greg/pcompress
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

Use openmp parallelism always for chunk hash computation during Global Dedupe.

Browse source
This commit is contained in:
Moinak Ghosh 2013-05-02 23:24:43 +05:30
parent 120877348c
commit c43e99f422
1 changed files with 2 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

4
rabin/rabin_dedup.c
View file

@ -766,7 +766,7 @@ process_blocks:
* First compute all the rabin chunk/block cryptographic hashes. * First compute all the rabin chunk/block cryptographic hashes.
*/ */
#if defined(_OPENMP) #if defined(_OPENMP)
# pragma omp parallel for if (mt) # pragma omp parallel for
#endif #endif
for (i=0; i<blknum; i++) { for (i=0; i<blknum; i++) {
compute_checksum(ctx->g_blocks[i].cksum, 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 * Now lookup blocks in index. First wait for our semaphore to be
* signaled. If the previous thread in sequence is using the index * 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 * predictable serialization of index access in a sequence of
* threads without locking. * threads without locking.
*/ */