Rubinius vs. CRuby

Comparison between Rubinius and CRuby held at LRUG 12 om 2014/12/18

Transcript

1. Rubinius vs. CRuby Michał Gutowski
2. Agenda • What is Rubinius? • How does it work? • Benchmarks
3. What is Rubinius? • Interpreter + ruby virtual machine based on LLVM • Written mostly in ruby • Has C-API for running C extensions • Implementation of Ruby which uses native threads
4. How does it work?
5. Parser stage • calls Melbourne parser, the same as in CRuby • using ruby code, creates AST trees
6. Abstract Syntax Trees • Trees created to execute program in proper order • Created recursively up the the level of primitives
7. Generator Stage • Generates byte code from an AST • Creates main Generator object and child Generators for block, methods etc. • Maps directly to Rubinius Instructions like push_int(number), goto_if_true(location) etc.
8. Encoder Stage • Creates instruction sequence - all instructions are put in proper place • Validates stack for underflow, overflow and so on
9. Packager Stage • During this stage Rubinius creates CompiledCode object • CompiledCode is an ruby object, which contains byte code created during encoder stage, variable count, variable names, arguments etc.
10. Writer Stage • writes method to file for future use • creates CompiledFile ruby object • returns input
11. Conclusion • Rubinius has modular structure • Every phase can be customized • Which can be done much easier thanks to modular structure
12. Benchmarks • https://github.com/acangiano/ruby-benchmark- suite • benchmarks were handles on mid-2010 macbook pro • Core 2 Duo 2,66 Ghz, 6 GB ram
13. bm_count_multithreaded.rb def count_high 100_000.times {} end Bench.run [1, 2, 4, 8, 16] do |n| threads = [] n.times{ threads << Thread.new { count_high }} threads.each{|t| t.join} end
14. bm_count_multithreaded.rb
15. bm_count_shared_thread.rb def count_high how_many how_many.times {} end Bench.run [1, 2, 4, 8, 16] do |n| threads = [] n.times{ threads << Thread.new { count_high(1_000_000 / n) }} threads.each{|t| t.join} end
16. bm_count_shared_thread.rb
17. bm_hilbert_matrix.rb def hilbert(dimension) rows = Array.new (1..dimension).each do |i| row = Array.new (1..dimension).each do |j| row.push(1/(i + j - 1)) end rows.push(row) end return(Matrix.rows(rows)) end Hilbert Matrix of size 5
18. bm_hilbert_matrix.rb
19. bm_dirp.rb Bench.run [10000] do |n| n.times do dirp = Dir.open(".") for f in dirp case f when /^\./, /~$/, /\.o/ # do not print else # don't print since we don't care about print speed #print f, "\n" end end dirp.close end end
20. bm_dirp.rb
21. bm_eval.rb Bench.run [1000000] do |n| n.times { eval "a = 33" } end
22. bm_eval.rb
23. bm_so_lists.rb def test_lists(size) li1 = (1..size).to_a li2 = li1.dup li3 = Array.new while (not li2.empty?) li3.push(li2.shift) end while (not li3.empty?) li2.push(li3.pop) end li1.reverse! if li1[0] != size then p "not size" 0 else if li1 != li2 then return(0) else li1.length end end end
24. bm_so_lists.rb
25. bm_gc_mb.rb Bench.run [500_000, 1_000_000, 3_000_000] do |n| a = [] n.times { a << []} # use up some RAM n.times {[]} end
26. bm_gc_mb.rb
27. Thanks for your attention