Scalable precise-dynamic-datarace-detection-for-structured-parallelism

Scalable and Precise Dynamic Datarace Detection for Structured Parallelism

Presented by Yunming ZhangRice University

Paper by Raghavan Raman, Jisheng Zhao, Vivek Sarkar, Martin Vechev, Eran Yahav

Tuesday, October 22, 13

Background

• Problems

• sequential

• large space overhead

• costly read-write checks

• Design a datarace detector

• works in parallel

• constant space per monitored memory location

• precise and sound

• no false positives for a given input

• if one or more races exist, at least one will be reported

Structured Parallelism

• Express a wide range of parallel programs succinctly with a few parallel constructs

• Exploit structured parallelism for better performance through scheduling

• Provide guarantee of deadlock-freedom

Structured Parallelism• Cilk

• spawn/sync model

• requires child/children to synchronize with the parent

• Habanero Java

• async/finish model

• requires descendants to synchronize with some ancestor

Structured Parallelismint main() {

Node * node = initialize node walk(node);

}ReducerList<Node *> output_list;void walk(Node *x){ if (x) { if (has_property(x)) output_list.push_back(x);

cilk_spawn walk(x->left); walk(x->right); cilk_sync; }}

Structured Parallelismint main() {

Node * node = initialize node finish { walk(node); }

}ReducerList<Node *> output_list;void walk(Node *x){ if (x) { if (has_property(x)) output_list.push_back(x);

async walk(x->left); async walk(x->right); cilk_sync; }}

Race Detector Components

• Data structure to determine whether two tasks may execute in parallel

• Cilk uses SP bags

• This paper: Dynamic Program Structure Tree

• Data structure that keeps track of accesses to the same memory location that may result in data race

• Cilk records one read and one write

Dynamic Program Structure Tree (DPST)• Node types

• async

• finish

• step: a maximal sequence of statements without any async or finish operation

• Edges between parent and child tasks

• Children ordered from left to right

Dynamic Program Structure Treefinish{

async{int i = 0;int j = 1;int k = i*j;...

step node

async nodefinish node

Dynamic Program Structure Tree

finish{//F1step1async{//A1

step2async{//A2

step3}step4

}step5async{//A3

step6}

}//finish F1

finish{step1async{//A1

step2async{//A2

step3}step4

}step5async{//A3

step6}

}//finish F1

step2async{//A2

step3}step4

}step5async{//A3

step6}

}//finish F1

step2async{//A2

step3}step4

}step5async{//A3

step6}

}//finish F1

step2async{//A2

step3}step4

}step5async{//A3

step6}

}//finish F1

step2async{//A2

step3}step4

}step5async{//A3

step6}

}//finish F1

step2async{//A2

step3}step4

}step5async{//A3

step6}

}//finish F1

step2async{//A2

step3}step4

}step5async{//A3

step6}

}//finish F1

step2async{//A2

step3}step4

}step5async{//A3

step6}

}//finish F1

Dynamic May Happen in Parallel (DMHP)• Need to determine if two tasks can

execute in parallel using DPST

• Theorem 1. Consider two leaf nodes (steps) S1 and S2 in a DPST, where S1 ≠ S2 and S1 is to the left of S2. S1 and S2 can execute in parallel if and only if the ancestor of S1 that is the child of LCA(S1,S2) is an async node.

Can happen in parallel

Lowest Common Ancestor LCA(S3, S4)

Ancestor of S3, Child of LCA(S3, S4)

Can happen in parallel

step2async{//A2

step3}step4

}step5async{//A3

step6}

}//finish F1

Cannot happen in parallel

Lowest Common Ancestor LCA(S2,S3)

Ancestor of S2, Child of LCA(S2,S3)

Cannot happen in parallel

step2async{//A2

step3}step4

}step5async{//A3

step6}

}//finish F1

Access Summary

• Maintains access summary in “shadow memory”

• for each data word m, an access summary is maintained in S(m)

• S(m) contains

• w: a reference to a step that wrote m

• r1: a reference to a step that read m

• r2: a reference to another step that read m

Access Summary

• Last write is straightforward

• Among the reads to a memory location, (a1, a2, ... ak) Only necessary to store two reads

• ai, aj , such that the subtree under LCA(ai, aj) includes all past reads

• any future read an that is in parallel with any prior read will be in parallel with at least one of ai or aj

Race Detection Algorithm

If replacing one ofthe reads with my step yields a higher LCA in DPST, then do so

Why two readsint[] x = new int[2];x[0] = 0;finish{//F1

async {//A1async {//A2

... = x[0]; //S1}... = x[0]; //S2

}async {//A3

... = x[0]; //S3}

... = x[0]; //S1}... = x[0]; //S2

}async {//A3

... = x[0]; //S3}

Why two reads

M.r1 = S1M.r2 = S2

... = x[0]; //S1}... = x[0]; //S2

}async {//A3

... = x[0]; //S3}

Why two readsIf replacing one ofthe reads with my step yields a higher LCA in DPST, then do so

M.r1 = S1M.r2 = S2S = S3

Why two reads

lca(S3,S1) > DPSTlca(S1,S2) lca(S3,S2) > DPSTlca(S1,S2)

M.r1 = S1M.r2 = S2M.r2 = S3

Keys to Performance

• Construct the DPST in parallel

• Relax Atomicity when updating access summary

Space and Time Complexity

• Definitions:

• T1: sequential execution time

• v: total number of variables

• n: number of nodes in the DPST

• Time Complexity

• O(T1* T∞)

• Space Complexity

• O(v + n)

Precise and Sound• Precise

• if Algorithm reports a data race on a memory location M during an execution of a program P with an input, then there exists at least one execution of P with the input in which data race exists

• Sound

• if Algorithms do not report a data race on a memory location M during an execution of a program P with an input, then no execution with the input will report data race

Performance Evaluation

Conclusions• Scalable

• parallel data race detection

• trading off parallelism for efficient DMHP mechanism and results show it is worth it

• Modest memory overhead

• store 3 access history per memory location

• Eraser maintained large lock sets and FastTrack uses vector clocks

• Precise and sound

Scalable precise-dynamic-datarace-detection-for-structured-parallelism

Technology

9. Alternative Konzepte: Parallele funktionale Programmierung Implicit Parallelism Controlled Parallelism Explicit Parallelism Data Parallelism Control

DataRace OSDI2010-v3

Proving parallelism

Datarace: IoT e Big Data (Italian)

Parallelism and Orders of Signification (Parallelism ...journal.oraltradition.org/files/articles/31ii/11_31.2.pdf · Parallelism and Orders of Signification (Parallelism Dynamics

Lecture 8: OpenMP. Parallel Programming Models Parallel Programming Models: Data parallelism / Task parallelism Explicit parallelism / Implicit parallelism

Parallelism 1

Parallelism - GoGMATlms.gogmat.com/files/2/gmat/Parallelism.pdf · Parallelism Arguably, the GMAT s favorite grammar topic is parallelism. According to the principle of parallelism,

Polyglot parallelism

Scalable and Precise Dynamic Datarace Detection for ... · Race Detector Components • Data structure to determine whether two tasks may execute in parallel • Cilk uses SP bags

Intermachine Parallelism

Kite: Braided Parallelism for Heterogeneous Systemsmixture of both task- and data-parallelism, a form of parallelism Lefohn [12] calls braided parallelism. This is only one frame;

Parallelism copy

Parallelism Orchestration using DoPE: the Degree of ...liberty.princeton.edu/Publications/pldi11_dope.pdf · Parallelism Orchestration using DoPE: the Degree of Parallelism Executive

Chapter 21: Parallel Databases. 21.2 Chapter 21: Parallel Databases Introduction I/O Parallelism Interquery Parallelism Intraquery Parallelism Intraoperation

Parallelism Ppt

web.cs.sunyit.edugandhaj/Data Parallelism in … · Web viewData Parallelism in SPARK

PARALLELISM PARALLELISM PARALLELISM

Scalable and Precise Dynamic Datarace Detection for Structured Parallelism

6.1 Parallelism