- ruby-core - ml.ruby-lang.org

[ruby-core:111009] [Ruby master Misc#19149] Minimal covered tests with the --enable-yjit case?
by jaruga (Jun Aruga) 25 Nov '22

25 Nov '22

Issue #19149 has been updated by jaruga (Jun Aruga). > > This will run YJIT with its default options, which is what most people would use when running YJIT. If you specify `--yjit-call-threshold=1 --yjit-verify-ctx`, it will make YJIT compile every single method and run extra verifications. This is what we do on the CI to be extra thorough, but it will be slower. Checking the commit <https://github.com/ruby/ruby/commit/f90549cd38518231a6a74432fe1168c943a7cc18> about YJIT, and reading a document about [Ruby JIT's document](https://developer.squareup.com/blog/rubys-new-jit/), the main feature of the YJIT is to create iseq (Instruction Sequence) uniquely for YJIT, and compile it into YARV code, right? The top main functions are below? * `rb_yjit_iseq_mark` at `iseq.c` * `rb_yjit_compile_iseq` at `vm.c` Running the command below with the Ruby above, the `compiled_iseq_count` was `0`, the `rb_yjit_iseq_mark` and `rb_yjit_compile_iseq` were not executed. ``` $ ruby --yjit --yjit-stats -e 'puts "a"' a ***YJIT: Printing YJIT statistics on exit*** method call exit reasons: (all relevant counters are zero) invokeblock exit reasons: (all relevant counters are zero) invokesuper exit reasons: (all relevant counters are zero) leave exit reasons: (all relevant counters are zero) getblockparamproxy exit reasons: (all relevant counters are zero) getinstancevariable exit reasons: (all relevant counters are zero) setinstancevariable exit reasons: (all relevant counters are zero) opt_aref exit reasons: (all relevant counters are zero) expandarray exit reasons: (all relevant counters are zero) opt_getinlinecache exit reasons: (all relevant counters are zero) invalidation reasons: (all relevant counters are zero) bindings_allocations: 73 bindings_set: 0 compiled_iseq_count: 0 compiled_block_count: 0 compiled_branch_count: 0 freed_iseq_count: 0 invalidation_count: 0 constant_state_bumps: 0 inline_code_size: 0 outlined_code_size: 84 freed_code_size: 0 code_region_size: 12288 yjit_alloc_size: 1763 live_page_count: 1 freed_page_count: 0 code_gc_count: 0 num_gc_obj_refs: 0 object_shape_count: 236 side_exit_count: 0 total_exit_count: 0 total_insns_count: 236517 vm_insns_count: 236517 yjit_insns_count: 0 ratio_in_yjit: 0.0% avg_len_in_yjit: NaN total_exits: 0 ``` Running the command below, the 2 functions above were executed. The `compiled_iseq_count` was `4` non-zero. I feel that the `--yjit-call-threshold=1` is necessary to avoid skipping the functions unintentionally in the tests. ``` $ ruby --yjit --yjit-stats --yjit-call-threshold=1 -e 'puts "a"' a ***YJIT: Printing YJIT statistics on exit*** method call exit reasons: (all relevant counters are zero) invokeblock exit reasons: (all relevant counters are zero) invokesuper exit reasons: (all relevant counters are zero) leave exit reasons: interp_return 1 (100.0%) getblockparamproxy exit reasons: (all relevant counters are zero) getinstancevariable exit reasons: (all relevant counters are zero) setinstancevariable exit reasons: (all relevant counters are zero) opt_aref exit reasons: (all relevant counters are zero) expandarray exit reasons: (all relevant counters are zero) opt_getinlinecache exit reasons: (all relevant counters are zero) invalidation reasons: (all relevant counters are zero) bindings_allocations: 73 bindings_set: 0 compiled_iseq_count: 4 compiled_block_count: 8 compiled_branch_count: 12 freed_iseq_count: 0 invalidation_count: 0 constant_state_bumps: 0 inline_code_size: 1074 outlined_code_size: 603 freed_code_size: 0 code_region_size: 12288 yjit_alloc_size: 21097 live_page_count: 1 freed_page_count: 0 code_gc_count: 0 num_gc_obj_refs: 9 object_shape_count: 236 side_exit_count: 0 total_exit_count: 1 total_insns_count: 236325 vm_insns_count: 236313 yjit_insns_count: 12 ratio_in_yjit: 0.0% avg_len_in_yjit: 12.0 total_exits: 0 ``` And it seems that the `--yjit-verify-ctx` option is defined at <https://github.com/ruby/ruby/blob/e15cd01149afe4924460f81cb6e27dd96de06657/…>. However the option is not defined at <https://github.com/ruby/ruby/blob/e15cd01149afe4924460f81cb6e27dd96de06657/…>. Is it intentionally hidden for users? in the `--help` document? ``` $ ruby -v ruby 3.2.0dev (2022-11-24T06:13:00Z master 66e5200ba4) [x86_64-linux] $ ruby --yjit-call-threshold=1 --yjit-verify-ctx -e 'puts "a"' a $ ruby --help | grep 'yjit' --jit enable JIT for the platform, same as --yjit (experimental) --yjit enable in-process JIT compiler (experimental) yjit in-process JIT compiler (default: disabled) --yjit-stats Enable collecting YJIT statistics --yjit-exec-mem-size=num --yjit-call-threshold=num --yjit-max-versions=num --yjit-greedy-versioning ``` ---------------------------------------- Misc #19149: Minimal covered tests with the --enable-yjit case? https://bugs.ruby-lang.org/issues/19149#change-100259 * Author: jaruga (Jun Aruga) * Status: Open * Priority: Normal ---------------------------------------- In the [Fedora Ruby's RPM recipe file](https://src.fedoraproject.org/rpms/ruby/blob/rawhide/f/ruby.spec), we were running the commands below. ``` $ ./autogen.sh $ ./configure ... $ make $ make check ``` ## What is the minimal covered tests? Now the we want to add the `--enable-yjit` option. Could you tell me what is the minimal covered test commands in the `--enable-yjit` case? ``` $ ./autogen.sh $ ./configure --enable-yjit ... $ make $ ?? ``` Then do we need to run the `make check` 2 times with both with yjit and without yjit as follows? The yjit command options `--yjit-call-threshold=1 --yjit-verify-ctx` comes from the <https://github.com/ruby/ruby/blob/d2fa67de81f66cb42cfeebc81a03c57a4621c09a/…>. ``` $ make check $ make check RUN_OPTS="--yjit-call-threshold=1 --yjit-verify-ctx" ``` Or is it good enough to run the `make check` and the specific tests with the yjit options as follows? ``` $ make check $ make test-all RUN_OPTS="--yjit-call-threshold=1 --yjit-verify-ctx" TESTS="test/ruby/{test_yjit_exit_locations.rb,test_yjit.rb}" ``` Or is it good enough to run the `make check` with the YJIT options as follows? ``` $ make check RUN_OPTS="--yjit-call-threshold=1 --yjit-verify-ctx" ``` ## YJIT command options Could you explain why the command options `--yjit-call-threshold=1 --yjit-verify-ctx` above is better to test the YJIT cases rather than just `--yjit`? ## Ideal situation I want to see the just running `make check` covers necessary cases in YJIT. Because it is convenience, and I think users tend to be satisfied with only running the `make check`. What do you think? ``` $ ./autogen.sh $ ./configure --enable-yjit ... $ make $ make check ``` I tried to inject the YJIT command options in a test file for that. Perhaps it might be like this. But so far I am not succeeded. ``` diff --git a/test/lib/jit_support.rb b/test/lib/jit_support.rb index 26f8542dc2..3fce402e32 100644 --- a/test/lib/jit_support.rb +++ b/test/lib/jit_support.rb @@ -69,8 +69,10 @@ def supported? end def yjit_supported? + return @yjit_supported if defined?(@yjit_supported) # e.g. x86_64-linux, x64-mswin64_140, x64-mingw32, x64-mingw-ucrt - RUBY_PLATFORM.match?(/^(x86_64|x64|arm64|aarch64)-/) + @yjit_supported = RbConfig::CONFIG["YJIT_SUPPORT"] != 'no' && + RUBY_PLATFORM.match?(/^(x86_64|x64|arm64|aarch64)-/) end def remove_mjit_logs(stderr) diff --git a/test/ruby/test_yjit.rb b/test/ruby/test_yjit.rb index 9ab058d97b..10c8e3b891 100644 --- a/test/ruby/test_yjit.rb +++ b/test/ruby/test_yjit.rb @@ -8,7 +8,7 @@ require 'tmpdir' require_relative '../lib/jit_support' -return unless defined?(RubyVM::YJIT) && RubyVM::YJIT.enabled? +return unless JITSupport.yjit_supported? # Tests for YJIT with assertions on compilation and side exits # insipired by the MJIT tests in test/ruby/test_mjit.rb ``` -- https://bugs.ruby-lang.org/

1 0

[ruby-core:111008] [Ruby master Misc#19149] Minimal covered tests with the --enable-yjit case?
by jaruga (Jun Aruga) 25 Nov '22

25 Nov '22

Issue #19149 has been updated by jaruga (Jun Aruga). maximecb (Maxime Chevalier-Boisvert) wrote in #note-1: > If you want to thoroughly test everything, you could run make check two times, once with YJIT, and once without. > > If you just run: > ``` > $ make check RUN_OPTS="--yjit" > ``` > > This will run YJIT with its default options, which is what most people would use when running YJIT. If you specify `--yjit-call-threshold=1 --yjit-verify-ctx`, it will make YJIT compile every single method and run extra verifications. This is what we do on the CI to be extra thorough, but it will be slower. > > If you just want to quickly verify that YJIT works on the host platform, you could just run `make btest` with `--yjit` for some quick minimal verification. We do already test YJIT pretty well on the GitHub CI and on `rubyci.org`. Thank you for clarifying it. I understood it. I am still trying to find a good balance of running the YJIT tests between the running time and the coverage. In the case of "MJIT", while the MJIT is supported as a default. And we don't run the tests with RUN_OPTS="--mjit" in the Feodra Ruby's RPM recipe file, while I notice there are tests with `RUN_OPTS="--mjit"` now at <https://github.com/ruby/ruby/blob/4ab89d57bbc569143e9833addb88b91db86ad057/…>. Because we can run the the following MJIT tests within the `make check` without `RUN_OPTS="--mjit"`. Here are the test results on the yesterday's latest master `66e5200ba435361624caa3e23db7962d906b70db`. ``` $ grep MJIT_SUPPORT rbconfig.rb CONFIG["MJIT_SUPPORT"] = "yes" $ make test-all TESTS=test/ruby/test_mjit.rb revision.h updated Run options: --seed=16704 "--ruby=./miniruby -I./lib -I. -I.ext/common ./tool/runruby.rb --extout=.ext -- --disable-gems" --excludes-dir=./test/excludes --name=!/memory_leak/ # Running tests: Finished tests in 132.081114s, 0.7950 tests/s, 4.2020 assertions/s. 105 tests, 555 assertions, 0 failures, 0 errors, 4 skips ruby -v: ruby 3.2.0dev (2022-11-24T06:13:00Z master 66e5200ba4) [x86_64-linux] ``` However, we cannot run the following YJIT tests without `RUN_OPTS="--yjit"`. The tests were skipped. ``` $ grep YJIT_SUPPORT rbconfig.rb CONFIG["YJIT_SUPPORT"] = "dev" $ make test-all TESTS=test/ruby/test_yjit.rb revision.h updated Run options: --seed=38677 "--ruby=./miniruby -I./lib -I. -I.ext/common ./tool/runruby.rb --extout=.ext -- --disable-gems" --excludes-dir=./test/excludes --name=!/memory_leak/ # Running tests: Finished tests in 0.017483s, 0.0000 tests/s, 0.0000 assertions/s. 0 tests, 0 assertions, 0 failures, 0 errors, 0 skips ruby -v: ruby 3.2.0dev (2022-11-24T06:13:00Z master 66e5200ba4) [x86_64-linux] ``` What does the difference of the ideas come from? ---------------------------------------- Misc #19149: Minimal covered tests with the --enable-yjit case? https://bugs.ruby-lang.org/issues/19149#change-100258 * Author: jaruga (Jun Aruga) * Status: Open * Priority: Normal ---------------------------------------- In the [Fedora Ruby's RPM recipe file](https://src.fedoraproject.org/rpms/ruby/blob/rawhide/f/ruby.spec), we were running the commands below. ``` $ ./autogen.sh $ ./configure ... $ make $ make check ``` ## What is the minimal covered tests? Now the we want to add the `--enable-yjit` option. Could you tell me what is the minimal covered test commands in the `--enable-yjit` case? ``` $ ./autogen.sh $ ./configure --enable-yjit ... $ make $ ?? ``` Then do we need to run the `make check` 2 times with both with yjit and without yjit as follows? The yjit command options `--yjit-call-threshold=1 --yjit-verify-ctx` comes from the <https://github.com/ruby/ruby/blob/d2fa67de81f66cb42cfeebc81a03c57a4621c09a/…>. ``` $ make check $ make check RUN_OPTS="--yjit-call-threshold=1 --yjit-verify-ctx" ``` Or is it good enough to run the `make check` and the specific tests with the yjit options as follows? ``` $ make check $ make test-all RUN_OPTS="--yjit-call-threshold=1 --yjit-verify-ctx" TESTS="test/ruby/{test_yjit_exit_locations.rb,test_yjit.rb}" ``` Or is it good enough to run the `make check` with the YJIT options as follows? ``` $ make check RUN_OPTS="--yjit-call-threshold=1 --yjit-verify-ctx" ``` ## YJIT command options Could you explain why the command options `--yjit-call-threshold=1 --yjit-verify-ctx` above is better to test the YJIT cases rather than just `--yjit`? ## Ideal situation I want to see the just running `make check` covers necessary cases in YJIT. Because it is convenience, and I think users tend to be satisfied with only running the `make check`. What do you think? ``` $ ./autogen.sh $ ./configure --enable-yjit ... $ make $ make check ``` I tried to inject the YJIT command options in a test file for that. Perhaps it might be like this. But so far I am not succeeded. ``` diff --git a/test/lib/jit_support.rb b/test/lib/jit_support.rb index 26f8542dc2..3fce402e32 100644 --- a/test/lib/jit_support.rb +++ b/test/lib/jit_support.rb @@ -69,8 +69,10 @@ def supported? end def yjit_supported? + return @yjit_supported if defined?(@yjit_supported) # e.g. x86_64-linux, x64-mswin64_140, x64-mingw32, x64-mingw-ucrt - RUBY_PLATFORM.match?(/^(x86_64|x64|arm64|aarch64)-/) + @yjit_supported = RbConfig::CONFIG["YJIT_SUPPORT"] != 'no' && + RUBY_PLATFORM.match?(/^(x86_64|x64|arm64|aarch64)-/) end def remove_mjit_logs(stderr) diff --git a/test/ruby/test_yjit.rb b/test/ruby/test_yjit.rb index 9ab058d97b..10c8e3b891 100644 --- a/test/ruby/test_yjit.rb +++ b/test/ruby/test_yjit.rb @@ -8,7 +8,7 @@ require 'tmpdir' require_relative '../lib/jit_support' -return unless defined?(RubyVM::YJIT) && RubyVM::YJIT.enabled? +return unless JITSupport.yjit_supported? # Tests for YJIT with assertions on compilation and side exits # insipired by the MJIT tests in test/ruby/test_mjit.rb ``` -- https://bugs.ruby-lang.org/

1 0

[ruby-core:111007] [Ruby master Misc#19122] Use MADV_DONTNEED instead of MADV_FREE when freeing a Fiber's stack
by smcgivern (Sean McGivern) 25 Nov '22

25 Nov '22

Issue #19122 has been updated by smcgivern (Sean McGivern). ioquatix (Samuel Williams) wrote in #note-2: > I don't have a strong opinion about this, but I'm generally against loosing performance. > > Maybe it can be controlled using environment variable. Having this user-controllable works for me, although I don't know about the Ruby project's general stance on toggles like this. @ioquatix is https://github.com/socketry/falcon-benchmark the right test suite to run here? I'm not sure if that's what you ran in https://bugs.ruby-lang.org/issues/15997#note-19, or if that was slightly different. I'm happy to try this out with the various options here to quantify the current state - just point me in the right direction :-) ---------------------------------------- Misc #19122: Use MADV_DONTNEED instead of MADV_FREE when freeing a Fiber's stack https://bugs.ruby-lang.org/issues/19122#change-100257 * Author: smcgivern (Sean McGivern) * Status: Open * Priority: Normal * Assignee: ioquatix (Samuel Williams) ---------------------------------------- I'd like to propose that Ruby stops using MADV_FREE when freeing a Fiber's stack, and switches to using MADV_DONTNEED even when MADV_FREE is supported. MADV_FREE is used in one place in the Ruby codebase, when freeing the stack of a freed Fiber: https://git.ruby-lang.org/ruby.git/tree/cont.c#n683 The comment for `fiber_pool_stack_free` says: ```c // We advise the operating system that the stack memory pages are no longer being used. // This introduce some performance overhead but allows system to relaim memory when there is pressure. ``` Where possible (i.e. on Linux 4.5 and later), `fiber_pool_stack_free` uses `MADV_FREE` over `MADV_DONTNEED`. This has the side effect that memory statistics such as RSS will not reduce until and unless the OS actually reclaims that memory. If that doesn't happen, then the reported memory usage via RSS will be much higher than the 'real' memory usage. If this was pervasive throughtout the Ruby codebase then that would be one thing, but currently this is just for Fiber. This means that: 1. A program that doesn't use Fiber will have somewhat reliable RSS statistics on recent Linux. 2. A program that heavily uses Fiber (such as something using Async::HTTP) will see an inflated RSS statistic. Go made a similar change to the one I'm proposing here for similar reasons: https://github.com/golang/go/issues/42330 > While `MADV_FREE` is somewhat faster than `MADV_DONTNEED`, it doesn't affect many of the statistics that `MADV_DONTNEED` does until the memory is actually reclaimed. This generally leads to poor user experience, like confusing stats in `top` and other monitoring tools; and bad integration with management systems that respond to memory usage. > [...] > I propose we change the default to prefer `MADV_DONTNEED` over `MADV_FREE`, to favor user-friendliness and minimal surprise over performance. I think it's become clear that Linux's implementation of `MADV_FREE` ultimately doesn't meet our needs. As an aside, MADV_FREE was not used in Ruby 3.1 (https://bugs.ruby-lang.org/issues/19101) and I haven't found any bugs filed about this behaviour other than that one. -- https://bugs.ruby-lang.org/

1 0

[ruby-core:111006] Test for footer
by Shugo Maeda 25 Nov '22

25 Nov '22

This is a test mail for footer change.

1 0

[ruby-core:111005] [Ruby master Misc#19030] [ANN] Migrate lists.ruby-lang.org to Google Groups
by hsbt (Hiroshi SHIBATA) 25 Nov '22

25 Nov '22

Issue #19030 has been updated by hsbt (Hiroshi SHIBATA). I migrated `ruby-core` list to mailmanlist and confirm to work redmine integration. Sorry to send debug message for this. ---------------------------------------- Misc #19030: [ANN] Migrate lists.ruby-lang.org to Google Groups https://bugs.ruby-lang.org/issues/19030#change-100256 * Author: hsbt (Hiroshi SHIBATA) * Status: Assigned * Priority: Normal * Assignee: hsbt (Hiroshi SHIBATA) ---------------------------------------- Our mailing-list server that is `lists.ruby-lang.org` is too old. And it's difficult to replace new server on AWS because building mail-service on AWS has a lot of limitations. I and @shugo decided to migrate lists.ruby-lang.org to Google Groups. * In Nov-Dec 2022, we migrate the current list member to Google Groups of our google workspace. * I hope to migrate to the last list-id, But I'm not sure we can do that. * What will be used as an archive viewer has yet to be TBD status. * blade is still down. * I prefer plain text viewer like blade instead of google groups. Should we build it? I will update this plan in this thread. -- https://bugs.ruby-lang.org/

1 0

[ruby-core:111004] [Ruby master Misc#19030] [ANN] Migrate lists.ruby-lang.org to Google Groups
by hsbt (Hiroshi SHIBATA) 25 Nov '22

25 Nov '22

Issue #19030 has been updated by hsbt (Hiroshi SHIBATA). This is test comment. ---------------------------------------- Misc #19030: [ANN] Migrate lists.ruby-lang.org to Google Groups https://bugs.ruby-lang.org/issues/19030#change-100255 * Author: hsbt (Hiroshi SHIBATA) * Status: Assigned * Priority: Normal * Assignee: hsbt (Hiroshi SHIBATA) ---------------------------------------- Our mailing-list server that is `lists.ruby-lang.org` is too old. And it's difficult to replace new server on AWS because building mail-service on AWS has a lot of limitations. I and @shugo decided to migrate lists.ruby-lang.org to Google Groups. * In Nov-Dec 2022, we migrate the current list member to Google Groups of our google workspace. * I hope to migrate to the last list-id, But I'm not sure we can do that. * What will be used as an archive viewer has yet to be TBD status. * blade is still down. * I prefer plain text viewer like blade instead of google groups. Should we build it? I will update this plan in this thread. -- https://bugs.ruby-lang.org/

1 0

[ruby-core:111003] [Ruby master Misc#19030] [ANN] Migrate lists.ruby-lang.org to Google Groups
by hsbt (Hiroshi SHIBATA) 25 Nov '22

25 Nov '22

Issue #19030 has been updated by hsbt (Hiroshi SHIBATA). This is test comment. ---------------------------------------- Misc #19030: [ANN] Migrate lists.ruby-lang.org to Google Groups https://bugs.ruby-lang.org/issues/19030#change-100252 * Author: hsbt (Hiroshi SHIBATA) * Status: Assigned * Priority: Normal * Assignee: hsbt (Hiroshi SHIBATA) ---------------------------------------- Our mailing-list server that is `lists.ruby-lang.org` is too old. And it's difficult to replace new server on AWS because building mail-service on AWS has a lot of limitations. I and @shugo decided to migrate lists.ruby-lang.org to Google Groups. * In Nov-Dec 2022, we migrate the current list member to Google Groups of our google workspace. * I hope to migrate to the last list-id, But I'm not sure we can do that. * What will be used as an archive viewer has yet to be TBD status. * blade is still down. * I prefer plain text viewer like blade instead of google groups. Should we build it? I will update this plan in this thread. -- https://bugs.ruby-lang.org/

1 0

[ruby-core:111002] Please verify your email forwarding address
by Google Domains 25 Nov '22

25 Nov '22

Google Domains <https://notifications.google.com/g/p/ACnX6LYAMnarZcSkRbueggZyEDCNqwDZIMF59i…> ----- You recently added email forwarding from ruby-core(a)ruby-lang.org to ruby-core(a)ml.ruby-lang.org Verify my email now <https://notifications.google.com/g/p/ACnX6LZifLhMmrDpnyref5v_aM_mghPYmz-RBV…> ----- Please verify your email address to start getting emails sent to ruby-core(a)ml.ruby-lang.org. See you online, Google Domains Team ----- Have a question? Visit the Help Center <https://notifications.google.com/g/p/ACnX6LZs0U93fyyvb7jL1ffmtXPSGx-Satdsxl…> ----- Google LLC 1600 Amphitheatre Parkway, Mountain View, CA 94043 This email was sent to ruby-core(a)ml.ruby-lang.org to update you about important information regarding your Google Domains account. Questions, comments, and support for Google Domains are available <https://notifications.google.com/g/p/ACnX6La6oLJAcAu8Sydnb_jQp1ZZ7-ajNi-_G6…>here.

1 0

[ruby-core:111001] [Ruby master Misc#19030] [ANN] Migrate lists.ruby-lang.org to Google Groups
by hsbt (Hiroshi SHIBATA) 25 Nov '22

25 Nov '22

Issue #19030 has been updated by hsbt (Hiroshi SHIBATA). This is test post for mailmanlist. ---------------------------------------- Misc #19030: [ANN] Migrate lists.ruby-lang.org to Google Groups https://bugs.ruby-lang.org/issues/19030#change-100251 * Author: hsbt (Hiroshi SHIBATA) * Status: Assigned * Priority: Normal * Assignee: hsbt (Hiroshi SHIBATA) ---------------------------------------- Our mailing-list server that is `lists.ruby-lang.org` is too old. And it's difficult to replace new server on AWS because building mail-service on AWS has a lot of limitations. I and @shugo decided to migrate lists.ruby-lang.org to Google Groups. * In Nov-Dec 2022, we migrate the current list member to Google Groups of our google workspace. * I hope to migrate to the last list-id, But I'm not sure we can do that. * What will be used as an archive viewer has yet to be TBD status. * blade is still down. * I prefer plain text viewer like blade instead of google groups. Should we build it? I will update this plan in this thread. -- https://bugs.ruby-lang.org/

1 0

[ruby-core:00006] test number
by SHIBATA Hiroshi 24 Nov '22

24 Nov '22

test test -- SHIBATA Hiroshi shibata.hiroshi(a)gmail.com http://www.hsbt.org/

2 1