| 26,7 → 26,7 |
|---|
| * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| */ |
| |
| /** @addtogroup genericmm |
| /** @addtogroup genericmm |
| * @{ |
| */ |
| |
| 83,7 → 83,7 |
|---|
| * The brutal reclaim removes all cached objects, even from CPU-bound |
| * magazines. |
| * |
| * TODO:@n |
| * @todo |
| * For better CPU-scaling the magazine allocation strategy should |
| * be extended. Currently, if the cache does not have magazine, it asks |
| * for non-cpu cached magazine cache to provide one. It might be feasible |
| 93,9 → 93,10 |
|---|
| * 'empty-magazine-list', which decreases competing for 1 per-system |
| * magazine cache. |
| * |
| * @li it might be good to add granularity of locks even to slab level, |
| * we could then try_spinlock over all partial slabs and thus improve |
| * scalability even on slab level |
| * @todo |
| * it might be good to add granularity of locks even to slab level, |
| * we could then try_spinlock over all partial slabs and thus improve |
| * scalability even on slab level |
| */ |
| |
| #include <synch/spinlock.h> |
| 907,6 → 908,5 |
|---|
| _slab_free(slab->cache, obj, slab); |
| } |
| |
| /** @} |
| /** @} |
| */ |
| |