AuroraMiddleware/zstd
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
4,084 Commits 1 Branch 0 Tags 27 MiB
232d62b637
Commit Graph

1692 Commits

Author SHA1 Message Date
Yann Collet
232d62b637 fixed a few headers that were too hastily copy/pasted during last license change 2017-08-21 11:24:32 -07:00
Yann Collet
7db552676e reduced pool queue to 0 to save memory 
fixed : pool performance when jobs are fires fast and queueSize==0
 2017-08-19 15:07:54 -07:00
Yann Collet
32fb407c9d updated a bunch of headers 
for the new license
 2017-08-18 16:52:05 -07:00
Yann Collet
f9e6590715 Merge pull request #796 from terrelln/is-error 
[FSE][HUF] Inline error checks
 2017-08-15 12:37:28 -07:00
Yann Collet
2dbcfc6994 Merge pull request #794 from terrelln/force-inline 
[libzstd] Fix FORCE_INLINE macro
 2017-08-15 12:03:44 -07:00
Nick Terrell
07c6ff588e [FSE][HUF] Inline error checks 
Caught by Clang's optimization remarks.
 2017-08-15 11:23:28 -07:00
Nick Terrell
565e925eb7 [libzstd] Fix FORCE_INLINE macro 2017-08-14 21:12:05 -07:00
Roman Gershman
b9d4f4fb74 Fix ZSTD_estimateDStreamSize function after ZSTD_DStream and ZSTD_DCtx were merged 2017-08-13 13:29:42 +03:00
Nick Terrell
9ba97182d1 [CI] Add gcc7build test 2017-08-08 13:28:56 -07:00
Yann Collet
d9f2893eb9 Merge pull request #782 from terrelln/dstSizeTooSmall 
Fix compression failure on incompressible data
 2017-08-07 14:52:02 -07:00
Yann Collet
8049556928 Merge pull request #778 from terrelln/bad-huff 
[libzstd] Fix bug in Huffman decompresser
 2017-08-07 14:05:58 -07:00
Nick Terrell
abe12b3399 [libzstd] Fix bug in Huffman decompresser 
The zstd format specification doesn't enforce that Huffman compressed
literals (including the table) have to be smaller than the uncompressed
literals. The compressor will never Huffman compress literals if the
compressed size is larger than the uncompressed size. The decompresser
doesn't accept Huffman compressed literals with 4 streams whose compressed
size is at least as large as the uncompressed size.

* Make the decompresser accept Huffman compressed literals whose size
  increases.
* Add a test case that exposes the bug. The compressed file has to be
  statically generated, since the compressor won't normally produce files
  that expose the bug.
 2017-08-07 12:37:48 -07:00
Nick Terrell
308047eb5d Fix compression failure on incompressible data 
If the destination buffer is the minimum allowed size in
`ZSTD_compressSequences()` (2^17), then if the block isn't compressible
compression might fail with `dstSize_tooSmall`, when it should instead emit
a raw uncompressed block.

Additionally, `ZSTD_compressLiterals()` implicitly called
`ZSTD_noCompressLiterals()` if Huffman compression failed. Make that
explicit.
 2017-08-07 11:45:24 -07:00
Stella Lau
73ba58955f Signal after finishing job when queueSize=0 2017-08-01 20:12:06 -07:00
Stella Lau
1d76da1d87 Replace marker with queueEmpty variable and update pool.h comment 2017-08-01 12:30:16 -07:00
Stella Lau
5adceeed01 Allow queueSize=0 in pool.c and update poolTests 2017-07-31 10:10:16 -07:00
Yann Collet
e1222544be Merge pull request #753 from paulcruz74/adapt-approach-3 
adaptive compression v1
 2017-07-27 10:00:10 -07:00
Nick Terrell
ae20d413da [libzstd] Fix CHECK_V_F macros 2017-07-25 12:52:01 -07:00
Yann Collet
a90b16e150 Visual blind fix 2 2017-07-20 15:57:55 -07:00
Yann Collet
b4d460f32c pool.c : blindfix for Visual warnings 2017-07-20 01:13:14 -07:00
Yann Collet
3974d2b38a blind fix for Windows Multithreading module 
adds a fake 0 return value for mutex/cond init
 2017-07-19 13:33:21 -07:00
Paul Cruz
6945b3c43d removed previous version of completion for compression 2017-07-19 11:51:50 -07:00
Yann Collet
b71363b967 check pthread_*_init() success condition 2017-07-19 01:05:40 -07:00
Nick Terrell
cc1522351f [libzstd] Fix bug in Huffman encoding 
Summary:
Huffman encoding with a bad dictionary can encode worse than the
HUF_BLOCKBOUND(srcSize), since we don't filter out incompressible
input, and even if we did, the dictionaries Huffman table could be
ill suited to compressing actual data.

The fast optimization doesn't seem to improve compression speed,
even when I hard coded fast = 1, the speed didn't improve over hard coding
it to 0.

Benchmarks:
$ ./zstd.dev -b1e5
Benchmarking levels from 1 to 5
 1#Synthetic 50%     :  10000000 ->   3139163 (3.186), 524.8 MB/s ,1890.0 MB/s
 2#Synthetic 50%     :  10000000 ->   3115138 (3.210), 372.6 MB/s ,1830.2 MB/s
 3#Synthetic 50%     :  10000000 ->   3222672 (3.103), 223.3 MB/s ,1400.2 MB/s
 4#Synthetic 50%     :  10000000 ->   3276678 (3.052), 198.0 MB/s ,1280.1 MB/s
 5#Synthetic 50%     :  10000000 ->   3271570 (3.057), 107.8 MB/s ,1200.0 MB/s
$ ./zstd -b1e5
Benchmarking levels from 1 to 5
 1#Synthetic 50%     :  10000000 ->   3139163 (3.186), 524.8 MB/s ,1870.2 MB/s
 2#Synthetic 50%     :  10000000 ->   3115138 (3.210), 370.0 MB/s ,1810.3 MB/s
 3#Synthetic 50%     :  10000000 ->   3222672 (3.103), 223.3 MB/s ,1380.1 MB/s
 4#Synthetic 50%     :  10000000 ->   3276678 (3.052), 196.1 MB/s ,1270.0 MB/s
 5#Synthetic 50%     :  10000000 ->   3271570 (3.057), 106.8 MB/s ,1180.1 MB/s
$ ./zstd.dev -b1e5 ../silesia.tar
Benchmarking levels from 1 to 5
 1#silesia.tar       : 211988480 ->  73651685 (2.878), 429.7 MB/s ,1096.5 MB/s
 2#silesia.tar       : 211988480 ->  70158785 (3.022), 321.2 MB/s ,1029.1 MB/s
 3#silesia.tar       : 211988480 ->  66993813 (3.164), 243.7 MB/s , 981.4 MB/s
 4#silesia.tar       : 211988480 ->  66306481 (3.197), 226.7 MB/s , 972.4 MB/s
 5#silesia.tar       : 211988480 ->  64757852 (3.274), 150.3 MB/s , 963.6 MB/s
$ ./zstd -b1e5 ../silesia.tar
Benchmarking levels from 1 to 5
 1#silesia.tar       : 211988480 ->  73651685 (2.878), 429.7 MB/s ,1087.1 MB/s
 2#silesia.tar       : 211988480 ->  70158785 (3.022), 318.8 MB/s ,1029.1 MB/s
 3#silesia.tar       : 211988480 ->  66993813 (3.164), 246.5 MB/s , 981.4 MB/s
 4#silesia.tar       : 211988480 ->  66306481 (3.197), 229.2 MB/s , 972.4 MB/s
 5#silesia.tar       : 211988480 ->  64757852 (3.274), 149.3 MB/s , 963.6 MB/s

Test Plan:
I added a test case to the fuzzer which crashed with ASAN before the patch
and succeeded after.
 2017-07-18 13:20:40 -07:00
Yann Collet
77d67fb167 Merge pull request #766 from terrelln/real-block-split 
[libzstd] Pull optimal parser state out of seqStore_t
 2017-07-18 08:26:24 -07:00
Yann Collet
14c83b05c7 Merge pull request #765 from terrelln/real-block-split 
[libzstd] Remove ZSTD_CCtx* argument of ZSTD_compressSequences()
 2017-07-17 19:25:55 -07:00
Nick Terrell
7a28b9e4a3 [libzstd] Pull optimal parser state out of seqStore_t 2017-07-17 15:29:11 -07:00
Yann Collet
3381bf4b84 Merge pull request #764 from terrelln/real-block-split 
[libzstd] Refactor ZSTD_compressSequences()
 2017-07-17 14:46:01 -07:00
Nick Terrell
e198230645 [libzstd] Remove ZSTD_CCtx* argument of ZSTD_compressSequences() 2017-07-17 12:27:24 -07:00
Nick Terrell
634f012420 [libzstd] Refactor ZSTD_compressSequences() 2017-07-17 11:36:11 -07:00
Paul Cruz
50ce4eaeb6 added error detection for pthread initialization, added compression completion measurement, fixed const values 2017-07-17 10:12:44 -07:00
Yann Collet
3b0cff3c33 fixed clang's -Wdocumentation 2017-07-13 18:58:30 -07:00
Yann Collet
2bd6440be0 pinned down error code enum values 
Note : all error codes are changed by this new version,
but it's expected to be the last change for existing codes.

Codes are now grouped by category, and receive a manually attributed value.
The objective is to guarantee that
error code values will not change in the future
when introducing new codes.
Intentionnal empty spaces and ranges are defined
in order to keep room for potential new codes.
 2017-07-13 17:12:16 -07:00
Nick Terrell
830ef4152a [libzstd] Increase granularity of FSECTable repeat mode 2017-07-13 12:45:39 -07:00
Yann Collet
d985319337 Merge pull request #759 from terrelln/real-block-split 
[libzstd] Pull CTables into sub-structure
 2017-07-13 10:24:19 -07:00
Yann Collet
3a60efd3a9 policy change : ZSTDMT automatically caps nbThreads to ZSTDMT_NBTHREADS_MAX (#760) 
Previously, ZSTDMT would refuse to create the compressor.
Also : increased ZSTDMT_NBTHREADS_MAX to 256,
updated doc,
and added relevant test
 2017-07-13 10:17:23 -07:00
Yann Collet
132e6efd76 switched ZSTDMT_compress_advanced() last argument to overlapLog 
overlapRLog (== 9 - overlapLog) was a bit "strange"
as all other public entry points use overlapLog
 2017-07-13 02:22:58 -07:00
Yann Collet
4e77f7761d clarified comment on ZSTD_p_contentSizeFlag 2017-07-13 02:09:07 -07:00
Nick Terrell
de0414b736 [libzstd] Pull CTables into sub-structure 2017-07-12 19:49:19 -07:00
Yann Collet
8ef666c325 slightly increased buffer pool, to cover normal "full load" scenarios 
2 buffers per active worker
+ 1 buffer for input loading
+ 1 buffer for "next input" when submitting current one
+ 1 buffer stuck in queue
 2017-07-12 14:23:34 -07:00
Yann Collet
052a95f77c fix : ZSTDMT_compress_advanced() correctly generates checksum 
when params.fParams.checksumFlag==1.
This use case used to be impossible when only ZSTD_compress() was available
 2017-07-11 17:18:26 -07:00
Yann Collet
2a62f48bf4 release input buffers from inside worker thread 
buffers are released sooner, which makes them available faster for next job.
=> decreases total nb of buffers necessary
 2017-07-11 15:56:40 -07:00
Yann Collet
57236184af buffer pool : all buffers have same size 
to reduce memory fragmentation.
They can be used for in or out, interchangeably.
 2017-07-11 15:17:25 -07:00
Yann Collet
34b2b95631 zstdmt : intermediate outBuffer allocated from within worker 
reduces total amount of memory needed,
since jobs in queue do not have an outBuffer pre-reserved now
 2017-07-11 14:59:10 -07:00
Yann Collet
16261e6951 buffer pool can be invoked from multiple threads 2017-07-11 14:14:07 -07:00
Yann Collet
ef0ff7fe7f zstdmt: removed margin for improved memory usage 2017-07-11 08:54:29 -07:00
Yann Collet
4616fad18b improved ZSTDMT_compress() memory usage 
does not need the input buffer for streaming operations

also : reduced a few tests time length
 2017-07-10 17:16:41 -07:00
Yann Collet
670b1fc547 optimized memory usage for ZSTDMT_compress() 
Previously, each job would reserve a CCtx right before being posted.
The CCtx would be "part of the job description",
and only released when the job is completed (aka flushed).
For ZSTDMT_compress(), which creates all jobs first and only join at the end,
that meant one CCtx per job.
The nb of jobs used to be == nb of threads,
but since latest modification,
which reduces the size of jobs in order to spread the load of difficult areas,
it also increases the nb of jobs for large sources / small compression level.
This resulted in many more CCtx being created.

In this new version, CCtx are reserved within the worker thread.
It guaranteea there cannot be more CCtx reserved than workers (<= nb threads).

To do that, it required to make the CCtx Pool multi-threading-safe :
it can now be called from multiple threads in parallel.
 2017-07-10 16:30:55 -07:00
Yann Collet
3510efb02d fix : custom allocator correctly propagated to child contexts 2017-07-10 14:21:40 -07:00
Yann Collet
88da8f1816 fix : propagate custom allocator to ZSTDMT though ZSTD_CCtx_setParameter() 
also : compile fuzzer with MT enabled
 2017-07-10 14:02:33 -07:00