diff --git a/.gitignore b/.gitignore
index 3e759b75..bc3df510 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1,330 +1,7 @@
-## Ignore Visual Studio temporary files, build results, and
-## files generated by popular Visual Studio add-ons.
-##
-## Get latest from https://github.com/github/gitignore/blob/master/VisualStudio.gitignore
-
-# User-specific files
-*.suo
-*.user
-*.userosscache
-*.sln.docstates
-
-# User-specific files (MonoDevelop/Xamarin Studio)
-*.userprefs
-
-# Build results
-[Dd]ebug/
-[Dd]ebugPublic/
-[Rr]elease/
-[Rr]eleases/
-x64/
-x86/
-bld/
-[Bb]in/
-[Oo]bj/
-[Ll]og/
-
-# Visual Studio 2015/2017 cache/options directory
-.vs/
-# Uncomment if you have tasks that create the project's static files in wwwroot
-#wwwroot/
-
-# Visual Studio 2017 auto generated files
-Generated\ Files/
-
-# MSTest test Results
-[Tt]est[Rr]esult*/
-[Bb]uild[Ll]og.*
-
-# NUNIT
-*.VisualState.xml
-TestResult.xml
-
-# Build Results of an ATL Project
-[Dd]ebugPS/
-[Rr]eleasePS/
-dlldata.c
-
-# Benchmark Results
-BenchmarkDotNet.Artifacts/
-
-# .NET Core
-project.lock.json
-project.fragment.lock.json
-artifacts/
-**/Properties/launchSettings.json
-
-# StyleCop
-StyleCopReport.xml
-
-# Files built by Visual Studio
-*_i.c
-*_p.c
-*_i.h
-*.ilk
-*.meta
-*.obj
-*.iobj
-*.pch
-*.pdb
-*.ipdb
-*.pgc
-*.pgd
-*.rsp
-*.sbr
-*.tlb
-*.tli
-*.tlh
-*.tmp
-*.tmp_proj
-*.log
-*.vspscc
-*.vssscc
-.builds
-*.pidb
-*.svclog
-*.scc
-
-# Chutzpah Test files
-_Chutzpah*
-
-# Visual C++ cache files
-ipch/
-*.aps
-*.ncb
-*.opendb
-*.opensdf
-*.sdf
-*.cachefile
-*.VC.db
-*.VC.VC.opendb
-
-# Visual Studio profiler
-*.psess
-*.vsp
-*.vspx
-*.sap
-
-# Visual Studio Trace Files
-*.e2e
-
-# TFS 2012 Local Workspace
-$tf/
-
-# Guidance Automation Toolkit
-*.gpState
-
-# ReSharper is a .NET coding add-in
-_ReSharper*/
-*.[Rr]e[Ss]harper
-*.DotSettings.user
-
-# JustCode is a .NET coding add-in
-.JustCode
-
-# TeamCity is a build add-in
-_TeamCity*
-
-# DotCover is a Code Coverage Tool
-*.dotCover
-
-# AxoCover is a Code Coverage Tool
-.axoCover/*
-!.axoCover/settings.json
-
-# Visual Studio code coverage results
-*.coverage
-*.coveragexml
-
-# NCrunch
-_NCrunch_*
-.*crunch*.local.xml
-nCrunchTemp_*
-
-# MightyMoose
-*.mm.*
-AutoTest.Net/
-
-# Web workbench (sass)
-.sass-cache/
-
-# Installshield output folder
-[Ee]xpress/
-
-# DocProject is a documentation generator add-in
-DocProject/buildhelp/
-DocProject/Help/*.HxT
-DocProject/Help/*.HxC
-DocProject/Help/*.hhc
-DocProject/Help/*.hhk
-DocProject/Help/*.hhp
-DocProject/Help/Html2
-DocProject/Help/html
-
-# Click-Once directory
-publish/
-
-# Publish Web Output
-*.[Pp]ublish.xml
-*.azurePubxml
-# Note: Comment the next line if you want to checkin your web deploy settings,
-# but database connection strings (with potential passwords) will be unencrypted
-*.pubxml
-*.publishproj
-
-# Microsoft Azure Web App publish settings. Comment the next line if you want to
-# checkin your Azure Web App publish settings, but sensitive information contained
-# in these scripts will be unencrypted
-PublishScripts/
-
-# NuGet Packages
-*.nupkg
-# The packages folder can be ignored because of Package Restore
-**/[Pp]ackages/*
-# except build/, which is used as an MSBuild target.
-!**/[Pp]ackages/build/
-# Uncomment if necessary however generally it will be regenerated when needed
-#!**/[Pp]ackages/repositories.config
-# NuGet v3's project.json files produces more ignorable files
-*.nuget.props
-*.nuget.targets
-
-# Microsoft Azure Build Output
-csx/
-*.build.csdef
-
-# Microsoft Azure Emulator
-ecf/
-rcf/
-
-# Windows Store app package directories and files
-AppPackages/
-BundleArtifacts/
-Package.StoreAssociation.xml
-_pkginfo.txt
-*.appx
-
-# Visual Studio cache files
-# files ending in .cache can be ignored
-*.[Cc]ache
-# but keep track of directories ending in .cache
-!*.[Cc]ache/
-
-# Others
-ClientBin/
-~$*
-*~
-*.dbmdl
-*.dbproj.schemaview
-*.jfm
-*.pfx
-*.publishsettings
-orleans.codegen.cs
-
-# Including strong name files can present a security risk
-# (https://github.com/github/gitignore/pull/2483#issue-259490424)
-#*.snk
-
-# Since there are multiple workflows, uncomment next line to ignore bower_components
-# (https://github.com/github/gitignore/pull/1529#issuecomment-104372622)
-#bower_components/
-
-# RIA/Silverlight projects
-Generated_Code/
-
-# Backup & report files from converting an old project file
-# to a newer Visual Studio version. Backup files are not needed,
-# because we have git ;-)
-_UpgradeReport_Files/
-Backup*/
-UpgradeLog*.XML
-UpgradeLog*.htm
-ServiceFabricBackup/
-*.rptproj.bak
-
-# SQL Server files
-*.mdf
-*.ldf
-*.ndf
-
-# Business Intelligence projects
-*.rdl.data
-*.bim.layout
-*.bim_*.settings
-*.rptproj.rsuser
-
-# Microsoft Fakes
-FakesAssemblies/
-
-# GhostDoc plugin setting file
-*.GhostDoc.xml
-
-# Node.js Tools for Visual Studio
-.ntvs_analysis.dat
-node_modules/
-
-# Visual Studio 6 build log
-*.plg
-
-# Visual Studio 6 workspace options file
-*.opt
-
-# Visual Studio 6 auto-generated workspace file (contains which files were open etc.)
-*.vbw
-
-# Visual Studio LightSwitch build output
-**/*.HTMLClient/GeneratedArtifacts
-**/*.DesktopClient/GeneratedArtifacts
-**/*.DesktopClient/ModelManifest.xml
-**/*.Server/GeneratedArtifacts
-**/*.Server/ModelManifest.xml
-_Pvt_Extensions
-
-# Paket dependency manager
-.paket/paket.exe
-paket-files/
-
-# FAKE - F# Make
-.fake/
-
-# JetBrains Rider
-.idea/
-*.sln.iml
-
-# CodeRush
-.cr/
-
-# Python Tools for Visual Studio (PTVS)
-__pycache__/
-*.pyc
-
-# Cake - Uncomment if you are using it
-# tools/**
-# !tools/packages.config
-
-# Tabs Studio
-*.tss
-
-# Telerik's JustMock configuration file
-*.jmconfig
-
-# BizTalk build output
-*.btp.cs
-*.btm.cs
-*.odx.cs
-*.xsd.cs
-
-# OpenCover UI analysis results
-OpenCover/
-
-# Azure Stream Analytics local run output
-ASALocalRun/
-
-# MSBuild Binary and Structured Log
-*.binlog
-
-# NVidia Nsight GPU debugger configuration file
-*.nvuser
-
-# MFractors (Xamarin productivity tool) working folder
-.mfractor/
+ide/vs2017/*.db
+ide/vs2017/*.opendb
+ide/vs2017/*.user
+ide/vs2017/.vs
+out/
+docs/
+*.zip
diff --git a/doc/doxyfile b/doc/doxyfile
index 3fa43cf6..11d71667 100644
--- a/doc/doxyfile
+++ b/doc/doxyfile
@@ -1235,18 +1235,7 @@ HTML_EXTRA_STYLESHEET = mimalloc-doxygen.css
# files will be copied as-is; there are no commands or markers available.
# This tag requires that the tag GENERATE_HTML is set to YES.
-HTML_EXTRA_FILES = bench-r5a-4xlarge-t1.png \
- bench-r5a-4xlarge-t2.png \
- bench-r5a-4xlarge-m1.png \
- bench-r5a-4xlarge-m2.png \
- bench-c5d-2xlarge-t1.png \
- bench-c5d-2xlarge-t2.png \
- bench-c5d-2xlarge-m1.png \
- bench-c5d-2xlarge-m2.png \
- bench-z4-win-t1.png \
- bench-z4-win-t2.png \
- bench-z4-win-m1.png \
- bench-z4-win-m2.png
+HTML_EXTRA_FILES =
# The HTML_COLORSTYLE_HUE tag controls the color of the HTML output. Doxygen
# will adjust the colors in the style sheet and background images according to
diff --git a/doc/mimalloc-doc.h b/doc/mimalloc-doc.h
index c3e41372..a8a22000 100644
--- a/doc/mimalloc-doc.h
+++ b/doc/mimalloc-doc.h
@@ -11,13 +11,19 @@ terms of the MIT license. A copy of the license can be found in the file
/*! \mainpage
This is the API documentation of the
-[mimalloc](https://github.com/koka-lang/mimalloc) allocator
+[mimalloc](https://github.com/microsoft/mimalloc) allocator
(pronounced "me-malloc") -- a
general purpose allocator with excellent [performance](bench.html)
characteristics. Initially
developed by Daan Leijen for the run-time systems of the
[Koka](https://github.com/koka-lang/koka) and [Lean](https://github.com/leanprover/lean) languages.
+It is a drop-in replacement for `malloc` and can be used in other programs
+without code changes, for example, on Unix you can use it as:
+```
+> LD_PRELOAD=/usr/bin/libmimalloc.so myprogram
+```
+
Notable aspects of the design include:
- __small and consistent__: the library is less than 3500 LOC using simple and
@@ -25,23 +31,32 @@ Notable aspects of the design include:
to integrate and adapt in other projects. For runtime systems it
provides hooks for a monotonic _heartbeat_ and deferred freeing (for
bounded worst-case times with reference counting).
-- __free list sharding__: "the big idea": instead of one big free list (per size class) we have
+- __free list sharding__: the big idea: instead of one big free list (per size class) we have
many smaller lists per memory "page" which both reduces fragmentation
and increases locality --
things that are allocated close in time get allocated close in memory.
- (A memory "page" in mimalloc contains blocks of one size class and is
- usually 64KB on a 64-bit system).
+ (A memory "page" in _mimalloc_ contains blocks of one size class and is
+ usually 64KiB on a 64-bit system).
- __eager page reset__: when a "page" becomes empty (with increased chance
due to free list sharding) the memory is marked to the OS as unused ("reset" or "purged")
reducing (real) memory pressure and fragmentation, especially in long running
programs.
-- __lazy initialization__: pages in a segment are lazily initialized so
- no memory is touched until it becomes allocated, reducing the resident
- memory and potential page faults.
+- __secure__: _mimalloc_ can be build in secure mode, adding guard pages,
+ randomized allocation, encrypted free lists, etc. to protect against various
+ heap vulnerabilities. The performance penalty is only around 3% on average
+ over our benchmarks.
+- __first-class heaps__: efficiently create and use multiple heaps to allocate across different regions.
+ A heap can be destroyed at once instead of deallocating each object separately.
- __bounded__: it does not suffer from _blowup_ \[1\], has bounded worst-case allocation
times (_wcat_), bounded space overhead (~0.2% meta-data, with at most 16.7% waste in allocation sizes),
- and has no internal points of contention using atomic operations almost
- everywhere.
+ and has no internal points of contention using only atomic operations.
+- __fast__: In our benchmarks (see [below](#performance)),
+ _mimalloc_ always outperforms all other leading allocators (_jemalloc_, _tcmalloc_, _Hoard_, etc),
+ and usually uses less memory (up to 25% more in the worst case). A nice property
+ is that it does consistently well over a wide range of benchmarks.
+
+You can read more on the design of _mimalloc_ in the upcoming technical report
+which also has detailed benchmark results.
Further information:
@@ -623,13 +638,13 @@ void mi_option_set_default(mi_option_t option, long value);
Checkout the sources from Github:
```
-git clone https://github.com/koka-lang/mimalloc.git
+git clone https://github.com/microsoft/mimalloc
```
## Windows
Open `ide/vs2017/mimalloc.sln` in Visual Studio 2017 and build.
-The `mimalloc` project builds a static library, while the
+The `mimalloc` project builds a static library (in `out/msvc-x64`), while the
`mimalloc-override` project builds a DLL for overriding malloc
in the entire program.
@@ -637,44 +652,50 @@ in the entire program.
We use [`cmake`](https://cmake.org)1 as the build system:
-- `mkdir -p out/release` (create a build directory)
-- `cd out/release` (go to it)
-- `cmake ../..` (generate the make file)
-- `make` (and build)
+```
+> mkdir -p out/release
+> cd out/release
+> cmake ../..
+> make
+```
+This builds the library as a shared (dynamic)
+library (`.so` or `.dylib`), a static library (`.a`), and
+as a single object file (`.o`).
- This will build the library as a shared (dynamic)
- library (`.so` or `.dylib`), a static library (`.a`), and
- as a single object file (`.o`).
-
-- `sudo make install` (install the library and header files in `/usr/lib` and `/usr/include`)
-
- Use the option `-DCMAKE_INSTALL_PREFIX=../local` (for example) to the `ccmake`
- command to install to a local directory to see what gets installed.
+`> sudo make install` (install the library and header files in `/usr/local/lib` and `/usr/local/include`)
You can build the debug version which does many internal checks and
maintains detailed statistics as:
-- `mkdir -p out/debug`
-- `cd out/debug`
-- `cmake -DCMAKE_BUILD_TYPE=Debug ../..`
-- `make`
-
- This will name the shared library as `libmimalloc-debug.so`.
-
-Or build with `clang`:
-
-- `CC=clang cmake ../..`
+```
+> mkdir -p out/debug
+> cd out/debug
+> cmake -DCMAKE_BUILD_TYPE=Debug ../..
+> make
+```
+This will name the shared library as `libmimalloc-debug.so`.
+Finally, you can build a _secure_ version that uses guard pages, encrypted
+free lists, etc, as:
+```
+> mkdir -p out/secure
+> cd out/secure
+> cmake -DSECURE=ON ../..
+> make
+```
+This will name the shared library as `libmimalloc-secure.so`.
Use `ccmake`2 instead of `cmake`
to see and customize all the available build options.
Notes:
1. Install CMake: `sudo apt-get install cmake`
2. Install CCMake: `sudo apt-get install cmake-curses-gui`
+
*/
/*! \page using Using the library
+
The preferred usage is including ``, linking with
the shared- or static library, and using the `mi_malloc` API exclusively for allocation. For example,
```
@@ -745,7 +766,7 @@ See \ref overrides for more info.
/*! \page overrides Overriding Malloc
-Overriding standard malloc can be done either _dynamically_ or _statically_.
+Overriding the standard `malloc` can be done either _dynamically_ or _statically_.
## Dynamic override
@@ -753,7 +774,7 @@ This is the recommended way to override the standard malloc interface.
### Unix, BSD, MacOSX
-On these system we preload the mimalloc shared
+On these systems we preload the mimalloc shared
library so all calls to the standard `malloc` interface are
resolved to the _mimalloc_ library.
@@ -770,7 +791,7 @@ env MIMALLOC_VERBOSE=1 LD_PRELOAD=/usr/lib/libmimalloc.so myprogram
```
or run with the debug version to get detailed statistics:
```
-env MIMALLOC_STATS=1 LD_PRELOAD=/usr/lib/libmimallocd.so myprogram
+env MIMALLOC_STATS=1 LD_PRELOAD=/usr/lib/libmimalloc-debug.so myprogram
```
### Windows
@@ -780,7 +801,7 @@ DLL, and use the C-runtime library as a DLL (the `/MD` or `/MDd` switch).
To ensure the mimalloc DLL gets loaded it is easiest to insert some
call to the mimalloc API in the `main` function, like `mi_version()`.
-Due to the way mimalloc overrides the standard malloc at runtime, it is best
+Due to the way mimalloc intercepts the standard malloc at runtime, it is best
to link to the mimalloc import library first on the command line so it gets
loaded right after the universal C runtime DLL (`ucrtbase`). See
the `mimalloc-override-test` project for an example.
@@ -788,9 +809,9 @@ the `mimalloc-override-test` project for an example.
## Static override
-You can also statically link with _mimalloc_ to override the standard
+On Unix systems, you can also statically link with _mimalloc_ to override the standard
malloc interface. The recommended way is to link the final program with the
-_mimalloc_ single object file (`mimalloc-override.o` (or `.obj`)). We use
+_mimalloc_ single object file (`mimalloc-override.o`). We use
an object file instead of a library file as linkers give preference to
that over archives to resolve symbols. To ensure that the standard
malloc interface resolves to the _mimalloc_ library, link it as the first
@@ -858,239 +879,19 @@ void _free_dbg(void* p, int block_type);
/*! \page bench Performance
+We tested _mimalloc_ against many other top allocators over a wide
+range of benchmarks, ranging from various real world programs to
+synthetic benchmarks that see how the allocator behaves under more
+extreme circumstances.
-tldr: In our benchmarks, mimalloc always outperforms
-all other leading allocators (jemalloc, tcmalloc, hoard, and glibc), and usually
-uses less memory (with less then 25% more in the worst case) (as of Jan 2019).
-A nice property is that it does consistently well over a wide range of benchmarks.
+In our benchmarks, _mimalloc_ always outperforms all other leading
+allocators (_jemalloc_, _tcmalloc_, _Hoard_, etc) (Apr 2019),
+and usually uses less memory (up to 25% more in the worst case).
+A nice property is that it does *consistently* well over the wide
+range of benchmarks.
-Disclaimer: allocators are interesting as there is no optimal algorithm -- for
-a given allocator one can always construct a workload where it does not do so well.
-The goal is thus to find an allocation strategy that performs well over a wide
-range of benchmarks without suffering from underperformance in less
-common situations (which is what our second benchmark set tests for).
-
-
-## Benchmarking
-
-We tested _mimalloc_ with 5 other allocators over 11 benchmarks.
-The tested allocators are:
-
-- **mi**: The mimalloc allocator (version tag `v1.0.0`).
-- **je**: [jemalloc](https://github.com/jemalloc/jemalloc), by [Jason Evans](https://www.facebook.com/notes/facebook-engineering/scalable-memory-allocation-using-jemalloc/480222803919) (Facebook);
- currently (2018) one of the leading allocators and is widely used, for example
- in BSD, Firefox, and at Facebook. Installed as package `libjemalloc-dev:amd64/bionic 3.6.0-11`.
-- **tc**: [tcmalloc](https://github.com/gperftools/gperftools), by Google as part of the performance tools.
- Highly performant and used in the Chrome browser. Installed as package `libgoogle-perftools-dev:amd64/bionic 2.5-2.2ubuntu3`.
-- **jx**: A compiled version of a more recent instance of [jemalloc](https://github.com/jemalloc/jemalloc).
- Using commit ` 7a815c1b` ([dev](https://github.com/jemalloc/jemalloc/tree/dev), 2019-01-15).
-- **hd**: [Hoard](https://github.com/emeryberger/Hoard), by Emery Berger \[1].
- One of the first multi-thread scalable allocators.
- ([master](https://github.com/emeryberger/Hoard), 2019-01-01, version tag `3.13`)
-- **mc**: The system allocator. Here we use the LibC allocator (which is originally based on
- PtMalloc). Using version 2.27. (Note that version 2.26 significantly improved scalability over
- earlier versions).
-
-All allocators run exactly the same benchmark programs and use `LD_PRELOAD` to override the system allocator.
-The wall-clock elapsed time and peak resident memory (_rss_) are
-measured with the `time` program. The best scores over 5 runs are used.
-Performance is reported relative to mimalloc, e.g. a time of 66% means that
-mimalloc ran 1.5× faster (i.e. that mimalloc finished in 66% of the time
-that the other allocator needed).
-
-## On a 16-core AMD EPYC running Linux
-
-Testing on a big Amazon EC2 instance ([r5a.4xlarge](https://aws.amazon.com/ec2/instance-types/))
-consisting of a 16-core AMD EPYC 7000 at 2.5GHz
-with 128GB ECC memory, running Ubuntu 18.04.1 with LibC 2.27 and GCC 7.3.0.
-
-
-The first benchmark set consists of programs that allocate a lot. Relative
-elapsed time:
-
-![bench-r5a-4xlarge-t1](bench-r5a-4xlarge-t1.png)
-
-and memory usage:
-
-![bench-r5a-4xlarge-m1](bench-r5a-4xlarge-m1.png)
-
-The benchmarks above are (with N=16 in our case):
-
-- __cfrac__: by Dave Barrett, implementation of continued fraction factorization:
- uses many small short-lived allocations. Factorizes as `./cfrac 175451865205073170563711388363274837927895`.
-- __espresso__: a programmable logic array analyzer \[3].
-- __barnes__: a hierarchical n-body particle solver \[4]. Simulates 163840 particles.
-- __leanN__: by Leonardo de Moura _et al_, the [lean](https://github.com/leanprover/lean)
- compiler, version 3.4.1, compiling its own standard library concurrently using N cores (`./lean --make -j N`).
- Big real-world workload with intensive allocation, takes about 1:40s when running on a
- single high-end core.
-- __redis__: running the [redis](https://redis.io/) 5.0.3 server on
- 1 million requests pushing 10 new list elements and then requesting the
- head 10 elements. Measures the requests handled per second.
-- __alloc-test__: a modern [allocator test](http://ithare.com/testing-memory-allocators-ptmalloc2-tcmalloc-hoard-jemalloc-while-trying-to-simulate-real-world-loads/)
- developed by by OLogN Technologies AG at [ITHare.com](http://ithare.com). Simulates intensive allocation workloads with a Pareto
- size distribution. The `alloc-testN` benchmark runs on N cores doing 100×106
- allocations per thread with objects up to 1KB in size.
- Using commit `94f6cb` ([master](https://github.com/node-dot-cpp/alloc-test), 2018-07-04)
-
-We can see mimalloc outperforms the other allocators moderately but all
-these modern allocators perform well.
-In `cfrac`, mimalloc is about 13%
-faster than jemalloc for many small and short-lived allocations.
-The `cfrac` and `espresso` programs do not use much
-memory (~1.5MB) so it does not matter too much, but still mimalloc uses about half the resident
-memory of tcmalloc (and almost 5× less than Hoard on `espresso`).
-
-_The `leanN` program is most interesting as a large realistic and concurrent
-workload and there is a 6% speedup over both tcmalloc and jemalloc. This is
-quite significant: if Lean spends (optimistically) 20% of its time in the allocator
-that means that mimalloc is 1.5× faster than the others._
-
-The `alloc-test` is very allocation intensive and we see the larger
-diffrerences here. Since all allocators perform almost identical on `alloc-test1`
-as `alloc-testN`, we can see that they are all excellent and scale (almost) linearly.
-
-The second benchmark set test specific aspects of the allocators and
-shows more extreme differences between allocators:
-
-![bench-r5a-4xlarge-t2](bench-r5a-4xlarge-t2.png)
-
-
-
-![bench-r5a-4xlarge-m2](bench-r5a-4xlarge-m2.png)
-
-The benchmarks in the second set are (again with N=16):
-
-- __larson__: by Larson and Krishnan \[2]. Simulates a server workload using 100
- separate threads where
- they allocate and free many objects but leave some objects to
- be freed by other threads. Larson and Krishnan observe this behavior
- (which they call _bleeding_) in actual server applications, and the
- benchmark simulates this.
-- __sh6bench__: by [MicroQuill](http://www.microquill.com) as part of SmartHeap. Stress test for
- single-threaded allocation where some of the objects are freed
- in a usual last-allocated, first-freed (LIFO) order, but others
- are freed in reverse order. Using the public [source](http://www.microquill.com/smartheap/shbench/bench.zip) (retrieved 2019-01-02)
-- __sh8bench__: by [MicroQuill](http://www.microquill.com) as part of SmartHeap. Stress test for
- multithreaded allocation (with N threads) where, just as in `larson`, some objects are freed
- by other threads, and some objects freed in reverse (as in `sh6bench`).
- Using the public [source](http://www.microquill.com/smartheap/SH8BENCH.zip) (retrieved 2019-01-02)
-- __cache-scratch__: by Emery Berger _et al_ \[1]. Introduced with the Hoard
- allocator to test for _passive-false_ sharing of cache lines: first some
- small objects are allocated and given to each thread; the threads free that
- object and allocate another one and access that repeatedly. If an allocator
- allocates objects from different threads close to each other this will
- lead to cache-line contention.
-
-In the `larson` server workload mimalloc is 2.5× faster than
-tcmalloc and jemalloc which is quite surprising -- probably due to the object
-migration between different threads. Also in `sh6bench` mimalloc does much
-better than the others (more than 4× faster than jemalloc). a
-We cannot explain this well but believe it may be
-caused in part by the "reverse" free-ing in `sh6bench`. Again in `sh8bench`
-the mimalloc allocator handles object migration between threads much better .
-
-The `cache-scratch` benchmark also demonstrates the different architectures
-of the allocators nicely. With a single thread they all perform the same, but when
-running with multiple threads the allocator induced false sharing of the
-cache lines causes large run-time differences, where mimalloc is up to
-20× faster than tcmalloc here. Only the original jemalloc does almost
-as well (but the most recent version, jxmalloc, regresses). The
-Hoard allocator is specifically designed to avoid this false sharing and we
-are not sure why it is not doing well here (although it runs still 5× as
-fast as tcmalloc and jxmalloc).
-
-
-## On a 8-core Intel Xeon running Linux
-
-Testing on a compute optimized Amazon EC2 instance ([c5d.2xlarge](https://aws.amazon.com/ec2/instance-types/))
-consisting of a 8-core Intel Xeon Platinum at 3GHz (up to 3.5GHz turbo boost)
-with 16GB ECC memory, running Ubuntu 18.04.1 with LibC 2.27 and GCC 7.3.0.
-
-First the regular workload benchmarks (with N=8):
-
-![bench-c5d-2xlarge-t1](bench-c5d-2xlarge-t1.png)
-
-
-
-![bench-c5d-2xlarge-m1](bench-c5d-2xlarge-m1.png)
-
-Most results are quite similar to the 16-core AMD machine except the
-the differences are less pronounced with all a bit closer to mimalloc performance.
-
-This is shown too in the second set of benchmarks:
-
-![bench-c5d-2xlarge-t2](bench-c5d-2xlarge-t2.png)
-
-
-
-![bench-c5d-2xlarge-m2](bench-c5d-2xlarge-m2.png)
-
-On the server workload of `larson` everyone does a bit better on the 8-cores
-than on 16. On the other benchmarks the performance does not improve though.
-
-
-## On Windows (4-core Intel Xeon)
-
-Testing on a HP Z4 G4 Workstation with a 4-core Intel® Xeon® W2123 at 3.6 GHz
-with 16GB ECC memory, running Windows 10 Pro (version 10.0.17134 Build 17134)
-with Visual Studio 2017 (version 15.8.9).
-
-Since we cannot use `LD_PRELOAD` on Windows we compiled a subset of our
-allocators and benchmarks and linked them statically. The **je** benchmark
-is therefore equivalent to the **jx** benchmark in the previous graphs.
-The **mc** allocator now refers to the standard Microsoft allocator.
-Unfortunately we could not get Hoard to work on Windows at this time.
-
-We used the Windows call `QueryPerformanceCounter` to measure elapsed wall-clock
-times, and `GetProcessMemoryInfo` to measure the peak working set (rss).
-
-First the regular workload benchmarks:
-
-![bench-z4-win-t1](bench-z4-win-t1.png)
-
-
-
-![bench-z4-win-m1](bench-z4-win-m1.png)
-
-Here mimalloc and tcmalloc perform very similar, and outperform the system
-allocator by a significant margin. Somehow jemalloc does much worse than
-running on Linux. It it not clear why yet, but it might be a compilation issue:
-when running through the profiler the `__chkstk` routine takes
-quite some time. This is a compiler inserted runtime function to check for enough
-stack space if there are many local variables or when the compiler cannot make
-a static estimate. Perhaps this is the culprit but it needs more investigation.
-
-The second set of benchmarks shows again more pronounced differences:
-
-![bench-z4-win-t2](bench-z4-win-t2.png)
-
-
-
-![bench-z4-win-m2](bench-z4-win-m2.png)
-
-In the `larson` server workload mimalloc is 25% faster than
-tcmalloc, and both significantly outperform the system allocator.
-(again probably due to the object
-migration between different threads).
-Also in `sh6bench` and `sh8bench`, mimalloc scales much
-better than the others.
-
-## References
-
-- \[1] Emery D. Berger, Kathryn S. McKinley, Robert D. Blumofe, and Paul R. Wilson.
- _Hoard: A Scalable Memory Allocator for Multithreaded Applications_
- the Ninth International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS-IX). Cambridge, MA, November 2000.
- [pdf](http://www.cs.utexas.edu/users/mckinley/papers/asplos-2000.pdf)
-
-- \[2] P. Larson and M. Krishnan. _Memory allocation for long-running server applications_. In ISMM, Vancouver, B.C., Canada, 1998.
- [pdf](http://citeseemi.ist.psu.edu/viewdoc/download;jsessionid=5F0BFB4F57832AEB6C11BF8257271088?doi=10.1.1.45.1947&rep=rep1&type=pdf)
-
-- \[3] D. Grunwald, B. Zorn, and R. Henderson.
- _Improving the cache locality of memory allocation_. In R. Cartwright, editor,
- Proceedings of the Conference on Programming Language Design and Implementation, pages 177–186, New York, NY, USA, June 1993.
- [pdf](http://citeseemi.ist.psu.edu/viewdoc/download?doi=10.1.1.43.6621&rep=rep1&type=pdf)
-
-- \[4] J. Barnes and P. Hut. _A hierarchical O(n*log(n)) force-calculation algorithm_. Nature, 324:446-449, 1986.
+See the [Performance](https://github.com/microsoft/mimalloc#Performance)
+section in the _mimalloc_ repository for benchmark results,
+or the the technical report for detailed benchmark results.
*/