Use openmp parallelism always for chunk hash computation during Global Dedupe.
This commit is contained in:
parent
120877348c
commit
c43e99f422
1 changed files with 2 additions and 2 deletions
|
@ -766,7 +766,7 @@ process_blocks:
|
|||
* First compute all the rabin chunk/block cryptographic hashes.
|
||||
*/
|
||||
#if defined(_OPENMP)
|
||||
# pragma omp parallel for if (mt)
|
||||
# pragma omp parallel for
|
||||
#endif
|
||||
for (i=0; i<blknum; i++) {
|
||||
compute_checksum(ctx->g_blocks[i].cksum,
|
||||
|
@ -797,7 +797,7 @@ process_blocks:
|
|||
/*
|
||||
* Now lookup blocks in index. First wait for our semaphore to be
|
||||
* signaled. If the previous thread in sequence is using the index
|
||||
* he will finish and then signal our semaphore. So we can have
|
||||
* it will finish and then signal our semaphore. So we can have
|
||||
* predictable serialization of index access in a sequence of
|
||||
* threads without locking.
|
||||
*/
|
||||
|
|
Loading…
Reference in a new issue