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 pull request #131 from VectorCamp/develop

Browse Source 
Prepare for new release 5.4.9
This commit is contained in:
Konstantinos Margaritis 2022-09-19 17:59:40 +03:00 committed by GitHub
commit 6d8599eece
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
7 changed files with 72 additions and 41 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

58
CMakeLists.txt
View File

@ -1,4 +1,5 @@
cmake_minimum_required (VERSION 2.8.11) cmake_minimum_required (VERSION 2.8.12)
project (vectorscan C CXX) project (vectorscan C CXX)
set (HS_MAJOR_VERSION 5) set (HS_MAJOR_VERSION 5)
@ -165,27 +166,49 @@ if (CMAKE_COMPILER_IS_GNUCC AND NOT CROSS_COMPILE)
# cpuid info and then chooses the best microarch it can (and replaces # cpuid info and then chooses the best microarch it can (and replaces
# the flag), so use that for tune. # the flag), so use that for tune.
set(TUNE_FLAG "mtune")
set(GNUCC_TUNE "")
message(STATUS "ARCH_FLAG '${ARCH_FLAG}' '${GNUCC_ARCH}', TUNE_FLAG '${TUNE_FLAG}' '${GNUCC_TUNE}' ")
# arg1 might exist if using ccache # arg1 might exist if using ccache
string (STRIP "${CMAKE_C_COMPILER_ARG1}" CC_ARG1) string (STRIP "${CMAKE_C_COMPILER_ARG1}" CC_ARG1)
set (EXEC_ARGS ${CC_ARG1} -c -Q --help=target -${ARCH_FLAG}=native -mtune=native) set (EXEC_ARGS ${CC_ARG1} -c -Q --help=target -${ARCH_FLAG}=native -${TUNE_FLAG}=native)
execute_process(COMMAND ${CMAKE_C_COMPILER} ${EXEC_ARGS} execute_process(COMMAND ${CMAKE_C_COMPILER} ${EXEC_ARGS}
OUTPUT_VARIABLE _GCC_OUTPUT) OUTPUT_VARIABLE _GCC_OUTPUT)
string(FIND "${_GCC_OUTPUT}" "${ARCH_FLAG}" POS) set(_GCC_OUTPUT_TUNE ${_GCC_OUTPUT})
string(FIND "${_GCC_OUTPUT}" "${ARCH_FLAG}=" POS)
string(SUBSTRING "${_GCC_OUTPUT}" ${POS} -1 _GCC_OUTPUT) string(SUBSTRING "${_GCC_OUTPUT}" ${POS} -1 _GCC_OUTPUT)
string(REGEX REPLACE "${ARCH_FLAG}=[ \t]*([^ \n]*)[ \n].*" "\\1" GNUCC_ARCH "${_GCC_OUTPUT}") string(REGEX REPLACE "${ARCH_FLAG}=[ \t]*([^ \n]*)[ \n].*" "\\1" GNUCC_ARCH "${_GCC_OUTPUT}")
string(FIND "${_GCC_OUTPUT_TUNE}" "${TUNE_FLAG}=" POS_TUNE)
string(SUBSTRING "${_GCC_OUTPUT_TUNE}" ${POS_TUNE} -1 _GCC_OUTPUT_TUNE)
string(REGEX REPLACE "${TUNE_FLAG}=[ \t]*([^ \n]*)[ \n].*" "\\1" GNUCC_TUNE "${_GCC_OUTPUT_TUNE}")
string(FIND "${GNUCC_ARCH}" "sve" POS_SVE)
string(FIND "${GNUCC_ARCH}" "sve2" POS_SVE2)
string(FIND "${GNUCC_ARCH}" "sve2-bitperm" POS_SVE2_BITPERM)
if (NOT POS_SVE EQUAL 0)
set(SVE_FOUND 1)
elseif(NOT POS_SVE2 EQUAL 0)
set(SVE2_FOUND 1)
elseif(NOT POS_SVE2_BITPERM EQUAL 0)
set(SVE2_BITPERM_FOUND 1)
endif()
message(STATUS "ARCH_FLAG '${ARCH_FLAG}' '${GNUCC_ARCH}', TUNE_FLAG '${TUNE_FLAG}' '${GNUCC_TUNE}' ")
# test the parsed flag # test the parsed flag
set (EXEC_ARGS ${CC_ARG1} -E - -mtune=${GNUCC_ARCH}) set (EXEC_ARGS ${CC_ARG1} -E - -${ARCH_FLAG}=${GNUCC_ARCH} -${TUNE_FLAG}=${GNUCC_TUNE})
execute_process(COMMAND ${CMAKE_C_COMPILER} ${EXEC_ARGS} execute_process(COMMAND ${CMAKE_C_COMPILER} ${EXEC_ARGS}
OUTPUT_QUIET ERROR_QUIET OUTPUT_QUIET ERROR_QUIET
INPUT_FILE /dev/null INPUT_FILE /dev/null
RESULT_VARIABLE GNUCC_TUNE_TEST) RESULT_VARIABLE GNUCC_TUNE_TEST)
if (NOT GNUCC_TUNE_TEST EQUAL 0) if (NOT GNUCC_TUNE_TEST EQUAL 0)
message(WARNING "Something went wrong determining gcc tune: -mtune=${GNUCC_ARCH} not valid, falling back to -mtune=native") message(WARNING "Something went wrong determining gcc tune: -mtune=${GNUCC_TUNE} not valid, falling back to -mtune=native")
set(TUNE_FLAG native) set(GNUCC_TUNE native)
else() else()
set(TUNE_FLAG ${GNUCC_ARCH}) set(GNUCC_TUNE ${GNUCC_TUNE})
message(STATUS "gcc will tune for ${GNUCC_ARCH}, ${TUNE_FLAG}") message(STATUS "gcc will tune for ${GNUCC_ARCH}, ${GNUCC_TUNE}")
endif() endif()
elseif (CMAKE_COMPILER_IS_CLANG AND NOT CROSS_COMPILE) elseif (CMAKE_COMPILER_IS_CLANG AND NOT CROSS_COMPILE)
if (ARCH_IA32 OR ARCH_X86_64) if (ARCH_IA32 OR ARCH_X86_64)
@ -226,22 +249,26 @@ if (ARCH_IA32 OR ARCH_X86_64)
endif() endif()
if (ARCH_AARCH64) if (ARCH_AARCH64)
if (BUILD_SVE2_BITPERM) if (BUILD_SVE2_BITPERM AND NOT SVE2_BITPERM_FOUND)
set(GNUCC_ARCH "${GNUCC_ARCH}+sve2-bitperm") set(GNUCC_ARCH "${GNUCC_ARCH}+sve2-bitperm")
elseif (BUILD_SVE2) elseif (BUILD_SVE2 AND NOT SVE2_FOUND)
set(GNUCC_ARCH "${GNUCC_ARCH}+sve2") set(GNUCC_ARCH "${GNUCC_ARCH}+sve2")
elseif (BUILD_SVE) elseif (BUILD_SVE AND NOT SVE_FOUND)
set(GNUCC_ARCH "${GNUCC_ARCH}+sve") set(GNUCC_ARCH "${GNUCC_ARCH}+sve")
endif () endif ()
endif(ARCH_AARCH64) endif(ARCH_AARCH64)
message(STATUS "ARCH_C_FLAGS : ${ARCH_C_FLAGS}") message(STATUS "ARCH_C_FLAGS : ${ARCH_C_FLAGS}")
message(STATUS "ARCH_CXX_FLAGS : ${ARCH_CXX_FLAGS}") message(STATUS "ARCH_CXX_FLAGS : ${ARCH_CXX_FLAGS}")
if (NOT FAT_RUNTIME) if (NOT FAT_RUNTIME)
if (GNUCC_TUNE)
set(ARCH_C_FLAGS "-${ARCH_FLAG}=${GNUCC_ARCH} -${TUNE_FLAG}=${GNUCC_TUNE}")
set(ARCH_CXX_FLAGS "-${ARCH_FLAG}=${GNUCC_ARCH} -${TUNE_FLAG}=${GNUCC_TUNE}")
else()
set(ARCH_C_FLAGS "-${ARCH_FLAG}=${GNUCC_ARCH} -mtune=${TUNE_FLAG} ${ARCH_C_FLAGS}") set(ARCH_C_FLAGS "-${ARCH_FLAG}=${GNUCC_ARCH} -mtune=${TUNE_FLAG} ${ARCH_C_FLAGS}")
set(ARCH_CXX_FLAGS "-${ARCH_FLAG}=${GNUCC_ARCH} -mtune=${TUNE_FLAG} ${ARCH_CXX_FLAGS}") set(ARCH_CXX_FLAGS "-${ARCH_FLAG}=${GNUCC_ARCH} -mtune=${TUNE_FLAG} ${ARCH_CXX_FLAGS}")
endif()
endif() endif()
#if (ARCH_IA32 OR ARCH_X86_64 OR ARCH_ARM32 OR ARCH_AARCH64) #if (ARCH_IA32 OR ARCH_X86_64 OR ARCH_ARM32 OR ARCH_AARCH64)
@ -296,6 +323,12 @@ if (NOT RELEASE_BUILD)
# release builds # release builds
set(EXTRA_C_FLAGS "${EXTRA_C_FLAGS} -Werror") set(EXTRA_C_FLAGS "${EXTRA_C_FLAGS} -Werror")
set(EXTRA_CXX_FLAGS "${EXTRA_CXX_FLAGS} -Werror") set(EXTRA_CXX_FLAGS "${EXTRA_CXX_FLAGS} -Werror")
if (CMAKE_COMPILER_IS_CLANG)
if (CMAKE_C_COMPILER_VERSION VERSION_GREATER "13.0")
set(EXTRA_C_FLAGS "${EXTRA_C_FLAGS} -Wno-unused-but-set-variable")
set(EXTRA_CXX_FLAGS "${EXTRA_CXX_FLAGS} -Wno-unused-but-set-variable")
endif()
endif()
endif() endif()
if (DISABLE_ASSERTS) if (DISABLE_ASSERTS)
@ -332,6 +365,7 @@ if (ARCH_IA32 OR ARCH_X86_64)
elseif (ARCH_ARM32 OR ARCH_AARCH64) elseif (ARCH_ARM32 OR ARCH_AARCH64)
CHECK_INCLUDE_FILE_CXX(arm_neon.h HAVE_C_ARM_NEON_H) CHECK_INCLUDE_FILE_CXX(arm_neon.h HAVE_C_ARM_NEON_H)
if (BUILD_SVE OR BUILD_SVE2 OR BUILD_SVE2_BITPERM) if (BUILD_SVE OR BUILD_SVE2 OR BUILD_SVE2_BITPERM)
set(CMAKE_REQUIRED_FLAGS ${ARCH_CXX_FLAGS})
CHECK_INCLUDE_FILE_CXX(arm_sve.h HAVE_C_ARM_SVE_H) CHECK_INCLUDE_FILE_CXX(arm_sve.h HAVE_C_ARM_SVE_H)
if (NOT HAVE_C_ARM_SVE_H) if (NOT HAVE_C_ARM_SVE_H)
message(FATAL_ERROR "arm_sve.h is required to build for SVE.") message(FATAL_ERROR "arm_sve.h is required to build for SVE.")

2
src/nfa/mcclellancompile.cpp
View File

@ -1484,12 +1484,12 @@ bytecode_ptr<NFA> mcclellanCompile_i(raw_dfa &raw, accel_dfa_build_strat &strat,
find_wide_state(info); find_wide_state(info);
} }
u16 total_daddy = 0;
bool any_cyclic_near_anchored_state bool any_cyclic_near_anchored_state
= is_cyclic_near(raw, raw.start_anchored); = is_cyclic_near(raw, raw.start_anchored);
// Sherman optimization // Sherman optimization
if (info.impl_alpha_size > 16) { if (info.impl_alpha_size > 16) {
u16 total_daddy = 0;
for (u32 i = 0; i < info.size(); i++) { for (u32 i = 0; i < info.size(); i++) {
if (info.is_widestate(i)) { if (info.is_widestate(i)) {
continue; continue;

3
src/nfagraph/ng_misc_opt.cpp
View File

@ -385,8 +385,7 @@ bool improveGraph(NGHolder &g, som_type som) {
const vector<NFAVertex> ordering = getTopoOrdering(g); const vector<NFAVertex> ordering = getTopoOrdering(g);
return enlargeCyclicCR(g, som, ordering) return enlargeCyclicCR(g, som, ordering) || enlargeCyclicCR_rev(g, ordering);
| enlargeCyclicCR_rev(g, ordering);
} }
/** finds a smaller reachability for a state by the reverse transformation of /** finds a smaller reachability for a state by the reverse transformation of

4
src/rose/rose_build_add.cpp
View File

@ -216,9 +216,9 @@ RoseRoleHistory selectHistory(const RoseBuildImpl &tbi, const RoseBuildData &bd,
const bool fixed_offset_src = g[u].fixedOffset(); const bool fixed_offset_src = g[u].fixedOffset();
const bool has_bounds = g[e].minBound || (g[e].maxBound != ROSE_BOUND_INF); const bool has_bounds = g[e].minBound || (g[e].maxBound != ROSE_BOUND_INF);
DEBUG_PRINTF("edge %zu->%zu, bounds=[%u,%u], fixed_u=%d, prefix=%d\n", /*DEBUG_PRINTF("edge %zu->%zu, bounds=[%u,%u], fixed_u=%d, prefix=%d\n",
g[u].index, g[v].index, g[e].minBound, g[e].maxBound, g[u].index, g[v].index, g[e].minBound, g[e].maxBound,
(int)g[u].fixedOffset(), (int)g[v].left); (int)g[u].fixedOffset(), (int)g[v].left);*/
if (g[v].left) { if (g[v].left) {
// Roles with prefix engines have their history handled by that prefix. // Roles with prefix engines have their history handled by that prefix.

34
src/util/arch/ppc64el/simd_utils.h
View File

@ -152,7 +152,7 @@ static really_inline u32 movemask128(m128 a) {
static uint8x16_t perm = { 16, 24, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }; static uint8x16_t perm = { 16, 24, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 };
uint8x16_t bitmask = vec_gb((uint8x16_t) a); uint8x16_t bitmask = vec_gb((uint8x16_t) a);
bitmask = (uint8x16_t) vec_perm(vec_splat_u8(0), bitmask, perm); bitmask = (uint8x16_t) vec_perm(vec_splat_u8(0), bitmask, perm);
u32 movemask; u32 ALIGN_ATTR(16) movemask;
vec_ste((uint32x4_t) bitmask, 0, &movemask); vec_ste((uint32x4_t) bitmask, 0, &movemask);
return movemask; return movemask;
} }
@ -285,27 +285,27 @@ m128 loadbytes128(const void *ptr, unsigned int n) {
return a; return a;
} }
#define CASE_ALIGN_VECTORS(a, b, offset) case offset: return (m128)vec_sld((int8x16_t)(b), (int8x16_t)(a), (16 - offset)); break; #define CASE_ALIGN_VECTORS(a, b, offset) case offset: return (m128)vec_sld((int8x16_t)(a), (int8x16_t)(b), (16 - offset)); break;
static really_really_inline static really_really_inline
m128 palignr_imm(m128 r, m128 l, int offset) { m128 palignr_imm(m128 r, m128 l, int offset) {
switch (offset) { switch (offset) {
case 0: return l; break; case 0: return l; break;
CASE_ALIGN_VECTORS(l, r, 1); CASE_ALIGN_VECTORS(r, l, 1);
CASE_ALIGN_VECTORS(l, r, 2); CASE_ALIGN_VECTORS(r, l, 2);
CASE_ALIGN_VECTORS(l, r, 3); CASE_ALIGN_VECTORS(r, l, 3);
CASE_ALIGN_VECTORS(l, r, 4); CASE_ALIGN_VECTORS(r, l, 4);
CASE_ALIGN_VECTORS(l, r, 5); CASE_ALIGN_VECTORS(r, l, 5);
CASE_ALIGN_VECTORS(l, r, 6); CASE_ALIGN_VECTORS(r, l, 6);
CASE_ALIGN_VECTORS(l, r, 7); CASE_ALIGN_VECTORS(r, l, 7);
CASE_ALIGN_VECTORS(l, r, 8); CASE_ALIGN_VECTORS(r, l, 8);
CASE_ALIGN_VECTORS(l, r, 9); CASE_ALIGN_VECTORS(r, l, 9);
CASE_ALIGN_VECTORS(l, r, 10); CASE_ALIGN_VECTORS(r, l, 10);
CASE_ALIGN_VECTORS(l, r, 11); CASE_ALIGN_VECTORS(r, l, 11);
CASE_ALIGN_VECTORS(l, r, 12); CASE_ALIGN_VECTORS(r, l, 12);
CASE_ALIGN_VECTORS(l, r, 13); CASE_ALIGN_VECTORS(r, l, 13);
CASE_ALIGN_VECTORS(l, r, 14); CASE_ALIGN_VECTORS(r, l, 14);
CASE_ALIGN_VECTORS(l, r, 15); CASE_ALIGN_VECTORS(r, l, 15);
case 16: return r; break; case 16: return r; break;
default: return zeroes128(); break; default: return zeroes128(); break;
} }

6
src/util/supervector/arch/ppc64el/impl.cpp
View File

@ -49,7 +49,7 @@ really_inline SuperVector<16>::SuperVector(SuperVector const &other)
template<> template<>
template<> template<>
really_inline SuperVector<16>::SuperVector(char __bool __vector v) really_inline SuperVector<16>::SuperVector(__vector __bool char v)
{ {
u.u8x16[0] = (uint8x16_t) v; u.u8x16[0] = (uint8x16_t) v;
}; };
@ -269,10 +269,10 @@ really_inline SuperVector<16> SuperVector<16>::eq(SuperVector<16> const &b) cons
template <> template <>
really_inline typename SuperVector<16>::comparemask_type really_inline typename SuperVector<16>::comparemask_type
SuperVector<16>::comparemask(void) const { SuperVector<16>::comparemask(void) const {
uint8x16_t bitmask = vec_gb( u.u8x16[0]);
static uint8x16_t perm = { 16, 24, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }; static uint8x16_t perm = { 16, 24, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 };
uint8x16_t bitmask = vec_gb(u.u8x16[0]);
bitmask = (uint8x16_t) vec_perm(vec_splat_u8(0), bitmask, perm); bitmask = (uint8x16_t) vec_perm(vec_splat_u8(0), bitmask, perm);
u32 movemask; u32 ALIGN_ATTR(16) movemask;
vec_ste((uint32x4_t) bitmask, 0, &movemask); vec_ste((uint32x4_t) bitmask, 0, &movemask);
return movemask; return movemask;
} }

2
src/util/supervector/arch/x86/impl.cpp
View File

@ -523,9 +523,7 @@ template <>
really_inline SuperVector<16> SuperVector<16>::loadu_maskz(void const *ptr, uint8_t const len) really_inline SuperVector<16> SuperVector<16>::loadu_maskz(void const *ptr, uint8_t const len)
{ {
SuperVector mask = Ones_vshr(16 -len); SuperVector mask = Ones_vshr(16 -len);
mask.print8("mask");
SuperVector v = _mm_loadu_si128((const m128 *)ptr); SuperVector v = _mm_loadu_si128((const m128 *)ptr);
v.print8("v");
return mask & v; return mask & v;
} }