src/liblzma/rangecoder/range_common.h

   1 // SPDX-License-Identifier: 0BSD
   2
   3 ///////////////////////////////////////////////////////////////////////////////
   4 //
   5 /// \file       range_common.h
   6 /// \brief      Common things for range encoder and decoder
   7 ///
   8 //  Authors:    Igor Pavlov
   9 //              Lasse Collin
  10 //
  11 ///////////////////////////////////////////////////////////////////////////////
  12
  13 #ifndef LZMA_RANGE_COMMON_H
  14 #define LZMA_RANGE_COMMON_H
  15
  16 // Skip common.h if building price_tablegen.c.
  17 #ifndef BUILDING_PRICE_TABLEGEN
  18 #       include "common.h"
  19 #endif
  20
  21
  22 ///////////////
  23 // Constants //
  24 ///////////////
  25
  26 #define RC_SHIFT_BITS 8
  27 #define RC_TOP_BITS 24
  28 #define RC_TOP_VALUE (UINT32_C(1) << RC_TOP_BITS)
  29 #define RC_BIT_MODEL_TOTAL_BITS 11
  30 #define RC_BIT_MODEL_TOTAL (UINT32_C(1) << RC_BIT_MODEL_TOTAL_BITS)
  31 #define RC_MOVE_BITS 5
  32
  33
  34 ////////////
  35 // Macros //
  36 ////////////
  37
  38 // Resets the probability so that both 0 and 1 have probability of 50 %
  39 #define bit_reset(prob) \
  40         prob = RC_BIT_MODEL_TOTAL >> 1
  41
  42 // This does the same for a complete bit tree.
  43 // (A tree represented as an array.)
  44 #define bittree_reset(probs, bit_levels) \
  45         for (uint32_t bt_i = 0; bt_i < (1 << (bit_levels)); ++bt_i) \
  46                 bit_reset((probs)[bt_i])
  47
  48
  49 //////////////////////
  50 // Type definitions //
  51 //////////////////////
  52
  53 /// \brief      Type of probabilities used with range coder
  54 ///
  55 /// This needs to be at least 12-bit integer, so uint16_t is a logical choice.
  56 /// However, on some architecture and compiler combinations, a bigger type
  57 /// may give better speed, because the probability variables are accessed
  58 /// a lot. On the other hand, bigger probability type increases cache
  59 /// footprint, since there are 2 to 14 thousand probability variables in
  60 /// LZMA (assuming the limit of lc + lp <= 4; with lc + lp <= 12 there
  61 /// would be about 1.5 million variables).
  62 ///
  63 /// With malicious files, the initialization speed of the LZMA decoder can
  64 /// become important. In that case, smaller probability variables mean that
  65 /// there is less bytes to write to RAM, which makes initialization faster.
  66 /// With big probability type, the initialization can become so slow that it
  67 /// can be a problem e.g. for email servers doing virus scanning.
  68 ///
  69 /// I will be sticking to uint16_t unless some specific architectures
  70 /// are *much* faster (20-50 %) with uint32_t.
  71 ///
  72 /// Update in 2024: The branchless C and x86-64 assembly was written so that
  73 /// probability is assumed to be uint16_t. (In contrast, LZMA SDK 23.01
  74 /// assembly supports both types.)
  75 typedef uint16_t probability;
  76
  77 #endif