The existing scalar implementation uses a 4-way pipelined histogram
calculation which is very efficient on out-of-order CPUs. However,
this can be further accelerated using the SVE2 HISTSEG instructions -
which compute a histogram for 16 byte chunks in a vector register.
On a system with 128-bit vectors (VL128) we need 16 HISTSEG executions
to compute the histogram for the whole symbol space (0..255) of 16
bytes input. However we can only accumulate 15 of such 16 byte strips
before possible overflow. So we need to extend and save the 8-bit
histogram accumulators to 16-bit after every 240 byte chunks of input.
To store all in registers we would need 32 128-bit registers. Longer
SVE2 vectors could help here, if such machines become available.
The maximum input block size in Zstd is 128 KiB, so 16-bit accumulators
would not be enough. However an LZ pass will prepend the histogram
calculation, so it is impossible (my assumption) to overflow the 16-bit
accumulators.
The symbol distribution is also not uniform, the lower values are more
common, so we used a 3 pass algorithm to prevent stack spilling. In the
first pass we only compute histograms for 64 symbols (4-way SIMD) while
also computing the maximum symbol value. If we have symbol values
larger than 64 we start the second pass to compute the next 96 elements
of the histogram. The final pass calculates the remaining part of the
histogram (256 symbols in total) if needed. This split of histogram
generation gave the best overall results for performance.
This implementation is the best performing of a number of different
cache blocking schemes tested.
Compression uplifts on a Neoverse V2 system, using Zstd-1.5.8
(e26dde3d) as a baseline, compiled with "-O3 -march=armv8.2-a+sve2":
Clang-20 GCC-14
1#silesia.tar: +6.173% +5.987%
2#silesia.tar: +5.200% +5.011%
3#silesia.tar: +4.332% +5.031%
4#silesia.tar: +2.789% +3.064%
5#silesia.tar: +2.028% +1.838%
6#silesia.tar: +1.562% +1.340%
7#silesia.tar: +1.160% +0.959%
```
for f in $(find . \( -path ./.git -o -path ./tests/fuzz/corpora \) -prune -o -type f);
do
sed -i 's/Facebook, Inc\./Meta Platforms, Inc. and affiliates./' $f;
done
```
* Switch to yearless copyright per FB policy
* Fix up SPDX-License-Identifier lines in `contrib/linux-kernel` sources
* Add zstd copyright/license header to the `contrib/linux-kernel` sources
* Update the `tests/test-license.py` to check for yearless copyright
* Improvements to `tests/test-license.py`
* Check `contrib/linux-kernel` in `tests/test-license.py`
The unused function definitions are hidden behind a
`#ifndef ZSTD_NO_UNUSED_FUNCTIONS` check.
Initially hiding all functions which are unused and take up more than
2KB of stack space, because these will show up as warnings in the
Linux Kernel build system.
new version easier to vectorize
leads to smaller code and faster execution
notably at the last recombination stage
(basically, fixed cost per block).
Assembly inspected with godbolt
On my laptop, with `clang` and `-mavx2` :
2K block : 1280 MB/s -> 1550 MB/s
8K block : 1750 MB/s -> 1860 MB/s
* All copyright lines now have -2020 instead of -present
* All copyright lines include "Facebook, Inc"
* All licenses are now standardized
The copyright in `threading.{h,c}` is not changed because it comes from
zstdmt.
The copyright and license of `divsufsort.{h,c}` is not changed.
changed workspace parameter convention
to always provide workspaceSize,
so that size can be explicitly checked.
Also, use more enum to make the meaning of some parameters more explicit.
Arpad Panyik