Part of the plan to reduce dependencies on malloc_usable_size
which costs us speed: https://bugs.ruby-lang.org/issues/10238
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@63481 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* array.c (rb_ary_collect): yield in block argument semantics
always to splat array elements to lambda, for the backward
compatibility. [ruby-core:86362] [Bug #14639]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@63030 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* In Enumerable, Enumerator::Lazy, Array, Hash and Set
[Feature #13784] [ruby-core:82285]
* Share specs for the various #select#select! methods and
reuse them for #filter/#filter!.
* Add corresponding filter tests for select tests.
* Update NEWS.
[Fix GH-1824]
From: Alexander Patrick <adp90@case.edu>
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@62575 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* array.c (rb_ary_sample): Adds examples for the use of the
optional parameter random for Array#sample and unifies the style
with the documentation of Array#shuffle.
[Fix GH-1825]
From: Alberto Almagro <alberto.almagro@rakuten.com>
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@62553 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* array.c (rb_ary_values_at): optimization range argument case.
bulk concatenation than pushing for each element.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@62413 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
which has been developed by Takashi Kokubun <takashikkbn@gmail> as
YARV-MJIT. Many of its bugs are fixed by wanabe <s.wanabe@gmail.com>.
This JIT compiler is designed to be a safe migration path to introduce
JIT compiler to MRI. So this commit does not include any bytecode
changes or dynamic instruction modifications, which are done in original
MJIT.
This commit even strips off some aggressive optimizations from
YARV-MJIT, and thus it's slower than YARV-MJIT too. But it's still
fairly faster than Ruby 2.5 in some benchmarks (attached below).
Note that this JIT compiler passes `make test`, `make test-all`, `make
test-spec` without JIT, and even with JIT. Not only it's perfectly safe
with JIT disabled because it does not replace VM instructions unlike
MJIT, but also with JIT enabled it stably runs Ruby applications
including Rails applications.
I'm expecting this version as just "initial" JIT compiler. I have many
optimization ideas which are skipped for initial merging, and you may
easily replace this JIT compiler with a faster one by just replacing
mjit_compile.c. `mjit_compile` interface is designed for the purpose.
common.mk: update dependencies for mjit_compile.c.
internal.h: declare `rb_vm_insn_addr2insn` for MJIT.
vm.c: exclude some definitions if `-DMJIT_HEADER` is provided to
compiler. This avoids to include some functions which take a long time
to compile, e.g. vm_exec_core. Some of the purpose is achieved in
transform_mjit_header.rb (see `IGNORED_FUNCTIONS`) but others are
manually resolved for now. Load mjit_helper.h for MJIT header.
mjit_helper.h: New. This is a file used only by JIT-ed code. I'll
refactor `mjit_call_cfunc` later.
vm_eval.c: add some #ifdef switches to skip compiling some functions
like Init_vm_eval.
win32/mkexports.rb: export thread/ec functions, which are used by MJIT.
include/ruby/defines.h: add MJIT_FUNC_EXPORTED macro alis to clarify
that a function is exported only for MJIT.
array.c: export a function used by MJIT.
bignum.c: ditto.
class.c: ditto.
compile.c: ditto.
error.c: ditto.
gc.c: ditto.
hash.c: ditto.
iseq.c: ditto.
numeric.c: ditto.
object.c: ditto.
proc.c: ditto.
re.c: ditto.
st.c: ditto.
string.c: ditto.
thread.c: ditto.
variable.c: ditto.
vm_backtrace.c: ditto.
vm_insnhelper.c: ditto.
vm_method.c: ditto.
I would like to improve maintainability of function exports, but I
believe this way is acceptable as initial merging if we clarify the
new exports are for MJIT (so that we can use them as TODO list to fix)
and add unit tests to detect unresolved symbols.
I'll add unit tests of JIT compilations in succeeding commits.
Author: Takashi Kokubun <takashikkbn@gmail.com>
Contributor: wanabe <s.wanabe@gmail.com>
Part of [Feature #14235]
---
* Known issues
* Code generated by gcc is faster than clang. The benchmark may be worse
in macOS. Following benchmark result is provided by gcc w/ Linux.
* Performance is decreased when Google Chrome is running
* JIT can work on MinGW, but it doesn't improve performance at least
in short running benchmark.
* Currently it doesn't perform well with Rails. We'll try to fix this
before release.
---
* Benchmark reslts
Benchmarked with:
Intel 4.0GHz i7-4790K with 16GB memory under x86-64 Ubuntu 8 Cores
- 2.0.0-p0: Ruby 2.0.0-p0
- r62186: Ruby trunk (early 2.6.0), before MJIT changes
- JIT off: On this commit, but without `--jit` option
- JIT on: On this commit, and with `--jit` option
** Optcarrot fps
Benchmark: https://github.com/mame/optcarrot
| |2.0.0-p0 |r62186 |JIT off |JIT on |
|:--------|:--------|:--------|:--------|:--------|
|fps |37.32 |51.46 |51.31 |58.88 |
|vs 2.0.0 |1.00x |1.38x |1.37x |1.58x |
** MJIT benchmarks
Benchmark: https://github.com/benchmark-driver/mjit-benchmarks
(Original: https://github.com/vnmakarov/ruby/tree/rtl_mjit_branch/MJIT-benchmarks)
| |2.0.0-p0 |r62186 |JIT off |JIT on |
|:----------|:--------|:--------|:--------|:--------|
|aread |1.00 |1.09 |1.07 |2.19 |
|aref |1.00 |1.13 |1.11 |2.22 |
|aset |1.00 |1.50 |1.45 |2.64 |
|awrite |1.00 |1.17 |1.13 |2.20 |
|call |1.00 |1.29 |1.26 |2.02 |
|const2 |1.00 |1.10 |1.10 |2.19 |
|const |1.00 |1.11 |1.10 |2.19 |
|fannk |1.00 |1.04 |1.02 |1.00 |
|fib |1.00 |1.32 |1.31 |1.84 |
|ivread |1.00 |1.13 |1.12 |2.43 |
|ivwrite |1.00 |1.23 |1.21 |2.40 |
|mandelbrot |1.00 |1.13 |1.16 |1.28 |
|meteor |1.00 |2.97 |2.92 |3.17 |
|nbody |1.00 |1.17 |1.15 |1.49 |
|nest-ntimes|1.00 |1.22 |1.20 |1.39 |
|nest-while |1.00 |1.10 |1.10 |1.37 |
|norm |1.00 |1.18 |1.16 |1.24 |
|nsvb |1.00 |1.16 |1.16 |1.17 |
|red-black |1.00 |1.02 |0.99 |1.12 |
|sieve |1.00 |1.30 |1.28 |1.62 |
|trees |1.00 |1.14 |1.13 |1.19 |
|while |1.00 |1.12 |1.11 |2.41 |
** Discourse's script/bench.rb
Benchmark: https://github.com/discourse/discourse/blob/v1.8.7/script/bench.rb
NOTE: Rails performance was somehow a little degraded with JIT for now.
We should fix this.
(At least I know opt_aref is performing badly in JIT and I have an idea
to fix it. Please wait for the fix.)
*** JIT off
Your Results: (note for timings- percentile is first, duration is second in millisecs)
categories_admin:
50: 17
75: 18
90: 22
99: 29
home_admin:
50: 21
75: 21
90: 27
99: 40
topic_admin:
50: 17
75: 18
90: 22
99: 32
categories:
50: 35
75: 41
90: 43
99: 77
home:
50: 39
75: 46
90: 49
99: 95
topic:
50: 46
75: 52
90: 56
99: 101
*** JIT on
Your Results: (note for timings- percentile is first, duration is second in millisecs)
categories_admin:
50: 19
75: 21
90: 25
99: 33
home_admin:
50: 24
75: 26
90: 30
99: 35
topic_admin:
50: 19
75: 20
90: 25
99: 30
categories:
50: 40
75: 44
90: 48
99: 76
home:
50: 42
75: 48
90: 51
99: 89
topic:
50: 49
75: 55
90: 58
99: 99
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@62197 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
This is redundant since r15206 /
ffe425ecaa as we no longer lock
the array during sort. Instead, fall back to Object#frozen?
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@62154 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* array.c (rb_check_to_array): conversion to array by to_a method.
returns nil if not possible.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@62072 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
r61827, r61947 was about to fix this. The proper way to allocate
memory region is called ALLOCV_N.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@61949 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
r61827 was about to fix this. The proper way to allocate memory
region is called ALLOCV_N.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@61947 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
It seems HUGE_VAL is already used. Why not eliminate INTINITY.
NAN is also float. That of double is called nan(). This is also
fixed.
Signed-off-by: Urabe, Shyouhei <shyouhei@ruby-lang.org>
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@61938 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Confusion of argument order ignores this st_free_table.
Results in garbaged table not GCed.
Easily noticable when you read the compiper warnings.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@61561 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
When doing the difference of a small array with a big one it is not
efficient in both time and memory to convert the second one to a hash.
From: Ana María Martínez Gómez <ammartinez@suse.de>
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@61330 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
and specialized Array#any? and Hash#any?
Based on patch by D.E. Akers [#11286]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@61098 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* array.c (rb_to_array_type): make public to share common code
internally.
* hash.c (rb_to_hash_type): make public to share common code
internally.
* symbol.c (rb_to_symbol_type): make public to share common code
internally.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@60438 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* vm_insnhelper.c (vm_opt_aref): optimize on other than fixnum
argument too.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@60333 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
[Feature #13884]
Reduce number of memory allocations for "and", "or" and "diff"
operations on small arrays
Very often, arrays are used to filter parameters and to select
interesting items from 2 collections and very often these
collections are small enough, for example:
```ruby
SAFE_COLUMNS = [:id, :title, :created_at]
def columns
@all_columns & SAFE_COLUMNS
end
```
In this patch, I got rid of unnecessary memory allocations for
small arrays when "and", "or" and "diff" operations are performed.
name | HEAD | PATCH
-----------------+------:+------:
array_small_and | 0.615 | 0.263
array_small_diff | 0.676 | 0.282
array_small_or | 0.953 | 0.463
name | PATCH
-----------------+------:
array_small_and | 2.343
array_small_diff | 2.392
array_small_or | 2.056
name | HEAD | PATCH
-----------------+------:+------:
array_small_and | 1.429 | 1.005
array_small_diff | 1.493 | 0.878
array_small_or | 1.672 | 1.152
name | PATCH
-----------------+------:
array_small_and | 1.422
array_small_diff | 1.700
array_small_or | 1.452
Author: Dmitry Bochkarev <dimabochkarev@gmail.com>
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@60057 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* array.c (binomial_coefficient): get rid of bignums by division
after each multiplications.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@59692 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* array.c (rb_ary_cycle_size, descending_factorial): use
rb_int_mul instead of rb_funcallv.
* array.c (binomial_coefficient): use rb_int_idiv instead of
rb_funcallv.
* array.c (rb_ary_repeated_permutation_size): use
rb_int_positive_pow.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@59689 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* array.c (ary_join_1): ignore encodings in nested arrays as an
initial encoding.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@59685 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* array.c (ary_join_1): copy the encoding of the converted string
of the first element by to_str too, as an initial encoding.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@59684 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
This patch is sent from @selmertsx (morioka shuhei).
[fix GH-1646]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@59070 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* array.c: [DOC] make example in the docs for Array#{sort,sort!}
match the call-seq and description by using the same block vars.
Based on a patch by Roque Pinel (repinel). [Fix GH-1628]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@59011 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
To convert the object implicitly, it has had two parts in convert_type() which are
1. lookink up the method's id
2. calling the method
Seems that strncmp() and strcmp() in convert_type() are slightly heavy to look up
the method's id for type conversion.
This patch will add and use internal APIs (rb_convert_type_with_id, rb_check_convert_type_with_id)
to call the method without looking up the method's id when convert the object.
Array#flatten -> 19 % up
Array#+ -> 3 % up
[ruby-dev:50024] [Bug #13341] [Fix GH-1537]
### Before
Array#flatten 104.119k (± 1.1%) i/s - 525.690k in 5.049517s
Array#+ 1.993M (± 1.8%) i/s - 10.010M in 5.024258s
### After
Array#flatten 124.005k (± 1.0%) i/s - 624.240k in 5.034477s
Array#+ 2.058M (± 4.8%) i/s - 10.302M in 5.019328s
### Test Code
require 'benchmark/ips'
class Foo
def to_ary
[1,2,3]
end
end
Benchmark.ips do |x|
ary = []
100.times { |i| ary << i }
array = [ary]
x.report "Array#flatten" do |i|
i.times { array.flatten }
end
x.report "Array#+" do |i|
obj = Foo.new
i.times { array + obj }
end
end
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@58978 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* array.c (rb_ary_concat_multi): concatenate the array without generating
temporary Array object if only one argument is given.
This is very similar with r58886.
Array#concat will be faster around 19%.
[Fix GH-1634]
### Before
Array#concat 2.187M (± 3.5%) i/s - 10.926M in 5.002829s
### After
Array#concat 2.598M (± 1.8%) i/s - 13.008M in 5.008201s
### Test code
require 'benchmark/ips'
Benchmark.ips do |x|
x.report "Array#concat" do |i|
other = [4]
i.times { [1, 2, 3].concat(other) }
end
end
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@58909 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* object.c (rb_equal): add optimized path to compare the objects using
rb_equal_opt(). Previously, if not same objects were given, rb_equal() would
call `==' method via rb_funcall() which took a long time.
rb_equal_opt() has provided faster comparing for Fixnum/Float/String objects.
Now, Time#eql? uses rb_equal() to compare with argument object and it will
be faster around 40% on 64-bit environment.
* array.c (rb_ary_index): remove redundant rb_equal_opt() calling.
Now, rb_equal() was optimized using rb_equal_opt().
If rb_equal_opt() returns Qundef, it will invoke rb_equal() -> rb_equal_opt(),
and it will cause the performance regression.
So, this patch will remove first redundant rb_equal_opt() calling.
* array.c (rb_ary_rindex): ditto.
* array.c (rb_ary_includes): ditto.
[ruby-core:80360] [Bug #13365] [Fix GH-#1552]
### Before
Time#eql? with other 7.309M (± 1.4%) i/s - 36.647M in 5.014964s
Array#index(val) 1.433M (± 1.2%) i/s - 7.207M in 5.030942s
Array#rindex(val) 1.418M (± 1.6%) i/s - 7.103M in 5.009164s
Array#include?(val) 1.451M (± 0.9%) i/s - 7.295M in 5.026392s
### After
Time#eql? with other 10.321M (± 1.9%) i/s - 51.684M in 5.009203s
Array#index(val) 1.474M (± 0.9%) i/s - 7.433M in 5.044384s
Array#rindex(val) 1.449M (± 1.7%) i/s - 7.292M in 5.034436s
Array#include?(val) 1.466M (± 1.7%) i/s - 7.373M in 5.030047s
### Test code
require 'benchmark/ips'
Benchmark.ips do |x|
t1 = Time.now
t2 = Time.now
x.report "Time#eql? with other" do |i|
i.times { t1.eql?(t2) }
end
# Benchmarks to check whether it didn't introduce the regression
obj = Object.new
x.report "Array#index(val)" do |i|
ary = [1, 2, true, false, obj]
i.times { ary.index(obj) }
end
x.report "Array#rindex(val)" do |i|
ary = [1, 2, true, false, obj].reverse
i.times { ary.rindex(obj) }
end
x.report "Array#include?(val)" do |i|
ary = [1, 2, true, false, obj]
i.times { ary.include?(obj) }
end
end
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@58880 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* array.c (ary_ensure_room_for_push): use rb_ary_modify_check() instead of
rb_ary_modify() to check whether the object can be modified for non shared
Array object. rb_ary_modify() has the codes for shared Array object too.
In here, it has condition branch for shared / non shared Array object and
it can use rb_ary_modify_check() which is smaller function than
rb_ary_modify() for non shared object.
rb_ary_modify_check() will be expand as inline function.
If it will compile with GCC, Array#<< will be faster around 8%.
[ruby-core:81082] [Bug #13553] [Fix GH-1609]
## Clang 802.0.42
### Before
Array#<< 9.353M (± 1.7%) i/s - 46.787M in 5.004123s
Array#push 7.702M (± 1.1%) i/s - 38.577M in 5.009338s
Array#values_at 6.133M (± 1.9%) i/s - 30.699M in 5.007772s
### After
Array#<< 9.458M (± 2.0%) i/s - 47.357M in 5.009069s
Array#push 7.921M (± 1.8%) i/s - 39.665M in 5.009151s
Array#values_at 6.377M (± 2.3%) i/s - 31.881M in 5.001888s
### Result
Array#<< -> 1.2% faster
Array#push -> 2.8% faster
Array#values_at -> 3.9% faster
## GCC 7.1.0
### Before
Array#<< 10.497M (± 1.1%) i/s - 52.665M in 5.017601s
Array#push 8.527M (± 1.6%) i/s - 42.777M in 5.018003s
Array#values_at 7.621M (± 1.7%) i/s - 38.152M in 5.007910s
### After
Array#<< 11.403M (± 1.3%) i/s - 57.028M in 5.001849s
Array#push 8.924M (± 1.3%) i/s - 44.609M in 4.999940s
Array#values_at 8.291M (± 1.4%) i/s - 41.487M in 5.004727s
### Result
Array#<< -> 8.3% faster
Array#push -> 4.3% faster
Array#values_at -> 8.7% faster
## Test code
require 'benchmark/ips'
Benchmark.ips do |x|
x.report "Array#<<" do |i|
i.times { [1,2] << 3 }
end
x.report "Array#push" do |i|
i.times { [1,2].push(3) }
end
x.report "Array#values_at" do |i|
ary = [1, 2, 3, 4, 5]
i.times { ary.values_at(0, 2, 4) }
end
end
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@58867 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* debug_counter.h: add the following counters to measure object types.
obj_free: freed count
obj_str_ptr: freed count of Strings they have extra buff.
obj_str_embed: freed count of Strings they don't have extra buff.
obj_str_shared: freed count of Strings they have shared extra buff.
obj_str_nofree: freed count of Strings they are marked as nofree.
obj_str_fstr: freed count of Strings they are marked as fstr.
obj_ary_ptr: freed count of Arrays they have extra buff.
obj_ary_embed: freed count of Arrays they don't have extra buff.
obj_obj_ptr: freed count of Objects (T_OBJECT) they have extra buff.
obj_obj_embed: freed count of Objects they don't have extra buff.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@58865 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* array.c (rb_ary_insert): fix the position in error message, when
it is less than -1.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@58694 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* array.c (rb_ary_insert): check position to insert even if no
elements to be inserted. [ruby-core:81125] [Bug #13558]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@58693 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Found a part where copy&paste can be eliminated. Reduces vm_exec_core
from 26,228 bytes to 26,176 bytes in size on my machine. I believe it
does not affect any runtime performance.
----
* array.c (rb_ary_tmp_new_from_values): extend existing
rb_ary_new_from_values function so that it can take
additional value for klass.
* array.c (rb_ary_new_from_values): use the new function.
* insns.def (toregexp): ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@58416 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
For a shared array root, struct RArray::as.heap.aux.capa stores the
number of Arrays holding reference to that T_ARRAY instead of the actual
heap-allocated capacity. Use ARY_CAPA() macro which handles this
appropriately.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@58332 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* array.c: [DOC] add example for Array.new with block and index.
Reported by Don Cruickshank. [ruby-core:68442] [Bug #10944]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@58037 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* array.c: [DOC] fix grammar in Array#sort, #sort!, #sort_by!,
move references below the code example, add a missing reference.
* enum.c: [DOC] fix grammar in Enumerable#sort, #sort_by.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@57771 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* array.c (finish_exact_sum): add 0 and the initial value to check
if the latter is numeric. [ruby-core:79572] [Bug #13222]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@57651 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
normal
shyouhei
nobu
eregon
k0kubun
kazu
marcandre
glass
duerst
stomar
watson1978
ko1
mrkn
rhe