mirror of
https://github.com/VectorCamp/vectorscan.git
synced 2025-06-28 16:41:01 +03:00
Merge pull request #237 from gtsoul-tech/feature/microbenchmarkingCSV
Microbenchmarking tool changed color output to csv output
This commit is contained in:
Konstantinos Margaritis
GitHub
commit
c2700cafd9
@ -26,32 +26,30 @@
|
||||
* 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"
|
||||
|
||||
#define MAX_LOOPS 1000000000
|
||||
#define MAX_MATCHES 5
|
||||
#define N 8
|
||||
#define MAX_LOOPS 1000000000
|
||||
#define MAX_MATCHES 5
|
||||
#define N 8
|
||||
|
||||
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,
|
||||
UNUSED struct hs_scratch *scratch) {
|
||||
static hwlmcb_rv_t hlmSimpleCallback(size_t to, u32 id,
|
||||
UNUSED struct hs_scratch *scratch) {
|
||||
DEBUG_PRINTF("match @%zu = %u\n", to, id);
|
||||
|
||||
ctxt.push_back(hlmMatchEntry(to, id));
|
||||
@ -59,8 +57,10 @@ hwlmcb_rv_t hlmSimpleCallback(size_t to, u32 id,
|
||||
return HWLM_CONTINUE_MATCHING;
|
||||
}
|
||||
|
||||
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) {
|
||||
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) {
|
||||
init(bench);
|
||||
double total_sec = 0.0;
|
||||
u64a total_size = 0;
|
||||
@ -70,29 +70,31 @@ static void run_benchmarks(int size, int loops, int max_matches, bool is_reverse
|
||||
double avg_time = 0.0;
|
||||
if (max_matches) {
|
||||
int pos = 0;
|
||||
for(int j = 0; j < max_matches - 1; j++) {
|
||||
for (int j = 0; j < max_matches - 1; j++) {
|
||||
bench.buf[pos] = 'b';
|
||||
pos = (j+1) *size / max_matches ;
|
||||
pos = (j + 1) * size / max_matches;
|
||||
bench.buf[pos] = 'a';
|
||||
u64a actual_size = 0;
|
||||
auto start = std::chrono::steady_clock::now();
|
||||
for(int i = 0; i < loops; i++) {
|
||||
for (int i = 0; i < loops; i++) {
|
||||
const u8 *res = func(bench);
|
||||
if (is_reverse)
|
||||
actual_size += bench.buf.data() + size - res;
|
||||
else
|
||||
actual_size += res - bench.buf.data();
|
||||
if (is_reverse)
|
||||
actual_size += bench.buf.data() + size - res;
|
||||
else
|
||||
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;
|
||||
/*std::cout << "act_size = " << act_size << std::endl;
|
||||
std::cout << "dt = " << dt << std::endl;
|
||||
std::cout << "bw = " << bw << std::endl;*/
|
||||
avg_bw += bw;
|
||||
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;*/
|
||||
avg_bw += bw;
|
||||
/*convert to MB/s*/
|
||||
max_bw = std::max(bw, max_bw);
|
||||
/*calculate average time*/
|
||||
@ -100,18 +102,20 @@ static void run_benchmarks(int size, int loops, int max_matches, bool is_reverse
|
||||
}
|
||||
avg_time /= max_matches;
|
||||
avg_bw /= max_matches;
|
||||
total_sec /= 1000000.0;
|
||||
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",
|
||||
bench.label, max_matches, size ,loops, total_sec, avg_time, max_bw, avg_bw);
|
||||
printf("%-18s, %-12u, %-10u, %-6u, %-10.3f, %-9.3f, %-8.3f, %-7.3f\n",
|
||||
bench.label, max_matches, size, loops, total_sec, avg_time,
|
||||
max_bw, avg_bw);
|
||||
} else {
|
||||
auto start = std::chrono::steady_clock::now();
|
||||
for (int i = 0; i < loops; i++) {
|
||||
const u8 *res = 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;
|
||||
/*calculate average time*/
|
||||
@ -122,117 +126,126 @@ 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, %-10u, %-6u, %-10.3f, %-9.3f, %-8.3f, %-7s\n",
|
||||
bench.label, "0", size, loops, total_sec, avg_time, max_bw, "0");
|
||||
}
|
||||
}
|
||||
|
||||
int main(){
|
||||
int main() {
|
||||
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++) {
|
||||
//we imitate the noodle unit tests
|
||||
// we imitate the noodle unit tests
|
||||
std::string str;
|
||||
const size_t char_len = 5;
|
||||
str.resize(char_len + 1);
|
||||
for (size_t j=0; j < char_len; j++) {
|
||||
srand (time(NULL));
|
||||
int key = rand() % + 36 ;
|
||||
for (size_t j = 0; j < char_len; j++) {
|
||||
srand(time(NULL));
|
||||
int key = rand() % +36;
|
||||
str[char_len] = charset[key];
|
||||
str[char_len + 1] = '\0';
|
||||
}
|
||||
|
||||
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);
|
||||
@ -242,10 +255,10 @@ int main(){
|
||||
assert(b.nt != 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;
|
||||
}
|
||||
);
|
||||
});
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@ -26,44 +26,32 @@
|
||||
* 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;
|
||||
char const *label;
|
||||
size_t size;
|
||||
|
||||
// Shufti/Truffle
|
||||
m128 lo, hi;
|
||||
ue2::CharReach chars;
|
||||
std::vector<u8> buf;
|
||||
// Shufti/Truffle
|
||||
m128 lo, hi;
|
||||
ue2::CharReach chars;
|
||||
std::vector<u8> buf;
|
||||
|
||||
// Noodle
|
||||
struct hs_scratch scratch;
|
||||
ue2::bytecode_ptr<noodTable> nt;
|
||||
// Noodle
|
||||
struct hs_scratch scratch;
|
||||
ue2::bytecode_ptr<noodTable> nt;
|
||||
|
||||
MicroBenchmark(char const *label_, size_t size_)
|
||||
:label(label_), size(size_), buf(size_) {
|
||||
};
|
||||
MicroBenchmark(char const *label_, size_t size_)
|
||||
: label(label_), size(size_), buf(size_){};
|
||||
};
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user