archive/ruby
1
0
Fork 0
mirror of https://github.com/ruby/ruby.git synced 2025-09-16 17:14:01 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 10

merge revision(s) 49376,49387,49389: [Backport #10761]

Browse source 
* st.c (st_numhash): mix float value for flonum

	* hash.c (rb_any_hash): ditto

	* benchmark/bm_hash_aref_flo.rb: new benchmark

	* benchmark/bm_hash_ident_flo.rb: ditto
	  [Bug #10761]

	* benchmark/bm_marshal_dump_flo.rb: new benchmark for [Bug #10761]

	* marshal.c (w_object, marshal_dump): use indetity tables for
	  arbitrary VALUE keys, because of performance of FLONUM.
	  [Bug #10761]

	* marshal.c (obj_alloc_by_klass, marshal_load): ditto.


git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/branches/ruby_2_2@49513 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
This commit is contained in:
naruse 2015-02-05 16:05:10 +00:00
parent b5623f9863
commit 665f7a2ff1
9 changed files with 67 additions and 8 deletions
Download patch file Download diff file Expand all files Collapse all files

9
st.c
View file

@ -1761,6 +1761,15 @@ st_numhash(st_data_t n)
* - (n << 3) was finally added to avoid losing bits for fixnums
* - avoid expensive modulo instructions, it is currently only
* shifts and bitmask operations.
* - flonum (on 64-bit) is pathologically bad, mix the actual
* float value in, but do not use the float value as-is since
* many integers get interpreted as 2.0 or -2.0 [Bug #10761]
*/
#ifdef USE_FLONUM /* RUBY */
if (FLONUM_P(n)) {
n ^= (st_data_t)rb_float_value(n);
}
#endif
return (st_index_t)((n>>(RUBY_SPECIAL_SHIFT+3)|(n<<3)) ^ (n>>3));
}