1035 lines
32 KiB
C
1035 lines
32 KiB
C
#ifdef DS_PERF
|
|
#define _CRT_SECURE_NO_WARNINGS
|
|
#define _CRT_NONSTDC_NO_DEPRECATE
|
|
#define _CRT_NON_CONFORMING_SWPRINTFS
|
|
|
|
|
|
//#define STBDS_INTERNAL_SMALL_BUCKET // make 64-bit bucket fit both keys and hash bits
|
|
//#define STBDS_SIPHASH_2_4 // performance test 1_3 against 2_4
|
|
//#define STBDS_INTERNAL_BUCKET_START // don't bother offseting differently within bucket for different hash values
|
|
//#define STBDS_FLUSH_CACHE (1u<<20) // do this much memory traffic to flush the cache between some benchmarking measurements
|
|
|
|
#include <stdio.h>
|
|
|
|
#define WIN32_LEAN_AND_MEAN
|
|
#include <windows.h>
|
|
#define STB_DEFINE
|
|
#define STB_NO_REGISTRY
|
|
#include "../stb.h"
|
|
#endif
|
|
|
|
#ifdef DS_TEST
|
|
#define STBDS_UNIT_TESTS
|
|
#define STBDS_SMALL_BUCKET
|
|
#endif
|
|
|
|
#ifdef DS_STATS
|
|
#define STBDS_STATISTICS
|
|
#endif
|
|
|
|
#ifndef DS_PERF
|
|
#define STBDS_ASSERT assert
|
|
#include <assert.h>
|
|
#endif
|
|
|
|
#define STB_DS_IMPLEMENTATION
|
|
#include "../stb_ds.h"
|
|
|
|
size_t churn_inserts, churn_deletes;
|
|
|
|
void churn(int a, int b, int count)
|
|
{
|
|
struct { int key,value; } *map=NULL;
|
|
int i,j,n,k;
|
|
for (i=0; i < a; ++i)
|
|
hmput(map,i,i+1);
|
|
for (n=0; n < count; ++n) {
|
|
for (j=a; j < b; ++j,++i) {
|
|
hmput(map,i,i+1);
|
|
}
|
|
assert(hmlen(map) == b);
|
|
for (j=a; j < b; ++j) {
|
|
k=i-j-1;
|
|
k = hmdel(map,k);
|
|
assert(k != 0);
|
|
}
|
|
assert(hmlen(map) == a);
|
|
}
|
|
hmfree(map);
|
|
churn_inserts = i;
|
|
churn_deletes = (b-a) * n;
|
|
}
|
|
|
|
#ifdef DS_TEST
|
|
#include <stdio.h>
|
|
int main(int argc, char **argv)
|
|
{
|
|
char *temp=NULL;
|
|
stbds_unit_tests();
|
|
arrins(temp, 0, 'a');
|
|
arrins(temp, arrlen(temp), 'b');
|
|
churn(0,100,1);
|
|
churn(3,7,50000);
|
|
churn(3,15,50000);
|
|
churn(16, 48, 25000);
|
|
churn(10, 15, 25000);
|
|
churn(200,500, 5000);
|
|
churn(2000,5000, 500);
|
|
churn(20000,50000, 50);
|
|
printf("Ok!");
|
|
return 0;
|
|
}
|
|
#endif
|
|
|
|
#ifdef DS_STATS
|
|
#define MAX(a,b) ((a) > (b) ? (a) : (b))
|
|
size_t max_hit_probes, max_miss_probes, total_put_probes, total_miss_probes, churn_misses;
|
|
void churn_stats(int a, int b, int count)
|
|
{
|
|
struct { int key,value; } *map=NULL;
|
|
int i,j,n,k;
|
|
churn_misses = 0;
|
|
for (i=0; i < a; ++i) {
|
|
hmput(map,i,i+1);
|
|
max_hit_probes = MAX(max_hit_probes, stbds_hash_probes);
|
|
total_put_probes += stbds_hash_probes;
|
|
stbds_hash_probes = 0;
|
|
}
|
|
|
|
for (n=0; n < count; ++n) {
|
|
for (j=a; j < b; ++j,++i) {
|
|
hmput(map,i,i+1);
|
|
max_hit_probes = MAX(max_hit_probes, stbds_hash_probes);
|
|
total_put_probes += stbds_hash_probes;
|
|
stbds_hash_probes = 0;
|
|
}
|
|
for (j=0; j < (b-a)*10; ++j) {
|
|
k=i+j;
|
|
(void) hmgeti(map,k); // miss
|
|
max_miss_probes = MAX(max_miss_probes, stbds_hash_probes);
|
|
total_miss_probes += stbds_hash_probes;
|
|
stbds_hash_probes = 0;
|
|
++churn_misses;
|
|
}
|
|
assert(hmlen(map) == b);
|
|
for (j=a; j < b; ++j) {
|
|
k=i-j-1;
|
|
k = hmdel(map,k);
|
|
stbds_hash_probes = 0;
|
|
assert(k);
|
|
}
|
|
assert(hmlen(map) == a);
|
|
}
|
|
hmfree(map);
|
|
churn_inserts = i;
|
|
churn_deletes = (b-a) * n;
|
|
}
|
|
|
|
void reset_stats(void)
|
|
{
|
|
stbds_array_grow=0,
|
|
stbds_hash_grow=0;
|
|
stbds_hash_shrink=0;
|
|
stbds_hash_rebuild=0;
|
|
stbds_hash_probes=0;
|
|
stbds_hash_alloc=0;
|
|
stbds_rehash_probes=0;
|
|
stbds_rehash_items=0;
|
|
max_hit_probes = 0;
|
|
max_miss_probes = 0;
|
|
total_put_probes = 0;
|
|
total_miss_probes = 0;
|
|
}
|
|
|
|
void print_churn_probe_stats(char *str)
|
|
{
|
|
printf("Probes: %3d max hit, %3d max miss, %4.2f avg hit, %4.2f avg miss: %s\n",
|
|
(int) max_hit_probes, (int) max_miss_probes, (float) total_put_probes / churn_inserts, (float) total_miss_probes / churn_misses, str);
|
|
reset_stats();
|
|
}
|
|
|
|
int main(int arg, char **argv)
|
|
{
|
|
churn_stats(0,500000,1); print_churn_probe_stats("Inserting 500000 items");
|
|
churn_stats(0,500000,1); print_churn_probe_stats("Inserting 500000 items");
|
|
churn_stats(0,500000,1); print_churn_probe_stats("Inserting 500000 items");
|
|
churn_stats(0,500000,1); print_churn_probe_stats("Inserting 500000 items");
|
|
churn_stats(49000,50000,500); print_churn_probe_stats("Deleting/Inserting 500000 items");
|
|
churn_stats(49000,50000,500); print_churn_probe_stats("Deleting/Inserting 500000 items");
|
|
churn_stats(49000,50000,500); print_churn_probe_stats("Deleting/Inserting 500000 items");
|
|
churn_stats(49000,50000,500); print_churn_probe_stats("Deleting/Inserting 500000 items");
|
|
return 0;
|
|
}
|
|
#endif
|
|
|
|
|
|
#ifdef DS_PERF
|
|
//char *strdup(const char *foo) { return 0; }
|
|
//int stricmp(const char *a, const char *b) { return 0; }
|
|
//int strnicmp(const char *a, const char *b, size_t n) { return 0; }
|
|
|
|
unsigned __int64 t0, xsum, mn,mx,count;
|
|
void begin(void)
|
|
{
|
|
LARGE_INTEGER m;
|
|
QueryPerformanceCounter(&m);
|
|
t0 = m.QuadPart;
|
|
xsum = 0;
|
|
count = 0;
|
|
mx = 0;
|
|
mn = ~(unsigned __int64) 0;
|
|
}
|
|
|
|
void measure(void)
|
|
{
|
|
unsigned __int64 t1, t;
|
|
LARGE_INTEGER m;
|
|
QueryPerformanceCounter(&m);
|
|
t1 = m.QuadPart;
|
|
t = t1-t0;
|
|
if (t1 < t0)
|
|
printf("ALERT: QueryPerformanceCounter was unordered!\n");
|
|
if (t < mn) mn = t;
|
|
if (t > mx) mx = t;
|
|
xsum += t;
|
|
++count;
|
|
t0 = t1;
|
|
}
|
|
|
|
void dont_measure(void)
|
|
{
|
|
LARGE_INTEGER m;
|
|
QueryPerformanceCounter(&m);
|
|
t0 = m.QuadPart;
|
|
}
|
|
|
|
double timer;
|
|
double end(void)
|
|
{
|
|
LARGE_INTEGER m;
|
|
QueryPerformanceFrequency(&m);
|
|
|
|
if (count > 3) {
|
|
// discard the highest and lowest
|
|
xsum -= mn;
|
|
xsum -= mx;
|
|
count -= 2;
|
|
}
|
|
timer = (double) (xsum) / count / m.QuadPart * 1000;
|
|
return timer;
|
|
}
|
|
|
|
void build(int a, int b, int count, int step)
|
|
{
|
|
struct { int key,value; } *map=NULL;
|
|
int i,n;
|
|
for (i=0; i < a; ++i) {
|
|
n = i*step;
|
|
hmput(map,n,i+1);
|
|
}
|
|
measure();
|
|
churn_inserts = i;
|
|
hmfree(map);
|
|
dont_measure();
|
|
}
|
|
|
|
#ifdef STB__INCLUDE_STB_H
|
|
void build_stb(int a, int b, int count, int step)
|
|
{
|
|
stb_idict *d = stb_idict_new_size(8);
|
|
int i;
|
|
for (i=0; i < a; ++i)
|
|
stb_idict_add(d, i*step, i+1);
|
|
measure();
|
|
churn_inserts = i;
|
|
stb_idict_destroy(d);
|
|
dont_measure();
|
|
}
|
|
|
|
void multibuild_stb(int a, int b, int count, int step, int tables)
|
|
{
|
|
stb_idict *d[50000];
|
|
int i,q;
|
|
for (q=0; q < tables; ++q)
|
|
d[q] = stb_idict_new_size(8);
|
|
dont_measure();
|
|
for (i=0; i < a; ++i)
|
|
for (q=0; q < tables; ++q)
|
|
stb_idict_add(d[q], i*step+q*771, i+1);
|
|
measure();
|
|
churn_inserts = i;
|
|
for (q=0; q < tables; ++q)
|
|
stb_idict_destroy(d[q]);
|
|
dont_measure();
|
|
}
|
|
|
|
int multisearch_stb(int a, int start, int end, int step, int tables)
|
|
{
|
|
stb_idict *d[50000];
|
|
int i,q,total=0,v;
|
|
for (q=0; q < tables; ++q)
|
|
d[q] = stb_idict_new_size(8);
|
|
for (q=0; q < tables; ++q)
|
|
for (i=0; i < a; ++i)
|
|
stb_idict_add(d[q], i*step+q*771, i+1);
|
|
dont_measure();
|
|
for (i=start; i < end; ++i)
|
|
for (q=0; q < tables; ++q)
|
|
if (stb_idict_get_flag(d[q], i*step+q*771, &v))
|
|
total += v;
|
|
measure();
|
|
churn_inserts = i;
|
|
for (q=0; q < tables; ++q)
|
|
stb_idict_destroy(d[q]);
|
|
dont_measure();
|
|
return total;
|
|
}
|
|
#endif
|
|
|
|
int multisearch(int a, int start, int end, int step, int tables)
|
|
{
|
|
struct { int key,value; } *hash[50000];
|
|
int i,q,total=0;
|
|
for (q=0; q < tables; ++q)
|
|
hash[q] = NULL;
|
|
for (q=0; q < tables; ++q)
|
|
for (i=0; i < a; ++i)
|
|
hmput(hash[q], i*step+q*771, i+1);
|
|
dont_measure();
|
|
for (i=start; i < end; ++i)
|
|
for (q=0; q < tables; ++q)
|
|
total += hmget(hash[q], i*step+q*771);
|
|
measure();
|
|
churn_inserts = i;
|
|
for (q=0; q < tables; ++q)
|
|
hmfree(hash[q]);
|
|
dont_measure();
|
|
return total;
|
|
}
|
|
|
|
void churn_skip(unsigned int a, unsigned int b, int count)
|
|
{
|
|
struct { unsigned int key,value; } *map=NULL;
|
|
unsigned int i,j,n,k;
|
|
for (i=0; i < a; ++i)
|
|
hmput(map,i,i+1);
|
|
dont_measure();
|
|
for (n=0; n < count; ++n) {
|
|
for (j=a; j < b; ++j,++i) {
|
|
hmput(map,i,i+1);
|
|
}
|
|
assert(hmlen(map) == b);
|
|
for (j=a; j < b; ++j) {
|
|
k=i-j-1;
|
|
k = hmdel(map,k);
|
|
assert(k != 0);
|
|
}
|
|
assert(hmlen(map) == a);
|
|
}
|
|
measure();
|
|
churn_inserts = i;
|
|
churn_deletes = (b-a) * n;
|
|
hmfree(map);
|
|
dont_measure();
|
|
}
|
|
|
|
typedef struct { int n[8]; } str32;
|
|
void churn32(int a, int b, int count, int include_startup)
|
|
{
|
|
struct { str32 key; int value; } *map=NULL;
|
|
int i,j,n;
|
|
str32 key = { 0 };
|
|
for (i=0; i < a; ++i) {
|
|
key.n[0] = i;
|
|
hmput(map,key,i+1);
|
|
}
|
|
if (!include_startup)
|
|
dont_measure();
|
|
for (n=0; n < count; ++n) {
|
|
for (j=a; j < b; ++j,++i) {
|
|
key.n[0] = i;
|
|
hmput(map,key,i+1);
|
|
}
|
|
assert(hmlen(map) == b);
|
|
for (j=a; j < b; ++j) {
|
|
key.n[0] = i-j-1;
|
|
hmdel(map,key);
|
|
}
|
|
assert(hmlen(map) == a);
|
|
}
|
|
measure();
|
|
hmfree(map);
|
|
churn_inserts = i;
|
|
churn_deletes = (b-a) * n;
|
|
dont_measure();
|
|
}
|
|
|
|
typedef struct { int n[32]; } str256;
|
|
void churn256(int a, int b, int count, int include_startup)
|
|
{
|
|
struct { str256 key; int value; } *map=NULL;
|
|
int i,j,n;
|
|
str256 key = { 0 };
|
|
for (i=0; i < a; ++i) {
|
|
key.n[0] = i;
|
|
hmput(map,key,i+1);
|
|
}
|
|
if (!include_startup)
|
|
dont_measure();
|
|
for (n=0; n < count; ++n) {
|
|
for (j=a; j < b; ++j,++i) {
|
|
key.n[0] = i;
|
|
hmput(map,key,i+1);
|
|
}
|
|
assert(hmlen(map) == b);
|
|
for (j=a; j < b; ++j) {
|
|
key.n[0] = i-j-1;
|
|
hmdel(map,key);
|
|
}
|
|
assert(hmlen(map) == a);
|
|
}
|
|
measure();
|
|
hmfree(map);
|
|
churn_inserts = i;
|
|
churn_deletes = (b-a) * n;
|
|
dont_measure();
|
|
}
|
|
|
|
void churn8(int a, int b, int count, int include_startup)
|
|
{
|
|
struct { size_t key,value; } *map=NULL;
|
|
int i,j,n,k;
|
|
for (i=0; i < a; ++i)
|
|
hmput(map,i,i+1);
|
|
if (!include_startup)
|
|
dont_measure();
|
|
for (n=0; n < count; ++n) {
|
|
for (j=a; j < b; ++j,++i) {
|
|
hmput(map,i,i+1);
|
|
}
|
|
assert(hmlen(map) == b);
|
|
for (j=a; j < b; ++j) {
|
|
k=i-j-1;
|
|
k = hmdel(map,k);
|
|
assert(k != 0);
|
|
}
|
|
assert(hmlen(map) == a);
|
|
}
|
|
measure();
|
|
hmfree(map);
|
|
churn_inserts = i;
|
|
churn_deletes = (b-a) * n;
|
|
dont_measure();
|
|
}
|
|
|
|
void multichurn4(int a, int b, int count, int include_startup, int tables)
|
|
{
|
|
struct { int key,value; } *map[50000];
|
|
int i,j,n,k,q;
|
|
for (q=0; q < tables; ++q)
|
|
map[q] = NULL;
|
|
dont_measure();
|
|
|
|
for (i=0; i < a; ++i)
|
|
for (q=0; q < tables; ++q)
|
|
hmput(map[q],i,i+1);
|
|
if (!include_startup)
|
|
dont_measure();
|
|
for (n=0; n < count; ++n) {
|
|
for (j=a; j < b; ++j,++i) {
|
|
for (q=0; q < tables; ++q)
|
|
hmput(map[q],i,i+1);
|
|
}
|
|
assert(hmlen(map[0]) == b);
|
|
for (j=a; j < b; ++j) {
|
|
k=i-j-1;
|
|
for (q=0; q < tables; ++q)
|
|
k = hmdel(map[q],k);
|
|
assert(k != 0);
|
|
}
|
|
assert(hmlen(map[0]) == a);
|
|
}
|
|
measure();
|
|
for (q=0; q < tables; ++q)
|
|
hmfree(map[q]);
|
|
churn_inserts = i * tables;
|
|
churn_deletes = (b-a) * n * tables;
|
|
dont_measure();
|
|
}
|
|
|
|
struct {
|
|
unsigned __int64 start;
|
|
unsigned __int64 end;
|
|
int table_size;
|
|
} mstats[32][4000];
|
|
|
|
const int first_step = 64;
|
|
const int last_step = 384-48; // 32M
|
|
|
|
void measure_build4(int step_log2)
|
|
{
|
|
double length;
|
|
int i,j,k=0;
|
|
int step = 1 << step_log2;
|
|
unsigned __int64 t0,t1;
|
|
struct { int key,value; } *map=NULL;
|
|
double rdtsc_scale;
|
|
begin();
|
|
t0 = __rdtsc();
|
|
|
|
mstats[0][0].start = __rdtsc();
|
|
for (i=0; i < 256; ++i) {
|
|
hmput(map,k,k+1);
|
|
k += step;
|
|
}
|
|
mstats[0][first_step-1].end = __rdtsc();
|
|
mstats[0][first_step-1].table_size = k >> step_log2;
|
|
for (j=first_step; j < last_step; ++j) {
|
|
for (i=0; i < (1<<(j>>4)); ++i) {
|
|
hmput(map, k,k+1);
|
|
k += step;
|
|
}
|
|
mstats[0][j].end = __rdtsc();
|
|
mstats[0][j].table_size = k >> step_log2;
|
|
}
|
|
t1 = __rdtsc();
|
|
measure();
|
|
hmfree(map);
|
|
length = end();
|
|
rdtsc_scale = length / (t1-t0) * 1000;
|
|
|
|
for (j=1; j < last_step; ++j)
|
|
mstats[0][j].start = mstats[0][0].start;
|
|
for (j=first_step-1; j < last_step; ++j) {
|
|
printf("%12.4f,%12d,%12d,0,0,0\n", (mstats[0][j].end - mstats[0][j].start) * rdtsc_scale, mstats[0][j].table_size, mstats[0][j].table_size);
|
|
}
|
|
}
|
|
|
|
#ifdef STBDS_FLUSH_CACHE
|
|
static int cache_index;
|
|
char dummy[8][STBDS_FLUSH_CACHE];
|
|
|
|
int flush_cache(void)
|
|
{
|
|
memmove(dummy[cache_index],dummy[cache_index]+1, sizeof(dummy[cache_index])-1);
|
|
cache_index = (cache_index+1)%8;
|
|
return dummy[cache_index][0];
|
|
}
|
|
#else
|
|
int flush_cache(void) { return 0; }
|
|
#endif
|
|
|
|
int measure_average_lookup4(int step_log2)
|
|
{
|
|
int total;
|
|
double length;
|
|
int i,j,k=0,q;
|
|
int step = 1 << step_log2;
|
|
unsigned __int64 t0,t1;
|
|
struct { int key,value; } *map=NULL;
|
|
double rdtsc_scale;
|
|
begin();
|
|
t0 = __rdtsc();
|
|
|
|
for (i=0; i < 128; ++i) {
|
|
hmput(map,k,k+1);
|
|
k += step;
|
|
}
|
|
for (j=first_step; j <= last_step; ++j) {
|
|
total += flush_cache();
|
|
mstats[0][j].start = __rdtsc();
|
|
for (q=i=0; i < 50000; ++i) {
|
|
total += hmget(map, q); // hit
|
|
if (++q == k) q = 0;
|
|
}
|
|
mstats[0][j].end = __rdtsc();
|
|
mstats[0][j].table_size = k;
|
|
total += flush_cache();
|
|
mstats[1][j].start = __rdtsc();
|
|
for (i=0; i < 50000; ++i) {
|
|
total += hmget(map, i+k); // miss
|
|
}
|
|
mstats[1][j].end = __rdtsc();
|
|
mstats[1][j].table_size = k;
|
|
|
|
// expand table
|
|
for (i=0; i < (1<<(j>>4)); ++i) {
|
|
hmput(map, k,k+1);
|
|
k += step;
|
|
}
|
|
}
|
|
|
|
t1 = __rdtsc();
|
|
measure();
|
|
hmfree(map);
|
|
length = end();
|
|
rdtsc_scale = length / (t1-t0) * 1000;
|
|
|
|
for (j=first_step; j <= last_step; ++j) {
|
|
// time,table_size,numins,numhit,nummiss,numperflush
|
|
printf("%12.4f,%12d,0,50000,0,0\n", (mstats[0][j].end - mstats[0][j].start) * rdtsc_scale, mstats[0][j].table_size);
|
|
}
|
|
for (j=first_step; j <= last_step; ++j) {
|
|
printf("%12.4f,%12d,0,0,50000,0\n", (mstats[1][j].end - mstats[1][j].start) * rdtsc_scale, mstats[1][j].table_size);
|
|
}
|
|
return total;
|
|
}
|
|
|
|
int measure_worst_lookup4_a(int step_log2)
|
|
{
|
|
int total;
|
|
double length;
|
|
int i,j,k=0,q,worst_q,n,z,attempts;
|
|
int step = 1 << step_log2;
|
|
unsigned __int64 t0,t1;
|
|
unsigned __int64 m0,m1,worst;
|
|
struct { int key,value; } *map=NULL;
|
|
double rdtsc_scale;
|
|
begin();
|
|
t0 = __rdtsc();
|
|
|
|
memset(mstats, 0, sizeof(mstats));
|
|
for (j=first_step; j <= last_step; ++j)
|
|
mstats[0][j].end = mstats[1][j].end = ~(unsigned __int64) 0;
|
|
|
|
for(attempts=0; attempts < 2; ++attempts) {
|
|
k = 0;
|
|
stbds_rand_seed(0); // force us to get the same table every time
|
|
for (i=0; i < 128; ++i) {
|
|
hmput(map,k,k+1);
|
|
k += step;
|
|
}
|
|
for (j=first_step; j <= last_step; ++j) {
|
|
unsigned __int64 times[32];
|
|
|
|
// find the worst hit time
|
|
for (z=0; z < 2; ++z) { // try the bisectioning measurement 4 times
|
|
worst = 0;
|
|
for (n=0; n < 10; ++n) { // test 400 keys total
|
|
// find the worst time to hit 20 keys
|
|
q=0;
|
|
worst_q = 0;
|
|
total += flush_cache();
|
|
m0 = __rdtsc();
|
|
for (i=0; i < 20; ++i) {
|
|
total += hmget(map, q); // hit
|
|
if (++q == k) q = 0;
|
|
}
|
|
m1 = __rdtsc();
|
|
// for each n, check if this is the worst lookup we've seen
|
|
if (m1 - m0 > worst) {
|
|
worst = m1-m0;
|
|
worst_q = q - i;
|
|
if (worst_q < 0) q += k;
|
|
}
|
|
}
|
|
// after 400 keys, take the worst 20 keys, and try each one
|
|
worst = 0;
|
|
q = worst_q;
|
|
for (i=0; i < 20; ++i) {
|
|
total += flush_cache();
|
|
m0 = __rdtsc();
|
|
total += hmget(map, q); // hit
|
|
m1 = __rdtsc();
|
|
if (m1 - m0 > worst)
|
|
worst = m1-m0;
|
|
if (++q == k) q = 0;
|
|
}
|
|
times[z] = worst;
|
|
}
|
|
// find the worst time in the bunch
|
|
worst = 0;
|
|
for (i=0; i < z; ++i)
|
|
if (times[i] > worst)
|
|
worst = times[i];
|
|
// take the best of 'attempts', to discard outliers
|
|
if (worst < mstats[0][j].end)
|
|
mstats[0][j].end = worst;
|
|
mstats[0][j].start = 0;
|
|
mstats[0][j].table_size = k >> step_log2;
|
|
|
|
// find the worst miss time
|
|
for (z=0; z < 8; ++z) { // try the bisectioning measurement 8 times
|
|
worst = 0;
|
|
for (n=0; n < 20; ++n) { // test 400 keys total
|
|
// find the worst time to hit 20 keys
|
|
q=k;
|
|
worst_q = 0;
|
|
total += flush_cache();
|
|
m0 = __rdtsc();
|
|
for (i=0; i < 20; ++i) {
|
|
total += hmget(map, q); // hit
|
|
}
|
|
m1 = __rdtsc();
|
|
// for each n, check if this is the worst lookup we've seen
|
|
if (m1 - m0 > worst) {
|
|
worst = m1-m0;
|
|
worst_q = q - i;
|
|
}
|
|
}
|
|
// after 400 keys, take the worst 20 keys, and try each one
|
|
worst = 0;
|
|
q = worst_q;
|
|
for (i=0; i < 20; ++i) {
|
|
total += flush_cache();
|
|
m0 = __rdtsc();
|
|
total += hmget(map, q); // hit
|
|
m1 = __rdtsc();
|
|
if (m1 - m0 > worst)
|
|
worst = m1-m0;
|
|
}
|
|
times[z] = worst;
|
|
}
|
|
// find the worst time in the bunch
|
|
worst = 0;
|
|
for (i=0; i < z; ++i)
|
|
if (times[i] > worst)
|
|
worst = times[i];
|
|
if (worst < mstats[1][j].end)
|
|
mstats[1][j].end = worst;
|
|
mstats[1][j].start = 0;
|
|
mstats[1][j].table_size = k >> step_log2;
|
|
|
|
// expand table
|
|
for (i=0; i < (1<<(j>>4)); ++i) {
|
|
hmput(map, k,k+1);
|
|
k += step;
|
|
}
|
|
}
|
|
hmfree(map);
|
|
}
|
|
|
|
t1 = __rdtsc();
|
|
measure();
|
|
length = end();
|
|
rdtsc_scale = length / (t1-t0) * 1000;
|
|
|
|
for (j=first_step; j <= last_step; ++j) {
|
|
printf("%12.4f,%12d,0,1,0,1\n", (mstats[0][j].end - mstats[0][j].start) * rdtsc_scale, mstats[0][j].table_size);
|
|
}
|
|
for (j=first_step; j <= last_step; ++j) {
|
|
printf("%12.4f,%12d,0,0,1,1\n", (mstats[1][j].end - mstats[1][j].start) * rdtsc_scale, mstats[1][j].table_size);
|
|
}
|
|
return total;
|
|
}
|
|
|
|
int measure_worst_lookup4_b(int step_log2)
|
|
{
|
|
int total;
|
|
double length;
|
|
int i,j,k=0,q,worst_q,n,z,attempts;
|
|
int step = 1 << step_log2;
|
|
unsigned __int64 t0,t1;
|
|
unsigned __int64 m0,m1,worst;
|
|
struct { int key,value; } *map=NULL;
|
|
double rdtsc_scale;
|
|
begin();
|
|
t0 = __rdtsc();
|
|
|
|
memset(mstats, 0, sizeof(mstats));
|
|
for (j=first_step; j <= last_step; ++j)
|
|
mstats[0][j].end = mstats[1][j].end = ~(unsigned __int64) 0;
|
|
|
|
k = 0;
|
|
stbds_rand_seed(0); // force us to get the same table every time
|
|
for (i=0; i < 128; ++i) {
|
|
hmput(map,k,k+1);
|
|
k += step;
|
|
}
|
|
for (j=first_step; j <= last_step; ++j) {
|
|
unsigned __int64 times[32];
|
|
|
|
// find the worst hit time
|
|
for (z=0; z < 8; ++z) { // try this 8 times
|
|
worst = 0;
|
|
q=0;
|
|
for (i=0; i < 5000; ++i) {
|
|
total += hmget(map, q);
|
|
m0 = __rdtsc();
|
|
total += hmget(map, q);
|
|
m1 = __rdtsc();
|
|
if (m1 - m0 > worst) {
|
|
worst = m1-m0;
|
|
worst_q = q - i;
|
|
}
|
|
if (++q == k) q = 0;
|
|
}
|
|
// now retry with the worst one, but find the shortest time for it
|
|
worst = ~(unsigned __int64) 0;
|
|
for (i=0; i < 4; ++i) {
|
|
total += flush_cache();
|
|
m0 = __rdtsc();
|
|
total += hmget(map,worst_q);
|
|
m1 = __rdtsc();
|
|
if (m1-m0 < worst)
|
|
worst = m1-m0;
|
|
}
|
|
times[z] = worst;
|
|
}
|
|
|
|
// find the worst of those
|
|
worst = 0;
|
|
for (i=0; i < z; ++i)
|
|
if (times[i] > worst)
|
|
worst = times[i];
|
|
mstats[0][j].start = 0;
|
|
mstats[0][j].end = worst;
|
|
mstats[0][j].table_size = k;
|
|
|
|
// find the worst miss time
|
|
for (z=0; z < 8; ++z) { // try this 8 times
|
|
worst = 0;
|
|
q=k;
|
|
for (i=0; i < 5000; ++i) {
|
|
total += hmget(map, q);
|
|
m0 = __rdtsc();
|
|
total += hmget(map, q);
|
|
m1 = __rdtsc();
|
|
if (m1 - m0 > worst) {
|
|
worst = m1-m0;
|
|
worst_q = q - i;
|
|
}
|
|
//printf("%6llu ", m1-m0);
|
|
}
|
|
// now retry with the worst one, but find the shortest time for it
|
|
worst = ~(unsigned __int64) 0;
|
|
for (i=0; i < 4; ++i) {
|
|
total += flush_cache();
|
|
m0 = __rdtsc();
|
|
total += hmget(map,worst_q);
|
|
m1 = __rdtsc();
|
|
if (m1-m0 < worst)
|
|
worst = m1-m0;
|
|
}
|
|
times[z] = worst;
|
|
}
|
|
|
|
// find the worst of those
|
|
worst = 0;
|
|
for (i=0; i < z; ++i)
|
|
if (times[i] > worst)
|
|
worst = times[i];
|
|
mstats[1][j].start = 0;
|
|
mstats[1][j].end = worst;
|
|
mstats[1][j].table_size = k;
|
|
|
|
// expand table
|
|
for (i=0; i < (1<<(j>>4)); ++i) {
|
|
hmput(map, k,k+1);
|
|
k += step;
|
|
}
|
|
}
|
|
hmfree(map);
|
|
|
|
t1 = __rdtsc();
|
|
measure();
|
|
length = end();
|
|
rdtsc_scale = length / (t1-t0) * 1000;
|
|
|
|
for (j=first_step+1; j <= last_step; ++j) {
|
|
printf("%12.4f,%12d,0,1,0,1\n", (mstats[0][j].end - mstats[0][j].start) * rdtsc_scale, mstats[0][j].table_size);
|
|
}
|
|
for (j=first_step+1; j <= last_step; ++j) {
|
|
printf("%12.4f,%12d,0,0,1,1\n", (mstats[1][j].end - mstats[1][j].start) * rdtsc_scale, mstats[1][j].table_size);
|
|
}
|
|
return total;
|
|
}
|
|
|
|
int measure_uncached_lookup4(int step_log2)
|
|
{
|
|
int total;
|
|
double length;
|
|
int i,j,k=0,q;
|
|
int step = 1 << step_log2;
|
|
unsigned __int64 t0,t1;
|
|
struct { int key,value; } *map=NULL;
|
|
double rdtsc_scale;
|
|
begin();
|
|
t0 = __rdtsc();
|
|
|
|
map = NULL;
|
|
for (i=0; i < 128; ++i) {
|
|
hmput(map,k,k+1);
|
|
k += step;
|
|
}
|
|
for (j=first_step; j <= last_step; ++j) {
|
|
mstats[0][j].start = __rdtsc();
|
|
mstats[0][j].end = 0;
|
|
for (q=i=0; i < 512; ++i) {
|
|
if ((i & 3) == 0) {
|
|
mstats[0][j].end += __rdtsc();
|
|
total += flush_cache();
|
|
mstats[0][j].start += __rdtsc();
|
|
}
|
|
total += hmget(map, q); // hit
|
|
if (++q == k) q = 0;
|
|
}
|
|
mstats[0][j].end += __rdtsc();
|
|
mstats[0][j].table_size = k;
|
|
total += flush_cache();
|
|
mstats[1][j].end = 0;
|
|
mstats[1][j].start = __rdtsc();
|
|
for (i=0; i < 512; ++i) {
|
|
if ((i & 3) == 0) {
|
|
mstats[1][j].end += __rdtsc();
|
|
total += flush_cache();
|
|
mstats[1][j].start += __rdtsc();
|
|
}
|
|
total += hmget(map, i+k); // miss
|
|
}
|
|
mstats[1][j].end += __rdtsc();
|
|
mstats[1][j].table_size = k;
|
|
|
|
// expand table
|
|
for (i=0; i < (1<<(j>>4)); ++i) {
|
|
hmput(map, k,k+1);
|
|
k += step;
|
|
}
|
|
}
|
|
hmfree(map);
|
|
|
|
t1 = __rdtsc();
|
|
measure();
|
|
length = end();
|
|
rdtsc_scale = length / (t1-t0) * 1000;
|
|
|
|
for (j=first_step; j <= last_step; ++j) {
|
|
printf("%12.4f,%12d,0,512,0,4\n", (mstats[0][j].end - mstats[0][j].start) * rdtsc_scale, mstats[0][j].table_size);
|
|
}
|
|
for (j=first_step; j <= last_step; ++j) {
|
|
printf("%12.4f,%12d,0,0,512,4\n", (mstats[1][j].end - mstats[1][j].start) * rdtsc_scale, mstats[1][j].table_size);
|
|
}
|
|
return total;
|
|
}
|
|
|
|
|
|
|
|
|
|
int main(int arg, char **argv)
|
|
{
|
|
int n,s,w;
|
|
double worst = 0;
|
|
|
|
printf("# size_t=%d,", (int) sizeof(size_t));
|
|
|
|
// number of cache-lines
|
|
#ifdef STBDS_SMALL_BUCKET
|
|
printf("cacheline=%d,", 1);
|
|
#else
|
|
printf("cacheline=%d,", sizeof(size_t)==8 ? 2 : 1);
|
|
#endif
|
|
#ifdef STBDS_FLUSH_CACHE
|
|
printf("%d,", (int) stbds_log2(STBDS_FLUSH_CACHE));
|
|
#else
|
|
printf("0,");
|
|
#endif
|
|
#ifdef STBDS_BUCKET_START // don't bother offseting differently within bucket for different hash values
|
|
printf("STBDS_BUCKET_START,");
|
|
#else
|
|
printf(",");
|
|
#endif
|
|
#ifdef STBDS_SIPHASH_2_4
|
|
printf("STBDS_SIPHASH_2_4,");
|
|
#else
|
|
printf(",");
|
|
#endif
|
|
printf("\n");
|
|
|
|
measure_worst_lookup4_b(0);
|
|
//measure_worst_lookup4_a(0);
|
|
measure_average_lookup4(0);
|
|
measure_uncached_lookup4(0);
|
|
measure_build4(0);
|
|
return 0;
|
|
|
|
#if 0
|
|
begin(); for (n=0; n < 2000; ++n) { build_stb(2000,0,0,1); } end(); printf(" // %7.2fms : 2,000 inserts creating 2K table\n", timer);
|
|
begin(); for (n=0; n < 500; ++n) { build_stb(20000,0,0,1); } end(); printf(" // %7.2fms : 20,000 inserts creating 20K table\n", timer);
|
|
begin(); for (n=0; n < 100; ++n) { build_stb(200000,0,0,1); } end(); printf(" // %7.2fms : 200,000 inserts creating 200K table\n", timer);
|
|
begin(); for (n=0; n < 10; ++n) { build_stb(2000000,0,0,1); } end(); printf(" // %7.2fms : 2,000,000 inserts creating 2M table\n", timer);
|
|
begin(); for (n=0; n < 5; ++n) { build_stb(20000000,0,0,1); } end(); printf(" // %7.2fms : 20,000,000 inserts creating 20M table\n", timer);
|
|
#endif
|
|
|
|
#if 0
|
|
begin(); for (n=0; n < 2000; ++n) { churn32(2000,0,0,1); } end(); printf(" // %7.2fms : 2,000 inserts creating 2K table w/ 32-byte key\n", timer);
|
|
begin(); for (n=0; n < 500; ++n) { churn32(20000,0,0,1); } end(); printf(" // %7.2fms : 20,000 inserts creating 20K table w/ 32-byte key\n", timer);
|
|
begin(); for (n=0; n < 100; ++n) { churn32(200000,0,0,1); } end(); printf(" // %7.2fms : 200,000 inserts creating 200K table w/ 32-byte key\n", timer);
|
|
begin(); for (n=0; n < 10; ++n) { churn32(2000000,0,0,1); } end(); printf(" // %7.2fms : 2,000,000 inserts creating 2M table w/ 32-byte key\n", timer);
|
|
begin(); for (n=0; n < 5; ++n) { churn32(20000000,0,0,1); } end(); printf(" // %7.2fms : 20,000,000 inserts creating 20M table w/ 32-byte key\n", timer);
|
|
|
|
begin(); for (n=0; n < 2000; ++n) { churn256(2000,0,0,1); } end(); printf(" // %7.2fms : 2,000 inserts creating 2K table w/ 256-byte key\n", timer);
|
|
begin(); for (n=0; n < 500; ++n) { churn256(20000,0,0,1); } end(); printf(" // %7.2fms : 20,000 inserts creating 20K table w/ 256-byte key\n", timer);
|
|
begin(); for (n=0; n < 100; ++n) { churn256(200000,0,0,1); } end(); printf(" // %7.2fms : 200,000 inserts creating 200K table w/ 256-byte key\n", timer);
|
|
begin(); for (n=0; n < 10; ++n) { churn256(2000000,0,0,1); } end(); printf(" // %7.2fms : 2,000,000 inserts creating 2M table w/ 256-byte key\n", timer);
|
|
begin(); for (n=0; n < 5; ++n) { churn256(20000000,0,0,1); } end(); printf(" // %7.2fms : 20,000,000 inserts creating 20M table w/ 256-byte key\n", timer);
|
|
#endif
|
|
|
|
begin(); for (n=0; n < 20; ++n) { multisearch_stb(2000,0,2000,1,1000); } end(); printf(" // %7.2fms : 2,000,000 hits on 1,000 2K table w/ 4-byte key\n", timer);
|
|
begin(); for (n=0; n < 10; ++n) { multisearch_stb(20000,0,2000,1,1000); } end(); printf(" // %7.2fms : 2,000,000 hits on 1,000 20K table w/ 4-byte key\n", timer);
|
|
begin(); for (n=0; n < 6; ++n) { multisearch_stb(200000,0,2000,1,1000); } end(); printf(" // %7.2fms : 2,000,000 hits on 1,000 200K table w/ 4-byte key\n", timer);
|
|
begin(); for (n=0; n < 2; ++n) { multisearch_stb(2000000,0,20000,1,100); } end(); printf(" // %7.2fms : 2,000,000 hits on 100 2M table w/ 4-byte key\n", timer);
|
|
|
|
begin(); for (n=0; n < 20; ++n) { multisearch (2000,0,2000,1,1000); } end(); printf(" // %7.2fms : 2,000,000 hits on 1,000 2K table w/ 4-byte key\n", timer);
|
|
begin(); for (n=0; n < 10; ++n) { multisearch (20000,0,2000,1,1000); } end(); printf(" // %7.2fms : 2,000,000 hits on 1,000 20K table w/ 4-byte key\n", timer);
|
|
begin(); for (n=0; n < 6; ++n) { multisearch (200000,0,2000,1,1000); } end(); printf(" // %7.2fms : 2,000,000 hits on 1,000 200K table w/ 4-byte key\n", timer);
|
|
begin(); for (n=0; n < 2; ++n) { multisearch (2000000,0,20000,1,100); } end(); printf(" // %7.2fms : 2,000,000 hits on 100 2M table w/ 4-byte key\n", timer);
|
|
|
|
|
|
#if 1
|
|
begin(); for (n=0; n < 2; ++n) { multibuild_stb(2000,0,0,1,10000); } end(); printf(" // %7.2fms : 20,000,000 inserts creating 10,000 2K table w/ 4-byte key\n", timer);
|
|
begin(); for (n=0; n < 2; ++n) { multibuild_stb(20000,0,0,1,1000); } end(); printf(" // %7.2fms : 20,000,000 inserts creating 1,000 20K table w/ 4-byte key\n", timer);
|
|
begin(); for (n=0; n < 2; ++n) { multibuild_stb(200000,0,0,1,100); } end(); printf(" // %7.2fms : 20,000,000 inserts creating 100 200K table w/ 4-byte key\n", timer);
|
|
begin(); for (n=0; n < 2; ++n) { multibuild_stb(2000000,0,0,1,10); } end(); printf(" // %7.2fms : 20,000,000 inserts creating 10 2M table w/ 4-byte key\n", timer);
|
|
|
|
begin(); for (n=0; n < 2; ++n) { multichurn4(2000,0,0,1,10000); } end(); printf(" // %7.2fms : 20,000,000 inserts creating 10,000 2K table w/ 4-byte key\n", timer);
|
|
begin(); for (n=0; n < 2; ++n) { multichurn4(20000,0,0,1,1000); } end(); printf(" // %7.2fms : 20,000,000 inserts creating 1,000 20K table w/ 4-byte key\n", timer);
|
|
begin(); for (n=0; n < 2; ++n) { multichurn4(200000,0,0,1,100); } end(); printf(" // %7.2fms : 20,000,000 inserts creating 100 200K table w/ 4-byte key\n", timer);
|
|
begin(); for (n=0; n < 2; ++n) { multichurn4(2000000,0,0,1,10); } end(); printf(" // %7.2fms : 20,000,000 inserts creating 10 2M table w/ 4-byte key\n", timer);
|
|
#endif
|
|
|
|
begin(); for (n=0; n < 2000; ++n) { build(2000,0,0,1); } end(); printf(" // %7.2fms : 2,000 inserts creating 2K table w/ 4-byte key\n", timer);
|
|
begin(); for (n=0; n < 500; ++n) { build(20000,0,0,1); } end(); printf(" // %7.2fms : 20,000 inserts creating 20K table w/ 4-byte key\n", timer);
|
|
begin(); for (n=0; n < 100; ++n) { build(200000,0,0,1); } end(); printf(" // %7.2fms : 200,000 inserts creating 200K table w/ 4-byte key\n", timer);
|
|
begin(); for (n=0; n < 10; ++n) { build(2000000,0,0,1); } end(); printf(" // %7.2fms : 2,000,000 inserts creating 2M table w/ 4-byte key\n", timer);
|
|
begin(); for (n=0; n < 5; ++n) { build(20000000,0,0,1); } end(); printf(" // %7.2fms : 20,000,000 inserts creating 20M table w/ 4-byte key\n", timer);
|
|
|
|
begin(); for (n=0; n < 2000; ++n) { churn8(2000,0,0,1); } end(); printf(" // %7.2fms : 2,000 inserts creating 2K table w/ 8-byte key\n", timer);
|
|
begin(); for (n=0; n < 500; ++n) { churn8(20000,0,0,1); } end(); printf(" // %7.2fms : 20,000 inserts creating 20K table w/ 8-byte key\n", timer);
|
|
begin(); for (n=0; n < 100; ++n) { churn8(200000,0,0,1); } end(); printf(" // %7.2fms : 200,000 inserts creating 200K table w/ 8-byte key\n", timer);
|
|
begin(); for (n=0; n < 10; ++n) { churn8(2000000,0,0,1); } end(); printf(" // %7.2fms : 2,000,000 inserts creating 2M table w/ 8-byte key\n", timer);
|
|
begin(); for (n=0; n < 5; ++n) { churn8(20000000,0,0,1); } end(); printf(" // %7.2fms : 20,000,000 inserts creating 20M table w/ 8-byte key\n", timer);
|
|
|
|
begin(); for (n=0; n < 60; ++n) { churn_skip(2000,2100,5000); } end(); printf(" // %7.2fms : 500,000 inserts & deletes in 2K table\n", timer);
|
|
begin(); for (n=0; n < 30; ++n) { churn_skip(20000,21000,500); } end(); printf(" // %7.2fms : 500,000 inserts & deletes in 20K table\n", timer);
|
|
begin(); for (n=0; n < 15; ++n) { churn_skip(200000,201000,500); } end(); printf(" // %7.2fms : 500,000 inserts & deletes in 200K table\n", timer);
|
|
begin(); for (n=0; n < 8; ++n) { churn_skip(2000000,2001000,500); } end(); printf(" // %7.2fms : 500,000 inserts & deletes in 2M table\n", timer);
|
|
begin(); for (n=0; n < 5; ++n) { churn_skip(20000000,20001000,500); } end(); printf(" // %7.2fms : 500,000 inserts & deletes in 20M table\n", timer);
|
|
begin(); for (n=0; n < 1; ++n) { churn_skip(200000000u,200001000u,500); } end(); printf(" // %7.2fms : 500,000 inserts & deletes in 200M table\n", timer);
|
|
// even though the above measures a roughly fixed amount of work, we still have to build the table n times, hence the fewer measurements each time
|
|
|
|
begin(); for (n=0; n < 60; ++n) { churn_skip(1000,3000,250); } end(); printf(" // %7.2fms : 500,000 inserts & deletes in 2K table\n", timer);
|
|
begin(); for (n=0; n < 15; ++n) { churn_skip(10000,30000,25); } end(); printf(" // %7.2fms : 500,000 inserts & deletes in 20K table\n", timer);
|
|
begin(); for (n=0; n < 7; ++n) { churn_skip(100000,300000,10); } end(); printf(" // %7.2fms : 2,000,000 inserts & deletes in 200K table\n", timer);
|
|
begin(); for (n=0; n < 2; ++n) { churn_skip(1000000,3000000,10); } end(); printf(" // %7.2fms : 20,000,000 inserts & deletes in 2M table\n", timer);
|
|
|
|
// search for bad intervals.. in practice this just seems to measure execution variance
|
|
for (s = 2; s < 64; ++s) {
|
|
begin(); for (n=0; n < 50; ++n) { build(200000,0,0,s); } end();
|
|
if (timer > worst) {
|
|
worst = timer;
|
|
w = s;
|
|
}
|
|
}
|
|
for (; s <= 1024; s *= 2) {
|
|
begin(); for (n=0; n < 50; ++n) { build(200000,0,0,s); } end();
|
|
if (timer > worst) {
|
|
worst = timer;
|
|
w = s;
|
|
}
|
|
}
|
|
printf(" // %7.2fms(%d) : Worst time from inserting 200,000 items with spacing %d.\n", worst, w, w);
|
|
|
|
return 0;
|
|
}
|
|
#endif
|