github_mirrors/vectorscan
3
0
Fork 1
mirror of https://github.com/VectorCamp/vectorscan.git synced 2025-06-28 16:41:01 +03:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'develop' into wip-cppcheck271-part2

Browse Source
This commit is contained in:
g. economou 2024-05-17 11:08:09 +03:00 committed by GitHub
commit 22c3e3da6e
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
211 changed files with 3167 additions and 2125 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
CMakeLists.txt
View File

@ -1221,11 +1221,17 @@ if (NOT BUILD_STATIC_LIBS)
endif ()
add_subdirectory(util)
add_subdirectory(unit)
if (EXISTS ${CMAKE_SOURCE_DIR}/tools/CMakeLists.txt)
option(BUILD_UNIT "Build Hyperscan unit tests (default TRUE)" TRUE)
if(BUILD_UNIT)
add_subdirectory(unit)
endif()
option(BUILD_TOOLS "Build Hyperscan tools (default TRUE)" TRUE)
if(EXISTS ${CMAKE_SOURCE_DIR}/tools/CMakeLists.txt AND BUILD_TOOLS)
add_subdirectory(tools)
endif()
if (EXISTS ${CMAKE_SOURCE_DIR}/chimera/CMakeLists.txt AND BUILD_CHIMERA)
add_subdirectory(chimera)
endif()
@ -1240,4 +1246,7 @@ if(BUILD_BENCHMARKS)
add_subdirectory(benchmarks)
endif()
option(BUILD_DOC "Build the Hyperscan documentation (default TRUE)" TRUE)
if(BUILD_DOC)
add_subdirectory(doc/dev-reference)
endif()

8
README.md
View File

@ -146,6 +146,7 @@ export CXX="/usr/pkg/gcc12/bin/g++"
```
In FreeBSD similarly, you might want to install a different compiler.
If you want to use gcc, it is recommended to use gcc12.
You will also, as in NetBSD, need to install cmake, sqlite, boost and ragel packages.
Using the example of gcc12 from pkg:
installing the desired compiler:
@ -164,7 +165,6 @@ the environment variables to point to this compiler:
export CC="/usr/local/bin/gcc"
export CXX="/usr/local/bin/g++"
```
A further note in FreeBSD, on the PowerPC and ARM platforms,
the gcc12 package installs to a slightly different name, on FreeBSD/ppc,
gcc12 will be found using:
@ -175,12 +175,6 @@ export CXX="/usr/local/bin/g++12"
Then continue with the build as below.
A note about running in FreeBSD: if you built a dynamically linked binary
with an alternative compiler, the libraries specific to the compiler that
built the binary will probably not be found and the base distro libraries
in /lib will be found instead. Adjust LD_LIBRARY_PATH appropriately. For
example, with gcc12 installed from pkg, one would want to use
```export LD_LIBRARY_PATH=/usr/local/lib/gcc12/```
## Configure & build

117
benchmarks/benchmarks.cpp
View File

@ -26,13 +26,13 @@
* POSSIBILITY OF SUCH DAMAGE.
*/
#include <iostream>
#include <chrono>
#include <cstdlib>
#include <cstring>
#include <ctime>
#include <cstdlib>
#include <memory>
#include <functional>
#include <iostream>
#include <memory>
#include "benchmarks.hpp"
@ -43,14 +43,12 @@
struct hlmMatchEntry {
size_t to;
u32 id;
hlmMatchEntry(size_t end, u32 identifier) :
to(end), id(identifier) {}
hlmMatchEntry(size_t end, u32 identifier) : to(end), id(identifier) {}
};
std::vector<hlmMatchEntry> ctxt;
static
hwlmcb_rv_t hlmSimpleCallback(size_t to, u32 id,
static hwlmcb_rv_t hlmSimpleCallback(size_t to, u32 id,
UNUSED struct hs_scratch *scratch) {
DEBUG_PRINTF("match @%zu = %u\n", to, id);
@ -60,15 +58,15 @@ hwlmcb_rv_t hlmSimpleCallback(size_t to, u32 id,
}
template <typename InitFunc, typename BenchFunc>
static void run_benchmarks(int size, int loops, int max_matches, bool is_reverse, MicroBenchmark &bench, InitFunc &&init, BenchFunc &&func) {
static void run_benchmarks(int size, int loops, int max_matches,
bool is_reverse, MicroBenchmark &bench,
InitFunc &&init, BenchFunc &&func) {
init(bench);
double total_sec = 0.0;
u64a total_size = 0;
double bw = 0.0;
double avg_bw = 0.0;
double max_bw = 0.0;
double avg_time = 0.0;
if (max_matches) {
double avg_bw = 0.0;
int pos = 0;
for (int j = 0; j < max_matches - 1; j++) {
bench.buf[pos] = 'b';
@ -84,11 +82,13 @@ static void run_benchmarks(int size, int loops, int max_matches, bool is_reverse
actual_size += res - bench.buf.data();
}
auto end = std::chrono::steady_clock::now();
double dt = std::chrono::duration_cast<std::chrono::microseconds>(end - start).count();
double dt = std::chrono::duration_cast<std::chrono::microseconds>(
end - start)
.count();
total_sec += dt;
/*convert microseconds to seconds*/
/*calculate bandwidth*/
bw = (actual_size / dt) * 1000000.0 / 1048576.0;
double bw = (actual_size / dt) * 1000000.0 / 1048576.0;
/*std::cout << "act_size = " << act_size << std::endl;
std::cout << "dt = " << dt << std::endl;
std::cout << "bw = " << bw << std::endl;*/
@ -102,18 +102,20 @@ static void run_benchmarks(int size, int loops, int max_matches, bool is_reverse
avg_bw /= max_matches;
total_sec /= 1000000.0;
/*convert average time to us*/
printf(KMAG "%s: %u matches, %u * %u iterations," KBLU " total elapsed time =" RST " %.3f s, "
KBLU "average time per call =" RST " %.3f μs," KBLU " max bandwidth = " RST " %.3f MB/s," KBLU " average bandwidth =" RST " %.3f MB/s \n",
printf("%-18s, %-12d, %-10d, %-6d, %-10.3f, %-9.3f, %-8.3f, %-7.3f\n",
bench.label, max_matches, size ,loops, total_sec, avg_time, max_bw, avg_bw);
} else {
u64a total_size = 0;
auto start = std::chrono::steady_clock::now();
for (int i = 0; i < loops; i++) {
const u8 *res = func(bench);
func(bench);
}
auto end = std::chrono::steady_clock::now();
total_sec += std::chrono::duration_cast<std::chrono::microseconds>(end - start).count();
total_sec +=
std::chrono::duration_cast<std::chrono::microseconds>(end - start)
.count();
/*calculate transferred size*/
total_size = size * loops;
total_size = (u64a)size * (u64a)loops;
/*calculate average time*/
avg_time = total_sec / loops;
/*convert microseconds to seconds*/
@ -122,101 +124,109 @@ static void run_benchmarks(int size, int loops, int max_matches, bool is_reverse
max_bw = total_size / total_sec;
/*convert to MB/s*/
max_bw /= 1048576.0;
printf(KMAG "%s: no matches, %u * %u iterations," KBLU " total elapsed time =" RST " %.3f s, "
KBLU "average time per call =" RST " %.3f μs ," KBLU " bandwidth = " RST " %.3f MB/s \n",
bench.label, size ,loops, total_sec, avg_time, max_bw );
printf("%-18s, %-12s, %-10d, %-6d, %-10.3f, %-9.3f, %-8.3f, %-7s\n",
bench.label, "0", size, loops, total_sec, avg_time, max_bw, "0");
}
}
int main(){
int matches[] = {0, MAX_MATCHES};
const int matches[] = {0, MAX_MATCHES};
std::vector<size_t> sizes;
for (size_t i = 0; i < N; i++) sizes.push_back(16000 << i*2);
for (size_t i = 0; i < N; i++)
sizes.push_back(16000 << i * 2);
const char charset[] = "aAaAaAaAAAaaaaAAAAaaaaAAAAAAaaaAAaaa";
printf("%-18s, %-12s, %-10s, %-6s, %-10s, %-9s, %-8s, %-7s\n", "Matcher",
"max_matches", "size", "loops", "total_sec", "avg_time", "max_bw",
"avg_bw");
for (int m = 0; m < 2; m++) {
for (size_t i = 0; i < std::size(sizes); i++) {
MicroBenchmark bench("Shufti", sizes[i]);
run_benchmarks(sizes[i], MAX_LOOPS / sizes[i], matches[m], false, bench,
run_benchmarks(
sizes[i], MAX_LOOPS / sizes[i], matches[m], false, bench,
[&](MicroBenchmark &b) {
b.chars.set('a');
ue2::shuftiBuildMasks(b.chars, (u8 *)&b.lo, (u8 *)&b.hi);
memset(b.buf.data(), 'b', b.size);
},
[&](MicroBenchmark &b) {
return shuftiExec(b.lo, b.hi, b.buf.data(), b.buf.data() + b.size);
}
);
return shuftiExec(b.lo, b.hi, b.buf.data(),
b.buf.data() + b.size);
});
}
for (size_t i = 0; i < std::size(sizes); i++) {
MicroBenchmark bench("Reverse Shufti", sizes[i]);
run_benchmarks(sizes[i], MAX_LOOPS / sizes[i], matches[m], true, bench,
run_benchmarks(
sizes[i], MAX_LOOPS / sizes[i], matches[m], true, bench,
[&](MicroBenchmark &b) {
b.chars.set('a');
ue2::shuftiBuildMasks(b.chars, (u8 *)&b.lo, (u8 *)&b.hi);
memset(b.buf.data(), 'b', b.size);
},
[&](MicroBenchmark &b) {
return rshuftiExec(b.lo, b.hi, b.buf.data(), b.buf.data() + b.size);
}
);
return rshuftiExec(b.lo, b.hi, b.buf.data(),
b.buf.data() + b.size);
});
}
for (size_t i = 0; i < std::size(sizes); i++) {
MicroBenchmark bench("Truffle", sizes[i]);
run_benchmarks(sizes[i], MAX_LOOPS / sizes[i], matches[m], false, bench,
run_benchmarks(
sizes[i], MAX_LOOPS / sizes[i], matches[m], false, bench,
[&](MicroBenchmark &b) {
b.chars.set('a');
ue2::truffleBuildMasks(b.chars, (u8 *)&b.lo, (u8 *)&b.hi);
memset(b.buf.data(), 'b', b.size);
},
[&](MicroBenchmark &b) {
return truffleExec(b.lo, b.hi, b.buf.data(), b.buf.data() + b.size);
}
);
return truffleExec(b.lo, b.hi, b.buf.data(),
b.buf.data() + b.size);
});
}
for (size_t i = 0; i < std::size(sizes); i++) {
MicroBenchmark bench("Reverse Truffle", sizes[i]);
run_benchmarks(sizes[i], MAX_LOOPS / sizes[i], matches[m], true, bench,
run_benchmarks(
sizes[i], MAX_LOOPS / sizes[i], matches[m], true, bench,
[&](MicroBenchmark &b) {
b.chars.set('a');
ue2::truffleBuildMasks(b.chars, (u8 *)&b.lo, (u8 *)&b.hi);
memset(b.buf.data(), 'b', b.size);
},
[&](MicroBenchmark &b) {
return rtruffleExec(b.lo, b.hi, b.buf.data(), b.buf.data() + b.size);
}
);
return rtruffleExec(b.lo, b.hi, b.buf.data(),
b.buf.data() + b.size);
});
}
for (size_t i = 0; i < std::size(sizes); i++) {
MicroBenchmark bench("Vermicelli", sizes[i]);
run_benchmarks(sizes[i], MAX_LOOPS / sizes[i], matches[m], false, bench,
run_benchmarks(
sizes[i], MAX_LOOPS / sizes[i], matches[m], false, bench,
[&](MicroBenchmark &b) {
b.chars.set('a');
ue2::truffleBuildMasks(b.chars, (u8 *)&b.lo, (u8 *)&b.hi);
memset(b.buf.data(), 'b', b.size);
},
[&](MicroBenchmark &b) {
return vermicelliExec('a', 'b', b.buf.data(), b.buf.data() + b.size);
}
);
return vermicelliExec('a', 'b', b.buf.data(),
b.buf.data() + b.size);
});
}
for (size_t i = 0; i < std::size(sizes); i++) {
MicroBenchmark bench("Reverse Vermicelli", sizes[i]);
run_benchmarks(sizes[i], MAX_LOOPS / sizes[i], matches[m], true, bench,
run_benchmarks(
sizes[i], MAX_LOOPS / sizes[i], matches[m], true, bench,
[&](MicroBenchmark &b) {
b.chars.set('a');
ue2::truffleBuildMasks(b.chars, (u8 *)&b.lo, (u8 *)&b.hi);
memset(b.buf.data(), 'b', b.size);
},
[&](MicroBenchmark &b) {
return rvermicelliExec('a', 'b', b.buf.data(), b.buf.data() + b.size);
}
);
return rvermicelliExec('a', 'b', b.buf.data(),
b.buf.data() + b.size);
});
}
for (size_t i = 0; i < std::size(sizes); i++) {
@ -232,20 +242,21 @@ int main(){
}
MicroBenchmark bench("Noodle", sizes[i]);
run_benchmarks(sizes[i], MAX_LOOPS / sizes[i], matches[m], false, bench,
run_benchmarks(
sizes[i], MAX_LOOPS / sizes[i], matches[m], false, bench,
[&](MicroBenchmark &b) {
ctxt.clear();
memset(b.buf.data(), 'a', b.size);
u32 id = 1000;
ue2::hwlmLiteral lit(str, true, id);
b.nt = ue2::noodBuildTable(lit);
assert(b.nt != nullptr);
assert(b.nt.get() != nullptr);
},
[&](MicroBenchmark &b) {
noodExec(b.nt.get(), b.buf.data(), b.size, 0, hlmSimpleCallback, &b.scratch);
noodExec(b.nt.get(), b.buf.data(), b.size, 0,
hlmSimpleCallback, &b.scratch);
return b.buf.data() + b.size;
}
);
});
}
}

26
benchmarks/benchmarks.hpp
View File

@ -26,30 +26,19 @@
* POSSIBILITY OF SUCH DAMAGE.
*/
#include "hwlm/hwlm_literal.h"
#include "hwlm/noodle_build.h"
#include "hwlm/noodle_engine.h"
#include "hwlm/noodle_internal.h"
#include "nfa/shufti.h"
#include "nfa/shufticompile.h"
#include "nfa/truffle.h"
#include "nfa/trufflecompile.h"
#include "nfa/vermicelli.hpp"
#include "hwlm/noodle_build.h"
#include "hwlm/noodle_engine.h"
#include "hwlm/noodle_internal.h"
#include "hwlm/hwlm_literal.h"
#include "util/bytecode_ptr.h"
#include "scratch.h"
#include "util/bytecode_ptr.h"
/*define colour control characters*/
#define RST "\x1B[0m"
#define KRED "\x1B[31m"
#define KGRN "\x1B[32m"
#define KYEL "\x1B[33m"
#define KBLU "\x1B[34m"
#define KMAG "\x1B[35m"
#define KCYN "\x1B[36m"
#define KWHT "\x1B[37m"
class MicroBenchmark
{
class MicroBenchmark {
public:
char const *label;
size_t size;
@ -64,6 +53,5 @@ public:
ue2::bytecode_ptr<noodTable> nt;
MicroBenchmark(char const *label_, size_t size_)
:label(label_), size(size_), buf(size_) {
};
: label(label_), size(size_), buf(size_){};
};

6
cmake/osdetection.cmake
View File

@ -6,10 +6,10 @@ if(CMAKE_SYSTEM_NAME MATCHES "FreeBSD")
set(FREEBSD true)
set(CMAKE_INSTALL_RPATH_USE_LINK_PATH TRUE)
#FIXME: find a nicer and more general way of doing this
if(CMAKE_C_COMPILER MATCHES "/usr/local/bin/gcc12")
set(CMAKE_BUILD_RPATH "/usr/local/lib/gcc12")
elseif(CMAKE_C_COMPILER MATCHES "/usr/local/bin/gcc13")
if(CMAKE_C_COMPILER MATCHES "/usr/local/bin/gcc13")
set(CMAKE_BUILD_RPATH "/usr/local/lib/gcc13")
elseif(ARCH_AARCH64 AND (CMAKE_C_COMPILER MATCHES "/usr/local/bin/gcc12"))
set(CMAKE_BUILD_RPATH "/usr/local/lib/gcc12")
endif()
endif(CMAKE_SYSTEM_NAME MATCHES "FreeBSD")

11
doc/dev-reference/CMakeLists.txt
View File

@ -19,6 +19,7 @@ else()
set(SPHINX_BUILD_DIR "${CMAKE_CURRENT_BINARY_DIR}/_build")
set(SPHINX_CACHE_DIR "${CMAKE_CURRENT_BINARY_DIR}/_doctrees")
set(SPHINX_HTML_DIR "${CMAKE_CURRENT_BINARY_DIR}/html")
set(SPHINX_MAN_DIR "${CMAKE_CURRENT_BINARY_DIR}/man")
configure_file("${CMAKE_CURRENT_SOURCE_DIR}/conf.py.in"
"${CMAKE_CURRENT_BINARY_DIR}/conf.py" @ONLY)
@ -32,4 +33,14 @@ add_custom_target(dev-reference
"${SPHINX_HTML_DIR}"
DEPENDS dev-reference-doxygen
COMMENT "Building HTML dev reference with Sphinx")
add_custom_target(dev-reference-man
${SPHINX_BUILD}
-b man
-c "${CMAKE_CURRENT_BINARY_DIR}"
-d "${SPHINX_CACHE_DIR}"
"${CMAKE_CURRENT_SOURCE_DIR}"
"${SPHINX_MAN_DIR}"
DEPENDS dev-reference-doxygen
COMMENT "Building man page reference with Sphinx")
endif()

22
doc/dev-reference/chimera.rst
View File

@ -11,10 +11,10 @@ Introduction
************
Chimera is a software regular expression matching engine that is a hybrid of
Hyperscan and PCRE. The design goals of Chimera are to fully support PCRE
syntax as well as to take advantage of the high performance nature of Hyperscan.
Vectorscan and PCRE. The design goals of Chimera are to fully support PCRE
syntax as well as to take advantage of the high performance nature of Vectorscan.
Chimera inherits the design guideline of Hyperscan with C APIs for compilation
Chimera inherits the design guideline of Vectorscan with C APIs for compilation
and scanning.
The Chimera API itself is composed of two major components:
@ -65,13 +65,13 @@ For a given database, Chimera provides several guarantees:
.. note:: Chimera is designed to have the same matching behavior as PCRE,
including greedy/ungreedy, capturing, etc. Chimera reports both
**start offset** and **end offset** for each match like PCRE. Different
from the fashion of reporting all matches in Hyperscan, Chimera only reports
from the fashion of reporting all matches in Vectorscan, Chimera only reports
non-overlapping matches. For example, the pattern :regexp:`/foofoo/` will
match ``foofoofoofoo`` at offsets (0, 6) and (6, 12).
.. note:: Since Chimera is a hybrid of Hyperscan and PCRE in order to support
.. note:: Since Chimera is a hybrid of Vectorscan and PCRE in order to support
full PCRE syntax, there will be extra performance overhead compared to
Hyperscan-only solution. Please always use Hyperscan for better performance
Vectorscan-only solution. Please always use Vectorscan for better performance
unless you must need full PCRE syntax support.
See :ref:`chruntime` for more details
@ -83,12 +83,12 @@ Requirements
The PCRE library (http://pcre.org/) version 8.41 is required for Chimera.
.. note:: Since Chimera needs to reference PCRE internal function, please place PCRE source
directory under Hyperscan root directory in order to build Chimera.
directory under Vectorscan root directory in order to build Chimera.
Beside this, both hardware and software requirements of Chimera are the same to Hyperscan.
Beside this, both hardware and software requirements of Chimera are the same to Vectorscan.
See :ref:`hardware` and :ref:`software` for more details.
.. note:: Building Hyperscan will automatically generate Chimera library.
.. note:: Building Vectorscan will automatically generate Chimera library.
Currently only static library is supported for Chimera, so please
use static build type when configure CMake build options.
@ -119,7 +119,7 @@ databases:
Compilation allows the Chimera library to analyze the given pattern(s) and
pre-determine how to scan for these patterns in an optimized fashion using
Hyperscan and PCRE.
Vectorscan and PCRE.
===============
Pattern Support
@ -134,7 +134,7 @@ Semantics
=========
Chimera supports the exact same semantics of PCRE library. Moreover, it supports
multiple simultaneous pattern matching like Hyperscan and the multiple matches
multiple simultaneous pattern matching like Vectorscan and the multiple matches
will be reported in order by end offset.
.. _chruntime:

92
doc/dev-reference/compilation.rst
View File

@ -9,7 +9,7 @@ Compiling Patterns
Building a Database
*******************
The Hyperscan compiler API accepts regular expressions and converts them into a
The Vectorscan compiler API accepts regular expressions and converts them into a
compiled pattern database that can then be used to scan data.
The API provides three functions that compile regular expressions into
@ -24,7 +24,7 @@ databases:
#. :c:func:`hs_compile_ext_multi`: compiles an array of expressions as above,
but allows :ref:`extparam` to be specified for each expression.
Compilation allows the Hyperscan library to analyze the given pattern(s) and
Compilation allows the Vectorscan library to analyze the given pattern(s) and
pre-determine how to scan for these patterns in an optimized fashion that would
be far too expensive to compute at run-time.
@ -48,10 +48,10 @@ To compile patterns to be used in streaming mode, the ``mode`` parameter of
block mode requires the use of :c:member:`HS_MODE_BLOCK` and vectored mode
requires the use of :c:member:`HS_MODE_VECTORED`. A pattern database compiled
for one mode (streaming, block or vectored) can only be used in that mode. The
version of Hyperscan used to produce a compiled pattern database must match the
version of Hyperscan used to scan with it.
version of Vectorscan used to produce a compiled pattern database must match the
version of Vectorscan used to scan with it.
Hyperscan provides support for targeting a database at a particular CPU
Vectorscan provides support for targeting a database at a particular CPU
platform; see :ref:`instr_specialization` for details.
=====================
@ -75,14 +75,14 @@ characters exist in regular grammar like ``[``, ``]``, ``(``, ``)``, ``{``,
While in pure literal case, all these meta characters lost extra meanings
expect for that they are just common ASCII codes.
Hyperscan is initially designed to process common regular expressions. It is
Vectorscan is initially designed to process common regular expressions. It is
hence embedded with a complex parser to do comprehensive regular grammar
interpretation. Particularly, the identification of above meta characters is the
basic step for the interpretation of far more complex regular grammars.
However in real cases, patterns may not always be regular expressions. They
could just be pure literals. Problem will come if the pure literals contain
regular meta characters. Supposing fed directly into traditional Hyperscan
regular meta characters. Supposing fed directly into traditional Vectorscan
compile API, all these meta characters will be interpreted in predefined ways,
which is unnecessary and the result is totally out of expectation. To avoid
such misunderstanding by traditional API, users have to preprocess these
@ -90,7 +90,7 @@ literal patterns by converting the meta characters into some other formats:
either by adding a backslash ``\`` before certain meta characters, or by
converting all the characters into a hexadecimal representation.
In ``v5.2.0``, Hyperscan introduces 2 new compile APIs for pure literal patterns:
In ``v5.2.0``, Vectorscan introduces 2 new compile APIs for pure literal patterns:
#. :c:func:`hs_compile_lit`: compiles a single pure literal into a pattern
database.
@ -106,7 +106,7 @@ content directly into these APIs without worrying about writing regular meta
characters in their patterns. No preprocessing work is needed any more.
For new APIs, the ``length`` of each literal pattern is a newly added parameter.
Hyperscan needs to locate the end position of the input expression via clearly
Vectorscan needs to locate the end position of the input expression via clearly
knowing each literal's length, not by simply identifying character ``\0`` of a
string.
@ -127,19 +127,19 @@ Supported flags: :c:member:`HS_FLAG_CASELESS`, :c:member:`HS_FLAG_SINGLEMATCH`,
Pattern Support
***************
Hyperscan supports the pattern syntax used by the PCRE library ("libpcre"),
Vectorscan supports the pattern syntax used by the PCRE library ("libpcre"),
described at <http://www.pcre.org/>. However, not all constructs available in
libpcre are supported. The use of unsupported constructs will result in
compilation errors.
The version of PCRE used to validate Hyperscan's interpretation of this syntax
The version of PCRE used to validate Vectorscan's interpretation of this syntax
is 8.41 or above.
====================
Supported Constructs
====================
The following regex constructs are supported by Hyperscan:
The following regex constructs are supported by Vectorscan:
* Literal characters and strings, with all libpcre quoting and character
escapes.
@ -177,7 +177,7 @@ The following regex constructs are supported by Hyperscan:
:c:member:`HS_FLAG_SINGLEMATCH` flag is on for that pattern.
* Lazy modifiers (:regexp:`?` appended to another quantifier, e.g.
:regexp:`\\w+?`) are supported but ignored (as Hyperscan reports all
:regexp:`\\w+?`) are supported but ignored (as Vectorscan reports all
matches).
* Parenthesization, including the named and unnamed capturing and
@ -219,15 +219,15 @@ The following regex constructs are supported by Hyperscan:
.. note:: At this time, not all patterns can be successfully compiled with the
:c:member:`HS_FLAG_SOM_LEFTMOST` flag, which enables per-pattern support for
:ref:`som`. The patterns that support this flag are a subset of patterns that
can be successfully compiled with Hyperscan; notably, many bounded repeat
forms that can be compiled with Hyperscan without the Start of Match flag
can be successfully compiled with Vectorscan; notably, many bounded repeat
forms that can be compiled with Vectorscan without the Start of Match flag
enabled cannot be compiled with the flag enabled.
======================
Unsupported Constructs
======================
The following regex constructs are not supported by Hyperscan:
The following regex constructs are not supported by Vectorscan:
* Backreferences and capturing sub-expressions.
* Arbitrary zero-width assertions.
@ -246,32 +246,32 @@ The following regex constructs are not supported by Hyperscan:
Semantics
*********
While Hyperscan follows libpcre syntax, it provides different semantics. The
While Vectorscan follows libpcre syntax, it provides different semantics. The
major departures from libpcre semantics are motivated by the requirements of
streaming and multiple simultaneous pattern matching.
The major departures from libpcre semantics are:
#. **Multiple pattern matching**: Hyperscan allows matches to be reported for
#. **Multiple pattern matching**: Vectorscan allows matches to be reported for
several patterns simultaneously. This is not equivalent to separating the
patterns by :regexp:`|` in libpcre, which evaluates alternations
left-to-right.
#. **Lack of ordering**: the multiple matches that Hyperscan produces are not
#. **Lack of ordering**: the multiple matches that Vectorscan produces are not
guaranteed to be ordered, although they will always fall within the bounds of
the current scan.
#. **End offsets only**: Hyperscan's default behaviour is only to report the end
#. **End offsets only**: Vectorscan's default behaviour is only to report the end
offset of a match. Reporting of the start offset can be enabled with
per-expression flags at pattern compile time. See :ref:`som` for details.
#. **"All matches" reported**: scanning :regexp:`/foo.*bar/` against
``fooxyzbarbar`` will return two matches from Hyperscan -- at the points
``fooxyzbarbar`` will return two matches from Vectorscan -- at the points
corresponding to the ends of ``fooxyzbar`` and ``fooxyzbarbar``. In contrast,
libpcre semantics by default would report only one match at ``fooxyzbarbar``
(greedy semantics) or, if non-greedy semantics were switched on, one match at
``fooxyzbar``. This means that switching between greedy and non-greedy
semantics is a no-op in Hyperscan.
semantics is a no-op in Vectorscan.
To support libpcre quantifier semantics while accurately reporting streaming
matches at the time they occur is impossible. For example, consider the pattern
@ -299,7 +299,7 @@ as in block 3 -- which would constitute a better match for the pattern.
Start of Match
==============
In standard operation, Hyperscan will only provide the end offset of a match
In standard operation, Vectorscan will only provide the end offset of a match
when the match callback is called. If the :c:member:`HS_FLAG_SOM_LEFTMOST` flag
is specified for a particular pattern, then the same set of matches is
returned, but each match will also provide the leftmost possible start offset
@ -308,7 +308,7 @@ corresponding to its end offset.
Using the SOM flag entails a number of trade-offs and limitations:
* Reduced pattern support: For many patterns, tracking SOM is complex and can
result in Hyperscan failing to compile a pattern with a "Pattern too
result in Vectorscan failing to compile a pattern with a "Pattern too
large" error, even if the pattern is supported in normal operation.
* Increased stream state: At scan time, state space is required to track
potential SOM offsets, and this must be stored in persistent stream state in
@ -316,20 +316,20 @@ Using the SOM flag entails a number of trade-offs and limitations:
required to match a pattern.
* Performance overhead: Similarly, there is generally a performance cost
associated with tracking SOM.
* Incompatible features: Some other Hyperscan pattern flags (such as
* Incompatible features: Some other Vectorscan pattern flags (such as
:c:member:`HS_FLAG_SINGLEMATCH` and :c:member:`HS_FLAG_PREFILTER`) can not be
used in combination with SOM. Specifying them together with
:c:member:`HS_FLAG_SOM_LEFTMOST` will result in a compilation error.
In streaming mode, the amount of precision delivered by SOM can be controlled
with the SOM horizon flags. These instruct Hyperscan to deliver accurate SOM
with the SOM horizon flags. These instruct Vectorscan to deliver accurate SOM
information within a certain distance of the end offset, and return a special
start offset of :c:member:`HS_OFFSET_PAST_HORIZON` otherwise. Specifying a
small or medium SOM horizon will usually reduce the stream state required for a
given database.
.. note:: In streaming mode, the start offset returned for a match may refer to
a point in the stream *before* the current block being scanned. Hyperscan
a point in the stream *before* the current block being scanned. Vectorscan
provides no facility for accessing earlier blocks; if the calling application
needs to inspect historical data, then it must store it itself.
@ -341,7 +341,7 @@ Extended Parameters
In some circumstances, more control over the matching behaviour of a pattern is
required than can be specified easily using regular expression syntax. For
these scenarios, Hyperscan provides the :c:func:`hs_compile_ext_multi` function
these scenarios, Vectorscan provides the :c:func:`hs_compile_ext_multi` function
that allows a set of "extended parameters" to be set on a per-pattern basis.
Extended parameters are specified using an :c:type:`hs_expr_ext_t` structure,
@ -383,18 +383,18 @@ section.
Prefiltering Mode
=================
Hyperscan provides a per-pattern flag, :c:member:`HS_FLAG_PREFILTER`, which can
be used to implement a prefilter for a pattern than Hyperscan would not
Vectorscan provides a per-pattern flag, :c:member:`HS_FLAG_PREFILTER`, which can
be used to implement a prefilter for a pattern than Vectorscan would not
ordinarily support.
This flag instructs Hyperscan to compile an "approximate" version of this
pattern for use in a prefiltering application, even if Hyperscan does not
This flag instructs Vectorscan to compile an "approximate" version of this
pattern for use in a prefiltering application, even if Vectorscan does not
support the pattern in normal operation.
The set of matches returned when this flag is used is guaranteed to be a
superset of the matches specified by the non-prefiltering expression.
If the pattern contains pattern constructs not supported by Hyperscan (such as
If the pattern contains pattern constructs not supported by Vectorscan (such as
zero-width assertions, back-references or conditional references) these
constructs will be replaced internally with broader constructs that may match
more often.
@ -404,7 +404,7 @@ back-reference :regexp:`\\1`. In prefiltering mode, this pattern might be
approximated by having its back-reference replaced with its referent, forming
:regexp:`/\\w+ again \\w+/`.
Furthermore, in prefiltering mode Hyperscan may simplify a pattern that would
Furthermore, in prefiltering mode Vectorscan may simplify a pattern that would
otherwise return a "Pattern too large" error at compile time, or for performance
reasons (subject to the matching guarantee above).
@ -422,22 +422,22 @@ matches for the pattern.
Instruction Set Specialization
******************************
Hyperscan is able to make use of several modern instruction set features found
Vectorscan is able to make use of several modern instruction set features found
on x86 processors to provide improvements in scanning performance.
Some of these features are selected when the library is built; for example,
Hyperscan will use the native ``POPCNT`` instruction on processors where it is
Vectorscan will use the native ``POPCNT`` instruction on processors where it is
available and the library has been optimized for the host architecture.
.. note:: By default, the Hyperscan runtime is built with the ``-march=native``
.. note:: By default, the Vectorscan runtime is built with the ``-march=native``
compiler flag and (where possible) will make use of all instructions known by
the host's C compiler.
To use some instruction set features, however, Hyperscan must build a
To use some instruction set features, however, Vectorscan must build a
specialized database to support them. This means that the target platform must
be specified at pattern compile time.
The Hyperscan compiler API functions all accept an optional
The Vectorscan compiler API functions all accept an optional
:c:type:`hs_platform_info_t` argument, which describes the target platform
for the database to be built. If this argument is NULL, the database will be
targeted at the current host platform.
@ -467,7 +467,7 @@ See :ref:`api_constants` for the full list of CPU tuning and feature flags.
Approximate matching
********************
Hyperscan provides an experimental approximate matching mode, which will match
Vectorscan provides an experimental approximate matching mode, which will match
patterns within a given edit distance. The exact matching behavior is defined as
follows:
@ -492,7 +492,7 @@ follows:
Here are a few examples of approximate matching:
* Pattern :regexp:`/foo/` can match ``foo`` when using regular Hyperscan
* Pattern :regexp:`/foo/` can match ``foo`` when using regular Vectorscan
matching behavior. With approximate matching within edit distance 2, the
pattern will produce matches when scanned against ``foo``, ``foooo``, ``f00``,
``f``, and anything else that lies within edit distance 2 of matching corpora
@ -513,7 +513,7 @@ matching support. Here they are, in a nutshell:
* Reduced pattern support:
* For many patterns, approximate matching is complex and can result in
Hyperscan failing to compile a pattern with a "Pattern too large" error,
Vectorscan failing to compile a pattern with a "Pattern too large" error,
even if the pattern is supported in normal operation.
* Additionally, some patterns cannot be approximately matched because they
reduce to so-called "vacuous" patterns (patterns that match everything). For
@ -548,7 +548,7 @@ Logical Combinations
********************
For situations when a user requires behaviour that depends on the presence or
absence of matches from groups of patterns, Hyperscan provides support for the
absence of matches from groups of patterns, Vectorscan provides support for the
logical combination of patterns in a given pattern set, with three operators:
``NOT``, ``AND`` and ``OR``.
@ -561,7 +561,7 @@ offset is *true* if the expression it refers to is *false* at this offset.
For example, ``NOT 101`` means that expression 101 has not yet matched at this
offset.
A logical combination is passed to Hyperscan at compile time as an expression.
A logical combination is passed to Vectorscan at compile time as an expression.
This combination expression will raise matches at every offset where one of its
sub-expressions matches and the logical value of the whole expression is *true*.
@ -603,7 +603,7 @@ In a logical combination expression:
* Whitespace is ignored.
To use a logical combination expression, it must be passed to one of the
Hyperscan compile functions (:c:func:`hs_compile_multi`,
Vectorscan compile functions (:c:func:`hs_compile_multi`,
:c:func:`hs_compile_ext_multi`) along with the :c:member:`HS_FLAG_COMBINATION` flag,
which identifies the pattern as a logical combination expression. The patterns
referred to in the logical combination expression must be compiled together in
@ -613,7 +613,7 @@ When an expression has the :c:member:`HS_FLAG_COMBINATION` flag set, it ignores
all other flags except the :c:member:`HS_FLAG_SINGLEMATCH` flag and the
:c:member:`HS_FLAG_QUIET` flag.
Hyperscan will accept logical combination expressions at compile time that
Vectorscan will accept logical combination expressions at compile time that
evaluate to *true* when no patterns have matched, and report the match for
combination at end of data if no patterns have matched; for example: ::

16
doc/dev-reference/conf.py.in
View File

@ -1,6 +1,6 @@
# -*- coding: utf-8 -*-
#
# Hyperscan documentation build configuration file, created by
# Vectorscan documentation build configuration file, created by
# sphinx-quickstart on Tue Sep 29 15:59:19 2015.
#
# This file is execfile()d with the current directory set to its
@ -43,8 +43,8 @@ source_suffix = '.rst'
master_doc = 'index'
# General information about the project.
project = u'Hyperscan'
copyright = u'2015-2018, Intel Corporation'
project = u'Vectorscan'
copyright = u'2015-2020, Intel Corporation; 2020-2024, VectorCamp; and other contributors'
# The version info for the project you're documenting, acts as replacement for
# |version| and |release|, also used in various other places throughout the
@ -202,7 +202,7 @@ latex_elements = {
# (source start file, target name, title,
# author, documentclass [howto, manual, or own class]).
latex_documents = [
('index', 'Hyperscan.tex', u'Hyperscan Documentation',
('index', 'Hyperscan.tex', u'Vectorscan Documentation',
u'Intel Corporation', 'manual'),
]
@ -232,8 +232,8 @@ latex_documents = [
# One entry per manual page. List of tuples
# (source start file, name, description, authors, manual section).
man_pages = [
('index', 'hyperscan', u'Hyperscan Documentation',
[u'Intel Corporation'], 1)
('index', 'vectorscan', u'Vectorscan Documentation',
[u'Intel Corporation'], 7)
]
# If true, show URL addresses after external links.
@ -246,8 +246,8 @@ man_pages = [
# (source start file, target name, title, author,
# dir menu entry, description, category)
texinfo_documents = [
('index', 'Hyperscan', u'Hyperscan Documentation',
u'Intel Corporation', 'Hyperscan', 'High-performance regular expression matcher.',
('index', 'Vectorscan', u'Vectorscan Documentation',
u'Intel Corporation; VectorCamp', 'Vectorscan', 'High-performance regular expression matcher.',
'Miscellaneous'),
]

159
doc/dev-reference/getting_started.rst
View File

@ -7,43 +7,41 @@ Getting Started
Very Quick Start
****************
#. Clone Hyperscan ::
#. Clone Vectorscan ::
cd <where-you-want-hyperscan-source>
git clone git://github.com/intel/hyperscan
cd <where-you-want-vectorscan-source>
git clone https://github.com/VectorCamp/vectorscan
#. Configure Hyperscan
#. Configure Vectorscan
Ensure that you have the correct :ref:`dependencies <software>` present,
and then:
::
cd <where-you-want-to-build-hyperscan>
cd <where-you-want-to-build-vectorscan>
mkdir <build-dir>
cd <build-dir>
cmake [-G <generator>] [options] <hyperscan-source-path>
cmake [-G <generator>] [options] <vectorscan-source-path>
Known working generators:
* ``Unix Makefiles`` --- make-compatible makefiles (default on Linux/FreeBSD/Mac OS X)
* ``Ninja`` --- `Ninja <http://martine.github.io/ninja/>`_ build files.
* ``Visual Studio 15 2017`` --- Visual Studio projects
Generators that might work include:
Unsupported generators that might work include:
* ``Xcode`` --- OS X Xcode projects.
#. Build Hyperscan
#. Build Vectorscan
Depending on the generator used:
* ``cmake --build .`` --- will build everything
* ``make -j<jobs>`` --- use makefiles in parallel
* ``ninja`` --- use Ninja build
* ``MsBuild.exe`` --- use Visual Studio MsBuild
* etc.
#. Check Hyperscan
#. Check Vectorscan
Run the Hyperscan unit tests: ::
Run the Vectorscan unit tests: ::
bin/unit-hyperscan
@ -55,20 +53,23 @@ Requirements
Hardware
========
Hyperscan will run on x86 processors in 64-bit (Intel\ |reg| 64 Architecture) and
32-bit (IA-32 Architecture) modes.
Vectorscan will run on x86 processors in 64-bit (Intel\ |reg| 64 Architecture) and
32-bit (IA-32 Architecture) modes as well as Arm v8.0+ aarch64, and POWER 8+ ppc64le
machines.
Hyperscan is a high performance software library that takes advantage of recent
Intel architecture advances. At a minimum, support for Supplemental Streaming
SIMD Extensions 3 (SSSE3) is required, which should be available on any modern
x86 processor.
architecture advances.
Additionally, Hyperscan can make use of:
Additionally, Vectorscan can make use of:
* Intel Streaming SIMD Extensions 4.2 (SSE4.2)
* the POPCNT instruction
* Bit Manipulation Instructions (BMI, BMI2)
* Intel Advanced Vector Extensions 2 (Intel AVX2)
* Arm NEON
* Arm SVE and SVE2
* Arm SVE2 BITPERM
* IBM Power8/Power9 VSX
if present.
@ -79,40 +80,34 @@ These can be determined at library compile time, see :ref:`target_arch`.
Software
========
As a software library, Hyperscan doesn't impose any particular runtime
software requirements, however to build the Hyperscan library we require a
modern C and C++ compiler -- in particular, Hyperscan requires C99 and C++11
As a software library, Vectorscan doesn't impose any particular runtime
software requirements, however to build the Vectorscan library we require a
modern C and C++ compiler -- in particular, Vectorscan requires C99 and C++17
compiler support. The supported compilers are:
* GCC, v4.8.1 or higher
* Clang, v3.4 or higher (with libstdc++ or libc++)
* Intel C++ Compiler v15 or higher
* Visual C++ 2017 Build Tools
* GCC, v9 or higher
* Clang, v5 or higher (with libstdc++ or libc++)
Examples of operating systems that Hyperscan is known to work on include:
Examples of operating systems that Vectorscan is known to work on include:
Linux:
* Ubuntu 14.04 LTS or newer
* Ubuntu 20.04 LTS or newer
* RedHat/CentOS 7 or newer
* Fedora 38 or newer
* Debian 10
FreeBSD:
* 10.0 or newer
Windows:
* 8 or newer
Mac OS X:
* 10.8 or newer, using XCode/Clang
Hyperscan *may* compile and run on other platforms, but there is no guarantee.
We currently have experimental support for Windows using Intel C++ Compiler
or Visual Studio 2017.
Vectorscan *may* compile and run on other platforms, but there is no guarantee.
In addition, the following software is required for compiling the Hyperscan library:
In addition, the following software is required for compiling the Vectorscan library:
======================================================= =========== ======================================
Dependency Version Notes
@ -132,20 +127,20 @@ Ragel, you may use Cygwin to build it from source.
Boost Headers
-------------
Compiling Hyperscan depends on a recent version of the Boost C++ header
Compiling Vectorscan depends on a recent version of the Boost C++ header
library. If the Boost libraries are installed on the build machine in the
usual paths, CMake will find them. If the Boost libraries are not installed,
the location of the Boost source tree can be specified during the CMake
configuration step using the ``BOOST_ROOT`` variable (described below).
Another alternative is to put a copy of (or a symlink to) the boost
subdirectory in ``<hyperscan-source-path>/include/boost``.
subdirectory in ``<vectorscanscan-source-path>/include/boost``.
For example: for the Boost-1.59.0 release: ::
ln -s boost_1_59_0/boost <hyperscan-source-path>/include/boost
ln -s boost_1_59_0/boost <vectorscan-source-path>/include/boost
As Hyperscan uses the header-only parts of Boost, it is not necessary to
As Vectorscan uses the header-only parts of Boost, it is not necessary to
compile the Boost libraries.
CMake Configuration
@ -168,11 +163,12 @@ Common options for CMake include:
| | Valid options are Debug, Release, RelWithDebInfo, |
| | and MinSizeRel. Default is RelWithDebInfo. |
+------------------------+----------------------------------------------------+
| BUILD_SHARED_LIBS | Build Hyperscan as a shared library instead of |
| BUILD_SHARED_LIBS | Build Vectorscan as a shared library instead of |
| | the default static library. |
| | Default: Off |
+------------------------+----------------------------------------------------+
| BUILD_STATIC_AND_SHARED| Build both static and shared Hyperscan libs. |
| | Default off. |
| BUILD_STATIC_LIBS | Build Vectorscan as a static library. |
| | Default: On |
+------------------------+----------------------------------------------------+
| BOOST_ROOT | Location of Boost source tree. |
+------------------------+----------------------------------------------------+
@ -180,12 +176,64 @@ Common options for CMake include:
+------------------------+----------------------------------------------------+
| FAT_RUNTIME | Build the :ref:`fat runtime<fat_runtime>`. Default |
| | true on Linux, not available elsewhere. |
| | Default: Off |
+------------------------+----------------------------------------------------+
| USE_CPU_NATIVE | Native CPU detection is off by default, however it |
| | is possible to build a performance-oriented non-fat|
| | library tuned to your CPU. |
| | Default: Off |
+------------------------+----------------------------------------------------+
| SANITIZE | Use libasan sanitizer to detect possible bugs. |
| | Valid options are address, memory and undefined. |
+------------------------+----------------------------------------------------+
| SIMDE_BACKEND | Enable SIMDe backend. If this is chosen all native |
| | (SSE/AVX/AVX512/Neon/SVE/VSX) backends will be |
| | disabled and a SIMDe SSE4.2 emulation backend will |
| | be enabled. This will enable Vectorscan to build |
| | and run on architectures without SIMD. |
| | Default: Off |
+------------------------+----------------------------------------------------+
| SIMDE_NATIVE | Enable SIMDe native emulation of x86 SSE4.2 |
| | intrinsics on the building platform. That is, |
| | SSE4.2 intrinsics will be emulated using Neon on |
| | an Arm platform, or VSX on a Power platform, etc. |
| | Default: Off |
+------------------------+----------------------------------------------------+
X86 platform specific options include:
+------------------------+----------------------------------------------------+
| Variable | Description |
+========================+====================================================+
| BUILD_AVX2 | Enable code for AVX2. |
+------------------------+----------------------------------------------------+
| BUILD_AVX512 | Enable code for AVX512. Implies BUILD_AVX2. |
+------------------------+----------------------------------------------------+
| BUILD_AVX512VBMI | Enable code for AVX512 with VBMI extension. Implies|
| | BUILD_AVX512. |
+------------------------+----------------------------------------------------+
Arm platform specific options include:
+------------------------+----------------------------------------------------+
| Variable | Description |
+========================+====================================================+
| BUILD_SVE | Enable code for SVE, like on AWS Graviton3 CPUs. |
| | Not much code is ported just for SVE , but enabling|
| | SVE code production, does improve code generation, |
| | see Benchmarks. |
+------------------------+----------------------------------------------------+
| BUILD_SVE2 | Enable code for SVE2, implies BUILD_SVE. Most |
| | non-Neon code is written for SVE2. |
+------------------------+----------------------------------------------------+
| BUILD_SVE2_BITPERM | Enable code for SVE2_BITPERM harwdare feature, |
| | implies BUILD_SVE2. |
+------------------------+----------------------------------------------------+
For example, to generate a ``Debug`` build: ::
cd <build-dir>
cmake -DCMAKE_BUILD_TYPE=Debug <hyperscan-source-path>
cmake -DCMAKE_BUILD_TYPE=Debug <vectorscan-source-path>
@ -193,7 +241,7 @@ Build Type
----------
CMake determines a number of features for a build based on the Build Type.
Hyperscan defaults to ``RelWithDebInfo``, i.e. "release with debugging
Vectorscan defaults to ``RelWithDebInfo``, i.e. "release with debugging
information". This is a performance optimized build without runtime assertions
but with debug symbols enabled.
@ -201,7 +249,7 @@ The other types of builds are:
* ``Release``: as above, but without debug symbols
* ``MinSizeRel``: a stripped release build
* ``Debug``: used when developing Hyperscan. Includes runtime assertions
* ``Debug``: used when developing Vectorscan. Includes runtime assertions
(which has a large impact on runtime performance), and will also enable
some other build features like building internal unit
tests.
@ -211,7 +259,7 @@ The other types of builds are:
Target Architecture
-------------------
Unless using the :ref:`fat runtime<fat_runtime>`, by default Hyperscan will be
Unless using the :ref:`fat runtime<fat_runtime>`, by default Vectorscan will be
compiled to target the instruction set of the processor of the machine that
being used for compilation. This is done via the use of ``-march=native``. The
result of this means that a library built on one machine may not work on a
@ -223,7 +271,7 @@ CMake, or ``CMAKE_C_FLAGS`` and ``CMAKE_CXX_FLAGS`` on the CMake command line. F
example, to set the instruction subsets up to ``SSE4.2`` using GCC 4.8: ::
cmake -DCMAKE_C_FLAGS="-march=corei7" \
-DCMAKE_CXX_FLAGS="-march=corei7" <hyperscan-source-path>
-DCMAKE_CXX_FLAGS="-march=corei7" <vectorscan-source-path>
For more information, refer to :ref:`instr_specialization`.
@ -232,17 +280,17 @@ For more information, refer to :ref:`instr_specialization`.
Fat Runtime
-----------
A feature introduced in Hyperscan v4.4 is the ability for the Hyperscan
A feature introduced in Hyperscan v4.4 is the ability for the Vectorscan
library to dispatch the most appropriate runtime code for the host processor.
This feature is called the "fat runtime", as a single Hyperscan library
This feature is called the "fat runtime", as a single Vectorscan library
contains multiple copies of the runtime code for different instruction sets.
.. note::
The fat runtime feature is only available on Linux. Release builds of
Hyperscan will default to having the fat runtime enabled where supported.
Vectorscan will default to having the fat runtime enabled where supported.
When building the library with the fat runtime, the Hyperscan runtime code
When building the library with the fat runtime, the Vectorscan runtime code
will be compiled multiple times for these different instruction sets, and
these compiled objects are combined into one library. There are no changes to
how user applications are built against this library.
@ -254,11 +302,11 @@ resolved so that the right version of each API function is used. There is no
impact on function call performance, as this check and resolution is performed
by the ELF loader once when the binary is loaded.
If the Hyperscan library is used on x86 systems without ``SSSE3``, the runtime
If the Vectorscan library is used on x86 systems without ``SSSE4.2``, the runtime
API functions will resolve to functions that return :c:member:`HS_ARCH_ERROR`
instead of potentially executing illegal instructions. The API function
:c:func:`hs_valid_platform` can be used by application writers to determine if
the current platform is supported by Hyperscan.
the current platform is supported by Vectorscan.
As of this release, the variants of the runtime that are built, and the CPU
capability that is required, are the following:
@ -299,6 +347,11 @@ capability that is required, are the following:
cmake -DBUILD_AVX512VBMI=on <...>
Vectorscan add support for Arm processors and SVE, SV2 and SVE2_BITPERM.
example: ::
cmake -DBUILD_SVE=ON -DBUILD_SVE2=ON -DBUILD_SVE2_BITPERM=ON <...>
As the fat runtime requires compiler, libc, and binutils support, at this time
it will only be enabled for Linux builds where the compiler supports the
`indirect function "ifunc" function attribute

2
doc/dev-reference/index.rst
View File

@ -1,5 +1,5 @@
###############################################
Hyperscan |version| Developer's Reference Guide
Vectorscan |version| Developer's Reference Guide
###############################################
-------

22
doc/dev-reference/intro.rst
View File

@ -5,11 +5,11 @@
Introduction
############
Hyperscan is a software regular expression matching engine designed with
Vectorscan is a software regular expression matching engine designed with
high performance and flexibility in mind. It is implemented as a library that
exposes a straightforward C API.
The Hyperscan API itself is composed of two major components:
The Vectorscan API itself is composed of two major components:
***********
Compilation
@ -17,7 +17,7 @@ Compilation
These functions take a group of regular expressions, along with identifiers and
option flags, and compile them into an immutable database that can be used by
the Hyperscan scanning API. This compilation process performs considerable
the Vectorscan scanning API. This compilation process performs considerable
analysis and optimization work in order to build a database that will match the
given expressions efficiently.
@ -36,8 +36,8 @@ See :ref:`compilation` for more detail.
Scanning
********
Once a Hyperscan database has been created, it can be used to scan data in
memory. Hyperscan provides several scanning modes, depending on whether the
Once a Vectorscan database has been created, it can be used to scan data in
memory. Vectorscan provides several scanning modes, depending on whether the
data to be scanned is available as a single contiguous block, whether it is
distributed amongst several blocks in memory at the same time, or whether it is
to be scanned as a sequence of blocks in a stream.
@ -45,7 +45,7 @@ to be scanned as a sequence of blocks in a stream.
Matches are delivered to the application via a user-supplied callback function
that is called synchronously for each match.
For a given database, Hyperscan provides several guarantees:
For a given database, Vectorscan provides several guarantees:
* No memory allocations occur at runtime with the exception of two
fixed-size allocations, both of which should be done ahead of time for
@ -56,7 +56,7 @@ For a given database, Hyperscan provides several guarantees:
call.
- **Stream state**: in streaming mode only, some state space is required to
store data that persists between scan calls for each stream. This allows
Hyperscan to track matches that span multiple blocks of data.
Vectorscan to track matches that span multiple blocks of data.
* The sizes of the scratch space and stream state (in streaming mode) required
for a given database are fixed and determined at database compile time. This
@ -64,7 +64,7 @@ For a given database, Hyperscan provides several guarantees:
time, and these structures can be pre-allocated if required for performance
reasons.
* Any pattern that has successfully been compiled by the Hyperscan compiler can
* Any pattern that has successfully been compiled by the Vectorscan compiler can
be scanned against any input. There are no internal resource limits or other
limitations at runtime that could cause a scan call to return an error.
@ -74,12 +74,12 @@ See :ref:`runtime` for more detail.
Tools
*****
Some utilities for testing and benchmarking Hyperscan are included with the
Some utilities for testing and benchmarking Vectorscan are included with the
library. See :ref:`tools` for more information.
************
Example Code
************
Some simple example code demonstrating the use of the Hyperscan API is
available in the ``examples/`` subdirectory of the Hyperscan distribution.
Some simple example code demonstrating the use of the Vectorscan API is
available in the ``examples/`` subdirectory of the Vectorscan distribution.

22
doc/dev-reference/performance.rst
View File

@ -4,7 +4,7 @@
Performance Considerations
##########################
Hyperscan supports a wide range of patterns in all three scanning modes. It is
Vectorscan supports a wide range of patterns in all three scanning modes. It is
capable of extremely high levels of performance, but certain patterns can
reduce performance markedly.
@ -25,7 +25,7 @@ For example, caseless matching of :regexp:`/abc/` can be written as:
* :regexp:`/(?i)abc(?-i)/`
* :regexp:`/abc/i`
Hyperscan is capable of handling all these constructs. Unless there is a
Vectorscan is capable of handling all these constructs. Unless there is a
specific reason otherwise, do not rewrite patterns from one form to another.
As another example, matching of :regexp:`/foo(bar|baz)(frotz)?/` can be
@ -41,24 +41,24 @@ Library usage
.. tip:: Do not hand-optimize library usage.
The Hyperscan library is capable of dealing with small writes, unusually large
The Vectorscan library is capable of dealing with small writes, unusually large
and small pattern sets, etc. Unless there is a specific performance problem
with some usage of the library, it is best to use Hyperscan in a simple and
with some usage of the library, it is best to use Vectorscan in a simple and
direct fashion. For example, it is unlikely for there to be much benefit in
buffering input to the library into larger blocks unless streaming writes are
tiny (say, 1-2 bytes at a time).
Unlike many other pattern matching products, Hyperscan will run faster with
Unlike many other pattern matching products, Vectorscan will run faster with
small numbers of patterns and slower with large numbers of patterns in a smooth
fashion (as opposed to, typically, running at a moderate speed up to some fixed
limit then either breaking or running half as fast).
Hyperscan also provides high-throughput matching with a single thread of
control per core; if a database runs at 3.0 Gbps in Hyperscan it means that a
Vectorscan also provides high-throughput matching with a single thread of
control per core; if a database runs at 3.0 Gbps in Vectorscan it means that a
3000-bit block of data will be scanned in 1 microsecond in a single thread of
control, not that it is required to scan 22 3000-bit blocks of data in 22
microseconds. Thus, it is not usually necessary to buffer data to supply
Hyperscan with available parallelism.
Vectorscan with available parallelism.
********************
Block-based matching
@ -72,7 +72,7 @@ accumulated before processing, it should be scanned in block rather than in
streaming mode.
Unnecessary use of streaming mode reduces the number of optimizations that can
be applied in Hyperscan and may make some patterns run slower.
be applied in Vectorscan and may make some patterns run slower.
If there is a mixture of 'block' and 'streaming' mode patterns, these should be
scanned in separate databases except in the case that the streaming patterns
@ -107,7 +107,7 @@ Allocate scratch ahead of time
Scratch allocation is not necessarily a cheap operation. Since it is the first
time (after compilation or deserialization) that a pattern database is used,
Hyperscan performs some validation checks inside :c:func:`hs_alloc_scratch` and
Vectorscan performs some validation checks inside :c:func:`hs_alloc_scratch` and
must also allocate memory.
Therefore, it is important to ensure that :c:func:`hs_alloc_scratch` is not
@ -329,7 +329,7 @@ Consequently, :regexp:`/foo.*bar/L` with a check on start of match values after
the callback is considerably more expensive and general than
:regexp:`/foo.{300}bar/`.
Similarly, the :c:member:`hs_expr_ext::min_length` extended parameter can be
Similarly, the :cpp:member:`hs_expr_ext::min_length` extended parameter can be
used to specify a lower bound on the length of the matches for a pattern. Using
this facility may be more lightweight in some circumstances than using the SOM
flag and post-confirming match length in the calling application.

18
doc/dev-reference/preface.rst
View File

@ -6,35 +6,35 @@ Preface
Overview
********
Hyperscan is a regular expression engine designed to offer high performance, the
Vectorscan is a regular expression engine designed to offer high performance, the
ability to match multiple expressions simultaneously and flexibility in
scanning operation.
Patterns are provided to a compilation interface which generates an immutable
pattern database. The scan interface then can be used to scan a target data
buffer for the given patterns, returning any matching results from that data
buffer. Hyperscan also provides a streaming mode, in which matches that span
buffer. Vectorscan also provides a streaming mode, in which matches that span
several blocks in a stream are detected.
This document is designed to facilitate code-level integration of the Hyperscan
This document is designed to facilitate code-level integration of the Vectorscan
library with existing or new applications.
:ref:`intro` is a short overview of the Hyperscan library, with more detail on
the Hyperscan API provided in the subsequent sections: :ref:`compilation` and
:ref:`intro` is a short overview of the Vectorscan library, with more detail on
the Vectorscan API provided in the subsequent sections: :ref:`compilation` and
:ref:`runtime`.
:ref:`perf` provides details on various factors which may impact the
performance of a Hyperscan integration.
performance of a Vectorscan integration.
:ref:`api_constants` and :ref:`api_files` provides a detailed summary of the
Hyperscan Application Programming Interface (API).
Vectorscan Application Programming Interface (API).
********
Audience
********
This guide is aimed at developers interested in integrating Hyperscan into an
application. For information on building the Hyperscan library, see the Quick
This guide is aimed at developers interested in integrating Vectorscan into an
application. For information on building the Vectorscan library, see the Quick
Start Guide.
***********

24
doc/dev-reference/runtime.rst
View File

@ -4,7 +4,7 @@
Scanning for Patterns
#####################
Hyperscan provides three different scanning modes, each with its own scan
Vectorscan provides three different scanning modes, each with its own scan
function beginning with ``hs_scan``. In addition, streaming mode has a number
of other API functions for managing stream state.
@ -33,8 +33,8 @@ See :c:type:`match_event_handler` for more information.
Streaming Mode
**************
The core of the Hyperscan streaming runtime API consists of functions to open,
scan, and close Hyperscan data streams:
The core of the Vectorscan streaming runtime API consists of functions to open,
scan, and close Vectorscan data streams:
* :c:func:`hs_open_stream`: allocates and initializes a new stream for scanning.
@ -57,14 +57,14 @@ will return immediately with :c:member:`HS_SCAN_TERMINATED`. The caller must
still call :c:func:`hs_close_stream` to complete the clean-up process for that
stream.
Streams exist in the Hyperscan library so that pattern matching state can be
Streams exist in the Vectorscan library so that pattern matching state can be
maintained across multiple blocks of target data -- without maintaining this
state, it would not be possible to detect patterns that span these blocks of
data. This, however, does come at the cost of requiring an amount of storage
per-stream (the size of this storage is fixed at compile time), and a slight
performance penalty in some cases to manage the state.
While Hyperscan does always support a strict ordering of multiple matches,
While Vectorscan does always support a strict ordering of multiple matches,
streaming matches will not be delivered at offsets before the current stream
write, with the exception of zero-width asserts, where constructs such as
:regexp:`\\b` and :regexp:`$` can cause a match on the final character of a
@ -76,7 +76,7 @@ Stream Management
=================
In addition to :c:func:`hs_open_stream`, :c:func:`hs_scan_stream`, and
:c:func:`hs_close_stream`, the Hyperscan API provides a number of other
:c:func:`hs_close_stream`, the Vectorscan API provides a number of other
functions for the management of streams:
* :c:func:`hs_reset_stream`: resets a stream to its initial state; this is
@ -98,10 +98,10 @@ A stream object is allocated as a fixed size region of memory which has been
sized to ensure that no memory allocations are required during scan
operations. When the system is under memory pressure, it may be useful to reduce
the memory consumed by streams that are not expected to be used soon. The
Hyperscan API provides calls for translating a stream to and from a compressed
Vectorscan API provides calls for translating a stream to and from a compressed
representation for this purpose. The compressed representation differs from the
full stream object as it does not reserve space for components which are not
required given the current stream state. The Hyperscan API functions for this
required given the current stream state. The Vectorscan API functions for this
functionality are:
* :c:func:`hs_compress_stream`: fills the provided buffer with a compressed
@ -157,7 +157,7 @@ scanned in block mode.
Scratch Space
*************
While scanning data, Hyperscan needs a small amount of temporary memory to store
While scanning data, Vectorscan needs a small amount of temporary memory to store
on-the-fly internal data. This amount is unfortunately too large to fit on the
stack, particularly for embedded applications, and allocating memory dynamically
is too expensive, so a pre-allocated "scratch" space must be provided to the
@ -170,7 +170,7 @@ databases, only a single scratch region is necessary: in this case, calling
will ensure that the scratch space is large enough to support scanning against
any of the given databases.
While the Hyperscan library is re-entrant, the use of scratch spaces is not.
While the Vectorscan library is re-entrant, the use of scratch spaces is not.
For example, if by design it is deemed necessary to run recursive or nested
scanning (say, from the match callback function), then an additional scratch
space is required for that context.
@ -219,11 +219,11 @@ For example:
Custom Allocators
*****************
By default, structures used by Hyperscan at runtime (scratch space, stream
By default, structures used by Vectorscan at runtime (scratch space, stream
state, etc) are allocated with the default system allocators, usually
``malloc()`` and ``free()``.
The Hyperscan API provides a facility for changing this behaviour to support
The Vectorscan API provides a facility for changing this behaviour to support
applications that use custom memory allocators.
These functions are:

20
doc/dev-reference/serialization.rst
View File

@ -4,7 +4,7 @@
Serialization
#############
For some applications, compiling Hyperscan pattern databases immediately prior
For some applications, compiling Vectorscan pattern databases immediately prior
to use is not an appropriate design. Some users may wish to:
* Compile pattern databases on a different host;
@ -14,9 +14,9 @@ to use is not an appropriate design. Some users may wish to:
* Control the region of memory in which the compiled database is located.
Hyperscan pattern databases are not completely flat in memory: they contain
Vectorscan pattern databases are not completely flat in memory: they contain
pointers and have specific alignment requirements. Therefore, they cannot be
copied (or otherwise relocated) directly. To enable these use cases, Hyperscan
copied (or otherwise relocated) directly. To enable these use cases, Vectorscan
provides functionality for serializing and deserializing compiled pattern
databases.
@ -40,10 +40,10 @@ The API provides the following functions:
returns a string containing information about the database. This call is
analogous to :c:func:`hs_database_info`.
.. note:: Hyperscan performs both version and platform compatibility checks
.. note:: Vectorscan performs both version and platform compatibility checks
upon deserialization. The :c:func:`hs_deserialize_database` and
:c:func:`hs_deserialize_database_at` functions will only permit the
deserialization of databases compiled with (a) the same version of Hyperscan
deserialization of databases compiled with (a) the same version of Vectorscan
and (b) platform features supported by the current host platform. See
:ref:`instr_specialization` for more information on platform specialization.
@ -51,17 +51,17 @@ The API provides the following functions:
The Runtime Library
===================
The main Hyperscan library (``libhs``) contains both the compiler and runtime
portions of the library. This means that in order to support the Hyperscan
The main Vectorscan library (``libhs``) contains both the compiler and runtime
portions of the library. This means that in order to support the Vectorscan
compiler, which is written in C++, it requires C++ linkage and has a
dependency on the C++ standard library.
Many embedded applications require only the scanning ("runtime") portion of the
Hyperscan library. In these cases, pattern compilation generally takes place on
Vectorscan library. In these cases, pattern compilation generally takes place on
another host, and serialized pattern databases are delivered to the application
for use.
To support these applications without requiring the C++ dependency, a
runtime-only version of the Hyperscan library, called ``libhs_runtime``, is also
runtime-only version of the Vectorscan library, called ``libhs_runtime``, is also
distributed. This library does not depend on the C++ standard library and
provides all Hyperscan functions other that those used to compile databases.
provides all Vectorscan functions other that those used to compile databases.

44
doc/dev-reference/tools.rst
View File

@ -4,14 +4,14 @@
Tools
#####
This section describes the set of utilities included with the Hyperscan library.
This section describes the set of utilities included with the Vectorscan library.
********************
Quick Check: hscheck
********************
The ``hscheck`` tool allows the user to quickly check whether Hyperscan supports
a group of patterns. If a pattern is rejected by Hyperscan's compiler, the
The ``hscheck`` tool allows the user to quickly check whether Vectorscan supports
a group of patterns. If a pattern is rejected by Vectorscan's compiler, the
compile error is provided on standard output.
For example, given the following three patterns (the last of which contains a
@ -34,7 +34,7 @@ syntax error) in a file called ``/tmp/test``::
Benchmarker: hsbench
********************
The ``hsbench`` tool provides an easy way to measure Hyperscan's performance
The ``hsbench`` tool provides an easy way to measure Vectorscan's performance
for a particular set of patterns and corpus of data to be scanned.
Patterns are supplied in the format described below in
@ -44,7 +44,7 @@ easy control of how a corpus is broken into blocks and streams.
.. note:: A group of Python scripts for constructing corpora databases from
various input types, such as PCAP network traffic captures or text files, can
be found in the Hyperscan source tree in ``tools/hsbench/scripts``.
be found in the Vectorscan source tree in ``tools/hsbench/scripts``.
Running hsbench
===============
@ -56,7 +56,7 @@ produce output like this::
$ hsbench -e /tmp/patterns -c /tmp/corpus.db
Signatures: /tmp/patterns
Hyperscan info: Version: 4.3.1 Features: AVX2 Mode: STREAM
Vectorscan info: Version: 5.4.11 Features: AVX2 Mode: STREAM
Expression count: 200
Bytecode size: 342,540 bytes
Database CRC: 0x6cd6b67c
@ -77,7 +77,7 @@ takes to perform all twenty scans. The number of repeats can be changed with the
``-n`` argument, and the results of each scan will be displayed if the
``--per-scan`` argument is specified.
To benchmark Hyperscan on more than one core, you can supply a list of cores
To benchmark Vectorscan on more than one core, you can supply a list of cores
with the ``-T`` argument, which will instruct ``hsbench`` to start one
benchmark thread per core given and compute the throughput from the time taken
to complete all of them.
@ -91,17 +91,17 @@ Correctness Testing: hscollider
*******************************
The ``hscollider`` tool, or Pattern Collider, provides a way to verify
Hyperscan's matching behaviour. It does this by compiling and scanning patterns
Vectorscan's matching behaviour. It does this by compiling and scanning patterns
(either singly or in groups) against known corpora and comparing the results
against another engine (the "ground truth"). Two sources of ground truth for
comparison are available:
* The PCRE library (http://pcre.org/).
* An NFA simulation run on Hyperscan's compile-time graph representation. This
* An NFA simulation run on Vectorscan's compile-time graph representation. This
is used if PCRE cannot support the pattern or if PCRE execution fails due to
a resource limit.
Much of Hyperscan's testing infrastructure is built on ``hscollider``, and the
Much of Vectorscan's testing infrastructure is built on ``hscollider``, and the
tool is designed to take advantage of multiple cores and provide considerable
flexibility in controlling the test. These options are described in the help
(``hscollider -h``) and include:
@ -116,11 +116,11 @@ flexibility in controlling the test. These options are described in the help
Using hscollider to debug a pattern
===================================
One common use-case for ``hscollider`` is to determine whether Hyperscan will
One common use-case for ``hscollider`` is to determine whether Vectorscan will
match a pattern in the expected location, and whether this accords with PCRE's
behaviour for the same case.
Here is an example. We put our pattern in a file in Hyperscan's pattern
Here is an example. We put our pattern in a file in Vectorscan's pattern
format::
$ cat /tmp/pat
@ -172,7 +172,7 @@ individual matches are displayed in the output::
Total elapsed time: 0.00522815 secs.
We can see from this output that both PCRE and Hyperscan find matches ending at
We can see from this output that both PCRE and Vectorscan find matches ending at
offset 33 and 45, and so ``hscollider`` considers this test case to have
passed.
@ -180,13 +180,13 @@ passed.
corpus alignment 0, and ``-T 1`` instructs us to only use one thread.)
.. note:: In default operation, PCRE produces only one match for a scan, unlike
Hyperscan's automata semantics. The ``hscollider`` tool uses libpcre's
"callout" functionality to match Hyperscan's semantics.
Vectorscan's automata semantics. The ``hscollider`` tool uses libpcre's
"callout" functionality to match Vectorscan's semantics.
Running a larger scan test
==========================
A set of patterns for testing purposes are distributed with Hyperscan, and these
A set of patterns for testing purposes are distributed with Vectorscan, and these
can be tested via ``hscollider`` on an in-tree build. Two CMake targets are
provided to do this easily:
@ -202,10 +202,10 @@ Debugging: hsdump
*****************
When built in debug mode (using the CMake directive ``CMAKE_BUILD_TYPE`` set to
``Debug``), Hyperscan includes support for dumping information about its
``Debug``), Vectorscan includes support for dumping information about its
internals during pattern compilation with the ``hsdump`` tool.
This information is mostly of use to Hyperscan developers familiar with the
This information is mostly of use to Vectorscan developers familiar with the
library's internal structure, but can be used to diagnose issues with patterns
and provide more information in bug reports.
@ -215,7 +215,7 @@ and provide more information in bug reports.
Pattern Format
**************
All of the Hyperscan tools accept patterns in the same format, read from plain
All of the Vectorscan tools accept patterns in the same format, read from plain
text files with one pattern per line. Each line looks like this:
* ``<integer id>:/<regex>/<flags>``
@ -227,12 +227,12 @@ For example::
3:/^.{10,20}hatstand/m
The integer ID is the value that will be reported when a match is found by
Hyperscan and must be unique.
Vectorscan and must be unique.
The pattern itself is a regular expression in PCRE syntax; see
:ref:`compilation` for more information on supported features.
The flags are single characters that map to Hyperscan flags as follows:
The flags are single characters that map to Vectorscan flags as follows:
========= ================================= ===========
Character API Flag Description
@ -256,7 +256,7 @@ between braces, separated by commas. For example::
1:/hatstand.*teakettle/s{min_offset=50,max_offset=100}
All Hyperscan tools will accept a pattern file (or a directory containing
All Vectorscan tools will accept a pattern file (or a directory containing
pattern files) with the ``-e`` argument. If no further arguments constraining
the pattern set are given, all patterns in those files are used.

11
examples/patbench.cc
View File

@ -202,7 +202,7 @@ struct FiveTuple {
unsigned int dstPort;
// Construct a FiveTuple from a TCP or UDP packet.
FiveTuple(const struct ip *iphdr) {
explicit FiveTuple(const struct ip *iphdr) {
// IP fields
protocol = iphdr->ip_p;
srcAddr = iphdr->ip_src.s_addr;
@ -391,7 +391,7 @@ public:
// Close all open Hyperscan streams (potentially generating any
// end-anchored matches)
void closeStreams() {
for (auto &stream : streams) {
for (const auto &stream : streams) {
hs_error_t err =
hs_close_stream(stream, scratch, onMatch, &matchCount);
if (err != HS_SUCCESS) {
@ -444,7 +444,7 @@ class Sigdata {
public:
Sigdata() {}
Sigdata(const char *filename) {
explicit Sigdata(const char *filename) {
parseFile(filename, patterns, flags, ids, originals);
}
@ -568,7 +568,7 @@ double measure_block_time(Benchmark &bench, unsigned int repeatCount) {
}
static
double eval_set(Benchmark &bench, Sigdata &sigs, unsigned int mode,
double eval_set(Benchmark &bench, const Sigdata &sigs, unsigned int mode,
unsigned repeatCount, Criterion criterion,
bool diagnose = true) {
double compileTime = 0;
@ -608,8 +608,9 @@ double eval_set(Benchmark &bench, Sigdata &sigs, unsigned int mode,
scan_time = measure_stream_time(bench, repeatCount);
}
size_t bytes = bench.bytes();
size_t matches = bench.matches();
if (diagnose) {
size_t matches = bench.matches();
std::ios::fmtflags f(cout.flags());
cout << "Scan time " << std::fixed << std::setprecision(3) << scan_time
<< " sec, Scanned " << bytes * repeatCount << " bytes, Throughput "

18
examples/pcapscan.cc
View File

@ -100,15 +100,14 @@ struct FiveTuple {
unsigned int dstPort;
// Construct a FiveTuple from a TCP or UDP packet.
FiveTuple(const struct ip *iphdr) {
explicit FiveTuple(const struct ip *iphdr) {
// IP fields
protocol = iphdr->ip_p;
srcAddr = iphdr->ip_src.s_addr;
dstAddr = iphdr->ip_dst.s_addr;
// UDP/TCP ports
const struct udphdr *uh =
(const struct udphdr *)(((const char *)iphdr) + (iphdr->ip_hl * 4));
const struct udphdr *uh = reinterpret_cast<const struct udphdr *>(iphdr) + (iphdr->ip_hl * 4);
srcPort = uh->uh_sport;
dstPort = uh->uh_dport;
}
@ -137,7 +136,7 @@ static
int onMatch(unsigned int id, unsigned long long from, unsigned long long to,
unsigned int flags, void *ctx) {
// Our context points to a size_t storing the match count
size_t *matches = (size_t *)ctx;
size_t *matches = static_cast<size_t *>(ctx);
(*matches)++;
return 0; // continue matching
}
@ -233,9 +232,8 @@ public:
}
// Valid TCP or UDP packet
const struct ip *iphdr = (const struct ip *)(pktData
+ sizeof(struct ether_header));
const char *payload = (const char *)pktData + offset;
const struct ip *iphdr = reinterpret_cast<const struct ip *>(pktData) + sizeof(struct ether_header);
const char *payload = reinterpret_cast<const char *>(pktData) + offset;
size_t id = stream_map.insert(std::make_pair(FiveTuple(iphdr),
stream_map.size())).first->second;
@ -281,7 +279,7 @@ public:
// Close all open Hyperscan streams (potentially generating any
// end-anchored matches)
void closeStreams() {
for (auto &stream : streams) {
for (const auto &stream : streams) {
hs_error_t err = hs_close_stream(stream, scratch, onMatch,
&matchCount);
if (err != HS_SUCCESS) {
@ -575,7 +573,7 @@ int main(int argc, char **argv) {
*/
static bool payloadOffset(const unsigned char *pkt_data, unsigned int *offset,
unsigned int *length) {
const ip *iph = (const ip *)(pkt_data + sizeof(ether_header));
const ip *iph = reinterpret_cast<const ip *>(pkt_data) + sizeof(ether_header);
const tcphdr *th = nullptr;
// Ignore packets that aren't IPv4
@ -594,7 +592,7 @@ static bool payloadOffset(const unsigned char *pkt_data, unsigned int *offset,
switch (iph->ip_p) {
case IPPROTO_TCP:
th = (const tcphdr *)((const char *)iph + ihlen);
th = reinterpret_cast<const tcphdr *>(iph) + ihlen;
thlen = th->th_off * 4;
break;
case IPPROTO_UDP:

4
examples/simplegrep.c
View File

@ -67,7 +67,7 @@
* to pass in the pattern that was being searched for so we can print it out.
*/
static int eventHandler(unsigned int id, unsigned long long from,
unsigned long long to, unsigned int flags, void *ctx) {
unsigned long long to, unsigned int flags, void *ctx) { // cppcheck-suppress constParameterCallback
printf("Match for pattern \"%s\" at offset %llu\n", (char *)ctx, to);
return 0;
}
@ -150,7 +150,7 @@ int main(int argc, char *argv[]) {
}
char *pattern = argv[1];
char *inputFN = argv[2];
const char *inputFN = argv[2];
/* First, we attempt to compile the pattern provided on the command line.
* We assume 'DOTALL' semantics, meaning that the '.' meta-character will

2
libhs.pc.in
View File

@ -4,7 +4,7 @@ libdir=@CMAKE_INSTALL_PREFIX@/@CMAKE_INSTALL_LIBDIR@
includedir=@CMAKE_INSTALL_PREFIX@/@CMAKE_INSTALL_INCLUDEDIR@
Name: libhs
Description: Intel(R) Hyperscan Library
Description: A portable fork of the high-performance regular expression matching library
Version: @HS_VERSION@
Libs: -L${libdir} -lhs
Cflags: -I${includedir}/hs

2
simde

@ -1 +1 @@
Subproject commit aae22459fa284e9fc2b7d4b8e4571afa0418125f
Subproject commit 416091ebdb9e901b29d026633e73167d6353a0b0

14
src/compiler/asserts.cpp
View File

@ -176,7 +176,8 @@ void replaceAssertVertex(NGHolder &g, NFAVertex t, const ExpressionInfo &expr,
auto ecit = edge_cache.find(cache_key);
if (ecit == edge_cache.end()) {
DEBUG_PRINTF("adding edge %zu %zu\n", g[u].index, g[v].index);
NFAEdge e = add_edge(u, v, g);
NFAEdge e;
std::tie(e, std::ignore) = add_edge(u, v, g);
edge_cache.emplace(cache_key, e);
g[e].assert_flags = flags;
if (++assert_edge_count > MAX_ASSERT_EDGES) {
@ -229,11 +230,12 @@ void checkForMultilineStart(ReportManager &rm, NGHolder &g,
/* we need to interpose a dummy dot vertex between v and accept if
* required so that ^ doesn't match trailing \n */
for (const auto &e : out_edges_range(v, g)) {
if (target(e, g) == g.accept) {
dead.emplace_back(e);
}
}
auto deads = [&g=g](const NFAEdge &e) {
return (target(e, g) == g.accept);
};
const auto &er = out_edges_range(v, g);
std::copy_if(begin(er), end(er), std::back_inserter(dead), deads);
/* assert has been resolved; clear flag */
g[v].assert_flags &= ~POS_FLAG_MULTILINE_START;
}

8
src/compiler/compiler.cpp
View File

@ -443,7 +443,7 @@ bytecode_ptr<RoseEngine> generateRoseEngine(NG &ng) {
if (!rose) {
DEBUG_PRINTF("error building rose\n");
assert(0);
return nullptr;
return bytecode_ptr<RoseEngine>(nullptr);
}
dumpReportManager(ng.rm, ng.cc.grey);
@ -478,7 +478,7 @@ hs_database_t *dbCreate(const char *in_bytecode, size_t len, u64a platform) {
DEBUG_PRINTF("db size %zu\n", db_len);
DEBUG_PRINTF("db platform %llx\n", platform);
struct hs_database *db = (struct hs_database *)hs_database_alloc(db_len);
struct hs_database *db = static_cast<struct hs_database *>(hs_database_alloc(db_len));
if (hs_check_alloc(db) != HS_SUCCESS) {
hs_database_free(db);
return nullptr;
@ -492,7 +492,7 @@ hs_database_t *dbCreate(const char *in_bytecode, size_t len, u64a platform) {
DEBUG_PRINTF("shift is %zu\n", shift);
db->bytecode = offsetof(struct hs_database, bytes) - shift;
char *bytecode = (char *)db + db->bytecode;
char *bytecode = reinterpret_cast<char *>(db) + db->bytecode;
assert(ISALIGNED_CL(bytecode));
db->magic = HS_DB_MAGIC;
@ -525,7 +525,7 @@ struct hs_database *build(NG &ng, unsigned int *length, u8 pureFlag) {
throw CompileError("Internal error.");
}
const char *bytecode = (const char *)(rose.get());
const char *bytecode = reinterpret_cast<const char *>(rose.get());
const platform_t p = target_to_platform(ng.cc.target_info);
struct hs_database *db = dbCreate(bytecode, *length, p);
if (!db) {

5
src/compiler/error.cpp
View File

@ -57,15 +57,14 @@ extern const hs_compile_error_t hs_badalloc = {
namespace ue2 {
hs_compile_error_t *generateCompileError(const string &err, int expression) {
hs_compile_error_t *ret =
(struct hs_compile_error *)hs_misc_alloc(sizeof(hs_compile_error_t));
hs_compile_error_t *ret = static_cast<struct hs_compile_error *>(hs_misc_alloc(sizeof(hs_compile_error_t)));
if (ret) {
hs_error_t e = hs_check_alloc(ret);
if (e != HS_SUCCESS) {
hs_misc_free(ret);
return const_cast<hs_compile_error_t *>(&hs_badalloc);
}
char *msg = (char *)hs_misc_alloc(err.size() + 1);
char *msg = static_cast<char *>(hs_misc_alloc(err.size() + 1));
if (msg) {
e = hs_check_alloc(msg);
if (e != HS_SUCCESS) {

5
src/crc32.c
View File

@ -542,14 +542,13 @@ u32 crc32c_sb8_64_bit(u32 running_crc, const unsigned char* p_buf,
// Main aligned loop, processes eight bytes at a time.
u32 term1, term2;
for (size_t li = 0; li < running_length/8; li++) {
u32 block = *(const u32 *)p_buf;
crc ^= block;
p_buf += 4;
term1 = crc_tableil8_o88[crc & 0x000000FF] ^
u32 term1 = crc_tableil8_o88[crc & 0x000000FF] ^
crc_tableil8_o80[(crc >> 8) & 0x000000FF];
term2 = crc >> 16;
u32 term2 = crc >> 16;
crc = term1 ^
crc_tableil8_o72[term2 & 0x000000FF] ^
crc_tableil8_o64[(term2 >> 8) & 0x000000FF];

9
src/database.h
View File

@ -79,21 +79,18 @@ static UNUSED
const platform_t hs_current_platform_no_avx2 = {
HS_PLATFORM_NOAVX2 |
HS_PLATFORM_NOAVX512 |
HS_PLATFORM_NOAVX512VBMI |
0,
HS_PLATFORM_NOAVX512VBMI
};
static UNUSED
const platform_t hs_current_platform_no_avx512 = {
HS_PLATFORM_NOAVX512 |
HS_PLATFORM_NOAVX512VBMI |
0,
HS_PLATFORM_NOAVX512VBMI
};
static UNUSED
const platform_t hs_current_platform_no_avx512vbmi = {
HS_PLATFORM_NOAVX512VBMI |
0,
HS_PLATFORM_NOAVX512VBMI
};
/*

249
src/dispatcher.c
View File

@ -1,5 +1,6 @@
/*
* Copyright (c) 2016-2020, Intel Corporation
* Copyright (c) 2024, VectorCamp PC
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
@ -30,6 +31,39 @@
#include "hs_common.h"
#include "hs_runtime.h"
#include "ue2common.h"
/* Streamlining the dispatch to eliminate runtime checking/branching:
* What we want to do is, first call to the function will run the resolve
* code and set the static resolved/dispatch pointer to point to the
* correct function. Subsequent calls to the function will go directly to
* the resolved ptr. The simplest way to accomplish this is, to
* initially set the pointer to the resolve function.
* To accomplish this in a manner invisible to the user,
* we do involve some rather ugly/confusing macros in here.
* There are four macros that assemble the code for each function
* we want to dispatch in this manner:
* CREATE_DISPATCH
* this generates the declarations for the candidate target functions,
* for the fat_dispatch function pointer, for the resolve_ function,
* points the function pointer to the resolve function, and contains
* most of the definition of the resolve function. The very end of the
* resolve function is completed by the next macro, because in the
* CREATE_DISPATCH macro we have the argument list with the arg declarations,
* which is needed to generate correct function signatures, but we
* can't generate from this, in a macro, a _call_ to one of those functions.
* CONNECT_ARGS_1
* this macro fills in the actual call at the end of the resolve function,
* with the correct arg list. hence the name connect args.
* CONNECT_DISPATCH_2
* this macro likewise gives up the beginning of the definition of the
* actual entry point function (the 'real name' that's called by the user)
* but again in the pass-through call, cannot invoke the target without
* getting the arg list , which is supplied by the final macro,
* CONNECT_ARGS_3
*
*/
#if defined(ARCH_IA32) || defined(ARCH_X86_64)
#include "util/arch/x86/cpuid_inline.h"
#include "util/join.h"
@ -57,30 +91,38 @@
return (RTYPE)HS_ARCH_ERROR; \
} \
\
/* resolver */ \
static RTYPE (*JOIN(resolve_, NAME)(void))(__VA_ARGS__) { \
if (check_avx512vbmi()) { \
return JOIN(avx512vbmi_, NAME); \
} \
if (check_avx512()) { \
return JOIN(avx512_, NAME); \
} \
if (check_avx2()) { \
return JOIN(avx2_, NAME); \
} \
if (check_sse42() && check_popcnt()) { \
return JOIN(corei7_, NAME); \
} \
if (check_ssse3()) { \
return JOIN(core2_, NAME); \
} \
/* anything else is fail */ \
return JOIN(error_, NAME); \
} \
/* dispatch routing pointer for this function */ \
/* initially point it at the resolve function */ \
static RTYPE JOIN(resolve_, NAME)(__VA_ARGS__); \
static RTYPE (* JOIN(fat_dispatch_, NAME))(__VA_ARGS__) = \
&JOIN(resolve_, NAME); \
\
/* function */ \
HS_PUBLIC_API \
RTYPE NAME(__VA_ARGS__) __attribute__((ifunc("resolve_" #NAME)))
/* resolver */ \
static RTYPE JOIN(resolve_, NAME)(__VA_ARGS__) { \
if (check_avx512vbmi()) { \
fat_dispatch_ ## NAME = &JOIN(avx512vbmi_, NAME); \
} \
else if (check_avx512()) { \
fat_dispatch_ ## NAME = &JOIN(avx512_, NAME); \
} \
else if (check_avx2()) { \
fat_dispatch_ ## NAME = &JOIN(avx2_, NAME); \
} \
else if (check_sse42() && check_popcnt()) { \
fat_dispatch_ ## NAME = &JOIN(corei7_, NAME); \
} \
else if (check_ssse3()) { \
fat_dispatch_ ## NAME = &JOIN(core2_, NAME); \
} else { \
/* anything else is fail */ \
fat_dispatch_ ## NAME = &JOIN(error_, NAME); \
} \
/* the rest of the function is completed in the CONNECT_ARGS_1 macro. */
#elif defined(ARCH_AARCH64)
#include "util/arch/arm/cpuid_inline.h"
@ -97,99 +139,226 @@
return (RTYPE)HS_ARCH_ERROR; \
} \
\
/* resolver */ \
static RTYPE (*JOIN(resolve_, NAME)(void))(__VA_ARGS__) { \
if (check_sve2()) { \
return JOIN(sve2_, NAME); \
} \
if (check_sve()) { \
return JOIN(sve_, NAME); \
} \
if (check_neon()) { \
return JOIN(neon_, NAME); \
} \
/* anything else is fail */ \
return JOIN(error_, NAME); \
} \
/* dispatch routing pointer for this function */ \
/* initially point it at the resolve function */ \
static RTYPE JOIN(resolve_, NAME)(__VA_ARGS__); \
static RTYPE (* JOIN(fat_dispatch_, NAME))(__VA_ARGS__) = \
&JOIN(resolve_, NAME); \
\
/* function */ \
HS_PUBLIC_API \
RTYPE NAME(__VA_ARGS__) __attribute__((ifunc("resolve_" #NAME)))
/* resolver */ \
static RTYPE JOIN(resolve_, NAME)(__VA_ARGS__) { \
if (check_sve2()) { \
fat_dispatch_ ## NAME = &JOIN(sve2_, NAME); \
} \
else if (check_sve()) { \
fat_dispatch_ ## NAME = &JOIN(sve_, NAME); \
} \
else if (check_neon()) { \
fat_dispatch_ ## NAME = &JOIN(neon_, NAME); \
} else { \
/* anything else is fail */ \
fat_dispatch_ ## NAME = &JOIN(error_, NAME); \
} \
/* the rest of the function is completed in the CONNECT_ARGS_1 macro. */
#endif
#define CONNECT_ARGS_1(RTYPE, NAME, ...) \
return (*fat_dispatch_ ## NAME)(__VA_ARGS__); \
} \
#define CONNECT_DISPATCH_2(RTYPE, NAME, ...) \
/* new function */ \
HS_PUBLIC_API \
RTYPE NAME(__VA_ARGS__) { \
#define CONNECT_ARGS_3(RTYPE, NAME, ...) \
return (*fat_dispatch_ ## NAME)(__VA_ARGS__); \
} \
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wunused-parameter"
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wunused-function"
/* this gets a bit ugly to compose the static redirect functions,
* as we necessarily need first the typed arg list and then just the arg
* names, twice in a row, to define the redirect function and the
* dispatch function call */
CREATE_DISPATCH(hs_error_t, hs_scan, const hs_database_t *db, const char *data,
unsigned length, unsigned flags, hs_scratch_t *scratch,
match_event_handler onEvent, void *userCtx);
CONNECT_ARGS_1(hs_error_t, hs_scan, db, data, length, flags, scratch, onEvent, userCtx);
CONNECT_DISPATCH_2(hs_error_t, hs_scan, const hs_database_t *db, const char *data,
unsigned length, unsigned flags, hs_scratch_t *scratch,
match_event_handler onEvent, void *userCtx);
CONNECT_ARGS_3(hs_error_t, hs_scan, db, data, length, flags, scratch, onEvent, userCtx);
CREATE_DISPATCH(hs_error_t, hs_stream_size, const hs_database_t *database,
size_t *stream_size);
CONNECT_ARGS_1(hs_error_t, hs_stream_size, database, stream_size);
CONNECT_DISPATCH_2(hs_error_t, hs_stream_size, const hs_database_t *database,
size_t *stream_size);
CONNECT_ARGS_3(hs_error_t, hs_stream_size, database, stream_size);
CREATE_DISPATCH(hs_error_t, hs_database_size, const hs_database_t *db,
size_t *size);
CONNECT_ARGS_1(hs_error_t, hs_database_size, db, size);
CONNECT_DISPATCH_2(hs_error_t, hs_database_size, const hs_database_t *db,
size_t *size);
CONNECT_ARGS_3(hs_error_t, hs_database_size, db, size);
CREATE_DISPATCH(hs_error_t, dbIsValid, const hs_database_t *db);
CONNECT_ARGS_1(hs_error_t, dbIsValid, db);
CONNECT_DISPATCH_2(hs_error_t, dbIsValid, const hs_database_t *db);
CONNECT_ARGS_3(hs_error_t, dbIsValid, db);
CREATE_DISPATCH(hs_error_t, hs_free_database, hs_database_t *db);
CONNECT_ARGS_1(hs_error_t, hs_free_database, db);
CONNECT_DISPATCH_2(hs_error_t, hs_free_database, hs_database_t *db);
CONNECT_ARGS_3(hs_error_t, hs_free_database, db);
CREATE_DISPATCH(hs_error_t, hs_open_stream, const hs_database_t *db,
unsigned int flags, hs_stream_t **stream);
CONNECT_ARGS_1(hs_error_t, hs_open_stream, db, flags, stream);
CONNECT_DISPATCH_2(hs_error_t, hs_open_stream, const hs_database_t *db,
unsigned int flags, hs_stream_t **stream);
CONNECT_ARGS_3(hs_error_t, hs_open_stream, db, flags, stream);
CREATE_DISPATCH(hs_error_t, hs_scan_stream, hs_stream_t *id, const char *data,
unsigned int length, unsigned int flags, hs_scratch_t *scratch,
match_event_handler onEvent, void *ctxt);
CONNECT_ARGS_1(hs_error_t, hs_scan_stream, id, data, length, flags, scratch, onEvent, ctxt);
CONNECT_DISPATCH_2(hs_error_t, hs_scan_stream, hs_stream_t *id, const char *data,
unsigned int length, unsigned int flags, hs_scratch_t *scratch,
match_event_handler onEvent, void *ctxt);
CONNECT_ARGS_3(hs_error_t, hs_scan_stream, id, data, length, flags, scratch, onEvent, ctxt);
CREATE_DISPATCH(hs_error_t, hs_close_stream, hs_stream_t *id,
hs_scratch_t *scratch, match_event_handler onEvent, void *ctxt);
CONNECT_ARGS_1(hs_error_t, hs_close_stream, id, scratch, onEvent, ctxt);
CONNECT_DISPATCH_2(hs_error_t, hs_close_stream, hs_stream_t *id,
hs_scratch_t *scratch, match_event_handler onEvent, void *ctxt);
CONNECT_ARGS_3(hs_error_t, hs_close_stream, id, scratch, onEvent, ctxt);
CREATE_DISPATCH(hs_error_t, hs_scan_vector, const hs_database_t *db,
const char *const *data, const unsigned int *length,
unsigned int count, unsigned int flags, hs_scratch_t *scratch,
match_event_handler onevent, void *context);
CONNECT_ARGS_1(hs_error_t, hs_scan_vector, db, data, length, count, flags, scratch, onevent, context);
CONNECT_DISPATCH_2(hs_error_t, hs_scan_vector, const hs_database_t *db,
const char *const *data, const unsigned int *length,
unsigned int count, unsigned int flags, hs_scratch_t *scratch,
match_event_handler onevent, void *context);
CONNECT_ARGS_3(hs_error_t, hs_scan_vector, db, data, length, count, flags, scratch, onevent, context);
CREATE_DISPATCH(hs_error_t, hs_database_info, const hs_database_t *db, char **info);
CONNECT_ARGS_1(hs_error_t, hs_database_info, db, info);
CONNECT_DISPATCH_2(hs_error_t, hs_database_info, const hs_database_t *db, char **info);
CONNECT_ARGS_3(hs_error_t, hs_database_info, db, info);
CREATE_DISPATCH(hs_error_t, hs_copy_stream, hs_stream_t **to_id,
const hs_stream_t *from_id);
CONNECT_ARGS_1(hs_error_t, hs_copy_stream, to_id, from_id);
CONNECT_DISPATCH_2(hs_error_t, hs_copy_stream, hs_stream_t **to_id,
const hs_stream_t *from_id);
CONNECT_ARGS_3(hs_error_t, hs_copy_stream, to_id, from_id);
CREATE_DISPATCH(hs_error_t, hs_reset_stream, hs_stream_t *id,
unsigned int flags, hs_scratch_t *scratch,
match_event_handler onEvent, void *context);
CONNECT_ARGS_1(hs_error_t, hs_reset_stream, id, flags, scratch, onEvent, context);
CONNECT_DISPATCH_2(hs_error_t, hs_reset_stream, hs_stream_t *id,
unsigned int flags, hs_scratch_t *scratch,
match_event_handler onEvent, void *context);
CONNECT_ARGS_3(hs_error_t, hs_reset_stream, id, flags, scratch, onEvent, context);
CREATE_DISPATCH(hs_error_t, hs_reset_and_copy_stream, hs_stream_t *to_id,
const hs_stream_t *from_id, hs_scratch_t *scratch,
match_event_handler onEvent, void *context);
CONNECT_ARGS_1(hs_error_t, hs_reset_and_copy_stream, to_id, from_id, scratch, onEvent, context);
CONNECT_DISPATCH_2(hs_error_t, hs_reset_and_copy_stream, hs_stream_t *to_id,
const hs_stream_t *from_id, hs_scratch_t *scratch,
match_event_handler onEvent, void *context);
CONNECT_ARGS_3(hs_error_t, hs_reset_and_copy_stream, to_id, from_id, scratch, onEvent, context);
CREATE_DISPATCH(hs_error_t, hs_serialize_database, const hs_database_t *db,
char **bytes, size_t *length);
CONNECT_ARGS_1(hs_error_t, hs_serialize_database, db, bytes, length);
CONNECT_DISPATCH_2(hs_error_t, hs_serialize_database, const hs_database_t *db,
char **bytes, size_t *length);
CONNECT_ARGS_3(hs_error_t, hs_serialize_database, db, bytes, length);
CREATE_DISPATCH(hs_error_t, hs_deserialize_database, const char *bytes,
const size_t length, hs_database_t **db);
CONNECT_ARGS_1(hs_error_t, hs_deserialize_database, bytes, length, db);
CONNECT_DISPATCH_2(hs_error_t, hs_deserialize_database, const char *bytes,
const size_t length, hs_database_t **db);
CONNECT_ARGS_3(hs_error_t, hs_deserialize_database, bytes, length, db);
CREATE_DISPATCH(hs_error_t, hs_deserialize_database_at, const char *bytes,
const size_t length, hs_database_t *db);
CONNECT_ARGS_1(hs_error_t, hs_deserialize_database_at, bytes, length, db);
CONNECT_DISPATCH_2(hs_error_t, hs_deserialize_database_at, const char *bytes,
const size_t length, hs_database_t *db);
CONNECT_ARGS_3(hs_error_t, hs_deserialize_database_at, bytes, length, db);
CREATE_DISPATCH(hs_error_t, hs_serialized_database_info, const char *bytes,
size_t length, char **info);
CONNECT_ARGS_1(hs_error_t, hs_serialized_database_info, bytes, length, info);
CONNECT_DISPATCH_2(hs_error_t, hs_serialized_database_info, const char *bytes,
size_t length, char **info);
CONNECT_ARGS_3(hs_error_t, hs_serialized_database_info, bytes, length, info);
CREATE_DISPATCH(hs_error_t, hs_serialized_database_size, const char *bytes,
const size_t length, size_t *deserialized_size);
CONNECT_ARGS_1(hs_error_t, hs_serialized_database_size, bytes, length, deserialized_size);
CONNECT_DISPATCH_2(hs_error_t, hs_serialized_database_size, const char *bytes,
const size_t length, size_t *deserialized_size);
CONNECT_ARGS_3(hs_error_t, hs_serialized_database_size, bytes, length, deserialized_size);
CREATE_DISPATCH(hs_error_t, hs_compress_stream, const hs_stream_t *stream,
char *buf, size_t buf_space, size_t *used_space);
CONNECT_ARGS_1(hs_error_t, hs_compress_stream, stream,
buf, buf_space, used_space);
CONNECT_DISPATCH_2(hs_error_t, hs_compress_stream, const hs_stream_t *stream,
char *buf, size_t buf_space, size_t *used_space);
CONNECT_ARGS_3(hs_error_t, hs_compress_stream, stream,
buf, buf_space, used_space);
CREATE_DISPATCH(hs_error_t, hs_expand_stream, const hs_database_t *db,
hs_stream_t **stream, const char *buf,size_t buf_size);
CONNECT_ARGS_1(hs_error_t, hs_expand_stream, db, stream, buf,buf_size);
CONNECT_DISPATCH_2(hs_error_t, hs_expand_stream, const hs_database_t *db,
hs_stream_t **stream, const char *buf,size_t buf_size);
CONNECT_ARGS_3(hs_error_t, hs_expand_stream, db, stream, buf,buf_size);
CREATE_DISPATCH(hs_error_t, hs_reset_and_expand_stream, hs_stream_t *to_stream,
const char *buf, size_t buf_size, hs_scratch_t *scratch,
match_event_handler onEvent, void *context);
CONNECT_ARGS_1(hs_error_t, hs_reset_and_expand_stream, to_stream,
buf, buf_size, scratch, onEvent, context);
CONNECT_DISPATCH_2(hs_error_t, hs_reset_and_expand_stream, hs_stream_t *to_stream,
const char *buf, size_t buf_size, hs_scratch_t *scratch,
match_event_handler onEvent, void *context);
CONNECT_ARGS_3(hs_error_t, hs_reset_and_expand_stream, to_stream,
buf, buf_size, scratch, onEvent, context);
/** INTERNALS **/
CREATE_DISPATCH(u32, Crc32c_ComputeBuf, u32 inCrc32, const void *buf, size_t bufLen);
CONNECT_ARGS_1(u32, Crc32c_ComputeBuf, inCrc32, buf, bufLen);
CONNECT_DISPATCH_2(u32, Crc32c_ComputeBuf, u32 inCrc32, const void *buf, size_t bufLen);
CONNECT_ARGS_3(u32, Crc32c_ComputeBuf, inCrc32, buf, bufLen);
#pragma GCC diagnostic pop
#pragma GCC diagnostic pop

4
src/fdr/fdr.c
View File

@ -298,7 +298,7 @@ void get_conf_stride_4(const u8 *itPtr, UNUSED const u8 *start_ptr,
static really_inline
void do_confirm_fdr(u64a *conf, u8 offset, hwlmcb_rv_t *control,
const u32 *confBase, const struct FDR_Runtime_Args *a,
const u8 *ptr, u32 *last_match_id, struct zone *z) {
const u8 *ptr, u32 *last_match_id, const struct zone *z) {
const u8 bucket = 8;
if (likely(!*conf)) {
@ -333,7 +333,7 @@ void do_confirm_fdr(u64a *conf, u8 offset, hwlmcb_rv_t *control,
}
static really_inline
void dumpZoneInfo(UNUSED struct zone *z, UNUSED size_t zone_id) {
void dumpZoneInfo(UNUSED const struct zone *z, UNUSED size_t zone_id) {
#ifdef DEBUG
DEBUG_PRINTF("zone: zone=%zu, bufPtr=%p\n", zone_id, z->buf);
DEBUG_PRINTF("zone: startPtr=%p, endPtr=%p, shift=%u\n",

4
src/fdr/fdr_compile.cpp
View File

@ -127,7 +127,7 @@ void andMask(u8 *dest, const u8 *a, const u8 *b, u32 num_bytes) {
}
void FDRCompiler::createInitialState(FDR *fdr) {
u8 *start = (u8 *)&fdr->start;
u8 *start = reinterpret_cast<u8 *>(&fdr->start);
/* initial state should to be 1 in each slot in the bucket up to bucket
* minlen - 1, and 0 thereafter */
@ -176,7 +176,7 @@ bytecode_ptr<FDR> FDRCompiler::setupFDR() {
auto fdr = make_zeroed_bytecode_ptr<FDR>(size, 64);
assert(fdr); // otherwise would have thrown std::bad_alloc
u8 *fdr_base = (u8 *)fdr.get();
u8 *fdr_base = reinterpret_cast<u8 *>(fdr.get());
// Write header.
fdr->size = size;

15
src/fdr/fdr_confirm_compile.cpp
View File

@ -58,7 +58,7 @@ u64a make_u64a_mask(const vector<u8> &v) {
u64a mask = 0;
size_t vlen = v.size();
size_t len = std::min(vlen, sizeof(mask));
unsigned char *m = (unsigned char *)&mask;
u8 *m = reinterpret_cast<u8 *>(&mask);
memcpy(m + sizeof(mask) - len, &v[vlen - len], len);
return mask;
}
@ -159,7 +159,7 @@ bytecode_ptr<FDRConfirm> getFDRConfirm(const vector<hwlmLiteral> &lits,
map<u32, vector<LiteralIndex> > res2lits;
hwlm_group_t gm = 0;
for (LiteralIndex i = 0; i < lits.size(); i++) {
LitInfo & li = tmpLitInfo[i];
const LitInfo & li = tmpLitInfo[i];
u32 hash = CONF_HASH_CALL(li.v, andmsk, mult, nBits);
DEBUG_PRINTF("%016llx --> %u\n", li.v, hash);
res2lits[hash].emplace_back(i);
@ -245,10 +245,10 @@ bytecode_ptr<FDRConfirm> getFDRConfirm(const vector<hwlmLiteral> &lits,
fdrc->groups = gm;
// After the FDRConfirm, we have the lit index array.
u8 *fdrc_base = (u8 *)fdrc.get();
u8 *fdrc_base = reinterpret_cast<u8 *>(fdrc.get());
u8 *ptr = fdrc_base + sizeof(*fdrc);
ptr = ROUNDUP_PTR(ptr, alignof(u32));
u32 *bitsToLitIndex = (u32 *)ptr;
u32 *bitsToLitIndex = reinterpret_cast<u32 *>(ptr);
ptr += bitsToLitIndexSize;
// After the lit index array, we have the LitInfo structures themselves,
@ -265,7 +265,7 @@ bytecode_ptr<FDRConfirm> getFDRConfirm(const vector<hwlmLiteral> &lits,
LiteralIndex litIdx = *i;
// Write LitInfo header.
LitInfo &finalLI = *(LitInfo *)ptr;
LitInfo &finalLI = *(reinterpret_cast<LitInfo *>(ptr));
finalLI = tmpLitInfo[litIdx];
ptr += sizeof(LitInfo); // String starts directly after LitInfo.
@ -294,9 +294,6 @@ setupFullConfs(const vector<hwlmLiteral> &lits,
const EngineDescription &eng,
const map<BucketIndex, vector<LiteralIndex>> &bucketToLits,
bool make_small) {
unique_ptr<TeddyEngineDescription> teddyDescr =
getTeddyDescription(eng.getID());
BC2CONF bc2Conf;
u32 totalConfirmSize = 0;
for (BucketIndex b = 0; b < eng.getNumBuckets(); b++) {
@ -321,7 +318,7 @@ setupFullConfs(const vector<hwlmLiteral> &lits,
auto buf = make_zeroed_bytecode_ptr<u8>(totalSize, 64);
assert(buf); // otherwise would have thrown std::bad_alloc
u32 *confBase = (u32 *)buf.get();
u32 *confBase = reinterpret_cast<u32 *>(buf.get());
u8 *ptr = buf.get() + totalConfSwitchSize;
assert(ISALIGNED_CL(ptr));

2
src/fdr/fdr_engine_description.cpp
View File

@ -71,7 +71,7 @@ u32 findDesiredStride(size_t num_lits, size_t min_len, size_t min_len_count) {
} else if (num_lits < 5000) {
// for larger but not huge sizes, go to stride 2 only if we have at
// least minlen 3
desiredStride = MIN(min_len - 1, 2);
desiredStride = std::min(min_len - 1, 2UL);
}
}

4
src/fdr/flood_compile.cpp
View File

@ -208,8 +208,8 @@ bytecode_ptr<u8> setupFDRFloodControl(const vector<hwlmLiteral> &lits,
auto buf = make_zeroed_bytecode_ptr<u8>(totalSize, 16);
assert(buf); // otherwise would have thrown std::bad_alloc
u32 *floodHeader = (u32 *)buf.get();
FDRFlood *layoutFlood = (FDRFlood *)(buf.get() + floodHeaderSize);
u32 *floodHeader = reinterpret_cast<u32 *>(buf.get());
FDRFlood *layoutFlood = reinterpret_cast<FDRFlood *>(buf.get() + floodHeaderSize);
u32 currentFloodIndex = 0;
for (const auto &m : flood2chars) {

8
src/fdr/teddy_compile.cpp
View File

@ -328,7 +328,7 @@ bool pack(const vector<hwlmLiteral> &lits,
static
void initReinforcedTable(u8 *rmsk) {
u64a *mask = (u64a *)rmsk;
u64a *mask = reinterpret_cast<u64a *>(rmsk);
fill_n(mask, N_CHARS, 0x00ffffffffffffffULL);
}
@ -576,8 +576,8 @@ bytecode_ptr<FDR> TeddyCompiler::build() {
auto fdr = make_zeroed_bytecode_ptr<FDR>(size, 64);
assert(fdr); // otherwise would have thrown std::bad_alloc
Teddy *teddy = (Teddy *)fdr.get(); // ugly
u8 *teddy_base = (u8 *)teddy;
Teddy *teddy = reinterpret_cast<Teddy *>(fdr.get()); // ugly
u8 *teddy_base = reinterpret_cast<u8 *>(teddy);
// Write header.
teddy->size = size;
@ -622,7 +622,7 @@ bytecode_ptr<FDR> TeddyCompiler::build() {
static
bool assignStringsToBuckets(
const vector<hwlmLiteral> &lits,
TeddyEngineDescription &eng,
const TeddyEngineDescription &eng,
map<BucketIndex, vector<LiteralIndex>> &bucketToLits) {
assert(eng.numMasks <= MAX_NUM_MASKS);
if (lits.size() > eng.getNumBuckets() * TEDDY_BUCKET_LOAD) {

16
src/fdr/teddy_engine_description.cpp
View File

@ -52,14 +52,14 @@ u32 TeddyEngineDescription::getDefaultFloodSuffixLength() const {
void getTeddyDescriptions(vector<TeddyEngineDescription> *out) {
static const TeddyEngineDef defns[] = {
{ 3, 0 | HS_CPU_FEATURES_AVX2, 1, 16, false },
{ 4, 0 | HS_CPU_FEATURES_AVX2, 1, 16, true },
{ 5, 0 | HS_CPU_FEATURES_AVX2, 2, 16, false },
{ 6, 0 | HS_CPU_FEATURES_AVX2, 2, 16, true },
{ 7, 0 | HS_CPU_FEATURES_AVX2, 3, 16, false },
{ 8, 0 | HS_CPU_FEATURES_AVX2, 3, 16, true },
{ 9, 0 | HS_CPU_FEATURES_AVX2, 4, 16, false },
{ 10, 0 | HS_CPU_FEATURES_AVX2, 4, 16, true },
{ 3, HS_CPU_FEATURES_AVX2, 1, 16, false },
{ 4, HS_CPU_FEATURES_AVX2, 1, 16, true },
{ 5, HS_CPU_FEATURES_AVX2, 2, 16, false },
{ 6, HS_CPU_FEATURES_AVX2, 2, 16, true },
{ 7, HS_CPU_FEATURES_AVX2, 3, 16, false },
{ 8, HS_CPU_FEATURES_AVX2, 3, 16, true },
{ 9, HS_CPU_FEATURES_AVX2, 4, 16, false },
{ 10, HS_CPU_FEATURES_AVX2, 4, 16, true },
{ 11, 0, 1, 8, false },
{ 12, 0, 1, 8, true },
{ 13, 0, 2, 8, false },

2
src/hs.cpp
View File

@ -589,7 +589,7 @@ hs_error_t hs_expression_info_int(const char *expression, unsigned int flags,
return HS_COMPILER_ERROR;
}
hs_expr_info *rv = (hs_expr_info *)hs_misc_alloc(sizeof(*rv));
hs_expr_info *rv = static_cast<hs_expr_info *>(hs_misc_alloc(sizeof(*rv)));
if (!rv) {
*error = const_cast<hs_compile_error_t *>(&hs_enomem);
return HS_COMPILER_ERROR;

2
src/hs_valid_platform.c
View File

@ -48,6 +48,8 @@ hs_error_t HS_CDECL hs_valid_platform(void) {
return HS_ARCH_ERROR;
}
#elif !defined(VS_SIMDE_BACKEND) && (defined(ARCH_ARM32) || defined(ARCH_AARCH64))
//check_neon returns true for now
// cppcheck-suppress knownConditionTrueFalse
if (check_neon()) {
return HS_SUCCESS;
} else {

3
src/hwlm/hwlm.c
View File

@ -170,8 +170,7 @@ void do_accel_streaming(const union AccelAux *aux, const u8 *hbuf, size_t hlen,
DEBUG_PRINTF("got %zu/%zu in 2nd buffer\n", delta, len);
*start += delta;
} else if (hlen) {
UNUSED size_t remaining = offset + ptr2 - found;
DEBUG_PRINTF("got %zu/%zu remaining in 1st buffer\n", remaining, hlen);
DEBUG_PRINTF("got %zu/%zu remaining in 1st buffer\n", offset + ptr2 - found, hlen);
}
}

9
src/hwlm/hwlm_build.cpp
View File

@ -143,7 +143,7 @@ bytecode_ptr<HWLM> hwlmBuild(const HWLMProto &proto, const CompileContext &cc,
}
if (!eng) {
return nullptr;
return bytecode_ptr<HWLM>(nullptr);
}
assert(engSize);
@ -155,6 +155,7 @@ bytecode_ptr<HWLM> hwlmBuild(const HWLMProto &proto, const CompileContext &cc,
auto h = make_zeroed_bytecode_ptr<HWLM>(hwlm_len, 64);
h->type = proto.engType;
// cppcheck-suppress cstyleCast
memcpy(HWLM_DATA(h.get()), eng.get(), engSize);
return h;
@ -218,10 +219,12 @@ size_t hwlmSize(const HWLM *h) {
switch (h->type) {
case HWLM_ENGINE_NOOD:
engSize = noodSize((const noodTable *)HWLM_C_DATA(h));
// cppcheck-suppress cstyleCast
engSize = noodSize(reinterpret_cast<const noodTable *>(HWLM_C_DATA(h)));
break;
case HWLM_ENGINE_FDR:
engSize = fdrSize((const FDR *)HWLM_C_DATA(h));
// cppcheck-suppress cstyleCast
engSize = fdrSize(reinterpret_cast<const FDR *>(HWLM_C_DATA(h)));
break;
}

2
src/hwlm/noodle_build.cpp
View File

@ -56,7 +56,7 @@ u64a make_u64a_mask(const vector<u8> &v) {
u64a mask = 0;
size_t len = v.size();
unsigned char *m = (unsigned char *)&mask;
u8 *m = reinterpret_cast<u8 *>(&mask);
DEBUG_PRINTF("making mask len %zu\n", len);
memcpy(m, &v[0], len);
return mask;

35
src/nfa/accel_dfa_build_strat.cpp
View File

@ -427,7 +427,7 @@ void
accel_dfa_build_strat::buildAccel(UNUSED dstate_id_t this_idx,
const AccelScheme &info,
void *accel_out) {
AccelAux *accel = (AccelAux *)accel_out;
AccelAux *accel = reinterpret_cast<AccelAux *>(accel_out);
DEBUG_PRINTF("accelerations scheme has offset s%u/d%u\n", info.offset,
info.double_offset);
@ -474,7 +474,8 @@ accel_dfa_build_strat::buildAccel(UNUSED dstate_id_t this_idx,
u8 c1 = info.double_byte.begin()->first & m1;
u8 c2 = info.double_byte.begin()->second & m2;
#ifdef HAVE_SVE2
if (vermicelliDoubleMasked16Build(c1, c2, m1, m2, (u8 *)&accel->mdverm16.mask)) {
if (vermicelliDoubleMasked16Build(c1, c2, m1, m2,
reinterpret_cast<u8 *>(&accel->mdverm16.mask))) {
accel->accel_type = ACCEL_DVERM16_MASKED;
accel->mdverm16.offset = verify_u8(info.double_offset);
accel->mdverm16.c1 = c1;
@ -483,8 +484,9 @@ accel_dfa_build_strat::buildAccel(UNUSED dstate_id_t this_idx,
c1, c2);
return;
} else if (info.double_byte.size() <= 8 &&
vermicelliDouble16Build(info.double_byte, (u8 *)&accel->dverm16.mask,
(u8 *)&accel->dverm16.firsts)) {
vermicelliDouble16Build(info.double_byte,
reinterpret_cast<u8 *>(&accel->dverm16.mask),
reinterpret_cast<u8 *>(&accel->dverm16.firsts))) {
accel->accel_type = ACCEL_DVERM16;
accel->dverm16.offset = verify_u8(info.double_offset);
DEBUG_PRINTF("building double16-vermicelli\n");
@ -504,8 +506,9 @@ accel_dfa_build_strat::buildAccel(UNUSED dstate_id_t this_idx,
}
#ifdef HAVE_SVE2
if (info.double_byte.size() <= 8 &&
vermicelliDouble16Build(info.double_byte, (u8 *)&accel->dverm16.mask,
(u8 *)&accel->dverm16.firsts)) {
vermicelliDouble16Build(info.double_byte,
reinterpret_cast<u8 *>(&accel->dverm16.mask),
reinterpret_cast<u8 *>(&accel->dverm16.firsts))) {
accel->accel_type = ACCEL_DVERM16;
accel->dverm16.offset = verify_u8(info.double_offset);
DEBUG_PRINTF("building double16-vermicelli\n");
@ -516,9 +519,11 @@ accel_dfa_build_strat::buildAccel(UNUSED dstate_id_t this_idx,
if (double_byte_ok(info) &&
shuftiBuildDoubleMasks(
info.double_cr, info.double_byte, (u8 *)&accel->dshufti.lo1,
(u8 *)&accel->dshufti.hi1, (u8 *)&accel->dshufti.lo2,
(u8 *)&accel->dshufti.hi2)) {
info.double_cr, info.double_byte,
reinterpret_cast<u8 *>(&accel->dshufti.lo1),
reinterpret_cast<u8 *>(&accel->dshufti.hi1),
reinterpret_cast<u8 *>(&accel->dshufti.lo2),
reinterpret_cast<u8 *>(&accel->dshufti.hi2))) {
accel->accel_type = ACCEL_DSHUFTI;
accel->dshufti.offset = verify_u8(info.double_offset);
DEBUG_PRINTF("state %hu is double shufti\n", this_idx);
@ -550,7 +555,7 @@ accel_dfa_build_strat::buildAccel(UNUSED dstate_id_t this_idx,
#ifdef HAVE_SVE2
if (info.cr.count() <= 16) {
accel->accel_type = ACCEL_VERM16;
vermicelli16Build(info.cr, (u8 *)&accel->verm16.mask);
vermicelli16Build(info.cr, reinterpret_cast<u8 *>(&accel->verm16.mask));
DEBUG_PRINTF("state %hu is vermicelli16\n", this_idx);
return;
}
@ -563,16 +568,18 @@ accel_dfa_build_strat::buildAccel(UNUSED dstate_id_t this_idx,
}
accel->accel_type = ACCEL_SHUFTI;
if (-1 != shuftiBuildMasks(info.cr, (u8 *)&accel->shufti.lo,
(u8 *)&accel->shufti.hi)) {
if (-1 != shuftiBuildMasks(info.cr,
reinterpret_cast<u8 *>(&accel->shufti.lo),
reinterpret_cast<u8 *>(&accel->shufti.hi))) {
DEBUG_PRINTF("state %hu is shufti\n", this_idx);
return;
}
assert(!info.cr.none());
accel->accel_type = ACCEL_TRUFFLE;
truffleBuildMasks(info.cr, (u8 *)&accel->truffle.mask1,
(u8 *)&accel->truffle.mask2);
truffleBuildMasks(info.cr,
reinterpret_cast<u8 *>(&accel->truffle.mask1),
reinterpret_cast<u8 *>(&accel->truffle.mask2));
DEBUG_PRINTF("state %hu is truffle\n", this_idx);
}

23
src/nfa/accelcompile.cpp
View File

@ -84,8 +84,9 @@ void buildAccelSingle(const AccelInfo &info, AccelAux *aux) {
#endif
DEBUG_PRINTF("attempting shufti for %zu chars\n", outs);
if (-1 != shuftiBuildMasks(info.single_stops, (u8 *)&aux->shufti.lo,
(u8 *)&aux->shufti.hi)) {
if (-1 != shuftiBuildMasks(info.single_stops,
reinterpret_cast<u8 *>(&aux->shufti.lo),
reinterpret_cast<u8 *>(&aux->shufti.hi))) {
aux->accel_type = ACCEL_SHUFTI;
aux->shufti.offset = offset;
DEBUG_PRINTF("shufti built OK\n");
@ -98,8 +99,9 @@ void buildAccelSingle(const AccelInfo &info, AccelAux *aux) {
DEBUG_PRINTF("building Truffle for %zu chars\n", outs);
aux->accel_type = ACCEL_TRUFFLE;
aux->truffle.offset = offset;
truffleBuildMasks(info.single_stops, (u8 *)&aux->truffle.mask1,
(u8 *)&aux->truffle.mask2);
truffleBuildMasks(info.single_stops,
reinterpret_cast<u8 *>(&aux->truffle.mask1),
reinterpret_cast<u8 *>(&aux->truffle.mask2));
return;
}
@ -219,8 +221,9 @@ void buildAccelDouble(const AccelInfo &info, AccelAux *aux) {
c1, c2);
return;
} else if (outs2 <= 8 &&
vermicelliDouble16Build(info.double_stop2, (u8 *)&aux->dverm16.mask,
(u8 *)&aux->dverm16.firsts)) {
vermicelliDouble16Build(info.double_stop2,
reinterpret_cast<u8 *>(&aux->dverm16.mask),
reinterpret_cast<u8 *>(&aux->dverm16.firsts))) {
aux->accel_type = ACCEL_DVERM16;
aux->dverm16.offset = offset;
DEBUG_PRINTF("building double16-vermicelli\n");
@ -254,9 +257,11 @@ void buildAccelDouble(const AccelInfo &info, AccelAux *aux) {
aux->accel_type = ACCEL_DSHUFTI;
aux->dshufti.offset = offset;
if (shuftiBuildDoubleMasks(
info.double_stop1, info.double_stop2, (u8 *)&aux->dshufti.lo1,
(u8 *)&aux->dshufti.hi1, (u8 *)&aux->dshufti.lo2,
(u8 *)&aux->dshufti.hi2)) {
info.double_stop1, info.double_stop2,
reinterpret_cast<u8 *>(&aux->dshufti.lo1),
reinterpret_cast<u8 *>(&aux->dshufti.hi1),
reinterpret_cast<u8 *>(&aux->dshufti.lo2),
reinterpret_cast<u8 *>(&aux->dshufti.hi2))) {
return;
}
}

62
src/nfa/castle.c
View File

@ -94,8 +94,8 @@ char subCastleReportCurrent(const struct Castle *c, struct mq *q,
const struct SubCastle *sub = getSubCastle(c, subIdx);
const struct RepeatInfo *info = getRepeatInfo(sub);
union RepeatControl *rctrl = getControl(q->state, sub);
char *rstate = (char *)q->streamState + sub->streamStateOffset +
const union RepeatControl *rctrl = getControl(q->state, sub);
const char *rstate = (char *)q->streamState + sub->streamStateOffset +
info->packedCtrlSize;
enum RepeatMatch match =
repeatHasMatch(info, rctrl, rstate, offset);
@ -118,10 +118,10 @@ int castleReportCurrent(const struct Castle *c, struct mq *q) {
if (c->exclusive) {
u8 *active = (u8 *)q->streamState;
u8 *groups = active + c->groupIterOffset;
const u8 *groups = active + c->groupIterOffset;
for (u32 i = mmbit_iterate(groups, c->numGroups, MMB_INVALID);
i != MMB_INVALID; i = mmbit_iterate(groups, c->numGroups, i)) {
u8 *cur = active + i * c->activeIdxSize;
const u8 *cur = active + i * c->activeIdxSize;
const u32 activeIdx = partial_load_u32(cur, c->activeIdxSize);
DEBUG_PRINTF("subcastle %u\n", activeIdx);
if (subCastleReportCurrent(c, q,
@ -156,8 +156,8 @@ char subCastleInAccept(const struct Castle *c, struct mq *q,
}
const struct RepeatInfo *info = getRepeatInfo(sub);
union RepeatControl *rctrl = getControl(q->state, sub);
char *rstate = (char *)q->streamState + sub->streamStateOffset +
const union RepeatControl *rctrl = getControl(q->state, sub);
const char *rstate = (char *)q->streamState + sub->streamStateOffset +
info->packedCtrlSize;
enum RepeatMatch match =
repeatHasMatch(info, rctrl, rstate, offset);
@ -180,10 +180,10 @@ char castleInAccept(const struct Castle *c, struct mq *q,
if (c->exclusive) {
u8 *active = (u8 *)q->streamState;
u8 *groups = active + c->groupIterOffset;
const u8 *groups = active + c->groupIterOffset;
for (u32 i = mmbit_iterate(groups, c->numGroups, MMB_INVALID);
i != MMB_INVALID; i = mmbit_iterate(groups, c->numGroups, i)) {
u8 *cur = active + i * c->activeIdxSize;
const u8 *cur = active + i * c->activeIdxSize;
const u32 activeIdx = partial_load_u32(cur, c->activeIdxSize);
DEBUG_PRINTF("subcastle %u\n", activeIdx);
if (subCastleInAccept(c, q, report, offset, activeIdx)) {
@ -213,8 +213,8 @@ void subCastleDeactivateStaleSubs(const struct Castle *c, const u64a offset,
const struct SubCastle *sub = getSubCastle(c, subIdx);
const struct RepeatInfo *info = getRepeatInfo(sub);
union RepeatControl *rctrl = getControl(full_state, sub);
char *rstate = (char *)stream_state + sub->streamStateOffset +
const union RepeatControl *rctrl = getControl(full_state, sub);
const char *rstate = (char *)stream_state + sub->streamStateOffset +
info->packedCtrlSize;
if (repeatHasMatch(info, rctrl, rstate, offset) == REPEAT_STALE) {
@ -242,10 +242,10 @@ void castleDeactivateStaleSubs(const struct Castle *c, const u64a offset,
if (c->exclusive) {
u8 *active = (u8 *)stream_state;
u8 *groups = active + c->groupIterOffset;
const u8 *groups = active + c->groupIterOffset;
for (u32 i = mmbit_iterate(groups, c->numGroups, MMB_INVALID);
i != MMB_INVALID; i = mmbit_iterate(groups, c->numGroups, i)) {
u8 *cur = active + i * c->activeIdxSize;
const u8 *cur = active + i * c->activeIdxSize;
const u32 activeIdx = partial_load_u32(cur, c->activeIdxSize);
DEBUG_PRINTF("subcastle %u\n", activeIdx);
subCastleDeactivateStaleSubs(c, offset, full_state,
@ -329,8 +329,8 @@ void subCastleFindMatch(const struct Castle *c, const u64a begin,
size_t *mloc, char *found, const u32 subIdx) {
const struct SubCastle *sub = getSubCastle(c, subIdx);
const struct RepeatInfo *info = getRepeatInfo(sub);
union RepeatControl *rctrl = getControl(full_state, sub);
char *rstate = (char *)stream_state + sub->streamStateOffset +
const union RepeatControl *rctrl = getControl(full_state, sub);
const char *rstate = (char *)stream_state + sub->streamStateOffset +
info->packedCtrlSize;
u64a match = repeatNextMatch(info, rctrl, rstate, begin);
@ -374,10 +374,10 @@ char castleFindMatch(const struct Castle *c, const u64a begin, const u64a end,
if (c->exclusive) {
u8 *active = (u8 *)stream_state;
u8 *groups = active + c->groupIterOffset;
const u8 *groups = active + c->groupIterOffset;
for (u32 i = mmbit_iterate(groups, c->numGroups, MMB_INVALID);
i != MMB_INVALID; i = mmbit_iterate(groups, c->numGroups, i)) {
u8 *cur = active + i * c->activeIdxSize;
const u8 *cur = active + i * c->activeIdxSize;
const u32 activeIdx = partial_load_u32(cur, c->activeIdxSize);
DEBUG_PRINTF("subcastle %u\n", activeIdx);
subCastleFindMatch(c, begin, end, full_state, stream_state, mloc,
@ -386,7 +386,7 @@ char castleFindMatch(const struct Castle *c, const u64a begin, const u64a end,
}
if (c->exclusive != PURE_EXCLUSIVE) {
u8 *active = (u8 *)stream_state + c->activeOffset;
const u8 *active = (u8 *)stream_state + c->activeOffset;
for (u32 i = mmbit_iterate(active, c->numRepeats, MMB_INVALID);
i != MMB_INVALID;
i = mmbit_iterate(active, c->numRepeats, i)) {
@ -400,8 +400,8 @@ char castleFindMatch(const struct Castle *c, const u64a begin, const u64a end,
}
static really_inline
u64a subCastleNextMatch(const struct Castle *c, void *full_state,
void *stream_state, const u64a loc,
u64a subCastleNextMatch(const struct Castle *c, const void *full_state,
const void *stream_state, const u64a loc,
const u32 subIdx) {
DEBUG_PRINTF("subcastle %u\n", subIdx);
const struct SubCastle *sub = getSubCastle(c, subIdx);
@ -489,15 +489,14 @@ char castleMatchLoop(const struct Castle *c, const u64a begin, const u64a end,
// full_state (scratch).
u64a offset = end; // min offset of next match
u32 activeIdx = 0;
mmbit_clear(matching, c->numRepeats);
if (c->exclusive) {
u8 *active = (u8 *)stream_state;
u8 *groups = active + c->groupIterOffset;
for (u32 i = mmbit_iterate(groups, c->numGroups, MMB_INVALID);
i != MMB_INVALID; i = mmbit_iterate(groups, c->numGroups, i)) {
u8 *cur = active + i * c->activeIdxSize;
activeIdx = partial_load_u32(cur, c->activeIdxSize);
const u8 *cur = active + i * c->activeIdxSize;
u32 activeIdx = partial_load_u32(cur, c->activeIdxSize);
u64a match = subCastleNextMatch(c, full_state, stream_state,
loc, activeIdx);
set_matching(c, match, groups, matching, c->numGroups, i,
@ -797,7 +796,7 @@ char nfaExecCastle_Q_i(const struct NFA *n, struct mq *q, s64a end,
char found = 0;
if (c->exclusive) {
u8 *groups = (u8 *)q->streamState + c->groupIterOffset;
const u8 *groups = (u8 *)q->streamState + c->groupIterOffset;
found = mmbit_any(groups, c->numGroups);
}
@ -864,7 +863,7 @@ char nfaExecCastle_Q_i(const struct NFA *n, struct mq *q, s64a end,
}
if (c->exclusive) {
u8 *groups = (u8 *)q->streamState + c->groupIterOffset;
const u8 *groups = (u8 *)q->streamState + c->groupIterOffset;
if (mmbit_any_precise(groups, c->numGroups)) {
return 1;
}
@ -884,7 +883,7 @@ char nfaExecCastle_Q2(const struct NFA *n, struct mq *q, s64a end) {
}
static
s64a castleLastKillLoc(const struct Castle *c, struct mq *q) {
s64a castleLastKillLoc(const struct Castle *c, const struct mq *q) {
assert(q_cur_type(q) == MQE_START);
assert(q_last_type(q) == MQE_END);
s64a sp = q_cur_loc(q);
@ -907,7 +906,6 @@ s64a castleLastKillLoc(const struct Castle *c, struct mq *q) {
if (castleRevScan(c, q->history, sp + hlen, ep + hlen, &loc)) {
return (s64a)loc - hlen;
}
ep = 0;
}
return sp - 1; /* the repeats are never killed */
@ -959,7 +957,7 @@ char nfaExecCastle_QR(const struct NFA *n, struct mq *q, ReportID report) {
char found = 0;
if (c->exclusive) {
u8 *groups = (u8 *)q->streamState + c->groupIterOffset;
const u8 *groups = (u8 *)q->streamState + c->groupIterOffset;
found = mmbit_any_precise(groups, c->numGroups);
}
@ -1007,10 +1005,10 @@ char nfaExecCastle_inAnyAccept(const struct NFA *n, struct mq *q) {
if (c->exclusive) {
u8 *active = (u8 *)q->streamState;
u8 *groups = active + c->groupIterOffset;
const u8 *groups = active + c->groupIterOffset;
for (u32 i = mmbit_iterate(groups, c->numGroups, MMB_INVALID);
i != MMB_INVALID; i = mmbit_iterate(groups, c->numGroups, i)) {
u8 *cur = active + i * c->activeIdxSize;
const u8 *cur = active + i * c->activeIdxSize;
const u32 activeIdx = partial_load_u32(cur, c->activeIdxSize);
DEBUG_PRINTF("subcastle %u\n", activeIdx);
const struct SubCastle *sub = getSubCastle(c, activeIdx);
@ -1079,7 +1077,7 @@ void subCastleQueueCompressState(const struct Castle *c, const u32 subIdx,
const struct mq *q, const u64a offset) {
const struct SubCastle *sub = getSubCastle(c, subIdx);
const struct RepeatInfo *info = getRepeatInfo(sub);
union RepeatControl *rctrl = getControl(q->state, sub);
const union RepeatControl *rctrl = getControl(q->state, sub);
char *packed = (char *)q->streamState + sub->streamStateOffset;
DEBUG_PRINTF("sub %u next match %llu\n", subIdx,
repeatNextMatch(info, rctrl,
@ -1100,10 +1098,10 @@ char nfaExecCastle_queueCompressState(const struct NFA *n, const struct mq *q,
DEBUG_PRINTF("offset=%llu\n", offset);
if (c->exclusive) {
u8 *active = (u8 *)q->streamState;
u8 *groups = active + c->groupIterOffset;
const u8 *groups = active + c->groupIterOffset;
for (u32 i = mmbit_iterate(groups, c->numGroups, MMB_INVALID);
i != MMB_INVALID; i = mmbit_iterate(groups, c->numGroups, i)) {
u8 *cur = active + i * c->activeIdxSize;
const u8 *cur = active + i * c->activeIdxSize;
const u32 activeIdx = partial_load_u32(cur, c->activeIdxSize);
DEBUG_PRINTF("packing state for sub %u\n", activeIdx);
subCastleQueueCompressState(c, activeIdx, q, offset);

57
src/nfa/castlecompile.cpp
View File

@ -106,25 +106,27 @@ void writeCastleScanEngine(const CharReach &cr, Castle *c) {
#ifdef HAVE_SVE2
if (cr.count() <= 16) {
c->type = CASTLE_NVERM16;
vermicelli16Build(cr, (u8 *)&c->u.verm16.mask);
vermicelli16Build(cr, reinterpret_cast<u8 *>(&c->u.verm16.mask));
return;
}
if (negated.count() <= 16) {
c->type = CASTLE_VERM16;
vermicelli16Build(negated, (u8 *)&c->u.verm16.mask);
vermicelli16Build(negated, reinterpret_cast<u8 *>(&c->u.verm16.mask));
return;
}
#endif // HAVE_SVE2
if (shuftiBuildMasks(negated, (u8 *)&c->u.shuf.mask_lo,
(u8 *)&c->u.shuf.mask_hi) != -1) {
if (shuftiBuildMasks(negated,
reinterpret_cast<u8 *>(&c->u.shuf.mask_lo),
reinterpret_cast<u8 *>(&c->u.shuf.mask_hi)) != -1) {
c->type = CASTLE_SHUFTI;
return;
}
c->type = CASTLE_TRUFFLE;
truffleBuildMasks(negated, (u8 *)(u8 *)&c->u.truffle.mask1,
(u8 *)&c->u.truffle.mask2);
truffleBuildMasks(negated,
reinterpret_cast<u8 *>(&c->u.truffle.mask1),
reinterpret_cast<u8 *>(&c->u.truffle.mask2));
}
static
@ -227,11 +229,13 @@ vector<u32> removeClique(CliqueGraph &cg) {
while (!graph_empty(cg)) {
const vector<u32> &c = cliquesVec.back();
vector<CliqueVertex> dead;
for (const auto &v : vertices_range(cg)) {
if (find(c.begin(), c.end(), cg[v].stateId) != c.end()) {
dead.emplace_back(v);
}
}
auto deads = [&c=c, &cg=cg](const CliqueVertex &v) {
return (find(c.begin(), c.end(), cg[v].stateId) != c.end());
};
const auto &vr = vertices_range(cg);
std::copy_if(begin(vr), end(vr), std::back_inserter(dead), deads);
for (const auto &v : dead) {
clear_vertex(v, cg);
remove_vertex(v, cg);
@ -294,7 +298,7 @@ vector<vector<u32>> checkExclusion(u32 &streamStateSize,
size_t lower = 0;
size_t total = 0;
while (lower < trigSize) {
vector<CliqueVertex> vertices;
vector<CliqueVertex> clvertices;
unique_ptr<CliqueGraph> cg = make_unique<CliqueGraph>();
vector<vector<size_t>> min_reset_dist;
@ -302,7 +306,7 @@ vector<vector<u32>> checkExclusion(u32 &streamStateSize,
// get min reset distance for each repeat
for (size_t i = lower; i < upper; i++) {
CliqueVertex v = add_vertex(CliqueVertexProps(i), *cg);
vertices.emplace_back(v);
clvertices.emplace_back(v);
const vector<size_t> &tmp_dist =
minResetDistToEnd(triggers[i], cr);
@ -311,11 +315,11 @@ vector<vector<u32>> checkExclusion(u32 &streamStateSize,
// find exclusive pair for each repeat
for (size_t i = lower; i < upper; i++) {
CliqueVertex s = vertices[i - lower];
CliqueVertex s = clvertices[i - lower];
for (size_t j = i + 1; j < upper; j++) {
if (findExclusivePair(i, j, lower, min_reset_dist,
triggers)) {
CliqueVertex d = vertices[j - lower];
CliqueVertex d = clvertices[j - lower];
add_edge(s, d, *cg);
}
}
@ -600,9 +604,9 @@ buildCastle(const CastleProto &proto,
nfa->minWidth = verify_u32(minWidth);
nfa->maxWidth = maxWidth.is_finite() ? verify_u32(maxWidth) : 0;
char * const base_ptr = (char *)nfa.get() + sizeof(NFA);
char * const base_ptr = reinterpret_cast<char *>(nfa.get()) + sizeof(NFA);
char *ptr = base_ptr;
Castle *c = (Castle *)ptr;
Castle *c = reinterpret_cast<Castle *>(ptr);
c->numRepeats = verify_u32(subs.size());
c->numGroups = exclusiveInfo.numGroups;
c->exclusive = verify_s8(exclusive);
@ -613,7 +617,7 @@ buildCastle(const CastleProto &proto,
writeCastleScanEngine(cr, c);
ptr += sizeof(Castle);
SubCastle *subCastles = ((SubCastle *)(ROUNDUP_PTR(ptr, alignof(u32))));
SubCastle *subCastles = reinterpret_cast<SubCastle *>(ROUNDUP_PTR(ptr, alignof(u32)));
copy(subs.begin(), subs.end(), subCastles);
u32 length = 0;
@ -623,16 +627,16 @@ buildCastle(const CastleProto &proto,
SubCastle *sub = &subCastles[i];
sub->repeatInfoOffset = offset;
ptr = (char *)sub + offset;
ptr = reinterpret_cast<char *>(sub) + offset;
memcpy(ptr, &infos[i], sizeof(RepeatInfo));
if (patchSize[i]) {
RepeatInfo *info = (RepeatInfo *)ptr;
u64a *table = ((u64a *)(ROUNDUP_PTR(((char *)(info) +
sizeof(*info)), alignof(u64a))));
RepeatInfo *info = reinterpret_cast<RepeatInfo *>(ptr);
u64a *table = reinterpret_cast<u64a *>(ROUNDUP_PTR(info +
sizeof(*info), alignof(u64a)));
copy(tables.begin() + tableIdx,
tables.begin() + tableIdx + patchSize[i], table);
u32 diff = (char *)table - (char *)info +
u32 diff = reinterpret_cast<ptrdiff_t>(table) - reinterpret_cast<ptrdiff_t>(info) +
sizeof(u64a) * patchSize[i];
info->length = diff;
length += diff;
@ -655,7 +659,6 @@ buildCastle(const CastleProto &proto,
if (!stale_iter.empty()) {
c->staleIterOffset = verify_u32(ptr - base_ptr);
copy_bytes(ptr, stale_iter);
ptr += byte_length(stale_iter);
}
return nfa;
@ -922,7 +925,7 @@ void addToHolder(NGHolder &g, u32 top, const PureRepeat &pr) {
u32 min_bound = pr.bounds.min; // always finite
if (min_bound == 0) { // Vacuous case, we can only do this once.
assert(!edge(g.start, g.accept, g).second);
NFAEdge e = add_edge(g.start, g.accept, g);
NFAEdge e = add_edge(g.start, g.accept, g).first;
g[e].tops.insert(top);
g[u].reports.insert(pr.reports.begin(), pr.reports.end());
min_bound = 1;
@ -931,7 +934,7 @@ void addToHolder(NGHolder &g, u32 top, const PureRepeat &pr) {
for (u32 i = 0; i < min_bound; i++) {
NFAVertex v = add_vertex(g);
g[v].char_reach = pr.reach;
NFAEdge e = add_edge(u, v, g);
NFAEdge e = add_edge(u, v, g).first;
if (u == g.start) {
g[e].tops.insert(top);
}
@ -950,7 +953,7 @@ void addToHolder(NGHolder &g, u32 top, const PureRepeat &pr) {
if (head != u) {
add_edge(head, v, g);
}
NFAEdge e = add_edge(u, v, g);
NFAEdge e = add_edge(u, v, g).first;
if (u == g.start) {
g[e].tops.insert(top);
}

1
src/nfa/dfa_min.cpp
View File

@ -305,6 +305,7 @@ void minimize_hopcroft(raw_dfa &rdfa, const Grey &grey) {
DEBUG_PRINTF("dfa is empty\n");
}
// cppcheck-suppress unreadVariable
UNUSED const size_t states_before = rdfa.states.size();
HopcroftInfo info(rdfa);

8
src/nfa/gough.c
View File

@ -978,14 +978,14 @@ char nfaExecGough16_initCompressedState(const struct NFA *nfa, u64a offset,
char nfaExecGough8_reportCurrent(const struct NFA *n, struct mq *q) {
const struct mcclellan *m = (const struct mcclellan *)getImplNfa(n);
NfaCallback cb = q->cb;
void *ctxt = q->context;
u8 s = *(u8 *)q->state;
u64a offset = q_cur_offset(q);
struct gough_som_info *som = getSomInfo(q->state);
const struct gough_som_info *som = getSomInfo(q->state);
assert(q_cur_type(q) == MQE_START);
assert(s);
if (s >= m->accept_limit_8) {
void *ctxt = q->context;
u32 cached_accept_id = 0;
u16 cached_accept_state = 0;
u32 cached_accept_som = 0;
@ -1000,16 +1000,16 @@ char nfaExecGough8_reportCurrent(const struct NFA *n, struct mq *q) {
char nfaExecGough16_reportCurrent(const struct NFA *n, struct mq *q) {
const struct mcclellan *m = (const struct mcclellan *)getImplNfa(n);
NfaCallback cb = q->cb;
void *ctxt = q->context;
u16 s = *(u16 *)q->state;
const struct mstate_aux *aux = get_aux(m, s);
u64a offset = q_cur_offset(q);
struct gough_som_info *som = getSomInfo(q->state);
const struct gough_som_info *som = getSomInfo(q->state);
assert(q_cur_type(q) == MQE_START);
DEBUG_PRINTF("state %hu\n", s);
assert(s);
if (aux->accept) {
void *ctxt = q->context;
u32 cached_accept_id = 0;
u16 cached_accept_state = 0;
u32 cached_accept_som = 0;

54
src/nfa/goughcompile.cpp
View File

@ -132,7 +132,7 @@ void GoughSSAVarMin::replace_input(GoughSSAVar *old_v, GoughSSAVar *new_v) {
}
static
void translateRawReports(UNUSED GoughGraph &cfg, UNUSED const raw_som_dfa &raw,
void translateRawReports(UNUSED const GoughGraph &cfg, UNUSED const raw_som_dfa &raw,
const flat_map<u32, GoughSSAVarJoin *> &joins_at_s,
UNUSED GoughVertex s,
const set<som_report> &reports_in,
@ -206,10 +206,6 @@ void makeCFG_top_edge(GoughGraph &cfg, const vector<GoughVertex> &vertices,
assert(contains(src_slots, slot_id));
shared_ptr<GoughSSAVarMin> vmin = make_shared<GoughSSAVarMin>();
if (!vmin) {
assert(0);
throw std::bad_alloc();
}
cfg[e].vars.emplace_back(vmin);
final_var = vmin.get();
@ -321,10 +317,6 @@ void makeCFG_edge(GoughGraph &cfg, const map<u32, u32> &som_creators,
DEBUG_PRINTF("bypassing min on join %u\n", slot_id);
} else {
shared_ptr<GoughSSAVarMin> vmin = make_shared<GoughSSAVarMin>();
if (!vmin) {
assert(0);
throw std::bad_alloc();
}
cfg[e].vars.emplace_back(vmin);
final_var = vmin.get();
@ -441,10 +433,11 @@ unique_ptr<GoughGraph> makeCFG(const raw_som_dfa &raw) {
}
static
// cppcheck-suppress constParameterReference
void copy_propagate_report_set(vector<pair<ReportID, GoughSSAVar *> > &rep) {
vector<pair<ReportID, GoughSSAVar *> >::iterator it = rep.begin();
while (it != rep.end()) {
GoughSSAVar *var = it->second;
const GoughSSAVar *var = it->second;
if (!var) {
++it;
continue;
@ -546,7 +539,7 @@ void remove_dead(GoughGraph &g) {
}
while (!queue.empty()) {
GoughSSAVar *v = queue.back();
const GoughSSAVar *v = queue.back();
queue.pop_back();
for (GoughSSAVar *var : v->get_inputs()) {
if (var->seen) {
@ -659,8 +652,8 @@ GoughSSAVar *GoughSSAVarJoin::get_input(const GoughEdge &prev) const {
return nullptr;
}
const flat_set<GoughEdge> &GoughSSAVarJoin::get_edges_for_input(
GoughSSAVar *input) const {
// cppcheck-suppress constParameterPointer
const flat_set<GoughEdge> &GoughSSAVarJoin::get_edges_for_input(GoughSSAVar *input) const {
return input_map.at(input);
}
@ -811,7 +804,7 @@ private:
static
void prep_joins_for_generation(const GoughGraph &g, GoughVertex v,
map<GoughEdge, edge_join_info> *edge_info) {
map<GoughEdge, edge_join_info> &edge_info) {
DEBUG_PRINTF("writing out joins for %u\n", g[v].state_id);
for (const auto &var : g[v].vars) {
u32 dest_slot = var->slot;
@ -822,7 +815,7 @@ void prep_joins_for_generation(const GoughGraph &g, GoughVertex v,
}
for (const GoughEdge &incoming_edge : var_edges.second) {
(*edge_info)[incoming_edge].insert(input, dest_slot);
edge_info[incoming_edge].insert(input, dest_slot);
DEBUG_PRINTF("need %u<-%u\n", dest_slot, input);
}
}
@ -920,7 +913,7 @@ void build_blocks(const GoughGraph &g,
}
map<GoughEdge, edge_join_info> eji;
prep_joins_for_generation(g, t, &eji);
prep_joins_for_generation(g, t, eji);
for (auto &m : eji) {
vector<gough_ins> &block = (*blocks)[gough_edge_id(g, m.first)];
@ -1019,7 +1012,7 @@ void update_accel_prog_offset(const gough_build_strat &gbs,
verts[gbs.gg[v].state_id] = v;
}
for (auto &m : gbs.built_accel) {
for (const auto &m : gbs.built_accel) {
gough_accel *ga = m.first;
assert(!ga->prog_offset);
GoughVertex v = verts[m.second];
@ -1052,7 +1045,7 @@ bytecode_ptr<NFA> goughCompile(raw_som_dfa &raw, u8 somPrecision,
|| !cc.streaming);
if (!cc.grey.allowGough) {
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
DEBUG_PRINTF("hello world\n");
@ -1083,11 +1076,12 @@ bytecode_ptr<NFA> goughCompile(raw_som_dfa &raw, u8 somPrecision,
auto basic_dfa = mcclellanCompile_i(raw, gbs, cc);
assert(basic_dfa);
if (!basic_dfa) {
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
u8 alphaShift
= ((const mcclellan *)getImplNfa(basic_dfa.get()))->alphaShift;
// cppcheck-suppress cstyleCast
const auto nfa = static_cast<const mcclellan *>(getImplNfa(basic_dfa.get()));
u8 alphaShift = nfa->alphaShift;
u32 edge_count = (1U << alphaShift) * raw.states.size();
u32 curr_offset = ROUNDUP_N(basic_dfa->length, 4);
@ -1128,8 +1122,8 @@ bytecode_ptr<NFA> goughCompile(raw_som_dfa &raw, u8 somPrecision,
u32 gough_size = ROUNDUP_N(curr_offset, 16);
auto gough_dfa = make_zeroed_bytecode_ptr<NFA>(gough_size);
memcpy(gough_dfa.get(), basic_dfa.get(), basic_dfa->length);
memcpy((char *)gough_dfa.get() + haig_offset, &gi, sizeof(gi));
memcpy(reinterpret_cast<char *>(gough_dfa.get()), basic_dfa.get(), basic_dfa->length);
memcpy(reinterpret_cast<char *>(gough_dfa.get()) + haig_offset, &gi, sizeof(gi));
if (gough_dfa->type == MCCLELLAN_NFA_16) {
gough_dfa->type = GOUGH_NFA_16;
} else {
@ -1142,18 +1136,19 @@ bytecode_ptr<NFA> goughCompile(raw_som_dfa &raw, u8 somPrecision,
gough_dfa->streamStateSize = base_state_size + slot_count * somPrecision;
gough_dfa->scratchStateSize = (u32)(16 + scratch_slot_count * sizeof(u64a));
mcclellan *m = (mcclellan *)getMutableImplNfa(gough_dfa.get());
// cppcheck-suppress cstyleCast
auto *m = reinterpret_cast<mcclellan *>(getMutableImplNfa(gough_dfa.get()));
m->haig_offset = haig_offset;
/* update nfa length, haig_info offset (leave mcclellan length alone) */
gough_dfa->length = gough_size;
/* copy in blocks */
copy_bytes((u8 *)gough_dfa.get() + edge_prog_offset, edge_blocks);
copy_bytes(reinterpret_cast<u8 *>(gough_dfa.get()) + edge_prog_offset, edge_blocks);
if (top_prog_offset) {
copy_bytes((u8 *)gough_dfa.get() + top_prog_offset, top_blocks);
copy_bytes(reinterpret_cast<u8 *>(gough_dfa.get()) + top_prog_offset, top_blocks);
}
copy_bytes((u8 *)gough_dfa.get() + prog_base_offset, temp_blocks);
copy_bytes(reinterpret_cast<u8 *>(gough_dfa.get()) + prog_base_offset, temp_blocks);
return gough_dfa;
}
@ -1186,7 +1181,7 @@ AccelScheme gough_build_strat::find_escape_strings(dstate_id_t this_idx) const {
void gough_build_strat::buildAccel(dstate_id_t this_idx, const AccelScheme &info,
void *accel_out) {
assert(mcclellan_build_strat::accelSize() == sizeof(AccelAux));
gough_accel *accel = (gough_accel *)accel_out;
gough_accel *accel = reinterpret_cast<gough_accel *>(accel_out);
/* build a plain accelaux so we can work out where we can get to */
mcclellan_build_strat::buildAccel(this_idx, info, &accel->accel);
DEBUG_PRINTF("state %hu is accel with type %hhu\n", this_idx,
@ -1324,7 +1319,8 @@ void raw_gough_report_info_impl::fillReportLists(NFA *n, size_t base_offset,
for (const raw_gough_report_list &r : rl) {
ro.emplace_back(base_offset);
gough_report_list *p = (gough_report_list *)((char *)n + base_offset);
u8 * n_ptr = reinterpret_cast<u8 *>(n);
gough_report_list *p = reinterpret_cast<gough_report_list *>(n_ptr + base_offset);
u32 i = 0;
for (const som_report &sr : r.reports) {

6
src/nfa/goughcompile_reg.cpp
View File

@ -195,7 +195,7 @@ void handle_pending_vars(GoughSSAVar *def, const GoughGraph &g,
if (contains(aux.containing_v, var)) {
/* def is used by join vertex, value only needs to be live on some
* incoming edges */
GoughSSAVarJoin *vj = (GoughSSAVarJoin *)var;
const GoughSSAVarJoin *vj = reinterpret_cast<const GoughSSAVarJoin *>(var);
const flat_set<GoughEdge> &live_edges
= vj->get_edges_for_input(def);
for (const auto &e : live_edges) {
@ -279,7 +279,7 @@ set<const GoughSSAVar *> live_during(GoughSSAVar *def, const GoughGraph &g,
template<typename VarP>
void set_initial_slots(const vector<VarP> &vars, u32 *next_slot) {
for (auto &var : vars) {
for (const auto &var : vars) {
assert(var->slot == INVALID_SLOT);
var->slot = (*next_slot)++;
}
@ -440,7 +440,7 @@ void create_slot_mapping(const GoughGraph &cfg, UNUSED u32 old_slot_count,
}
static
void update_local_slots(GoughGraph &g, set<GoughSSAVar *> &locals,
void update_local_slots(GoughGraph &g, const set<GoughSSAVar *> &locals,
u32 local_base) {
DEBUG_PRINTF("%zu local variables\n", locals.size());
/* local variables only occur on edges (joins are never local) */

2
src/nfa/limex.h
View File

@ -56,7 +56,7 @@ extern "C"
char gf_name##_Q(const struct NFA *n, struct mq *q, s64a end); \
char gf_name##_Q2(const struct NFA *n, struct mq *q, s64a end); \
char gf_name##_QR(const struct NFA *n, struct mq *q, ReportID report); \
char gf_name##_reportCurrent(const struct NFA *n, struct mq *q); \
char gf_name##_reportCurrent(const struct NFA *n, const struct mq *q); \
char gf_name##_inAccept(const struct NFA *n, ReportID report, \
struct mq *q); \
char gf_name##_inAnyAccept(const struct NFA *n, struct mq *q); \

4
src/nfa/limex_common_impl.h
View File

@ -332,7 +332,7 @@ void EXPIRE_ESTATE_FN(const IMPL_NFA_T *limex, struct CONTEXT_T *ctx,
// UE-1636) need to guard cyclic tug-accepts as well.
static really_inline
char LIMEX_INACCEPT_FN(const IMPL_NFA_T *limex, STATE_T state,
union RepeatControl *repeat_ctrl, char *repeat_state,
const union RepeatControl *repeat_ctrl, const char *repeat_state,
u64a offset, ReportID report) {
assert(limex);
@ -382,7 +382,7 @@ char LIMEX_INACCEPT_FN(const IMPL_NFA_T *limex, STATE_T state,
static really_inline
char LIMEX_INANYACCEPT_FN(const IMPL_NFA_T *limex, STATE_T state,
union RepeatControl *repeat_ctrl, char *repeat_state,
const union RepeatControl *repeat_ctrl, const char *repeat_state,
u64a offset) {
assert(limex);

28
src/nfa/limex_compile.cpp
View File

@ -290,7 +290,7 @@ void maskSetBits(Mask &m, const NFAStateSet &bits) {
template<class Mask>
bool isMaskZero(Mask &m) {
u8 *m8 = (u8 *)&m;
const u8 *m8 = (u8 *)&m;
for (u32 i = 0; i < sizeof(m); i++) {
if (m8[i]) {
return false;
@ -329,11 +329,11 @@ void buildReachMapping(const build_info &args, vector<NFAStateSet> &reach,
// Build a list of vertices with a state index assigned.
vector<NFAVertex> verts;
verts.reserve(args.num_states);
for (auto v : vertices_range(h)) {
if (state_ids.at(v) != NO_STATE) {
verts.emplace_back(v);
}
}
auto sidat = [&state_ids=state_ids](const NFAVertex &v) {
return (state_ids.at(v) != NO_STATE);
};
const auto &vr = vertices_range(h);
std::copy_if(begin(vr), end(vr), std::back_inserter(verts), sidat);
// Build a mapping from set-of-states -> reachability.
map<NFAStateSet, CharReach> mapping;
@ -556,7 +556,8 @@ void filterAccelStates(NGHolder &g, const map<u32, set<NFAVertex>> &tops,
// Similarly, connect (start, startDs) if necessary.
if (!edge(g.start, g.startDs, g).second) {
NFAEdge e = add_edge(g.start, g.startDs, g);
NFAEdge e;
std::tie(e, std::ignore) = add_edge(g.start, g.startDs, g);
tempEdges.emplace_back(e); // Remove edge later.
}
@ -1485,6 +1486,7 @@ u32 buildExceptionMap(const build_info &args, ReportListCache &reports_cache,
continue;
}
u32 j = args.state_ids.at(w);
// j can be NO_STATE if args.state_ids.at(w) returns NO_STATE
if (j == NO_STATE) {
continue;
}
@ -1576,7 +1578,7 @@ u32 findMaxVarShift(const build_info &args, u32 nShifts) {
static
int getLimexScore(const build_info &args, u32 nShifts) {
const NGHolder &h = args.h;
u32 maxVarShift = nShifts;
u32 maxVarShift;
int score = 0;
score += SHIFT_COST * nShifts;
@ -1704,7 +1706,7 @@ struct Factory {
static
void allocState(NFA *nfa, u32 repeatscratchStateSize,
u32 repeatStreamState) {
implNFA_t *limex = (implNFA_t *)getMutableImplNfa(nfa);
const implNFA_t *limex = (implNFA_t *)getMutableImplNfa(nfa);
// LimEx NFAs now store the following in state:
// 1. state bitvector (always present)
@ -2222,7 +2224,7 @@ struct Factory {
static
bytecode_ptr<NFA> generateNfa(const build_info &args) {
if (args.num_states > NFATraits<dtype>::maxStates) {
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
// Build bounded repeat structures.
@ -2581,7 +2583,7 @@ bytecode_ptr<NFA> generate(NGHolder &h,
if (!cc.grey.allowLimExNFA) {
DEBUG_PRINTF("limex not allowed\n");
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
// If you ask for a particular type, it had better be an NFA.
@ -2616,7 +2618,7 @@ bytecode_ptr<NFA> generate(NGHolder &h,
if (scores.empty()) {
DEBUG_PRINTF("No NFA returned a valid score for this case.\n");
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
// Sort acceptable models in priority order, lowest score first.
@ -2635,7 +2637,7 @@ bytecode_ptr<NFA> generate(NGHolder &h,
}
DEBUG_PRINTF("NFA build failed.\n");
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
u32 countAccelStates(NGHolder &h,

6
src/nfa/limex_exceptional.h
View File

@ -302,8 +302,8 @@ int PE_FN(STATE_ARG, ESTATE_ARG, UNUSED u32 diffmask, STATE_T *succ,
}
#else
// A copy of the estate as an array of GPR-sized chunks.
CHUNK_T chunks[sizeof(STATE_T) / sizeof(CHUNK_T)];
CHUNK_T emask_chunks[sizeof(STATE_T) / sizeof(CHUNK_T)];
CHUNK_T chunks[sizeof(STATE_T) / sizeof(CHUNK_T)]; // cppcheck-suppress duplicateExpression
CHUNK_T emask_chunks[sizeof(STATE_T) / sizeof(CHUNK_T)]; // cppcheck-suppress duplicateExpression
#ifdef ESTATE_ON_STACK
memcpy(chunks, &estate, sizeof(STATE_T));
#else
@ -311,7 +311,7 @@ int PE_FN(STATE_ARG, ESTATE_ARG, UNUSED u32 diffmask, STATE_T *succ,
#endif
memcpy(emask_chunks, &limex->exceptionMask, sizeof(STATE_T));
u32 base_index[sizeof(STATE_T) / sizeof(CHUNK_T)];
u32 base_index[sizeof(STATE_T) / sizeof(CHUNK_T)]; // cppcheck-suppress duplicateExpression
base_index[0] = 0;
for (s32 i = 0; i < (s32)ARRAY_LENGTH(base_index) - 1; i++) {
base_index[i + 1] = base_index[i] + POPCOUNT_FN(emask_chunks[i]);

14
src/nfa/limex_runtime_impl.h
View File

@ -927,7 +927,7 @@ char JOIN(LIMEX_API_ROOT, _testEOD)(const struct NFA *n, const char *state,
context);
}
char JOIN(LIMEX_API_ROOT, _reportCurrent)(const struct NFA *n, struct mq *q) {
char JOIN(LIMEX_API_ROOT, _reportCurrent)(const struct NFA *n, const struct mq *q) {
const IMPL_NFA_T *limex = getImplNfa(n);
REPORTCURRENT_FN(limex, q);
return 1;
@ -984,9 +984,9 @@ char JOIN(LIMEX_API_ROOT, _inAccept)(const struct NFA *nfa,
assert(q->state && q->streamState);
const IMPL_NFA_T *limex = getImplNfa(nfa);
union RepeatControl *repeat_ctrl =
const union RepeatControl *repeat_ctrl =
getRepeatControlBase(q->state, sizeof(STATE_T));
char *repeat_state = q->streamState + limex->stateSize;
const char *repeat_state = q->streamState + limex->stateSize;
STATE_T state = *(STATE_T *)q->state;
u64a offset = q->offset + q_last_loc(q) + 1;
@ -999,9 +999,9 @@ char JOIN(LIMEX_API_ROOT, _inAnyAccept)(const struct NFA *nfa, struct mq *q) {
assert(q->state && q->streamState);
const IMPL_NFA_T *limex = getImplNfa(nfa);
union RepeatControl *repeat_ctrl =
const union RepeatControl *repeat_ctrl =
getRepeatControlBase(q->state, sizeof(STATE_T));
char *repeat_state = q->streamState + limex->stateSize;
const char *repeat_state = q->streamState + limex->stateSize;
STATE_T state = *(STATE_T *)q->state;
u64a offset = q->offset + q_last_loc(q) + 1;
@ -1020,9 +1020,9 @@ enum nfa_zombie_status JOIN(LIMEX_API_ROOT, _zombie_status)(
if (limex->repeatCount) {
u64a offset = q->offset + loc + 1;
union RepeatControl *repeat_ctrl =
const union RepeatControl *repeat_ctrl =
getRepeatControlBase(q->state, sizeof(STATE_T));
char *repeat_state = q->streamState + limex->stateSize;
const char *repeat_state = q->streamState + limex->stateSize;
SQUASH_UNTUG_BR_FN(limex, repeat_ctrl, repeat_state, offset, &state);
}

26
src/nfa/mcclellancompile.cpp
View File

@ -177,7 +177,7 @@ static
mstate_aux *getAux(NFA *n, dstate_id_t i) {
assert(isMcClellanType(n->type));
mcclellan *m = (mcclellan *)getMutableImplNfa(n);
const mcclellan *m = (mcclellan *)getMutableImplNfa(n);
mstate_aux *aux_base = (mstate_aux *)((char *)n + m->aux_offset);
mstate_aux *aux = aux_base + i;
@ -203,7 +203,7 @@ void markEdges(NFA *n, u16 *succ_table, const dfa_info &info) {
continue;
}
mstate_aux *aux = getAux(n, succ_table[c_prime]);
const mstate_aux *aux = getAux(n, succ_table[c_prime]);
if (aux->accept) {
succ_table[c_prime] |= ACCEPT_FLAG;
@ -232,7 +232,7 @@ void markEdges(NFA *n, u16 *succ_table, const dfa_info &info) {
continue;
}
mstate_aux *aux = getAux(n, succ_i);
const mstate_aux *aux = getAux(n, succ_i);
if (aux->accept) {
succ_i |= ACCEPT_FLAG;
@ -262,7 +262,7 @@ void markEdges(NFA *n, u16 *succ_table, const dfa_info &info) {
// check successful transition
u16 next = unaligned_load_u16((u8 *)trans);
if (next < wide_limit) {
mstate_aux *aux = getAux(n, next);
const mstate_aux *aux = getAux(n, next);
if (aux->accept) {
next |= ACCEPT_FLAG;
}
@ -279,7 +279,7 @@ void markEdges(NFA *n, u16 *succ_table, const dfa_info &info) {
if (next_k >= wide_limit) {
continue;
}
mstate_aux *aux_k = getAux(n, next_k);
const mstate_aux *aux_k = getAux(n, next_k);
if (aux_k->accept) {
next_k |= ACCEPT_FLAG;
}
@ -362,7 +362,7 @@ struct raw_report_list {
raw_report_list(const flat_set<ReportID> &reports_in,
const ReportManager &rm, bool do_remap) {
if (do_remap) {
for (auto &id : reports_in) {
for (const auto &id : reports_in) {
reports.insert(rm.getProgramOffset(id));
}
} else {
@ -546,7 +546,7 @@ size_t calcWideRegionSize(const dfa_info &info) {
static
void fillInAux(mstate_aux *aux, dstate_id_t i, const dfa_info &info,
const vector<u32> &reports, const vector<u32> &reports_eod,
vector<u32> &reportOffsets) {
const vector<u32> &reportOffsets) {
const dstate &raw_state = info.states[i];
aux->accept = raw_state.reports.empty() ? 0 : reportOffsets[reports[i]];
aux->accept_eod = raw_state.reports_eod.empty() ? 0
@ -631,7 +631,7 @@ bytecode_ptr<NFA> mcclellanCompile16(dfa_info &info, const CompileContext &cc,
if (!allocateFSN16(info, &count_real_states, &wide_limit)) {
DEBUG_PRINTF("failed to allocate state numbers, %zu states total\n",
info.size());
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
DEBUG_PRINTF("count_real_states: %d\n", count_real_states);
@ -800,8 +800,8 @@ bytecode_ptr<NFA> mcclellanCompile16(dfa_info &info, const CompileContext &cc,
}
for (size_t i : order) {
vector<dstate_id_t> &state_chain = info.wide_state_chain[i];
vector<symbol_t> &symbol_chain = info.wide_symbol_chain[i];
const vector<dstate_id_t> &state_chain = info.wide_state_chain[i];
const vector<symbol_t> &symbol_chain = info.wide_symbol_chain[i];
u16 width = verify_u16(symbol_chain.size());
*(u16 *)(curr_wide_entry + WIDE_WIDTH_OFFSET) = width;
@ -1373,11 +1373,11 @@ bool store_chain_longest(vector<vector<dstate_id_t>> &candidate_chain,
/* \brief Generate wide_symbol_chain from wide_state_chain. */
static
void generate_symbol_chain(dfa_info &info, vector<symbol_t> &chain_tail) {
raw_dfa &rdfa = info.raw;
const raw_dfa &rdfa = info.raw;
assert(chain_tail.size() == info.wide_state_chain.size());
for (size_t i = 0; i < info.wide_state_chain.size(); i++) {
vector<dstate_id_t> &state_chain = info.wide_state_chain[i];
const vector<dstate_id_t> &state_chain = info.wide_state_chain[i];
vector<symbol_t> symbol_chain;
info.extra[state_chain[0]].wideHead = true;
@ -1385,7 +1385,6 @@ void generate_symbol_chain(dfa_info &info, vector<symbol_t> &chain_tail) {
for (size_t j = 0; j < width; j++) {
dstate_id_t curr_id = state_chain[j];
dstate_id_t next_id = state_chain[j + 1];
// The last state of the chain doesn't belong to a wide state.
info.extra[curr_id].wideState = true;
@ -1394,6 +1393,7 @@ void generate_symbol_chain(dfa_info &info, vector<symbol_t> &chain_tail) {
if (j == width - 1) {
symbol_chain.emplace_back(chain_tail[i]);
} else {
dstate_id_t next_id = state_chain[j + 1];
for (symbol_t sym = 0; sym < info.impl_alpha_size; sym++) {
if (rdfa.states[curr_id].next[sym] == next_id) {
symbol_chain.emplace_back(sym);

115
src/nfa/mcsheng_compile.cpp
View File

@ -144,11 +144,11 @@ u8 dfa_info::getAlphaShift() const {
static
mstate_aux *getAux(NFA *n, dstate_id_t i) {
mcsheng *m = (mcsheng *)getMutableImplNfa(n);
mstate_aux *aux_base = (mstate_aux *)((char *)n + m->aux_offset);
const mcsheng *m = reinterpret_cast<const mcsheng *>(getMutableImplNfa(n));
mstate_aux *aux_base = reinterpret_cast<mstate_aux *>(reinterpret_cast<u8 *>(n) + m->aux_offset);
mstate_aux *aux = aux_base + i;
assert((const char *)aux < (const char *)n + m->length);
assert(reinterpret_cast<const char *>(aux) < reinterpret_cast<const char *>(n) + m->length);
return aux;
}
@ -192,8 +192,8 @@ void createShuffleMasks(mcsheng *m, const dfa_info &info,
}
for (u32 i = 0; i < N_CHARS; i++) {
assert(info.alpha_remap[i] != info.alpha_remap[TOP]);
memcpy((u8 *)&m->sheng_masks[i],
(u8 *)masks[info.alpha_remap[i]].data(), sizeof(m128));
memcpy(reinterpret_cast<u8 *>(&m->sheng_masks[i]),
reinterpret_cast<u8 *>(masks[info.alpha_remap[i]].data()), sizeof(m128));
}
m->sheng_end = sheng_end;
m->sheng_accel_limit = sheng_end - 1;
@ -223,7 +223,7 @@ void populateBasicInfo(size_t state_size, const dfa_info &info,
nfa->type = MCSHENG_NFA_16;
}
mcsheng *m = (mcsheng *)getMutableImplNfa(nfa);
mcsheng *m = reinterpret_cast<mcsheng *>(getMutableImplNfa(nfa));
for (u32 i = 0; i < 256; i++) {
m->remap[i] = verify_u8(info.alpha_remap[i]);
}
@ -244,11 +244,11 @@ void populateBasicInfo(size_t state_size, const dfa_info &info,
static
mstate_aux *getAux64(NFA *n, dstate_id_t i) {
mcsheng64 *m = (mcsheng64 *)getMutableImplNfa(n);
mstate_aux *aux_base = (mstate_aux *)((char *)n + m->aux_offset);
const mcsheng64 *m = reinterpret_cast<const mcsheng64 *>(getMutableImplNfa(n));
mstate_aux *aux_base = reinterpret_cast<mstate_aux *>(reinterpret_cast<u8 *>(n) + m->aux_offset);
mstate_aux *aux = aux_base + i;
assert((const char *)aux < (const char *)n + m->length);
assert(reinterpret_cast<const char *>(aux) < reinterpret_cast<const char *>(n) + m->length);
return aux;
}
@ -292,8 +292,8 @@ void createShuffleMasks64(mcsheng64 *m, const dfa_info &info,
}
for (u32 i = 0; i < N_CHARS; i++) {
assert(info.alpha_remap[i] != info.alpha_remap[TOP]);
memcpy((u8 *)&m->sheng_succ_masks[i],
(u8 *)masks[info.alpha_remap[i]].data(), sizeof(m512));
memcpy(reinterpret_cast<u8 *>(&m->sheng_succ_masks[i]),
reinterpret_cast<u8 *>(masks[info.alpha_remap[i]].data()), sizeof(m512));
}
m->sheng_end = sheng_end;
m->sheng_accel_limit = sheng_end - 1;
@ -323,7 +323,7 @@ void populateBasicInfo64(size_t state_size, const dfa_info &info,
nfa->type = MCSHENG_64_NFA_16;
}
mcsheng64 *m = (mcsheng64 *)getMutableImplNfa(nfa);
mcsheng64 *m = reinterpret_cast<mcsheng64 *>(getMutableImplNfa(nfa));
for (u32 i = 0; i < 256; i++) {
m->remap[i] = verify_u8(info.alpha_remap[i]);
}
@ -534,7 +534,7 @@ double leakiness(const RdfaGraph &g, dfa_info &info,
static
dstate_id_t find_sheng_states(dfa_info &info,
map<dstate_id_t, AccelScheme> &accel_escape_info,
const map<dstate_id_t, AccelScheme> &accel_escape_info,
size_t max_sheng_states) {
RdfaGraph g(info.raw);
auto cyclics = find_vertices_in_cycles(g);
@ -650,7 +650,7 @@ void fill_in_aux_info(NFA *nfa, const dfa_info &info,
const vector<u32> &reports_eod,
u32 report_base_offset,
const raw_report_info &ri) {
mcsheng *m = (mcsheng *)getMutableImplNfa(nfa);
mcsheng *m = reinterpret_cast<mcsheng *>(getMutableImplNfa(nfa));
vector<u32> reportOffsets;
@ -667,14 +667,14 @@ void fill_in_aux_info(NFA *nfa, const dfa_info &info,
assert(accel_offset <= accel_end_offset);
assert(ISALIGNED_N(accel_offset, alignof(union AccelAux)));
info.strat.buildAccel(i, accel_escape_info.at(i),
(void *)((char *)m + this_aux->accel_offset));
reinterpret_cast<void *>(reinterpret_cast<char *>(m) + this_aux->accel_offset));
}
}
}
static
u16 get_edge_flags(NFA *nfa, dstate_id_t target_impl_id) {
mstate_aux *aux = getAux(nfa, target_impl_id);
const mstate_aux *aux = getAux(nfa, target_impl_id);
u16 flags = 0;
if (aux->accept) {
@ -692,7 +692,7 @@ static
void fill_in_succ_table_16(NFA *nfa, const dfa_info &info,
dstate_id_t sheng_end,
UNUSED dstate_id_t sherman_base) {
u16 *succ_table = (u16 *)((char *)nfa + sizeof(NFA) + sizeof(mcsheng));
u16 *succ_table = reinterpret_cast<u16 *>(reinterpret_cast<char *>(nfa) + sizeof(NFA) + sizeof(mcsheng));
u8 alphaShift = info.getAlphaShift();
assert(alphaShift <= 8);
@ -724,7 +724,7 @@ void fill_in_aux_info64(NFA *nfa, const dfa_info &info,
const vector<u32> &reports_eod,
u32 report_base_offset,
const raw_report_info &ri) {
mcsheng64 *m = (mcsheng64 *)getMutableImplNfa(nfa);
mcsheng64 *m = reinterpret_cast<mcsheng64 *>(getMutableImplNfa(nfa));
vector<u32> reportOffsets;
@ -741,14 +741,14 @@ void fill_in_aux_info64(NFA *nfa, const dfa_info &info,
assert(accel_offset <= accel_end_offset);
assert(ISALIGNED_N(accel_offset, alignof(union AccelAux)));
info.strat.buildAccel(i, accel_escape_info.at(i),
(void *)((char *)m + this_aux->accel_offset));
reinterpret_cast<void *>(reinterpret_cast<char *>(m) + this_aux->accel_offset));
}
}
}
static
u16 get_edge_flags64(NFA *nfa, dstate_id_t target_impl_id) {
mstate_aux *aux = getAux64(nfa, target_impl_id);
const mstate_aux *aux = getAux64(nfa, target_impl_id);
u16 flags = 0;
if (aux->accept) {
@ -766,7 +766,7 @@ static
void fill_in_succ_table_64_16(NFA *nfa, const dfa_info &info,
dstate_id_t sheng_end,
UNUSED dstate_id_t sherman_base) {
u16 *succ_table = (u16 *)((char *)nfa + sizeof(NFA) + sizeof(mcsheng64));
u16 *succ_table = reinterpret_cast<u16 *>(reinterpret_cast<char *>(nfa) + sizeof(NFA) + sizeof(mcsheng64));
u8 alphaShift = info.getAlphaShift();
assert(alphaShift <= 8);
@ -955,9 +955,9 @@ bool is_cyclic_near(const raw_dfa &raw, dstate_id_t root) {
}
static
void fill_in_sherman(NFA *nfa, dfa_info &info, UNUSED u16 sherman_limit) {
char *nfa_base = (char *)nfa;
mcsheng *m = (mcsheng *)getMutableImplNfa(nfa);
void fill_in_sherman(NFA *nfa, const dfa_info &info, UNUSED u16 sherman_limit) {
char *nfa_base = reinterpret_cast<char *>(nfa);
mcsheng *m = reinterpret_cast<mcsheng *>(getMutableImplNfa(nfa));
char *sherman_table = nfa_base + m->sherman_offset;
assert(ISALIGNED_16(sherman_table));
@ -978,10 +978,10 @@ void fill_in_sherman(NFA *nfa, dfa_info &info, UNUSED u16 sherman_limit) {
assert(len <= 9);
dstate_id_t d = info.states[i].daddy;
*(u8 *)(curr_sherman_entry + SHERMAN_TYPE_OFFSET) = SHERMAN_STATE;
*(u8 *)(curr_sherman_entry + SHERMAN_LEN_OFFSET) = len;
*(u16 *)(curr_sherman_entry + SHERMAN_DADDY_OFFSET) = info.implId(d);
u8 *chars = (u8 *)(curr_sherman_entry + SHERMAN_CHARS_OFFSET);
*(reinterpret_cast<u8 *>(curr_sherman_entry + SHERMAN_TYPE_OFFSET)) = SHERMAN_STATE;
*(reinterpret_cast<u8 *>(curr_sherman_entry + SHERMAN_LEN_OFFSET)) = len;
*(reinterpret_cast<u16 *>(curr_sherman_entry + SHERMAN_DADDY_OFFSET)) = info.implId(d);
u8 *chars = reinterpret_cast<u8 *>(curr_sherman_entry + SHERMAN_CHARS_OFFSET);
for (u16 s = 0; s < info.impl_alpha_size; s++) {
if (info.states[i].next[s] != info.states[d].next[s]) {
@ -989,7 +989,7 @@ void fill_in_sherman(NFA *nfa, dfa_info &info, UNUSED u16 sherman_limit) {
}
}
u16 *states = (u16 *)(curr_sherman_entry + SHERMAN_STATES_OFFSET(len));
u16 *states = reinterpret_cast<u16 *>(curr_sherman_entry + SHERMAN_STATES_OFFSET(len));
for (u16 s = 0; s < info.impl_alpha_size; s++) {
if (info.states[i].next[s] != info.states[d].next[s]) {
DEBUG_PRINTF("s overrider %hu dad %hu char next %hu\n", fs,
@ -997,7 +997,7 @@ void fill_in_sherman(NFA *nfa, dfa_info &info, UNUSED u16 sherman_limit) {
info.implId(info.states[i].next[s]));
u16 entry_val = info.implId(info.states[i].next[s]);
entry_val |= get_edge_flags(nfa, entry_val);
unaligned_store_u16((u8 *)states++, entry_val);
unaligned_store_u16(reinterpret_cast<u8 *>(states++), entry_val);
}
}
}
@ -1018,12 +1018,16 @@ bytecode_ptr<NFA> mcshengCompile16(dfa_info &info, dstate_id_t sheng_end,
// Sherman optimization
if (info.impl_alpha_size > 16) {
#ifdef DEBUG
u16 total_daddy = 0;
#endif // DEBUG
for (u32 i = 0; i < info.size(); i++) {
find_better_daddy(info, i,
is_cyclic_near(info.raw, info.raw.start_anchored),
grey);
#ifdef DEBUG
total_daddy += info.extra[i].daddytaken;
#endif // DEBUG
}
DEBUG_PRINTF("daddy %hu/%zu states=%zu alpha=%hu\n", total_daddy,
@ -1035,7 +1039,7 @@ bytecode_ptr<NFA> mcshengCompile16(dfa_info &info, dstate_id_t sheng_end,
if (!allocateImplId16(info, sheng_end, &sherman_limit)) {
DEBUG_PRINTF("failed to allocate state numbers, %zu states total\n",
info.size());
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
u16 count_real_states = sherman_limit - sheng_end;
@ -1059,7 +1063,7 @@ bytecode_ptr<NFA> mcshengCompile16(dfa_info &info, dstate_id_t sheng_end,
assert(ISALIGNED_N(accel_offset, alignof(union AccelAux)));
auto nfa = make_zeroed_bytecode_ptr<NFA>(total_size);
mcsheng *m = (mcsheng *)getMutableImplNfa(nfa.get());
mcsheng *m = reinterpret_cast<mcsheng *>(getMutableImplNfa(nfa.get()));
populateBasicInfo(sizeof(u16), info, total_size, aux_offset, accel_offset,
accel_escape_info.size(), arb, single, nfa.get());
@ -1087,7 +1091,7 @@ bytecode_ptr<NFA> mcshengCompile16(dfa_info &info, dstate_id_t sheng_end,
static
void fill_in_succ_table_8(NFA *nfa, const dfa_info &info,
dstate_id_t sheng_end) {
u8 *succ_table = (u8 *)nfa + sizeof(NFA) + sizeof(mcsheng);
u8 *succ_table = reinterpret_cast<u8 *>(reinterpret_cast<char *>(nfa) + sizeof(NFA) + sizeof(mcsheng));
u8 alphaShift = info.getAlphaShift();
assert(alphaShift <= 8);
@ -1109,9 +1113,9 @@ void fill_in_succ_table_8(NFA *nfa, const dfa_info &info,
}
static
void fill_in_sherman64(NFA *nfa, dfa_info &info, UNUSED u16 sherman_limit) {
char *nfa_base = (char *)nfa;
mcsheng64 *m = (mcsheng64 *)getMutableImplNfa(nfa);
void fill_in_sherman64(NFA *nfa, const dfa_info &info, UNUSED u16 sherman_limit) {
char *nfa_base = reinterpret_cast<char *>(nfa);
mcsheng *m = reinterpret_cast<mcsheng *>(getMutableImplNfa(nfa));
char *sherman_table = nfa_base + m->sherman_offset;
assert(ISALIGNED_16(sherman_table));
@ -1132,10 +1136,10 @@ void fill_in_sherman64(NFA *nfa, dfa_info &info, UNUSED u16 sherman_limit) {
assert(len <= 9);
dstate_id_t d = info.states[i].daddy;
*(u8 *)(curr_sherman_entry + SHERMAN_TYPE_OFFSET) = SHERMAN_STATE;
*(u8 *)(curr_sherman_entry + SHERMAN_LEN_OFFSET) = len;
*(u16 *)(curr_sherman_entry + SHERMAN_DADDY_OFFSET) = info.implId(d);
u8 *chars = (u8 *)(curr_sherman_entry + SHERMAN_CHARS_OFFSET);
*(reinterpret_cast<u8 *>(curr_sherman_entry + SHERMAN_TYPE_OFFSET)) = SHERMAN_STATE;
*(reinterpret_cast<u8 *>(curr_sherman_entry + SHERMAN_LEN_OFFSET)) = len;
*(reinterpret_cast<u16 *>(curr_sherman_entry + SHERMAN_DADDY_OFFSET)) = info.implId(d);
u8 *chars = reinterpret_cast<u8 *>(curr_sherman_entry + SHERMAN_CHARS_OFFSET);
for (u16 s = 0; s < info.impl_alpha_size; s++) {
if (info.states[i].next[s] != info.states[d].next[s]) {
@ -1143,7 +1147,7 @@ void fill_in_sherman64(NFA *nfa, dfa_info &info, UNUSED u16 sherman_limit) {
}
}
u16 *states = (u16 *)(curr_sherman_entry + SHERMAN_STATES_OFFSET(len));
u16 *states = reinterpret_cast<u16 *>(curr_sherman_entry + SHERMAN_STATES_OFFSET(len));
for (u16 s = 0; s < info.impl_alpha_size; s++) {
if (info.states[i].next[s] != info.states[d].next[s]) {
DEBUG_PRINTF("s overrider %hu dad %hu char next %hu\n", fs,
@ -1151,7 +1155,7 @@ void fill_in_sherman64(NFA *nfa, dfa_info &info, UNUSED u16 sherman_limit) {
info.implId(info.states[i].next[s]));
u16 entry_val = info.implId(info.states[i].next[s]);
entry_val |= get_edge_flags64(nfa, entry_val);
unaligned_store_u16((u8 *)states++, entry_val);
unaligned_store_u16(reinterpret_cast<u8 *>(states++), entry_val);
}
}
}
@ -1172,12 +1176,16 @@ bytecode_ptr<NFA> mcsheng64Compile16(dfa_info&info, dstate_id_t sheng_end,
// Sherman optimization
if (info.impl_alpha_size > 16) {
#ifdef DEBUG
u16 total_daddy = 0;
#endif // DEBUG
for (u32 i = 0; i < info.size(); i++) {
find_better_daddy(info, i,
is_cyclic_near(info.raw, info.raw.start_anchored),
grey);
#ifdef DEBUG
total_daddy += info.extra[i].daddytaken;
#endif // DEBUG
}
DEBUG_PRINTF("daddy %hu/%zu states=%zu alpha=%hu\n", total_daddy,
@ -1189,7 +1197,7 @@ bytecode_ptr<NFA> mcsheng64Compile16(dfa_info&info, dstate_id_t sheng_end,
if (!allocateImplId16(info, sheng_end, &sherman_limit)) {
DEBUG_PRINTF("failed to allocate state numbers, %zu states total\n",
info.size());
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
u16 count_real_states = sherman_limit - sheng_end;
@ -1213,7 +1221,7 @@ bytecode_ptr<NFA> mcsheng64Compile16(dfa_info&info, dstate_id_t sheng_end,
assert(ISALIGNED_N(accel_offset, alignof(union AccelAux)));
auto nfa = make_zeroed_bytecode_ptr<NFA>(total_size);
mcsheng64 *m = (mcsheng64 *)getMutableImplNfa(nfa.get());
mcsheng64 *m = reinterpret_cast<mcsheng64 *>(getMutableImplNfa(nfa.get()));
populateBasicInfo64(sizeof(u16), info, total_size, aux_offset, accel_offset,
accel_escape_info.size(), arb, single, nfa.get());
@ -1241,7 +1249,7 @@ bytecode_ptr<NFA> mcsheng64Compile16(dfa_info&info, dstate_id_t sheng_end,
static
void fill_in_succ_table_64_8(NFA *nfa, const dfa_info &info,
dstate_id_t sheng_end) {
u8 *succ_table = (u8 *)nfa + sizeof(NFA) + sizeof(mcsheng64);
u8 *succ_table = reinterpret_cast<u8 *>(reinterpret_cast<char *>(nfa) + sizeof(NFA) + sizeof(mcsheng));
u8 alphaShift = info.getAlphaShift();
assert(alphaShift <= 8);
@ -1339,7 +1347,7 @@ bytecode_ptr<NFA> mcshengCompile8(dfa_info &info, dstate_id_t sheng_end,
assert(ISALIGNED_N(accel_offset, alignof(union AccelAux)));
auto nfa = make_zeroed_bytecode_ptr<NFA>(total_size);
mcsheng *m = (mcsheng *)getMutableImplNfa(nfa.get());
mcsheng *m = reinterpret_cast<mcsheng *>(getMutableImplNfa(nfa.get()));
allocateImplId8(info, sheng_end, accel_escape_info, &m->accel_limit_8,
&m->accept_limit_8);
@ -1392,7 +1400,7 @@ bytecode_ptr<NFA> mcsheng64Compile8(dfa_info &info, dstate_id_t sheng_end,
assert(ISALIGNED_N(accel_offset, alignof(union AccelAux)));
auto nfa = make_zeroed_bytecode_ptr<NFA>(total_size);
mcsheng64 *m = (mcsheng64 *)getMutableImplNfa(nfa.get());
mcsheng64 *m = reinterpret_cast<mcsheng64 *>(getMutableImplNfa(nfa.get()));
allocateImplId8(info, sheng_end, accel_escape_info, &m->accel_limit_8,
&m->accept_limit_8);
@ -1414,7 +1422,7 @@ bytecode_ptr<NFA> mcsheng64Compile8(dfa_info &info, dstate_id_t sheng_end,
bytecode_ptr<NFA> mcshengCompile(raw_dfa &raw, const CompileContext &cc,
const ReportManager &rm) {
if (!cc.grey.allowMcSheng) {
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
mcclellan_build_strat mbs(raw, rm, false);
@ -1430,12 +1438,10 @@ bytecode_ptr<NFA> mcshengCompile(raw_dfa &raw, const CompileContext &cc,
map<dstate_id_t, AccelScheme> accel_escape_info
= info.strat.getAccelInfo(cc.grey);
auto old_states = info.states;
dstate_id_t sheng_end = find_sheng_states(info, accel_escape_info, MAX_SHENG_STATES);
if (sheng_end <= DEAD_STATE + 1) {
info.states = old_states;
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
bytecode_ptr<NFA> nfa;
@ -1447,7 +1453,6 @@ bytecode_ptr<NFA> mcshengCompile(raw_dfa &raw, const CompileContext &cc,
}
if (!nfa) {
info.states = old_states;
return nfa;
}
@ -1462,12 +1467,12 @@ bytecode_ptr<NFA> mcshengCompile(raw_dfa &raw, const CompileContext &cc,
bytecode_ptr<NFA> mcshengCompile64(raw_dfa &raw, const CompileContext &cc,
const ReportManager &rm) {
if (!cc.grey.allowMcSheng) {
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
if (!cc.target_info.has_avx512vbmi()) {
DEBUG_PRINTF("McSheng64 failed, no HS_CPU_FEATURES_AVX512VBMI!\n");
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
mcclellan_build_strat mbs(raw, rm, false);
@ -1488,7 +1493,7 @@ bytecode_ptr<NFA> mcshengCompile64(raw_dfa &raw, const CompileContext &cc,
sheng_end64 = find_sheng_states(info, accel_escape_info, MAX_SHENG64_STATES);
if (sheng_end64 <= DEAD_STATE + 1) {
return nullptr;
return bytecode_ptr<NFA>(nullptr);
} else {
using64state = true;
}

14
src/nfa/mpv.c
View File

@ -512,7 +512,7 @@ size_t find_last_bad(const struct mpv_kilopuff *kp, const u8 *buf,
verm_restart:;
assert(buf[curr] == kp->u.verm.c);
size_t test = curr;
size_t test;
if (curr + min_rep < length) {
test = curr + min_rep;
} else {
@ -534,7 +534,7 @@ size_t find_last_bad(const struct mpv_kilopuff *kp, const u8 *buf,
m128 hi = kp->u.shuf.mask_hi;
shuf_restart:
assert(do_single_shufti(lo, hi, buf[curr]));
size_t test = curr;
size_t test;
if (curr + min_rep < length) {
test = curr + min_rep;
} else {
@ -556,7 +556,7 @@ size_t find_last_bad(const struct mpv_kilopuff *kp, const u8 *buf,
const m128 mask1 = kp->u.truffle.mask1;
const m128 mask2 = kp->u.truffle.mask2;
truffle_restart:;
size_t test = curr;
size_t test;
if (curr + min_rep < length) {
test = curr + min_rep;
} else {
@ -582,7 +582,7 @@ size_t find_last_bad(const struct mpv_kilopuff *kp, const u8 *buf,
nverm_restart:;
assert(buf[curr] != kp->u.verm.c);
size_t test = curr;
size_t test;
if (curr + min_rep < length) {
test = curr + min_rep;
} else {
@ -607,7 +607,7 @@ size_t find_last_bad(const struct mpv_kilopuff *kp, const u8 *buf,
}
static really_inline
void restartKilo(const struct mpv *m, UNUSED u8 *active, u8 *reporters,
void restartKilo(const struct mpv *m, UNUSED const u8 *active, u8 *reporters,
struct mpv_decomp_state *dstate, struct mpv_pq_item *pq,
const u8 *buf, u64a prev_limit, size_t buf_length, u32 i) {
const struct mpv_kilopuff *kp = (const void *)(m + 1);
@ -1074,7 +1074,7 @@ s64a nfaExecMpv_QueueExecRaw(const struct NFA *nfa, struct mq *q, s64a end) {
return 0;
} else {
const struct mpv *m = getImplNfa(nfa);
u8 *reporters = (u8 *)q->state + m->reporter_offset;
const u8 *reporters = (u8 *)q->state + m->reporter_offset;
if (mmbit_any_precise(reporters, m->kilo_count)) {
DEBUG_PRINTF("next byte\n");
@ -1087,7 +1087,7 @@ s64a nfaExecMpv_QueueExecRaw(const struct NFA *nfa, struct mq *q, s64a end) {
next_event = q->items[q->cur].location;
}
struct mpv_decomp_state *s = (struct mpv_decomp_state *)q->state;
const struct mpv_decomp_state *s = (struct mpv_decomp_state *)q->state;
struct mpv_pq_item *pq
= (struct mpv_pq_item *)(q->state + m->pq_offset);
if (s->pq_size) {

6
src/nfa/nfa_api_queue.h
View File

@ -167,7 +167,7 @@ void pushQueueNoMerge(struct mq * restrict q, u32 e, s64a loc) {
// We assert that the event is different from its predecessor. If it's a
// dupe, you should have used the ordinary pushQueue call.
if (q->end) {
UNUSED struct mq_item *prev = &q->items[q->end - 1];
UNUSED const struct mq_item *prev = &q->items[q->end - 1];
assert(prev->type != e || prev->location != loc);
}
#endif
@ -251,6 +251,10 @@ void q_skip_forward_to(struct mq *q, s64a min_loc) {
// Dump the contents of the given queue.
static never_inline UNUSED
void debugQueue(const struct mq *q) {
if (q == nullptr) {
DEBUG_PRINTF("q=NULL!\n");
return;
}
DEBUG_PRINTF("q=%p, nfa=%p\n", q, q->nfa);
DEBUG_PRINTF("q offset=%llu, buf={%p, len=%zu}, history={%p, len=%zu}\n",
q->offset, q->buffer, q->length, q->history, q->hlength);

2
src/nfa/nfa_api_util.h
View File

@ -68,7 +68,7 @@ void pushQueueAt(struct mq * restrict q, u32 pos, u32 e, s64a loc) {
// We assert that the event is different from its predecessor. If it's a
// dupe, you should have used the ordinary pushQueue call.
if (q->end) {
UNUSED struct mq_item *prev = &q->items[q->end - 1];
UNUSED const struct mq_item *prev = &q->items[q->end - 1];
assert(prev->type != e || prev->location != loc);
}
#endif

9
src/nfa/nfa_internal.h
View File

@ -1,5 +1,6 @@
/*
* Copyright (c) 2015-2020, Intel Corporation
* Copyright (c) 2024, VectorCamp PC
* Copyright (c) 2021, Arm Limited
*
* Redistribution and use in source and binary forms, with or without
@ -133,6 +134,7 @@ struct ALIGN_CL_DIRECTIVE NFA {
/* Note: implementation (e.g. a LimEx) directly follows struct in memory */
} ;
#ifndef __cplusplus
// Accessor macro for the implementation NFA: we do things this way to avoid
// type-punning warnings.
#define getImplNfa(nfa) \
@ -140,6 +142,13 @@ struct ALIGN_CL_DIRECTIVE NFA {
// Non-const version of the above, used at compile time.
#define getMutableImplNfa(nfa) ((char *)(nfa) + sizeof(struct NFA))
#else
// Same versions without C casts to avoid Cppcheck warnings
#define getImplNfa(nfa) \
(reinterpret_cast<const void *>(reinterpret_cast<const char *>(nfa) + sizeof(struct NFA)))
#define getMutableImplNfa(nfa) (reinterpret_cast<char *>(nfa) + sizeof(struct NFA))
#endif
static really_inline u32 nfaAcceptsEod(const struct NFA *nfa) {
return nfa->flags & NFA_ACCEPTS_EOD;

2
src/nfa/rdfa_graph.h
View File

@ -45,7 +45,7 @@ struct RdfaEdgeProps {
};
struct RdfaGraph : public ue2_graph<RdfaGraph, RdfaVertexProps, RdfaEdgeProps> {
RdfaGraph(const raw_dfa &rdfa);
explicit RdfaGraph(const raw_dfa &rdfa);
};

4
src/nfa/repeat.c
View File

@ -785,7 +785,7 @@ enum RepeatMatch repeatHasMatchRange(const struct RepeatInfo *info,
if (diff > info->repeatMax) {
DEBUG_PRINTF("range list is stale\n");
return REPEAT_STALE;
} else if (diff >= info->repeatMin && diff <= info->repeatMax) {
} else if (diff >= info->repeatMin) {
return REPEAT_MATCH;
}
@ -836,7 +836,7 @@ enum RepeatMatch repeatHasMatchBitmap(const struct RepeatInfo *info,
if (diff > info->repeatMax) {
DEBUG_PRINTF("stale\n");
return REPEAT_STALE;
} else if (diff >= info->repeatMin && diff <= info->repeatMax) {
} else if (diff >= info->repeatMin) {
return REPEAT_MATCH;
}

9
src/nfa/repeatcompile.cpp
View File

@ -94,9 +94,6 @@ u32 repeatRecurTable(struct RepeatStateInfo *info, const depth &repeatMax,
static
u32 findOptimalPatchSize(struct RepeatStateInfo *info, const depth &repeatMax,
const u32 minPeriod, u32 rv) {
u32 cnt = 0;
u32 patch_bits = 0;
u32 total_size = 0;
u32 min = ~0U;
u32 patch_len = 0;
@ -105,11 +102,11 @@ u32 findOptimalPatchSize(struct RepeatStateInfo *info, const depth &repeatMax,
}
for (u32 i = minPeriod; i <= rv; i++) {
cnt = ((u32)repeatMax + (i - 1)) / i + 1;
u32 cnt = ((u32)repeatMax + (i - 1)) / i + 1;
// no bit packing version
patch_bits = calcPackedBits(info->table[i]);
total_size = (patch_bits + 7U) / 8U * cnt;
u32 patch_bits = calcPackedBits(info->table[i]);
u32 total_size = (patch_bits + 7U) / 8U * cnt;
if (total_size < min) {
patch_len = i;

10
src/nfa/sheng.c
View File

@ -154,7 +154,7 @@ char fireReports(const struct sheng *sh, NfaCallback cb, void *ctxt,
return MO_CONTINUE_MATCHING; /* continue execution */
}
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
// Sheng32
static really_inline
const struct sheng32 *get_sheng32(const struct NFA *n) {
@ -351,7 +351,7 @@ char fireReports64(const struct sheng64 *sh, NfaCallback cb, void *ctxt,
}
return MO_CONTINUE_MATCHING; /* continue execution */
}
#endif // end of HAVE_AVX512VBMI
#endif // end of HAVE_AVX512VBMI || HAVE_SVE
/* include Sheng function definitions */
#include "sheng_defs.h"
@ -814,7 +814,6 @@ char nfaExecSheng_testEOD(const struct NFA *nfa, const char *state,
char nfaExecSheng_reportCurrent(const struct NFA *n, struct mq *q) {
const struct sheng *sh = (const struct sheng *)getImplNfa(n);
NfaCallback cb = q->cb;
void *ctxt = q->context;
u8 s = *(u8 *)q->state;
const struct sstate_aux *aux = get_aux(sh, s);
u64a offset = q_cur_offset(q);
@ -823,6 +822,7 @@ char nfaExecSheng_reportCurrent(const struct NFA *n, struct mq *q) {
assert(q_cur_type(q) == MQE_START);
if (aux->accept) {
void *ctxt = q->context;
if (sh->flags & SHENG_FLAG_SINGLE_REPORT) {
fireSingleReport(cb, ctxt, sh->report, offset);
} else {
@ -871,7 +871,7 @@ char nfaExecSheng_expandState(UNUSED const struct NFA *nfa, void *dest,
return 0;
}
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
// Sheng32
static really_inline
char runSheng32Cb(const struct sheng32 *sh, NfaCallback cb, void *ctxt,
@ -1874,4 +1874,4 @@ char nfaExecSheng64_expandState(UNUSED const struct NFA *nfa, void *dest,
*(u8 *)dest = *(const u8 *)src;
return 0;
}
#endif // end of HAVE_AVX512VBMI
#endif // end of HAVE_AVX512VBMI || HAVE_SVE

8
src/nfa/sheng.h
View File

@ -58,7 +58,7 @@ char nfaExecSheng_reportCurrent(const struct NFA *n, struct mq *q);
char nfaExecSheng_B(const struct NFA *n, u64a offset, const u8 *buffer,
size_t length, NfaCallback cb, void *context);
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#define nfaExecSheng32_B_Reverse NFA_API_NO_IMPL
#define nfaExecSheng32_zombie_status NFA_API_ZOMBIE_NO_IMPL
@ -106,8 +106,7 @@ char nfaExecSheng64_reportCurrent(const struct NFA *n, struct mq *q);
char nfaExecSheng64_B(const struct NFA *n, u64a offset, const u8 *buffer,
size_t length, NfaCallback cb, void *context);
#else // !HAVE_AVX512VBMI
#else // !HAVE_AVX512VBMI && !HAVE_SVE
#define nfaExecSheng32_B_Reverse NFA_API_NO_IMPL
#define nfaExecSheng32_zombie_status NFA_API_ZOMBIE_NO_IMPL
@ -138,6 +137,7 @@ char nfaExecSheng64_B(const struct NFA *n, u64a offset, const u8 *buffer,
#define nfaExecSheng64_testEOD NFA_API_NO_IMPL
#define nfaExecSheng64_reportCurrent NFA_API_NO_IMPL
#define nfaExecSheng64_B NFA_API_NO_IMPL
#endif // end of HAVE_AVX512VBMI
#endif // end of HAVE_AVX512VBMI || defined(HAVE_SVE)
#endif /* SHENG_H_ */

70
src/nfa/sheng_defs.h
View File

@ -52,7 +52,7 @@ u8 hasInterestingStates(const u8 a, const u8 b, const u8 c, const u8 d) {
return (a | b | c | d) & (SHENG_STATE_FLAG_MASK);
}
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
static really_inline
u8 isDeadState32(const u8 a) {
return a & SHENG32_STATE_DEAD;
@ -108,7 +108,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#define SHENG_IMPL sheng_cod
#define DEAD_FUNC isDeadState
#define ACCEPT_FUNC isAcceptState
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#define SHENG32_IMPL sheng32_cod
#define DEAD_FUNC32 isDeadState32
#define ACCEPT_FUNC32 isAcceptState32
@ -121,7 +121,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#undef SHENG_IMPL
#undef DEAD_FUNC
#undef ACCEPT_FUNC
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#undef SHENG32_IMPL
#undef DEAD_FUNC32
#undef ACCEPT_FUNC32
@ -135,7 +135,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#define SHENG_IMPL sheng_co
#define DEAD_FUNC dummyFunc
#define ACCEPT_FUNC isAcceptState
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#define SHENG32_IMPL sheng32_co
#define DEAD_FUNC32 dummyFunc
#define ACCEPT_FUNC32 isAcceptState32
@ -148,7 +148,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#undef SHENG_IMPL
#undef DEAD_FUNC
#undef ACCEPT_FUNC
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#undef SHENG32_IMPL
#undef DEAD_FUNC32
#undef ACCEPT_FUNC32
@ -162,7 +162,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#define SHENG_IMPL sheng_samd
#define DEAD_FUNC isDeadState
#define ACCEPT_FUNC isAcceptState
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#define SHENG32_IMPL sheng32_samd
#define DEAD_FUNC32 isDeadState32
#define ACCEPT_FUNC32 isAcceptState32
@ -175,7 +175,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#undef SHENG_IMPL
#undef DEAD_FUNC
#undef ACCEPT_FUNC
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#undef SHENG32_IMPL
#undef DEAD_FUNC32
#undef ACCEPT_FUNC32
@ -189,7 +189,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#define SHENG_IMPL sheng_sam
#define DEAD_FUNC dummyFunc
#define ACCEPT_FUNC isAcceptState
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#define SHENG32_IMPL sheng32_sam
#define DEAD_FUNC32 dummyFunc
#define ACCEPT_FUNC32 isAcceptState32
@ -202,7 +202,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#undef SHENG_IMPL
#undef DEAD_FUNC
#undef ACCEPT_FUNC
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#undef SHENG32_IMPL
#undef DEAD_FUNC32
#undef ACCEPT_FUNC32
@ -216,7 +216,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#define SHENG_IMPL sheng_nmd
#define DEAD_FUNC isDeadState
#define ACCEPT_FUNC dummyFunc
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#define SHENG32_IMPL sheng32_nmd
#define DEAD_FUNC32 isDeadState32
#define ACCEPT_FUNC32 dummyFunc
@ -229,7 +229,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#undef SHENG_IMPL
#undef DEAD_FUNC
#undef ACCEPT_FUNC
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#undef SHENG32_IMPL
#undef DEAD_FUNC32
#undef ACCEPT_FUNC32
@ -243,7 +243,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#define SHENG_IMPL sheng_nm
#define DEAD_FUNC dummyFunc
#define ACCEPT_FUNC dummyFunc
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#define SHENG32_IMPL sheng32_nm
#define DEAD_FUNC32 dummyFunc
#define ACCEPT_FUNC32 dummyFunc
@ -256,7 +256,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#undef SHENG_IMPL
#undef DEAD_FUNC
#undef ACCEPT_FUNC
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#undef SHENG32_IMPL
#undef DEAD_FUNC32
#undef ACCEPT_FUNC32
@ -277,7 +277,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#define INNER_ACCEL_FUNC isAccelState
#define OUTER_ACCEL_FUNC dummyFunc
#define ACCEPT_FUNC isAcceptState
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#define SHENG32_IMPL sheng32_4_coda
#define INTERESTING_FUNC32 hasInterestingStates32
#define INNER_DEAD_FUNC32 isDeadState32
@ -296,7 +296,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#undef INNER_ACCEL_FUNC
#undef OUTER_ACCEL_FUNC
#undef ACCEPT_FUNC
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#undef SHENG32_IMPL
#undef INTERESTING_FUNC32
#undef INNER_DEAD_FUNC32
@ -316,7 +316,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#define INNER_ACCEL_FUNC dummyFunc
#define OUTER_ACCEL_FUNC dummyFunc
#define ACCEPT_FUNC isAcceptState
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#define SHENG32_IMPL sheng32_4_cod
#define INTERESTING_FUNC32 hasInterestingStates32
#define INNER_DEAD_FUNC32 isDeadState32
@ -339,7 +339,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#undef INNER_ACCEL_FUNC
#undef OUTER_ACCEL_FUNC
#undef ACCEPT_FUNC
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#undef SHENG32_IMPL
#undef INTERESTING_FUNC32
#undef INNER_DEAD_FUNC32
@ -363,7 +363,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#define INNER_ACCEL_FUNC isAccelState
#define OUTER_ACCEL_FUNC dummyFunc
#define ACCEPT_FUNC isAcceptState
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#define SHENG32_IMPL sheng32_4_coa
#define INTERESTING_FUNC32 hasInterestingStates32
#define INNER_DEAD_FUNC32 dummyFunc
@ -382,7 +382,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#undef INNER_ACCEL_FUNC
#undef OUTER_ACCEL_FUNC
#undef ACCEPT_FUNC
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#undef SHENG32_IMPL
#undef INTERESTING_FUNC32
#undef INNER_DEAD_FUNC32
@ -402,7 +402,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#define INNER_ACCEL_FUNC dummyFunc
#define OUTER_ACCEL_FUNC dummyFunc
#define ACCEPT_FUNC isAcceptState
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#define SHENG32_IMPL sheng32_4_co
#define INTERESTING_FUNC32 hasInterestingStates32
#define INNER_DEAD_FUNC32 dummyFunc
@ -425,7 +425,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#undef INNER_ACCEL_FUNC
#undef OUTER_ACCEL_FUNC
#undef ACCEPT_FUNC
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#undef SHENG32_IMPL
#undef INTERESTING_FUNC32
#undef INNER_DEAD_FUNC32
@ -449,7 +449,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#define INNER_ACCEL_FUNC isAccelState
#define OUTER_ACCEL_FUNC dummyFunc
#define ACCEPT_FUNC isAcceptState
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#define SHENG32_IMPL sheng32_4_samda
#define INTERESTING_FUNC32 hasInterestingStates32
#define INNER_DEAD_FUNC32 isDeadState32
@ -468,7 +468,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#undef INNER_ACCEL_FUNC
#undef OUTER_ACCEL_FUNC
#undef ACCEPT_FUNC
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#undef SHENG32_IMPL
#undef INTERESTING_FUNC32
#undef INNER_DEAD_FUNC32
@ -488,7 +488,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#define INNER_ACCEL_FUNC dummyFunc
#define OUTER_ACCEL_FUNC dummyFunc
#define ACCEPT_FUNC isAcceptState
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#define SHENG32_IMPL sheng32_4_samd
#define INTERESTING_FUNC32 hasInterestingStates32
#define INNER_DEAD_FUNC32 isDeadState32
@ -511,7 +511,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#undef INNER_ACCEL_FUNC
#undef OUTER_ACCEL_FUNC
#undef ACCEPT_FUNC
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#undef SHENG32_IMPL
#undef INTERESTING_FUNC32
#undef INNER_DEAD_FUNC32
@ -535,7 +535,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#define INNER_ACCEL_FUNC isAccelState
#define OUTER_ACCEL_FUNC dummyFunc
#define ACCEPT_FUNC isAcceptState
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#define SHENG32_IMPL sheng32_4_sama
#define INTERESTING_FUNC32 hasInterestingStates32
#define INNER_DEAD_FUNC32 dummyFunc
@ -554,7 +554,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#undef INNER_ACCEL_FUNC
#undef OUTER_ACCEL_FUNC
#undef ACCEPT_FUNC
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#undef SHENG32_IMPL
#undef INTERESTING_FUNC32
#undef INNER_DEAD_FUNC32
@ -574,7 +574,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#define INNER_ACCEL_FUNC dummyFunc
#define OUTER_ACCEL_FUNC dummyFunc
#define ACCEPT_FUNC isAcceptState
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#define SHENG32_IMPL sheng32_4_sam
#define INTERESTING_FUNC32 hasInterestingStates32
#define INNER_DEAD_FUNC32 dummyFunc
@ -597,7 +597,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#undef INNER_ACCEL_FUNC
#undef OUTER_ACCEL_FUNC
#undef ACCEPT_FUNC
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#undef SHENG32_IMPL
#undef INTERESTING_FUNC32
#undef INNER_DEAD_FUNC32
@ -623,7 +623,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#define INNER_ACCEL_FUNC dummyFunc
#define OUTER_ACCEL_FUNC isAccelState
#define ACCEPT_FUNC dummyFunc
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#define SHENG32_IMPL sheng32_4_nmda
#define INTERESTING_FUNC32 dummyFunc4
#define INNER_DEAD_FUNC32 dummyFunc
@ -642,7 +642,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#undef INNER_ACCEL_FUNC
#undef OUTER_ACCEL_FUNC
#undef ACCEPT_FUNC
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#undef SHENG32_IMPL
#undef INTERESTING_FUNC32
#undef INNER_DEAD_FUNC32
@ -662,7 +662,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#define INNER_ACCEL_FUNC dummyFunc
#define OUTER_ACCEL_FUNC dummyFunc
#define ACCEPT_FUNC dummyFunc
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#define SHENG32_IMPL sheng32_4_nmd
#define INTERESTING_FUNC32 dummyFunc4
#define INNER_DEAD_FUNC32 dummyFunc
@ -685,7 +685,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#undef INNER_ACCEL_FUNC
#undef OUTER_ACCEL_FUNC
#undef ACCEPT_FUNC
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#undef SHENG32_IMPL
#undef INTERESTING_FUNC32
#undef INNER_DEAD_FUNC32
@ -712,7 +712,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#define INNER_ACCEL_FUNC dummyFunc
#define OUTER_ACCEL_FUNC dummyFunc
#define ACCEPT_FUNC dummyFunc
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#define SHENG32_IMPL sheng32_4_nm
#define INTERESTING_FUNC32 dummyFunc4
#define INNER_DEAD_FUNC32 dummyFunc
@ -735,7 +735,7 @@ u8 dummyFunc(UNUSED const u8 a) {
#undef INNER_ACCEL_FUNC
#undef OUTER_ACCEL_FUNC
#undef ACCEPT_FUNC
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
#undef SHENG32_IMPL
#undef INTERESTING_FUNC32
#undef INNER_DEAD_FUNC32

38
src/nfa/sheng_impl.h
View File

@ -96,7 +96,7 @@ char SHENG_IMPL(u8 *state, NfaCallback cb, void *ctxt, const struct sheng *s,
return MO_CONTINUE_MATCHING;
}
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
static really_inline
char SHENG32_IMPL(u8 *state, NfaCallback cb, void *ctxt,
const struct sheng32 *s,
@ -114,14 +114,28 @@ char SHENG32_IMPL(u8 *state, NfaCallback cb, void *ctxt,
}
DEBUG_PRINTF("Scanning %lli bytes\n", (s64a)(end - start));
#if defined(HAVE_SVE)
const svbool_t lane_pred_32 = svwhilelt_b8(0, 32);
svuint8_t cur_state = svdup_u8(*state);
svuint8_t tbl_mask = svdup_u8((unsigned char)0x1F);
const m512 *masks = s->succ_masks;
#else
m512 cur_state = set1_64x8(*state);
const m512 *masks = s->succ_masks;
#endif
while (likely(cur_buf != end)) {
const u8 c = *cur_buf;
#if defined(HAVE_SVE)
svuint8_t succ_mask = svld1(lane_pred_32, (const u8*)(masks + c));
cur_state = svtbl(succ_mask, svand_x(svptrue_b8(), tbl_mask, cur_state));
const u8 tmp = svlastb(lane_pred_32, cur_state);
#else
const m512 succ_mask = masks[c];
cur_state = vpermb512(cur_state, succ_mask);
const u8 tmp = movd512(cur_state);
#endif
DEBUG_PRINTF("c: %02hhx '%c'\n", c, ourisprint(c) ? c : '?');
DEBUG_PRINTF("s: %u (flag: %u)\n", tmp & SHENG32_STATE_MASK,
@ -153,7 +167,11 @@ char SHENG32_IMPL(u8 *state, NfaCallback cb, void *ctxt,
}
cur_buf++;
}
#if defined(HAVE_SVE)
*state = svlastb(lane_pred_32, cur_state);
#else
*state = movd512(cur_state);
#endif
*scan_end = cur_buf;
return MO_CONTINUE_MATCHING;
}
@ -175,14 +193,28 @@ char SHENG64_IMPL(u8 *state, NfaCallback cb, void *ctxt,
}
DEBUG_PRINTF("Scanning %lli bytes\n", (s64a)(end - start));
#if defined(HAVE_SVE)
const svbool_t lane_pred_64 = svwhilelt_b8(0, 64);
svuint8_t cur_state = svdup_u8(*state);
svuint8_t tbl_mask = svdup_u8((unsigned char)0x3F);
const m512 *masks = s->succ_masks;
#else
m512 cur_state = set1_64x8(*state);
const m512 *masks = s->succ_masks;
#endif
while (likely(cur_buf != end)) {
const u8 c = *cur_buf;
#if defined(HAVE_SVE)
svuint8_t succ_mask = svld1(lane_pred_64, (const u8*)(masks + c));
cur_state = svtbl(succ_mask, svand_x(svptrue_b8(), tbl_mask, cur_state));
const u8 tmp = svlastb(lane_pred_64, cur_state);
#else
const m512 succ_mask = masks[c];
cur_state = vpermb512(cur_state, succ_mask);
const u8 tmp = movd512(cur_state);
#endif
DEBUG_PRINTF("c: %02hhx '%c'\n", c, ourisprint(c) ? c : '?');
DEBUG_PRINTF("s: %u (flag: %u)\n", tmp & SHENG64_STATE_MASK,
@ -214,7 +246,11 @@ char SHENG64_IMPL(u8 *state, NfaCallback cb, void *ctxt,
}
cur_buf++;
}
#if defined(HAVE_SVE)
*state = svlastb(lane_pred_64, cur_state);
#else
*state = movd512(cur_state);
#endif
*scan_end = cur_buf;
return MO_CONTINUE_MATCHING;
}

61
src/nfa/sheng_impl4.h
View File

@ -283,7 +283,7 @@ char SHENG_IMPL(u8 *state, NfaCallback cb, void *ctxt, const struct sheng *s,
return MO_CONTINUE_MATCHING;
}
#if defined(HAVE_AVX512VBMI)
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
static really_inline
char SHENG32_IMPL(u8 *state, NfaCallback cb, void *ctxt,
const struct sheng32 *s,
@ -320,8 +320,15 @@ char SHENG32_IMPL(u8 *state, NfaCallback cb, void *ctxt,
return MO_CONTINUE_MATCHING;
}
#if defined(HAVE_SVE)
const svbool_t lane_pred_32 = svwhilelt_b8(0, 32);
svuint8_t cur_state = svdup_u8(*state);
svuint8_t tbl_mask = svdup_u8((unsigned char)0x1F);
const m512 *masks = s->succ_masks;
#else
m512 cur_state = set1_64x8(*state);
const m512 *masks = s->succ_masks;
#endif
while (likely(end - cur_buf >= 4)) {
const u8 *b1 = cur_buf;
@ -333,6 +340,23 @@ char SHENG32_IMPL(u8 *state, NfaCallback cb, void *ctxt,
const u8 c3 = *b3;
const u8 c4 = *b4;
#if defined(HAVE_SVE)
svuint8_t succ_mask1 = svld1(lane_pred_32, (const u8*)(masks+c1));
cur_state = svtbl(succ_mask1, svand_x(svptrue_b8(), tbl_mask, cur_state));
const u8 a1 = svlastb(lane_pred_32, cur_state);
svuint8_t succ_mask2 = svld1(lane_pred_32, (const u8*)(masks+c2));
cur_state = svtbl(succ_mask2, svand_x(svptrue_b8(), tbl_mask, cur_state));
const u8 a2 = svlastb(lane_pred_32, cur_state);
svuint8_t succ_mask3 = svld1(lane_pred_32, (const u8*)(masks+c3));
cur_state = svtbl(succ_mask3, svand_x(svptrue_b8(), tbl_mask, cur_state));
const u8 a3 = svlastb(lane_pred_32, cur_state);
svuint8_t succ_mask4 = svld1(lane_pred_32, (const u8*)(masks+c4));
cur_state = svtbl(succ_mask4, svand_x(svptrue_b8(), tbl_mask, cur_state));
const u8 a4 = svlastb(lane_pred_32, cur_state);
#else
const m512 succ_mask1 = masks[c1];
cur_state = vpermb512(cur_state, succ_mask1);
const u8 a1 = movd512(cur_state);
@ -348,6 +372,7 @@ char SHENG32_IMPL(u8 *state, NfaCallback cb, void *ctxt,
const m512 succ_mask4 = masks[c4];
cur_state = vpermb512(cur_state, succ_mask4);
const u8 a4 = movd512(cur_state);
#endif
DEBUG_PRINTF("c: %02hhx '%c'\n", c1, ourisprint(c1) ? c1 : '?');
DEBUG_PRINTF("s: %u (flag: %u)\n", a1 & SHENG32_STATE_MASK,
@ -517,7 +542,11 @@ char SHENG32_IMPL(u8 *state, NfaCallback cb, void *ctxt,
};
cur_buf += 4;
}
#if defined(HAVE_SVE)
*state = svlastb(lane_pred_32, cur_state);
#else
*state = movd512(cur_state);
#endif
*scan_end = cur_buf;
return MO_CONTINUE_MATCHING;
}
@ -541,9 +570,15 @@ char SHENG64_IMPL(u8 *state, NfaCallback cb, void *ctxt,
*scan_end = end;
return MO_CONTINUE_MATCHING;
}
#if defined(HAVE_SVE)
const svbool_t lane_pred_64 = svwhilelt_b8(0, 64);
svuint8_t cur_state = svdup_u8(*state);
svuint8_t tbl_mask = svdup_u8((unsigned char)0x3F);
const m512 *masks = s->succ_masks;
#else
m512 cur_state = set1_64x8(*state);
const m512 *masks = s->succ_masks;
#endif
while (likely(end - cur_buf >= 4)) {
const u8 *b1 = cur_buf;
@ -555,6 +590,23 @@ char SHENG64_IMPL(u8 *state, NfaCallback cb, void *ctxt,
const u8 c3 = *b3;
const u8 c4 = *b4;
#if defined(HAVE_SVE)
svuint8_t succ_mask1 = svld1(lane_pred_64, (const u8*)(masks+c1));
cur_state = svtbl(succ_mask1, svand_x(svptrue_b8(), tbl_mask, cur_state));
const u8 a1 = svlastb(lane_pred_64, cur_state);
svuint8_t succ_mask2 = svld1(lane_pred_64, (const u8*)(masks+c2));
cur_state = svtbl(succ_mask2, svand_x(svptrue_b8(), tbl_mask, cur_state));
const u8 a2 = svlastb(lane_pred_64, cur_state);
svuint8_t succ_mask3 = svld1(lane_pred_64, (const u8*)(masks+c3));
cur_state = svtbl(succ_mask3, svand_x(svptrue_b8(), tbl_mask, cur_state));
const u8 a3 = svlastb(lane_pred_64, cur_state);
svuint8_t succ_mask4 = svld1(lane_pred_64, (const u8*)(masks+c4));
cur_state = svtbl(succ_mask4, svand_x(svptrue_b8(), tbl_mask, cur_state));
const u8 a4 = svlastb(lane_pred_64, cur_state);
#else
const m512 succ_mask1 = masks[c1];
cur_state = vpermb512(cur_state, succ_mask1);
const u8 a1 = movd512(cur_state);
@ -570,6 +622,7 @@ char SHENG64_IMPL(u8 *state, NfaCallback cb, void *ctxt,
const m512 succ_mask4 = masks[c4];
cur_state = vpermb512(cur_state, succ_mask4);
const u8 a4 = movd512(cur_state);
#endif
DEBUG_PRINTF("c: %02hhx '%c'\n", c1, ourisprint(c1) ? c1 : '?');
DEBUG_PRINTF("s: %u (flag: %u)\n", a1 & SHENG64_STATE_MASK,
@ -703,7 +756,11 @@ char SHENG64_IMPL(u8 *state, NfaCallback cb, void *ctxt,
}
cur_buf += 4;
}
#if defined(HAVE_SVE)
*state = svlastb(lane_pred_64, cur_state);
#else
*state = movd512(cur_state);
#endif
*scan_end = cur_buf;
return MO_CONTINUE_MATCHING;
}

84
src/nfa/shengcompile.cpp
View File

@ -99,7 +99,7 @@ struct dfa_info {
return next(idx, TOP);
}
dstate &next(dstate_id_t idx, u16 chr) {
auto &src = (*this)[idx];
const auto &src = (*this)[idx];
auto next_id = src.next[raw.alpha_remap[chr]];
return states[next_id];
}
@ -109,7 +109,7 @@ struct dfa_info {
// if DFA can't die, shift all indices left by 1
return can_die ? idx : idx + 1;
}
bool isDead(dstate &state) {
bool isDead(const dstate &state) {
return raw_id(state.impl_id) == DEAD_STATE;
}
bool isDead(dstate_id_t idx) {
@ -117,7 +117,7 @@ struct dfa_info {
}
private:
static bool dfaCanDie(raw_dfa &rdfa) {
static bool dfaCanDie(const raw_dfa &rdfa) {
for (unsigned chr = 0; chr < 256; chr++) {
for (dstate_id_t state = 0; state < rdfa.states.size(); state++) {
auto succ = rdfa.states[state].next[rdfa.alpha_remap[chr]];
@ -138,7 +138,7 @@ struct raw_report_list {
raw_report_list(const flat_set<ReportID> &reports_in,
const ReportManager &rm, bool do_remap) {
if (do_remap) {
for (auto &id : reports_in) {
for (const auto &id : reports_in) {
reports.insert(rm.getProgramOffset(id));
}
} else {
@ -334,14 +334,14 @@ void fillAccelOut(const map<dstate_id_t, AccelScheme> &accel_escape_info,
template <typename T>
static
u8 getShengState(UNUSED dstate &state, UNUSED dfa_info &info,
UNUSED map<dstate_id_t, AccelScheme> &accelInfo) {
u8 getShengState(UNUSED const dstate &state, UNUSED dfa_info &info,
UNUSED const map<dstate_id_t, AccelScheme> &accelInfo) {
return 0;
}
template <>
u8 getShengState<sheng>(dstate &state, dfa_info &info,
map<dstate_id_t, AccelScheme> &accelInfo) {
u8 getShengState<sheng>(const dstate &state, dfa_info &info,
const map<dstate_id_t, AccelScheme> &accelInfo) {
u8 s = state.impl_id;
if (!state.reports.empty()) {
s |= SHENG_STATE_ACCEPT;
@ -356,8 +356,8 @@ u8 getShengState<sheng>(dstate &state, dfa_info &info,
}
template <>
u8 getShengState<sheng32>(dstate &state, dfa_info &info,
map<dstate_id_t, AccelScheme> &accelInfo) {
u8 getShengState<sheng32>(const dstate &state, dfa_info &info,
const map<dstate_id_t, AccelScheme> &accelInfo) {
u8 s = state.impl_id;
if (!state.reports.empty()) {
s |= SHENG32_STATE_ACCEPT;
@ -372,8 +372,8 @@ u8 getShengState<sheng32>(dstate &state, dfa_info &info,
}
template <>
u8 getShengState<sheng64>(dstate &state, dfa_info &info,
UNUSED map<dstate_id_t, AccelScheme> &accelInfo) {
u8 getShengState<sheng64>(const dstate &state, dfa_info &info,
UNUSED const map<dstate_id_t, AccelScheme> &accelInfo) {
u8 s = state.impl_id;
if (!state.reports.empty()) {
s |= SHENG64_STATE_ACCEPT;
@ -409,8 +409,8 @@ void fillAccelAux(struct NFA *n, dfa_info &info,
template <typename T>
static
void populateBasicInfo(UNUSED struct NFA *n, UNUSED dfa_info &info,
UNUSED map<dstate_id_t, AccelScheme> &accelInfo,
void populateBasicInfo(UNUSED struct NFA *n, UNUSED dfa_info &info, // cppcheck-suppress constParameterPointer
UNUSED const map<dstate_id_t, AccelScheme> &accelInfo,
UNUSED u32 aux_offset, UNUSED u32 report_offset,
UNUSED u32 accel_offset, UNUSED u32 total_size,
UNUSED u32 dfa_size) {
@ -418,7 +418,7 @@ void populateBasicInfo(UNUSED struct NFA *n, UNUSED dfa_info &info,
template <>
void populateBasicInfo<sheng>(struct NFA *n, dfa_info &info,
map<dstate_id_t, AccelScheme> &accelInfo,
const map<dstate_id_t, AccelScheme> &accelInfo,
u32 aux_offset, u32 report_offset,
u32 accel_offset, u32 total_size,
u32 dfa_size) {
@ -443,7 +443,7 @@ void populateBasicInfo<sheng>(struct NFA *n, dfa_info &info,
template <>
void populateBasicInfo<sheng32>(struct NFA *n, dfa_info &info,
map<dstate_id_t, AccelScheme> &accelInfo,
const map<dstate_id_t, AccelScheme> &accelInfo,
u32 aux_offset, u32 report_offset,
u32 accel_offset, u32 total_size,
u32 dfa_size) {
@ -468,7 +468,7 @@ void populateBasicInfo<sheng32>(struct NFA *n, dfa_info &info,
template <>
void populateBasicInfo<sheng64>(struct NFA *n, dfa_info &info,
map<dstate_id_t, AccelScheme> &accelInfo,
const map<dstate_id_t, AccelScheme> &accelInfo,
u32 aux_offset, u32 report_offset,
u32 accel_offset, u32 total_size,
u32 dfa_size) {
@ -551,19 +551,19 @@ void fillSingleReport(NFA *n, ReportID r_id) {
template <typename T>
static
bool createShuffleMasks(UNUSED T *s, UNUSED dfa_info &info,
UNUSED map<dstate_id_t, AccelScheme> &accelInfo) {
bool createShuffleMasks(UNUSED T *s, UNUSED dfa_info &info, // cppcheck-suppress constParameterPointer
UNUSED const map<dstate_id_t, AccelScheme> &accelInfo) {
return true;
}
template <>
bool createShuffleMasks<sheng>(sheng *s, dfa_info &info,
map<dstate_id_t, AccelScheme> &accelInfo) {
const map<dstate_id_t, AccelScheme> &accelInfo) {
for (u16 chr = 0; chr < 256; chr++) {
u8 buf[16] = {0};
for (dstate_id_t idx = 0; idx < info.size(); idx++) {
auto &succ_state = info.next(idx, chr);
const auto &succ_state = info.next(idx, chr);
buf[idx] = getShengState<sheng>(succ_state, info, accelInfo);
}
@ -577,13 +577,13 @@ bool createShuffleMasks<sheng>(sheng *s, dfa_info &info,
template <>
bool createShuffleMasks<sheng32>(sheng32 *s, dfa_info &info,
map<dstate_id_t, AccelScheme> &accelInfo) {
const map<dstate_id_t, AccelScheme> &accelInfo) {
for (u16 chr = 0; chr < 256; chr++) {
u8 buf[64] = {0};
assert(info.size() <= 32);
for (dstate_id_t idx = 0; idx < info.size(); idx++) {
auto &succ_state = info.next(idx, chr);
const auto &succ_state = info.next(idx, chr);
buf[idx] = getShengState<sheng32>(succ_state, info, accelInfo);
buf[32 + idx] = buf[idx];
@ -598,13 +598,13 @@ bool createShuffleMasks<sheng32>(sheng32 *s, dfa_info &info,
template <>
bool createShuffleMasks<sheng64>(sheng64 *s, dfa_info &info,
map<dstate_id_t, AccelScheme> &accelInfo) {
const map<dstate_id_t, AccelScheme> &accelInfo) {
for (u16 chr = 0; chr < 256; chr++) {
u8 buf[64] = {0};
assert(info.size() <= 64);
for (dstate_id_t idx = 0; idx < info.size(); idx++) {
auto &succ_state = info.next(idx, chr);
const auto &succ_state = info.next(idx, chr);
if (accelInfo.find(info.raw_id(succ_state.impl_id))
!= accelInfo.end()) {
@ -690,7 +690,7 @@ bytecode_ptr<NFA> shengCompile_int(raw_dfa &raw, const CompileContext &cc,
}
if (!createShuffleMasks<T>((T *)getMutableImplNfa(nfa.get()), info, accelInfo)) {
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
return nfa;
@ -701,7 +701,7 @@ bytecode_ptr<NFA> shengCompile(raw_dfa &raw, const CompileContext &cc,
set<dstate_id_t> *accel_states) {
if (!cc.grey.allowSheng) {
DEBUG_PRINTF("Sheng is not allowed!\n");
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
sheng_build_strat strat(raw, rm, only_accel_init);
@ -716,7 +716,7 @@ bytecode_ptr<NFA> shengCompile(raw_dfa &raw, const CompileContext &cc,
info.can_die ? "can" : "cannot", info.size());
if (info.size() > 16) {
DEBUG_PRINTF("Too many states\n");
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
return shengCompile_int<sheng>(raw, cc, accel_states, strat, info);
@ -727,13 +727,20 @@ bytecode_ptr<NFA> sheng32Compile(raw_dfa &raw, const CompileContext &cc,
set<dstate_id_t> *accel_states) {
if (!cc.grey.allowSheng) {
DEBUG_PRINTF("Sheng is not allowed!\n");
return nullptr;
bytecode_ptr<NFA>(nullptr);
}
#ifdef HAVE_SVE
if (svcntb()<32) {
DEBUG_PRINTF("Sheng32 failed, SVE width is too small!\n");
bytecode_ptr<NFA>(nullptr);
}
#else
if (!cc.target_info.has_avx512vbmi()) {
DEBUG_PRINTF("Sheng32 failed, no HS_CPU_FEATURES_AVX512VBMI!\n");
return nullptr;
bytecode_ptr<NFA>(nullptr);
}
#endif
sheng_build_strat strat(raw, rm, only_accel_init);
dfa_info info(strat);
@ -748,7 +755,7 @@ bytecode_ptr<NFA> sheng32Compile(raw_dfa &raw, const CompileContext &cc,
assert(info.size() > 16);
if (info.size() > 32) {
DEBUG_PRINTF("Too many states\n");
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
return shengCompile_int<sheng32>(raw, cc, accel_states, strat, info);
@ -759,13 +766,20 @@ bytecode_ptr<NFA> sheng64Compile(raw_dfa &raw, const CompileContext &cc,
set<dstate_id_t> *accel_states) {
if (!cc.grey.allowSheng) {
DEBUG_PRINTF("Sheng is not allowed!\n");
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
#ifdef HAVE_SVE
if (svcntb()<64) {
DEBUG_PRINTF("Sheng64 failed, SVE width is too small!\n");
return bytecode_ptr<NFA>(nullptr);
}
#else
if (!cc.target_info.has_avx512vbmi()) {
DEBUG_PRINTF("Sheng64 failed, no HS_CPU_FEATURES_AVX512VBMI!\n");
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
#endif
sheng_build_strat strat(raw, rm, only_accel_init);
dfa_info info(strat);
@ -780,13 +794,13 @@ bytecode_ptr<NFA> sheng64Compile(raw_dfa &raw, const CompileContext &cc,
assert(info.size() > 32);
if (info.size() > 64) {
DEBUG_PRINTF("Too many states\n");
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
vector<dstate> old_states;
old_states = info.states;
auto nfa = shengCompile_int<sheng64>(raw, cc, accel_states, strat, info);
if (!nfa) {
info.states = old_states;
info.states = old_states; // cppcheck-suppress unreadVariable
}
return nfa;
}

4
src/nfa/shufti_simd.hpp
View File

@ -264,7 +264,7 @@ const u8 *shuftiDoubleExecReal(m128 mask1_lo, m128 mask1_hi, m128 mask2_lo, m128
const u8 *shuftiExec(m128 mask_lo, m128 mask_hi, const u8 *buf,
const u8 *buf_end) {
if (buf_end - buf < VECTORSIZE) {
return shuftiFwdSlow((const u8 *)&mask_lo, (const u8 *)&mask_hi, buf, buf_end);
return shuftiFwdSlow(reinterpret_cast<const u8 *>(&mask_lo), reinterpret_cast<const u8 *>(&mask_hi), buf, buf_end);
}
return shuftiExecReal<VECTORSIZE>(mask_lo, mask_hi, buf, buf_end);
}
@ -272,7 +272,7 @@ const u8 *shuftiExec(m128 mask_lo, m128 mask_hi, const u8 *buf,
const u8 *rshuftiExec(m128 mask_lo, m128 mask_hi, const u8 *buf,
const u8 *buf_end) {
if (buf_end - buf < VECTORSIZE) {
return shuftiRevSlow((const u8 *)&mask_lo, (const u8 *)&mask_hi, buf, buf_end);
return shuftiRevSlow(reinterpret_cast<const u8 *>(&mask_lo), reinterpret_cast<const u8 *>(&mask_hi), buf, buf_end);
}
return rshuftiExecReal<VECTORSIZE>(mask_lo, mask_hi, buf, buf_end);
}

6
src/nfa/tamaramacompile.cpp
View File

@ -32,6 +32,8 @@
*/
#include "config.h"
#include <numeric>
#include "tamaramacompile.h"
@ -129,14 +131,10 @@ buildTamarama(const TamaInfo &tamaInfo, const u32 queue,
sizeof(u32) * subSize + 64; // offsets to subengines in bytecode and
// padding for subengines
auto subl = [](size_t z, NFA *sub) {
return z + (size_t)(ROUNDUP_CL(sub->length));
};
total_size += std::accumulate(tamaInfo.subengines.begin(), tamaInfo.subengines.end(), 0, subl);
// for (const auto &sub : tamaInfo.subengines) {
// total_size += ROUNDUP_CL(sub->length);
// }
// use subSize as a sentinel value for no active subengines,
// so add one to subSize here

2
src/nfa/truffle_simd.hpp
View File

@ -227,7 +227,7 @@ const u8 *fwdBlock(SuperVector<S> shuf_mask_lo_highclear, SuperVector<S> shuf_ma
}
template <uint16_t S>
const u8 *truffleExecReal(m128 &shuf_mask_lo_highclear, m128 shuf_mask_lo_highset, const u8 *buf, const u8 *buf_end) {
const u8 *truffleExecReal(const m128 &shuf_mask_lo_highclear, m128 shuf_mask_lo_highset, const u8 *buf, const u8 *buf_end) {
assert(buf && buf_end);
assert(buf < buf_end);
DEBUG_PRINTF("truffle %p len %zu\n", buf, buf_end - buf);

5
src/nfagraph/ng.cpp
View File

@ -193,9 +193,6 @@ void reduceGraph(NGHolder &g, som_type som, bool utf8,
if (!som) {
mergeCyclicDotStars(g);
}
if (!som) {
removeSiblingsOfStartDotStar(g);
}
}
@ -292,7 +289,7 @@ bool addComponent(NG &ng, NGHolder &g, const ExpressionInfo &expr,
// Returns true if all components have been added.
static
bool processComponents(NG &ng, ExpressionInfo &expr,
bool processComponents(NG &ng, const ExpressionInfo &expr,
deque<unique_ptr<NGHolder>> &g_comp,
const som_type som) {
const u32 num_components = g_comp.size();

20
src/nfagraph/ng_anchored_dots.cpp
View File

@ -166,9 +166,9 @@ void reformAnchoredRepeatsComponent(NGHolder &g,
return;
}
NFAVertex dotV = NGHolder::null_vertex();
set<NFAVertex> otherV;
dotV = findReformable(g, compAnchoredStarts, otherV);
NFAVertex dotV = findReformable(g, compAnchoredStarts, otherV);
if (dotV == NGHolder::null_vertex()) {
DEBUG_PRINTF("no candidate reformable dot found.\n");
return;
@ -258,7 +258,7 @@ void reformAnchoredRepeatsComponent(NGHolder &g,
static
void reformUnanchoredRepeatsComponent(NGHolder &g,
set<NFAVertex> &compAnchoredStarts,
const set<NFAVertex> &compAnchoredStarts,
set<NFAVertex> &compUnanchoredStarts,
set<NFAVertex> &dead,
depth *startBegin, depth *startEnd) {
@ -269,9 +269,9 @@ void reformUnanchoredRepeatsComponent(NGHolder &g,
}
while (true) {
NFAVertex dotV = NGHolder::null_vertex();
set<NFAVertex> otherV;
dotV = findReformable(g, compUnanchoredStarts, otherV);
NFAVertex dotV = findReformable(g, compUnanchoredStarts, otherV);
if (dotV == NGHolder::null_vertex()) {
DEBUG_PRINTF("no candidate reformable dot found.\n");
return;
@ -488,15 +488,15 @@ void collapseVariableDotRepeat(NGHolder &g, NFAVertex start,
// Collect all the other optional dot vertices and the successor vertices
// by walking down the graph from initialDot
set<NFAVertex> dots, succ;
if (!gatherParticipants(g, start, initialDot, dots, succ)) {
set<NFAVertex> dots, succr;
if (!gatherParticipants(g, start, initialDot, dots, succr)) {
DEBUG_PRINTF("gatherParticipants failed\n");
return;
}
DEBUG_PRINTF("optional dot repeat with %zu participants, "
"terminating in %zu non-dot nodes\n",
dots.size(), succ.size());
dots.size(), succr.size());
// Remove all the participants and set the start offset
dead.insert(dots.begin(), dots.end());
@ -512,7 +512,7 @@ void collapseVariableDotRepeat(NGHolder &g, NFAVertex start,
assert(startEnd->is_reachable());
// Connect our successor vertices to both start and startDs.
for (auto v : succ) {
for (auto v : succr) {
add_edge_if_not_present(g.start, v, g);
add_edge_if_not_present(g.startDs, v, g);
}
@ -558,7 +558,7 @@ void collapseVariableRepeats(NGHolder &g, depth *startBegin, depth *startEnd) {
}
static
void addDotsBetween(NGHolder &g, NFAVertex lhs, vector<NFAVertex> &rhs,
void addDotsBetween(NGHolder &g, NFAVertex lhs, const vector<NFAVertex> &rhs,
depth min_repeat, depth max_repeat) {
const bool unbounded = max_repeat.is_infinite();
if (unbounded) {

24
src/nfagraph/ng_asserts.cpp
View File

@ -92,11 +92,12 @@ static const CharReach CHARREACH_NONWORD_UCP_PRE(CHARREACH_NONWORD);
static
vector<NFAEdge> getAsserts(const NGHolder &g) {
vector<NFAEdge> out;
for (const auto &e : edges_range(g)) {
if (g[e].assert_flags) {
out.emplace_back(e);
}
}
auto assertflags = [&g=g](const NFAEdge &e) {
return (g[e].assert_flags);
};
const auto &er = edges_range(g);
std::copy_if(begin(er), end(er), std::back_inserter(out), assertflags);
return out;
}
@ -384,7 +385,10 @@ void resolveEdges(ReportManager &rm, NGHolder &g, const ExpressionInfo &expr,
/* there may already be a different edge from start to eod if so
* we need to make it unconditional and alive
*/
if (NFAEdge start_eod = edge(u, g.acceptEod, g)) {
NFAEdge start_eod;
bool exists;
std::tie(start_eod, exists) = edge(u, g.acceptEod, g);
if (exists) {
g[start_eod].assert_flags = 0;
dead->erase(start_eod);
} else {
@ -437,7 +441,10 @@ void resolveEdges(ReportManager &rm, NGHolder &g, const ExpressionInfo &expr,
/* there may already be a different edge from start to eod if so
* we need to make it unconditional and alive
*/
if (NFAEdge start_eod = edge(u, g.acceptEod, g)) {
NFAEdge start_eod;
bool exists;
std::tie(start_eod, exists) = edge(u, g.acceptEod, g);
if (exists) {
g[start_eod].assert_flags = 0;
dead->erase(start_eod);
} else {
@ -496,7 +503,8 @@ void ensureCodePointStart(ReportManager &rm, NGHolder &g,
* boundaries. Assert resolution handles the badness coming from asserts.
* The only other source of trouble is startDs->accept connections.
*/
NFAEdge orig = edge(g.startDs, g.accept, g);
NFAEdge orig;
std::tie(orig, std::ignore) = edge(g.startDs, g.accept, g);
if (expr.utf8 && orig) {
DEBUG_PRINTF("rectifying %u\n", expr.report);
Report ir = rm.getBasicInternalReport(expr);

12
src/nfagraph/ng_edge_redundancy.cpp
View File

@ -514,17 +514,17 @@ bool removeSiblingsOfStartDotStar(NGHolder &g) {
* for SOM mode. (see UE-1544) */
bool optimiseVirtualStarts(NGHolder &g) {
vector<NFAEdge> dead;
auto deads = [&g=g](const NFAEdge &e) {
return (!is_any_start(source(e, g), g));
};
for (auto v : adjacent_vertices_range(g.startDs, g)) {
u32 flags = g[v].assert_flags;
if (!(flags & POS_FLAG_VIRTUAL_START)) {
continue;
}
for (const auto &e : in_edges_range(v, g)) {
if (!is_any_start(source(e, g), g)) {
dead.emplace_back(e);
}
}
const auto &e = in_edges_range(v, g);
std::copy_if(begin(e), end(e), std::back_inserter(dead), deads);
}
if (dead.empty()) {

21
src/nfagraph/ng_equivalence.cpp
View File

@ -98,9 +98,9 @@ class ClassInfo {
public:
struct ClassDepth {
ClassDepth() {}
ClassDepth(const NFAVertexDepth &d)
explicit ClassDepth(const NFAVertexDepth &d)
: d1(d.fromStart), d2(d.fromStartDotStar) {}
ClassDepth(const NFAVertexRevDepth &rd)
explicit ClassDepth(const NFAVertexRevDepth &rd)
: d1(rd.toAccept), d2(rd.toAcceptEod) {}
DepthMinMax d1;
DepthMinMax d2;
@ -159,7 +159,7 @@ public:
return id;
}
void append(WorkQueue &other) {
void append(const WorkQueue &other) {
for (const auto &e : other) {
push(e);
}
@ -193,7 +193,7 @@ private:
}
static
bool outIsIrreducible(NFAVertex &v, const NGHolder &g) {
bool outIsIrreducible(const NFAVertex &v, const NGHolder &g) {
unsigned nonSpecialVertices = 0;
for (auto w : adjacent_vertices_range(v, g)) {
if (!is_special(w, g) && w != v) {
@ -205,7 +205,7 @@ bool outIsIrreducible(NFAVertex &v, const NGHolder &g) {
}
static
bool inIsIrreducible(NFAVertex &v, const NGHolder &g) {
bool inIsIrreducible(const NFAVertex &v, const NGHolder &g) {
unsigned nonSpecialVertices = 0;
for (auto u : inv_adjacent_vertices_range(v, g)) {
if (!is_special(u, g) && u != v) {
@ -339,9 +339,9 @@ vector<VertexInfoSet> partitionGraph(vector<unique_ptr<VertexInfo>> &infos,
ClassInfo::ClassDepth depth;
if (eq == LEFT_EQUIVALENCE) {
depth = depths[vi->vert_index];
depth = ClassInfo::ClassDepth(depths[vi->vert_index]);
} else {
depth = rdepths[vi->vert_index];
depth = ClassInfo::ClassDepth(rdepths[vi->vert_index]);
}
ClassInfo ci(g, *vi, depth, eq);
@ -549,8 +549,8 @@ void mergeClass(vector<unique_ptr<VertexInfo>> &infos, NGHolder &g,
pred_info->succ.erase(old_vertex_info);
// if edge doesn't exist, create it
NFAEdge e = add_edge_if_not_present(pred_info->v, new_v, g);
NFAEdge e;
std::tie(e, std::ignore) = add_edge_if_not_present(pred_info->v, new_v, g);
// put edge tops, if applicable
if (!edgetops.empty()) {
assert(g[e].tops.empty() || g[e].tops == edgetops);
@ -560,7 +560,8 @@ void mergeClass(vector<unique_ptr<VertexInfo>> &infos, NGHolder &g,
pred_info->succ.insert(new_vertex_info);
if (new_v_eod) {
NFAEdge ee = add_edge_if_not_present(pred_info->v, new_v_eod,
NFAEdge ee;
std::tie(ee, std::ignore) = add_edge_if_not_present(pred_info->v, new_v_eod,
g);
// put edge tops, if applicable

51
src/nfagraph/ng_extparam.cpp
View File

@ -432,7 +432,7 @@ NFAVertex findSingleCyclic(const NGHolder &g) {
}
static
bool hasOffsetAdjust(const ReportManager &rm, NGHolder &g,
bool hasOffsetAdjust(const ReportManager &rm, const NGHolder &g,
int *adjust) {
const auto &reports = all_reports(g);
if (reports.empty()) {
@ -509,14 +509,14 @@ bool transformMinLengthToRepeat(NGHolder &g, ReportManager &rm) {
while (v != cyclic) {
DEBUG_PRINTF("vertex %zu\n", g[v].index);
width++;
auto succ = succs(v, g);
if (contains(succ, cyclic)) {
if (succ.size() == 1) {
auto s = succs(v, g);
if (contains(s, cyclic)) {
if (s.size() == 1) {
v = cyclic;
} else if (succ.size() == 2) {
} else if (s.size() == 2) {
// Cyclic and jump edge.
succ.erase(cyclic);
NFAVertex v2 = *succ.begin();
s.erase(cyclic);
NFAVertex v2 = *s.begin();
if (!edge(cyclic, v2, g).second) {
DEBUG_PRINTF("bad form\n");
return false;
@ -527,11 +527,11 @@ bool transformMinLengthToRepeat(NGHolder &g, ReportManager &rm) {
return false;
}
} else {
if (succ.size() != 1) {
if (s.size() != 1) {
DEBUG_PRINTF("bad form\n");
return false;
}
v = *succ.begin();
v = *s.begin();
}
}
@ -547,12 +547,12 @@ bool transformMinLengthToRepeat(NGHolder &g, ReportManager &rm) {
while (!is_any_accept(v, g)) {
DEBUG_PRINTF("vertex %zu\n", g[v].index);
width++;
auto succ = succs(v, g);
if (succ.size() != 1) {
auto s = succs(v, g);
if (s.size() != 1) {
DEBUG_PRINTF("bad form\n");
return false;
}
v = *succ.begin();
v = *s.begin();
}
int offsetAdjust = 0;
@ -572,27 +572,28 @@ bool transformMinLengthToRepeat(NGHolder &g, ReportManager &rm) {
return true;
}
vector<NFAVertex> preds;
vector<NFAVertex> predcs;
vector<NFAEdge> dead;
auto deads = [&g=g](const NFAEdge &e) {
return (target(e, g) != g.startDs);
};
for (auto u : inv_adjacent_vertices_range(cyclic, g)) {
DEBUG_PRINTF("pred %zu\n", g[u].index);
if (u == cyclic) {
continue;
}
preds.emplace_back(u);
predcs.emplace_back(u);
// We want to delete the out-edges of each predecessor, but need to
// make sure we don't delete the startDs self loop.
for (const auto &e : out_edges_range(u, g)) {
if (target(e, g) != g.startDs) {
dead.emplace_back(e);
}
}
const auto &e = out_edges_range(u, g);
std::copy_if(begin(e), end(e), std::back_inserter(dead), deads);
}
remove_edges(dead, g);
assert(!preds.empty());
assert(!predcs.empty());
const CharReach &cr = g[cyclic].char_reach;
@ -600,14 +601,14 @@ bool transformMinLengthToRepeat(NGHolder &g, ReportManager &rm) {
v = add_vertex(g);
g[v].char_reach = cr;
for (auto u : preds) {
for (auto u : predcs) {
add_edge(u, v, g);
}
preds.clear();
preds.emplace_back(v);
predcs.clear();
predcs.emplace_back(v);
}
assert(!preds.empty());
for (auto u : preds) {
assert(!predcs.empty());
for (auto u : predcs) {
add_edge(u, cyclic, g);
}

8
src/nfagraph/ng_fixed_width.cpp
View File

@ -66,15 +66,15 @@ bool findMask(const NGHolder &g, vector<CharReach> *mask, bool *anchored,
return false;
}
set<NFAVertex> &succs = *anchored ? s_succ : sds_succ;
succs.erase(g.startDs);
if (succs.size() != 1) {
set<NFAVertex> &succrs = *anchored ? s_succ : sds_succ;
succrs.erase(g.startDs);
if (succrs.size() != 1) {
DEBUG_PRINTF("branchy root\n");
return false;
}
NFAVertex u = *anchored ? g.start : g.startDs;
NFAVertex v = *succs.begin();
NFAVertex v = *succrs.begin();
while (true) {
DEBUG_PRINTF("validating vertex %zu\n", g[v].index);

25
src/nfagraph/ng_fuzzy.cpp
View File

@ -71,13 +71,13 @@ vector<flat_set<NFAVertex>> gatherSuccessorsByDepth(const NGHolder &g,
continue;
}
for (auto succ : adjacent_vertices_range(v, g)) {
for (auto succr : adjacent_vertices_range(v, g)) {
// ignore self-loops
if (v == succ) {
if (v == succr) {
continue;
}
DEBUG_PRINTF("Node %zu depth %u\n", g[succ].index, d + 1);
next.insert(succ);
next.insert(succr);
}
}
result[d] = next;
@ -113,13 +113,13 @@ vector<flat_set<NFAVertex>> gatherPredecessorsByDepth(const NGHolder &g,
for (unsigned d = 1; d < depth; d++) {
// collect all successors for all current level vertices
for (auto v : cur) {
for (auto pred : inv_adjacent_vertices_range(v, g)) {
for (auto predc : inv_adjacent_vertices_range(v, g)) {
// ignore self-loops
if (v == pred) {
if (v == predc) {
continue;
}
DEBUG_PRINTF("Node %zu depth %u\n", g[pred].index, d + 1);
next.insert(pred);
next.insert(predc);
}
}
result[d] = next;
@ -582,11 +582,11 @@ private:
// set up all reports
bool clone = false;
for (auto &pair : reports_to_vertices) {
for (const auto &pair : reports_to_vertices) {
const auto &reports = pair.first;
const auto &vertices = pair.second;
const auto &svertices = pair.second;
for (auto src : vertices) {
for (auto src : svertices) {
// get all predecessors up to edit distance
auto src_vertices_by_depth =
gatherPredecessorsByDepth(g, src, edit_distance);
@ -594,7 +594,8 @@ private:
// find which accepts source vertex connects to
flat_set<NFAVertex> targets;
for (const auto &accept : accepts) {
NFAEdge e = edge(src, accept, g);
NFAEdge e;
std::tie(e, std::ignore) = edge(src, accept, g);
if (e) {
targets.insert(accept);
}
@ -602,8 +603,8 @@ private:
assert(targets.size());
for (unsigned d = 0; d < src_vertices_by_depth.size(); d++) {
const auto &preds = src_vertices_by_depth[d];
for (auto v : preds) {
const auto &predcs = src_vertices_by_depth[d];
for (auto v : predcs) {
// only clone a node if it already contains reports
if (clone && !g[v].reports.empty()) {
create_clone(v, reports, edit_distance - d,

4
src/nfagraph/ng_haig.cpp
View File

@ -514,12 +514,12 @@ static
bool doHaig(const NGHolder &g, som_type som,
const vector<vector<CharReach>> &triggers, bool unordered_som,
raw_som_dfa *rdfa) {
u32 state_limit = HAIG_FINAL_DFA_STATE_LIMIT; /* haig never backs down from
a fight */
using StateSet = typename Auto::StateSet;
vector<StateSet> nfa_state_map;
Auto n(g, som, triggers, unordered_som);
try {
u32 state_limit = HAIG_FINAL_DFA_STATE_LIMIT; /* haig never backs down from
a fight */
if (!determinise(n, rdfa->states, state_limit, &nfa_state_map)) {
DEBUG_PRINTF("state limit exceeded\n");
return false;

22
src/nfagraph/ng_lbr.cpp
View File

@ -154,7 +154,7 @@ bytecode_ptr<NFA> buildLbrDot(const CharReach &cr, const depth &repeatMin,
const depth &repeatMax, u32 minPeriod,
bool is_reset, ReportID report) {
if (!cr.all()) {
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
enum RepeatType rtype = chooseRepeatType(repeatMin, repeatMax, minPeriod,
@ -176,7 +176,7 @@ bytecode_ptr<NFA> buildLbrVerm(const CharReach &cr, const depth &repeatMin,
const CharReach escapes(~cr);
if (escapes.count() != 1) {
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
enum RepeatType rtype = chooseRepeatType(repeatMin, repeatMax, minPeriod,
@ -199,7 +199,7 @@ bytecode_ptr<NFA> buildLbrNVerm(const CharReach &cr, const depth &repeatMin,
const CharReach escapes(cr);
if (escapes.count() != 1) {
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
enum RepeatType rtype = chooseRepeatType(repeatMin, repeatMax, minPeriod,
@ -228,7 +228,7 @@ bytecode_ptr<NFA> buildLbrShuf(const CharReach &cr, const depth &repeatMin,
minPeriod, rtype);
if (shuftiBuildMasks(~cr, (u8 *)&ls->mask_lo, (u8 *)&ls->mask_hi) == -1) {
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
DEBUG_PRINTF("built shuf lbr\n");
@ -296,7 +296,7 @@ bytecode_ptr<NFA> constructLBR(const CharReach &cr, const depth &repeatMin,
if (!nfa) {
assert(0);
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
return nfa;
@ -307,11 +307,11 @@ bytecode_ptr<NFA> constructLBR(const CastleProto &proto,
const CompileContext &cc,
const ReportManager &rm) {
if (!cc.grey.allowLbr) {
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
if (proto.repeats.size() != 1) {
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
const PureRepeat &repeat = proto.repeats.begin()->second;
@ -319,7 +319,7 @@ bytecode_ptr<NFA> constructLBR(const CastleProto &proto,
if (repeat.reports.size() != 1) {
DEBUG_PRINTF("too many reports\n");
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
bool is_reset;
@ -346,16 +346,16 @@ bytecode_ptr<NFA> constructLBR(const NGHolder &g,
const CompileContext &cc,
const ReportManager &rm) {
if (!cc.grey.allowLbr) {
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
PureRepeat repeat;
if (!isPureRepeat(g, repeat)) {
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
if (repeat.reports.size() != 1) {
DEBUG_PRINTF("too many reports\n");
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
CastleProto proto(g.kind, repeat);

4
src/nfagraph/ng_lbr_sve.hpp
View File

@ -39,7 +39,7 @@ bytecode_ptr<NFA> buildLbrVerm16(const CharReach &cr, const depth &repeatMin,
const CharReach escapes(~cr);
if (escapes.count() > 16) {
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
enum RepeatType rtype = chooseRepeatType(repeatMin, repeatMax, minPeriod,
@ -62,7 +62,7 @@ bytecode_ptr<NFA> buildLbrNVerm16(const CharReach &cr, const depth &repeatMin,
const CharReach escapes(cr);
if (escapes.count() > 16) {
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
enum RepeatType rtype = chooseRepeatType(repeatMin, repeatMax, minPeriod,

22
src/nfagraph/ng_limex.cpp
View File

@ -342,7 +342,7 @@ void attemptToUseAsStart(const NGHolder &g, NFAVertex u,
map<NFAVertex, flat_set<u32>> &unhandled_succ_tops,
map<u32, set<NFAVertex>> &tops_out) {
flat_set<u32> top_inter = unhandled_succ_tops.at(u);
flat_set<NFAVertex> succs;
flat_set<NFAVertex> f_succs;
for (NFAVertex v : adjacent_vertices_range(u, g)) {
if (!contains(unhandled_succ_tops, v)) {
return;
@ -360,7 +360,7 @@ void attemptToUseAsStart(const NGHolder &g, NFAVertex u,
set_intersection(top_inter.begin(), top_inter.end(),
v_tops.begin(), v_tops.end(), ni_inserter);
top_inter = std::move(new_inter);
succs.insert(v);
f_succs.insert(v);
}
if (top_inter.empty()) {
@ -373,7 +373,7 @@ void attemptToUseAsStart(const NGHolder &g, NFAVertex u,
}
DEBUG_PRINTF("reusing %zu is a start vertex\n", g[u].index);
markTopSuccAsHandled(u, top_inter, succs, tops_out, unhandled_top_succs,
markTopSuccAsHandled(u, top_inter, f_succs, tops_out, unhandled_top_succs,
unhandled_succ_tops);
}
@ -389,11 +389,11 @@ void reusePredsAsStarts(const NGHolder &g, const map<u32, CharReach> &top_reach,
/* create list of candidates first, to avoid issues of iter invalidation */
DEBUG_PRINTF("attempting to reuse vertices for top starts\n");
vector<NFAVertex> cand_starts;
for (NFAVertex u : unhandled_succ_tops | map_keys) {
if (hasSelfLoop(u, g)) {
cand_starts.emplace_back(u);
}
}
auto cands = [&g=g](const NFAVertex &u) {
return (hasSelfLoop(u, g));
};
const auto &u = unhandled_succ_tops | map_keys;
std::copy_if(begin(u), end(u), std::back_inserter(cand_starts), cands);
for (NFAVertex u : cand_starts) {
if (!contains(unhandled_succ_tops, u)) {
@ -652,7 +652,7 @@ constructNFA(const NGHolder &h_in, const ReportManager *rm,
u32 numStates = countStates(state_ids);
if (numStates > NFA_MAX_STATES) {
DEBUG_PRINTF("Can't build an NFA with %u states\n", numStates);
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
map<NFAVertex, BoundedRepeatSummary> br_cyclic;
@ -722,14 +722,14 @@ bytecode_ptr<NFA> constructReversedNFA_i(const NGHolder &h_in, u32 hint,
assert(h.kind == NFA_REV_PREFIX); /* triggered, raises internal callbacks */
// Do state numbering.
auto state_ids = numberStates(h, {});
auto state_ids = numberStates(h, flat_set<graph_detail::vertex_descriptor<ue2_graph<NGHolder, NFAGraphVertexProps, NFAGraphEdgeProps>>>());
// Quick exit: if we've got an embarrassment of riches, i.e. more states
// than we can implement in our largest NFA model, bail here.
u32 numStates = countStates(state_ids);
if (numStates > NFA_MAX_STATES) {
DEBUG_PRINTF("Can't build an NFA with %u states\n", numStates);
return nullptr;
return bytecode_ptr<NFA>(nullptr);
}
assert(sanityCheckGraph(h, state_ids));

18
src/nfagraph/ng_limex_accel.cpp
View File

@ -62,12 +62,12 @@ namespace ue2 {
static
void findAccelFriendGeneration(const NGHolder &g, const CharReach &cr,
const flat_set<NFAVertex> &cands,
const flat_set<NFAVertex> &preds,
const flat_set<NFAVertex> &f_preds,
flat_set<NFAVertex> *next_cands,
flat_set<NFAVertex> *next_preds,
flat_set<NFAVertex> *friends) {
for (auto v : cands) {
if (contains(preds, v)) {
if (contains(f_preds, v)) {
continue;
}
@ -80,7 +80,7 @@ void findAccelFriendGeneration(const NGHolder &g, const CharReach &cr,
}
for (auto u : inv_adjacent_vertices_range(v, g)) {
if (!contains(preds, u)) {
if (!contains(f_preds, u)) {
DEBUG_PRINTF("bad pred\n");
goto next_cand;
}
@ -116,8 +116,8 @@ void findAccelFriends(const NGHolder &g, NFAVertex v,
u32 friend_depth = offset + 1;
flat_set<NFAVertex> preds;
insert(&preds, inv_adjacent_vertices(v, g));
flat_set<NFAVertex> f_preds;
insert(&f_preds, inv_adjacent_vertices(v, g));
const CharReach &cr = g[v].char_reach;
flat_set<NFAVertex> cands;
@ -126,9 +126,9 @@ void findAccelFriends(const NGHolder &g, NFAVertex v,
flat_set<NFAVertex> next_preds;
flat_set<NFAVertex> next_cands;
for (u32 i = 0; i < friend_depth; i++) {
findAccelFriendGeneration(g, cr, cands, preds, &next_cands, &next_preds,
findAccelFriendGeneration(g, cr, cands, f_preds, &next_cands, &next_preds,
friends);
preds.insert(next_preds.begin(), next_preds.end());
f_preds.insert(next_preds.begin(), next_preds.end());
next_preds.clear();
cands.swap(next_cands);
next_cands.clear();
@ -321,7 +321,7 @@ struct DAccelScheme {
bool cd_a = buildDvermMask(a.double_byte);
bool cd_b = buildDvermMask(b.double_byte);
if (cd_a != cd_b) {
return cd_a > cd_b;
return cd_a;
}
}
@ -811,11 +811,9 @@ depth_done:
return true;
}
}
}
// Second option: a two-byte shufti (i.e. less than eight 2-byte
// literals)
if (depth > 1) {
for (unsigned int i = 0; i < (depth - 1); i++) {
if (depthReach[i].count() * depthReach[i+1].count()
<= DOUBLE_SHUFTI_LIMIT) {

6
src/nfagraph/ng_literal_analysis.cpp
View File

@ -490,9 +490,9 @@ vector<LitEdge> add_reverse_edges_and_index(LitGraph &lg) {
const size_t edge_count = num_edges(lg);
vector<LitEdge> fwd_edges;
fwd_edges.reserve(edge_count);
for (const auto &e : edges_range(lg)) {
fwd_edges.push_back(e);
}
const auto &e = edges_range(lg);
std::copy(begin(e), end(e), std::back_inserter(fwd_edges));
vector<LitEdge> rev_map(2 * edge_count);

2
src/nfagraph/ng_literal_analysis.h
View File

@ -70,7 +70,7 @@ bool bad_mixed_sensitivity(const ue2_literal &s);
* Score all the edges in the given graph, returning them in \p scores indexed
* by edge_index. */
std::vector<u64a> scoreEdges(const NGHolder &h,
const flat_set<NFAEdge> &known_bad = {});
const flat_set<NFAEdge> &known_bad = flat_set<NFAEdge>());
/** Returns a score for a literal set. Lower scores are better. */
u64a scoreSet(const std::set<ue2_literal> &s);

2
src/nfagraph/ng_literal_component.cpp
View File

@ -98,7 +98,7 @@ void addToString(string &s, const NGHolder &g, NFAVertex v) {
}
static
bool splitOffLiteral(NG &ng, NGHolder &g, NFAVertex v, const bool anchored,
bool splitOffLiteral(NG &ng, const NGHolder &g, NFAVertex v, const bool anchored,
set<NFAVertex> &dead) {
DEBUG_PRINTF("examine vertex %zu\n", g[v].index);
bool nocase = false, casefixed = false;

10
src/nfagraph/ng_mcclellan_internal.h
View File

@ -94,7 +94,7 @@ void transition_graph(autom &nfa, const std::vector<NFAVertex> &vByStateId,
/* generate top transitions, false -> top = selfloop */
bool top_allowed = is_triggered(graph);
StateSet succ = nfa.dead;
StateSet succr = nfa.dead;
for (size_t i = in.find_first(); i != in.npos; i = in.find_next(i)) {
NFAVertex u = vByStateId[i];
@ -102,7 +102,7 @@ void transition_graph(autom &nfa, const std::vector<NFAVertex> &vByStateId,
if (contains(unused, v)) {
continue;
}
succ.set(graph[v].index);
succr.set(graph[v].index);
}
if (top_allowed && !nfa.toppable.test(i)) {
@ -112,15 +112,15 @@ void transition_graph(autom &nfa, const std::vector<NFAVertex> &vByStateId,
}
}
StateSet active_squash = succ & squash;
StateSet active_squash = succr & squash;
if (active_squash.any()) {
for (size_t j = active_squash.find_first(); j != active_squash.npos;
j = active_squash.find_next(j)) {
succ &= squash_mask.find(j)->second;
succr &= squash_mask.find(j)->second;
}
}
for (size_t j = succ.find_first(); j != succ.npos; j = succ.find_next(j)) {
for (size_t j = succr.find_first(); j != succr.npos; j = succr.find_next(j)) {
const CharReach &cr = cr_by_index[j];
for (size_t s = cr.find_first(); s != cr.npos; s = cr.find_next(s)) {
next[s].set(j); /* already alpha'ed */

24
src/nfagraph/ng_misc_opt.cpp
View File

@ -404,19 +404,19 @@ CharReach reduced_cr(NFAVertex v, const NGHolder &g,
return v_cr;
}
NFAVertex pred = getSoleSourceVertex(g, v);
assert(pred);
NFAVertex s_pred = getSoleSourceVertex(g, v);
assert(s_pred);
/* require pred to be fed by one vertex OR (start + startDS) */
/* require s_pred to be fed by one vertex OR (start + startDS) */
NFAVertex predpred;
size_t idp = in_degree(pred, g);
if (hasSelfLoop(pred, g)) {
size_t idp = in_degree(s_pred, g);
if (hasSelfLoop(s_pred, g)) {
return v_cr; /* not cliche */
} else if (idp == 1) {
predpred = getSoleSourceVertex(g, pred);
predpred = getSoleSourceVertex(g, s_pred);
} else if (idp == 2
&& edge(g.start, pred, g).second
&& edge(g.startDs, pred, g).second) {
&& edge(g.start, s_pred, g).second
&& edge(g.startDs, s_pred, g).second) {
predpred = g.startDs;
} else {
return v_cr; /* not cliche */
@ -425,7 +425,7 @@ CharReach reduced_cr(NFAVertex v, const NGHolder &g,
assert(predpred);
/* require predpred to be cyclic and its cr to be a superset of
pred and v */
s_pred and v */
if (!hasSelfLoop(predpred, g)) {
return v_cr; /* not cliche */
}
@ -435,7 +435,7 @@ CharReach reduced_cr(NFAVertex v, const NGHolder &g,
return v_cr; /* fake cyclic */
}
const CharReach &p_cr = g[pred].char_reach;
const CharReach &p_cr = g[s_pred].char_reach;
const CharReach &pp_cr = g[predpred].char_reach;
if (!v_cr.isSubsetOf(pp_cr) || !p_cr.isSubsetOf(pp_cr)) {
return v_cr; /* not cliche */
@ -446,7 +446,7 @@ CharReach reduced_cr(NFAVertex v, const NGHolder &g,
set<NFAVertex> v_succ;
insert(&v_succ, adjacent_vertices(v, g));
set<NFAVertex> p_succ;
insert(&p_succ, adjacent_vertices(pred, g));
insert(&p_succ, adjacent_vertices(s_pred, g));
if (!is_subset_of(v_succ, p_succ)) {
DEBUG_PRINTF("fail\n");
@ -456,7 +456,7 @@ CharReach reduced_cr(NFAVertex v, const NGHolder &g,
if (contains(v_succ, g.accept) || contains(v_succ, g.acceptEod)) {
/* need to check that reports of v are a subset of p's */
if (!is_subset_of(g[v].reports,
g[pred].reports)) {
g[s_pred].reports)) {
DEBUG_PRINTF("fail - reports not subset\n");
return v_cr; /* not cliche */
}

3
src/nfagraph/ng_netflow.cpp
View File

@ -93,7 +93,8 @@ void addReverseEdges(NGHolder &g, vector<NFAEdge> &reverseEdge,
if (it == allEdges.end()) {
// No reverse edge, add one.
NFAVertex u = source(fwd, g), v = target(fwd, g);
NFAEdge rev = add_edge(v, u, g);
NFAEdge rev;
std::tie(rev, std::ignore) = add_edge(v, u, g);
it = allEdges.insert(make_pair(make_pair(vidx, uidx), rev)).first;
// Add to capacity map.
u32 revIndex = g[rev].index;

22
src/nfagraph/ng_prune.cpp
View File

@ -62,11 +62,13 @@ void pruneUnreachable(NGHolder &g) {
&& edge(g.accept, g.acceptEod, g).second) {
// Trivial case: there are no in-edges to our accepts (other than
// accept->acceptEod), so all non-specials are unreachable.
for (auto v : vertices_range(g)) {
if (!is_special(v, g)) {
dead.emplace_back(v);
}
}
auto deads = [&g=g](const NFAVertex &v) {
return (!is_special(v, g));
};
const auto &vr = vertices_range(g);
std::copy_if(begin(vr), end(vr), std::back_inserter(dead), deads);
} else {
// Walk a reverse graph from acceptEod with Boost's depth_first_visit
// call.
@ -199,17 +201,17 @@ void pruneHighlanderAccepts(NGHolder &g, const ReportManager &rm) {
}
vector<NFAEdge> dead;
auto deads = [&g=g](const NFAEdge &e) {
return (!is_any_accept(target(e, g), g));
};
for (auto u : inv_adjacent_vertices_range(g.accept, g)) {
if (is_special(u, g)) {
continue;
}
// We can prune any out-edges that aren't accepts
for (const auto &e : out_edges_range(u, g)) {
if (!is_any_accept(target(e, g), g)) {
dead.emplace_back(e);
}
}
const auto &er = out_edges_range(u, g);
std::copy_if(begin(er), end(er), std::back_inserter(dead), deads);
}
if (dead.empty()) {

6
src/nfagraph/ng_puff.cpp
View File

@ -244,7 +244,7 @@ u32 allowedSquashDistance(const CharReach &cr, u32 min_width, const NGHolder &g,
/** Gives a stronger puff trigger when the trigger is connected to a wide
* cyclic state (aside from sds) */
static
void improveHead(NGHolder &g, NFAVertex *a, vector<NFAVertex> *nodes) {
void improveHead(const NGHolder &g, NFAVertex *a, vector<NFAVertex> *nodes) {
DEBUG_PRINTF("attempting to improve puff trigger\n");
assert(!nodes->empty());
const CharReach &puff_cr = g[nodes->back()].char_reach;
@ -263,7 +263,7 @@ void improveHead(NGHolder &g, NFAVertex *a, vector<NFAVertex> *nodes) {
}
static
void constructPuff(NGHolder &g, const NFAVertex a, const NFAVertex puffv,
void constructPuff(const NGHolder &g, const NFAVertex a, const NFAVertex puffv,
const CharReach &cr, const ReportID report, u32 width,
bool fixed_depth, bool unbounded, bool auto_restart,
RoseBuild &rose, ReportManager &rm,
@ -361,9 +361,7 @@ bool doComponent(RoseBuild &rose, ReportManager &rm, NGHolder &g, NFAVertex a,
// single report ID on a vertex
if (is_match_vertex(a, g)) {
DEBUG_PRINTF("stop puffing due to vertex that leads to accept\n");
if (!nodes.empty()) {
nodes.pop_back();
}
break;
}
}

18
src/nfagraph/ng_redundancy.cpp
View File

@ -307,13 +307,15 @@ void markForRemoval(const NFAVertex v, VertexInfoMap &infoMap,
static
bool hasInEdgeTops(const NGHolder &g, NFAVertex v) {
NFAEdge e = edge(g.start, v, g);
NFAEdge e;
std::tie(e, std::ignore) = edge(g.start, v, g);
return e && !g[e].tops.empty();
}
/** Transform (1), removal of redundant vertices. */
static
bool doUselessMergePass(NGHolder &g, som_type som, VertexInfoMap &infoMap,
bool doUselessMergePass(const NGHolder &g, const som_type som, VertexInfoMap &infoMap,
set<NFAVertex> &removable) {
/* useless merges can be done in any order, no need to take any care with
* ordering */
@ -323,7 +325,7 @@ bool doUselessMergePass(NGHolder &g, som_type som, VertexInfoMap &infoMap,
bool changed = false;
for (auto v : vertices_range(g)) {
VertexInfo &info = infoMap[v];
const VertexInfo &info = infoMap[v];
if (info.isRemoved) {
continue;
@ -439,7 +441,7 @@ bool doUselessMergePass(NGHolder &g, som_type som, VertexInfoMap &infoMap,
continue; // Conservatively skip anything with nonzero tops.
}
CharReach &otherReach = g[t].char_reach;
const CharReach &otherReach = g[t].char_reach;
if (currReach.isSubsetOf(otherReach)) {
DEBUG_PRINTF("removing redundant vertex %zu (keeping %zu)\n",
g[v].index, g[t].index);
@ -636,12 +638,12 @@ bool reversePathReachSubset(const NFAEdge &e, const NFAVertex &dom,
NFAVertex start = source(e, g);
using RevGraph = boost::reverse_graph<NGHolder, const NGHolder &>;
map<RevGraph::vertex_descriptor, boost::default_color_type> vertexColor;
// Walk the graph backwards from v, examining each node. We fail (return
// false) if we encounter a node with reach NOT a subset of domReach, and
// we stop searching at dom.
try {
map<RevGraph::vertex_descriptor, boost::default_color_type> vertexColor;
depth_first_visit(RevGraph(g), start,
ReachSubsetVisitor(domReach),
make_assoc_property_map(vertexColor),
@ -664,12 +666,12 @@ bool forwardPathReachSubset(const NFAEdge &e, const NFAVertex &dom,
}
NFAVertex start = target(e, g);
map<NFAVertex, boost::default_color_type> vertexColor;
// Walk the graph forward from v, examining each node. We fail (return
// false) if we encounter a node with reach NOT a subset of domReach, and
// we stop searching at dom.
try {
map<NFAVertex, boost::default_color_type> vertexColor;
depth_first_visit(g, start, ReachSubsetVisitor(domReach),
make_assoc_property_map(vertexColor),
VertexIs<NGHolder, NFAVertex>(dom));
@ -748,7 +750,7 @@ u32 findCyclic(const NGHolder &g, vector<bool> &cyclic) {
}
static
void findCyclicDom(NGHolder &g, vector<bool> &cyclic,
void findCyclicDom(const NGHolder &g, vector<bool> &cyclic,
set<NFAEdge> &dead, som_type som) {
auto dominators = findDominators(g);
@ -792,7 +794,7 @@ void findCyclicDom(NGHolder &g, vector<bool> &cyclic,
}
static
void findCyclicPostDom(NGHolder &g, vector<bool> &cyclic,
void findCyclicPostDom(const NGHolder &g, vector<bool> &cyclic,
set<NFAEdge> &dead) {
auto postdominators = findPostDominators(g);

52
src/nfagraph/ng_repeat.cpp
View File

@ -393,9 +393,9 @@ void checkReachSubgraphs(const NGHolder &g, vector<ReachSubgraph> &rs,
unordered_set<NFAVertex> involved(rsi.vertices.begin(),
rsi.vertices.end());
unordered_set<NFAVertex> tail(involved); // to look for back-edges.
unordered_set<NFAVertex> pred, succ;
proper_pred(g, rsi.vertices.front(), pred);
proper_succ(g, rsi.vertices.back(), succ);
unordered_set<NFAVertex> v_pred, v_succ;
proper_pred(g, rsi.vertices.front(), v_pred);
proper_succ(g, rsi.vertices.back(), v_succ);
flat_set<ReportID> reports;
findFirstReports(g, rsi, reports);
@ -406,7 +406,7 @@ void checkReachSubgraphs(const NGHolder &g, vector<ReachSubgraph> &rs,
for (auto v : rsi.vertices) {
tail.erase(v); // now contains all vertices _after_ this one.
if (vertexIsBad(g, v, involved, tail, pred, succ, reports)) {
if (vertexIsBad(g, v, involved, tail, v_pred, v_succ, reports)) {
recalc = true;
continue;
}
@ -793,10 +793,10 @@ void replaceSubgraphWithSpecial(NGHolder &g, ReachSubgraph &rsi,
const unordered_set<NFAVertex> involved(rsi.vertices.begin(),
rsi.vertices.end());
vector<NFAVertex> succs;
getSuccessors(g, rsi, &succs);
vector<NFAVertex> g_succs;
getSuccessors(g, rsi, &g_succs);
unpeelNearEnd(g, rsi, depths, &succs);
unpeelNearEnd(g, rsi, depths, &g_succs);
// Create our replacement cyclic state with the same reachability and
// report info as the last vertex in our topo-ordered list.
@ -824,7 +824,7 @@ void replaceSubgraphWithSpecial(NGHolder &g, ReachSubgraph &rsi,
// Wire cyclic state to tug trigger states built from successors.
vector<NFAVertex> tugs;
for (auto v : succs) {
for (auto v : g_succs) {
buildTugTrigger(g, cyclic, v, involved, depths, tugs);
}
created.insert(tugs.begin(), tugs.end());
@ -860,10 +860,8 @@ void replaceSubgraphWithLazySpecial(NGHolder &g, ReachSubgraph &rsi,
DEBUG_PRINTF("entry\n");
const unordered_set<NFAVertex> involved(rsi.vertices.begin(),
rsi.vertices.end());
vector<NFAVertex> succs;
getSuccessors(g, rsi, &succs);
vector<NFAVertex> g_succs;
getSuccessors(g, rsi, &g_succs);
// Create our replacement cyclic state with the same reachability and
// report info as the last vertex in our topo-ordered list.
@ -892,15 +890,15 @@ void replaceSubgraphWithLazySpecial(NGHolder &g, ReachSubgraph &rsi,
// In the rose case, our tug is our cyclic, and it's wired to our
// successors (which should be just the accept).
vector<NFAVertex> tugs;
assert(succs.size() == 1);
for (auto v : succs) {
assert(g_succs.size() == 1);
for (auto v : g_succs) {
add_edge(cyclic, v, g);
}
// Wire pos trigger to accept if min repeat is one -- this deals with cases
// where we can get a pos and tug trigger on the same byte.
if (rsi.repeatMin == depth(1)) {
for (auto v : succs) {
for (auto v : g_succs) {
add_edge(pos_trigger, v, g);
g[pos_trigger].reports = g[cyclic].reports;
}
@ -1144,7 +1142,7 @@ NFAVertex buildTriggerStates(NGHolder &g, const vector<CharReach> &trigger,
g[v].char_reach = cr;
add_edge(u, v, g);
if (u == g.start) {
g[edge(u, v, g)].tops.insert(top);
g[edge(u, v, g).first].tops.insert(top);
}
u = v;
}
@ -1467,9 +1465,9 @@ struct StrawWalker {
}
if (ai != ae) {
DEBUG_PRINTF("more than one succ\n");
set<NFAVertex> succs;
insert(&succs, adjacent_vertices(v, g));
succs.erase(v);
set<NFAVertex> a_succs;
insert(&a_succs, adjacent_vertices(v, g));
a_succs.erase(v);
for (tie(ai, ae) = adjacent_vertices(v, g); ai != ae; ++ai) {
next = *ai;
DEBUG_PRINTF("checking %zu\n", g[next].index);
@ -1479,7 +1477,7 @@ struct StrawWalker {
set<NFAVertex> lsuccs;
insert(&lsuccs, adjacent_vertices(next, g));
if (lsuccs != succs) {
if (lsuccs != a_succs) {
continue;
}
@ -1887,7 +1885,7 @@ void buildFeeder(NGHolder &g, const BoundedRepeatData &rd,
* offset.
*/
static
bool improveLeadingRepeat(NGHolder &g, BoundedRepeatData &rd,
bool improveLeadingRepeat(NGHolder &g, const BoundedRepeatData &rd,
unordered_set<NFAVertex> &created,
const vector<BoundedRepeatData> &all_repeats) {
assert(edge(g.startDs, g.startDs, g).second);
@ -1908,9 +1906,9 @@ bool improveLeadingRepeat(NGHolder &g, BoundedRepeatData &rd,
}
vector<NFAVertex> straw;
NFAVertex pred =
NFAVertex w_pred =
walkStrawToCyclicRev(g, rd.pos_trigger, all_repeats, straw);
if (pred != g.startDs) {
if (w_pred != g.startDs) {
DEBUG_PRINTF("straw walk doesn't lead to startDs\n");
return false;
}
@ -1958,7 +1956,7 @@ bool improveLeadingRepeat(NGHolder &g, BoundedRepeatData &rd,
}
static
vector<NFAVertex> makeOwnStraw(NGHolder &g, BoundedRepeatData &rd,
vector<NFAVertex> makeOwnStraw(NGHolder &g, const BoundedRepeatData &rd,
const vector<NFAVertex> &straw) {
// Straw runs from startDs to our pos trigger.
assert(!straw.empty());
@ -1992,7 +1990,7 @@ vector<NFAVertex> makeOwnStraw(NGHolder &g, BoundedRepeatData &rd,
* rewire the straw to start instead of removing the startDs self-loop.
*/
static
bool improveLeadingRepeatOutfix(NGHolder &g, BoundedRepeatData &rd,
bool improveLeadingRepeatOutfix(NGHolder &g, const BoundedRepeatData &rd,
unordered_set<NFAVertex> &created,
const vector<BoundedRepeatData> &all_repeats) {
assert(g.kind == NFA_OUTFIX);
@ -2013,9 +2011,9 @@ bool improveLeadingRepeatOutfix(NGHolder &g, BoundedRepeatData &rd,
}
vector<NFAVertex> straw;
NFAVertex pred =
NFAVertex w_pred =
walkStrawToCyclicRev(g, rd.pos_trigger, all_repeats, straw);
if (pred != g.startDs) {
if (w_pred != g.startDs) {
DEBUG_PRINTF("straw walk doesn't lead to startDs\n");
return false;
}

4
src/nfagraph/ng_restructuring.cpp
View File

@ -54,8 +54,8 @@ void wireStartToTops(NGHolder &g, const flat_set<NFAVertex> &tops,
vector<NFAEdge> &tempEdges) {
for (NFAVertex v : tops) {
assert(!isLeafNode(v, g));
const NFAEdge &e = add_edge(g.start, v, g);
auto edge_result = add_edge(g.start, v, g);
const NFAEdge &e = edge_result.first;
tempEdges.emplace_back(e);
}
}

36
src/nfagraph/ng_som.cpp
View File

@ -877,18 +877,18 @@ bool beginsWithDotStar(const NGHolder &g) {
// We can ignore the successors of start, as matches that begin there will
// necessarily have a SOM of 0.
set<NFAVertex> succ;
insert(&succ, adjacent_vertices(g.startDs, g));
succ.erase(g.startDs);
set<NFAVertex> a_succ;
insert(&a_succ, adjacent_vertices(g.startDs, g));
a_succ.erase(g.startDs);
for (auto v : succ) {
for (auto v : a_succ) {
// We want 'dot' states that aren't virtual starts.
if (g[v].char_reach.all() &&
!g[v].assert_flags) {
hasDot = true;
set<NFAVertex> dotsucc;
insert(&dotsucc, adjacent_vertices(v, g));
if (dotsucc != succ) {
if (dotsucc != a_succ) {
DEBUG_PRINTF("failed dot-star succ check\n");
return false;
}
@ -1178,7 +1178,7 @@ void expandGraph(NGHolder &g, unordered_map<NFAVertex, u32> &regions,
}
static
bool doTreePlanningIntl(NGHolder &g,
bool doTreePlanningIntl(const NGHolder &g,
const unordered_map<NFAVertex, u32> &regions,
const map<u32, region_info> &info,
map<u32, region_info>::const_iterator picked, u32 bad_region,
@ -1293,7 +1293,7 @@ bool doTreePlanningIntl(NGHolder &g,
DEBUG_PRINTF("add mapped reporters for region %u\n", it->first);
addMappedReporterVertices(it->second, g, copy_to_orig,
plan.back().reporters);
} while (it->second.optional && it != info.rend() &&
} while (it != info.rend() && it->second.optional &&
(++it)->first > furthest->first);
return true;
@ -1409,7 +1409,7 @@ bool doSomPlanning(NGHolder &g, bool stuck_in,
/* Need to verify how far the lock covers */
u32 bad_region;
NGHolder *ap_pref = plan.back().prefix.get();
const NGHolder *ap_pref = plan.back().prefix.get();
NGHolder ap_temp;
if (hasBigCycles(*ap_pref)) {
fillRoughMidfix(&ap_temp, g, regions, info, picked);
@ -1552,7 +1552,7 @@ bool doSomPlanning(NGHolder &g, bool stuck_in,
DEBUG_PRINTF("region %u contributes reporters to last plan\n",
it->first);
addReporterVertices(it->second, g, plan.back().reporters);
} while (it->second.optional && it != info.rend() &&
} while (it != info.rend() && it->second.optional &&
(++it)->first > furthest->first);
DEBUG_PRINTF("done!\n");
@ -1856,7 +1856,7 @@ bool doSomRevNfa(NG &ng, NGHolder &g, const CompileContext &cc) {
}
static
u32 doSomRevNfaPrefix(NG &ng, const ExpressionInfo &expr, NGHolder &g,
u32 doSomRevNfaPrefix(NG &ng, const ExpressionInfo &expr, const NGHolder &g,
const CompileContext &cc) {
depth maxWidth = findMaxWidth(g);
@ -2012,7 +2012,7 @@ void setReportOnHaigPrefix(RoseBuild &rose, NGHolder &h) {
}
static
bool tryHaig(RoseBuild &rose, NGHolder &g,
bool tryHaig(RoseBuild &rose, const NGHolder &g,
const unordered_map<NFAVertex, u32> &regions,
som_type som, u32 somPrecision,
map<u32, region_info>::const_iterator picked,
@ -2444,13 +2444,9 @@ void makeReportsSomPass(ReportManager &rm, NGHolder &g) {
}
static
bool doLitHaigSom(NG &ng, NGHolder &g, som_type som) {
bool doLitHaigSom(NG &ng, const NGHolder &g, som_type som) {
ue2_literal lit;
shared_ptr<NGHolder> rhs = make_shared<NGHolder>();
if (!rhs) {
assert(0);
throw std::bad_alloc();
}
if (!ng.cc.grey.allowLitHaig) {
return false;
}
@ -2515,10 +2511,6 @@ bool doHaigLitHaigSom(NG &ng, NGHolder &g,
ue2_literal lit;
shared_ptr<NGHolder> rhs = make_shared<NGHolder>();
shared_ptr<NGHolder> lhs = make_shared<NGHolder>();
if (!rhs || !lhs) {
assert(0);
throw std::bad_alloc();
}
if (!splitOffBestLiteral(g, regions, &lit, &*lhs, &*rhs, ng.cc)) {
return false;
@ -2661,7 +2653,7 @@ bool doHaigLitHaigSom(NG &ng, NGHolder &g,
}
static
bool doMultiLitHaigSom(NG &ng, NGHolder &g, som_type som) {
bool doMultiLitHaigSom(NG &ng, const NGHolder &g, som_type som) {
set<ue2_literal> lits;
shared_ptr<NGHolder> rhs = make_shared<NGHolder>();
if (!ng.cc.grey.allowLitHaig) {
@ -3135,7 +3127,7 @@ sombe_rv doSomWithHaig(NG &ng, NGHolder &g, const ExpressionInfo &expr,
// try a redundancy pass.
if (addSomRedundancy(g, depths)) {
depths = getDistancesFromSOM(g);
depths = getDistancesFromSOM(g); // cppcheck-suppress unreadVariable
}
auto regions = assignRegions(g);

8
src/nfagraph/ng_som_add_redundancy.cpp
View File

@ -113,9 +113,9 @@ bool forkVertex(NFAVertex v, NGHolder &g, vector<DepthMinMax> &depths,
}
*numNewVertices += predGroups.size();
for (auto &group : predGroups) {
for (const auto &group : predGroups) {
const depth &predDepth = group.first;
const vector<NFAEdge> &preds = group.second;
const vector<NFAEdge> &gspreds = group.second;
// Clone v for this depth with all its associated out-edges.
u32 clone_idx = depths.size(); // next index to be used
@ -131,8 +131,8 @@ bool forkVertex(NFAVertex v, NGHolder &g, vector<DepthMinMax> &depths,
add_edge(clone, target(e, g), g[e], g);
}
// Add in-edges from preds in this group.
for (const auto &e : preds) {
// Add in-edges from gspreds in this group.
for (const auto &e : gspreds) {
add_edge(source(e, g), clone, g[e], g);
}
}

23
src/nfagraph/ng_som_util.cpp
View File

@ -58,11 +58,12 @@ vector<DepthMinMax> getDistancesFromSOM(const NGHolder &g_orig) {
cloneHolder(g, g_orig, &vmap);
vector<NFAVertex> vstarts;
for (auto v : vertices_range(g)) {
if (is_virtual_start(v, g)) {
vstarts.emplace_back(v);
}
}
auto vstart = [&g=g](const NFAVertex &v) {
return (is_virtual_start(v, g));
};
const auto &vr = vertices_range(g);
std::copy_if(begin(vr), end(vr), std::back_inserter(vstarts), vstart);
vstarts.emplace_back(g.startDs);
// wire the successors of every virtual start or startDs to g.start.
@ -269,18 +270,6 @@ bool somMayGoBackwards(NFAVertex u, const NGHolder &g,
boost::depth_first_search(c_g, visitor(backEdgeVisitor)
.root_vertex(c_g.start));
for (const auto &e : be) {
NFAVertex s = source(e, c_g);
NFAVertex t = target(e, c_g);
DEBUG_PRINTF("back edge %zu %zu\n", c_g[s].index, c_g[t].index);
if (s != t) {
assert(0);
DEBUG_PRINTF("eek big cycle\n");
rv = true; /* big cycle -> eek */
goto exit;
}
}
DEBUG_PRINTF("checking acyclic+selfloop graph\n");
rv = !firstMatchIsFirst(c_g);

3
src/nfagraph/ng_split.cpp
View File

@ -151,7 +151,8 @@ void splitRHS(const NGHolder &base, const vector<NFAVertex> &pivots,
for (auto pivot : pivots) {
assert(contains(*rhs_map, pivot));
NFAEdge e = add_edge(rhs->start, (*rhs_map)[pivot], *rhs);
auto edge_result = add_edge(rhs->start, (*rhs_map)[pivot], *rhs);
NFAEdge e = edge_result.first;
(*rhs)[e].tops.insert(DEFAULT_TOP);
}

28
src/nfagraph/ng_squash.cpp
View File

@ -255,19 +255,19 @@ void buildSquashMask(NFAStateSet &mask, const NGHolder &g, NFAVertex v,
}
static
void buildSucc(NFAStateSet &succ, const NGHolder &g, NFAVertex v) {
void buildSucc(NFAStateSet &ssucc, const NGHolder &g, NFAVertex v) {
for (auto w : adjacent_vertices_range(v, g)) {
if (!is_special(w, g)) {
succ.set(g[w].index);
ssucc.set(g[w].index);
}
}
}
static
void buildPred(NFAStateSet &pred, const NGHolder &g, NFAVertex v) {
void buildPred(NFAStateSet &spred, const NGHolder &g, NFAVertex v) {
for (auto u : inv_adjacent_vertices_range(v, g)) {
if (!is_special(u, g)) {
pred.set(g[u].index);
spred.set(g[u].index);
}
}
}
@ -409,19 +409,19 @@ unordered_map<NFAVertex, NFAStateSet> findSquashers(const NGHolder &g,
DEBUG_PRINTF("state %u is cyclic\n", i);
NFAStateSet mask(numStates), succ(numStates), pred(numStates);
NFAStateSet mask(numStates), ssucc(numStates), spred(numStates);
buildSquashMask(mask, g, v, cr, initStates, vByIndex, pdom_tree, som,
som_depths, region_map, cache);
buildSucc(succ, g, v);
buildPred(pred, g, v);
buildSucc(ssucc, g, v);
buildPred(spred, g, v);
const auto &reports = g[v].reports;
for (size_t j = succ.find_first(); j != succ.npos;
j = succ.find_next(j)) {
for (size_t j = ssucc.find_first(); j != ssucc.npos;
j = ssucc.find_next(j)) {
NFAVertex vj = vByIndex[j];
NFAStateSet pred2(numStates);
buildPred(pred2, g, vj);
if (pred2 == pred) {
if (pred2 == spred) {
DEBUG_PRINTF("adding the sm from %zu to %u's sm\n", j, i);
NFAStateSet tmp(numStates);
buildSquashMask(tmp, g, vj, cr, initStates, vByIndex, pdom_tree,
@ -430,14 +430,14 @@ unordered_map<NFAVertex, NFAStateSet> findSquashers(const NGHolder &g,
}
}
for (size_t j = pred.find_first(); j != pred.npos;
j = pred.find_next(j)) {
for (size_t j = spred.find_first(); j != spred.npos;
j = spred.find_next(j)) {
NFAVertex vj = vByIndex[j];
NFAStateSet succ2(numStates);
buildSucc(succ2, g, vj);
/* we can use j as a basis for squashing if its succs are a subset
* of ours */
if ((succ2 & ~succ).any()) {
if ((succ2 & ~ssucc).any()) {
continue;
}
@ -590,7 +590,7 @@ void getHighlanderReporters(const NGHolder &g, const NFAVertex accept,
verts.insert(v);
next_vertex:
continue;
;
}
}

Some files were not shown because too many files have changed in this diff Show More