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
vectorscan/unit/internal/sheng.cpp
Konstantinos Margaritis f7d5546fe5
Bugfix/fix avx512vbmi regressions (#335) 
Multiple AVX512VBMI-related fixes:

src/nfa/mcsheng_compile.cpp: No need for an assert here, impl_id can be set to 0
src/nfa/nfa_api_queue.h: Make sure this compiles on both C++ and C
src/nfagraph/ng_fuzzy.cpp: Fix compilation error when DEBUG_OUTPUT=on
src/runtime.c: Fix crash when data == NULL
unit/internal/sheng.cpp: Unit test has to enable AVX512VBMI manually as autodetection does not get trigger, this causes test to fail
src/fdr/teddy_fat.cpp: AVX512 loads need to be 64-bit aligned, caused a crash on clang-18
2025-05-30 21:08:55 +03:00

721 lines
24 KiB
C++
Raw Permalink Blame History

/*
* Copyright (c) 2024, Arm ltd
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
*
* * Redistributions of source code must retain the above copyright notice,
* this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* * Neither the name of Intel Corporation nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
* LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
*/
#include "config.h"
#include "gtest/gtest.h"
#include "nfa/shengcompile.h"
#include "nfa/rdfa.h"
#include "util/bytecode_ptr.h"
#include "util/compile_context.h"
#include "util/report_manager.h"
extern "C" {
#include "hs_compile.h"
#include "nfa/nfa_api.h"
#include "nfa/nfa_api_queue.h"
#include "nfa/nfa_api_util.h"
#include "nfa/nfa_internal.h"
#include "nfa/rdfa.h"
#include "nfa/sheng.h"
#include "ue2common.h"
}
namespace {
struct callback_context {
unsigned int period;
unsigned int match_count;
unsigned int pattern_length;
};
int dummy_callback(u64a start, u64a end, ReportID id, void *context) {
(void) context;
printf("callback %llu %llu %u\n", start, end, id);
return 1; /* 0 stops matching, !0 continue */
}
int periodic_pattern_callback(u64a start, u64a end, ReportID id, void *raw_context) {
struct callback_context *context = (struct callback_context*) raw_context;
(void) start;
(void) id;
EXPECT_EQ(context->period * context->match_count, end - context->pattern_length);
context->match_count++;
return 1; /* 0 stops matching, !0 continue */
}
/**
* @brief Fill the state matrix with a diagonal pattern: accept the Nth character to go to the N+1 state
*/
static void fill_straight_regex_sequence(struct ue2::raw_dfa *dfa, int start_state, int end_state, int state_count)
{
for (int state = start_state; state < end_state; state++) {
dfa->states[state].next.assign(state_count ,1);
dfa->states[state].next[0] = 2;
dfa->states[state].next[1] = 2;
dfa->states[state].next[state] = state+1;
}
}
static void init_raw_dfa16(struct ue2::raw_dfa *dfa, const ReportID rID)
{
dfa->start_anchored = 1;
dfa->start_floating = 1;
dfa->alpha_size = 8;
int nb_state = 8;
for(int i = 0; i < nb_state; i++) {
struct ue2::dstate state(dfa->alpha_size);
state.next = std::vector<ue2::dstate_id_t>(nb_state);
state.daddy = 0;
state.impl_id = i; /* id of the state */
state.reports = ue2::flat_set<ReportID>();
state.reports_eod = ue2::flat_set<ReportID>();
dfa->states.push_back(state);
}
/* add a report to every accept state */
dfa->states[7].reports.insert(rID);
/**
* [a,b][c-e]{3}of
* (1) -a,b-> (2) -c,d,e-> (3) -c,d,e-> (4) -c,d,e-> (5) -o-> (6) -f-> ((7))
* (0) = dead
*/
for(int i = 0; i < ue2::ALPHABET_SIZE; i++) {
dfa->alpha_remap[i] = 0;
}
dfa->alpha_remap['a'] = 0;
dfa->alpha_remap['b'] = 1;
dfa->alpha_remap['c'] = 2;
dfa->alpha_remap['d'] = 3;
dfa->alpha_remap['e'] = 4;
dfa->alpha_remap['o'] = 5;
dfa->alpha_remap['f'] = 6;
dfa->alpha_remap[256] = 7; /* for some reason there's a check that run on dfa->alpha_size-1 */
/* a b c d e o f */
dfa->states[0].next = {0,0,0,0,0,0,0};
dfa->states[1].next = {2,2,1,1,1,1,1}; /* nothing */
dfa->states[2].next = {2,2,3,3,3,1,1}; /* [a,b] */
dfa->states[3].next = {2,2,4,4,4,1,1}; /* [a,b][c-e]{1} */
dfa->states[4].next = {2,2,5,5,5,1,1}; /* [a,b][c-e]{2} */
fill_straight_regex_sequence(dfa, 5, 7, 7); /* [a,b][c-e]{3}o */
dfa->states[7].next = {2,2,1,1,1,1,1}; /* [a,b][c-e]{3}of */
}
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
/* We need more than 16 states to run sheng32, so make the graph longer */
static void init_raw_dfa32(struct ue2::raw_dfa *dfa, const ReportID rID)
{
dfa->start_anchored = 1;
dfa->start_floating = 1;
dfa->alpha_size = 18;
int nb_state = 18;
for(int i = 0; i < nb_state; i++) {
struct ue2::dstate state(dfa->alpha_size);
state.next = std::vector<ue2::dstate_id_t>(nb_state);
state.daddy = 0;
state.impl_id = i; /* id of the state */
state.reports = ue2::flat_set<ReportID>();
state.reports_eod = ue2::flat_set<ReportID>();
dfa->states.push_back(state);
}
/* add a report to every accept state */
dfa->states[17].reports.insert(rID);
/**
* [a,b][c-e]{3}of0123456789
* (1) -a,b-> (2) -c,d,e-> (3) -c,d,e-> (4) -c,d,e-> (5) -o-> (6) -f-> (7) -<numbers>-> ((17))
* (0) = dead
*/
for(int i = 0; i < ue2::ALPHABET_SIZE; i++) {
dfa->alpha_remap[i] = 0;
}
dfa->alpha_remap['a'] = 0;
dfa->alpha_remap['b'] = 1;
dfa->alpha_remap['c'] = 2;
dfa->alpha_remap['d'] = 3;
dfa->alpha_remap['e'] = 4;
dfa->alpha_remap['o'] = 5;
dfa->alpha_remap['f'] = 6;
// maps 0 to 9
for (int i = 0; i < 10; i ++) {
dfa->alpha_remap[i + '0'] = i + 7;
}
dfa->alpha_remap[256] = 17; /* for some reason there's a check that run on dfa->alpha_size-1 */
/* a b c d e o f 0 1 2 3 4 5 6 7 8 9 */
dfa->states[0].next = {0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0};
dfa->states[1].next = {2,2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}; /* nothing */
dfa->states[2].next = {2,2,3,3,3,1,1,1,1,1,1,1,1,1,1,1,1}; /* [a,b] */
dfa->states[3].next = {2,2,4,4,4,1,1,1,1,1,1,1,1,1,1,1,1}; /* [a,b][c-e]{1} */
dfa->states[4].next = {2,2,5,5,5,1,1,1,1,1,1,1,1,1,1,1,1}; /* [a,b][c-e]{2} */
fill_straight_regex_sequence(dfa, 5, 17, 17); /* [a,b][c-e]{3}of012345678 */
dfa->states[17].next = {2,2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}; /* [a,b][c-e]{3}of0123456789 */
}
#endif /* defined(HAVE_AVX512VBMI) || defined(HAVE_SVE) */
typedef ue2::bytecode_ptr<NFA> (*sheng_compile_ptr)(ue2::raw_dfa&,
const ue2::CompileContext&,
const ue2::ReportManager&,
bool,
std::set<ue2::dstate_id_t>*);
typedef void (*init_raw_dfa_ptr)(struct ue2::raw_dfa*, const ReportID);
static inline void init_nfa(struct NFA **out_nfa, sheng_compile_ptr compile_function, init_raw_dfa_ptr init_dfa_function) {
ue2::Grey *g = new ue2::Grey();
#if defined(HAVE_AVX512VBMI)
hs_platform_info plat_info = {0, HS_CPU_FEATURES_AVX512VBMI, 0, 0};
#else
hs_platform_info plat_info = {0, 0, 0, 0};
#endif
ue2::CompileContext *cc = new ue2::CompileContext(false, false, ue2::target_t(plat_info), *g);
ue2::ReportManager *rm = new ue2::ReportManager(*g);
ue2::Report *report = new ue2::Report(ue2::EXTERNAL_CALLBACK, 0);
ReportID rID = rm->getInternalId(*report);
rm->setProgramOffset(0, 0);
struct ue2::raw_dfa *dfa = new ue2::raw_dfa(ue2::NFA_OUTFIX);
init_dfa_function(dfa, rID);
*out_nfa = (compile_function(*dfa, *cc, *rm, false, nullptr)).release();
ASSERT_NE(nullptr, *out_nfa);
delete report;
delete rm;
delete cc;
delete g;
}
static void init_nfa16(struct NFA **out_nfa) {
init_nfa(out_nfa, ue2::shengCompile, init_raw_dfa16);
}
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
static void init_nfa32(struct NFA **out_nfa) {
init_nfa(out_nfa, ue2::sheng32Compile, init_raw_dfa32);
}
#endif /* defined(HAVE_AVX512VBMI) || defined(HAVE_SVE) */
static char state_buffer;
static inline void init_sheng_queue(struct mq **out_q, uint8_t *buffer, size_t max_size, void (*init_nfa_func)(struct NFA **out_nfa) ) {
struct NFA* nfa;
init_nfa_func(&nfa);
assert(nfa);
struct mq *q = new mq();
memset(q, 0, sizeof(struct mq));
q->nfa = nfa;
q->state = &state_buffer;
q->cb = dummy_callback;
q->buffer = buffer;
q->length = max_size; /* setting this as the max length scanable */
if (nfa != q->nfa) {
printf("Something went wrong while initializing sheng.\n");
}
nfaQueueInitState(nfa, q);
pushQueueAt(q, 0, MQE_START, 0);
pushQueueAt(q, 1, MQE_END, q->length );
*out_q = q;
}
static void init_sheng_queue16(struct mq **out_q, uint8_t *buffer ,size_t max_size) {
init_sheng_queue(out_q, buffer, max_size, init_nfa16);
}
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
static void init_sheng_queue32(struct mq **out_q, uint8_t *buffer, size_t max_size) {
init_sheng_queue(out_q, buffer, max_size, init_nfa32);
}
#endif /* defined(HAVE_AVX512VBMI) || defined(HAVE_SVE) */
static
void fill_pattern(u8* buf, size_t buffer_size, unsigned int start_offset, unsigned int period, const char *pattern, unsigned int pattern_length) {
memset(buf, '_', buffer_size);
for (unsigned int i = 0; i < buffer_size - 8; i+= 8) {
/* filling with some junk, including some character used for a valid state, to prevent the use of shufti */
memcpy(buf + i, "jgohcxbf", 8);
}
for (unsigned int i = start_offset; i < buffer_size - pattern_length; i += period) {
memcpy(buf + i, pattern, pattern_length);
}
}
/* Generate ground truth to compare to */
struct NFA *get_expected_nfa_header(u8 type, unsigned int length, unsigned int nposition) {
struct NFA *expected_nfa_header = new struct NFA();
memset(expected_nfa_header, 0, sizeof(struct NFA));
expected_nfa_header->length = length;
expected_nfa_header->type = type;
expected_nfa_header->nPositions = nposition;
expected_nfa_header->scratchStateSize = 1;
expected_nfa_header->streamStateSize = 1;
return expected_nfa_header;
}
struct NFA *get_expected_nfa16_header() {
return get_expected_nfa_header(SHENG_NFA, 4736, 8); /* size recorded in 04/2024 */
}
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
struct NFA *get_expected_nfa32_header() {
return get_expected_nfa_header(SHENG_NFA_32, 17216, 18); /* size recorded in 04/2024 */
}
#endif /* defined(HAVE_AVX512VBMI) || defined(HAVE_SVE) */
void test_nfa_equal(const NFA& l, const NFA& r)
{
/**
* The length is meant to be a sanity test: it's not 0 (we compiled something) and that it roughly fit the
* expected size for a given sheng implementation (we don't feed compiled sheng32 into sheng16).
* Changes in other nfa algorithms may affect the sheng length, so we accept small variations.
*/
int relative_difference = std::abs((float)(l.length) - r.length) / ((l.length + r.length) / 2);
EXPECT_LE(relative_difference, 0.1); /* same +-10% */
EXPECT_EQ(l.flags, r.flags);
EXPECT_EQ(l.type, r.type);
EXPECT_EQ(l.rAccelType, r.rAccelType);
EXPECT_EQ(l.rAccelOffset, r.rAccelOffset);
EXPECT_EQ(l.maxBiAnchoredWidth, r.maxBiAnchoredWidth);
EXPECT_EQ(l.rAccelData.dc, r.rAccelData.dc);
EXPECT_EQ(l.queueIndex, r.queueIndex);
EXPECT_EQ(l.nPositions, r.nPositions);
EXPECT_EQ(l.scratchStateSize, r.scratchStateSize);
EXPECT_EQ(l.streamStateSize, r.streamStateSize);
EXPECT_EQ(l.maxWidth, r.maxWidth);
EXPECT_EQ(l.minWidth, r.minWidth);
EXPECT_EQ(l.maxOffset, r.maxOffset);
}
/* Start of actual tests */
/*
* Runs shengCompile and compares its outputs to previously recorded outputs.
*/
TEST(Sheng16, std_compile_header) {
ue2::Grey *g = new ue2::Grey();
hs_platform_info plat_info = {0, 0, 0, 0};
ue2::CompileContext *cc = new ue2::CompileContext(false, false, ue2::target_t(plat_info), *g);
ue2::ReportManager *rm = new ue2::ReportManager(*g);
ue2::Report *report = new ue2::Report(ue2::EXTERNAL_CALLBACK, 0);
ReportID rID = rm->getInternalId(*report);
rm->setProgramOffset(0, 0);
struct ue2::raw_dfa *dfa = new ue2::raw_dfa(ue2::NFA_OUTFIX);
init_raw_dfa16(dfa, rID);
struct NFA *nfa = (shengCompile(*dfa, *cc, *rm, false)).release();
EXPECT_NE(nullptr, nfa);
EXPECT_NE(0, nfa->length);
EXPECT_EQ(SHENG_NFA, nfa->type);
struct NFA *expected_nfa = get_expected_nfa16_header();
test_nfa_equal(*expected_nfa, *nfa);
delete expected_nfa;
delete report;
delete rm;
delete cc;
delete g;
}
/*
* nfaExecSheng_B is the most basic of the sheng variants. It simply calls the core of the algorithm.
* We test it with a buffer having a few matches at fixed intervals and check that it finds them all.
*/
TEST(Sheng16, std_run_B) {
struct mq *q;
unsigned int pattern_length = 6;
unsigned int period = 128;
const size_t buf_size = 200;
unsigned int expected_matches = buf_size/128 + 1;
u8 buf[buf_size];
struct callback_context context = {period, 0, pattern_length};
struct NFA* nfa;
init_nfa16(&nfa);
ASSERT_NE(nullptr, nfa);
fill_pattern(buf, buf_size, 0, period, "acecof", pattern_length);
char ret_val;
unsigned int offset = 0;
unsigned int loop_count = 0;
for (; loop_count < expected_matches + 1; loop_count++) {
ASSERT_LT(offset, buf_size);
ret_val = nfaExecSheng_B(nfa,
offset,
buf + offset,
(s64a) buf_size - offset,
periodic_pattern_callback,
&context);
offset = (context.match_count - 1) * context.period + context.pattern_length;
if(unlikely(ret_val != MO_ALIVE)) {
break;
}
}
/*check normal return*/
EXPECT_EQ(MO_ALIVE, ret_val);
/*check that we don't find additional match nor crash when no match are found*/
EXPECT_EQ(expected_matches + 1, loop_count);
/*check that we have all the matches*/
EXPECT_EQ(expected_matches, context.match_count);
}
/*
* nfaExecSheng_Q runs like the _B version (callback), but exercises the message queue logic.
* We test it with a buffer having a few matches at fixed intervals and check that it finds them all.
*/
TEST(Sheng16, std_run_Q) {
struct mq *q;
unsigned int pattern_length = 6;
unsigned int period = 128;
const size_t buf_size = 200;
unsigned int expected_matches = buf_size/128 + 1;
u8 buf[buf_size];
struct callback_context context = {period, 0, pattern_length};
init_sheng_queue16(&q, buf, buf_size);
fill_pattern(buf, buf_size, 0, period, "acecof", pattern_length);
q->cur = 0;
q->items[q->cur].location = 0;
q->context = &context;
q->cb = periodic_pattern_callback;
nfaExecSheng_Q(q->nfa, q, (s64a) buf_size);
/*check that we have all the matches*/
EXPECT_EQ(expected_matches, context.match_count);
delete q;
}
/*
* nfaExecSheng_Q2 uses the message queue, but stops at match instead of using a callback.
* We test it with a buffer having a few matches at fixed intervals and check that it finds them all.
*/
TEST(Sheng16, std_run_Q2) {
struct mq *q;
unsigned int pattern_length = 6;
unsigned int period = 128;
const size_t buf_size = 200;
unsigned int expected_matches = buf_size/128 + 1;
u8 buf[buf_size];
init_sheng_queue16(&q, buf, buf_size);
fill_pattern(buf, buf_size, 0, period, "acecof", pattern_length);
q->cur = 0;
q->items[q->cur].location = 0;
char ret_val;
int location;
unsigned int loop_count = 0;
do {
ret_val = nfaExecSheng_Q2(q->nfa, q, (s64a) buf_size);
location = q->items[q->cur].location;
loop_count++;
} while(likely((ret_val == MO_MATCHES_PENDING) && (location < (int)buf_size) && ((location % period) == pattern_length)));
/*check if it's a spurious match*/
EXPECT_EQ(0, (ret_val == MO_MATCHES_PENDING) && ((location % period) != pattern_length));
/*check that we have all the matches*/
EXPECT_EQ(expected_matches, loop_count-1);
delete q;
}
/*
* The message queue can also run on the "history" buffer. We test it the same way as the normal
* buffer, expecting the same behavior.
* We test it with a buffer having a few matches at fixed intervals and check that it finds them all.
*/
TEST(Sheng16, history_run_Q2) {
struct mq *q;
unsigned int pattern_length = 6;
unsigned int period = 128;
const size_t buf_size = 200;
unsigned int expected_matches = buf_size/128 + 1;
u8 buf[buf_size];
init_sheng_queue16(&q, buf, buf_size);
fill_pattern(buf, buf_size, 0, period, "acecof", pattern_length);
q->history = buf;
q->hlength = buf_size;
q->cur = 0;
q->items[q->cur].location = -200;
char ret_val;
int location;
unsigned int loop_count = 0;
do {
ret_val = nfaExecSheng_Q2(q->nfa, q, 0);
location = q->items[q->cur].location;
loop_count++;
} while(likely((ret_val == MO_MATCHES_PENDING) && (location > -(int)buf_size) && (location < 0) && (((buf_size + location) % period) == pattern_length)));
/*check if it's a spurious match*/
EXPECT_EQ(0, (ret_val == MO_MATCHES_PENDING) && (((buf_size + location) % period) != pattern_length));
/*check that we have all the matches*/
EXPECT_EQ(expected_matches, loop_count-1);
delete q;
}
/**
* Those tests only covers the basic paths. More tests can cover:
* - running for history buffer to current buffer in Q2
* - running while expecting no match
* - nfaExecSheng_QR
* - run sheng when it should call an accelerator and confirm it call them
*/
#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
/*
* Runs sheng32Compile and compares its outputs to previously recorded outputs.
*/
TEST(Sheng32, std_compile_header) {
#if defined(HAVE_SVE)
if(svcntb()<32) {
return;
}
#endif
ue2::Grey *g = new ue2::Grey();
hs_platform_info plat_info = {0, HS_CPU_FEATURES_AVX512VBMI, 0, 0};
ue2::CompileContext *cc = new ue2::CompileContext(false, false, ue2::target_t(plat_info), *g);
ue2::ReportManager *rm = new ue2::ReportManager(*g);
ue2::Report *report = new ue2::Report(ue2::EXTERNAL_CALLBACK, 0);
ReportID rID = rm->getInternalId(*report);
rm->setProgramOffset(0, 0);
struct ue2::raw_dfa *dfa = new ue2::raw_dfa(ue2::NFA_OUTFIX);
init_raw_dfa32(dfa, rID);
struct NFA *nfa = (sheng32Compile(*dfa, *cc, *rm, false)).release();
EXPECT_NE(nullptr, nfa);
EXPECT_NE(0, nfa->length);
EXPECT_EQ(SHENG_NFA_32, nfa->type);
struct NFA *expected_nfa = get_expected_nfa32_header();
test_nfa_equal(*expected_nfa, *nfa);
delete expected_nfa;
delete report;
delete rm;
delete cc;
delete g;
}
/*
* nfaExecSheng32_B is the most basic of the sheng variants. It simply calls the core of the algorithm.
* We test it with a buffer having a few matches at fixed intervals and check that it finds them all.
*/
TEST(Sheng32, std_run_B) {
#if defined(HAVE_SVE)
if(svcntb()<32) {
return;
}
#endif
struct mq *q;
unsigned int pattern_length = 16;
unsigned int period = 128;
const size_t buf_size = 200;
unsigned int expected_matches = buf_size/128 + 1;
u8 buf[buf_size];
struct callback_context context = {period, 0, pattern_length};
struct NFA* nfa;
init_nfa32(&nfa);
ASSERT_NE(nullptr, nfa);
fill_pattern(buf, buf_size, 0, period, "acecof0123456789", pattern_length);
char ret_val;
unsigned int offset = 0;
unsigned int loop_count = 0;
for (; loop_count < expected_matches + 1; loop_count++) {
ASSERT_LT(offset, buf_size);
ret_val = nfaExecSheng32_B(nfa,
offset,
buf + offset,
(s64a) buf_size - offset,
periodic_pattern_callback,
&context);
offset = (context.match_count - 1) * context.period + context.pattern_length;
if(unlikely(ret_val != MO_ALIVE)) {
break;
}
}
/*check normal return*/
EXPECT_EQ(MO_ALIVE, ret_val);
/*check that we don't find additional match nor crash when no match are found*/
EXPECT_EQ(expected_matches + 1, loop_count);
/*check that we have all the matches*/
EXPECT_EQ(expected_matches, context.match_count);
}
/*
* nfaExecSheng32_Q runs like the _B version (callback), but exercises the message queue logic.
* We test it with a buffer having a few matches at fixed intervals and check that it finds them all.
*/
TEST(Sheng32, std_run_Q) {
#if defined(HAVE_SVE)
if(svcntb()<32) {
return;
}
#endif
struct mq *q;
unsigned int pattern_length = 16;
unsigned int period = 128;
const size_t buf_size = 200;
unsigned int expected_matches = buf_size/128 + 1;
u8 buf[buf_size];
struct callback_context context = {period, 0, pattern_length};
init_sheng_queue32(&q, buf, buf_size);
fill_pattern(buf, buf_size, 0, period, "acecof0123456789", pattern_length);
q->cur = 0;
q->items[q->cur].location = 0;
q->context = &context;
q->cb = periodic_pattern_callback;
nfaExecSheng32_Q(q->nfa, q, (s64a) buf_size);
/*check that we have all the matches*/
EXPECT_EQ(expected_matches, context.match_count);
delete q;
}
/*
* nfaExecSheng32_Q2 uses the message queue, but stops at match instead of using a callback.
* We test it with a buffer having a few matches at fixed intervals and check that it finds them all.
*/
TEST(Sheng32, std_run_Q2) {
#if defined(HAVE_SVE)
if(svcntb()<32) {
return;
}
#endif
struct mq *q;
unsigned int pattern_length = 16;
unsigned int period = 128;
const size_t buf_size = 200;
unsigned int expected_matches = buf_size/128 + 1;
u8 buf[buf_size];
init_sheng_queue32(&q, buf, buf_size);
fill_pattern(buf, buf_size, 0, period, "acecof0123456789", pattern_length);
q->cur = 0;
q->items[q->cur].location = 0;
char ret_val;
int location;
unsigned int loop_count = 0;
do {
ret_val = nfaExecSheng32_Q2(q->nfa, q, (s64a) buf_size);
location = q->items[q->cur].location;
loop_count++;
} while(likely((ret_val == MO_MATCHES_PENDING) && (location < (int)buf_size) && ((location % period) == pattern_length)));
/*check if it's a spurious match*/
EXPECT_EQ(0, (ret_val == MO_MATCHES_PENDING) && ((location % period) != pattern_length));
/*check that we have all the matches*/
EXPECT_EQ(expected_matches, loop_count-1);
delete q;
}
/*
* The message queue can also runs on the "history" buffer. We test it the same way as the normal
* buffer, expecting the same behavior.
* We test it with a buffer having a few matches at fixed intervals and check that it finds them all.
*/
TEST(Sheng32, history_run_Q2) {
#if defined(HAVE_SVE)
if(svcntb()<32) {
return;
}
#endif
struct mq *q;
unsigned int pattern_length = 16;
unsigned int period = 128;
const size_t buf_size = 200;
unsigned int expected_matches = buf_size/128 + 1;
u8 buf[buf_size];
init_sheng_queue32(&q, buf, buf_size);
fill_pattern(buf, buf_size, 0, period, "acecof0123456789", pattern_length);
q->history = buf;
q->hlength = buf_size;
q->cur = 0;
q->items[q->cur].location = -200;
char ret_val;
int location;
unsigned int loop_count = 0;
do {
ret_val = nfaExecSheng32_Q2(q->nfa, q, 0);
location = q->items[q->cur].location;
loop_count++;
} while(likely((ret_val == MO_MATCHES_PENDING) && (location > -(int)buf_size) && (location < 0) && (((buf_size + location) % period) == pattern_length)));
/*check if it's a spurious match*/
EXPECT_EQ(0, (ret_val == MO_MATCHES_PENDING) && (((buf_size + location) % period) != pattern_length));
/*check that we have all the matches*/
EXPECT_EQ(expected_matches, loop_count-1);
delete q;
}
#endif /* defined(HAVE_AVX512VBMI) || defined(HAVE_SVE) */
} /* namespace */
Reference in New Issue View Git Blame Copy Permalink