Page Index - laurynas-biveinis/unodb GitHub Wiki

56 page(s) in this GitHub Wiki:

Home
Concurrent ART flavor overhead in single threaded workloads
Effect of annotating pmr_allocate with attribute((malloc,returns_nonnull))
Effect of AVX2 for Node48 insert position search
Effect of branching exactly once in remove code path
Effect of fully inlining try_read_unlock
Effect of initializing Node48 child_indexes at declaration
Effect of jemalloc sdallocx sized free
Effect of LD_PRELOAD'ing jemalloc
Effect of lock free QSBR step 1
Effect of lock free QSBR step 10
Effect of lock free QSBR step 11
Effect of lock free QSBR step 12
Effect of lock free QSBR step 2
Effect of lock free QSBR step 3
Effect of lock free QSBR step 4
Effect of lock free QSBR step 5
Effect of lock free QSBR step 6
Effect of lock free QSBR step 7
Effect of lock free QSBR step 8
Effect of lock free QSBR step 9
Effect of merging find_child and add node type dispatch
Effect of Node16 NEON search
Effect of Node16 NEON search replacing uint128_t with a ternary
Effect of Node4 NEON search
Effect of Node48 AVX2 insert position search unrolling x2
Effect of Node48 AVX2 insert position search unrolling x4
Effect of Node48 NEON search
Effect of Node48 NEON search unrolled x2
Effect of optimistic_lock::try_read_lock testing once in the straight path
Effect of padding OLC ART counters
Effect of removing branching on children_count in N4 insert_position_search
Effect of removing is_full dispatch on add node path
Effect of removing node type check before growing inode
Effect of removing Node256 fullness check on add
Effect of removing polymorphism on inode entry removal
Effect of removing pool PMR allocators
Effect of removing redundant nullptr checks in deleters
Effect of removing std::uncaught_exceptions calls on OL obsoleting critical section enter and exit
Effect of replacing RMW atomics with load non atomic modify store in OL critical sections
Effect of replacing RMW with W in optimistic lock obsoletion
Effect of replacing RMW with W in optimistic lock write unlock
Effect of SIMD'ing Node4 insert position search
Effect of simpler Node4 add_two_to_empty key2_i
Effect of tagged node pointers
Performance comparison with @justinasvd art_map
Performance of GCC 11 vs GCC 10
Performance of GCC 12 vs GCC 11
Performance of LLVM 12 vs LLVM 11
Performance of LLVM 12 vs LLVM 11, take 2
Performance of LLVM 13 vs LLVM 12
Performance of LLVM 14 vs LLVM 13
Performance of LLVM 15 vs LLVM 14
Performance of LLVM 16 vs LLVM 15
Performance of LLVM 17 vs LLVM 16
Performance Tests