refactor newhash (revision 58463 another try) #1600
Conversation
* st.c (rb_hash_bulk_insert): new API to bulk insert entries into a hash. Given arguments are first inserted into the table at once, then reindexed. This is faster than inserting things using rb_hash_aset() one by one. This arrangement (rb_ prefixed function placed in st.c) is unavoidable because it both touches table internal and write barrier at once. * internal.h: delcare the new function. * hash.c (rb_hash_s_create): use the new function. * vm.c (core_hash_merge): ditto. * insns.def (newhash): ditto. ----------------------------------------------------------- benchmark results: minimum results in each 7 measurements. Execution time (sec) name before after loop_whileloop2 0.136 0.137 vm2_bighash* 1.249 0.623 Speedup ratio: compare with the result of `before' (greater is better) name after loop_whileloop2 0.996 vm2_bighash* 2.004
|
@vnmakarov I would like to merge this hash bulk insert feature. Can you take a look at st.c diff? |
|
Shyouhei, thank you for asking my opinion. About half year ago I thought about this too to improve vm2_bighash and overall hash tables benchmark score. To be honest I did not expect such big improvement of vm2_bighash which you reported here. That time my approach would be simply using st_insert2 with providing its inlining. Basically it would be analog of your rb_hash_bulk_insert with st_insert2 (more accurately a static inline function which could be used in st_insert2 too) inside rb_hash_bulk loop. Your implementation is probably faster as it has more specialized code although I don't know how much faster it is. I have no time and hardly will have time to work on hash tables more. So I am really glad that you implemented it. I've read your code. It looks OK for me. If you need my support to put your changes into st.c. I gladly give it to you. |
|
Thank you for reviewing. Will merge this. FYI, although reverted due to random bugs, I tried using st_insert before (cf: 4ee09d9). It was a bit slower than what is proposed in this pull request (1.8 times faster then versus 2.0 times faster here). |
st.c (rb_hash_bulk_insert): new API to bulk insert entries
into a hash. Given arguments are first inserted into the
table at once, then reindexed. This is faster than inserting
things using rb_hash_aset() one by one.
This arrangement (rb_ prefixed function placed in st.c) is
unavoidable because it both touches table internal and write
barrier at once.
internal.h: delcare the new function.
hash.c (rb_hash_s_create): use the new function.
vm.c (core_hash_merge): ditto.
insns.def (newhash): ditto.