A Rails Developer’s Guide To The Ruby VM

RailsConf 2022

Transcript

1. A Rails Developer’s Guide to the Ruby VM Maple Ong, Shopify RailsConf 2022 ❤
2. Compiled vs. Interpreted
3. ❓❓❓
5. *some Ruby & Rails InfrastructureTeam members not depicted here #
6. The Outline so you know what to expect • The Ruby Execution Process • What’s a YARV? • How The Ruby VM Works • Why?? • Practical Uses to Rails Apps
7. Tokenizer Parser Compiler Your Ruby Code AST nodes Tokens Bytecode Virtual Machine
8. 1. Tokenize / Lexical Analysis
9. def foo(a, b) puts a + b end foo(9, 8)
10. foo_example = <<~HEREDOC def foo(a, b) puts a + b end foo(9, 8) HEREDOC puts Ripper.lex(foo_example)
11. [[[1, 0], :on_kw, "def", FNAME], [[1, 3], :on_sp, " ", FNAME], [[1, 4], :on_ident, "foo", ENDFN], [[1, 7], :on_lparen, "(", BEG|LABEL], [[1, 8], :on_ident, "a", ARG], [[1, 9], :on_comma, ",", BEG|LABEL], [[1, 10], :on_sp, " ", BEG|LABEL], [[1, 11], :on_ident, "b", ARG], [[1, 12], :on_rparen, ")", ENDFN], [[1, 13], :on_ignored_nl, "\n", BEG], [[2, 0], :on_sp, " ", BEG], [[2, 2], :on_ident, "puts", CMDARG], [[2, 6], :on_sp, " ", CMDARG], [[2, 7], :on_ident, "a", END|LABEL], [[2, 8], :on_sp, " ", END|LABEL], [[2, 9], :on_op, "+", BEG], [[2, 10], :on_sp, " ", BEG], [[2, 11], :on_ident, "b", END|LABEL], [[2, 12], :on_nl, "\n", BEG], [[3, 0], :on_kw, "end", END], [[3, 3], :on_nl, "\n", BEG], [[4, 0], :on_ignored_nl, "\n", BEG], [[5, 0], :on_ident, "foo", CMDARG], [[5, 3], :on_lparen, "(", BEG|LABEL], [[5, 4], :on_int, "9", END], [[5, 5], :on_comma, ",", BEG|LABEL], [[5, 6], :on_sp, " ", BEG|LABEL], [[5, 7], :on_int, "8", END], [[5, 8], :on_rparen, ")", ENDFN], [[5, 9], :on_nl, "\n", BEG]]
12. [[[1, 0], :on_kw, "def", FNAME], [[1, 3], :on_sp, " ", FNAME], [[1, 4], :on_ident, "foo", ENDFN], [[1, 7], :on_lparen, "(", BEG|LABEL], [[1, 8], :on_ident, "a", ARG], [[1, 9], :on_comma, ",", BEG|LABEL], [[1, 10], :on_sp, " ", BEG|LABEL], [[1, 11], :on_ident, "b", ARG], [[1, 12], :on_rparen, ")", ENDFN], [[1, 13], :on_ignored_nl, "\n", BEG], …
13. [[1, 0], :on_kw, "def", FNAME] Line Number Column Number Token Type Lexeme Token State
14. Tokenizer Your Ruby Code Tokens [[[1, 0], :on_kw, "def", FNAME], [[1, 3], :on_sp, " ", FNAME], [[1, 4], :on_ident, "foo", ENDFN], [[1, 7], :on_lparen, "(", BEG|LABEL], [[1, 8], :on_ident, "a", ARG], [[1, 9], :on_comma, ",", BEG|LABEL], [[1, 10], :on_sp, " ", BEG|LABEL], [[1, 11], :on_ident, "b", ARG], [[1, 12], :on_rparen, ")", ENDFN], [[1, 13], :on_ignored_nl, "\n", BEG], [[2, 0], :on_sp, " ", BEG], [[2, 2], :on_ident, "puts", CMDARG], [[2, 6], :on_sp, " ", CMDARG], [[2, 7], :on_ident, "a", END|LABEL], [[2, 8], :on_sp, " ", END|LABEL], [[2, 9], :on_op, "+", BEG], [[2, 10], :on_sp, " ", BEG], [[2, 11], :on_ident, "b", END|LABEL], [[2, 12], :on_nl, "\n", BEG],
15. 2. Parse
16. [[[1, 0], :on_kw, "def", FNAME], [[1, 3], :on_sp, " ", FNAME], [[1, 4], :on_ident, "foo", ENDFN], [[1, 7], :on_lparen, "(", BEG|LABEL], [[1, 8], :on_ident, "a", ARG], [[1, 9], :on_comma, ",", BEG|LABEL], [[1, 10], :on_sp, " ", BEG|LABEL], [[1, 11], :on_ident, "b", ARG], [[1, 12], :on_rparen, ")", ENDFN], [[1, 13], :on_ignored_nl, "\n", BEG], [[2, 0], :on_sp, " ", BEG], [[2, 2], :on_ident, "puts", CMDARG], [[2, 6], :on_sp, " ", CMDARG], [[2, 7], :on_ident, "a", END|LABEL], [[2, 8], :on_sp, " ", END|LABEL], [[2, 9], :on_op, "+", BEG], [[2, 10], :on_sp, " ", BEG], [[2, 11], :on_ident, "b", END|LABEL], [[2, 12], :on_nl, "\n", BEG], [[3, 0], :on_kw, "end", END], [[3, 3], :on_nl, "\n", BEG], [[4, 0], :on_ignored_nl, "\n", BEG], [[5, 0], :on_ident, "foo", CMDARG], [[5, 3], :on_lparen, "(", BEG|LABEL], [[5, 4], :on_int, "9", END], [[5, 5], :on_comma, ",", BEG|LABEL], [[5, 6], :on_sp, " ", BEG|LABEL], [[5, 7], :on_int, "8", END], [[5, 8], :on_rparen, ")", ENDFN], [[5, 9], :on_nl, "\n", BEG]]
17. foo_example = <<~HEREDOC def foo(a, b) puts a + b end foo(9, 8) HEREDOC puts Ripper.sexp(foo_example) e
18. command args_add_block ident “puts” binary var_ref var_ref ident “b” ident “a” + Abstract Syntax Tree (AST) Representation [:command, [:@ident, "puts", [2, 2]], [:args_add_block, [[:binary, [:var_ref, [:@ident, "a", [2, 7]]], :+, [:var_ref, [:@ident, "b", [2, 11]] ] ] ], false]] Symbolic Expression (S-exp) Representation puts a + b Source Code
19. Tokenizer Parser Your Ruby Code AST nodes Tokens
20. 3. ????
21. Tokenizer Parser Your Ruby Code Tokens Interpreter AST nodes $ < Ruby 1.8
22. Tokenizer Parser Compiler Your Ruby Code AST nodes Tokens Interpreter AST nodes % $ < Ruby 1.8 Ruby 1.9 +
23. 3. Compile
24. Compiler AST nodes Bytecode command args_add_block ident “puts” binary var_ref var_ref ident “b” ident “a” + ❓❓❓
25. foo_example = <<~HEREDOC def foo(a, b) puts a + b end foo(9, 8) HEREDOC puts RubyVM::InstructionSequence.compile(foo_example, specialized_instruction: false, instructions_unification: false, operands_unification: false ).to_a
26. foo_example = <<~HEREDOC def foo(a, b) puts a + b end foo(9, 8) HEREDOC puts RubyVM::InstructionSequence.compile(foo_example, specialized_instruction: false, instructions_unification: false, operands_unification: false ).to_a
27. ["YARVInstructionSequence/SimpleDataFormat", 3, 1, 1, {:arg_size=>0, :local_size=>0, :stack_max=>3, :node_id=>19, :code_location=>[1, 0, 5, 9], :node_ids=>[1, 12, 13, 15, 12, -1]}, "<compiled>", "<compiled>", "<compiled>", 1, :top, [], {}, [], [1, :RUBY_EVENT_LINE, [:definemethod, :foo, ["YARVInstructionSequence/SimpleDataFormat", 3, 1, 1, {:arg_size=>2, :local_size=>2, :stack_max=>3, :node_id=>11, :code_location=>[1, 0, 3, 3], :node_ids=>[5, 6, 7, 9, 5, -1]}, "foo", "<compiled>", "<compiled>", 1, :method, [:a, :b], {:lead_num=>2}, [], [2, :RUBY_EVENT_LINE, :RUBY_EVENT_CALL, [:putself], [:getlocal, 4, 0], [:getlocal, 3, 0], [:send, {:mid=>:+, :flag=>16, :orig_argc=>1}, nil], [:send, {:mid=>:puts, :flag=>20, :orig_argc=>1}, nil], 3, :RUBY_EVENT_RETURN, [:leave]]]], 5, :RUBY_EVENT_LINE, [:putself], [:putobject, 9], [:putobject, 8], [:send, {:mid=>:foo, :flag=>20, :orig_argc=>2}, nil], [:leave]]]
28. foo_example = <<~HEREDOC def foo(a, b) puts a + b end foo(9, 8) HEREDOC puts RubyVM::InstructionSequence.compile(foo_example, specialized_instruction: false, instructions_unification: false, operands_unification: false ).disasm
29. == disasm: #<ISeq:<compiled>@<compiled>:1 (1,0)-(5,9)> (catch: FALSE) 0000 definemethod :foo, foo ( 1)[Li] 0003 putself ( 5)[Li] 0004 putobject 9 0006 putobject 8 0008 send <calldata!mid:foo, argc:2, FCALL|ARGS_SIMPLE>, nil 0011 leave == disasm: #<ISeq:foo@<compiled>:1 (1,0)-(3,3)> (catch: FALSE) local table (size: 2, argc: 2 [opts: 0, rest: -1, post: 0, block: -1, kw: -1@-1, kwrest: -1]) [ 2] a@0<Arg> [ 1] b@1<Arg> 0000 putself ( 2)[LiCa] 0001 getlocal a@0, 0 0004 getlocal b@1, 0 0007 send <calldata!mid:+, argc:1, ARGS_SIMPLE>, nil 0010 send <calldata!mid:puts, argc:1, FCALL|ARGS_SIMPLE>, nil 0013 leave ( 3)[Re]
30. == disasm: #<ISeq:<compiled>@<compiled>:1 (1,0)-(5,9)> (catch: FALSE) 0000 definemethod :foo, foo ( 1)[Li] 0003 putself ( 5)[Li] 0004 putobject 9 0006 putobject 8 0008 send <calldata! mid:foo, argc:2, FCALL|ARGS_SIMPLE>, nil 0011 leave == disasm: #<ISeq:foo@<compiled>:1 (1,0)-(3,3)> (catch: FALSE) local table (size: 2, argc: 2 [opts: 0, rest: -1, post: 0, block: -1, kw: -1@-1, kwrest: -1]) [ 2] a@0<Arg> [ 1] b@1<Arg> 0000 putself ( 2)[LiCa] 0001 getlocal a@0, 0 0004 getlocal b@1, 0 0007 send <calldata!mid:+, argc:1, ARGS_SIMPLE>, nil 0010 send <calldata!mid:puts, argc:1, FCALL|ARGS_SIMPLE>, nil 0013 leave ( 3)[Re] … puts a + b … … def foo(a, b) … end foo(9, 8)
31. Compiler AST nodes Bytecode == disasm: #<ISeq:<compiled>@<compiled>:1 (1,0)- (5,9)> (catch: FALSE) 0000 definemethod :foo, foo ( 1)[Li] 0003 putself ( 5)[Li] 0004 putobject 9 0006 putobject 8 0008 send <calldata!mid:foo, argc:2, FCALL|ARGS_SIMPLE>, nil 0011 leave com args_add ident binar var_re var_re ident ident +
32. Tokenizer Parser Compiler Your Ruby Code AST nodes Tokens Bytecode
33. 4. Interpret
34. ➡ & Interpret Tokenizer Parser Compiler Your Ruby Code AST nodes Tokens Bytecode Virtual Machine
35. YARV
36. Yet Another Ruby Virtual Machine
37. How does YARV work?
38. YARV is a stack-based virtual machine.
39. YARV maintains two stacks.
40. 1. Internal stack 2. Control frame stack
41. 1. Internal stack 2. Control frame stack Stores values: returns, local variables, arguments
42. 1. Internal stack 2. Control frame stack Stores values: returns, local variables, arguments Tracks Ruby’s call stack (methods and blocks)
43. puts 9 + 8
44. == disasm: #<ISeq:<compiled>@<compiled>:1 (1,0)-(1,10)> (catch: FALSE) 0000 putself ( 1)[Li] 0001 putobject 9 0003 putobject 8 0005 send <calldata!mid:+, argc:1, ARGS_SIMPLE>, nil 0008 send <calldata!mid:puts, argc:1, FCALL|ARGS_SIMPLE>, nil 0011 leave Output: puts RubyVM::InstructionSequence.compile(“puts 9 + 8”, specialized_instruction: false, instructions_unification: false, operands_unification: false ).disasm
45. putself putobject(9) putobject(8) send(:+, argc: 1) send(:puts, argc: 1) leave Simpli ied Bytecode puts 9 + 8 Source Code
46. putself putobject(9) putobject(8) send(:+, argc: 1) send(:puts, argc: 1) leave Simpli ied Bytecode puts 9 + 8 Source Code
47. leave send(:puts, argc: 1) send(:+, argc: 1) putobject(8) putobject(9) putself Program Counter self YARV Instruction Sequence YARV’s Internal Stack Stack Pointer Instruction being executed: putself
48. leave send(:puts, argc: 1) send(:+, argc: 1) putobject(8) putobject(9) putself Program Counter self 9 YARV Instruction Sequence YARV’s Internal Stack Stack Pointer Instruction being executed: putobject(9)
49. leave send(:puts, argc: 1) send(:+, argc: 1) putobject(8) putobject(9) putself Program Counter self 9 8 YARV Instruction Sequence YARV’s Internal Stack Stack Pointer Instruction being executed: putobject(8)
50. leave send(:puts, argc: 1) send(:+, argc: 1) putobject(8) putobject(9) putself Program Counter self 9 8 YARV Instruction Sequence YARV’s Internal Stack Stack Pointer Instruction being executed: send(:+, argc: 1)
51. leave send(:puts, argc: 1) send(:+, argc: 1) putobject(8) putobject(9) putself Program Counter self YARV Instruction Sequence YARV’s Internal Stack Stack Pointer Instruction being executed: 9.send(:+, 8) 17
52. leave send(:puts, argc: 1) send(:+, argc: 1) putobject(8) putobject(9) putself Program Counter self YARV Instruction Sequence YARV’s Internal Stack Stack Pointer Instruction being executed: 17 send(:puts, argc: 1)
53. leave send(:puts, argc: 1) send(:+, argc: 1) putobject(8) putobject(9) putself Program Counter YARV Instruction Sequence YARV’s Internal Stack Stack Pointer Instruction being executed: self.send(:puts, 17) nil
54. leave send(:puts, argc: 1) send(:+, argc: 1) putobject(8) putobject(9) putself Program Counter YARV Instruction Sequence YARV’s Internal Stack Stack Pointer Instruction being executed: nil leave
55. def foo(a, b) puts a + b end foo(9, 8)
56. == disasm: #<ISeq:<compiled>@<compiled>:1 (1,0)-(5,9)> (catch: FALSE) 0000 definemethod :foo, foo ( 1)[Li] 0003 putself ( 5)[Li] 0004 putobject 9 0006 putobject 8 0008 send <calldata!mid:foo, argc:2, FCALL|ARGS_SIMPLE>, nil 0011 leave == disasm: #<ISeq:foo@<compiled>:1 (1,0)-(3,3)> (catch: FALSE) local table (size: 2, argc: 2 [opts: 0, rest: -1, post: 0, block: -1, kw: -1@-1, kwrest: -1]) [ 2] a@0<Arg> [ 1] b@1<Arg> 0000 putself ( 2)[LiCa] 0001 getlocal a@0, 0 0004 getlocal b@1, 0 0007 send <calldata!mid:+, argc:1, ARGS_SIMPLE>, nil 0010 send <calldata!mid:puts, argc:1, FCALL|ARGS_SIMPLE>, nil 0013 leave ( 3)[Re]
57. == disasm: #<ISeq:<compiled>@<compiled>:1 (1,0)-(5,9)> (catch: FALSE) 0000 definemethod :foo, foo ( 1)[Li] 0003 putself ( 5)[Li] 0004 putobject 9 0006 putobject 8 0008 send <calldata!mid:foo, argc:2, FCALL| ARGS_SIMPLE>, nil 0011 leave YARV Instruction Sequence For Main Scope
58. definemethod(:foo) putself putobject(9) putobject(8) send(:foo, argc: 2) leave Simpli ied Bytecode: Main Scope def foo(a, b) … end foo(9, 8) Source Code
59. leave send(:foo, argc: 2) putobject(8) putobject(9) putself definemethod(:foo) Program Counter YARV instructions for main scope YARV’s Internal Stack Stack Pointer Instruction being executed: definemethod(:foo)
60. leave send(:foo, argc: 2) putobject(8) putobject(9) putself definemethod(:foo) Program Counter self YARV instructions for main scope YARV’s Internal Stack Stack Pointer Instruction being executed: putself
61. leave send(:foo, argc: 2) putobject(8) putobject(9) putself definemethod(:foo) Program Counter self 9 YARV instructions for main scope YARV’s Internal Stack Stack Pointer Instruction being executed: putobject(9)
62. leave send(:foo, argc: 2) putobject(8) putobject(9) putself definemethod(:foo) Program Counter self 9 8 YARV instructions for main scope YARV’s Internal Stack Stack Pointer Instruction being executed: putobject(8)
63. leave send(:foo, argc: 2) putobject(8) putobject(9) putself definemethod(:foo) Program Counter self 9 8 YARV instructions for main scope YARV’s Internal Stack Stack Pointer Instruction being executed: send(:foo, argc: 2)
64. == disasm: #<ISeq:foo@<compiled>:1 (1,0)-(3,3)> (catch: FALSE) local table (size: 2, argc: 2 [opts: 0, rest: -1, post: 0, block: -1, kw: -1@-1, kwrest: -1]) [ 2] a@0<Arg> [ 1] b@1<Arg> 0000 putself ( 2)[LiCa] 0001 getlocal a@0, 0 0004 getlocal b@1, 0 0007 send <calldata!mid:+, argc:1, ARGS_SIMPLE>, nil 0010 send <calldata!mid:puts, argc:1, FCALL|ARGS_SIMPLE>, nil 0013 leave ( 3)[Re] YARV Instruction Sequence For Method #foo
65. == disasm: #<ISeq:foo@<compiled>:1 (1,0)-(3,3)> (catch: FALSE) local table (size: 2, argc: 2 [opts: 0, rest: -1, post: 0, block: -1, kw: -1@-1, kwrest: -1]) [ 2] a@0<Arg> [ 1] b@1<Arg> 0000 putself ( 2)[LiCa] 0001 getlocal a@0, 0 0004 getlocal b@1, 0 0007 send <calldata!mid:+, argc:1, ARGS_SIMPLE>, nil 0010 send <calldata!mid:puts, argc:1, FCALL|ARGS_SIMPLE>, nil 0013 leave ( 3)[Re] YARV Instruction Sequence For Method #foo
66. putself getlocal(a, index: 0) getlocal(b, index: 1) send(:+, argc: 1) send(:puts, argc: 1) leave Simpli ied Bytecode: #foo Method Scope a@0 <Arg> b@1 <Arg> Local Table … puts a + b … … Source Code
67. leave send(:puts, argc: 1) send(:+, argc: 1) getlocal(b, index: 1) getlocal(a, index: 0) putself YARV instructions for method scope self 9 8 YARV’s Internal Stack Instruction being executed: Stack Pointer
68. leave send(:puts, argc: 1) send(:+, argc: 1) getlocal(b, index: 1) getlocal(a, index: 0) putself YARV instructions for method scope self 9 8 YARV’s Internal Stack Environment Pointer Instruction being executed: Special 3 Special 2 Special 1 Stack Pointer
69. leave send(:puts, argc: 1) send(:+, argc: 1) getlocal(b, index: 1) getlocal(a, index: 0) putself Program Counter YARV instructions for method scope self 9 8 YARV’s Internal Stack Environment Pointer Instruction being executed: Special 3 Special 2 Special 1 Stack Pointer putself self
70. leave send(:puts, argc: 1) send(:+, argc: 1) getlocal(b, index: 1) getlocal(a, index: 0) putself Program Counter YARV instructions for method scope self 9 8 YARV’s Internal Stack Environment Pointer Instruction being executed: Special 3 Special 2 Special 1 Stack Pointer self 9 getlocal(a, index: 0)
71. leave send(:puts, argc: 1) send(:+, argc: 1) getlocal(b, index: 1) getlocal(a, index: 0) putself Program Counter YARV instructions for method scope self 9 8 YARV’s Internal Stack Environment Pointer Instruction being executed: Special 3 Special 2 Special 1 Stack Pointer self 9 8 getlocal(b, index: 1)
72. leave send(:puts, argc: 1) send(:+, argc: 1) getlocal(b, index: 1) getlocal(a, index: 0) putself Program Counter YARV instructions for method scope self 9 8 YARV’s Internal Stack Environment Pointer Instruction being executed: Special 3 Special 2 Special 1 Stack Pointer self 9 8 send(:+, argc: 1)
73. leave send(:puts, argc: 1) send(:+, argc: 1) getlocal(b, index: 1) getlocal(a, index: 0) putself Program Counter YARV instructions for method scope self 9 8 YARV’s Internal Stack Environment Pointer Instruction being executed: Special 3 Special 2 Special 1 Stack Pointer self 17 9.send(:+, 8)
74. leave send(:puts, argc: 1) send(:+, argc: 1) getlocal(b, index: 1) getlocal(a, index: 0) putself Program Counter YARV instructions for method scope self 9 8 YARV’s Internal Stack Environment Pointer Instruction being executed: Special 3 Special 2 Special 1 Stack Pointer self 17 send(:puts, argc: 1)
75. leave send(:puts, argc: 1) send(:+, argc: 1) getlocal(b, index: 1) getlocal(a, index: 0) putself Program Counter YARV instructions for method scope self 9 8 YARV’s Internal Stack Environment Pointer Instruction being executed: Special 3 Special 2 Special 1 Stack Pointer self.send(:puts, 17) nil
76. leave send(:puts, argc: 1) send(:+, argc: 1) getlocal(b, index: 1) getlocal(a, index: 0) putself Program Counter YARV instructions for method scope self 9 8 YARV’s Internal Stack Environment Pointer Instruction being executed: Special 3 Special 2 Special 1 Stack Pointer nil leave
77. Why do we care about the Ruby VM?
78. Why use a virtual machine? • Better performance, especially for long-running programs • e.g. loops with large number of iterations • Additional optimizations performed on bytecode • Expandable / future-proof (e.g. JITs) • Cache bytecode instructions for use at a later time
79. Why use a virtual machine? • Better performance, especially for long-running programs • e.g. loops with large number of iterations • Additional optimizations performed on bytecode • Expandable / future-proof (e.g. JITs) • Cache bytecode instructions for use at a later time
80. Tokenizer Parser Compiler Your Ruby Code AST nodes Tokens Bytecode VM Interpret CPU JIT Compiler Machine code
81. Tokenizer Parser Compiler Your Ruby Code AST nodes Tokens Bytecode VM Interpret CPU JIT Compiler Machine code
82. 1. MJIT ✗ ruby --jit foo.rb 2. YJIT ✗ ruby --yjit foo.rb
83. Why use a virtual machine? • Better performance, especially for long-running programs • e.g. loops with large number of iterations • Additional optimizations performed on bytecode • Expandable / future-proof (e.g. JITs) • Cache bytecode instructions for use at a later time
86. Takeaways '
87. The Ruby execution process involves both compilation and interpretation.
88. The Ruby VM makes Ruby performant.
89. The Ruby VM makes Ruby a more future-proof programming language!
90. Get involved! (
91. • Ruby parser & formatter: ruby-syntax-tree/syntax_tree • YARV emulator: kddnewton/yarv • JIT for Ruby (educational): chrisseaton/rhizome • JIT for Ruby (YJIT): tenderlove/tenderjit Projects Written in Ruby • YJIT: shopify/yjit • Bootsnap: shopify/bootsnap Mentioned Projects
92. Thank you! @OngMaple on Twitter )