ceph/src/isa-l/Release_notes.txt

   1 v2.18 Intel Intelligent Storage Acceleration Library Release Notes
   2 ==================================================================
   3
   4 RELEASE NOTE CONTENTS
   5 1. KNOWN ISSUES
   6 2. FIXED ISSUES
   7 3. CHANGE LOG & FEATURES ADDED
   8
   9 1. KNOWN ISSUES
  10 ----------------
  11
  12 * Perf tests do not run in Windows environment.
  13
  14 * 32-bit lib is not supported in Windows.
  15
  16 2. FIXED ISSUES
  17 ---------------
  18
  19 v2.18
  20
  21 * Mac OS X/darwin systems no longer require the --target=darwin config option.
  22   The autoconf canonical build should detect.
  23
  24 v2.17
  25
  26 * Fix igzip using 32K window and a shared object
  27
  28 * Fix igzip undefined instruction error on Nehalem.
  29
  30 * Fixed issue in crc performance tests where OS optimizations turned cold cache
  31   tests into warm tests.
  32
  33 v2.15
  34
  35 * Fix for windows register save in gf_6vect_mad_avx2.asm.  Only affects windows
  36   versions of ec_encode_data_update() running with AVX2.  A GP register was not
  37   properly restored resulting in corruption on return.
  38
  39 v2.14
  40
  41 * Building in unit directories is no longer supported removing the issue of
  42   leftover object files causing the top-level make build to fail.
  43
  44 v2.10
  45
  46 * Fix for windows register save overlap in gf_{3-6}vect_dot_prod_sse.asm. Only
  47   affects windows versions of erasure code.  GP register saves/restore were
  48   pushed to same stack area as XMM.
  49
  50 3. CHANGE LOG & FEATURES ADDED
  51 ------------------------------
  52
  53 v2.18
  54
  55 * New 2-pass fully-dynamic deflate compression (level -1).  ISA-L fast deflate
  56   now has two levels.  Level 0 (default) is the same as previous generations.
  57   Setting to level 1 will switch to the fully-dynamic compression that will
  58   typically reach higher compression ratios.
  59
  60 * RAID AVX512 functions.
  61
  62 v2.17
  63
  64 * New fast decompression (inflate)
  65
  66 * Compression improvements (deflate)
  67   - Speed and compression ratio improvements.
  68   - Fast custom Huffman code generation.
  69   - New features:
  70     * Run-time option of gzip crc calculation and headers/trailer.
  71     * Choice of static header (BTYPE 01) blocks.
  72     * LARGE_WINDOW, 32K history, now default.
  73     * Stateless full flush mode.
  74
  75 * CRC64
  76   - Six new 64-bit polynomials supported. Normal and reflected versions of ECMA,
  77     ISO and Jones polynomials.
  78
  79 v2.16
  80
  81 * Units added: crc, raid, igzip (deflate compression).
  82
  83 v2.15
  84
  85 * Erasure code updates. New AVX512 versions.
  86
  87 * Nasm support.  ISA-L ported to build with nasm or yasm assembler.
  88
  89 * Windows DLL support.  Windows builds DLL by default.
  90
  91 v2.14
  92
  93 * Autoconf and autotools build allows easier porting to additional systems.
  94   Previous make system still available to embedded users with Makefile.unx.
  95
  96 * Includes update for building on Mac OS X/darwin systems. Add --target=darwin
  97   to ./configure step.
  98
  99 v2.13
 100
 101 * Erasure code improvments
 102   - 32-bit port of optimized gf_vect_dot_prod() functions.  This makes
 103     ec_encode_data() functions much faster on 32-bit processors.
 104   - Avoton performance improvements.  Performance on Avoton for
 105     gf_vect_dot_prod() and ec_encode_data() can improve by as much as 20%.
 106
 107 v2.11
 108
 109 * Incremental erasure code.  New functions added to erasure code to handle
 110   single source update of code blocks.  The function ec_encode_data_update()
 111   works with parameters similar to ec_encode_data() but are called incrementally
 112   with each source block.  These versions are useful when source blocks are not
 113   all available at once.
 114
 115 v2.10
 116
 117 * Erasure code updates
 118   - New AVX and AVX2 support functions.
 119   - Changes min len requirement on gf_vect_dot_prod() to 32 from 16.
 120   - Tests include both source and parity recovery with ec_encode_data().
 121   - New encoding examples with Vandermonde or Cauchy matrix.
 122
 123 v2.8
 124
 125 * First open release of erasure code unit that is part of ISA-L.