github_mirrors/vectorscan
3
0
Fork 1
mirror of https://github.com/VectorCamp/vectorscan.git synced 2025-11-18 02:00:36 +03:00
Code Issues Packages Projects Releases Wiki Activity
Files
6a2700f1d44a8214ea575469ee8b01d0188ea664
vectorscan/src/nfa
History
Danila Kutenin eb7b0bb50c Optimize vectorscan for aarch64 by using shrn instruction 
This optimization is based on the thread
https://twitter.com/Danlark1/status/1539344279268691970 and uses
shift right and narrow by 4 instruction https://developer.arm.com/documentation/ddi0596/2020-12/SIMD-FP-Instructions/SHRN--SHRN2--Shift-Right-Narrow--immediate--

To achieve that, I needed to redesign a little movemask into comparemask
and have an additional step towards mask iteration. Our benchmarks
showed 10-15% improvement on average for long matches.
2022-06-26 22:55:45 +00:00
..
arm
Fix all ASAN issues in vectorscan
2022-02-18 17:14:51 +00:00
ppc64el
Fix all ASAN issues in vectorscan
2022-02-18 17:14:51 +00:00
x86
Fix all ASAN issues in vectorscan
2022-02-18 17:14:51 +00:00
accel_dfa_build_strat.cpp
Implement new DoubleVermicelli16 acceleration functions using SVE2
2021-10-12 11:51:34 +03:00
accel_dfa_build_strat.h
DFA state compression: 16-bit wide and sherman co-exist
2019-01-21 09:56:37 +08:00
accel_dump.cpp
util: switch from Boost to std::unordered set/map
2017-08-21 11:14:55 +10:00
accel_dump.h
Initial commit of Hyperscan
2015-10-20 09:13:35 +11:00
accel.c
remove vermicelli.h and replace it with vermicelli.hpp
2021-11-02 22:30:53 +02:00
accel.h
Implement new DoubleVermicelli16 acceleration functions using SVE2
2021-10-12 11:51:34 +03:00
accelcompile.cpp
Implement new DoubleVermicelli16 acceleration functions using SVE2
2021-10-12 11:51:34 +03:00
accelcompile.h
util: switch from Boost to std::unordered set/map
2017-08-21 11:14:55 +10:00
callback.h
nfa: standardise callback start, end naming
2016-07-08 11:02:05 +10:00
castle_dump.cpp
castle_dump: use StdioFile
2017-08-21 11:12:16 +10:00
castle_dump.h
have single dump function per engine
2016-12-02 11:32:36 +11:00
castle_internal.h
Implement new Vermicelli16 acceleration functions using SVE2.
2021-10-12 11:51:34 +03:00
castle_sve.h
Move SVE functions into their own files.
2021-10-12 11:51:34 +03:00
castle.c
remove vermicelli.h and replace it with vermicelli.hpp
2021-11-02 22:30:53 +02:00
castle.h
nfa_api: remove subtype from dispatch
2016-12-02 11:32:28 +11:00
castlecompile.cpp
Implement new Vermicelli16 acceleration functions using SVE2.
2021-10-12 11:51:34 +03:00
castlecompile.h
mergeCastle: merge common repeats from the castles
2017-08-21 11:18:54 +10:00
dfa_build_strat.cpp
rdfa: move raw_dfa member functions to rdfa.cpp
2017-08-21 11:12:26 +10:00
dfa_build_strat.h
McClellan refactor
2016-08-10 15:10:25 +10:00
dfa_min.cpp
replace push_back by emplace_back where possible
2021-10-12 11:51:33 +03:00
dfa_min.h
dfa_min: clean up and improve minimize code
2017-05-30 13:57:32 +10:00
gough_internal.h
Initial commit of Hyperscan
2015-10-20 09:13:35 +11:00
gough.c
remove code preventing firing callbacks in the history buffer
2016-10-28 14:46:34 +11:00
gough.h
nfa: unify NfaCallback and SomNfaCallback
2016-07-08 11:01:56 +10:00
goughcompile_accel.cpp
Initial commit of Hyperscan
2015-10-20 09:13:35 +11:00
goughcompile_dump.cpp
replace push_back by emplace_back where possible
2021-10-12 11:51:33 +03:00
goughcompile_dump.h
Initial commit of Hyperscan
2015-10-20 09:13:35 +11:00
goughcompile_internal.h
tidy: "ue2::flat_set/map" -> "flat_set/map"
2017-08-21 11:14:59 +10:00
goughcompile_reg.cpp
replace push_back by emplace_back where possible
2021-10-12 11:51:33 +03:00
goughcompile.cpp
use STL make_unique, remove wrapper header, breaks C++17 compilation
2021-10-12 11:51:34 +03:00
goughcompile.h
util: switch from Boost to std::unordered set/map
2017-08-21 11:14:55 +10:00
goughdump.cpp
goughdump: use StdioFile
2017-08-21 11:12:16 +10:00
goughdump.h
have single dump function per engine
2016-12-02 11:32:36 +11:00
lbr_common_impl.h
nfa: unify NfaCallback and SomNfaCallback
2016-07-08 11:01:56 +10:00
lbr_dump.cpp
lbr_dump: use StdioFile
2017-08-21 11:12:16 +10:00
lbr_dump.h
have single dump function per engine
2016-12-02 11:32:36 +11:00
lbr_internal.h
Implement new Vermicelli16 acceleration functions using SVE2.
2021-10-12 11:51:34 +03:00
lbr_sve.h
Move SVE functions into their own files.
2021-10-12 11:51:34 +03:00
lbr.c
remove vermicelli.h and replace it with vermicelli.hpp
2021-11-02 22:30:53 +02:00
lbr.h
Implement new Vermicelli16 acceleration functions using SVE2.
2021-10-12 11:51:34 +03:00
limex_64.c
Introduce a 64-bit LimEx model.
2016-10-28 14:44:12 +11:00
limex_accel.c
remove vermicelli.h and replace it with vermicelli.hpp
2021-11-02 22:30:53 +02:00
limex_accel.h
Introduce a 64-bit LimEx model.
2016-10-28 14:44:12 +11:00
limex_common_impl.h
limex: make NFAAccept::squash rel to LimEx base
2016-10-28 14:46:23 +11:00
limex_compile.cpp
replace push_back by emplace_back where possible
2021-10-12 11:51:33 +03:00
limex_compile.h
limex: add fast NFA check
2021-01-25 14:13:13 +02:00
limex_context.h
Introduce a 64-bit LimEx model.
2016-10-28 14:44:12 +11:00
limex_dump.cpp
replace push_back by emplace_back where possible
2021-10-12 11:51:33 +03:00
limex_exceptional.h
use correct function names for AVX512, fix build failure
2021-02-15 13:54:19 +02:00
limex_internal.h
Limex: exception handling with AVX512
2021-01-25 14:13:13 +02:00
limex_limits.h
Create combo tops for trigger limexes
2016-12-02 11:22:23 +11:00
limex_native.c
limex: rework accept handling
2016-10-28 14:46:19 +11:00
limex_ring.h
Initial commit of Hyperscan
2015-10-20 09:13:35 +11:00
limex_runtime_impl.h
gcc7: comments for falling through a switch case
2017-05-30 13:59:48 +10:00
limex_runtime.h
limex: rework accept handling
2016-10-28 14:46:19 +11:00
limex_shuffle.h
Shuffle simd and SuperVector implementetions as well as their test realy fixed
2021-10-25 09:19:30 +03:00
limex_shuffle.hpp
Optimize vectorscan for aarch64 by using shrn instruction
2022-06-26 22:55:45 +00:00
limex_simd128.c
Introduce a 64-bit LimEx model.
2016-10-28 14:44:12 +11:00
limex_simd256.c
Introduce a 64-bit LimEx model.
2016-10-28 14:44:12 +11:00
limex_simd384.c
Introduce a 64-bit LimEx model.
2016-10-28 14:44:12 +11:00
limex_simd512.c
Introduce a 64-bit LimEx model.
2016-10-28 14:44:12 +11:00
limex_state_impl.h
Introduce a 64-bit LimEx model.
2016-10-28 14:44:12 +11:00
limex.h
have single dump function per engine
2016-12-02 11:32:36 +11:00
mcclellan_common_impl.h
fix names, use own intrinsic instead of explicit _mm* ones
2020-09-23 11:51:21 +03:00
mcclellan_internal.h
mcclellan: wide state fixes for sanitisers and accept state construction
2019-01-21 09:58:18 +08:00
mcclellan.c
prefetch works best when addresses are 64-byte aligned
2021-10-12 11:50:32 +03:00
mcclellan.h
nfa: unify NfaCallback and SomNfaCallback
2016-07-08 11:01:56 +10:00
mcclellancompile_util.cpp
replace push_back by emplace_back where possible
2021-10-12 11:51:33 +03:00
mcclellancompile_util.h
dfa: don't clear states, just mark dfa dead
2017-08-21 11:10:11 +10:00
mcclellancompile.cpp
use STL make_unique, remove wrapper header, breaks C++17 compilation
2021-10-12 11:51:34 +03:00
mcclellancompile.h
DFA state compression: 16-bit wide and sherman co-exist
2019-01-21 09:56:37 +08:00
mcclellandump.cpp
mcclellan: fix dump issue in wide-state case.
2019-01-21 09:59:29 +08:00
mcclellandump.h
have single dump function per engine
2016-12-02 11:32:36 +11:00
mcsheng_compile.cpp
use STL make_unique, remove wrapper header, breaks C++17 compilation
2021-10-12 11:51:34 +03:00
mcsheng_compile.h
MCSHENG64: extend to 64-state based on mcsheng
2021-01-25 14:13:13 +02:00
mcsheng_data.c
Discard HAVE_AVX512VBMI checks at Sheng/McSheng compile time.
2021-01-25 14:13:13 +02:00
mcsheng_dump.cpp
Discard HAVE_AVX512VBMI checks at Sheng/McSheng compile time.
2021-01-25 14:13:13 +02:00
mcsheng_dump.h
MCSHENG64: extend to 64-state based on mcsheng
2021-01-25 14:13:13 +02:00
mcsheng_internal.h
Discard HAVE_AVX512VBMI checks at Sheng/McSheng compile time.
2021-01-25 14:13:13 +02:00
mcsheng.c
prefetch works best when addresses are 64-byte aligned
2021-10-12 11:50:32 +03:00
mcsheng.h
Discard HAVE_AVX512VBMI checks at Sheng/McSheng compile time.
2021-01-25 14:13:13 +02:00
mpv_dump.cpp
mpv_dump: use StdioFile
2017-08-21 11:12:16 +10:00
mpv_dump.h
have single dump function per engine
2016-12-02 11:32:36 +11:00
mpv_internal.h
Implement new Vermicelli16 acceleration functions using SVE2.
2021-10-12 11:51:34 +03:00
mpv.c
remove vermicelli.h and replace it with vermicelli.hpp
2021-11-02 22:30:53 +02:00
mpv.h
nfa_api: remove subtype from dispatch
2016-12-02 11:32:28 +11:00
mpvcompile.cpp
Implement new Vermicelli16 acceleration functions using SVE2.
2021-10-12 11:51:34 +03:00
mpvcompile.h
mpvcompile: use bytecode_ptr
2017-04-26 15:19:36 +10:00
nfa_api_dispatch.c
Implement new Vermicelli16 acceleration functions using SVE2.
2021-10-12 11:51:34 +03:00
nfa_api_queue.h
nfa: unify NfaCallback and SomNfaCallback
2016-07-08 11:01:56 +10:00
nfa_api_util.h
Initial commit of Hyperscan
2015-10-20 09:13:35 +11:00
nfa_api.h
dfa: adding new Sheng engine
2016-08-10 15:10:46 +10:00
nfa_build_util.cpp
Implement new Vermicelli16 acceleration functions using SVE2.
2021-10-12 11:51:34 +03:00
nfa_build_util.h
limex: add fast NFA check
2021-01-25 14:13:13 +02:00
nfa_dump_api.h
have single dump function per engine
2016-12-02 11:32:36 +11:00
nfa_dump_dispatch.cpp
MCSHENG64: extend to 64-state based on mcsheng
2021-01-25 14:13:13 +02:00
nfa_dump_internal.cpp
Initial commit of Hyperscan
2015-10-20 09:13:35 +11:00
nfa_dump_internal.h
Initial commit of Hyperscan
2015-10-20 09:13:35 +11:00
nfa_internal.h
remove Windows/ICC support
2021-10-12 11:51:34 +03:00
nfa_kind.h
violet: initial implementation
2016-08-10 15:01:08 +10:00
nfa_rev_api.h
remove vermicelli.h and replace it with vermicelli.hpp
2021-11-02 22:30:53 +02:00
rdfa_graph.cpp
replace push_back by emplace_back where possible
2021-10-12 11:51:33 +03:00
rdfa_graph.h
introduce Sheng-McClellan hybrid
2016-12-14 15:27:18 +11:00
rdfa_merge.cpp
use STL make_unique, remove wrapper header, breaks C++17 compilation
2021-10-12 11:51:34 +03:00
rdfa_merge.h
Initial commit of Hyperscan
2015-10-20 09:13:35 +11:00
rdfa.cpp
rdfa: make getImplAlphaSize() inline
2017-09-18 13:25:52 +10:00
rdfa.h
rdfa: make getImplAlphaSize() inline
2017-09-18 13:25:52 +10:00
repeat_internal.h
Introduce REPEAT_ALWAYS model for {0,} castle repeats
2016-03-01 11:10:20 +11:00
repeat.c
Check if any tugs are alive when compressing/expanding repeats
2017-03-01 13:05:20 +11:00
repeat.h
Introduce REPEAT_ALWAYS model for {0,} castle repeats
2016-03-01 11:10:20 +11:00
repeatcompile.cpp
replace push_back by emplace_back where possible
2021-10-12 11:51:33 +03:00
repeatcompile.h
Introduce REPEAT_ALWAYS model for {0,} castle repeats
2016-03-01 11:10:20 +11:00
sheng_defs.h
SHENG64: 64-state 1-byte shuffle based DFA.
2021-01-25 14:13:13 +02:00
sheng_impl4.h
use correct function names for AVX512, fix build failure
2021-02-15 13:54:19 +02:00
sheng_impl.h
use correct function names for AVX512, fix build failure
2021-02-15 13:54:19 +02:00
sheng_internal.h
Discard HAVE_AVX512VBMI checks at Sheng/McSheng compile time.
2021-01-25 14:13:13 +02:00
sheng.c
SHENG64: 64-state 1-byte shuffle based DFA.
2021-01-25 14:13:13 +02:00
sheng.h
SHENG64: 64-state 1-byte shuffle based DFA.
2021-01-25 14:13:13 +02:00
shengcompile.cpp
use STL make_unique, remove wrapper header, breaks C++17 compilation
2021-10-12 11:51:34 +03:00
shengcompile.h
Discard HAVE_AVX512VBMI checks at Sheng/McSheng compile time.
2021-01-25 14:13:13 +02:00
shengdump.cpp
Discard HAVE_AVX512VBMI checks at Sheng/McSheng compile time.
2021-01-25 14:13:13 +02:00
shengdump.h
SHENG64: 64-state 1-byte shuffle based DFA.
2021-01-25 14:13:13 +02:00
shufti_simd.hpp
Fix all ASAN issues in vectorscan
2022-02-18 17:14:51 +00:00
shufti_sve.hpp
Use SVE for double shufti.
2021-10-12 11:51:34 +03:00
shufti.cpp
remove asserts, as they are not needed
2021-10-12 11:51:35 +03:00
shufti.h
refactor shufti algorithm to use SuperVector class, WIP
2021-10-12 11:51:34 +03:00
shufticompile.cpp
replace push_back by emplace_back where possible
2021-10-12 11:51:33 +03:00
shufticompile.h
util: switch from Boost to std::unordered set/map
2017-08-21 11:14:55 +10:00
tamarama_dump.cpp
tamarama_dump: fix typo in comment
2017-08-21 11:12:16 +10:00
tamarama_dump.h
have single dump function per engine
2016-12-02 11:32:36 +11:00
tamarama_internal.h
tamarama: add container engine for exclusive nfas
2016-07-08 11:01:34 +10:00
tamarama.c
nfa_api: remove subtype from dispatch
2016-12-02 11:32:28 +11:00
tamarama.h
nfa_api: remove subtype from dispatch
2016-12-02 11:32:28 +11:00
tamaramacompile.cpp
replace push_back by emplace_back where possible
2021-10-12 11:51:33 +03:00
tamaramacompile.h
nfa: switch to using bytecode_ptr<NFA>
2017-04-26 15:19:36 +10:00
truffle_simd.hpp
Fix all ASAN issues in vectorscan
2022-02-18 17:14:51 +00:00
truffle.cpp
Add missing copyright info from tampered files
2021-10-12 11:51:35 +03:00
truffle.h
Adding AVX2 version of truffle
2016-03-01 11:21:39 +11:00
trufflecompile.cpp
util: switch from Boost to std::unordered set/map
2017-08-21 11:14:55 +10:00
trufflecompile.h
Convert compile-time code to not require SIMD
2016-12-14 15:29:01 +11:00
vermicelli_run.h
remove vermicelli.h and replace it with vermicelli.hpp
2021-11-02 22:30:53 +02:00
vermicelli_simd.cpp
Fix a couple of tests
2022-02-18 19:31:03 +00:00
vermicelli_sve.h
fix SVE2 build after the changes
2021-11-25 18:48:24 +02:00
vermicelli.hpp
fix SVE2 build after the changes
2021-11-25 18:48:24 +02:00
vermicellicompile.cpp
Implement new DoubleVermicelli16 acceleration functions using SVE2
2021-10-12 11:51:34 +03:00
vermicellicompile.h
Implement new DoubleVermicelli16 acceleration functions using SVE2
2021-10-12 11:51:34 +03:00