/**
* Do a given number of zipfian reads on a cog.
*
* @param cog - the given cog
* @param alpha - zipfian rate of decay
* @param number - number of reads to do on a cog
* @param range - the key range for reads
* @return the resulting BTree
*/
struct cog *zipfianReads(struct cog *cog, double alpha, long number, long range) {
for (long i = 0; i < number; i++) {
long a = zipf(alpha, range);
long b = zipf(alpha, range);
long low = a <= b ? a : b;
long high = a > b ? a : b;
cog = crack(cog, low, high);
}
return cog;
}
int main(int argc, char *argv[])
{
int k = 100000;
int exp = 1;
// double c0 = 2, c1 = 3, c2 = 1;
// double c0 = 1.3, c1 = 8, c2 = 1.5;
std::CommandLine cmd;
cmd.AddValue ("exp", "", exp);
cmd.Parse (argc, argv);
// Set sample size for test
std::vector<int> testN;
long n = 1000;
for ( int i = 0; i < 15; i++ )
{
n *= 2;
}
testN.push_back( n );
// Set distribution for test
std::vector<double> p;
switch(exp)
{
case 0: p = uniform(k); break;
case 1: p = zipf(k); break;
case 2: p = zipfd5(k); break;
case 3: p = mixgeozipf(k); break;
}
// Set estimator
Entropy entropy( k );
entropy.setDegree( 18 );
entropy.setInterval( 40 );
entropy.setThreshold( 18 );
printf("Alphabet size=%d.\n", entropy.getAlphabetSize());
printf("Polynoimal degree=%d.\n", entropy.getDegree());
printf("Approximation interval=[0,%.2f/n].\n", entropy.getInterval());
printf("Plug-in threshold=%d.\n",(int)floor(entropy.getThreshold())+1);
printf("Unit: bits\n");
// TEST_fixed_P(p, entropy, testN);
const int trials = 50;
TEST_fixed_P_RMSE(p, entropy, testN, trials);
return 0;
}
//Retorna se um evento acontece ou não, pela probabilidade Zipf
//u = quantidade de universidades, i = Ui, p = tamanho do vetor a ser preenchido com as probabilidades
void DistribZipf(int vet[], int u, int i, int p) {
//Aloca vetor de u posiçoes
double rand_prob;
int size = i-p+1;
int j;
double *probs = (double *) malloc(size*sizeof(double)); //cria vetor para cópia dos valores não ordenados
if (probs == NULL) exit(1);
//Prenche vetor com a probabilidade de cada i
int k;
for (k=0;k<u;k++){
probs[k]=zipf(u,i);
}
//Gera uma probabilidade aleatória entre 0 e 1
//Ordena vetor de double
//verifica em área a probabilidade gerada se encaixa
//Gera p números
for (j=1;j<p;j++){
rand_prob = (double)rand()/(double)RAND_MAX;
for (k=1;k<u;k++){
if (rand_prob < probs[k]) vet[k] = 1;
else vet[k] = 0;
}
}
//Verifica se há repetição
//Vetor com as probabilidades
free(probs);
////-------------------///
if (rand_prob >= probs[k]) return;
return ;
}
BaseQuery * YCSBQueryGenerator::gen_requests_zipf(uint64_t home_partition_id, Workload * h_wl) {
YCSBQuery * query = (YCSBQuery*) mem_allocator.alloc(sizeof(YCSBQuery));
new(query) YCSBQuery();
query->requests.init(g_req_per_query);
uint64_t access_cnt = 0;
set<uint64_t> all_keys;
set<uint64_t> partitions_accessed;
uint64_t table_size = g_synth_table_size / g_part_cnt;
double r_twr = (double)(mrand->next() % 10000) / 10000;
int rid = 0;
for (UInt32 i = 0; i < g_req_per_query; i ++) {
double r = (double)(mrand->next() % 10000) / 10000;
uint64_t partition_id;
if ( FIRST_PART_LOCAL && rid == 0) {
partition_id = home_partition_id;;
} else {
partition_id = mrand->next() % g_part_cnt;
if(g_strict_ppt && g_part_per_txn <= g_part_cnt) {
while( (partitions_accessed.size() < g_part_per_txn && partitions_accessed.count(partition_id) > 0) ||
(partitions_accessed.size() == g_part_per_txn && partitions_accessed.count(partition_id) == 0)) {
partition_id = mrand->next() % g_part_cnt;
}
}
}
ycsb_request * req = (ycsb_request*) mem_allocator.alloc(sizeof(ycsb_request));
if (r_twr < g_txn_read_perc || r < g_tup_read_perc)
req->acctype = RD;
else
req->acctype = WR;
uint64_t row_id = zipf(table_size - 1, g_zipf_theta);;
assert(row_id < table_size);
uint64_t primary_key = row_id * g_part_cnt + partition_id;
assert(primary_key < g_synth_table_size);
req->key = primary_key;
req->value = mrand->next() % (1<<8);
// Make sure a single row is not accessed twice
if (all_keys.find(req->key) == all_keys.end()) {
all_keys.insert(req->key);
access_cnt ++;
} else {
// Need to have the full g_req_per_query amount
i--;
continue;
}
partitions_accessed.insert(partition_id);
rid ++;
query->requests.add(req);
}
assert(query->requests.size() == g_req_per_query);
// Sort the requests in key order.
if (g_key_order) {
for(uint64_t i = 0; i < query->requests.size(); i++) {
for(uint64_t j = query->requests.size() - 1; j > i ; j--) {
if(query->requests[j]->key < query->requests[j-1]->key) {
query->requests.swap(j,j-1);
}
}
}
//std::sort(query->requests.begin(),query->requests.end(),[](ycsb_request lhs, ycsb_request rhs) { return lhs.key < rhs.key;});
}
query->partitions.init(partitions_accessed.size());
for(auto it = partitions_accessed.begin(); it != partitions_accessed.end(); ++it) {
query->partitions.add(*it);
}
//query->print();
return query;
}
int spec_file() {
return FILE_ORDER[ zipf(FILE_ZIPF) ];
}
int spec_dir() {
return zipf(DIR_ZIPF);
}
int
main(int argc, char *argv[])
{
int i, which = 0, cnt = 100, trunc = 0, verbose = 0, weight = 1;
float alpha = 0, k = 0, value = 0, scale = 1.0, limit = 0;
float total = 0, median = -1, min = 0, max = 0;
for (i=1 ; i<argc ; i++) {
if (strcmp(argv[i], "-zipf") == 0) {
which = DIST_ZIPF;
if (++i >= argc || *argv[i] == '-') usage();
alpha = atof(argv[i]);
if (verbose) {
fprintf(stderr, "zipf(a=%0.2f)\n",
alpha);
}
}
else if (strcmp(argv[i], "-pareto") == 0) {
which = DIST_PARETO;
if (++i >= argc || *argv[i] == '-') usage();
alpha = atof(argv[i]);
if (++i >= argc || *argv[i] == '-') usage();
k = atof(argv[i]);
if (verbose) {
fprintf(stderr, "pareto(a=%0.2f, k=%0.2f)\n",
alpha, k);
}
}
else if (strcmp(argv[i], "-uniform") == 0) {
which = DIST_UNIFORM;
if (verbose) {
fprintf(stderr, "uniform()\n");
}
}
else if (strcmp(argv[i], "-cnt") == 0) {
if (++i >= argc || *argv[i] == '-') usage();
cnt = atoi(argv[i]);
if (verbose) {
fprintf(stderr, "cnt=%d\n", cnt);
}
}
else if (strcmp(argv[i], "-scale") == 0) {
if (++i >= argc || *argv[i] == '-') usage();
scale = atof(argv[i]);
if (verbose) {
fprintf(stderr, "scale=%0.2f\n", scale);
}
}
else if (strcmp(argv[i], "-limit") == 0) {
if (++i >= argc || *argv[i] == '-') usage();
limit = atof(argv[i]);
if (verbose) {
fprintf(stderr, "limit=%0.2f\n", limit);
}
}
else if (strcmp(argv[i], "-trunc") == 0) {
trunc = 1;
if (verbose) {
fprintf(stderr, "truncating values\n");
}
}
else if (strcmp(argv[i], "-weight") == 0) {
if (++i >= argc || *argv[i] == '-') usage();
weight = atoi(argv[i]);
if (verbose) {
fprintf(stderr, "weight=%d\n", weight);
}
}
else if (strcmp(argv[i], "-v") == 0) {
verbose = 1;
if (verbose) {
fprintf(stderr, "verbose output\n");
}
}
else {
fprintf(stderr, "unknown option \"%s\"\n", argv[i]);
usage();
}
}
for (i=1 ; i<=cnt ; i++) {
switch (which) {
case DIST_ZIPF:
value = zipf(i, alpha);
break;
case DIST_PARETO:
value = pareto(i, alpha, k);
break;
case DIST_UNIFORM:
value = uniform(i, cnt);
break;
default:
usage();
break;
}
/*
* scale value.
*/
value = value * scale;
/*
* optionally truncate values exceeding limit.
*/
if (limit && value > limit) {
value = limit;
}
/*
* optionally truncate values to integers.
*/
if (trunc) {
value = (int)value;
}
/*
* print value with proper decimal format.
*/
if (trunc) {
printf("%d\t%d\n", (int)value, weight);
} else {
printf("%0.2f\t%d\n", value, weight);
}
/*
* statistics: max, median, min, and total (for avg).
*/
if (i == 1) {
max = value;
}
if (i == cnt / 2) {
median = value;
}
if (i == cnt) {
min = value;
}
total += value;
}
if (verbose) {
fprintf(stderr, "total = %0.2f\n", total);
fprintf(stderr, "min = %0.2f (%0.2f%% of total)\n",
min, min / total * 100.0);
fprintf(stderr, "max = %0.2f (%0.2f%% of total)\n",
max, max / total * 100.0);
fprintf(stderr, "average = %0.2f (%0.2f%% of total)\n",
total/(float)cnt, (total/(float)cnt) / total * 100.0);
fprintf(stderr, "median = %0.2f (%0.2f%% of total)\n",
median, median / total * 100.0);
}
return 0;
}
//===== Main program ========================================================
void main(void)
{
FILE *fp; // File pointer to output file
char file_name[256]; // Output file name string
char temp_string[256]; // Temporary string variable
double alpha; // Alpha parameter
double n; // N parameter
int num_values; // Number of values
int zipf_rv; // Zipf random variable
int i; // Loop counter
// Output banner
printf("---------------------------------------- genzipf.c ----- \n");
printf("- Program to generate Zipf random variables - \n");
printf("-------------------------------------------------------- \n");
// Prompt for output filename and then create/open the file
printf("Output file name ===================================> ");
scanf("%s", file_name);
fp = fopen(file_name, "w");
if (fp == NULL)
{
printf("ERROR in creating output file (%s) \n", file_name);
exit(1);
}
// Prompt for random number seed and then use it
printf("Random number seed (greater than 0) ================> ");
scanf("%s", temp_string);
rand_val((int) atoi(temp_string));
// Prompt for alpha value
printf("Alpha value ========================================> ");
scanf("%s", temp_string);
alpha = atof(temp_string);
// Prompt for N value
printf("N value ============================================> ");
scanf("%s", temp_string);
n = atoi(temp_string);
// Prompt for number of values to generate
printf("Number of values to generate =======================> ");
scanf("%s", temp_string);
num_values = atoi(temp_string);
// Output "generating" message
printf("-------------------------------------------------------- \n");
printf("- Generating samples to file - \n");
printf("-------------------------------------------------------- \n");
// Generate and output zipf random variables
for (i=0; i<num_values; i++)
{
zipf_rv = zipf(alpha, n);
fprintf(fp, "%d \n", zipf_rv);
}
// Output "done" message and close the output file
printf("-------------------------------------------------------- \n");
printf("- Done! \n");
printf("-------------------------------------------------------- \n");
fclose(fp);
}
void ycsb_query::gen_requests(uint64_t thd_id, workload * h_wl) {
#if CC_ALG == HSTORE
assert(g_virtual_part_cnt == g_part_cnt);
#endif
int access_cnt = 0;
set<uint64_t> all_keys;
part_num = 0;
double r = 0;
int64_t rint64 = 0;
drand48_r(&_query_thd->buffer, &r);
lrand48_r(&_query_thd->buffer, &rint64);
if (r < g_perc_multi_part) {
for (UInt32 i = 0; i < g_part_per_txn; i++) {
if (i == 0 && FIRST_PART_LOCAL)
part_to_access[part_num] = thd_id % g_virtual_part_cnt;
else {
part_to_access[part_num] = rint64 % g_virtual_part_cnt;
}
UInt32 j;
for (j = 0; j < part_num; j++)
if ( part_to_access[part_num] == part_to_access[j] )
break;
if (j == part_num)
part_num ++;
}
} else {
part_num = 1;
if (FIRST_PART_LOCAL)
part_to_access[0] = thd_id % g_part_cnt;
else
part_to_access[0] = rint64 % g_part_cnt;
}
int rid = 0;
for (UInt32 tmp = 0; tmp < g_req_per_query; tmp ++) {
double r;
drand48_r(&_query_thd->buffer, &r);
ycsb_request * req = &requests[rid];
if (r < g_read_perc) {
req->rtype = RD;
} else if (r >= g_read_perc && r <= g_write_perc + g_read_perc) {
req->rtype = WR;
} else {
req->rtype = SCAN;
req->scan_len = SCAN_LEN;
}
// the request will access part_id.
uint64_t ith = tmp * part_num / g_req_per_query;
uint64_t part_id =
part_to_access[ ith ];
uint64_t table_size = g_synth_table_size / g_virtual_part_cnt;
uint64_t row_id = zipf(table_size - 1, g_zipf_theta);
assert(row_id < table_size);
uint64_t primary_key = row_id * g_virtual_part_cnt + part_id;
req->key = primary_key;
int64_t rint64;
lrand48_r(&_query_thd->buffer, &rint64);
req->value = rint64 % (1<<8);
// Make sure a single row is not accessed twice
if (req->rtype == RD || req->rtype == WR) {
if (all_keys.find(req->key) == all_keys.end()) {
all_keys.insert(req->key);
access_cnt ++;
} else continue;
} else {
bool conflict = false;
for (UInt32 i = 0; i < req->scan_len; i++) {
primary_key = (row_id + i) * g_part_cnt + part_id;
if (all_keys.find( primary_key )
!= all_keys.end())
conflict = true;
}
if (conflict) continue;
else {
for (UInt32 i = 0; i < req->scan_len; i++)
all_keys.insert( (row_id + i) * g_part_cnt + part_id);
access_cnt += SCAN_LEN;
}
}
rid ++;
}
request_cnt = rid;
// Sort the requests in key order.
if (g_key_order) {
for (int i = request_cnt - 1; i > 0; i--)
for (int j = 0; j < i; j ++)
if (requests[j].key > requests[j + 1].key) {
ycsb_request tmp = requests[j];
requests[j] = requests[j + 1];
requests[j + 1] = tmp;
}
for (UInt32 i = 0; i < request_cnt - 1; i++)
assert(requests[i].key < requests[i + 1].key);
}
}