ruby-changes:46313
From: shyouhei <ko1@a...>
Date: Fri, 21 Apr 2017 13:21:18 +0900 (JST)
Subject: [ruby-changes:46313] shyouhei:r58427 (trunk): refactor hash literal
shyouhei 2017-04-21 13:21:14 +0900 (Fri, 21 Apr 2017) New Revision: 58427 https://svn.ruby-lang.org/cgi-bin/viewvc.cgi?view=revision&revision=58427 Log: refactor hash literal Same as rb_ary_tmp_new_from_values(), it reduces vm_exec_core binary size from 26,176 bytes to 26,080 bytes. But this time, also with a bit of optimizations: - Because we are allocating a new hash and no back references are introduced at all, we can safely skip write barriers. - Also, the iteration never recurs. We can avoid complicated function callbacks by using st_insert instead of st_update. ---- * hash.c (rb_hash_new_from_values): refactor extract the bulk insert into a function. * hash.c (rb_hash_new_from_object): also refactor. * hash.c (rb_hash_s_create): use the new functions. * insns.def (newhash): ditto. * vm.c (core_hash_from_ary): ditto. * iternal.h: export the new function. ----------------------------------------------------------- benchmark results: minimum results in each 7 measurements. Execution time (sec) name before after loop_whileloop2 0.135 0.134 vm2_bighash* 1.236 0.687 Speedup ratio: compare with the result of `before' (greater is better) name after loop_whileloop2 1.008 vm2_bighash* 1.798 Modified files: trunk/hash.c trunk/insns.def trunk/internal.h trunk/vm.c Index: vm.c =================================================================== --- vm.c (revision 58426) +++ vm.c (revision 58427) @@ -2646,7 +2646,6 @@ m_core_set_postexe(VALUE self) https://github.com/ruby/ruby/blob/trunk/vm.c#L2646 return Qnil; } -static VALUE core_hash_merge_ary(VALUE hash, VALUE ary); static VALUE core_hash_from_ary(VALUE ary); static VALUE core_hash_merge_kwd(int argc, VALUE *argv); @@ -2656,7 +2655,6 @@ core_hash_merge(VALUE hash, long argc, c https://github.com/ruby/ruby/blob/trunk/vm.c#L2655 long i; Check_Type(hash, T_HASH); - VM_ASSERT(argc % 2 == 0); for (i=0; i<argc; i+=2) { rb_hash_aset(hash, argv[i], argv[i+1]); } @@ -2674,27 +2672,18 @@ m_core_hash_from_ary(VALUE self, VALUE a https://github.com/ruby/ruby/blob/trunk/vm.c#L2672 static VALUE core_hash_from_ary(VALUE ary) { - VALUE hash = rb_hash_new(); + long n; - RUBY_DTRACE_CREATE_HOOK(HASH, (Check_Type(ary, T_ARRAY), RARRAY_LEN(ary))); - return core_hash_merge_ary(hash, ary); -} - -#if 0 -static VALUE -m_core_hash_merge_ary(VALUE self, VALUE hash, VALUE ary) -{ - REWIND_CFP(core_hash_merge_ary(hash, ary)); - return hash; -} -#endif - -static VALUE -core_hash_merge_ary(VALUE hash, VALUE ary) -{ Check_Type(ary, T_ARRAY); - core_hash_merge(hash, RARRAY_LEN(ary), RARRAY_CONST_PTR(ary)); - return hash; + n = RARRAY_LEN(ary); + RUBY_DTRACE_CREATE_HOOK(HASH, n); + if (n) { + VM_ASSERT(n % 2 == 0); + return rb_hash_new_from_values(n, RARRAY_PTR(ary)); + } + else { + return rb_hash_new(); + } } static VALUE Index: insns.def =================================================================== --- insns.def (revision 58426) +++ insns.def (revision 58427) @@ -492,16 +492,13 @@ newhash https://github.com/ruby/ruby/blob/trunk/insns.def#L492 (...) (VALUE val) // inc += 1 - num; { - rb_num_t i; - RUBY_DTRACE_CREATE_HOOK(HASH, num); - val = rb_hash_new(); - - for (i = num; i > 0; i -= 2) { - const VALUE v = TOPN(i - 2); - const VALUE k = TOPN(i - 1); - rb_hash_aset(val, k, v); + if (num) { + val = rb_hash_new_from_values(num, STACK_ADDR_FROM_TOP(num)); + } + else { + val = rb_hash_new(); } POPN(num); } Index: hash.c =================================================================== --- hash.c (revision 58426) +++ hash.c (revision 58427) @@ -626,6 +626,103 @@ rb_hash_initialize(int argc, VALUE *argv https://github.com/ruby/ruby/blob/trunk/hash.c#L626 return hash; } +static inline VALUE +hash_alloc_from_st(VALUE klass, st_table *ntbl) +{ + VALUE h = hash_alloc(klass); + RHASH(h)->ntbl = ntbl; + return h; +} + +static inline int +hash_insert_raw(st_table *tbl, VALUE key, VALUE val) +{ + st_data_t v = (st_data_t)val; + st_data_t k = (rb_obj_class(k) == rb_cString) ? + (st_data_t)rb_str_new_frozen(key) : + (st_data_t)key; + + return st_insert(tbl, k, v); +} + +static VALUE +rb_hash_new_from_values_with_klass(long argc, const VALUE *argv, VALUE klass) +{ + long i; + st_table *t; + + if (argc % 2) { + rb_raise(rb_eArgError, "odd number of arguments for Hash"); + } + + t = st_init_table_with_size(&objhash, argc / 2); + for (i = 0; i < argc; /* */) { + VALUE key = argv[i++]; + VALUE val = argv[i++]; + + hash_insert_raw(t, key, val); + } + return hash_alloc_from_st(klass, t); +} + +VALUE +rb_hash_new_from_values(long argc, const VALUE *argv) +{ + return rb_hash_new_from_values_with_klass(argc, argv, rb_cHash); +} + +static VALUE +rb_hash_new_from_object(VALUE klass, VALUE obj) +{ + VALUE tmp = rb_hash_s_try_convert(Qnil, obj); + if (!NIL_P(tmp)) { + VALUE hash = hash_alloc(klass); + if (RHASH(tmp)->ntbl) { + RHASH(hash)->ntbl = st_copy(RHASH(tmp)->ntbl); + } + return hash; + } + + tmp = rb_check_array_type(obj); + if (!NIL_P(tmp)) { + long i; + long len = RARRAY_LEN(tmp); + st_table *tbl = len ? st_init_table_with_size(&objhash, len) : NULL; + for (i = 0; i < len; ++i) { + VALUE e = RARRAY_AREF(tmp, i); + VALUE v = rb_check_array_type(e); + VALUE key, val = Qnil; + + if (NIL_P(v)) { +#if 0 /* refix in the next release */ + rb_raise(rb_eArgError, "wrong element type %s at %ld (expected array)", + rb_builtin_class_name(e), i); + +#else + rb_warn("wrong element type %s at %ld (expected array)", + rb_builtin_class_name(e), i); + rb_warn("ignoring wrong elements is deprecated, remove them explicitly"); + rb_warn("this causes ArgumentError in the next release"); + continue; +#endif + } + switch (RARRAY_LEN(v)) { + default: + rb_raise(rb_eArgError, "invalid number of elements (%ld for 1..2)", + RARRAY_LEN(v)); + case 2: + val = RARRAY_AREF(v, 1); + case 1: + key = RARRAY_AREF(v, 0); + hash_insert_raw(tbl, key, val); + } + } + return hash_alloc_from_st(klass, tbl); + } + + rb_raise(rb_eArgError, "odd number of arguments for Hash"); +} + /* * call-seq: * Hash[ key, value, ... ] -> new_hash @@ -649,69 +746,11 @@ rb_hash_initialize(int argc, VALUE *argv https://github.com/ruby/ruby/blob/trunk/hash.c#L746 static VALUE rb_hash_s_create(int argc, VALUE *argv, VALUE klass) { - VALUE hash, tmp; - int i; - - if (argc == 1) { - tmp = rb_hash_s_try_convert(Qnil, argv[0]); - if (!NIL_P(tmp)) { - hash = hash_alloc(klass); - if (RHASH(tmp)->ntbl) { - RHASH(hash)->ntbl = st_copy(RHASH(tmp)->ntbl); - } - return hash; - } - - tmp = rb_check_array_type(argv[0]); - if (!NIL_P(tmp)) { - long i; - - hash = hash_alloc(klass); - for (i = 0; i < RARRAY_LEN(tmp); ++i) { - VALUE e = RARRAY_AREF(tmp, i); - VALUE v = rb_check_array_type(e); - VALUE key, val = Qnil; - - if (NIL_P(v)) { -#if 0 /* refix in the next release */ - rb_raise(rb_eArgError, "wrong element type %s at %ld (expected array)", - rb_builtin_class_name(e), i); - -#else - rb_warn("wrong element type %s at %ld (expected array)", - rb_builtin_class_name(e), i); - rb_warn("ignoring wrong elements is deprecated, remove them explicitly"); - rb_warn("this causes ArgumentError in the next release"); - continue; -#endif - } - switch (RARRAY_LEN(v)) { - default: - rb_raise(rb_eArgError, "invalid number of elements (%ld for 1..2)", - RARRAY_LEN(v)); - case 2: - val = RARRAY_AREF(v, 1); - case 1: - key = RARRAY_AREF(v, 0); - rb_hash_aset(hash, key, val); - } - } - return hash; - } + switch (argc) { + case 0: return hash_alloc(klass); + case 1: return rb_hash_new_from_object(klass, argv[0]); + default: return rb_hash_new_from_values_with_klass(argc, argv, klass); } - if (argc % 2 != 0) { - rb_raise(rb_eArgError, "odd number of arguments for Hash"); - } - - hash = hash_alloc(klass); - if (argc > 0) { - RHASH(hash)->ntbl = st_init_table_with_size(&objhash, argc / 2); - } - for (i=0; i<argc; i+=2) { - rb_hash_aset(hash, argv[i], argv[i + 1]); - } - - return hash; } static VALUE Index: internal.h =================================================================== --- internal.h (revision 58426) +++ internal.h (revision 58427) @@ -1236,6 +1236,7 @@ VALUE rb_hash_rehash(VALUE hash); https://github.com/ruby/ruby/blob/trunk/internal.h#L1236 int rb_hash_add_new_element(VALUE hash, VALUE key, VALUE val); #define HASH_DELETED FL_USER1 #define HASH_PROC_DEFAULT FL_USER2 +extern VALUE rb_hash_new_from_values(long, const VALUE *); /* inits.c */ void rb_call_inits(void); -- ML: ruby-changes@q... Info: http://www.atdot.net/~ko1/quickml/