1 v2.24 Intel Intelligent Storage Acceleration Library Release Notes
2 ==================================================================
7 3. CHANGE LOG & FEATURES ADDED
12 * Perf tests do not run in Windows environment.
14 * 32-bit lib is not supported in Windows.
20 * Fix for crc32_iscsi(). Potential read-over for small buffer. For an input
21 buffer length of less than 8 bytes and aligned to an 8 byte boundary, function
22 could read past length. Previously had the possibility to cause a seg fault
23 only for length 0 and invalid buffer passed. Calculated CRC is unchanged.
25 * Fix for compression/decompression of > 4GB files. For streaming compression
26 of extremely large files, the total_out parameter would wrap and could
27 potentially flag an otherwise valid lookback distance as being invalid.
28 Total_out is still 32bit for zlib compatibility. No inconsistent compressed
29 buffers were generated by the issue.
33 * Fix for histogram generation base function.
34 * Fix library build warnings on macOS.
35 * Fix igzip to use bsf instruction when tzcnt is not available.
39 * Fix ISA-L builds for other architectures. Base function and examples
40 sanitized for non-IA builds.
42 * Fix fuzz test script to work with llvm 6.0 builtin libFuzz.
46 * Inflate total_out behavior corrected for in-progress decompression.
47 Previously total_out represented the total bytes decompressed into the output
48 buffer or temp internal buffer. This is changed to be only the bytes put into
51 * Fixed issue with isal_create_hufftables_subset. Affects semi-dynamic
52 compression use case when explicitly creating hufftables from histogram. The
53 _hufftables_subset function could fail to generate length symbols for any
54 length that were never seen.
58 * Fix erasure code test that violates rs matrix bounds.
60 * Fix 0 length file and looping errors in igzip_inflate_test.
64 * Mac OS X/darwin systems no longer require the --target=darwin config option.
65 The autoconf canonical build should detect.
69 * Fix igzip using 32K window and a shared object
71 * Fix igzip undefined instruction error on Nehalem.
73 * Fixed issue in crc performance tests where OS optimizations turned cold cache
74 tests into warm tests.
78 * Fix for windows register save in gf_6vect_mad_avx2.asm. Only affects windows
79 versions of ec_encode_data_update() running with AVX2. A GP register was not
80 properly restored resulting in corruption on return.
84 * Building in unit directories is no longer supported removing the issue of
85 leftover object files causing the top-level make build to fail.
89 * Fix for windows register save overlap in gf_{3-6}vect_dot_prod_sse.asm. Only
90 affects windows versions of erasure code. GP register saves/restore were
91 pushed to same stack area as XMM.
93 3. CHANGE LOG & FEATURES ADDED
94 ------------------------------
97 * Igzip small file performance improvements and new features.
98 - Better performance on small files.
99 - New gzip/zlib header and trailer handling.
100 - New gzip/zlib header parsing helper functions.
101 - New user-space compression/decompression tool igzip.
103 * New mem unit added with first function isal_zero_detect().
107 * Igzip inflate (decompression) performance improvements.
108 - Implemented multi-byte decode for inflate. Decode can pack up to three
109 symbols into the decode table making some compressed streams decompress much
110 faster depending on the prevalence of short codes.
114 * Igzip: AVX2 version of level 3 compression added.
116 * Erasure code examples
117 - New examples for standard EC encode and decode.
118 - Example of piggyback EC encode and decode.
123 - New compression levels added. ISA-L fast deflate now has more levels to
124 balance speed vs. target compression level. Level 0, 1 are as in previous
125 generations. New levels 2 & 3 target higher compression roughly comparable
126 to zlib levels 2-3. Level 3 is currently only optimized for processors with
129 * New T10dif & copy function - crc16_t10dif_copy()
130 - CRC and copy was added to emulate T10dif operations such as DIF insert and
131 strip. This function stitches together CRC and memcpy operations
132 eliminating an extra data read.
134 * CRC32 iscsi performance improvements
135 - Fixes issue under some distributions where warm cache performance was
141 - Optimized deflate_hash in compression functions.
142 Improves performance of using preset dictionary.
143 - Removed alignment restrictions on input structure.
149 - Add optimized Adler-32 checksum.
151 - Implement zlib compression format.
153 - Add stateful dictionary support.
155 - Add struct reset functions for both deflate and inflate.
157 * Reflected IEEE format CRC32 is released out. Function interface is named
160 * Exact work condition of Erasure Code Reed-Solomon Matrix is determined by new
161 added program gen_rs_matrix_limits.
165 * New 2-pass fully-dynamic deflate compression (level -1). ISA-L fast deflate
166 now has two levels. Level 0 (default) is the same as previous generations.
167 Setting to level 1 will switch to the fully-dynamic compression that will
168 typically reach higher compression ratios.
170 * RAID AVX512 functions.
174 * New fast decompression (inflate)
176 * Compression improvements (deflate)
177 - Speed and compression ratio improvements.
178 - Fast custom Huffman code generation.
180 * Run-time option of gzip crc calculation and headers/trailer.
181 * Choice of static header (BTYPE 01) blocks.
182 * LARGE_WINDOW, 32K history, now default.
183 * Stateless full flush mode.
186 - Six new 64-bit polynomials supported. Normal and reflected versions of ECMA,
187 ISO and Jones polynomials.
191 * Units added: crc, raid, igzip (deflate compression).
195 * Erasure code updates. New AVX512 versions.
197 * Nasm support. ISA-L ported to build with nasm or yasm assembler.
199 * Windows DLL support. Windows builds DLL by default.
203 * Autoconf and autotools build allows easier porting to additional systems.
204 Previous make system still available to embedded users with Makefile.unx.
206 * Includes update for building on Mac OS X/darwin systems. Add --target=darwin
211 * Erasure code improvments
212 - 32-bit port of optimized gf_vect_dot_prod() functions. This makes
213 ec_encode_data() functions much faster on 32-bit processors.
214 - Avoton performance improvements. Performance on Avoton for
215 gf_vect_dot_prod() and ec_encode_data() can improve by as much as 20%.
219 * Incremental erasure code. New functions added to erasure code to handle
220 single source update of code blocks. The function ec_encode_data_update()
221 works with parameters similar to ec_encode_data() but are called incrementally
222 with each source block. These versions are useful when source blocks are not
223 all available at once.
227 * Erasure code updates
228 - New AVX and AVX2 support functions.
229 - Changes min len requirement on gf_vect_dot_prod() to 32 from 16.
230 - Tests include both source and parity recovery with ec_encode_data().
231 - New encoding examples with Vandermonde or Cauchy matrix.
235 * First open release of erasure code unit that is part of ISA-L.