So, I'm working with some memory bound applications and I have:
1 - Two arrays of structs that simulates tables on a vertical database. One of them just with keys (1.5M 32-bits integer keys) and another one with integer keys and double payloads (150k tuples). The two of then dynamically allocated
2 - An array of 2^15 64-bits unsigned integers
3 - An array of 2^10 32-bits unsigned integers
And I need to allocate dynamically an array of 32-bits integers which I will know the size just on runtime.
The problem is: I've been able to allocate this array using malloc, BUT when I initialize values to zero, it just subscribes the values of the 150k tuples table. Which means, I`m losing data. The worst thing that could happen to a databases researcher.
Allocation of the "tables"
tamCustomer = countLines("customer.tbl");
c_customer = malloc(tamCustomer*sizeof(column_customer));
readCustomerColumn("customer.tbl", c_customer);
tamOrders = countLines("orders.tbl");
c_orders = malloc(tamOrders*sizeof(column_orders));
readOrdersColumn("orders.tbl", c_orders, sel);
Allocation of the problematic array
cht->tamHT = actualPopCounter;
cht->HT = malloc(sizeof(uint32_t)*cht->tamHT);
if (cht->HT == NULL)
printf("deu merda\n");
for (int i=0; i<cht->tamHT; i++)
cht->HT[i] = 0;
So, after this point, half of the table c_customer gets lost, subscribed by zeros.
What can I do to avoid that?
EDIT: structs definitions:
/******** VETOR DE STRUCTS COLUMN CUSTOMER *********/
typedef struct customer_c
{
unsigned int C_CUSTKEY;
float C_ACCTBAL;
} column_customer;
column_customer *c_customer;
/******** VETOR DE STRUCTS COLUMN ORDERS ***********/
typedef struct orders_c
{
unsigned int O_CUSTKEY;
} column_orders;
column_orders *c_orders;
CHT definition:
typedef struct CHT
{
uint64_t bitmap[CHT_BMP_SIZE];
bucket OHT[CHT_OHT_SIZE];
bucket *HT;
uint32_t tamHT;
} CHT;
And thats pretty much the function where it occurs. This is not a small application and I
ve been so focused on this problem that I can`t think properly right now (sorry).
inline void generateCHT(column_customer *c_customer, int tamCustomer, CHT * cht)
{
uint32_t ohtOcc=0;
uint32_t chtOcc=0;
uint32_t ohtOccBMP=0;
uint32_t chtOccBMP=0;
uint64_t actualPopCounter;
uint64_t oldPopCounter;
//Allocate CHT
cht->tamHT = 0;
//Initialize OHT and bitmap
for (int i=0; i<CHT_OHT_SIZE;i++)
{
cht->OHT[i]=0;
cht->bitmap[i]=0;
}
for (int i=0; i<tamCustomer; i++)
{
switch (chtInsertBitmap(c_customer[i].C_CUSTKEY, tamCustomer, cht))
{
case 0:
printf("ERROR: Something went wrong while inserting the key %u on the CHT\n", c_customer[i].C_CUSTKEY);
break;
case 1:
chtOccBMP++;
break;
case 2:
ohtOccBMP++;
break;
}
}
//count Population
actualPopCounter = 0;
for (int i=0; i<CHT_BMP_SIZE;i++)
{
oldPopCounter = popCount(cht->bitmap[i]>>32);
cht->bitmap[i] = cht->bitmap[i] | actualPopCounter;
actualPopCounter = actualPopCounter + oldPopCounter;
}
cht->tamHT = actualPopCounter;
cht->HT = malloc(sizeof(uint32_t)*cht->tamHT);
if (cht->HT == NULL)
printf("deu merda\n");
for (int i=0; i<cht->tamHT; i++)
cht->HT[i] = 0;
for (int i=0; i<tamCustomer; i++)
{
if (chtInsertConciseTable(c_customer[i].C_CUSTKEY, cht, tamCustomer) == 0)
ohtOcc++;
else
chtOcc++;
}
printf("OHT has %d occupied buckets and %d on the bitmap \n", ohtOcc, ohtOccBMP);
printf("CHT has %d occupied buckets and %d on the bitmap \n", chtOcc, chtOccBMP);
}
You're possibly walking off the end of the cht->HT
array you allocated.
bucket *HT;
...
...
cht->HT = malloc(sizeof(uint32_t)*cht->tamHT);
Try sizeof(bucket)
instead.