vernier 0.5.1 → 0.7.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: da42588d9f2409a9d82a79439014490ecd81f00cb74c975267d6568a6e343c46
-  data.tar.gz: 921543f2b8cfec4e5803c4a47c8953df1acff567b88ddce2e1a65faad9e7c895
+  metadata.gz: 4988794520691870cb34178385be791e28fa4806eb1f6934b0d0abf52c8a070c
+  data.tar.gz: b1e63c6f5398f61fb68d634f17f33dc6aa7461344cb8acf446dd75ca52395818
 SHA512:
-  metadata.gz: a674536e944ef26f2db5893821d9d643fc8f651d45b7a72923ae9ab7c01f270431f0e5fadea76e3bb3feeabe30edde203b520310cbbbae6a53f03edb148db570
-  data.tar.gz: 1f39e9f04a2fd1c80525a72ce27aca6e8d614dfa9a4ba9f9721beff08f11a41431f6690c6215f0592dcc7923e49e516a4a0ddb812b493820877ceca81b4373ab
+  metadata.gz: '082561b2a381f88c540e607d1dd9c0d42d841a0de82fc61471fb2d67754ee14a4fd1e87fa71896b586077664e61fd1db623fd250cd13064d198c9dc91d3f13ed'
+  data.tar.gz: a28c197d405e4f52d999db3ae73187047886e1fcfb7d353d13f26ad5ba4b61f0675daf617119b37d72ad3760bd0dffcae234f2dcb0a0e960a60046dfb87dd784

data/Gemfile

@@ -8,5 +8,6 @@ gemspec
 gem "rake", "~> 13.0"
 
 gem "rake-compiler"
+gem "benchmark-ips"
 
 gem "minitest", "~> 5.0"

data/README.md

@@ -2,6 +2,8 @@
 
 Next-generation Ruby 3.2.1+ sampling profiler. Tracks multiple threads, GVL activity, GC pauses, idle time, and more.
 
+<img width="500" alt="Screenshot 2024-02-29 at 22 47 43" src="https://github.com/jhawthorn/vernier/assets/131752/aa995a41-d74f-405f-8ada-2522dd72c2c8">
+
 ## Examples
 
 [Livestreamed demo: Pairin' with Aaron (YouTube)](https://www.youtube.com/watch?v=9nvX3OHykGQ#t=27m43)
@@ -28,17 +30,33 @@ gem 'vernier'
 
 ## Usage
 
+The output can be viewed in the web app at https://vernier.prof or locally using the [`profile-viewer` gem](https://github.com/tenderlove/profiler/tree/ruby) (both are lightly customized versions of the firefox profiler frontend, which profiles are also compatible with).
+
+- **Flame Graph**: Shows proportionally how much time is spent within particular stack frames. Frames are grouped together, which means that x-axis / left-to-right order is not meaningful.
+- **Stack Chart**: Shows the stack at each sample with the x-axis representing time and can be read left-to-right.
+
 
 ### Time
 
+
+#### Command line
+
+The easiest way to record a program or script is via the CLI
+
 ```
-Vernier.trace(out: "time_profile.json") { some_slow_method }
+$ vernier run -- ruby -e 'sleep 1'
+starting profiler with interval 500
+#<Vernier::Result 1.001589 seconds, 1 threads, 2002 samples, 1 unique>
+written to /tmp/profile20240328-82441-gkzffc.vernier.json
 ```
 
-The output can then be viewed in the Firefox Profiler (demo) or the [`profile-viewer` gem](https://github.com/tenderlove/profiler/tree/ruby) (a Ruby-customized version of the firefox profiler.
+### Block of code
 
-- **Flame Graph**: Shows proportionally how much time is spent within particular stack frames. Frames are grouped together, which means that x-axis / left-to-right order is not meaningful.
-- **Stack Chart**: Shows the stack at each sample with the x-axis representing time and can be read left-to-right.
+``` ruby
+Vernier.run(out: "time_profile.json") do
+  some_slow_method
+end
+```
 
 ### Retained memory
 

data/examples/gvl_sleep.rb

@@ -0,0 +1,62 @@
+# Different (bad) ways to sleep
+
+File.write("#{__dir__}/my_sleep.c", <<~EOF)
+#include <time.h>
+#include <sys/errno.h>
+
+void my_sleep() {
+  struct timespec ts;
+  ts.tv_sec = 1;
+  ts.tv_nsec = 0;
+
+  int rc;
+  do {
+    rc = nanosleep(&ts, &ts);
+  } while (rc < 0 && errno == EINTR);
+}
+EOF
+
+soext = RbConfig::CONFIG["SOEXT"]
+system("gcc", "-shared", "-fPIC", "#{__dir__}/my_sleep.c", "-o", "#{__dir__}/my_sleep.#{soext}")
+
+require "fiddle"
+
+SLEEP_LIB = Fiddle.dlopen("./my_sleep.#{soext}")
+
+def cfunc_sleep_gvl
+  Fiddle::Function.new(SLEEP_LIB['my_sleep'], [], Fiddle::TYPE_VOID, need_gvl: true).call
+end
+
+def cfunc_sleep_idle
+  Fiddle::Function.new(SLEEP_LIB['my_sleep'], [], Fiddle::TYPE_VOID, need_gvl: true).call
+end
+
+def ruby_sleep_gvl
+  target = Process.clock_gettime(Process::CLOCK_MONOTONIC, :millisecond) + 1000
+  while Process.clock_gettime(Process::CLOCK_MONOTONIC, :millisecond) < target
+    i = 0
+    while i < 1000
+      i += 1
+    end
+  end
+end
+
+def sleep_idle
+  sleep 1
+end
+
+def run(name)
+  STDOUT.print "#{name}..."
+  STDOUT.flush
+
+  before = Process.clock_gettime(Process::CLOCK_MONOTONIC)
+  send(name)
+  after = Process.clock_gettime(Process::CLOCK_MONOTONIC)
+
+  STDOUT.puts " %.2fs" % (after - before)
+end
+
+run(:cfunc_sleep_gvl)
+run(:cfunc_sleep_idle)
+run(:ruby_sleep_gvl)
+run(:sleep_idle)

data/examples/measure_overhead.rb

@@ -0,0 +1,39 @@
+require "benchmark/ips"
+require "vernier"
+
+def compare(**options, &block)
+  block.call
+
+  Benchmark.ips do |x|
+    x.report "no profiler" do |n|
+      n.times do
+        block.call
+      end
+    end
+
+    x.report "vernier" do |n|
+      Vernier.profile(**options) do
+        n.times do
+          block.call
+        end
+      end
+    end
+
+    x.compare!
+  end
+end
+
+compare do
+  i = 0
+  while i < 10_000
+    i += 1
+  end
+end
+
+compare(allocation_sample_rate: 1000) do
+  Object.new
+end
+
+compare(allocation_sample_rate: 1) do
+  Object.new
+end

data/examples/rails.rb

@@ -59,12 +59,17 @@ end
 
 silence do
   ActiveRecord::Schema.define do
+    create_table :users, force: true do |t|
+      t.string :login
+    end
     create_table :posts, force: true do |t|
+      t.belongs_to :user
       t.string :title
       t.text :body
       t.integer :likes
     end
     create_table :comments, force: true do |t|
+      t.belongs_to :user
       t.belongs_to :post
       t.string :title
       t.text :body
@@ -73,25 +78,33 @@ silence do
   end
 end
 
+class User < ActiveRecord::Base
+end
+
 class Post < ActiveRecord::Base
   has_many :comments
+  belongs_to :user
 end
 
 class Comment < ActiveRecord::Base
   belongs_to :post
+  belongs_to :user
 end
 
+users = 0.upto(100).map do |i|
+  User.create!(login: "user#{i}")
+end
 0.upto(100) do |i|
-  post = Post.create!(title: "Post number #{i}", body: "blog " * 50, likes: ((i * 1337) % 30))
+  post = Post.create!(title: "Post number #{i}", body: "blog " * 50, likes: ((i * 1337) % 30), user: users.sample)
   5.times do
-    post.comments.create!(post: post, title: "nice post!", body: "keep it up!", posted_at: Time.now)
+    post.comments.create!(post: post, title: "nice post!", body: "keep it up!", posted_at: Time.now, user: users.sample)
   end
 end
 
 class HomeController < ActionController::Base
   def show
-    posts = Post.order(likes: :desc).includes(:comments).first(10)
-    render json: posts
+    posts = Post.order(likes: :desc).includes(:user, comments: :user).first(10)
+    render json: posts, include: [:user, comments: { include: :user } ]
   end
 end
 
@@ -110,15 +123,7 @@ end
 # warm up
 make_request.call
 
-Vernier.trace(out: "rails.json") do |collector|
-  ActiveSupport::Notifications.monotonic_subscribe do |name, start, finish, id, payload|
-    collector.add_marker(
-      name:,
-      start: (start * 1_000_000_000).to_i,
-      finish: (finish * 1_000_000_000).to_i,
-    )
-  end
-
+Vernier.trace(out: "rails.json", hooks: [:rails], allocation_sample_rate: 100) do |collector|
   1000.times do
     make_request.call
   end

data/examples/threaded_http_requests.rb

@@ -1,6 +1,6 @@
 require "vernier"
 
-Vernier.trace(out: "http_requests.json") do
+Vernier.trace(out: "http_requests.json", allocation_sample_rate: 100) do
 
 require "net/http"
 require "uri"

data/exe/vernier

@@ -16,6 +16,9 @@ parser = OptionParser.new(banner) do |o|
   o.on('--interval [MICROSECONDS]', Integer, "sampling interval (default 500)") do |i|
     options[:interval] = i
   end
+  o.on('--allocation_sample_rate [ALLOCATIONS]', Integer, "allocation sampling interval (default 0 disabled)") do |i|
+    options[:allocation_sample_rate] = i
+  end
   o.on('--signal [NAME]', String, "specify a signal to start and stop the profiler") do |s|
     options[:signal] = s
   end