e xploiting s emantic c ommutativity in h ardware s
play

E XPLOITING S EMANTIC C OMMUTATIVITY IN H ARDWARE S PECULATION G - PowerPoint PPT Presentation

Oct 18, 2022 •541 likes •1k views

E XPLOITING S EMANTIC C OMMUTATIVITY IN H ARDWARE S PECULATION G UOWEI Z HANG , V IRGINIA C HIU , D ANIEL S ANCHEZ MICRO 2016 Executive summary 2 Exploiting commutativity benefits update-heavy apps Software techniques that exploit

  1. E XPLOITING S EMANTIC C OMMUTATIVITY IN H ARDWARE S PECULATION G UOWEI Z HANG , V IRGINIA C HIU , D ANIEL S ANCHEZ MICRO 2016

  2. Executive summary 2 ¨ Exploiting commutativity benefits update-heavy apps ¤ Software techniques that exploit commutativity incur high run- time overheads (STM is 2-6x slower than HTM) ¤ Prior hardware exploits only single-instruction commutative operations (e.g., addition) ¨ CommTM exploits multi-instruction commutativity ¤ Extends coherence protocol to perform commutative operations locally and concurrently ¤ Leverages HTM to support multi-instruction updates ¤ Benefits speculative execution by reducing conflicts ¤ Accelerates full applications by up to 3.4x at 128 cores

  3. Commutativity 3 ¨ Commutative operations produce equivalent results when reordered ¤ No true data dependence à No need for communication ¤ Software exploits commutativity but incurs high run-time overheads Multi-instruction commutativity Single-instruction Top-K insertion commutativity Set insertion ADD MIN OR Ordered put Coup CommTM [Zhang et al, MICRO 2015]

  4. Commutativity 4 ¨ Commutative operations produce equivalent results when reordered ¤ No true data dependence à No need for communication ¤ Software exploits commutativity but incurs high run-time overheads ¤ Multi-instruction example: set (linked-list) insertion head null insert( a ); insert( b ); a b head null insert( b ); insert( a ); b a head null Different but semantically equivalent states

  5. Example: addition in conventional HTM 6 void add (int* counter, int delta) { tx_begin(); int v = load(counter) ; int nv = v + delta; store(counter, nv) ; tx_end(); }

  6. Example: addition in conventional HTM 6 void add (int* counter, int delta) { tx_begin(); int v = load(counter) ; int nv = v + delta; store(counter, nv) ; tx_end(); }

  7. Example: addition in conventional HTM 6 void add (int* counter, int delta) { A: 20 tx_begin(); int v = load(counter) ; int nv = v + delta; Core 0 Core 1 store(counter, nv) ; tx_end(); } Core 0 Core 1 add(A, 1); add(A, 1); add(A, 1); add(A, 1);

  8. Example: addition in conventional HTM 6 void add (int* counter, int delta) { A: 20 tx_begin(); int v = load(counter) ; read read int nv = v + delta; Core 0 Core 1 store(counter, nv) ; tx_end(); add(A, 1): Txn 0 load A add(A, 1): Txn 1 } load A Core 0 Core 1 add(A, 1); add(A, 1); add(A, 1); add(A, 1);

  9. Example: addition in conventional HTM 6 void add (int* counter, int delta) { A: 20 tx_begin(); int v = load(counter) ; read write read Conflict! int nv = v + delta; Core 0 Core 1 store(counter, nv) ; tx_end(); add(A, 1): Txn 0 load A add(A, 1): Txn 1 } load A store A Core 0 Core 1 add(A, 1); add(A, 1); add(A, 1); add(A, 1);

  10. Example: addition in conventional HTM 6 void add (int* counter, int delta) { A: 20 tx_begin(); int v = load(counter) ; read write int nv = v + delta; Core 0 Core 1 store(counter, nv) ; tx_end(); add(A, 1): Txn 0 load A add(A, 1): Txn 1 } load A store A a b o r t Core 0 Core 1 add(A, 1); add(A, 1); add(A, 1); add(A, 1);

  11. Example: addition in conventional HTM 6 void add (int* counter, int delta) { A: 24 tx_begin(); int v = load(counter) ; int nv = v + delta; Core 0 Core 1 store(counter, nv) ; tx_end(); add(A, 1): Txn 0 load A add(A, 1): Txn 1 } load A store A a b o r t commit Core 0 Core 1 load A restart add(A, 1): Txn 2 load A store A add(A, 1); add(A, 1); abort commit add(A, 1); add(A, 1); add(A, 1): Txn 3 restart load A load A store A a b o r t commit restart load A store A commit

  12. Example: addition in conventional HTM 6 void add (int* counter, int delta) { A: 24 tx_begin(); int v = load(counter) ; int nv = v + delta; Core 0 Core 1 store(counter, nv) ; tx_end(); add(A, 1): Txn 0 load A add(A, 1): Txn 1 } load A store A a b o r t commit Core 0 Core 1 load A restart add(A, 1): Txn 2 load A store A add(A, 1); add(A, 1); abort commit add(A, 1); add(A, 1); add(A, 1): Txn 3 restart load A load A store A a b o r t commit Traffic restart load A Serialization store A commit Wasted transactional work

  13. Example: addition in CommTM 7 void add (int* counter, int delta) { A: 20 tx_begin(); int v = load(counter) ; int nv = v + delta; Core 0 Core 1 store(counter, nv) ; tx_end(); } Core 0 Core 1 add(A, 1); add(A, 1); add(A, 1); add(A, 1);

  14. Example: addition in CommTM 7 void add (int* counter, int delta) { A: 20 tx_begin(); int v = load[ADD](counter) ; int nv = v + delta; Core 0 Core 1 store[ADD](counter, nv) ; tx_end(); } Core 0 Core 1 add(A, 1); add(A, 1); add(A, 1); add(A, 1);

  15. Example: addition in CommTM 7 void add (int* counter, int delta) { ADD ADD A: 20 A: 0 tx_begin(); int v = load[ADD](counter) ; int nv = v + delta; Core 0 Core 1 store[ADD](counter, nv) ; tx_end(); } Core 0 Core 1 add(A, 1); add(A, 1); add(A, 1); add(A, 1);

  16. Example: addition in CommTM 7 void add (int* counter, int delta) { ADD ADD A: 20 A: 0 tx_begin(); int v = load[ADD](counter) ; read read int nv = v + delta; Core 0 Core 1 store[ADD](counter, nv) ; tx_end(); add(A, 1): Txn1 add(A, 1): Txn0 } load[ADD] A load[ADD] A Core 0 Core 1 add(A, 1); add(A, 1); add(A, 1); add(A, 1);

  17. Example: addition in CommTM 7 void add (int* counter, int delta) { ADD ADD A: 20 A: 0 tx_begin(); int v = load[ADD](counter) ; read write read write int nv = v + delta; Core 0 Core 1 store[ADD](counter, nv) ; tx_end(); add(A, 1): Txn1 add(A, 1): Txn0 } load[ADD] A load[ADD] A store[ADD] A store[ADD] A Core 0 Core 1 add(A, 1); add(A, 1); add(A, 1); add(A, 1);

  18. Example: addition in CommTM 7 void add (int* counter, int delta) { ADD ADD A: 21 A: 1 tx_begin(); int v = load[ADD](counter) ; int nv = v + delta; Core 0 Core 1 store[ADD](counter, nv) ; tx_end(); add(A, 1): Txn1 add(A, 1): Txn0 } load[ADD] A load[ADD] A store[ADD] A store[ADD] A commit commit Core 0 Core 1 add(A, 1); add(A, 1); add(A, 1); add(A, 1);

  19. Example: addition in CommTM 7 void add (int* counter, int delta) { ADD ADD A: 22 A: 2 tx_begin(); int v = load[ADD](counter) ; int nv = v + delta; Core 0 Core 1 store[ADD](counter, nv) ; tx_end(); add(A, 1): Txn1 add(A, 1): Txn0 } load[ADD] A load[ADD] A store[ADD] A store[ADD] A commit commit Core 0 Core 1 add(A, 1): Txn3 add(A, 1): Txn2 load[ADD] A load[ADD] A add(A, 1); add(A, 1); store[ADD] A store[ADD] A add(A, 1); add(A, 1); commit commit

  20. Example: addition in CommTM 7 void add (int* counter, int delta) { ADD ADD A: 22 A: 2 tx_begin(); int v = load[ADD](counter) ; int nv = v + delta; Core 0 Core 1 store[ADD](counter, nv) ; tx_end(); add(A, 1): Txn1 add(A, 1): Txn0 } load[ADD] A load[ADD] A store[ADD] A store[ADD] A commit commit Core 0 Core 1 add(A, 1): Txn3 add(A, 1): Txn2 load[ADD] A load[ADD] A add(A, 1); add(A, 1); store[ADD] A store[ADD] A add(A, 1); add(A, 1); commit commit load A

  21. Example: addition in CommTM 7 void add (int* counter, int delta) { A: 24 tx_begin(); int v = load[ADD](counter) ; int nv = v + delta; reduction Core 0 Core 1 store[ADD](counter, nv) ; tx_end(); add(A, 1): Txn1 add(A, 1): Txn0 } load[ADD] A load[ADD] A store[ADD] A store[ADD] A commit commit Core 0 Core 1 add(A, 1): Txn3 add(A, 1): Txn2 load[ADD] A load[ADD] A add(A, 1); add(A, 1); store[ADD] A store[ADD] A add(A, 1); add(A, 1); commit commit load A User-defined reduction

  22. Example: addition in CommTM 7 void add (int* counter, int delta) { A: 24 tx_begin(); int v = load[ADD](counter) ; int nv = v + delta; reduction Core 0 Core 1 store[ADD](counter, nv) ; tx_end(); add(A, 1): Txn1 add(A, 1): Txn0 } load[ADD] A load[ADD] A store[ADD] A store[ADD] A commit commit Core 0 Core 1 add(A, 1): Txn3 add(A, 1): Txn2 load[ADD] A load[ADD] A add(A, 1); add(A, 1); store[ADD] A store[ADD] A add(A, 1); add(A, 1); commit commit load A Less traffic User-defined reduction Concurrent updates Less wasted transactional work Less run-time/memory overheads than STM

  23. CommTM

  24. Programming interface 9 Transactional update void add (int* counter, int delta) { tx_begin(); int v = load[ADD](counter) ; Labeled loads/stores int nv = v + delta; store[ADD](counter, nv) ; tx_end(); } Non-transactional reduction handler counter 16 + void reduce[ADD] (int* counter, int delta) { int v = load[ADD](counter); 20 delta int nv = v + delta; store[ADD](counter, nv); reduce[ADD] } 36 counter

  25. Handling arbitrary object sizes 10 ¨ For objects smaller than a cache line, assume lines are full of aligned elements and reduce all of them void reduce[ADD] (int* counterLine, int[] deltas) { for (int i = 0; I < intsPerCacheLine; i++) { int v = load[ADD](counterLine[i]); int nv = v + deltas[i]; store[ADD](counterLine[i], nv); } }

  26. Handling arbitrary object sizes 10 ¨ For objects smaller than a cache line, assume lines are full of aligned elements and reduce all of them counterLine void reduce[ADD] (int* counterLine, int[] deltas) { for (int i = 0; I < intsPerCacheLine; i++) { int v = load[ADD](counterLine[i]); deltas int nv = v + deltas[i]; store[ADD](counterLine[i], nv); } }

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend

S emantic Web Architecture Vitaly Vlasov inxaoc@ gmail.com Agenda 1. About S emantic Web,

S emantic Web Architecture Vitaly Vlasov inxaoc@ gmail.com Agenda 1. About S emantic Web,

S emantic Web Architecture Vitaly Vlasov inxaoc@ gmail.com Agenda 1. About S emantic Web, OWL, RDF review 2. Metadata on HTML-document 3. Pains 4. Architecture 5. Our Initiatives The S emantic Web (Tim Berners-Lee, May 17, 2001) The

555 views • 23 slides
THE C OST OF U PDATES TO S HARED D ATA IN C ACHE -C OHERENT S YSTEMS G UOWEI Z HANG , W EBB H ORN ,

THE C OST OF U PDATES TO S HARED D ATA IN C ACHE -C OHERENT S YSTEMS G UOWEI Z HANG , W EBB H ORN ,

E XPLOITING C OMMUTATIVITY TO R EDUCE THE C OST OF U PDATES TO S HARED D ATA IN C ACHE -C OHERENT S YSTEMS G UOWEI Z HANG , W EBB H ORN , D ANIEL S ANCHEZ MICRO 2015 Executive summary 2 Updates to shared data limit parallelism in current

313 views • 17 slides
H ARDWARE P REPROCESSING F RAMEWORK (HPF) Traditional hardware Hardware preprocessing description

H ARDWARE P REPROCESSING F RAMEWORK (HPF) Traditional hardware Hardware preprocessing description

M AMBA : C LOSING THE P ERFORMANCE G AP IN P RODUCTIVE H ARDWARE D EVELOPMENT F RAMEWORKS Shunning Jiang , Berkin Ilbeyi, Christopher Batten School of Electrical and Computer Engineering Cornell University 0/17 T HE T RADITIONAL F LOW * HDL:

544 views • 43 slides
A PPLICATION : S EARCH IN T OURISM (S KY S CANNER ) Goal: search for hotels/flights/trips using

A PPLICATION : S EARCH IN T OURISM (S KY S CANNER ) Goal: search for hotels/flights/trips using

Q UERYING S EMANTIC B IG D ATA AND I TS A PPLICATIONS Boris Motik University of Oxford November 16, 2015 T ABLE OF C ONTENTS B IG D ATA A PPLICATIONS OF S EMANTIC F ORMALISMS 1 RDF OX : P ARALLEL M ATERIALISATION -B ASED D ATALOG R EASONER 2 A

650 views • 50 slides
M ULTICORE H ARDWARE S HARED R ESOURCES : U NDERSTANDING OF THE S TATE OF THE A RT Gabriel

M ULTICORE H ARDWARE S HARED R ESOURCES : U NDERSTANDING OF THE S TATE OF THE A RT Gabriel

C ONTENTION IN M ULTICORE H ARDWARE S HARED R ESOURCES : U NDERSTANDING OF THE S TATE OF THE A RT Gabriel Fernandez 1 , Jaume Abella 2 , Eduardo Quiones 2 , Christine Rochange 3 , Tullio Vardanega 4 and Francisco J. Cazorla 2,4 1 2 3 5 4 14

648 views • 23 slides
T IME T RAVELING H ARDWARE AND S OFTWARE S YSTEMS Xiangyao Yu, Srini Devadas CSAIL, MIT F OR F

T IME T RAVELING H ARDWARE AND S OFTWARE S YSTEMS Xiangyao Yu, Srini Devadas CSAIL, MIT F OR F

T IME T RAVELING H ARDWARE AND S OFTWARE S YSTEMS Xiangyao Yu, Srini Devadas CSAIL, MIT F OR F IFTY YEARS , WE HAVE RIDDEN M OORE S L AW Moores Law and the scaling of clock frequency = printing press for the currency of performance T

948 views • 61 slides
care ? &amp; why do I what are they RDF S emantic eb W OWL S eparate eb W machines

care ? &amp; why do I what are they RDF S emantic eb W OWL S eparate eb W machines

microformats what are they care ? &amp; why do I what are they RDF S emantic eb W OWL S eparate eb W machines W tandards eb S (X)HTML wemantic s eb microformats wame s eb humans low hanging fruit where relationships

450 views • 29 slides
HARDWARE H ARDWARE T YPES Microcontroller (MCU) Arduino, ESP8266, Particle Single Board

HARDWARE H ARDWARE T YPES Microcontroller (MCU) Arduino, ESP8266, Particle Single Board

I O T L AB P AUL L ANGDON S ATURN 2016 HARDWARE H ARDWARE T YPES Microcontroller (MCU) Arduino, ESP8266, Particle Single Board Computer (SBC) Intel Edison, Raspberry Pi, C.H.I.P. Excellent Guide:

218 views • 10 slides
S EMANTIC S OLUTIONS FOR O IL &amp; G AS : R OLES AND R ESPONSIBILITIES

S EMANTIC S OLUTIONS FOR O IL &amp; G AS : R OLES AND R ESPONSIBILITIES

InformationIntegrationIntelligence: Solutions S EMANTIC S OLUTIONS FOR O IL &amp; G AS : R OLES AND R ESPONSIBILITIES

348 views • 24 slides
ISO-T IME ML: A N I NTERNATIONAL S TANDARD FOR S EMANTIC ANNOTATION James Pustejovsky*, Kiyong

ISO-T IME ML: A N I NTERNATIONAL S TANDARD FOR S EMANTIC ANNOTATION James Pustejovsky*, Kiyong

ISO-T IME ML: A N I NTERNATIONAL S TANDARD FOR S EMANTIC ANNOTATION James Pustejovsky*, Kiyong Lee, Harry Bunt, Laurent Romary ISO *Computer Science Department Brandeis University ISO LREC 2010 ISO Malta May19, 2010 ISO O UTLINE

722 views • 44 slides
E XTENDING S EMANTIC AND E PISODIC M EMORY TO S UPPORT R OBUST D ECISION M AKING

E XTENDING S EMANTIC AND E PISODIC M EMORY TO S UPPORT R OBUST D ECISION M AKING

E XTENDING S EMANTIC AND E PISODIC M EMORY TO S UPPORT R OBUST D ECISION M AKING (FA2386-10-1-4127) PI: John E. Laird (University of Michigan) Graduate Students: Nate Derbinsky, Mitchell Bloch, Mazin Assanie AFOSR Program Review: Mathematical

760 views • 46 slides
Natural S emantics Based Tools for S emantic Web with Application to Product Models CUGS

Natural S emantics Based Tools for S emantic Web with Application to Product Models CUGS

Natural S emantics Based Tools for S emantic Web with Application to Product Models CUGS thesis proposal Adrian Pop Programming Environments Laboratory (PELAB) Department of Computer and Information S cience (IDA) Linkping University

461 views • 22 slides
S emantic A utomated D iscovery and I ntegration http://sadiframework.org Summary SADI is a

S emantic A utomated D iscovery and I ntegration http://sadiframework.org Summary SADI is a

S emantic A utomated D iscovery and I ntegration http://sadiframework.org Summary SADI is a set of conventions for creating Semantic Web Services that can be automatically discovered and orchestrated . SADI does not create new technologies

774 views • 58 slides
S EMANTIC -B ASED M ULTILINGUAL D OCUMENT C LUSTERING VIA T ENSOR M ODELING Salvatore Romeo 1 ,

S EMANTIC -B ASED M ULTILINGUAL D OCUMENT C LUSTERING VIA T ENSOR M ODELING Salvatore Romeo 1 ,

EMNLP 2014 C ONFERENCE ON E MPIRICAL M ETHODS IN N ATURAL L ANGUAGE P ROCESSING D OHA , Q ATAR . O CTOBER 2529, 2014 S EMANTIC -B ASED M ULTILINGUAL D OCUMENT C LUSTERING VIA T ENSOR M ODELING Salvatore Romeo 1 , Andrea Tagarelli 1 , and Dino

594 views • 34 slides
S EM F IX : P ROGRAM R EPAIR VIA S EMANTIC A NALYSIS CREST Workshop, Jan 2014 H.D.T. Nguyen,

S EM F IX : P ROGRAM R EPAIR VIA S EMANTIC A NALYSIS CREST Workshop, Jan 2014 H.D.T. Nguyen,

S EM F IX : P ROGRAM R EPAIR VIA S EMANTIC A NALYSIS CREST Workshop, Jan 2014 H.D.T. Nguyen, Dawei Qi, Abhik Roychoudhury National University of Singapore , &amp; Satish Chandra Samsung 1 Talk given at 30 th CREST Workshop, London, Jan 2014. W

490 views • 29 slides
WAVES B IG D ATA P LATFORM FOR R EAL - TIME S EMANTIC S TREAM M ANAGEMENT WAVES ATOS SE OUTLINE

WAVES B IG D ATA P LATFORM FOR R EAL - TIME S EMANTIC S TREAM M ANAGEMENT WAVES ATOS SE OUTLINE

WAVES B IG D ATA P LATFORM FOR R EAL - TIME S EMANTIC S TREAM M ANAGEMENT WAVES ATOS SE OUTLINE What is WAVES? Why WAVES? How WAVES? Achievements Contact 2 WAVES ATOS SE WHAT IS WAVES? Massive Semantic Streams empowering Innovative Big

573 views • 17 slides
BinRec: Dynamic Binary Lifting and Recompilation Anil Altinay , Joseph Nash , Taddeus

BinRec: Dynamic Binary Lifting and Recompilation Anil Altinay , Joseph Nash , Taddeus

BinRec: Dynamic Binary Lifting and Recompilation Anil Altinay , Joseph Nash , Taddeus Kroes Prabhu Rajasekaran, Dixin Zhou, Adrian Dabrowski, David Gens, Yeoul Na, Stijn Volckaert, Cristiano Giuffrida, Herbert Bos, Michael Franz

457 views • 32 slides
CPSC 213 Introduction to Computer Systems Unit 2c Synchronization 1 Reading Companion 6

CPSC 213 Introduction to Computer Systems Unit 2c Synchronization 1 Reading Companion 6

CPSC 213 Introduction to Computer Systems Unit 2c Synchronization 1 Reading Companion 6 (Synchronization) Text Shared Variables in a Threaded Program, Synchronizing Threads with Semaphores, Using Threads for Parallelism, Other

821 views • 58 slides
make world Chris Smowton University of Cambridge spell-rite /usr/share/real_words ~/nonsense

make world Chris Smowton University of Cambridge spell-rite /usr/share/real_words ~/nonsense

make world Chris Smowton University of Cambridge spell-rite /usr/share/real_words ~/nonsense spell-rite /usr/share/real_words ~/nonsense spell-rite /usr/share/real_words ~/nonsense /usr/share/real_words ~/nonsense /usr/share/real_words

725 views • 37 slides
Objects (cont.) Deian Stefan (Adopted from my &amp; Edward Yangs CS242 slides) Today

Objects (cont.) Deian Stefan (Adopted from my &amp; Edward Yangs CS242 slides) Today

Objects (cont.) Deian Stefan (Adopted from my &amp; Edward Yangs CS242 slides) Today Statically-typed OO languages: C++ vtables Closer look at subtyping Why talk about C++? C++ is an OO extension of C Efficiency and

572 views • 33 slides
CS 241: Systems Programming Lecture 25. Function Pointers Spring 2020 Prof. Stephen Checkoway 1

CS 241: Systems Programming Lecture 25. Function Pointers Spring 2020 Prof. Stephen Checkoway 1

CS 241: Systems Programming Lecture 25. Function Pointers Spring 2020 Prof. Stephen Checkoway 1 Function pointers Function pointers are pointers that point tofunctions Syntax return_type (*var)(parameters); int (*f1)( void ); // f1 is

486 views • 21 slides
DescribingLinkedDatasets OntheDesignandUsageof voiD ,

DescribingLinkedDatasets OntheDesignandUsageof voiD ,

KeithAlexander(Talis),RichardCyganiak(DERI), MichaelHausenblas(DERI)andJunZhao(UniversityofOxford) DescribingLinkedDatasets OntheDesignandUsageof voiD ,

417 views • 31 slides
Introduction to CUDA C What is CUDA? CUDA Architecture Expose general-purpose GPU

Introduction to CUDA C What is CUDA? CUDA Architecture Expose general-purpose GPU

Introduction to CUDA C What is CUDA? CUDA Architecture Expose general-purpose GPU computing as first-class capability Retain traditional DirectX/OpenGL graphics performance CUDA C Based on industry-standard C A handful of

1.31k views • 62 slides
Roeder A First Program Using brings in a namespace, which is an using System; abstract container

Roeder A First Program Using brings in a namespace, which is an using System; abstract container

Slides adapted from 4week course at Cornell by Tom Roeder A First Program Using brings in a namespace, which is an using System; abstract container of symbols namespace Test { int a = 137 class Hello { public static void Main(string[] args) {

783 views • 54 slides

More recommend