tokens as it didn't negatively affect the size */
+         pBestMatch = pCompressor->improved_match - nPreviousBlockSize;
+      }
+   }
+
+   nResult = lzsa_write_block_v1(pCompressor, pBestMatch, pInWindow, nPreviousBlockSize, nPreviousBlockSize + nInDataSize, pOutData, nMaxOutDataSize);
+   if (nResult < 0 && pCompressor->flags & LZSA_FLAG_RAW_BLOCK) {
+      nResult = lzsa_write_raw_uncompressed_block_v1(pCompressor, pInWindow, nPreviousBlockSize, nPreviousBlockSize + nInDataSize, pOutData, nMaxOutDataSize);
+   }
+
+   return nResult;
+}

loader/tools/lzsa/src/shrink_block_v1.h

@@ -0,0 +1,53 @@
+/*
+ * shrink_block_v1.h - LZSA1 block compressor definitions
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#ifndef _SHRINK_BLOCK_V1_H
+#define _SHRINK_BLOCK_V1_H
+
+/* Forward declarations */
+typedef struct _lzsa_compressor lzsa_compressor;
+
+/**
+ * Select the most optimal matches, reduce the token count if possible, and then emit a block of compressed LZSA1 data
+ *
+ * @param pCompressor compression context
+ * @param pInWindow pointer to input data window (previously compressed bytes + bytes to compress)
+ * @param nPreviousBlockSize number of previously compressed bytes (or 0 for none)
+ * @param nInDataSize number of input bytes to compress
+ * @param pOutData pointer to output buffer
+ * @param nMaxOutDataSize maximum size of output buffer, in bytes
+ *
+ * @return size of compressed data in output buffer, or -1 if the data is uncompressible
+ */
+int lzsa_optimize_and_write_block_v1(lzsa_compressor *pCompressor, const unsigned char *pInWindow, const int nPreviousBlockSize, const int nInDataSize, unsigned char *pOutData, const int nMaxOutDataSize);
+
+#endif /* _SHRINK_BLOCK_V1_H */

loader/tools/lzsa/src/shrink_block_v2.h

@@ -0,0 +1,53 @@
+/*
+ * shrink_block_v2.h - LZSA2 block compressor definitions
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#ifndef _SHRINK_BLOCK_V2_H
+#define _SHRINK_BLOCK_V2_H
+
+/* Forward declarations */
+typedef struct _lzsa_compressor lzsa_compressor;
+
+/**
+ * Select the most optimal matches, reduce the token count if possible, and then emit a block of compressed LZSA2 data
+ *
+ * @param pCompressor compression context
+ * @param pInWindow pointer to input data window (previously compressed bytes + bytes to compress)
+ * @param nPreviousBlockSize number of previously compressed bytes (or 0 for none)
+ * @param nInDataSize number of input bytes to compress
+ * @param pOutData pointer to output buffer
+ * @param nMaxOutDataSize maximum size of output buffer, in bytes
+ *
+ * @return size of compressed data in output buffer, or -1 if the data is uncompressible
+ */
+int lzsa_optimize_and_write_block_v2(lzsa_compressor *pCompressor, const unsigned char *pInWindow, const int nPreviousBlockSize, const int nInDataSize, unsigned char *pOutData, const int nMaxOutDataSize);
+
+#endif /* _SHRINK_BLOCK_V2_H */

loader/tools/lzsa/src/shrink_context.c

@@ -0,0 +1,264 @@
+/*
+ * shrink_context.c - compression context implementation
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#include <stdlib.h>
+#include <string.h>
+#include "shrink_context.h"
+#include "shrink_block_v1.h"
+#include "shrink_block_v2.h"
+#include "format.h"
+#include "matchfinder.h"
+#include "lib.h"
+
+/**
+ * Initialize compression context
+ *
+ * @param pCompressor compression context to initialize
+ * @param nMaxWindowSize maximum size of input data window (previously compressed bytes + bytes to compress)
+ * @param nMinMatchSize minimum match size (cannot be less than MIN_MATCH_SIZE)
+ * @param nFlags compression flags
+ *
+ * @return 0 for success, non-zero for failure
+ */
+int lzsa_compressor_init(lzsa_compressor *pCompressor, const int nMaxWindowSize, const int nMinMatchSize, const int nFormatVersion, const int nFlags) {
+   int nResult;
+   int nMinMatchSizeForFormat = (nFormatVersion == 1) ? MIN_MATCH_SIZE_V1 : MIN_MATCH_SIZE_V2;
+   int nMaxMinMatchForFormat = (nFormatVersion == 1) ? 5 : 3;
+
+   nResult = divsufsort_init(&pCompressor->divsufsort_context);
+   pCompressor->intervals = NULL;
+   pCompressor->pos_data = NULL;
+   pCompressor->open_intervals = NULL;
+   pCompressor->match = NULL;
+   pCompressor->best_match = NULL;
+   pCompressor->improved_match = NULL;
+   pCompressor->arrival = NULL;
+   pCompressor->rep_slot_handled_mask = NULL;
+   pCompressor->rep_len_handled_mask = NULL;
+   pCompressor->first_offset_for_byte = NULL;
+   pCompressor->next_offset_for_pos = NULL;
+   pCompressor->min_match_size = nMinMatchSize;
+   if (pCompressor->min_match_size < nMinMatchSizeForFormat)
+      pCompressor->min_match_size = nMinMatchSizeForFormat;
+   else if (pCompressor->min_match_size > nMaxMinMatchForFormat)
+      pCompressor->min_match_size = nMaxMinMatchForFormat;
+   pCompressor->format_version = nFormatVersion;
+   pCompressor->flags = nFlags;
+   pCompressor->safe_dist = 0;
+   pCompressor->num_commands = 0;
+   
+   memset(&pCompressor->stats, 0, sizeof(pCompressor->stats));
+   pCompressor->stats.min_literals = -1;
+   pCompressor->stats.min_match_len = -1;
+   pCompressor->stats.min_offset = -1;
+   pCompressor->stats.min_rle1_len = -1;
+   pCompressor->stats.min_rle2_len = -1;
+   pCompressor->end_position = -1;
+
+   if (!nResult) {
+      pCompressor->intervals = (unsigned int *)malloc(nMaxWindowSize * sizeof(unsigned int));
+
+      if (pCompressor->intervals) {
+         pCompressor->pos_data = (unsigned int *)malloc(nMaxWindowSize * sizeof(unsigned int));
+
+         if (pCompressor->pos_data) {
+            pCompressor->open_intervals = (unsigned int *)malloc((LCP_AND_TAG_MAX + 1) * sizeof(unsigned int));
+
+            if (pCompressor->open_intervals) {
+               pCompressor->arrival = (lzsa_arrival *)malloc(((BLOCK_SIZE + 1) << ARRIVALS_PER_POSITION_SHIFT) * sizeof(lzsa_arrival));
+   
+               if (pCompressor->arrival) {
+                  pCompressor->best_match = (lzsa_match *)malloc(BLOCK_SIZE * sizeof(lzsa_match));
+
+                  if (pCompressor->best_match) {
+                     pCompressor->improved_match = (lzsa_match *)malloc(BLOCK_SIZE * sizeof(lzsa_match));
+
+                     if (pCompressor->improved_match) {
+                        if (pCompressor->format_version == 2)
+                           pCompressor->match = (lzsa_match *)malloc(BLOCK_SIZE * NMATCHES_PER_INDEX_V2 * sizeof(lzsa_match));
+                        else
+                           pCompressor->match = (lzsa_match *)malloc(BLOCK_SIZE * NMATCHES_PER_INDEX_V1 * sizeof(lzsa_match));
+                        if (pCompressor->match) {
+                           if (pCompressor->format_version == 2) {
+                              pCompressor->rep_slot_handled_mask = (char*)malloc(NARRIVALS_PER_POSITION_V2_BIG * ((LCP_MAX + 1) / 8) * sizeof(char));
+                              if (pCompressor->rep_slot_handled_mask) {
+                                 pCompressor->rep_len_handled_mask = (char*)malloc(((LCP_MAX + 1) / 8) * sizeof(char));
+                                 if (pCompressor->rep_len_handled_mask) {
+                                    pCompressor->first_offset_for_byte = (int*)malloc(65536 * sizeof(int));
+                                    if (pCompressor->first_offset_for_byte) {
+                                       pCompressor->next_offset_for_pos = (int*)malloc(BLOCK_SIZE * sizeof(int));
+                                       if (pCompressor->next_offset_for_pos) {
+                                          return 0;
+                                       }
+                                    }
+                                 }
+                              }
+                           }
+                           else {
+                              return 0;
+                           }
+                        }
+                     }
+                  }
+               }
+            }
+         }
+      }
+   }
+
+   lzsa_compressor_destroy(pCompressor);
+   return 100;
+}
+
+/**
+ * Clean up compression context and free up any associated resources
+ *
+ * @param pCompressor compression context to clean up
+ */
+void lzsa_compressor_destroy(lzsa_compressor *pCompressor) {
+   divsufsort_destroy(&pCompressor->divsufsort_context);
+
+   if (pCompressor->next_offset_for_pos) {
+      free(pCompressor->next_offset_for_pos);
+      pCompressor->next_offset_for_pos = NULL;
+   }
+
+   if (pCompressor->first_offset_for_byte) {
+      free(pCompressor->first_offset_for_byte);
+      pCompressor->first_offset_for_byte = NULL;
+   }
+
+   if (pCompressor->rep_len_handled_mask) {
+      free(pCompressor->rep_len_handled_mask);
+      pCompressor->rep_len_handled_mask = NULL;
+   }
+
+   if (pCompressor->rep_slot_handled_mask) {
+      free(pCompressor->rep_slot_handled_mask);
+      pCompressor->rep_slot_handled_mask = NULL;
+   }
+
+   if (pCompressor->match) {
+      free(pCompressor->match);
+      pCompressor->match = NULL;
+   }
+
+   if (pCompressor->improved_match) {
+      free(pCompressor->improved_match);
+      pCompressor->improved_match = NULL;
+   }
+
+   if (pCompressor->arrival) {
+      free(pCompressor->arrival);
+      pCompressor->arrival = NULL;
+   }
+
+   if (pCompressor->best_match) {
+      free(pCompressor->best_match);
+      pCompressor->best_match = NULL;
+   }
+
+   if (pCompressor->open_intervals) {
+      free(pCompressor->open_intervals);
+      pCompressor->open_intervals = NULL;
+   }
+
+   if (pCompressor->pos_data) {
+      free(pCompressor->pos_data);
+      pCompressor->pos_data = NULL;
+   }
+
+   if (pCompressor->intervals) {
+      free(pCompressor->intervals);
+      pCompressor->intervals = NULL;
+   }
+}
+
+/**
+ * Compress one block of data
+ *
+ * @param pCompressor compression context
+ * @param pInWindow pointer to input data window (previously compressed bytes + bytes to compress)
+ * @param nPreviousBlockSize number of previously compressed bytes (or 0 for none)
+ * @param nInDataSize number of input bytes to compress
+ * @param pOutData pointer to output buffer
+ * @param nMaxOutDataSize maximum size of output buffer, in bytes
+ *
+ * @return size of compressed data in output buffer, or -1 if the data is uncompressible
+ */
+int lzsa_compressor_shrink_block(lzsa_compressor *pCompressor, unsigned char *pInWindow, const int nPreviousBlockSize, const int nInDataSize, unsigned char *pOutData, const int nMaxOutDataSize) {
+   int nCompressedSize;
+
+   if (pCompressor->flags & LZSA_FLAG_RAW_BACKWARD) {
+      lzsa_reverse_buffer(pInWindow + nPreviousBlockSize, nInDataSize);
+   }
+
+   if (lzsa_build_suffix_array(pCompressor, pInWindow, nPreviousBlockSize + nInDataSize))
+      nCompressedSize = -1;
+   else {
+      if (nPreviousBlockSize) {
+         lzsa_skip_matches(pCompressor, 0, nPreviousBlockSize);
+      }
+      lzsa_find_all_matches(pCompressor, (pCompressor->format_version == 2) ? NMATCHES_PER_INDEX_V2 : NMATCHES_PER_INDEX_V1, nPreviousBlockSize, nPreviousBlockSize + nInDataSize);
+
+      if (pCompressor->format_version == 1) {
+         nCompressedSize = lzsa_optimize_and_write_block_v1(pCompressor, pInWindow, nPreviousBlockSize, nInDataSize, pOutData, nMaxOutDataSize);
+         if (nCompressedSize != -1 && (pCompressor->flags & LZSA_FLAG_RAW_BACKWARD)) {
+            lzsa_reverse_buffer(pOutData, nCompressedSize);
+         }
+      }
+      else if (pCompressor->format_version == 2) {
+         nCompressedSize = lzsa_optimize_and_write_block_v2(pCompressor, pInWindow, nPreviousBlockSize, nInDataSize, pOutData, nMaxOutDataSize);
+         if (nCompressedSize != -1 && (pCompressor->flags & LZSA_FLAG_RAW_BACKWARD)) {
+            lzsa_reverse_buffer(pOutData, nCompressedSize);
+         }
+      }
+      else {
+         nCompressedSize = -1;
+      }
+   }
+
+   if (pCompressor->flags & LZSA_FLAG_RAW_BACKWARD) {
+      lzsa_reverse_buffer(pInWindow + nPreviousBlockSize, nInDataSize);
+   }
+
+   return nCompressedSize;
+}
+
+/**
+ * Get the number of compression commands issued in compressed data blocks
+ *
+ * @return number of commands
+ */
+int lzsa_compressor_get_command_count(lzsa_compressor *pCompressor) {
+   return pCompressor->num_commands;
+}

loader/tools/lzsa/src/shrink_context.h

@@ -0,0 +1,184 @@
+/*
+ * shrink_context.h - compression context definitions
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#ifndef _SHRINK_CONTEXT_H
+#define _SHRINK_CONTEXT_H
+
+#include "divsufsort.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define LCP_BITS 14
+#define TAG_BITS 4
+#define LCP_MAX ((1U<<(LCP_BITS - TAG_BITS)) - 1)
+#define LCP_AND_TAG_MAX (1U<<(LCP_BITS - 1))
+#define LCP_SHIFT (31-LCP_BITS)
+#define LCP_MASK (((1U<<LCP_BITS) - 1) << LCP_SHIFT)
+#define POS_MASK ((1U<<LCP_SHIFT) - 1)
+#define VISITED_FLAG 0x80000000
+#define EXCL_VISITED_MASK  0x7fffffff
+
+#define NARRIVALS_PER_POSITION_V1 8
+#define NARRIVALS_PER_POSITION_V2_SMALL 9
+#define NARRIVALS_PER_POSITION_V2_BIG 32
+#define ARRIVALS_PER_POSITION_SHIFT 5
+
+#define NMATCHES_PER_INDEX_V1 8
+#define MATCHES_PER_INDEX_SHIFT_V1 3
+
+#define NMATCHES_PER_INDEX_V2 64
+#define MATCHES_PER_INDEX_SHIFT_V2 6
+
+#define LEAVE_ALONE_MATCH_SIZE 300
+#define LEAVE_ALONE_MATCH_SIZE_SMALL 1000
+
+#define MODESWITCH_PENALTY 3
+
+/** One match */
+typedef struct _lzsa_match {
+   unsigned short length;
+   unsigned short offset;
+} lzsa_match;
+
+/** Forward arrival slot */
+typedef struct {
+   int cost;
+   unsigned short rep_offset;
+   short from_slot;
+
+   int from_pos;
+   unsigned short rep_len;
+   unsigned short match_len;
+   int rep_pos;
+   int num_literals;
+   int score;
+} lzsa_arrival;
+
+/** Compression statistics */
+typedef struct _lzsa_stats {
+   int min_literals;
+   int max_literals;
+   int total_literals;
+
+   int min_offset;
+   int max_offset;
+   int num_rep_offsets;
+   int total_offsets;
+
+   int min_match_len;
+   int max_match_len;
+   int total_match_lens;
+
+   int min_rle1_len;
+   int max_rle1_len;
+   int total_rle1_lens;
+
+   int min_rle2_len;
+   int max_rle2_len;
+   int total_rle2_lens;
+
+   int literals_divisor;
+   int match_divisor;
+   int rle1_divisor;
+   int rle2_divisor;
+} lzsa_stats;
+
+/** Compression context */
+typedef struct _lzsa_compressor {
+   divsufsort_ctx_t divsufsort_context;
+   unsigned int *intervals;
+   unsigned int *pos_data;
+   unsigned int *open_intervals;
+   lzsa_match *match;
+   lzsa_match *best_match;
+   lzsa_match *improved_match;
+   lzsa_arrival *arrival;
+   char *rep_slot_handled_mask;
+   char *rep_len_handled_mask;
+   int *first_offset_for_byte;
+   int *next_offset_for_pos;
+   int min_match_size;
+   int format_version;
+   int flags;
+   int safe_dist;
+   int num_commands;
+   int end_position;
+   lzsa_stats stats;
+} lzsa_compressor;
+
+/**
+ * Initialize compression context
+ *
+ * @param pCompressor compression context to initialize
+ * @param nMaxWindowSize maximum size of input data window (previously compressed bytes + bytes to compress)
+ * @param nMinMatchSize minimum match size (cannot be less than MIN_MATCH_SIZE)
+ * @param nFlags compression flags
+ *
+ * @return 0 for success, non-zero for failure
+ */
+int lzsa_compressor_init(lzsa_compressor *pCompressor, const int nMaxWindowSize, const int nMinMatchSize, const int nFormatVersion, const int nFlags);
+
+/**
+ * Clean up compression context and free up any associated resources
+ *
+ * @param pCompressor compression context to clean up
+ */
+void lzsa_compressor_destroy(lzsa_compressor *pCompressor);
+
+/**
+ * Compress one block of data
+ *
+ * @param pCompressor compression context
+ * @param pInWindow pointer to input data window (previously compressed bytes + bytes to compress)
+ * @param nPreviousBlockSize number of previously compressed bytes (or 0 for none)
+ * @param nInDataSize number of input bytes to compress
+ * @param pOutData pointer to output buffer
+ * @param nMaxOutDataSize maximum size of output buffer, in bytes
+ *
+ * @return size of compressed data in output buffer, or -1 if the data is uncompressible
+ */
+int lzsa_compressor_shrink_block(lzsa_compressor *pCompressor, unsigned char *pInWindow, const int nPreviousBlockSize, const int nInDataSize, unsigned char *pOutData, const int nMaxOutDataSize);
+
+/**
+ * Get the number of compression commands issued in compressed data blocks
+ *
+ * @return number of commands
+ */
+int lzsa_compressor_get_command_count(lzsa_compressor *pCompressor);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SHRINK_CONTEXT_H */

loader/tools/lzsa/src/shrink_inmem.c

@@ -0,0 +1,185 @@
+/*
+ * shrink_inmem.c - in-memory compression implementation
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#include <stdlib.h>
+#include <string.h>
+#include "shrink_inmem.h"
+#include "shrink_context.h"
+#include "frame.h"
+#include "format.h"
+#include "lib.h"
+
+/**
+ * Get maximum compressed size of input(source) data
+ *
+ * @param nInputSize input(source) size in bytes
+ *
+ * @return maximum compressed size
+ */
+size_t lzsa_get_max_compressed_size_inmem(size_t nInputSize) {
+   return lzsa_get_header_size() + ((nInputSize + (BLOCK_SIZE - 1)) >> 16) * lzsa_get_frame_size() + nInputSize + lzsa_get_frame_size() /* footer */;
+}
+
+/**
+ * Compress memory
+ *
+ * @param pInputData pointer to input(source) data to compress
+ * @param pOutBuffer buffer for compressed data
+ * @param nInputSize input(source) size in bytes
+ * @param nMaxOutBufferSize maximum capacity of compression buffer
+ * @param nFlags compression flags (LZSA_FLAG_xxx)
+ * @param nMinMatchSize minimum match size
+ * @param nFormatVersion version of format to use (1-2)
+ *
+ * @return actual compressed size, or -1 for error
+ */
+size_t lzsa_compress_inmem(unsigned char *pInputData, unsigned char *pOutBuffer, size_t nInputSize, size_t nMaxOutBufferSize,
+                             const unsigned int nFlags, const int nMinMatchSize, const int nFormatVersion) {
+   lzsa_compressor compressor;
+   size_t nOriginalSize = 0;
+   size_t nCompressedSize = 0L;
+   int nResult;
+   int nError = 0;
+
+   nResult = lzsa_compressor_init(&compressor, BLOCK_SIZE * 2, nMinMatchSize, nFormatVersion, nFlags);
+   if (nResult != 0) {
+      return -1;
+   }
+
+   if ((nFlags & LZSA_FLAG_RAW_BLOCK) == 0) {
+      int nHeaderSize = lzsa_encode_header(pOutBuffer, (int)nMaxOutBufferSize, nFormatVersion);
+      if (nHeaderSize < 0)
+         nError = LZSA_ERROR_COMPRESSION;
+      else {
+         nCompressedSize += nHeaderSize;
+      }
+   }
+
+   int nPreviousBlockSize = 0;
+   int nNumBlocks = 0;
+
+   while (nOriginalSize < nInputSize && !nError) {
+      int nInDataSize;
+
+      nInDataSize = (int)(nInputSize - nOriginalSize);
+      if (nInDataSize > BLOCK_SIZE)
+         nInDataSize = BLOCK_SIZE;
+
+      if (nInDataSize > 0) {
+         if ((nFlags & LZSA_FLAG_RAW_BLOCK) != 0 && nNumBlocks) {
+            nError = LZSA_ERROR_RAW_TOOLARGE;
+            break;
+         }
+
+         int nOutDataSize;
+         int nOutDataEnd = (int)(nMaxOutBufferSize - (lzsa_get_frame_size() + nCompressedSize + lzsa_get_frame_size() /* footer */));
+         int nFrameSize = lzsa_get_frame_size();
+
+         if ((nFlags & LZSA_FLAG_RAW_BLOCK) != 0) {
+            nFrameSize = 0;
+            nOutDataEnd = (int)(nMaxOutBufferSize - nCompressedSize);
+         }
+
+         if (nOutDataEnd > BLOCK_SIZE)
+            nOutDataEnd = BLOCK_SIZE;
+
+         nOutDataSize = lzsa_compressor_shrink_block(&compressor, pInputData + nOriginalSize - nPreviousBlockSize, nPreviousBlockSize, nInDataSize, pOutBuffer + nFrameSize + nCompressedSize, nOutDataEnd);
+         if (nOutDataSize >= 0) {
+            /* Write compressed block */
+
+            if ((nFlags & LZSA_FLAG_RAW_BLOCK) == 0) {
+               int nBlockheaderSize = lzsa_encode_compressed_block_frame(pOutBuffer + nCompressedSize, (int)(nMaxOutBufferSize - nCompressedSize), nOutDataSize);
+               if (nBlockheaderSize < 0)
+                  nError = LZSA_ERROR_COMPRESSION;
+               else {
+                  nCompressedSize += nBlockheaderSize;
+               }
+            }
+
+            if (!nError) {
+               nOriginalSize += nInDataSize;
+               nCompressedSize += nOutDataSize;
+            }
+         }
+         else {
+            /* Write uncompressible, literal block */
+
+            if ((nFlags & LZSA_FLAG_RAW_BLOCK) != 0) {
+               nError = LZSA_ERROR_RAW_UNCOMPRESSED;
+               break;
+            }
+
+            int nBlockheaderSize = lzsa_encode_uncompressed_block_frame(pOutBuffer + nCompressedSize, (int)(nMaxOutBufferSize - nCompressedSize), nInDataSize);
+            if (nBlockheaderSize < 0)
+               nError = LZSA_ERROR_COMPRESSION;
+            else {
+               if ((size_t)nInDataSize > (nMaxOutBufferSize - (nCompressedSize + nBlockheaderSize)))
+                  nError = LZSA_ERROR_DST;
+               else {
+                  memcpy(pOutBuffer + nBlockheaderSize + nCompressedSize, pInputData + nOriginalSize, nInDataSize);
+
+                  nOriginalSize += nInDataSize;
+                  nCompressedSize += nBlockheaderSize + nInDataSize;
+               }
+            }
+         }
+
+         nPreviousBlockSize = nInDataSize;
+         nNumBlocks++;
+      }
+   }
+
+   if (!nError) {
+      int nFooterSize;
+
+      if ((nFlags & LZSA_FLAG_RAW_BLOCK) != 0) {
+         nFooterSize = 0;
+      }
+      else {
+         nFooterSize = lzsa_encode_footer_frame(pOutBuffer + nCompressedSize, (int)(nMaxOutBufferSize - nCompressedSize));
+         if (nFooterSize < 0)
+            nError = LZSA_ERROR_COMPRESSION;
+      }
+
+      nCompressedSize += nFooterSize;
+   }
+
+   lzsa_compressor_destroy(&compressor);
+
+   if (nError) {
+      return -1;
+   }
+   else {
+      return nCompressedSize;
+   }
+}
+

loader/tools/lzsa/src/shrink_inmem.h

@@ -0,0 +1,71 @@
+/*
+ * shrink_inmem.h - in-memory compression definitions
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#ifndef _SHRINK_INMEM_H
+#define _SHRINK_INMEM_H
+
+#include <stdlib.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * Get maximum compressed size of input(source) data
+ *
+ * @param nInputSize input(source) size in bytes
+ *
+ * @return maximum compressed size
+ */
+size_t lzsa_get_max_compressed_size_inmem(size_t nInputSize);
+
+/**
+ * Compress memory
+ *
+ * @param pInputData pointer to input(source) data to compress
+ * @param pOutBuffer buffer for compressed data
+ * @param nInputSize input(source) size in bytes
+ * @param nMaxOutBufferSize maximum capacity of compression buffer
+ * @param nFlags compression flags (LZSA_FLAG_xxx)
+ * @param nMinMatchSize minimum match size
+ * @param nFormatVersion version of format to use (1-2)
+ *
+ * @return actual compressed size, or -1 for error
+ */
+size_t lzsa_compress_inmem(unsigned char *pInputData, unsigned char *pOutBuffer, size_t nInputSize, size_t nMaxOutBufferSize,
+   const unsigned int nFlags, const int nMinMatchSize, const int nFormatVersion);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SHRINK_INMEM_H */

loader/tools/lzsa/src/shrink_streaming.c

@@ -0,0 +1,346 @@
+/*
+ * shrink_streaming.c - streaming compression implementation
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#include <stdlib.h>
+#include <string.h>
+#include "shrink_streaming.h"
+#include "format.h"
+#include "frame.h"
+#include "lib.h"
+#ifdef _WIN32
+#include <windows.h>
+#else
+#include <stdio.h>
+#endif
+
+/**
+ * Delete file
+ *
+ * @param pszInFilename name of file to delete
+ */
+static void lzsa_delete_file(const char *pszInFilename) {
+#ifdef _WIN32
+   DeleteFileA(pszInFilename);
+#else
+   remove(pszInFilename);
+#endif
+}
+
+/*-------------- File API -------------- */
+
+/**
+ * Compress file
+ *
+ * @param pszInFilename name of input(source) file to compress
+ * @param pszOutFilename name of output(compressed) file to generate
+ * @param pszDictionaryFilename name of dictionary file, or NULL for none
+ * @param nFlags compression flags (LZSA_FLAG_xxx)
+ * @param nMinMatchSize minimum match size
+ * @param nFormatVersion version of format to use (1-2)
+ * @param progress progress function, called after compressing each block, or NULL for none
+ * @param pOriginalSize pointer to returned input(source) size, updated when this function is successful
+ * @param pCompressedSize pointer to returned output(compressed) size, updated when this function is successful
+ * @param pCommandCount pointer to returned token(compression commands) count, updated when this function is successful
+ * @param pSafeDist pointer to return safe distance for raw blocks, updated when this function is successful
+ * @param pStats pointer to compression stats that are filled if this function is successful, or NULL
+ *
+ * @return LZSA_OK for success, or an error value from lzsa_status_t
+ */
+lzsa_status_t lzsa_compress_file(const char *pszInFilename, const char *pszOutFilename, const char *pszDictionaryFilename, const unsigned int nFlags, const int nMinMatchSize, const int nFormatVersion,
+      void(*progress)(long long nOriginalSize, long long nCompressedSize), long long *pOriginalSize, long long *pCompressedSize, int *pCommandCount, int *pSafeDist, lzsa_stats *pStats) {
+   lzsa_stream_t inStream, outStream;
+   void *pDictionaryData = NULL;
+   int nDictionaryDataSize = 0;
+   int nLoadAddress = 0;
+   unsigned char pInData[2] = { 0 };
+   lzsa_status_t nStatus;
+
+   if (lzsa_filestream_open(&inStream, pszInFilename, "rb") < 0) {
+      return LZSA_ERROR_SRC;
+   }
+
+   if (lzsa_filestream_open(&outStream, pszOutFilename, "wb") < 0) {
+      inStream.close(&inStream);
+      return LZSA_ERROR_DST;
+   }
+
+   nStatus = lzsa_dictionary_load(pszDictionaryFilename, &pDictionaryData, &nDictionaryDataSize);
+
+   if (nStatus) {
+      outStream.close(&outStream);
+      inStream.close(&inStream);
+      lzsa_delete_file(pszOutFilename);
+      return nStatus;
+   }
+
+   int nInDataSize;
+   nInDataSize = inStream.read(&inStream, pInData, 2);
+   if (nInDataSize == 2) {
+      nLoadAddress = pInData[0] | (pInData[1] << 8);
+   } else {
+      return LZSA_ERROR_SRC;
+   }
+
+   nStatus = lzsa_compress_stream(&inStream, &outStream, pDictionaryData, nDictionaryDataSize, nFlags, nMinMatchSize, nFormatVersion, progress, pOriginalSize, pCompressedSize, pCommandCount, pSafeDist, pStats, nLoadAddress);
+
+   lzsa_dictionary_free(&pDictionaryData);
+   outStream.close(&outStream);
+   inStream.close(&inStream);
+
+   if (nStatus) {
+      lzsa_delete_file(pszOutFilename);
+   }
+
+   return nStatus;
+}
+
+/*-------------- Streaming API -------------- */
+
+/**
+ * Compress stream
+ *
+ * @param pInStream input(source) stream to compress
+ * @param pOutStream output(compressed) stream to write to
+ * @param pDictionaryData dictionary contents, or NULL for none
+ * @param nDictionaryDataSize size of dictionary contents, or 0
+ * @param nFlags compression flags (LZSA_FLAG_xxx)
+ * @param nMinMatchSize minimum match size
+ * @param nFormatVersion version of format to use (1-2)
+ * @param progress progress function, called after compressing each block, or NULL for none
+ * @param pOriginalSize pointer to returned input(source) size, updated when this function is successful
+ * @param pCompressedSize pointer to returned output(compressed) size, updated when this function is successful
+ * @param pCommandCount pointer to returned token(compression commands) count, updated when this function is successful
+ * @param pSafeDist pointer to return safe distance for raw blocks, updated when this function is successful
+ * @param pStats pointer to compression stats that are filled if this function is successful, or NULL
+ *
+ * @return LZSA_OK for success, or an error value from lzsa_status_t
+ */
+lzsa_status_t lzsa_compress_stream(lzsa_stream_t *pInStream, lzsa_stream_t *pOutStream, const void *pDictionaryData, int nDictionaryDataSize,
+                                   const unsigned int nFlags, const int nMinMatchSize, const int nFormatVersion,
+                                   void(*progress)(long long nOriginalSize, long long nCompressedSize), long long *pOriginalSize, long long *pCompressedSize, int *pCommandCount, int *pSafeDist, lzsa_stats *pStats, int nLoadAddress) {
+   unsigned char *pInData, *pOutData;
+   lzsa_compressor compressor;
+   long long nOriginalSize = 0LL, nCompressedSize = 0LL;
+   int nResult;
+   unsigned char cFrameData[16];
+   int nError = 0;
+   int nRawPadding = (nFlags & LZSA_FLAG_RAW_BLOCK) ? 8 : 0;
+   int nSafetyMargin = 0;
+
+   if ((nFlags & LZSA_FLAG_RAW_BLOCK) != 0) {
+      nSafetyMargin = 3;
+   } else {
+      nSafetyMargin = 0;
+   }
+
+   pInData = (unsigned char*)malloc(BLOCK_SIZE * 2);
+   if (!pInData) {
+      return LZSA_ERROR_MEMORY;
+   }
+   memset(pInData, 0, BLOCK_SIZE * 2);
+
+   pOutData = (unsigned char*)malloc(BLOCK_SIZE);
+   if (!pOutData) {
+      free(pInData);
+      pInData = NULL;
+
+      return LZSA_ERROR_MEMORY;
+   }
+   memset(pOutData, 0, BLOCK_SIZE);
+
+   nResult = lzsa_compressor_init(&compressor, BLOCK_SIZE * 2, nMinMatchSize, nFormatVersion, nFlags);
+   if (nResult != 0) {
+      free(pOutData);
+      pOutData = NULL;
+
+      free(pInData);
+      pInData = NULL;
+
+      return LZSA_ERROR_MEMORY;
+   }
+
+   int nPreviousBlockSize = 0;
+   int nNumBlocks = 0;
+
+   while (!pInStream->eof(pInStream) && !nError) {
+      int nInDataSize;
+
+      if (nPreviousBlockSize) {
+         memcpy(pInData + BLOCK_SIZE - nPreviousBlockSize, pInData + BLOCK_SIZE, nPreviousBlockSize);
+      }
+      else if (nDictionaryDataSize && pDictionaryData) {
+         nPreviousBlockSize = nDictionaryDataSize;
+         memcpy(pInData + BLOCK_SIZE - nPreviousBlockSize, pDictionaryData, nPreviousBlockSize);
+      }
+
+      nInDataSize = (int)pInStream->read(pInStream, pInData + BLOCK_SIZE, BLOCK_SIZE);
+      if (nInDataSize > 0) {
+         if ((nFlags & LZSA_FLAG_RAW_BLOCK) != 0 && nNumBlocks) {
+            nError = LZSA_ERROR_RAW_TOOLARGE;
+            break;
+         }
+         nDictionaryDataSize = 0;
+
+         int nOutDataSize;
+
+         nOutDataSize = lzsa_compressor_shrink_block(&compressor, pInData + BLOCK_SIZE - nPreviousBlockSize, nPreviousBlockSize, nInDataSize, pOutData, ((nInDataSize + nRawPadding) >= BLOCK_SIZE) ? BLOCK_SIZE : (nInDataSize + nRawPadding));
+         if (nOutDataSize >= 0) {
+//            if ((nFlags & LZSA_FLAG_RAW_BLOCK) == 0) {
+
+               int nHeaderSize = 0;
+               int nRealLoadAddress = nInDataSize - nOutDataSize + nLoadAddress;
+               printf("%x %x\n", nOutDataSize, compressor.end_position);
+               /* New inplace LoadAddress */
+               nHeaderSize = lzsa_encode_addr_le(cFrameData, 16, nRealLoadAddress + nSafetyMargin - 4);	//-4 as 4 header bytes are added
+               if (nHeaderSize < 0)
+                  nError = LZSA_ERROR_COMPRESSION;
+               else {
+                  if (pOutStream->write(pOutStream, cFrameData, nHeaderSize) != nHeaderSize)
+                     nError = LZSA_ERROR_DST;
+                  nCompressedSize += (long long)nHeaderSize;
+               }
+               /* EndAddress */
+               nHeaderSize = lzsa_encode_addr_be(cFrameData, 16, nRealLoadAddress + compressor.end_position);
+               if (nHeaderSize < 0)
+                  nError = LZSA_ERROR_COMPRESSION;
+               else {
+                  if (pOutStream->write(pOutStream, cFrameData, nHeaderSize) != nHeaderSize)
+                     nError = LZSA_ERROR_DST;
+                  nCompressedSize += (long long)nHeaderSize;
+               }
+               /* DestAddress */
+               nHeaderSize = lzsa_encode_addr_be(cFrameData, 16, nLoadAddress);
+               if (nHeaderSize < 0)
+                  nError = LZSA_ERROR_COMPRESSION;
+               else {
+                  if (pOutStream->write(pOutStream, cFrameData, nHeaderSize) != nHeaderSize)
+                     nError = LZSA_ERROR_DST;
+                  nCompressedSize += (long long)nHeaderSize;
+               }
+ //           }
+
+            if (!nError) {
+               if (pOutStream->write(pOutStream, pOutData, (size_t)nOutDataSize) != (size_t)nOutDataSize) {
+                  nError = LZSA_ERROR_DST;
+               }
+               else {
+                  nOriginalSize += (long long)nInDataSize;
+                  nCompressedSize += (long long)nOutDataSize;
+               }
+            }
+         }
+         else {
+            /* Write uncompressible, literal block */
+
+            if ((nFlags & LZSA_FLAG_RAW_BLOCK) != 0) {
+               nError = LZSA_ERROR_RAW_UNCOMPRESSED;
+               break;
+            }
+
+            int nBlockheaderSize = lzsa_encode_uncompressed_block_frame(cFrameData, 16, nInDataSize);
+            if (nBlockheaderSize < 0)
+               nError = LZSA_ERROR_COMPRESSION;
+            else {
+               if (pOutStream->write(pOutStream, cFrameData, nBlockheaderSize) != (size_t)nBlockheaderSize) {
+                  nError = LZSA_ERROR_DST;
+               }
+               else {
+                  if (pOutStream->write(pOutStream, pInData + BLOCK_SIZE, (size_t)nInDataSize) != (size_t)nInDataSize) {
+                     nError = LZSA_ERROR_DST;
+                  }
+                  else {
+                     nOriginalSize += (long long)nInDataSize;
+                     nCompressedSize += (long long)nBlockheaderSize + (long long)nInDataSize;
+                  }
+               }
+            }
+         }
+
+         nPreviousBlockSize = nInDataSize;
+         nNumBlocks++;
+      }
+
+      if (!nError && !pInStream->eof(pInStream)) {
+         if (progress)
+            progress(nOriginalSize, nCompressedSize);
+      }
+   }
+
+   if (!nError) {
+      int nFooterSize = 0;
+
+//      if ((nFlags & LZSA_FLAG_RAW_BLOCK) != 0) {
+//         nFooterSize = 0;
+//      }
+//      else {
+//         nFooterSize = lzsa_encode_footer_frame(cFrameData, 16);
+//         if (nFooterSize < 0)
+//            nError = LZSA_ERROR_COMPRESSION;
+//      }
+
+      if (pOutStream->write(pOutStream, cFrameData, nFooterSize) != nFooterSize)
+         nError = LZSA_ERROR_DST;
+      nCompressedSize += (long long)nFooterSize;
+   }
+
+   if (progress)
+      progress(nOriginalSize, nCompressedSize);
+
+   int nCommandCount = lzsa_compressor_get_command_count(&compressor);
+   int nSafeDist = compressor.safe_dist;
+
+   if (pStats)
+      *pStats = compressor.stats;
+
+   lzsa_compressor_destroy(&compressor);
+
+   free(pOutData);
+   pOutData = NULL;
+
+   free(pInData);
+   pInData = NULL;
+
+   if (nError) {
+      return nError;
+   }
+   else {
+      if (pOriginalSize)
+         *pOriginalSize = nOriginalSize;
+      if (pCompressedSize)
+         *pCompressedSize = nCompressedSize;
+      if (pCommandCount)
+         *pCommandCount = nCommandCount;
+      if (pSafeDist)
+         *pSafeDist = nSafeDist;
+      return LZSA_OK;
+   }
+}

loader/tools/lzsa/src/shrink_streaming.h

@@ -0,0 +1,99 @@
+/*
+ * shrink_streaming.h - streaming compression definitions
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#ifndef _SHRINK_STREAMING_H
+#define _SHRINK_STREAMING_H
+
+#include "stream.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/* Forward declaration */
+typedef enum _lzsa_status_t lzsa_status_t;
+typedef struct _lzsa_stats lzsa_stats;
+
+/*-------------- File API -------------- */
+
+/**
+ * Compress file
+ *
+ * @param pszInFilename name of input(source) file to compress
+ * @param pszOutFilename name of output(compressed) file to generate
+ * @param pszDictionaryFilename name of dictionary file, or NULL for none
+ * @param nFlags compression flags (LZSA_FLAG_xxx)
+ * @param nMinMatchSize minimum match size
+ * @param nFormatVersion version of format to use (1-2)
+ * @param progress progress function, called after compressing each block, or NULL for none
+ * @param pOriginalSize pointer to returned input(source) size, updated when this function is successful
+ * @param pCompressedSize pointer to returned output(compressed) size, updated when this function is successful
+ * @param pCommandCount pointer to returned token(compression commands) count, updated when this function is successful
+ * @param pSafeDist pointer to return safe distance for raw blocks, updated when this function is successful
+ * @param pStats pointer to compression stats that are filled if this function is successful, or NULL
+ *
+ * @return LZSA_OK for success, or an error value from lzsa_status_t
+ */
+lzsa_status_t lzsa_compress_file(const char *pszInFilename, const char *pszOutFilename, const char *pszDictionaryFilename,
+   const unsigned int nFlags, const int nMinMatchSize, const int nFormatVersion,
+   void(*progress)(long long nOriginalSize, long long nCompressedSize), long long *pOriginalSize, long long *pCompressedSize, int *pCommandCount, int *pSafeDist, lzsa_stats *pStats);
+
+/*-------------- Streaming API -------------- */
+
+/**
+ * Compress stream
+ *
+ * @param pInStream input(source) stream to compress
+ * @param pOutStream output(compressed) stream to write to
+ * @param pDictionaryData dictionary contents, or NULL for none
+ * @param nDictionaryDataSize size of dictionary contents, or 0
+ * @param nFlags compression flags (LZSA_FLAG_xxx)
+ * @param nMinMatchSize minimum match size
+ * @param nFormatVersion version of format to use (1-2)
+ * @param progress progress function, called after compressing each block, or NULL for none
+ * @param pOriginalSize pointer to returned input(source) size, updated when this function is successful
+ * @param pCompressedSize pointer to returned output(compressed) size, updated when this function is successful
+ * @param pCommandCount pointer to returned token(compression commands) count, updated when this function is successful
+ * @param pSafeDist pointer to return safe distance for raw blocks, updated when this function is successful
+ * @param pStats pointer to compression stats that are filled if this function is successful, or NULL
+ *
+ * @return LZSA_OK for success, or an error value from lzsa_status_t
+ */
+lzsa_status_t lzsa_compress_stream(lzsa_stream_t *pInStream, lzsa_stream_t *pOutStream, const void *pDictionaryData, int nDictionaryDataSize,
+   const unsigned int nFlags, const int nMinMatchSize, const int nFormatVersion,
+   void(*progress)(long long nOriginalSize, long long nCompressedSize), long long *pOriginalSize, long long *pCompressedSize, int *pCommandCount, int *pSafeDist, lzsa_stats *pStats, int nLoadAddress);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SHRINK_STREAMING_H */

loader/tools/lzsa/src/stream.c

@@ -0,0 +1,111 @@
+/*
+ * stream.c - streaming I/O  implementation
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include "stream.h"
+
+/**
+ * Close file stream
+ *
+ * @param stream stream
+ */
+static void lzsa_filestream_close(lzsa_stream_t *stream) {
+   if (stream->obj) {
+      fclose((FILE*)stream->obj);
+      stream->obj = NULL;
+      stream->read = NULL;
+      stream->write = NULL;
+      stream->eof = NULL;
+      stream->close = NULL;
+   }
+}
+
+/**
+ * Read from file stream
+ *
+ * @param stream stream
+ * @param ptr buffer to read into
+ * @param size number of bytes to read
+ *
+ * @return number of bytes read
+ */
+static size_t lzsa_filestream_read(lzsa_stream_t *stream, void *ptr, size_t size) {
+   return fread(ptr, 1, size, (FILE*)stream->obj);
+}
+
+/**
+ * Write to file stream
+ *
+ * @param stream stream
+ * @param ptr buffer to write from
+ * @param size number of bytes to write
+ *
+ * @return number of bytes written
+ */
+static size_t lzsa_filestream_write(lzsa_stream_t *stream, void *ptr, size_t size) {
+   return fwrite(ptr, 1, size, (FILE*)stream->obj);
+}
+
+/**
+ * Check if file stream has reached the end of the data
+ *
+ * @param stream stream
+ *
+ * @return nonzero if the end of the data has been reached, 0 if there is more data
+ */
+static int lzsa_filestream_eof(lzsa_stream_t *stream) {
+   return feof((FILE*)stream->obj);
+}
+
+/**
+ * Open file and create an I/O stream from it
+ *
+ * @param stream stream to fill out
+ * @param pszInFilename filename
+ * @param pszMode open mode, as with fopen()
+ *
+ * @return 0 for success, nonzero for failure
+ */
+int lzsa_filestream_open(lzsa_stream_t *stream, const char *pszInFilename, const char *pszMode) {
+   stream->obj = (void*)fopen(pszInFilename, pszMode);
+   if (stream->obj) {
+      stream->read = lzsa_filestream_read;
+      stream->write = lzsa_filestream_write;
+      stream->eof = lzsa_filestream_eof;
+      stream->close = lzsa_filestream_close;
+      return 0;
+   }
+   else
+      return -1;
+}

loader/tools/lzsa/src/stream.h

@@ -0,0 +1,103 @@
+/*
+ * stream.h - streaming I/O definitions
+ *
+ * Copyright (C) 2019 Emmanuel Marty
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/*
+ * Uses the libdivsufsort library Copyright (c) 2003-2008 Yuta Mori
+ *
+ * Inspired by LZ4 by Yann Collet. https://github.com/lz4/lz4
+ * With help, ideas, optimizations and speed measurements by spke <zxintrospec@gmail.com>
+ * With ideas from Lizard by Przemyslaw Skibinski and Yann Collet. https://github.com/inikep/lizard
+ * Also with ideas from smallz4 by Stephan Brumme. https://create.stephan-brumme.com/smallz4/
+ *
+ */
+
+#ifndef _STREAM_H
+#define _STREAM_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/* Forward declaration */
+typedef struct _lzsa_stream_t lzsa_stream_t;
+
+/* I/O stream */
+typedef struct _lzsa_stream_t {
+   /** Opaque stream-specific pointer */
+   void *obj;
+
+   /**
+    * Read from stream
+    *
+    * @param stream stream
+    * @param ptr buffer to read into
+    * @param size number of bytes to read
+    *
+    * @return number of bytes read
+    */
+   size_t(*read)(lzsa_stream_t *stream, void *ptr, size_t size);
+
+   /**
+    * Write to stream
+    *
+    * @param stream stream
+    * @param ptr buffer to write from
+    * @param size number of bytes to write
+    *
+    * @return number of bytes written
+    */
+   size_t(*write)(lzsa_stream_t *stream, void *ptr, size_t size);
+
+
+   /**
+    * Check if stream has reached the end of the data
+    *
+    * @param stream stream
+    *
+    * @return nonzero if the end of the data has been reached, 0 if there is more data
+    */
+   int(*eof)(lzsa_stream_t *stream);
+
+   /**
+    * Close stream
+    *
+    * @param stream stream
+    */
+   void(*close)(lzsa_stream_t *stream);
+} lzsa_stream_t;
+
+/**
+ * Open file and create an I/O stream from it
+ *
+ * @param stream stream to fill out
+ * @param pszInFilename filename
+ * @param pszMode open mode, as with fopen()
+ *
+ * @return 0 for success, nonzero for failure
+ */
+int lzsa_filestream_open(lzsa_stream_t *stream, const char *pszInFilename, const char *pszMode);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _STREAM_H */