dcache RCU reduces system time for kernel compile by 10%

Martin J. Bligh (fletch@aracnet.com)
Sat, 21 Sep 2002 13:09:12 -0700


Installing dcache rcu drops system time for a kernel compile
from 52.7s to 47.3s - a little better than 10% reduction.
Total elapsed time doesn't change much, but I think this is
due to the lack of parallelism in the benchmark ... I need to
start running 32 parallel kernel compiles ;-)

Results were from a 16 cpu NUMA-Q. Comparison done on 2.5.36-mm1

Profile head before RCU:

6964 do_anonymous_page
4297 .text.lock.namei
4189 page_remove_rmap
2261 page_add_rmap
2245 __generic_copy_from_user
2217 .text.lock.dec_and_lock
1685 vm_enough_memory
1539 zap_pte_range
1503 file_read_actor
1389 rmqueue
1139 find_get_page
1074 get_empty_filp
1063 do_no_page
975 link_path_walk
863 __fput
838 path_lookup
833 __free_pages_ok
810 atomic_dec_and_lock
793 __d_lookup
737 do_page_fault
670 .text.lock.file_table
593 dput
577 pte_alloc_one
554 __generic_copy_to_user

profile head after RCU:

6841 do_anonymous_page
4171 page_remove_rmap
2423 __generic_copy_from_user
2321 page_add_rmap
2247 d_lookup
1773 vm_enough_memory
1566 zap_pte_range
1527 file_read_actor
1342 rmqueue
1281 .text.lock.file_table
1153 find_get_page
1149 get_empty_filp
1049 do_no_page
891 atomic_dec_and_lock
775 do_page_fault
752 __free_pages_ok
667 follow_mount
549 __fput
535 pte_alloc_one
529 do_generic_file_read
521 path_lookup
517 schedule
509 release_pages
504 do_softirq

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/