Attachment 341976 Details for Bug 186293 – patch

[patch] patch

b-backup.diff (text/plain), 8.51 KB, created by Saam Barati on 2018-06-05 11:00:04 PDT

(hide)

Description:

Filename:

MIME Type:

Creator: Saam Barati

Created: 2018-06-05 11:00:04 PDT

Size: 8.51 KB

patch

obsolete

>Index: Tools/ChangeLog
>===================================================================
>--- Tools/ChangeLog	(revision 232513)
>+++ Tools/ChangeLog	(working copy)
>@@ -1,3 +1,20 @@
>+2018-06-05  Saam Barati  <sbarati@apple.com>
>+
>+        run-testmem should have a dry run option and an option to parse the stdout of executing the dry run
>+        https://bugs.webkit.org/show_bug.cgi?id=186293
>+
>+        Reviewed by NOBODY (OOPS!).
>+
>+        This makes it easier to execute run-testmem on a device that doesn't have ruby
>+        on it. run-testmem now has a --dry option that will output commands to execute
>+        in bash. To run on a device that doesn't have ruby, just put its output into
>+        a bash script, and copy the PerformanceTests/testmem directory onto that device,
>+        and execute the bash script. Running that bash script will give you raw output.
>+        Make a file with that output and pass it into run-testmem using the --parse option.
>+        run-testmem will parse the raw output and compute the benchmark score.
>+
>+        * Scripts/run-testmem:
>+
> 2018-06-04  Frederic Wang  <fwang@igalia.com>
> 
>         import-w3c-tests should rely on <meta name="flags"> to detect CSS manual tests
>Index: Tools/Scripts/run-testmem
>===================================================================
>--- Tools/Scripts/run-testmem	(revision 232486)
>+++ Tools/Scripts/run-testmem	(working copy)
>@@ -31,10 +31,13 @@ require 'getoptlong'
> 
> def usage
>     puts "run-testmem [options]"
>-    puts "--build-dir  (-b)   pass in a path to your build directory, e.g, WebKitBuild/Release"
>-    puts "--verbose    (-v)   print more information as the benchmark runs"
>-    puts "--count      (-c)   number of outer iterations to run the benchmark for"
>-    puts "--help       (-h)   print this message"
>+    puts "--build-dir   (-b)   pass in a path to your build directory, e.g, WebKitBuild/Release"
>+    puts "--verbose     (-v)   print more information as the benchmark runs"
>+    puts "--count       (-c)   number of outer iterations to run the benchmark for"
>+    puts "--dry         (-d)   print shell output that can be run as a bash script on a different device. When using this option, provide the --script-path and --build-dir options."
>+    puts "--script-path (-s)   The path to the directory where you expect the testmem tests to live. Use this when doing a dry run with --dry."
>+    puts "--parse       (-p)   After executing the dry run, capture its stdout and write it to a file. Pass the path to that file for this option and run-testmem will compute the results of the benchmark run."
>+    puts "--help        (-h)   print this message"
> end
> 
> THIS_SCRIPT_PATH = Pathname.new(__FILE__).realpath
>@@ -43,10 +46,16 @@ SCRIPTS_PATH = THIS_SCRIPT_PATH.dirname
> $buildDir = nil
> $verbose = false
> $outerIterations = 3
>+$dryRun = false
>+$scriptPath = nil
>+$parsePath = nil
> 
> GetoptLong.new(["--build-dir", "-b", GetoptLong::REQUIRED_ARGUMENT],
>                ["--verbose", "-v", GetoptLong::NO_ARGUMENT],
>                ["--count", "-c", GetoptLong::REQUIRED_ARGUMENT],
>+               ["--dry", "-d", GetoptLong::NO_ARGUMENT],
>+               ["--script-path", "-s", GetoptLong::REQUIRED_ARGUMENT],
>+               ["--parse", "-p", GetoptLong::REQUIRED_ARGUMENT],
>                ["--help", "-h", GetoptLong::NO_ARGUMENT],
>                ).each {
>     | opt, arg |
>@@ -61,12 +70,23 @@ GetoptLong.new(["--build-dir", "-b", Get
>             puts "--count must be > 0"
>             exit 1
>         end
>+    when "--dry"
>+        $dryRun = true
>+    when "--script-path"
>+        $scriptPath = arg
>+    when "--parse"
>+        $parsePath = arg
>     when "--help"
>         usage
>         exit 1
>     end
> }
> 
>+if $scriptPath && !$dryRun
>+    puts "--script-path is only supported when you are doing a --dry run"
>+    exit 1
>+end
>+
> def getBuildDirectory
>     if $buildDir != nil
>         return $buildDir
>@@ -89,7 +109,7 @@ end
> 
> def getTestmemPath
>     path = Pathname.new(getBuildDirectory).join("testmem").to_s
>-    if !File.exists?(path)
>+    if !File.exists?(path) && !$dryRun
>         puts "Error: no testmem binary found in <build>/Release"
>         exit 1
>     end
>@@ -115,20 +135,30 @@ def getTests
>         | filename |
>         next unless filename =~ /\.js$/
>         filePath = dirPath.join(filename).to_s
>+        filePath = Pathname.new($scriptPath).join(filename).to_s if $scriptPath
>         files.push([filePath, iterationCount(filePath)])
>     }
> 
>     files.sort_by { | (path) | File.basename(path) }
> end
> 
>-def geomean(arr)
>-    score = arr.inject(1.0, :*)
>-    score ** (1.0 / arr.length)
>-end
>+def processRunOutput(stdout, path)
>+    time, peakFootprint, footprintAtEnd = stdout.split("\n")
>+    raise unless time.slice!("time:")
>+    raise unless peakFootprint.slice!("peak footprint:")
>+    raise unless footprintAtEnd.slice!("footprint at end:")
>+    time = time.to_f
>+    peakFootprint = peakFootprint.to_f
>+    footprintAtEnd = footprintAtEnd.to_f
> 
>-def mean(arr)
>-    sum = arr.inject(0.0, :+)
>-    sum / arr.length
>+    if $verbose
>+        puts path
>+        puts "time: #{time}"
>+        puts "peak footprint: #{peakFootprint/1024/1024} MB"
>+        puts "end footprint:  #{footprintAtEnd/1024/1024} MB\n"
>+    end
>+
>+    {"time"=>time, "peak"=>peakFootprint, "end"=>footprintAtEnd}
> end
> 
> def runTest(path, iters)
>@@ -138,6 +168,16 @@ def runTest(path, iters)
>         "JSC_useJIT" => "false",
>         "JSC_useRegExpJIT" => "false",
>     }
>+
>+    if $dryRun
>+        environment.each { | key, value |
>+            command = "#{key}=#{value} #{command}"
>+        }
>+        puts "echo \"#{command}\""
>+        puts command
>+        return
>+    end
>+
>     stdout, stderr, exitCode = Open3.capture3(environment, command)
> 
>     if $verbose
>@@ -152,39 +192,20 @@ def runTest(path, iters)
>         exit 1
>     end
> 
>-    time, peakFootprint, footprintAtEnd = stdout.split("\n")
>-    raise unless time.slice!("time:")
>-    raise unless peakFootprint.slice!("peak footprint:")
>-    raise unless footprintAtEnd.slice!("footprint at end:")
>-    time = time.to_f
>-    peakFootprint = peakFootprint.to_f
>-    footprintAtEnd = footprintAtEnd.to_f
>-
>-    if $verbose
>-        puts path
>-        puts "time: #{time}"
>-        puts "peak footprint: #{peakFootprint/1024/1024} MB"
>-        puts "end footprint:  #{footprintAtEnd/1024/1024} MB\n"
>-    end
>-
>-    {"time"=>time, "peak"=>peakFootprint, "end"=>footprintAtEnd}
>+    processRunOutput(stdout, path)
> end
> 
>-def run
>-    tests = getTests
>-    scores = {}
>-    tests.each { | (path) | scores[path] = [] }
>-    count = $outerIterations
>-    (0..(count-1)).each { | currentIter |
>-        tests.each { | (path, iters) | 
>-            statusToPrint = "iteration #{currentIter + 1}: #{File.basename(path, ".js")}"
>-            print "#{statusToPrint}\r"
>-            scores[path].push(runTest(path, iters))
>-            print "#{" ".rjust(statusToPrint.length)}\r"
>+def geomean(arr)
>+    score = arr.inject(1.0, :*)
>+    score ** (1.0 / arr.length)
>+end
> 
>-        }
>-    }
>+def mean(arr)
>+    sum = arr.inject(0.0, :+)
>+    sum / arr.length
>+end
> 
>+def processScores(scores)
>     peakScore = []
>     endScore = []
>     timeScore = []
>@@ -216,4 +237,56 @@ def run
>     puts JSON.pretty_generate(scores) if $verbose
> end
> 
>-run
>+def run
>+    tests = getTests
>+    scores = {}
>+    tests.each { | (path) | scores[path] = [] }
>+    count = $outerIterations
>+
>+    if $dryRun
>+        (0..(count-1)).each { | currentIter |
>+            tests.each { | (path, iters) | 
>+                runTest(path, iters)
>+            }
>+        }
>+        return
>+    end
>+
>+    (0..(count-1)).each { | currentIter |
>+        tests.each { | (path, iters) | 
>+            statusToPrint = "iteration #{currentIter + 1}: #{File.basename(path, ".js")}"
>+            print "#{statusToPrint}\r"
>+            scores[path].push(runTest(path, iters))
>+            print "#{" ".rjust(statusToPrint.length)}\r"
>+
>+        }
>+    }
>+
>+    processScores(scores)
>+end
>+
>+def parseResultOfDryRun(path)
>+    contents = IO.read(path).split("\n")
>+    if !contents.length || contents.length % 4 != 0
>+        puts "Bad input, expect multiple of 4 number of lines from output of running the result of --dry"
>+        exit 1
>+    end
>+
>+    scores = {}
>+    i = 0
>+    while i < contents.length
>+        path = contents[i + 0].split(" ")[-2]
>+        scores[path] = [] if !scores[path]
>+        stdout = [contents[i + 1], contents[i + 2], contents[i + 3]].join("\n")
>+        scores[path].push(processRunOutput(stdout, path))
>+        i += 4
>+    end
>+
>+    processScores(scores)
>+end
>+
>+if $parsePath
>+    parseResultOfDryRun($parsePath)
>+else
>+    run
>+end

Flags:

mark.lam: review+

Actions: View | Formatted Diff | Diff

Attachments on bug 186293: 341976 | 341978