ruby--ruby/yjit.rb

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

316 lines
12 KiB
Ruby
Raw Permalink Normal View History

# frozen_string_literal: true
# This module allows for introspection of YJIT, CRuby's experimental in-process
# just-in-time compiler. This module exists only to help develop YJIT, as such,
# everything in the module is highly implementation specific and comes with no
# API stability guarantee whatsoever.
#
# This module may not exist if YJIT does not support the particular platform
# for which CRuby is built. There is also no API stability guarantee as to in
# what situations this module is defined.
module RubyVM::YJIT
Rust YJIT In December 2021, we opened an [issue] to solicit feedback regarding the porting of the YJIT codebase from C99 to Rust. There were some reservations, but this project was given the go ahead by Ruby core developers and Matz. Since then, we have successfully completed the port of YJIT to Rust. The new Rust version of YJIT has reached parity with the C version, in that it passes all the CRuby tests, is able to run all of the YJIT benchmarks, and performs similarly to the C version (because it works the same way and largely generates the same machine code). We've even incorporated some design improvements, such as a more fine-grained constant invalidation mechanism which we expect will make a big difference in Ruby on Rails applications. Because we want to be careful, YJIT is guarded behind a configure option: ```shell ./configure --enable-yjit # Build YJIT in release mode ./configure --enable-yjit=dev # Build YJIT in dev/debug mode ``` By default, YJIT does not get compiled and cargo/rustc is not required. If YJIT is built in dev mode, then `cargo` is used to fetch development dependencies, but when building in release, `cargo` is not required, only `rustc`. At the moment YJIT requires Rust 1.60.0 or newer. The YJIT command-line options remain mostly unchanged, and more details about the build process are documented in `doc/yjit/yjit.md`. The CI tests have been updated and do not take any more resources than before. The development history of the Rust port is available at the following commit for interested parties: https://github.com/Shopify/ruby/commit/1fd9573d8b4b65219f1c2407f30a0a60e537f8be Our hope is that Rust YJIT will be compiled and included as a part of system packages and compiled binaries of the Ruby 3.2 release. We do not anticipate any major problems as Rust is well supported on every platform which YJIT supports, but to make sure that this process works smoothly, we would like to reach out to those who take care of building systems packages before the 3.2 release is shipped and resolve any issues that may come up. [issue]: https://bugs.ruby-lang.org/issues/18481 Co-authored-by: Maxime Chevalier-Boisvert <maximechevalierb@gmail.com> Co-authored-by: Noah Gibbs <the.codefolio.guy@gmail.com> Co-authored-by: Kevin Newton <kddnewton@gmail.com>
2022-04-19 18:40:21 +00:00
# Check if YJIT is enabled
def self.enabled?
Primitive.cexpr! 'RBOOL(rb_yjit_enabled_p())'
end
Rust YJIT In December 2021, we opened an [issue] to solicit feedback regarding the porting of the YJIT codebase from C99 to Rust. There were some reservations, but this project was given the go ahead by Ruby core developers and Matz. Since then, we have successfully completed the port of YJIT to Rust. The new Rust version of YJIT has reached parity with the C version, in that it passes all the CRuby tests, is able to run all of the YJIT benchmarks, and performs similarly to the C version (because it works the same way and largely generates the same machine code). We've even incorporated some design improvements, such as a more fine-grained constant invalidation mechanism which we expect will make a big difference in Ruby on Rails applications. Because we want to be careful, YJIT is guarded behind a configure option: ```shell ./configure --enable-yjit # Build YJIT in release mode ./configure --enable-yjit=dev # Build YJIT in dev/debug mode ``` By default, YJIT does not get compiled and cargo/rustc is not required. If YJIT is built in dev mode, then `cargo` is used to fetch development dependencies, but when building in release, `cargo` is not required, only `rustc`. At the moment YJIT requires Rust 1.60.0 or newer. The YJIT command-line options remain mostly unchanged, and more details about the build process are documented in `doc/yjit/yjit.md`. The CI tests have been updated and do not take any more resources than before. The development history of the Rust port is available at the following commit for interested parties: https://github.com/Shopify/ruby/commit/1fd9573d8b4b65219f1c2407f30a0a60e537f8be Our hope is that Rust YJIT will be compiled and included as a part of system packages and compiled binaries of the Ruby 3.2 release. We do not anticipate any major problems as Rust is well supported on every platform which YJIT supports, but to make sure that this process works smoothly, we would like to reach out to those who take care of building systems packages before the 3.2 release is shipped and resolve any issues that may come up. [issue]: https://bugs.ruby-lang.org/issues/18481 Co-authored-by: Maxime Chevalier-Boisvert <maximechevalierb@gmail.com> Co-authored-by: Noah Gibbs <the.codefolio.guy@gmail.com> Co-authored-by: Kevin Newton <kddnewton@gmail.com>
2022-04-19 18:40:21 +00:00
def self.stats_enabled?
Primitive.rb_yjit_stats_enabled_p
end
Add ability to trace exit locations in yjit (#5970) When running with `--yjit-stats` turned on, yjit can inform the user what the most common exits are. While this is useful information it doesn't tell you the source location of the code that exited or what the code that exited looks like. This change intends to fix that. To use the feature, run yjit with the `--yjit-trace-exits` option, which will record the backtrace for every exit that occurs. This functionality requires the stats feature to be turned on. Calling `--yjit-trace-exits` will automatically set the `--yjit-stats` option. Users must call `RubyVM::YJIT.dump_exit_locations(filename)` which will Marshal dump the contents of `RubyVM::YJIT.exit_locations` into a file based on the passed filename. *Example usage:* Given the following script, we write to a file called `concat_array.dump` the results of `RubyVM::YJIT.exit_locations`. ```ruby def concat_array ["t", "r", *x = "u", "e"].join end 1000.times do concat_array end RubyVM::YJIT.dump_exit_locations("concat_array.dump") ``` When we run the file with this branch and the appropriate flags the stacktrace will be recorded. Note Stackprof needs to be installed or you need to point to the library directly. ``` ./ruby --yjit --yjit-call-threshold=1 --yjit-trace-exits -I/Users/eileencodes/open_source/stackprof/lib test.rb ``` We can then read the dump file with Stackprof: ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump ``` Results will look similar to the following: ``` ================================== Mode: () Samples: 1817 (0.00% miss rate) GC: 0 (0.00%) ================================== TOTAL (pct) SAMPLES (pct) FRAME 1001 (55.1%) 1001 (55.1%) concatarray 335 (18.4%) 335 (18.4%) invokeblock 178 (9.8%) 178 (9.8%) send 140 (7.7%) 140 (7.7%) opt_getinlinecache ...etc... ``` Simply inspecting the `concatarray` method will give `SOURCE UNAVAILABLE` because the source is insns.def. ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump --method concatarray ``` Result: ``` concatarray (nonexistent.def:1) samples: 1001 self (55.1%) / 1001 total (55.1%) callers: 1000 ( 99.9%) Object#concat_array 1 ( 0.1%) Gem.suffixes callees (0 total): code: SOURCE UNAVAILABLE ``` However if we go deeper to the callee we can see the exact source of the `concatarray` exit. ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump --method Object#concat_array ``` ``` Object#concat_array (/Users/eileencodes/open_source/rust_ruby/test.rb:1) samples: 0 self (0.0%) / 1000 total (55.0%) callers: 1000 ( 100.0%) block in <main> callees (1000 total): 1000 ( 100.0%) concatarray code: | 1 | def concat_array 1000 (55.0%) | 2 | ["t", "r", *x = "u", "e"].join | 3 | end ``` The `--walk` option is recommended for this feature as it make it easier to traverse the tree of exits. *Goals of this feature:* This feature is meant to give more information when working on YJIT. The idea is that if we know what code is exiting we can decide what areas to prioritize when fixing exits. In some cases this means adding prioritizing avoiding certain exits in yjit. In more complex cases it might mean changing the Ruby code to be more performant when run with yjit. Ultimately the more information we have about what code is exiting AND why, the better we can make yjit. *Known limitations:* * Due to tracing exits, running this on large codebases like Rails can be quite slow. * On complex methods it can still be difficult to pinpoint the exact cause of an exit. * Stackprof is a requirement to to view the backtrace information from the dump file. Co-authored-by: Aaron Patterson <tenderlove@ruby-lang.org> Co-authored-by: Aaron Patterson <tenderlove@ruby-lang.org>
2022-06-09 16:59:39 +00:00
# Check if rb_yjit_trace_exit_locations_enabled_p is enabled.
def self.trace_exit_locations_enabled?
Primitive.rb_yjit_trace_exit_locations_enabled_p
end
Rust YJIT In December 2021, we opened an [issue] to solicit feedback regarding the porting of the YJIT codebase from C99 to Rust. There were some reservations, but this project was given the go ahead by Ruby core developers and Matz. Since then, we have successfully completed the port of YJIT to Rust. The new Rust version of YJIT has reached parity with the C version, in that it passes all the CRuby tests, is able to run all of the YJIT benchmarks, and performs similarly to the C version (because it works the same way and largely generates the same machine code). We've even incorporated some design improvements, such as a more fine-grained constant invalidation mechanism which we expect will make a big difference in Ruby on Rails applications. Because we want to be careful, YJIT is guarded behind a configure option: ```shell ./configure --enable-yjit # Build YJIT in release mode ./configure --enable-yjit=dev # Build YJIT in dev/debug mode ``` By default, YJIT does not get compiled and cargo/rustc is not required. If YJIT is built in dev mode, then `cargo` is used to fetch development dependencies, but when building in release, `cargo` is not required, only `rustc`. At the moment YJIT requires Rust 1.60.0 or newer. The YJIT command-line options remain mostly unchanged, and more details about the build process are documented in `doc/yjit/yjit.md`. The CI tests have been updated and do not take any more resources than before. The development history of the Rust port is available at the following commit for interested parties: https://github.com/Shopify/ruby/commit/1fd9573d8b4b65219f1c2407f30a0a60e537f8be Our hope is that Rust YJIT will be compiled and included as a part of system packages and compiled binaries of the Ruby 3.2 release. We do not anticipate any major problems as Rust is well supported on every platform which YJIT supports, but to make sure that this process works smoothly, we would like to reach out to those who take care of building systems packages before the 3.2 release is shipped and resolve any issues that may come up. [issue]: https://bugs.ruby-lang.org/issues/18481 Co-authored-by: Maxime Chevalier-Boisvert <maximechevalierb@gmail.com> Co-authored-by: Noah Gibbs <the.codefolio.guy@gmail.com> Co-authored-by: Kevin Newton <kddnewton@gmail.com>
2022-04-19 18:40:21 +00:00
# Discard statistics collected for --yjit-stats.
def self.reset_stats!
Primitive.rb_yjit_reset_stats_bang
end
Add ability to trace exit locations in yjit (#5970) When running with `--yjit-stats` turned on, yjit can inform the user what the most common exits are. While this is useful information it doesn't tell you the source location of the code that exited or what the code that exited looks like. This change intends to fix that. To use the feature, run yjit with the `--yjit-trace-exits` option, which will record the backtrace for every exit that occurs. This functionality requires the stats feature to be turned on. Calling `--yjit-trace-exits` will automatically set the `--yjit-stats` option. Users must call `RubyVM::YJIT.dump_exit_locations(filename)` which will Marshal dump the contents of `RubyVM::YJIT.exit_locations` into a file based on the passed filename. *Example usage:* Given the following script, we write to a file called `concat_array.dump` the results of `RubyVM::YJIT.exit_locations`. ```ruby def concat_array ["t", "r", *x = "u", "e"].join end 1000.times do concat_array end RubyVM::YJIT.dump_exit_locations("concat_array.dump") ``` When we run the file with this branch and the appropriate flags the stacktrace will be recorded. Note Stackprof needs to be installed or you need to point to the library directly. ``` ./ruby --yjit --yjit-call-threshold=1 --yjit-trace-exits -I/Users/eileencodes/open_source/stackprof/lib test.rb ``` We can then read the dump file with Stackprof: ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump ``` Results will look similar to the following: ``` ================================== Mode: () Samples: 1817 (0.00% miss rate) GC: 0 (0.00%) ================================== TOTAL (pct) SAMPLES (pct) FRAME 1001 (55.1%) 1001 (55.1%) concatarray 335 (18.4%) 335 (18.4%) invokeblock 178 (9.8%) 178 (9.8%) send 140 (7.7%) 140 (7.7%) opt_getinlinecache ...etc... ``` Simply inspecting the `concatarray` method will give `SOURCE UNAVAILABLE` because the source is insns.def. ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump --method concatarray ``` Result: ``` concatarray (nonexistent.def:1) samples: 1001 self (55.1%) / 1001 total (55.1%) callers: 1000 ( 99.9%) Object#concat_array 1 ( 0.1%) Gem.suffixes callees (0 total): code: SOURCE UNAVAILABLE ``` However if we go deeper to the callee we can see the exact source of the `concatarray` exit. ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump --method Object#concat_array ``` ``` Object#concat_array (/Users/eileencodes/open_source/rust_ruby/test.rb:1) samples: 0 self (0.0%) / 1000 total (55.0%) callers: 1000 ( 100.0%) block in <main> callees (1000 total): 1000 ( 100.0%) concatarray code: | 1 | def concat_array 1000 (55.0%) | 2 | ["t", "r", *x = "u", "e"].join | 3 | end ``` The `--walk` option is recommended for this feature as it make it easier to traverse the tree of exits. *Goals of this feature:* This feature is meant to give more information when working on YJIT. The idea is that if we know what code is exiting we can decide what areas to prioritize when fixing exits. In some cases this means adding prioritizing avoiding certain exits in yjit. In more complex cases it might mean changing the Ruby code to be more performant when run with yjit. Ultimately the more information we have about what code is exiting AND why, the better we can make yjit. *Known limitations:* * Due to tracing exits, running this on large codebases like Rails can be quite slow. * On complex methods it can still be difficult to pinpoint the exact cause of an exit. * Stackprof is a requirement to to view the backtrace information from the dump file. Co-authored-by: Aaron Patterson <tenderlove@ruby-lang.org> Co-authored-by: Aaron Patterson <tenderlove@ruby-lang.org>
2022-06-09 16:59:39 +00:00
# If --yjit-trace-exits is enabled parse the hashes from
# Primitive.rb_yjit_get_exit_locations into a format readable
# by Stackprof. This will allow us to find the exact location of a
# side exit in YJIT based on the instruction that is exiting.
def self.exit_locations
return unless trace_exit_locations_enabled?
results = Primitive.rb_yjit_get_exit_locations
raw_samples = results[:raw].dup
line_samples = results[:lines].dup
frames = results[:frames].dup
samples_count = 0
# Loop through the instructions and set the frame hash with the data.
# We use nonexistent.def for the file name, otherwise insns.def will be displayed
# and that information isn't useful in this context.
RubyVM::INSTRUCTION_NAMES.each_with_index do |name, frame_id|
frame_hash = { samples: 0, total_samples: 0, edges: {}, name: name, file: "nonexistent.def", line: nil, lines: {} }
Add ability to trace exit locations in yjit (#5970) When running with `--yjit-stats` turned on, yjit can inform the user what the most common exits are. While this is useful information it doesn't tell you the source location of the code that exited or what the code that exited looks like. This change intends to fix that. To use the feature, run yjit with the `--yjit-trace-exits` option, which will record the backtrace for every exit that occurs. This functionality requires the stats feature to be turned on. Calling `--yjit-trace-exits` will automatically set the `--yjit-stats` option. Users must call `RubyVM::YJIT.dump_exit_locations(filename)` which will Marshal dump the contents of `RubyVM::YJIT.exit_locations` into a file based on the passed filename. *Example usage:* Given the following script, we write to a file called `concat_array.dump` the results of `RubyVM::YJIT.exit_locations`. ```ruby def concat_array ["t", "r", *x = "u", "e"].join end 1000.times do concat_array end RubyVM::YJIT.dump_exit_locations("concat_array.dump") ``` When we run the file with this branch and the appropriate flags the stacktrace will be recorded. Note Stackprof needs to be installed or you need to point to the library directly. ``` ./ruby --yjit --yjit-call-threshold=1 --yjit-trace-exits -I/Users/eileencodes/open_source/stackprof/lib test.rb ``` We can then read the dump file with Stackprof: ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump ``` Results will look similar to the following: ``` ================================== Mode: () Samples: 1817 (0.00% miss rate) GC: 0 (0.00%) ================================== TOTAL (pct) SAMPLES (pct) FRAME 1001 (55.1%) 1001 (55.1%) concatarray 335 (18.4%) 335 (18.4%) invokeblock 178 (9.8%) 178 (9.8%) send 140 (7.7%) 140 (7.7%) opt_getinlinecache ...etc... ``` Simply inspecting the `concatarray` method will give `SOURCE UNAVAILABLE` because the source is insns.def. ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump --method concatarray ``` Result: ``` concatarray (nonexistent.def:1) samples: 1001 self (55.1%) / 1001 total (55.1%) callers: 1000 ( 99.9%) Object#concat_array 1 ( 0.1%) Gem.suffixes callees (0 total): code: SOURCE UNAVAILABLE ``` However if we go deeper to the callee we can see the exact source of the `concatarray` exit. ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump --method Object#concat_array ``` ``` Object#concat_array (/Users/eileencodes/open_source/rust_ruby/test.rb:1) samples: 0 self (0.0%) / 1000 total (55.0%) callers: 1000 ( 100.0%) block in <main> callees (1000 total): 1000 ( 100.0%) concatarray code: | 1 | def concat_array 1000 (55.0%) | 2 | ["t", "r", *x = "u", "e"].join | 3 | end ``` The `--walk` option is recommended for this feature as it make it easier to traverse the tree of exits. *Goals of this feature:* This feature is meant to give more information when working on YJIT. The idea is that if we know what code is exiting we can decide what areas to prioritize when fixing exits. In some cases this means adding prioritizing avoiding certain exits in yjit. In more complex cases it might mean changing the Ruby code to be more performant when run with yjit. Ultimately the more information we have about what code is exiting AND why, the better we can make yjit. *Known limitations:* * Due to tracing exits, running this on large codebases like Rails can be quite slow. * On complex methods it can still be difficult to pinpoint the exact cause of an exit. * Stackprof is a requirement to to view the backtrace information from the dump file. Co-authored-by: Aaron Patterson <tenderlove@ruby-lang.org> Co-authored-by: Aaron Patterson <tenderlove@ruby-lang.org>
2022-06-09 16:59:39 +00:00
results[:frames][frame_id] = frame_hash
frames[frame_id] = frame_hash
end
# Loop through the raw_samples and build the hashes for StackProf.
# The loop is based off an example in the StackProf documentation and therefore
# this functionality can only work with that library.
#
# Raw Samples:
# [ length, frame1, frame2, frameN, ..., instruction, count
#
# Line Samples
# [ length, line_1, line_2, line_n, ..., dummy value, count
i = 0
while i < raw_samples.length
stack_length = raw_samples[i] + 1
i += 1 # consume the stack length
Add ability to trace exit locations in yjit (#5970) When running with `--yjit-stats` turned on, yjit can inform the user what the most common exits are. While this is useful information it doesn't tell you the source location of the code that exited or what the code that exited looks like. This change intends to fix that. To use the feature, run yjit with the `--yjit-trace-exits` option, which will record the backtrace for every exit that occurs. This functionality requires the stats feature to be turned on. Calling `--yjit-trace-exits` will automatically set the `--yjit-stats` option. Users must call `RubyVM::YJIT.dump_exit_locations(filename)` which will Marshal dump the contents of `RubyVM::YJIT.exit_locations` into a file based on the passed filename. *Example usage:* Given the following script, we write to a file called `concat_array.dump` the results of `RubyVM::YJIT.exit_locations`. ```ruby def concat_array ["t", "r", *x = "u", "e"].join end 1000.times do concat_array end RubyVM::YJIT.dump_exit_locations("concat_array.dump") ``` When we run the file with this branch and the appropriate flags the stacktrace will be recorded. Note Stackprof needs to be installed or you need to point to the library directly. ``` ./ruby --yjit --yjit-call-threshold=1 --yjit-trace-exits -I/Users/eileencodes/open_source/stackprof/lib test.rb ``` We can then read the dump file with Stackprof: ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump ``` Results will look similar to the following: ``` ================================== Mode: () Samples: 1817 (0.00% miss rate) GC: 0 (0.00%) ================================== TOTAL (pct) SAMPLES (pct) FRAME 1001 (55.1%) 1001 (55.1%) concatarray 335 (18.4%) 335 (18.4%) invokeblock 178 (9.8%) 178 (9.8%) send 140 (7.7%) 140 (7.7%) opt_getinlinecache ...etc... ``` Simply inspecting the `concatarray` method will give `SOURCE UNAVAILABLE` because the source is insns.def. ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump --method concatarray ``` Result: ``` concatarray (nonexistent.def:1) samples: 1001 self (55.1%) / 1001 total (55.1%) callers: 1000 ( 99.9%) Object#concat_array 1 ( 0.1%) Gem.suffixes callees (0 total): code: SOURCE UNAVAILABLE ``` However if we go deeper to the callee we can see the exact source of the `concatarray` exit. ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump --method Object#concat_array ``` ``` Object#concat_array (/Users/eileencodes/open_source/rust_ruby/test.rb:1) samples: 0 self (0.0%) / 1000 total (55.0%) callers: 1000 ( 100.0%) block in <main> callees (1000 total): 1000 ( 100.0%) concatarray code: | 1 | def concat_array 1000 (55.0%) | 2 | ["t", "r", *x = "u", "e"].join | 3 | end ``` The `--walk` option is recommended for this feature as it make it easier to traverse the tree of exits. *Goals of this feature:* This feature is meant to give more information when working on YJIT. The idea is that if we know what code is exiting we can decide what areas to prioritize when fixing exits. In some cases this means adding prioritizing avoiding certain exits in yjit. In more complex cases it might mean changing the Ruby code to be more performant when run with yjit. Ultimately the more information we have about what code is exiting AND why, the better we can make yjit. *Known limitations:* * Due to tracing exits, running this on large codebases like Rails can be quite slow. * On complex methods it can still be difficult to pinpoint the exact cause of an exit. * Stackprof is a requirement to to view the backtrace information from the dump file. Co-authored-by: Aaron Patterson <tenderlove@ruby-lang.org> Co-authored-by: Aaron Patterson <tenderlove@ruby-lang.org>
2022-06-09 16:59:39 +00:00
prev_frame_id = nil
stack_length.times do |idx|
idx += i
frame_id = raw_samples[idx]
Add ability to trace exit locations in yjit (#5970) When running with `--yjit-stats` turned on, yjit can inform the user what the most common exits are. While this is useful information it doesn't tell you the source location of the code that exited or what the code that exited looks like. This change intends to fix that. To use the feature, run yjit with the `--yjit-trace-exits` option, which will record the backtrace for every exit that occurs. This functionality requires the stats feature to be turned on. Calling `--yjit-trace-exits` will automatically set the `--yjit-stats` option. Users must call `RubyVM::YJIT.dump_exit_locations(filename)` which will Marshal dump the contents of `RubyVM::YJIT.exit_locations` into a file based on the passed filename. *Example usage:* Given the following script, we write to a file called `concat_array.dump` the results of `RubyVM::YJIT.exit_locations`. ```ruby def concat_array ["t", "r", *x = "u", "e"].join end 1000.times do concat_array end RubyVM::YJIT.dump_exit_locations("concat_array.dump") ``` When we run the file with this branch and the appropriate flags the stacktrace will be recorded. Note Stackprof needs to be installed or you need to point to the library directly. ``` ./ruby --yjit --yjit-call-threshold=1 --yjit-trace-exits -I/Users/eileencodes/open_source/stackprof/lib test.rb ``` We can then read the dump file with Stackprof: ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump ``` Results will look similar to the following: ``` ================================== Mode: () Samples: 1817 (0.00% miss rate) GC: 0 (0.00%) ================================== TOTAL (pct) SAMPLES (pct) FRAME 1001 (55.1%) 1001 (55.1%) concatarray 335 (18.4%) 335 (18.4%) invokeblock 178 (9.8%) 178 (9.8%) send 140 (7.7%) 140 (7.7%) opt_getinlinecache ...etc... ``` Simply inspecting the `concatarray` method will give `SOURCE UNAVAILABLE` because the source is insns.def. ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump --method concatarray ``` Result: ``` concatarray (nonexistent.def:1) samples: 1001 self (55.1%) / 1001 total (55.1%) callers: 1000 ( 99.9%) Object#concat_array 1 ( 0.1%) Gem.suffixes callees (0 total): code: SOURCE UNAVAILABLE ``` However if we go deeper to the callee we can see the exact source of the `concatarray` exit. ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump --method Object#concat_array ``` ``` Object#concat_array (/Users/eileencodes/open_source/rust_ruby/test.rb:1) samples: 0 self (0.0%) / 1000 total (55.0%) callers: 1000 ( 100.0%) block in <main> callees (1000 total): 1000 ( 100.0%) concatarray code: | 1 | def concat_array 1000 (55.0%) | 2 | ["t", "r", *x = "u", "e"].join | 3 | end ``` The `--walk` option is recommended for this feature as it make it easier to traverse the tree of exits. *Goals of this feature:* This feature is meant to give more information when working on YJIT. The idea is that if we know what code is exiting we can decide what areas to prioritize when fixing exits. In some cases this means adding prioritizing avoiding certain exits in yjit. In more complex cases it might mean changing the Ruby code to be more performant when run with yjit. Ultimately the more information we have about what code is exiting AND why, the better we can make yjit. *Known limitations:* * Due to tracing exits, running this on large codebases like Rails can be quite slow. * On complex methods it can still be difficult to pinpoint the exact cause of an exit. * Stackprof is a requirement to to view the backtrace information from the dump file. Co-authored-by: Aaron Patterson <tenderlove@ruby-lang.org> Co-authored-by: Aaron Patterson <tenderlove@ruby-lang.org>
2022-06-09 16:59:39 +00:00
if prev_frame_id
prev_frame = frames[prev_frame_id]
prev_frame[:edges][frame_id] ||= 0
prev_frame[:edges][frame_id] += 1
end
frame_info = frames[frame_id]
frame_info[:total_samples] += 1
frame_info[:lines][line_samples[idx]] ||= [0, 0]
frame_info[:lines][line_samples[idx]][0] += 1
Add ability to trace exit locations in yjit (#5970) When running with `--yjit-stats` turned on, yjit can inform the user what the most common exits are. While this is useful information it doesn't tell you the source location of the code that exited or what the code that exited looks like. This change intends to fix that. To use the feature, run yjit with the `--yjit-trace-exits` option, which will record the backtrace for every exit that occurs. This functionality requires the stats feature to be turned on. Calling `--yjit-trace-exits` will automatically set the `--yjit-stats` option. Users must call `RubyVM::YJIT.dump_exit_locations(filename)` which will Marshal dump the contents of `RubyVM::YJIT.exit_locations` into a file based on the passed filename. *Example usage:* Given the following script, we write to a file called `concat_array.dump` the results of `RubyVM::YJIT.exit_locations`. ```ruby def concat_array ["t", "r", *x = "u", "e"].join end 1000.times do concat_array end RubyVM::YJIT.dump_exit_locations("concat_array.dump") ``` When we run the file with this branch and the appropriate flags the stacktrace will be recorded. Note Stackprof needs to be installed or you need to point to the library directly. ``` ./ruby --yjit --yjit-call-threshold=1 --yjit-trace-exits -I/Users/eileencodes/open_source/stackprof/lib test.rb ``` We can then read the dump file with Stackprof: ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump ``` Results will look similar to the following: ``` ================================== Mode: () Samples: 1817 (0.00% miss rate) GC: 0 (0.00%) ================================== TOTAL (pct) SAMPLES (pct) FRAME 1001 (55.1%) 1001 (55.1%) concatarray 335 (18.4%) 335 (18.4%) invokeblock 178 (9.8%) 178 (9.8%) send 140 (7.7%) 140 (7.7%) opt_getinlinecache ...etc... ``` Simply inspecting the `concatarray` method will give `SOURCE UNAVAILABLE` because the source is insns.def. ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump --method concatarray ``` Result: ``` concatarray (nonexistent.def:1) samples: 1001 self (55.1%) / 1001 total (55.1%) callers: 1000 ( 99.9%) Object#concat_array 1 ( 0.1%) Gem.suffixes callees (0 total): code: SOURCE UNAVAILABLE ``` However if we go deeper to the callee we can see the exact source of the `concatarray` exit. ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump --method Object#concat_array ``` ``` Object#concat_array (/Users/eileencodes/open_source/rust_ruby/test.rb:1) samples: 0 self (0.0%) / 1000 total (55.0%) callers: 1000 ( 100.0%) block in <main> callees (1000 total): 1000 ( 100.0%) concatarray code: | 1 | def concat_array 1000 (55.0%) | 2 | ["t", "r", *x = "u", "e"].join | 3 | end ``` The `--walk` option is recommended for this feature as it make it easier to traverse the tree of exits. *Goals of this feature:* This feature is meant to give more information when working on YJIT. The idea is that if we know what code is exiting we can decide what areas to prioritize when fixing exits. In some cases this means adding prioritizing avoiding certain exits in yjit. In more complex cases it might mean changing the Ruby code to be more performant when run with yjit. Ultimately the more information we have about what code is exiting AND why, the better we can make yjit. *Known limitations:* * Due to tracing exits, running this on large codebases like Rails can be quite slow. * On complex methods it can still be difficult to pinpoint the exact cause of an exit. * Stackprof is a requirement to to view the backtrace information from the dump file. Co-authored-by: Aaron Patterson <tenderlove@ruby-lang.org> Co-authored-by: Aaron Patterson <tenderlove@ruby-lang.org>
2022-06-09 16:59:39 +00:00
prev_frame_id = frame_id
end
i += stack_length # consume the stack
top_frame_id = prev_frame_id
Add ability to trace exit locations in yjit (#5970) When running with `--yjit-stats` turned on, yjit can inform the user what the most common exits are. While this is useful information it doesn't tell you the source location of the code that exited or what the code that exited looks like. This change intends to fix that. To use the feature, run yjit with the `--yjit-trace-exits` option, which will record the backtrace for every exit that occurs. This functionality requires the stats feature to be turned on. Calling `--yjit-trace-exits` will automatically set the `--yjit-stats` option. Users must call `RubyVM::YJIT.dump_exit_locations(filename)` which will Marshal dump the contents of `RubyVM::YJIT.exit_locations` into a file based on the passed filename. *Example usage:* Given the following script, we write to a file called `concat_array.dump` the results of `RubyVM::YJIT.exit_locations`. ```ruby def concat_array ["t", "r", *x = "u", "e"].join end 1000.times do concat_array end RubyVM::YJIT.dump_exit_locations("concat_array.dump") ``` When we run the file with this branch and the appropriate flags the stacktrace will be recorded. Note Stackprof needs to be installed or you need to point to the library directly. ``` ./ruby --yjit --yjit-call-threshold=1 --yjit-trace-exits -I/Users/eileencodes/open_source/stackprof/lib test.rb ``` We can then read the dump file with Stackprof: ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump ``` Results will look similar to the following: ``` ================================== Mode: () Samples: 1817 (0.00% miss rate) GC: 0 (0.00%) ================================== TOTAL (pct) SAMPLES (pct) FRAME 1001 (55.1%) 1001 (55.1%) concatarray 335 (18.4%) 335 (18.4%) invokeblock 178 (9.8%) 178 (9.8%) send 140 (7.7%) 140 (7.7%) opt_getinlinecache ...etc... ``` Simply inspecting the `concatarray` method will give `SOURCE UNAVAILABLE` because the source is insns.def. ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump --method concatarray ``` Result: ``` concatarray (nonexistent.def:1) samples: 1001 self (55.1%) / 1001 total (55.1%) callers: 1000 ( 99.9%) Object#concat_array 1 ( 0.1%) Gem.suffixes callees (0 total): code: SOURCE UNAVAILABLE ``` However if we go deeper to the callee we can see the exact source of the `concatarray` exit. ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump --method Object#concat_array ``` ``` Object#concat_array (/Users/eileencodes/open_source/rust_ruby/test.rb:1) samples: 0 self (0.0%) / 1000 total (55.0%) callers: 1000 ( 100.0%) block in <main> callees (1000 total): 1000 ( 100.0%) concatarray code: | 1 | def concat_array 1000 (55.0%) | 2 | ["t", "r", *x = "u", "e"].join | 3 | end ``` The `--walk` option is recommended for this feature as it make it easier to traverse the tree of exits. *Goals of this feature:* This feature is meant to give more information when working on YJIT. The idea is that if we know what code is exiting we can decide what areas to prioritize when fixing exits. In some cases this means adding prioritizing avoiding certain exits in yjit. In more complex cases it might mean changing the Ruby code to be more performant when run with yjit. Ultimately the more information we have about what code is exiting AND why, the better we can make yjit. *Known limitations:* * Due to tracing exits, running this on large codebases like Rails can be quite slow. * On complex methods it can still be difficult to pinpoint the exact cause of an exit. * Stackprof is a requirement to to view the backtrace information from the dump file. Co-authored-by: Aaron Patterson <tenderlove@ruby-lang.org> Co-authored-by: Aaron Patterson <tenderlove@ruby-lang.org>
2022-06-09 16:59:39 +00:00
top_frame_line = 1
sample_count = raw_samples[i]
frames[top_frame_id][:samples] += sample_count
Add ability to trace exit locations in yjit (#5970) When running with `--yjit-stats` turned on, yjit can inform the user what the most common exits are. While this is useful information it doesn't tell you the source location of the code that exited or what the code that exited looks like. This change intends to fix that. To use the feature, run yjit with the `--yjit-trace-exits` option, which will record the backtrace for every exit that occurs. This functionality requires the stats feature to be turned on. Calling `--yjit-trace-exits` will automatically set the `--yjit-stats` option. Users must call `RubyVM::YJIT.dump_exit_locations(filename)` which will Marshal dump the contents of `RubyVM::YJIT.exit_locations` into a file based on the passed filename. *Example usage:* Given the following script, we write to a file called `concat_array.dump` the results of `RubyVM::YJIT.exit_locations`. ```ruby def concat_array ["t", "r", *x = "u", "e"].join end 1000.times do concat_array end RubyVM::YJIT.dump_exit_locations("concat_array.dump") ``` When we run the file with this branch and the appropriate flags the stacktrace will be recorded. Note Stackprof needs to be installed or you need to point to the library directly. ``` ./ruby --yjit --yjit-call-threshold=1 --yjit-trace-exits -I/Users/eileencodes/open_source/stackprof/lib test.rb ``` We can then read the dump file with Stackprof: ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump ``` Results will look similar to the following: ``` ================================== Mode: () Samples: 1817 (0.00% miss rate) GC: 0 (0.00%) ================================== TOTAL (pct) SAMPLES (pct) FRAME 1001 (55.1%) 1001 (55.1%) concatarray 335 (18.4%) 335 (18.4%) invokeblock 178 (9.8%) 178 (9.8%) send 140 (7.7%) 140 (7.7%) opt_getinlinecache ...etc... ``` Simply inspecting the `concatarray` method will give `SOURCE UNAVAILABLE` because the source is insns.def. ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump --method concatarray ``` Result: ``` concatarray (nonexistent.def:1) samples: 1001 self (55.1%) / 1001 total (55.1%) callers: 1000 ( 99.9%) Object#concat_array 1 ( 0.1%) Gem.suffixes callees (0 total): code: SOURCE UNAVAILABLE ``` However if we go deeper to the callee we can see the exact source of the `concatarray` exit. ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump --method Object#concat_array ``` ``` Object#concat_array (/Users/eileencodes/open_source/rust_ruby/test.rb:1) samples: 0 self (0.0%) / 1000 total (55.0%) callers: 1000 ( 100.0%) block in <main> callees (1000 total): 1000 ( 100.0%) concatarray code: | 1 | def concat_array 1000 (55.0%) | 2 | ["t", "r", *x = "u", "e"].join | 3 | end ``` The `--walk` option is recommended for this feature as it make it easier to traverse the tree of exits. *Goals of this feature:* This feature is meant to give more information when working on YJIT. The idea is that if we know what code is exiting we can decide what areas to prioritize when fixing exits. In some cases this means adding prioritizing avoiding certain exits in yjit. In more complex cases it might mean changing the Ruby code to be more performant when run with yjit. Ultimately the more information we have about what code is exiting AND why, the better we can make yjit. *Known limitations:* * Due to tracing exits, running this on large codebases like Rails can be quite slow. * On complex methods it can still be difficult to pinpoint the exact cause of an exit. * Stackprof is a requirement to to view the backtrace information from the dump file. Co-authored-by: Aaron Patterson <tenderlove@ruby-lang.org> Co-authored-by: Aaron Patterson <tenderlove@ruby-lang.org>
2022-06-09 16:59:39 +00:00
frames[top_frame_id][:lines] ||= {}
frames[top_frame_id][:lines][top_frame_line] ||= [0, 0]
frames[top_frame_id][:lines][top_frame_line][1] += sample_count
Add ability to trace exit locations in yjit (#5970) When running with `--yjit-stats` turned on, yjit can inform the user what the most common exits are. While this is useful information it doesn't tell you the source location of the code that exited or what the code that exited looks like. This change intends to fix that. To use the feature, run yjit with the `--yjit-trace-exits` option, which will record the backtrace for every exit that occurs. This functionality requires the stats feature to be turned on. Calling `--yjit-trace-exits` will automatically set the `--yjit-stats` option. Users must call `RubyVM::YJIT.dump_exit_locations(filename)` which will Marshal dump the contents of `RubyVM::YJIT.exit_locations` into a file based on the passed filename. *Example usage:* Given the following script, we write to a file called `concat_array.dump` the results of `RubyVM::YJIT.exit_locations`. ```ruby def concat_array ["t", "r", *x = "u", "e"].join end 1000.times do concat_array end RubyVM::YJIT.dump_exit_locations("concat_array.dump") ``` When we run the file with this branch and the appropriate flags the stacktrace will be recorded. Note Stackprof needs to be installed or you need to point to the library directly. ``` ./ruby --yjit --yjit-call-threshold=1 --yjit-trace-exits -I/Users/eileencodes/open_source/stackprof/lib test.rb ``` We can then read the dump file with Stackprof: ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump ``` Results will look similar to the following: ``` ================================== Mode: () Samples: 1817 (0.00% miss rate) GC: 0 (0.00%) ================================== TOTAL (pct) SAMPLES (pct) FRAME 1001 (55.1%) 1001 (55.1%) concatarray 335 (18.4%) 335 (18.4%) invokeblock 178 (9.8%) 178 (9.8%) send 140 (7.7%) 140 (7.7%) opt_getinlinecache ...etc... ``` Simply inspecting the `concatarray` method will give `SOURCE UNAVAILABLE` because the source is insns.def. ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump --method concatarray ``` Result: ``` concatarray (nonexistent.def:1) samples: 1001 self (55.1%) / 1001 total (55.1%) callers: 1000 ( 99.9%) Object#concat_array 1 ( 0.1%) Gem.suffixes callees (0 total): code: SOURCE UNAVAILABLE ``` However if we go deeper to the callee we can see the exact source of the `concatarray` exit. ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump --method Object#concat_array ``` ``` Object#concat_array (/Users/eileencodes/open_source/rust_ruby/test.rb:1) samples: 0 self (0.0%) / 1000 total (55.0%) callers: 1000 ( 100.0%) block in <main> callees (1000 total): 1000 ( 100.0%) concatarray code: | 1 | def concat_array 1000 (55.0%) | 2 | ["t", "r", *x = "u", "e"].join | 3 | end ``` The `--walk` option is recommended for this feature as it make it easier to traverse the tree of exits. *Goals of this feature:* This feature is meant to give more information when working on YJIT. The idea is that if we know what code is exiting we can decide what areas to prioritize when fixing exits. In some cases this means adding prioritizing avoiding certain exits in yjit. In more complex cases it might mean changing the Ruby code to be more performant when run with yjit. Ultimately the more information we have about what code is exiting AND why, the better we can make yjit. *Known limitations:* * Due to tracing exits, running this on large codebases like Rails can be quite slow. * On complex methods it can still be difficult to pinpoint the exact cause of an exit. * Stackprof is a requirement to to view the backtrace information from the dump file. Co-authored-by: Aaron Patterson <tenderlove@ruby-lang.org> Co-authored-by: Aaron Patterson <tenderlove@ruby-lang.org>
2022-06-09 16:59:39 +00:00
samples_count += sample_count
i += 1
Add ability to trace exit locations in yjit (#5970) When running with `--yjit-stats` turned on, yjit can inform the user what the most common exits are. While this is useful information it doesn't tell you the source location of the code that exited or what the code that exited looks like. This change intends to fix that. To use the feature, run yjit with the `--yjit-trace-exits` option, which will record the backtrace for every exit that occurs. This functionality requires the stats feature to be turned on. Calling `--yjit-trace-exits` will automatically set the `--yjit-stats` option. Users must call `RubyVM::YJIT.dump_exit_locations(filename)` which will Marshal dump the contents of `RubyVM::YJIT.exit_locations` into a file based on the passed filename. *Example usage:* Given the following script, we write to a file called `concat_array.dump` the results of `RubyVM::YJIT.exit_locations`. ```ruby def concat_array ["t", "r", *x = "u", "e"].join end 1000.times do concat_array end RubyVM::YJIT.dump_exit_locations("concat_array.dump") ``` When we run the file with this branch and the appropriate flags the stacktrace will be recorded. Note Stackprof needs to be installed or you need to point to the library directly. ``` ./ruby --yjit --yjit-call-threshold=1 --yjit-trace-exits -I/Users/eileencodes/open_source/stackprof/lib test.rb ``` We can then read the dump file with Stackprof: ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump ``` Results will look similar to the following: ``` ================================== Mode: () Samples: 1817 (0.00% miss rate) GC: 0 (0.00%) ================================== TOTAL (pct) SAMPLES (pct) FRAME 1001 (55.1%) 1001 (55.1%) concatarray 335 (18.4%) 335 (18.4%) invokeblock 178 (9.8%) 178 (9.8%) send 140 (7.7%) 140 (7.7%) opt_getinlinecache ...etc... ``` Simply inspecting the `concatarray` method will give `SOURCE UNAVAILABLE` because the source is insns.def. ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump --method concatarray ``` Result: ``` concatarray (nonexistent.def:1) samples: 1001 self (55.1%) / 1001 total (55.1%) callers: 1000 ( 99.9%) Object#concat_array 1 ( 0.1%) Gem.suffixes callees (0 total): code: SOURCE UNAVAILABLE ``` However if we go deeper to the callee we can see the exact source of the `concatarray` exit. ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump --method Object#concat_array ``` ``` Object#concat_array (/Users/eileencodes/open_source/rust_ruby/test.rb:1) samples: 0 self (0.0%) / 1000 total (55.0%) callers: 1000 ( 100.0%) block in <main> callees (1000 total): 1000 ( 100.0%) concatarray code: | 1 | def concat_array 1000 (55.0%) | 2 | ["t", "r", *x = "u", "e"].join | 3 | end ``` The `--walk` option is recommended for this feature as it make it easier to traverse the tree of exits. *Goals of this feature:* This feature is meant to give more information when working on YJIT. The idea is that if we know what code is exiting we can decide what areas to prioritize when fixing exits. In some cases this means adding prioritizing avoiding certain exits in yjit. In more complex cases it might mean changing the Ruby code to be more performant when run with yjit. Ultimately the more information we have about what code is exiting AND why, the better we can make yjit. *Known limitations:* * Due to tracing exits, running this on large codebases like Rails can be quite slow. * On complex methods it can still be difficult to pinpoint the exact cause of an exit. * Stackprof is a requirement to to view the backtrace information from the dump file. Co-authored-by: Aaron Patterson <tenderlove@ruby-lang.org> Co-authored-by: Aaron Patterson <tenderlove@ruby-lang.org>
2022-06-09 16:59:39 +00:00
end
results[:samples] = samples_count
# Set missed_samples and gc_samples to 0 as their values
# don't matter to us in this context.
results[:missed_samples] = 0
results[:gc_samples] = 0
results
end
# Marshal dumps exit locations to the given filename.
#
# Usage:
#
# In a script call:
#
# RubyVM::YJIT.dump_exit_locations("my_file.dump")
#
# Then run the file with the following options:
#
# ruby --yjit --yjit-stats --yjit-trace-exits test.rb
#
# Once the code is done running, use Stackprof to read the dump file.
# See Stackprof documentation for options.
def self.dump_exit_locations(filename)
unless trace_exit_locations_enabled?
raise ArgumentError, "--yjit-trace-exits must be enabled to use dump_exit_locations."
end
File.binwrite(filename, Marshal.dump(RubyVM::YJIT.exit_locations))
Add ability to trace exit locations in yjit (#5970) When running with `--yjit-stats` turned on, yjit can inform the user what the most common exits are. While this is useful information it doesn't tell you the source location of the code that exited or what the code that exited looks like. This change intends to fix that. To use the feature, run yjit with the `--yjit-trace-exits` option, which will record the backtrace for every exit that occurs. This functionality requires the stats feature to be turned on. Calling `--yjit-trace-exits` will automatically set the `--yjit-stats` option. Users must call `RubyVM::YJIT.dump_exit_locations(filename)` which will Marshal dump the contents of `RubyVM::YJIT.exit_locations` into a file based on the passed filename. *Example usage:* Given the following script, we write to a file called `concat_array.dump` the results of `RubyVM::YJIT.exit_locations`. ```ruby def concat_array ["t", "r", *x = "u", "e"].join end 1000.times do concat_array end RubyVM::YJIT.dump_exit_locations("concat_array.dump") ``` When we run the file with this branch and the appropriate flags the stacktrace will be recorded. Note Stackprof needs to be installed or you need to point to the library directly. ``` ./ruby --yjit --yjit-call-threshold=1 --yjit-trace-exits -I/Users/eileencodes/open_source/stackprof/lib test.rb ``` We can then read the dump file with Stackprof: ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump ``` Results will look similar to the following: ``` ================================== Mode: () Samples: 1817 (0.00% miss rate) GC: 0 (0.00%) ================================== TOTAL (pct) SAMPLES (pct) FRAME 1001 (55.1%) 1001 (55.1%) concatarray 335 (18.4%) 335 (18.4%) invokeblock 178 (9.8%) 178 (9.8%) send 140 (7.7%) 140 (7.7%) opt_getinlinecache ...etc... ``` Simply inspecting the `concatarray` method will give `SOURCE UNAVAILABLE` because the source is insns.def. ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump --method concatarray ``` Result: ``` concatarray (nonexistent.def:1) samples: 1001 self (55.1%) / 1001 total (55.1%) callers: 1000 ( 99.9%) Object#concat_array 1 ( 0.1%) Gem.suffixes callees (0 total): code: SOURCE UNAVAILABLE ``` However if we go deeper to the callee we can see the exact source of the `concatarray` exit. ``` ./ruby -I/Users/eileencodes/open_source/stackprof/lib/ /Users/eileencodes/open_source/stackprof/bin/stackprof --text concat_array.dump --method Object#concat_array ``` ``` Object#concat_array (/Users/eileencodes/open_source/rust_ruby/test.rb:1) samples: 0 self (0.0%) / 1000 total (55.0%) callers: 1000 ( 100.0%) block in <main> callees (1000 total): 1000 ( 100.0%) concatarray code: | 1 | def concat_array 1000 (55.0%) | 2 | ["t", "r", *x = "u", "e"].join | 3 | end ``` The `--walk` option is recommended for this feature as it make it easier to traverse the tree of exits. *Goals of this feature:* This feature is meant to give more information when working on YJIT. The idea is that if we know what code is exiting we can decide what areas to prioritize when fixing exits. In some cases this means adding prioritizing avoiding certain exits in yjit. In more complex cases it might mean changing the Ruby code to be more performant when run with yjit. Ultimately the more information we have about what code is exiting AND why, the better we can make yjit. *Known limitations:* * Due to tracing exits, running this on large codebases like Rails can be quite slow. * On complex methods it can still be difficult to pinpoint the exact cause of an exit. * Stackprof is a requirement to to view the backtrace information from the dump file. Co-authored-by: Aaron Patterson <tenderlove@ruby-lang.org> Co-authored-by: Aaron Patterson <tenderlove@ruby-lang.org>
2022-06-09 16:59:39 +00:00
end
# Return a hash for statistics generated for the --yjit-stats command line option.
# Return nil when option is not passed or unavailable.
def self.runtime_stats
stats = Primitive.rb_yjit_get_stats
return stats if stats.nil? || !Primitive.rb_yjit_stats_enabled_p
side_exits = total_exit_count(stats)
total_exits = side_exits + stats[:leave_interp_return]
# Number of instructions that finish executing in YJIT.
# See :count-placement: about the subtraction.
retired_in_yjit = stats[:exec_instruction] - side_exits
# Average length of instruction sequences executed by YJIT
avg_len_in_yjit = retired_in_yjit.to_f / total_exits
# Proportion of instructions that retire in YJIT
total_insns_count = retired_in_yjit + stats[:vm_insns_count]
yjit_ratio_pct = 100.0 * retired_in_yjit.to_f / total_insns_count
# Make those stats available in RubyVM::YJIT.runtime_stats as well
stats[:side_exit_count] = side_exits
stats[:total_exit_count] = total_exits
stats[:ratio_in_yjit] = yjit_ratio_pct
stats[:avg_len_in_yjit] = avg_len_in_yjit
stats
end
Rust YJIT In December 2021, we opened an [issue] to solicit feedback regarding the porting of the YJIT codebase from C99 to Rust. There were some reservations, but this project was given the go ahead by Ruby core developers and Matz. Since then, we have successfully completed the port of YJIT to Rust. The new Rust version of YJIT has reached parity with the C version, in that it passes all the CRuby tests, is able to run all of the YJIT benchmarks, and performs similarly to the C version (because it works the same way and largely generates the same machine code). We've even incorporated some design improvements, such as a more fine-grained constant invalidation mechanism which we expect will make a big difference in Ruby on Rails applications. Because we want to be careful, YJIT is guarded behind a configure option: ```shell ./configure --enable-yjit # Build YJIT in release mode ./configure --enable-yjit=dev # Build YJIT in dev/debug mode ``` By default, YJIT does not get compiled and cargo/rustc is not required. If YJIT is built in dev mode, then `cargo` is used to fetch development dependencies, but when building in release, `cargo` is not required, only `rustc`. At the moment YJIT requires Rust 1.60.0 or newer. The YJIT command-line options remain mostly unchanged, and more details about the build process are documented in `doc/yjit/yjit.md`. The CI tests have been updated and do not take any more resources than before. The development history of the Rust port is available at the following commit for interested parties: https://github.com/Shopify/ruby/commit/1fd9573d8b4b65219f1c2407f30a0a60e537f8be Our hope is that Rust YJIT will be compiled and included as a part of system packages and compiled binaries of the Ruby 3.2 release. We do not anticipate any major problems as Rust is well supported on every platform which YJIT supports, but to make sure that this process works smoothly, we would like to reach out to those who take care of building systems packages before the 3.2 release is shipped and resolve any issues that may come up. [issue]: https://bugs.ruby-lang.org/issues/18481 Co-authored-by: Maxime Chevalier-Boisvert <maximechevalierb@gmail.com> Co-authored-by: Noah Gibbs <the.codefolio.guy@gmail.com> Co-authored-by: Kevin Newton <kddnewton@gmail.com>
2022-04-19 18:40:21 +00:00
# Produce disassembly for an iseq
def self.disasm(iseq)
# If a method or proc is passed in, get its iseq
iseq = RubyVM::InstructionSequence.of(iseq)
if self.enabled?
# Produce the disassembly string
# Include the YARV iseq disasm in the string for additional context
iseq.disasm + "\n" + Primitive.rb_yjit_disasm_iseq(iseq)
else
iseq.disasm
end
end
Rust YJIT In December 2021, we opened an [issue] to solicit feedback regarding the porting of the YJIT codebase from C99 to Rust. There were some reservations, but this project was given the go ahead by Ruby core developers and Matz. Since then, we have successfully completed the port of YJIT to Rust. The new Rust version of YJIT has reached parity with the C version, in that it passes all the CRuby tests, is able to run all of the YJIT benchmarks, and performs similarly to the C version (because it works the same way and largely generates the same machine code). We've even incorporated some design improvements, such as a more fine-grained constant invalidation mechanism which we expect will make a big difference in Ruby on Rails applications. Because we want to be careful, YJIT is guarded behind a configure option: ```shell ./configure --enable-yjit # Build YJIT in release mode ./configure --enable-yjit=dev # Build YJIT in dev/debug mode ``` By default, YJIT does not get compiled and cargo/rustc is not required. If YJIT is built in dev mode, then `cargo` is used to fetch development dependencies, but when building in release, `cargo` is not required, only `rustc`. At the moment YJIT requires Rust 1.60.0 or newer. The YJIT command-line options remain mostly unchanged, and more details about the build process are documented in `doc/yjit/yjit.md`. The CI tests have been updated and do not take any more resources than before. The development history of the Rust port is available at the following commit for interested parties: https://github.com/Shopify/ruby/commit/1fd9573d8b4b65219f1c2407f30a0a60e537f8be Our hope is that Rust YJIT will be compiled and included as a part of system packages and compiled binaries of the Ruby 3.2 release. We do not anticipate any major problems as Rust is well supported on every platform which YJIT supports, but to make sure that this process works smoothly, we would like to reach out to those who take care of building systems packages before the 3.2 release is shipped and resolve any issues that may come up. [issue]: https://bugs.ruby-lang.org/issues/18481 Co-authored-by: Maxime Chevalier-Boisvert <maximechevalierb@gmail.com> Co-authored-by: Noah Gibbs <the.codefolio.guy@gmail.com> Co-authored-by: Kevin Newton <kddnewton@gmail.com>
2022-04-19 18:40:21 +00:00
# Produce a list of instructions compiled by YJIT for an iseq
def self.insns_compiled(iseq)
# If a method or proc is passed in, get its iseq
iseq = RubyVM::InstructionSequence.of(iseq)
Rust YJIT In December 2021, we opened an [issue] to solicit feedback regarding the porting of the YJIT codebase from C99 to Rust. There were some reservations, but this project was given the go ahead by Ruby core developers and Matz. Since then, we have successfully completed the port of YJIT to Rust. The new Rust version of YJIT has reached parity with the C version, in that it passes all the CRuby tests, is able to run all of the YJIT benchmarks, and performs similarly to the C version (because it works the same way and largely generates the same machine code). We've even incorporated some design improvements, such as a more fine-grained constant invalidation mechanism which we expect will make a big difference in Ruby on Rails applications. Because we want to be careful, YJIT is guarded behind a configure option: ```shell ./configure --enable-yjit # Build YJIT in release mode ./configure --enable-yjit=dev # Build YJIT in dev/debug mode ``` By default, YJIT does not get compiled and cargo/rustc is not required. If YJIT is built in dev mode, then `cargo` is used to fetch development dependencies, but when building in release, `cargo` is not required, only `rustc`. At the moment YJIT requires Rust 1.60.0 or newer. The YJIT command-line options remain mostly unchanged, and more details about the build process are documented in `doc/yjit/yjit.md`. The CI tests have been updated and do not take any more resources than before. The development history of the Rust port is available at the following commit for interested parties: https://github.com/Shopify/ruby/commit/1fd9573d8b4b65219f1c2407f30a0a60e537f8be Our hope is that Rust YJIT will be compiled and included as a part of system packages and compiled binaries of the Ruby 3.2 release. We do not anticipate any major problems as Rust is well supported on every platform which YJIT supports, but to make sure that this process works smoothly, we would like to reach out to those who take care of building systems packages before the 3.2 release is shipped and resolve any issues that may come up. [issue]: https://bugs.ruby-lang.org/issues/18481 Co-authored-by: Maxime Chevalier-Boisvert <maximechevalierb@gmail.com> Co-authored-by: Noah Gibbs <the.codefolio.guy@gmail.com> Co-authored-by: Kevin Newton <kddnewton@gmail.com>
2022-04-19 18:40:21 +00:00
if self.enabled?
Primitive.rb_yjit_insns_compiled(iseq)
else
Qnil
end
2021-06-25 20:14:39 +00:00
end
# Free and recompile all existing JIT code
def self.code_gc
Primitive.rb_yjit_code_gc
end
def self.simulate_oom!
Rust YJIT In December 2021, we opened an [issue] to solicit feedback regarding the porting of the YJIT codebase from C99 to Rust. There were some reservations, but this project was given the go ahead by Ruby core developers and Matz. Since then, we have successfully completed the port of YJIT to Rust. The new Rust version of YJIT has reached parity with the C version, in that it passes all the CRuby tests, is able to run all of the YJIT benchmarks, and performs similarly to the C version (because it works the same way and largely generates the same machine code). We've even incorporated some design improvements, such as a more fine-grained constant invalidation mechanism which we expect will make a big difference in Ruby on Rails applications. Because we want to be careful, YJIT is guarded behind a configure option: ```shell ./configure --enable-yjit # Build YJIT in release mode ./configure --enable-yjit=dev # Build YJIT in dev/debug mode ``` By default, YJIT does not get compiled and cargo/rustc is not required. If YJIT is built in dev mode, then `cargo` is used to fetch development dependencies, but when building in release, `cargo` is not required, only `rustc`. At the moment YJIT requires Rust 1.60.0 or newer. The YJIT command-line options remain mostly unchanged, and more details about the build process are documented in `doc/yjit/yjit.md`. The CI tests have been updated and do not take any more resources than before. The development history of the Rust port is available at the following commit for interested parties: https://github.com/Shopify/ruby/commit/1fd9573d8b4b65219f1c2407f30a0a60e537f8be Our hope is that Rust YJIT will be compiled and included as a part of system packages and compiled binaries of the Ruby 3.2 release. We do not anticipate any major problems as Rust is well supported on every platform which YJIT supports, but to make sure that this process works smoothly, we would like to reach out to those who take care of building systems packages before the 3.2 release is shipped and resolve any issues that may come up. [issue]: https://bugs.ruby-lang.org/issues/18481 Co-authored-by: Maxime Chevalier-Boisvert <maximechevalierb@gmail.com> Co-authored-by: Noah Gibbs <the.codefolio.guy@gmail.com> Co-authored-by: Kevin Newton <kddnewton@gmail.com>
2022-04-19 18:40:21 +00:00
Primitive.rb_yjit_simulate_oom_bang
end
# Avoid calling a method here to not interfere with compilation tests
Rust YJIT In December 2021, we opened an [issue] to solicit feedback regarding the porting of the YJIT codebase from C99 to Rust. There were some reservations, but this project was given the go ahead by Ruby core developers and Matz. Since then, we have successfully completed the port of YJIT to Rust. The new Rust version of YJIT has reached parity with the C version, in that it passes all the CRuby tests, is able to run all of the YJIT benchmarks, and performs similarly to the C version (because it works the same way and largely generates the same machine code). We've even incorporated some design improvements, such as a more fine-grained constant invalidation mechanism which we expect will make a big difference in Ruby on Rails applications. Because we want to be careful, YJIT is guarded behind a configure option: ```shell ./configure --enable-yjit # Build YJIT in release mode ./configure --enable-yjit=dev # Build YJIT in dev/debug mode ``` By default, YJIT does not get compiled and cargo/rustc is not required. If YJIT is built in dev mode, then `cargo` is used to fetch development dependencies, but when building in release, `cargo` is not required, only `rustc`. At the moment YJIT requires Rust 1.60.0 or newer. The YJIT command-line options remain mostly unchanged, and more details about the build process are documented in `doc/yjit/yjit.md`. The CI tests have been updated and do not take any more resources than before. The development history of the Rust port is available at the following commit for interested parties: https://github.com/Shopify/ruby/commit/1fd9573d8b4b65219f1c2407f30a0a60e537f8be Our hope is that Rust YJIT will be compiled and included as a part of system packages and compiled binaries of the Ruby 3.2 release. We do not anticipate any major problems as Rust is well supported on every platform which YJIT supports, but to make sure that this process works smoothly, we would like to reach out to those who take care of building systems packages before the 3.2 release is shipped and resolve any issues that may come up. [issue]: https://bugs.ruby-lang.org/issues/18481 Co-authored-by: Maxime Chevalier-Boisvert <maximechevalierb@gmail.com> Co-authored-by: Noah Gibbs <the.codefolio.guy@gmail.com> Co-authored-by: Kevin Newton <kddnewton@gmail.com>
2022-04-19 18:40:21 +00:00
if Primitive.rb_yjit_stats_enabled_p
at_exit { _print_stats }
end
class << self
private
# Format and print out counters
def _print_stats
stats = runtime_stats
return unless stats
$stderr.puts("***YJIT: Printing YJIT statistics on exit***")
print_counters(stats, prefix: 'send_', prompt: 'method call exit reasons: ')
print_counters(stats, prefix: 'invokeblock_', prompt: 'invokeblock exit reasons: ')
print_counters(stats, prefix: 'invokesuper_', prompt: 'invokesuper exit reasons: ')
print_counters(stats, prefix: 'leave_', prompt: 'leave exit reasons: ')
print_counters(stats, prefix: 'gbpp_', prompt: 'getblockparamproxy exit reasons: ')
print_counters(stats, prefix: 'getivar_', prompt: 'getinstancevariable exit reasons:')
print_counters(stats, prefix: 'setivar_', prompt: 'setinstancevariable exit reasons:')
print_counters(stats, prefix: 'oaref_', prompt: 'opt_aref exit reasons: ')
2021-07-07 18:45:48 +00:00
print_counters(stats, prefix: 'expandarray_', prompt: 'expandarray exit reasons: ')
print_counters(stats, prefix: 'opt_getinlinecache_', prompt: 'opt_getinlinecache exit reasons: ')
print_counters(stats, prefix: 'invalidate_', prompt: 'invalidation reasons: ')
# Number of failed compiler invocations
compilation_failure = stats[:compilation_failure]
$stderr.puts "bindings_allocations: " + ("%10d" % stats[:binding_allocations])
$stderr.puts "bindings_set: " + ("%10d" % stats[:binding_set])
$stderr.puts "compilation_failure: " + ("%10d" % compilation_failure) if compilation_failure != 0
$stderr.puts "compiled_block_count: " + ("%10d" % stats[:compiled_block_count])
$stderr.puts "compiled_iseq_count: " + ("%10d" % stats[:compiled_iseq_count])
$stderr.puts "freed_iseq_count: " + ("%10d" % stats[:freed_iseq_count])
$stderr.puts "invalidation_count: " + ("%10d" % stats[:invalidation_count])
$stderr.puts "constant_state_bumps: " + ("%10d" % stats[:constant_state_bumps])
$stderr.puts "inline_code_size: " + ("%10d" % stats[:inline_code_size])
$stderr.puts "outlined_code_size: " + ("%10d" % stats[:outlined_code_size])
$stderr.puts "freed_code_size: " + ("%10d" % stats[:freed_code_size])
$stderr.puts "live_page_count: " + ("%10d" % stats[:live_page_count])
$stderr.puts "freed_page_count: " + ("%10d" % stats[:freed_page_count])
$stderr.puts "code_gc_count: " + ("%10d" % stats[:code_gc_count])
$stderr.puts "num_gc_obj_refs: " + ("%10d" % stats[:num_gc_obj_refs])
$stderr.puts "side_exit_count: " + ("%10d" % stats[:side_exit_count])
$stderr.puts "total_exit_count: " + ("%10d" % stats[:side_exit_count])
$stderr.puts "total_insns_count: " + ("%10d" % stats[:total_exit_count])
$stderr.puts "vm_insns_count: " + ("%10d" % stats[:vm_insns_count])
$stderr.puts "yjit_insns_count: " + ("%10d" % stats[:exec_instruction])
$stderr.puts "ratio_in_yjit: " + ("%9.1f" % stats[:ratio_in_yjit]) + "%"
$stderr.puts "avg_len_in_yjit: " + ("%10.1f" % stats[:avg_len_in_yjit])
print_sorted_exit_counts(stats, prefix: "exit_")
end
def print_sorted_exit_counts(stats, prefix:, how_many: 20, left_pad: 4)
exits = []
stats.each do |k, v|
if k.start_with?(prefix)
exits.push [k.to_s.delete_prefix(prefix), v]
end
end
exits = exits.sort_by { |name, count| -count }[0...how_many]
total_exits = total_exit_count(stats)
if total_exits > 0
top_n_total = exits.map { |name, count| count }.sum
top_n_exit_pct = 100.0 * top_n_total / total_exits
$stderr.puts "Top-#{how_many} most frequent exit ops (#{"%.1f" % top_n_exit_pct}% of exits):"
longest_insn_name_len = exits.map { |name, count| name.length }.max
exits.each do |name, count|
padding = longest_insn_name_len + left_pad
padded_name = "%#{padding}s" % name
padded_count = "%10d" % count
percent = 100.0 * count / total_exits
formatted_percent = "%.1f" % percent
$stderr.puts("#{padded_name}: #{padded_count} (#{formatted_percent}%)" )
end
else
$stderr.puts "total_exits: " + ("%10d" % total_exits)
end
end
def total_exit_count(stats, prefix: "exit_")
total = 0
stats.each do |k,v|
total += v if k.start_with?(prefix)
end
total
end
def print_counters(counters, prefix:, prompt:)
$stderr.puts(prompt)
counters = counters.filter { |key, _| key.start_with?(prefix) }
counters.filter! { |_, value| value != 0 }
counters.transform_keys! { |key| key.to_s.delete_prefix(prefix) }
if counters.empty?
$stderr.puts(" (all relevant counters are zero)")
return
end
counters = counters.to_a
counters.sort_by! { |(_, counter_value)| counter_value }
longest_name_length = counters.max_by { |(name, _)| name.length }.first.length
total = counters.sum { |(_, counter_value)| counter_value }
counters.reverse_each do |(name, value)|
percentage = value.fdiv(total) * 100
$stderr.printf(" %*s %10d (%4.1f%%)\n", longest_name_length, name, value, percentage);
end
end
end
Directly link libcapstone for easier development This lets us use libcapstone directly from miniruby so we don't need a Ruby Gem to to dev work. Example usage: ```ruby def foo(x) if x < 1 "wow" else "neat" end end iseq = RubyVM::InstructionSequence.of(method(:foo)) puts UJIT.disasm(iseq) 100.times { foo 1 } puts UJIT.disasm(iseq) ``` Then in the terminal ``` $ ./miniruby test.rb == disasm: #<ISeq:foo@test.rb:1 (1,0)-(7,3)> (catch: FALSE) local table (size: 1, argc: 1 [opts: 0, rest: -1, post: 0, block: -1, kw: -1@-1, kwrest: -1]) [ 1] x@0<Arg> 0000 getlocal_WC_0 x@0 ( 2)[LiCa] 0002 putobject_INT2FIX_1_ 0003 opt_lt <calldata!mid:<, argc:1, ARGS_SIMPLE> 0005 branchunless 10 0007 putstring "wow" ( 3)[Li] 0009 leave ( 7)[Re] 0010 putstring "neat" ( 5)[Li] 0012 leave ( 7)[Re] == ISEQ RANGE: 10 -> 10 ======================================================== 0x0: movabs rax, 0x7fe816e2d1a0 0xa: mov qword ptr [rdi], rax 0xd: mov r8, rax 0x10: mov r9, rax 0x13: mov r11, r12 0x16: jmp qword ptr [rax] == ISEQ RANGE: 0 -> 7 ========================================================== 0x0: mov rax, qword ptr [rdi + 0x20] 0x4: mov rax, qword ptr [rax - 0x18] 0x8: mov qword ptr [rdx], rax 0xb: mov qword ptr [rdx + 8], 3 0x13: movabs rax, 0x7fe817808200 0x1d: test byte ptr [rax + 0x3e6], 1 0x24: jne 0x3ffff7b 0x2a: test byte ptr [rdx], 1 0x2d: je 0x3ffff7b 0x33: test byte ptr [rdx + 8], 1 0x37: je 0x3ffff7b 0x3d: mov rax, qword ptr [rdx] 0x40: cmp rax, qword ptr [rdx + 8] 0x44: movabs rax, 0 0x4e: movabs rcx, 0x14 0x58: cmovl rax, rcx 0x5c: mov qword ptr [rdx], rax 0x5f: test qword ptr [rdx], -9 0x66: jne 0x3ffffd5 ``` Make sure to `brew install pkg-config capstone`
2021-01-22 18:43:26 +00:00
end