A NEW TYPE OF PHP: HHVM AND HACK Jeremy Mikola jmikola

A Brief History of Time

2008 HipHop PHP compiler (HPHPc) PHP → AST → C++ → x64 Developer tools Interpreter (HPHPi) Debugger (HPHPd)

2010 Facebook open-sources HPHPc HipHop Virtual Machine (HHVM) PHP → AST → HHBC HHBC → x64 (JIT)

2011 Facebook open-sources HHVM HHVM replaces HPHPi for development

2012

2013 HHVM replaces HPHPc in production HHVM 2.3.0 adds support for FastCGI Travis CI adds HHVM runtime Performance and parity lockdowns

2014 Facebook open-sources Hack language HHVM 3.0.0 adds support for Hack 27 PHP framework test suites at 100% Tracking PHP HEAD

HHVM

Why Are We Still Using PHP? : “objective issues… which make me sad” PHP: A Fractal of Bad Design PHP Sadness

The Elephant in the Room It’s had a role in significant projects Wikipedia, Drupal, Wordpress Facebook: x * 10⁵ files, y * 10⁷ LOC It gets some things right Developer workflow: tight feedback loop State: each request is a blank slate Concurrency: share nothing Taking PHP Seriously – Keith Adams, Facebook

Lifetime of a Request

Bytecode Interpreter

Native Code JIT

Repo Authoritative Pre-compile HHBC: h h v m - - h p h p - t a n a l y z e

Benchmarking HHVM

Test Environment Symfony2, ElasticSearch (HTTP API) JSON → PHP serialization PHP 5.3 (APC), PHP 5.5 (OpCache), HHVM 2.3.0 nginx web server with FastCGI ApacheBench for load testing Results from: , Christian Stocker Liip AG

Small Responses (7 KB) Requests / Second Response Time (ms) https://blog.liip.ch/archive/2013/10/29/hhvm-and-symfony2.html

Medium Responses (80 KB) Requests / Second Response Time (ms) https://blog.liip.ch/archive/2013/10/29/hhvm-and-symfony2.html

Large Responses (220 KB) Requests / Second Response Time (ms) https://blog.liip.ch/archive/2013/10/29/hhvm-and-symfony2.html

Some Conclusions PHP 5.5 is up to 1.5 times faster than PHP 5.3 HHVM is up to 3.5 times faster than PHP 5.3 HHVM is up to twice as fast as PHP 5.5 HHVM shines under high load, larger requests Stop using PHP 5.3

Installing HHVM

Platform Support 64-bit Linux: It just works™ Mac OS X: interpreter only Windows and ARM in development

Linux Binaries Ubuntu (LTS, recent releases) Debian 8 Fedora 20 Mint 16 CentOS (community supported)

Building from Source Fair amount of dependencies A must for extension development $ g i t c l o n e g i t : / / g i t h u b . c o m / f a c e b o o k / h h v m . g i t $ c d h h v m $ g i t c h e c k o u t H H V M - 3 . 1 . 0 $ g i t s u b m o d u l e u p d a t e - - i n i t - - r e c u r s i v e $ c m a k e . $ m a k e - j [ n u m _ c o r e s ] $ s u d o m a k e i n s t a l l

Grab a Snickers…

HACK < ? h h

In a Nutshell Static type analysis Local type inference, meaningful errors Tight feedback loop (h h _ c l i e n t , h h _ s e r v e r ) Checked before runtime New language features, core classes Some PHP ‘features’ removed Catch bugs early, develop faster

Wonderful World of Types

Type Hinting Scalars bool, int, float, num, string Array elements (and keys) a r r a y < i n t > , a r r a y < s t r i n g , i n t > Properties and return types Nullable and m i x e d types Generics lend to code reuse

Type Hinting c l a s s A d d e r { p r i v a t e i n t $ x ; p u b l i c f u n c t i o n _ _ c o n s t r u c t ( i n t $ x ) : v o i d { $ t h i s - > x = $ x ; } p u b l i c f u n c t i o n a d d ( i n t $ y ) : v o i d { $ t h i s - > x + = $ y ; } p u b l i c f u n c t i o n g e t ( ) : i n t { r e t u r n $ t h i s - > x ; } }

Constructor Arg Promotion c l a s s A d d e r { p u b l i c f u n c t i o n _ _ c o n s t r u c t ( p r i v a t e i n t $ x ) : v o i d { } p u b l i c f u n c t i o n a d d ( i n t $ y ) : v o i d { $ t h i s - > x + = $ y ; } p u b l i c f u n c t i o n g e t ( ) : i n t { r e t u r n $ t h i s - > x ; } }

Generics c l a s s A d d e r < T > { p u b l i c f u n c t i o n _ _ c o n s t r u c t ( p r i v a t e T $ x ) : v o i d { } p u b l i c f u n c t i o n a d d ( T $ y ) : v o i d { $ t h i s - > x + = $ y ; } p u b l i c f u n c t i o n g e t ( ) : T { r e t u r n $ t h i s - > x ; } }

Type Aliasing Declared outside of classes, functions Opaque aliasing abstracts underlying type / / t y p e s . p h p t y p e M y I n t = i n t ; n e w t y p e M y O p a q u e I n t = i n t ; r e q u i r e ' t y p e s . p h p ' ; f u n c t i o n f o o ( M y I n t $ x ) : M y I n t { r e t u r n $ x + 1 ; } f u n c t i o n b a r ( M y O p a q u e I n t $ x ) : M y O p a q u e I n t { / / E r r o r : c a n n o t a s s u m e i n t e g e r b e h a v i o r o u t s i d e o f t y p e s . p h p r e t u r n $ x + 1 ; }

Type Checking is Configurable Modes: strict, partial, decl < ? h h / / m o d e Partial is most flexible, the default Classes and functions can be partially typed Allows calling into untyped code (i.e. PHP) Decl useful for calling into legacy code from strict Type checker looks at signatures, but not code

Stop Using PHP Arrays (ab)

Tuples Typed, fixed-length lists Size and types are immutable Useful for custom types, multiple return values t y p e P o i n t 2 D = ( i n t , i n t ) ; f u n c t i o n c r e a t e P o i n t ( i n t $ x , i n t $ y ) : P o i n t 2 D { r e t u r n t u p l e ( $ x , $ y ) ; } f u n c t i o n g e t D a t e A n d T i m e s t a m p ( ) : ( s t r i n g , i n t ) { r e t u r n t u p l e ( d a t e ( ' c ' ) , t i m e ( ) ) ; }

Shapes Typed, structured dictionaries Alternative to classes with public properties Useful for simple data structures, argument bags t y p e P a g e V i e w = s h a p e ( ' u r l ' = > s t r i n g , ' c o u n t ' = > i n t , ) ; f u n c t i o n c r e a t e P a g e V i e w ( s t r i n g $ u r l , i n t $ c o u n t ) : P a g e V i e w { r e t u r n s h a p e ( ' u r l ' = > $ u r l , ' c o u n t ' = > $ c o u n t ) ; }

Collections Specialized array objects Vector, Map, Set, Pair Ordered, immutable or mutable Accessible, iterable like PHP arrays Literal syntax, familiar collection API

Vectors vs. PHP Arrays $ v e c t o r = V e c t o r { 5 , 1 0 , 1 5 , 2 0 } ; $ v e c t o r - > r e m o v e K e y ( 2 ) ; f o r e a c h ( $ v e c t o r a s $ k e y = > $ v a l u e ) { p r i n t f ( " % d = > % d \ n " , $ k e y , $ v a l u e ) ; } $ a r r a y = [ 5 , 1 0 , 1 5 , 2 0 ] ; u n s e t ( $ a r r a y [ 2 ] ) ; f o r e a c h ( $ a r r a y a s $ k e y = > $ v a l u e ) { p r i n t f ( " % d = > % d \ n " , $ k e y , $ v a l u e ) ; } [ 5 , 1 0 , 2 0 ] vs. { " 0 " : 5 , " 1 " : 1 0 , " 3 " : 2 0 }

These May Be Useful, Too

Lambda Expressions PHP has closures, but… Parent scope is not implicitly inherited Can be verbose for simple expressions $ i n c = f u n c t i o n ( $ x ) { r e t u r n $ x + 1 ; } $ i n c ( 1 1 ) ; / / r e t u r n s 1 2 $ f = f u n c t i o n ( $ x ) { r e t u r n f u n c t i o n ( $ y ) u s e ( $ x ) { r e t u r n $ x + $ y ; } ; } ; $ g = $ f ( 4 ) ; $ g ( 5 ) ; / / r e t u r n s 9

Lambda Expressions Parent scope is inherited automatically Concise syntax well-suited to callback APIs $ i n c = $ x = = > $ x + 1 ; $ i n c ( 1 1 ) ; / / r e t u r n s 1 2 $ f = $ x = = > $ y = = > $ x + $ y ; $ g = $ f ( 4 ) ; $ g ( 5 ) ; / / r e t u r n s 9

Lambdas and Collections $ a r r a y = [ 2 , 4 , 6 , 8 ] ; $ i n c = 3 ; a r r a y _ f i l t e r ( a r r a y _ m a p ( f u n c t i o n ( $ x ) u s e ( $ i n c ) { r e t u r n $ x + $ i n c ; } , $ a r r a y ) , f u n c t i o n ( $ x ) { r e t u r n $ x > 1 0 ; } ) ; vs. $ v e c t o r = V e c t o r { 2 , 4 , 6 , 8 } ; $ i n c = 3 ; $ v e c t o r - > m a p ( $ x = = > $ x + $ i n c ) - > f i l t e r ( $ x = = > $ x > 1 0 ) ;

Continuations Like generators in PHP 5.5, slight differences Must be primed with n e x t ( ) y i e l d must always return a value f u n c t i o n i n f i n i t e ( ) : C o n t i n u a t i o n < i n t > { $ i = 0 ; w h i l e ( t r u e ) { y i e l d $ i + + ; } } f o r e a c h ( i n f i n i t e ( ) a s $ i ) { p r i n t f ( " % d \ n " , $ i ) ; }

Async Functions Functions use a s y n c modifier, return Awaitables Callers opt in with a w a i t keyword a s y n c f u n c t i o n c o m p l e x C a l c u l a t i o n ( … ) : A w a i t a b l e < R e s u l t > { / / q u e r y d a t a b a s e s , c a l l s e r v i c e s , e t c . r e t u r n n e w R e s u l t ( … ) ; } a s y n c f u n c t i o n f e t c h U s e r ( i n t $ i d ) : A w a i t a b l e < U s e r > { $ a r g s = [ ' i d ' = > $ i d , ' t y p e ' = > ' u s e r ' ] ; $ a = c o m p l e x C a l c u l a t i o n ( $ a r g s ) ; / / $ a r e p r e s e n t s u n s t a r t e d s t a t e o f a c o m p u t a t i o n $ r e s u l t = a w a i t $ a ; / / b l o c k u n t i l $ a i s r e s o l v e d r e t u r n c o n v e r t R e s u l t T o U s e r ( $ r e s u l t ) ; }

Async Functions Parallelism by awaiting multiple things at once Top-level context can j o i n ( ) (block on results) a s y n c f u n c t i o n g e n e r a t e R e s p o n s e ( ) : A w a i t a b l e < R e s p o n s e > { l i s t ( $ a , $ b , $ c ) = a w a i t G e n A r r a y W a i t H a n d l e : : c r e a t e ( [ f e t c h U s e r ( … ) , q u e r y C a c h e ( … ) , c a l l S e r v i c e ( … ) , ] ) ; r e t u r n n e w R e s p o n s e ( $ a , $ b , $ c ) ; } a s y n c f u n c t i o n f e t c h U s e r ( … ) : A w a i t a b l e < … > { } a s y n c f u n c t i o n q u e r y C a c h e ( … ) : A w a i t a b l e < … > { } a s y n c f u n c t i o n c a l l S e r v i c e ( … ) : A w a i t a b l e < … > { } $ r e s p o n s e = g e n e r a t e R e s p o n s e ( ) - > j o i n ( ) ;

Additional Features User attributes (accessible via Reflection) Trait requirements (support static analysis) Override attribute (child counterpart to abstract) Variadic arguments (not type-checked) XHP (templated XML fragments in PHP)

Extending HHVM

Creating an Extension Bootstrap PHP in SystemLib HHVM Native Interface (HNI) Dynamically Loaded Objects

HHVM Native Interface (PHP) Implement as much as possible in PHP Annotated methods point to C++ implementations c l a s s M o n g o C l i e n t { < < _ _ N a t i v e > > p u b l i c f u n c t i o n _ _ c o n s t r u c t ( s t r i n g $ s e r v e r = ' m o n g o d b : / / l o c a l h o s t : 2 7 0 1 7 ' , a r r a y $ o p t i o n s = [ ' c o n n e c t ' = > t r u e ] ) : v o i d ; p u b l i c f u n c t i o n s e l e c t D B ( s t r i n g $ n a m e ) : M o n g o D B { r e t u r n n e w M o n g o D B ( $ t h i s , $ n a m e ) ; } / / … } https://github.com/10gen-labs/mongo-hhvm-driver/

HHVM Native Interface (C++) Method macros similar to PHP extension API 1:1 mapping between PHP and C++ types s t a t i c v o i d H H V M _ M E T H O D ( M o n g o C l i e n t , _ _ c o n s t r u c t , c o n s t S t r i n g & u r i , A r r a y o p t i o n s ) { m o n g o c _ c l i e n t _ t * c l i e n t = m o n g o c _ c l i e n t _ n e w ( u r i . c _ s t r ( ) ) ; / / w r a p m o n g o c _ c l i e n t _ t i n a n H H V M r e s o u r c e o b j e c t t h i s _ - > o _ s e t ( " _ r e s o u r c e " , c l i e n t R e s o u r c e , " C l i e n t R e s o u r c e " ) ; } https://github.com/10gen-labs/mongo-hhvm-driver/

Undocumented Territory Learn by example, consult prior art Core extensions (h p h p / r u n t i m e / e x t / ) Community extensions (about a dozen) Become an IRC regular

Looking Ahead PHP iterating faster than ever Multiple implementations is a plus Competition drives progress Hack features and overlapping PHP RFCs Facebook still active on php-internals

Getting the Community on Board http://hhvm.h4cc.de/

PHPNG (Next Generation) JIT based on Fair amount of internal refactoring Some behavioral changes, BC breaks Progress since January 2014 1.5x execution speed-up 30–60% less CPU instructions LLVM

THANKS! Any Questions? and and # h h v m on Freenode jmikola.net/blog/hack-dev-day/ github.com/facebook/hhvm wiki hhvm.com hacklang.org Thanks to: , , , and Elizabeth Smith Sara Golemon Christian Stocker Chris Heng

Image Credits id Software (Doom character sprites) http://xkcd.com/303/