From: vmakarov@... Date: 2016-03-17T14:33:30+00:00 Subject: [ruby-core:74426] [Ruby trunk Feature#12142] Hash tables with open addressing Issue #12142 has been updated by Vladimir Makarov. Koichi Sasada wrote: > On 2016/03/17 15:00, shyouhei@ruby-lang.org wrote: > > About 64bit versus 32bit index: several developers discussed this on this month's developer meeting. Consensus there was that we do not want to limit Hash size, while it is true that over 99% of hashes are smaller than 4G entries, and it is definitely a good idea to optimize to them. We did not reach a consensus as to how should we achieve that, though. Proposed ideas include: switch index type using configure, have 8/16/32/64bit versions in parallel and switch smoothly with increasing hash size. > > Compliment: > > (1) switch index type using configure > > Easy. However, we can not test both versions. > > (2) have 8/16/32/64bit versions in parallel and switch smoothly with > increasing hash size. > > There is one more: > > (3) st only supports 32bit, but Hash object support 64bit using another > technique (maybe having multiple tables). > > Thank you for the clarification. I'll investigate the possibility of usage less bits for indexes (and may be hashes) without introducing constraints on the table size. Right now I can say that using 32-bit for 64-bit Haswell improves the average performance of my tables by about 2% on MRI hash benchmarks. On my estimations, the size should be decreased by 12.5%. So it is worth to investigate this. > > Anyway, we concluded that this optimization should be introduced for Rub > 2.4 even if it only supports 64bit. Optimization with 32bit length index > is nice, but we can optimize later. I'll work on it. I thought that my patch was a final variant of the table when I submitted it first. But having all the feedback I believe I should work more on the code. As I wrote I'll probably submit a few patches because some work I've already done might seem controversial. I think about the new hash table patch, new hash functions patch, and a patch for the tables dealing with a denial attack. Yura Sokolov could submit his patch for using alternative allocation. I planning to do some research, actual coding and submit the patches definitely before the summer but it might happen earlier (it depends on how busy I am with my major work). I believe we will have enough time to introduce the new hash tables in 2.4. ---------------------------------------- Feature #12142: Hash tables with open addressing https://bugs.ruby-lang.org/issues/12142#change-57547 * Author: Vladimir Makarov * Status: Open * Priority: Normal * Assignee: ---------------------------------------- ~~~ Hello, the following patch contains a new implementation of hash tables (major files st.c and include/ruby/st.h). Modern processors have several levels of cache. Usually,the CPU reads one or a few lines of the cache from memory (or another level of cache). So CPU is much faster at reading data stored close to each other. The current implementation of Ruby hash tables does not fit well to modern processor cache organization, which requires better data locality for faster program speed. The new hash table implementation achieves a better data locality mainly by o switching to open addressing hash tables for access by keys. Removing hash collision lists lets us avoid *pointer chasing*, a common problem that produces bad data locality. I see a tendency to move from chaining hash tables to open addressing hash tables due to their better fit to modern CPU memory organizations. CPython recently made such switch (https://hg.python.org/cpython/file/ff1938d12240/Objects/dictobject.c). PHP did this a bit earlier https://nikic.github.io/2014/12/22/PHPs-new-hashtable-implementation.html. GCC has widely-used such hash tables (https://gcc.gnu.org/svn/gcc/trunk/libiberty/hashtab.c) internally for more than 15 years. o removing doubly linked lists and putting the elements into an array for accessing to elements by their inclusion order. That also removes pointer chaising on the doubly linked lists used for traversing elements by their inclusion order. A more detailed description of the proposed implementation can be found in the top comment of the file st.c. The new implementation was benchmarked on 21 MRI hash table benchmarks for two most widely used targets x86-64 (Intel 4.2GHz i7-4790K) and ARM (Exynos 5410 - 1.6GHz Cortex-A15): make benchmark-each ITEM=bm_hash OPTS='-r 3 -v' COMPARE_RUBY='' Here the results for x86-64: hash_aref_dsym 1.094 hash_aref_dsym_long 1.383 hash_aref_fix 1.048 hash_aref_flo 1.860 hash_aref_miss 1.107 hash_aref_str 1.107 hash_aref_sym 1.191 hash_aref_sym_long 1.113 hash_flatten 1.258 hash_ident_flo 1.627 hash_ident_num 1.045 hash_ident_obj 1.143 hash_ident_str 1.127 hash_ident_sym 1.152 hash_keys 2.714 hash_shift 2.209 hash_shift_u16 1.442 hash_shift_u24 1.413 hash_shift_u32 1.396 hash_to_proc 2.831 hash_values 2.701 The average performance improvement is more 50%. ARM results are analogous -- no any benchmark performance degradation and about the same average improvement. The patch can be seen as https://github.com/vnmakarov/ruby/compare/trunk...hash_tables_with_open_addressing.patch or in a less convenient way as pull request changes https://github.com/ruby/ruby/pull/1264/files This is my first patch for MRI and may be my proposal and implementation have pitfalls. But I am keen to learn and work on inclusion of this code into MRI. ~~~ ---Files-------------------------------- 0001-st.c-use-array-for-storing-st_table_entry.patch (46.7 KB) 0001-st.c-change-st_table-implementation.patch (59.4 KB) -- https://bugs.ruby-lang.org/ Unsubscribe: