naiveproxy/tools/binary_size/README.md
2018-08-14 22:19:20 +00:00

12 KiB

Tools for Analyzing Chrome's Binary Size

These tools currently focus on supporting Android. They somewhat work with Linux builds. As for Windows, some great tools already exist and are documented here:

There is also a dedicated mailing-list for binary size discussions:

Bugs are tracked here:

[TOC]

diagnose_bloat.py

Determine the cause of binary size bloat between two commits. Works for Android and Linux (although Linux symbol diffs have issues, as noted below).

How it Works

  1. Builds multiple revisions using release GN args.
    • Default is to build just two revisions (before & after commit)
    • Rather than building, can fetch build artifacts and .size files from perf bots (--cloud)
  2. Measures all outputs using resource_size.py and supersize.
  3. Saves & displays a breakdown of the difference in binary sizes.

Example Usage

# Build and diff monochrome_public_apk HEAD^ and HEAD.
tools/binary_size/diagnose_bloat.py HEAD -v

# Build and diff monochrome_apk HEAD^ and HEAD.
tools/binary_size/diagnose_bloat.py HEAD --enable-chrome-android-internal -v

# Build and diff monochrome_public_apk HEAD^ and HEAD without is_official_build.
tools/binary_size/diagnose_bloat.py HEAD --gn-args="is_official_build=false" -v

# Diff BEFORE_REV and AFTER_REV using build artifacts downloaded from perf bots.
tools/binary_size/diagnose_bloat.py AFTER_REV --reference-rev BEFORE_REV --cloud -v

# Fetch a .size, libmonochrome.so, and MonochromePublic.apk from perf bots (Googlers only):
tools/binary_size/diagnose_bloat.py AFTER_REV --cloud --unstripped --single

# Build and diff all contiguous revs in range BEFORE_REV..AFTER_REV for src/v8.
tools/binary_size/diagnose_bloat.py AFTER_REV --reference-rev BEFORE_REV --subrepo v8 --all -v

# Display detailed usage info (there are many options).
tools/binary_size/diagnose_bloat.py -h

Super Size

Collect, archive, and analyze Chrome's binary size. Supports Android and Linux (although Linux has issues).

.size files are archived on perf builders so that regressions can be quickly analyzed (via diagnose_bloat.py --cloud).

.size files are archived on official builders so that symbols can be diff'ed between milestones.

Technical Details

What's in a .size File?

.size files are gzipped plain text files that contain:

  1. A list of section sizes, including:
    • .so sections as reported by readelf -S
    • .pak and .dex sections for apk files
  2. Metadata (apk size, GN args, filenames, timestamps, git revision, build id),
  3. A list of symbols, including name, address, size, padding (caused by alignment), and associated source/object files.

How are Symbols Collected?

Native Symbols
  1. Symbol list is Extracted from linker .map file.
    • Map files contain some unique pieces of information compared to nm output, such as ** merge strings entries, and some unnamed symbols (which although unnamed, contain the .o path).
  2. .o files are mapped to .cc files by parsing .ninja files.
    • This means that .h files are never listed as sources. No information about inlined symbols is gathered.
  3. ** merge strings symbols are further broken down into individual string literal symbols. This is done by reading string literals from .o files, and then searching for them within the ** merge strings sections.
  4. Symbol aliases:
    • Aliases have the same address and size, but report their .pss as .size / .num_aliases.
    • Type 1: Different names. Caused by identical code folding.
      • These are collected from debug information via nm elf-file.
    • Type 2: Same names, different paths. Caused by inline functions defined in .h files.
      • These are collected by running nm on each .o file.
      • Normally represented using one alias per path, but are sometimes collapsed into a single symbol with a path of {shared}/$SYMBOL_COUNT. This collapsing is done only for symbols owned by a large number of paths.
    • Type 3: String literals that are de-duped at link-time.
      • These are found as part of the string literal extraction process.
Pak Symbols
  1. Grit creates a mapping between numeric id and textual id for grd files.
    • A side effect of pak whitelist generation is a mapping of .cc to numeric id.
    • A complete per-apk mapping of numeric id to textual id is stored in the output_dir/size-info dir.
  2. supersize uses these two mappings to find associated source files for the pak entries found in all of the apk's .pak files.
    • Pak entries with the same name are merged into a single symbol.
      • This is the case of pak files for translations.
    • The original grd file paths are stored in the full name of each symbol.
Dex Symbols
  1. Java compile targets create a mapping between java fully qualified names (FQN) and source files.
    • For .java files the FQN of the public class is mapped to the file.
    • For .srcjar files the FQN of the public class is mapped to the .srcjar file path.
    • A complete per-apk class FQN to source mapping is stored in the output_dir/size-info dir.
  2. The apkanalyzer sdk tool is used to find the size and FQN of entries in the dex file.
    • If a proguard .mapping file is available, that is used to get back the original FQN.
  3. The output from apkanalyzer is used by supersize along with the mapping file to find associated source files for the dex entries found in all of the apk's .dex files.
Common Symbols
  1. Shared bytes are stored in symbols with names starting with Overhead: .
    • Elf file, dex file, pak files, apk files all have compression overhead.
    • These are treated as padding-only symbols to de-emphasize them in diffs.
    • It is expected that these symbols have minor fluctuations since they are affected by changes in compressibility.
  2. All other files in an apk have one symbol each under the .other section with their corresponding path in the apk as their associated path.

What Other Processing Happens?

  1. Path normalization:

    • Prefixes are removed: out/Release/, gen/, obj/
    • Archive names made more pathy: foo/bar.a(baz.o) -> foo/bar.a/baz.o
    • Shared symbols do not store the complete source paths. Instead, the common ancestor is computed and stored as the path.
      • Example: base/{shared}/3 (the "3" means three different files contain the symbol)
  2. Name normalization:

    • (anonymous::) is removed from names (and stored as a symbol flag).
    • [clone] suffix removed (and stored as a symbol flag).
    • vtable for FOO -> Foo [vtable]
    • Mangling done by linkers is undone (e.g. prefixing with "unlikely.")
    • Names are processed into:
      • name: Name without template and argument parameters
      • template_name: Name without argument parameters.
      • full_name: Name with all parameters.
  3. Clustering:

    • Compiler & linker optimizations can cause symbols to be broken into multiple parts to become candidates for inlining ("partial inlining").
    • These symbols are sometimes suffixed with "[clone]" (removed by normalization).
    • Clustering creates groups containing all pieces of a symbol (in the case where multiple pieces remain after inlining).
    • Clustering is done by default on SizeInfo.symbols. To view unclustered symbols, use SizeInfo.raw_symbols.
  4. Diffing:

    • Some heuristics for matching up before/after symbols.
  5. Simulated compression:

    • Only some .pak files are compressed and others are kept uncompressed.
    • To get a reasonable idea of actual impact to final apk size, we use a constant compression factor for all the compressed .pak files.
      • This prevents swings in compressed sizes for all symbols when new entries are added or old entries are removed.
      • The constant is chosen so that it minimizes overall discrepancy with actual total compressed sizes.

Is Super Size a Generic Tool?

No. Most of the logic is would could work for any ELF executable. However, being a generic tool is not a goal. Some examples of existing Chrome-specific logic:

  • Assumes .ninja build rules are available.
  • Heuristic for locating .so given .apk.
  • Requires size-info dir in output directory to analyze .pak and .dex files.

Usage: archive

Collect size information and dump it into a .size file.

*** note Note: Refer to diagnose_bloat.py for list of GN args to build a Release binary (or just use the tool with --single).

Googlers: If you just want a .size for a commit on master:

GIT_REV="HEAD~200"
tools/binary_size/diagnose_bloat.py --single --cloud --unstripped $GIT_REV

Example Usage:

# Android:
ninja -C out/Release -j 1000 apks/ChromePublic.apk
tools/binary_size/supersize archive chrome.size --apk-file out/Release/apks/ChromePublic.apk -v

# Linux:
ninja -C out/Release -j 1000 chrome
tools/binary_size/supersize archive chrome.size --elf-file out/Release/chrome -v

Usage: html_report

Creates an interactive size breakdown (by source path) as a stand-alone html report.

Example output: https://notwoods.github.io/chrome-supersize-reports/

Example Usage:

# Creates the data file ./report.ndjson, generated based on ./chrome.size
tools/binary_size/supersize html_report chrome.size --report-file report.ndjson -v

# Includes every symbol in the data file, although it will take longer to load.
tools/binary_size/supersize html_report chrome.size --report-file report.ndjson --all-symbols

# Create a data file showing a diff between two .size files.
tools/binary_size/supersize html_report after.size --diff-with before.size --report-file report.ndjson

Usage: start_server

Locally view the data file generated by html_report, by starting a web server that links to a data file.

Example Usage:

# Starts a local server to view the data in ./report.ndjson
tools/binary_size/supersize start_server report.ndjson

# Set a custom address and port.
tools/binary_size/supersize start_server report.ndjson -a localhost -p 8080

Usage: diff

A convenience command equivalent to: console before.size after.size --query='Print(Diff(size_info1, size_info2))'

Example Usage:

tools/binary_size/supersize diff before.size after.size --all

Usage: console

Starts a Python interpreter where you can run custom queries, or run pre-made queries from canned_queries.py.

Example Usage:

# Prints size infomation and exits (does not enter interactive mode).
tools/binary_size/supersize console chrome.size --query='Print(size_info)'

# Enters a Python REPL (it will print more guidance).
tools/binary_size/supersize console chrome.size

Example session:

>>> ShowExamples()  # Get some inspiration.
...
>>> sorted = size_info.symbols.WhereInSection('t').Sorted()
>>> Print(sorted)  # Have a look at the largest symbols.
...
>>> sym = sorted.WhereNameMatches('TrellisQuantizeBlock')[0]
>>> Disassemble(sym)  # Time to learn assembly.
...
>>> help(canned_queries)
...
>>> Print(canned_queries.TemplatesByName(depth=-1))
...
>>> syms = size_info.symbols.WherePathMatches(r'skia').Sorted()
>>> Print(syms, verbose=True)  # Show full symbol names with parameter types.
...
>>> # Dump all string literals from skia files to "strings.txt".
>>> Print((t[1] for t in ReadStringLiterals(syms)), to_file='strings.txt')

Roadmap

  1. Better Linux support (clang+lld+lto vs gcc+gold).
  2. More console features:
    • Add SplitByName() - Like GroupByName(), but recursive.
    • A canned query, that does what ShowGlobals does (as described in Windows Binary Sizes).
  3. Integrate with resource_sizes.py so that it tracks size of major components separately: chrome vs blink vs skia vs v8.
  4. Add dependency graph info, perhaps just on a per-file basis.
    • No idea how to do this, but Windows can do it via tools\win\linker_verbose_tracking.py