Tweet
Tweet

Slide 1

Slide 1 text

Hierarchical Parallel Dynamic Dependence Analysis for Recursively Task-Parallel Programs IPDPS 2016, 26th of May 2016 Nikolaos Papakonstantinou, Foivos S. Zakkak, and Polyvios Pratikakis Except where otherwise noted, this presentation is licensed under the Creative Commons Attribution-ShareAlike 4.0 International License. Third party marks and brands are the property of their respective holders.

Slide 2

Slide 2 text

1 Introduction 1.1 Contributions 1/13 What it this talk about? PARTEE, a runtime system that: ■ Runs task-parallel applications efficiently ■ Detects and resolves dependencies ■ Features parallel and hierarchical dynamic dependence analysis F. Zakkak - [email protected]

Slide 3

Slide 3 text

1 Introduction 1.2 The Programming Model 2/13 The OMPSs programming model An example void baz(int *k, int *l) {*k = *l;} void foo(int *x, int *y, int *z) { #pragma omp task in(z) out(x) baz(x, z); #pragma omp task in(z) out(y) baz(y, z); } void bar(int *k, int *l) { #pragma omp task in(l) out(k) baz(k, l); } int main(void) { // ... #pragma omp task in(z) out(x, y) foo(x, y, z); #pragma omp task in(x) out(k) bar(k, x); #pragma omp task in(m) out(l) baz(l, m); // ... } Spawn Graph foo(x, y, z) out: x, y in: z Task 1 baz(y, z) out: y in: z Task 1.1 baz(x, z) out: x in: z Task 1.2 bar(k, x) out: k in: x Task 2 baz(k, x) out: k in: x Task 2.1 baz(l, m) out: l in: m Task 3 main() inout: x,y,z,k,l,m Root Task Spawn Dependency F. Zakkak - [email protected]

Slide 4

Slide 4 text

1 Introduction 1.2 The Programming Model 2/13 The OMPSs programming model An example void baz(int *k, int *l) {*k = *l;} void foo(int *x, int *y, int *z) { #pragma omp task in(z) out(x) baz(x, z); #pragma omp task in(z) out(y) baz(y, z); } void bar(int *k, int *l) { #pragma omp task in(l) out(k) baz(k, l); } int main(void) { // ... #pragma omp task in(z) out(x, y) foo(x, y, z); #pragma omp task in(x) out(k) bar(k, x); #pragma omp task in(m) out(l) baz(l, m); // ... } Spawn Graph foo(x, y, z) out: x, y in: z Task 1 baz(y, z) out: y in: z Task 1.1 baz(x, z) out: x in: z Task 1.2 bar(k, x) out: k in: x Task 2 baz(k, x) out: k in: x Task 2.1 baz(l, m) out: l in: m Task 3 main() inout: x,y,z,k,l,m Root Task Spawn Dependency F. Zakkak - [email protected]

Slide 5

Slide 5 text

1 Introduction 1.2 The Programming Model 2/13 The OMPSs programming model An example void baz(int *k, int *l) {*k = *l;} void foo(int *x, int *y, int *z) { #pragma omp task in(z) out(x) baz(x, z); #pragma omp task in(z) out(y) baz(y, z); } void bar(int *k, int *l) { #pragma omp task in(l) out(k) baz(k, l); } int main(void) { // ... #pragma omp task in(z) out(x, y) foo(x, y, z); #pragma omp task in(x) out(k) bar(k, x); #pragma omp task in(m) out(l) baz(l, m); // ... } Spawn Graph foo(x, y, z) out: x, y in: z Task 1 baz(y, z) out: y in: z Task 1.1 baz(x, z) out: x in: z Task 1.2 bar(k, x) out: k in: x Task 2 baz(k, x) out: k in: x Task 2.1 baz(l, m) out: l in: m Task 3 main() inout: x,y,z,k,l,m Root Task Spawn Dependency F. Zakkak - [email protected]

Slide 6

Slide 6 text

1 Introduction 1.2 The Programming Model 2/13 The OMPSs programming model An example void baz(int *k, int *l) {*k = *l;} void foo(int *x, int *y, int *z) { #pragma omp task in(z) out(x) baz(x, z); #pragma omp task in(z) out(y) baz(y, z); } void bar(int *k, int *l) { #pragma omp task in(l) out(k) baz(k, l); } int main(void) { // ... #pragma omp task in(z) out(x, y) foo(x, y, z); #pragma omp task in(x) out(k) bar(k, x); #pragma omp task in(m) out(l) baz(l, m); // ... } Spawn Graph foo(x, y, z) out: x, y in: z Task 1 baz(y, z) out: y in: z Task 1.1 baz(x, z) out: x in: z Task 1.2 bar(k, x) out: k in: x Task 2 baz(k, x) out: k in: x Task 2.1 baz(l, m) out: l in: m Task 3 main() inout: x,y,z,k,l,m Root Task Spawn Dependency F. Zakkak - [email protected]

Slide 7

Slide 7 text

1 Introduction 1.2 The Programming Model 2/13 The OMPSs programming model An example void baz(int *k, int *l) {*k = *l;} void foo(int *x, int *y, int *z) { #pragma omp task in(z) out(x) baz(x, z); #pragma omp task in(z) out(y) baz(y, z); } void bar(int *k, int *l) { #pragma omp task in(l) out(k) baz(k, l); } int main(void) { // ... #pragma omp task in(z) out(x, y) foo(x, y, z); #pragma omp task in(x) out(k) bar(k, x); #pragma omp task in(m) out(l) baz(l, m); // ... } Spawn Graph foo(x, y, z) out: x, y in: z Task 1 baz(y, z) out: y in: z Task 1.1 baz(x, z) out: x in: z Task 1.2 bar(k, x) out: k in: x Task 2 baz(k, x) out: k in: x Task 2.1 baz(l, m) out: l in: m Task 3 main() inout: x,y,z,k,l,m Root Task Spawn Dependency F. Zakkak - [email protected]

Slide 8

Slide 8 text

1 Introduction 1.2 The Programming Model 2/13 The OMPSs programming model An example void baz(int *k, int *l) {*k = *l;} void foo(int *x, int *y, int *z) { #pragma omp task in(z) out(x) baz(x, z); #pragma omp task in(z) out(y) baz(y, z); } void bar(int *k, int *l) { #pragma omp task in(l) out(k) baz(k, l); } int main(void) { // ... #pragma omp task in(z) out(x, y) foo(x, y, z); #pragma omp task in(x) out(k) bar(k, x); #pragma omp task in(m) out(l) baz(l, m); // ... } Spawn Graph foo(x, y, z) out: x, y in: z Task 1 baz(y, z) out: y in: z Task 1.1 baz(x, z) out: x in: z Task 1.2 bar(k, x) out: k in: x Task 2 baz(k, x) out: k in: x Task 2.1 baz(l, m) out: l in: m Task 3 main() inout: x,y,z,k,l,m Root Task Spawn Dependency F. Zakkak - [email protected]

Slide 9

Slide 9 text

1 Introduction 1.2 The Programming Model 2/13 The OMPSs programming model An example void baz(int *k, int *l) {*k = *l;} void foo(int *x, int *y, int *z) { #pragma omp task in(z) out(x) baz(x, z); #pragma omp task in(z) out(y) baz(y, z); } void bar(int *k, int *l) { #pragma omp task in(l) out(k) baz(k, l); } int main(void) { // ... #pragma omp task in(z) out(x, y) foo(x, y, z); #pragma omp task in(x) out(k) bar(k, x); #pragma omp task in(m) out(l) baz(l, m); // ... } Spawn Graph foo(x, y, z) out: x, y in: z Task 1 baz(y, z) out: y in: z Task 1.1 baz(x, z) out: x in: z Task 1.2 bar(k, x) out: k in: x Task 2 baz(k, x) out: k in: x Task 2.1 baz(l, m) out: l in: m Task 3 main() inout: x,y,z,k,l,m Root Task Spawn Dependency F. Zakkak - [email protected]

Slide 10

Slide 10 text

1 Introduction 1.2 The Programming Model 2/13 The OMPSs programming model An example void baz(int *k, int *l) {*k = *l;} void foo(int *x, int *y, int *z) { #pragma omp task in(z) out(x) baz(x, z); #pragma omp task in(z) out(y) baz(y, z); } void bar(int *k, int *l) { #pragma omp task in(l) out(k) baz(k, l); } int main(void) { // ... #pragma omp task in(z) out(x, y) foo(x, y, z); #pragma omp task in(x) out(k) bar(k, x); #pragma omp task in(m) out(l) baz(l, m); // ... } Spawn Graph foo(x, y, z) out: x, y in: z Task 1 baz(y, z) out: y in: z Task 1.1 baz(x, z) out: x in: z Task 1.2 bar(k, x) out: k in: x Task 2 baz(k, x) out: k in: x Task 2.1 baz(l, m) out: l in: m Task 3 main() inout: x,y,z,k,l,m Root Task Spawn Dependency F. Zakkak - [email protected]

Slide 11

Slide 11 text

1 Introduction 1.2 The Programming Model 2/13 The OMPSs programming model An example void baz(int *k, int *l) {*k = *l;} void foo(int *x, int *y, int *z) { #pragma omp task in(z) out(x) baz(x, z); #pragma omp task in(z) out(y) baz(y, z); } void bar(int *k, int *l) { #pragma omp task in(l) out(k) baz(k, l); } int main(void) { // ... #pragma omp task in(z) out(x, y) foo(x, y, z); #pragma omp task in(x) out(k) bar(k, x); #pragma omp task in(m) out(l) baz(l, m); // ... } Spawn Graph foo(x, y, z) out: x, y in: z Task 1 baz(y, z) out: y in: z Task 1.1 baz(x, z) out: x in: z Task 1.2 bar(k, x) out: k in: x Task 2 baz(k, x) out: k in: x Task 2.1 baz(l, m) out: l in: m Task 3 main() inout: x,y,z,k,l,m Root Task Spawn Dependency F. Zakkak - [email protected]

Slide 12

Slide 12 text

1 Introduction 1.2 The Programming Model 2/13 The OMPSs programming model An example void baz(int *k, int *l) {*k = *l;} void foo(int *x, int *y, int *z) { #pragma omp task in(z) out(x) baz(x, z); #pragma omp task in(z) out(y) baz(y, z); } void bar(int *k, int *l) { #pragma omp task in(l) out(k) baz(k, l); } int main(void) { // ... #pragma omp task in(z) out(x, y) foo(x, y, z); #pragma omp task in(x) out(k) bar(k, x); #pragma omp task in(m) out(l) baz(l, m); // ... } Spawn Graph foo(x, y, z) out: x, y in: z Task 1 baz(y, z) out: y in: z Task 1.1 baz(x, z) out: x in: z Task 1.2 bar(k, x) out: k in: x Task 2 baz(k, x) out: k in: x Task 2.1 baz(l, m) out: l in: m Task 3 main() inout: x,y,z,k,l,m Root Task Spawn Dependency F. Zakkak - [email protected]

Slide 13

Slide 13 text

1 Introduction 1.2 The Programming Model 2/13 The OMPSs programming model An example void baz(int *k, int *l) {*k = *l;} void foo(int *x, int *y, int *z) { #pragma omp task in(z) out(x) baz(x, z); #pragma omp task in(z) out(y) baz(y, z); } void bar(int *k, int *l) { #pragma omp task in(l) out(k) baz(k, l); } int main(void) { // ... #pragma omp task in(z) out(x, y) foo(x, y, z); #pragma omp task in(x) out(k) bar(k, x); #pragma omp task in(m) out(l) baz(l, m); // ... } Spawn Graph foo(x, y, z) out: x, y in: z Task 1 baz(y, z) out: y in: z Task 1.1 baz(x, z) out: x in: z Task 1.2 bar(k, x) out: k in: x Task 2 baz(k, x) out: k in: x Task 2.1 baz(l, m) out: l in: m Task 3 main() inout: x,y,z,k,l,m Root Task Spawn Dependency F. Zakkak - [email protected]

Slide 14

Slide 14 text

1 Introduction 1.2 The Programming Model 3/13 Properties of Tasks ■ Parents’ include their children’s memory footprints ■ Children’s spawns depend on their parents’ scheduling ■ Parents wait for their children to complete (not in OMPSs) ■ Two tasks are dependent when one is not a descendant-task of the other, and the intersection of their memory footprints is not the empty set foo(x, y, z) out: x, y in: z Task 1 baz(y, z) out: y in: z Task 1.1 baz(x, z) out: x in: z Task 1.2 bar(k, x) out: k in: x Task 2 baz(k, x) out: k in: x Task 2.1 baz(l, m) out: l in: m Task 3 main() inout: x,y,z,k,l,m Root Task Spawn Dependency PM Dependency F. Zakkak - [email protected]

Slide 15

Slide 15 text

1 Introduction 1.2 The Programming Model 3/13 Properties of Tasks ■ Parents’ include their children’s memory footprints ■ Children’s spawns depend on their parents’ scheduling ■ Parents wait for their children to complete (not in OMPSs) ■ Two tasks are dependent when one is not a descendant-task of the other, and the intersection of their memory footprints is not the empty set foo(x, y, z) out: x, y in: z Task 1 baz(y, z) out: y in: z Task 1.1 baz(x, z) out: x in: z Task 1.2 bar(k, x) out: k in: x Task 2 baz(k, x) out: k in: x Task 2.1 baz(l, m) out: l in: m Task 3 main() inout: x,y,z,k,l,m Root Task Spawn Dependency PM Dependency F. Zakkak - [email protected]

Slide 16

Slide 16 text

1 Introduction 1.2 The Programming Model 4/13 Observation Dependencies need to be resolved only among sibling tasks! F. Zakkak - [email protected]

Slide 17

Slide 17 text

2 The algorithm 2.1 The Base Algorithm 5/13 High Level Description The algorithm consists of 2 phases Phase 1 ■ Creates dependencies at task spawns ■ Run by parent task Phase 2 ■ Resolves dependencies at task completion ■ Run by finishing task F. Zakkak - [email protected]

Slide 18

Slide 18 text

2 The algorithm 2.1 The Base Algorithm 6/13 Metadata For each task we keep: ■ a dependencies counter ■ a resolved dependencies counter ■ a notify list ■ the last access type ■ the last owner ■ a readers’ list foo(x, y, z) out: x, y in: z Task 1 1 3 NL Task 2 Task 3 … z type: in owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 19

Slide 19 text

2 The algorithm 2.1 The Base Algorithm 6/13 Metadata For each task we keep: ■ a dependencies counter ■ a resolved dependencies counter ■ a notify list ■ the last access type ■ the last owner ■ a readers’ list foo(x, y, z) out: x, y in: z Task 1 1 3 NL Task 2 Task 3 … z type: in owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 20

Slide 20 text

2 The algorithm 2.1 The Base Algorithm 6/13 Metadata For each task we keep: ■ a dependencies counter ■ a resolved dependencies counter ■ a notify list ■ the last access type ■ the last owner ■ a readers’ list foo(x, y, z) out: x, y in: z Task 1 1 3 NL Task 2 Task 3 … z type: in owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 21

Slide 21 text

2 The algorithm 2.1 The Base Algorithm 6/13 Metadata For each task we keep: ■ a dependencies counter ■ a resolved dependencies counter ■ a notify list ■ the last access type ■ the last owner ■ a readers’ list foo(x, y, z) out: x, y in: z Task 1 1 3 NL Task 2 Task 3 … z type: in owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 22

Slide 22 text

2 The algorithm 2.1 The Base Algorithm 6/13 Metadata For each task we keep: ■ a dependencies counter ■ a resolved dependencies counter ■ a notify list For each argument we keep: ■ the last access type ■ the last owner ■ a readers’ list foo(x, y, z) out: x, y in: z Task 1 1 3 NL Task 2 Task 3 … z type: in owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 23

Slide 23 text

2 The algorithm 2.1 The Base Algorithm 6/13 Metadata For each task we keep: ■ a dependencies counter ■ a resolved dependencies counter ■ a notify list For each argument we keep: ■ the last access type ■ the last owner ■ a readers’ list foo(x, y, z) out: x, y in: z Task 1 1 3 NL Task 2 Task 3 … z type: in owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 24

Slide 24 text

2 The algorithm 2.1 The Base Algorithm 6/13 Metadata For each task we keep: ■ a dependencies counter ■ a resolved dependencies counter ■ a notify list For each argument we keep: ■ the last access type ■ the last owner ■ a readers’ list foo(x, y, z) out: x, y in: z Task 1 1 3 NL Task 2 Task 3 … z type: in owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 25

Slide 25 text

2 The algorithm 2.1 The Base Algorithm 6/13 Metadata For each task we keep: ■ a dependencies counter ■ a resolved dependencies counter ■ a notify list For each argument we keep: ■ the last access type ■ the last owner ■ a readers’ list foo(x, y, z) out: x, y in: z Task 1 1 3 NL Task 2 Task 3 … z type: in owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 26

Slide 26 text

2 The algorithm 2.1 The Base Algorithm 7/13 Phase 1: Dependency Creation x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: in type: out type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 0 NL 1 2 readwrite(x) inout: x Task 4 0 0 NL 1 print(x) in: x Task 5 0 0 1 NL read(x) in: x Task 6 0 0 1 NL F. Zakkak - [email protected]

Slide 27

Slide 27 text

2 The algorithm 2.1 The Base Algorithm 7/13 Phase 1: Dependency Creation x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: in type: out type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 0 NL 1 2 readwrite(x) inout: x Task 4 0 0 NL 1 print(x) in: x Task 5 0 0 1 NL read(x) in: x Task 6 0 0 1 NL F. Zakkak - [email protected]

Slide 28

Slide 28 text

2 The algorithm 2.1 The Base Algorithm 7/13 Phase 1: Dependency Creation x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: in type: out type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 0 NL 1 2 readwrite(x) inout: x Task 4 0 0 NL 1 print(x) in: x Task 5 0 0 1 NL read(x) in: x Task 6 0 0 1 NL F. Zakkak - [email protected]

Slide 29

Slide 29 text

2 The algorithm 2.1 The Base Algorithm 7/13 Phase 1: Dependency Creation x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: in type: out type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 0 NL 1 2 readwrite(x) inout: x Task 4 0 0 NL 1 print(x) in: x Task 5 0 0 1 NL read(x) in: x Task 6 0 0 1 NL F. Zakkak - [email protected]

Slide 30

Slide 30 text

2 The algorithm 2.1 The Base Algorithm 7/13 Phase 1: Dependency Creation x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: in type: out type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 0 NL 1 2 readwrite(x) inout: x Task 4 0 0 NL 1 print(x) in: x Task 5 0 0 1 NL read(x) in: x Task 6 0 0 1 NL F. Zakkak - [email protected]

Slide 31

Slide 31 text

2 The algorithm 2.1 The Base Algorithm 7/13 Phase 1: Dependency Creation x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: in type: out type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 0 NL 1 2 readwrite(x) inout: x Task 4 0 0 NL 1 print(x) in: x Task 5 0 0 1 NL read(x) in: x Task 6 0 0 1 NL F. Zakkak - [email protected]

Slide 32

Slide 32 text

2 The algorithm 2.1 The Base Algorithm 7/13 Phase 1: Dependency Creation x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: in type: out type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 0 NL 1 2 readwrite(x) inout: x Task 4 0 0 NL 1 print(x) in: x Task 5 0 0 1 NL read(x) in: x Task 6 0 0 1 NL F. Zakkak - [email protected]

Slide 33

Slide 33 text

2 The algorithm 2.1 The Base Algorithm 7/13 Phase 1: Dependency Creation x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: in type: out type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 0 NL 1 2 readwrite(x) inout: x Task 4 0 0 NL 1 print(x) in: x Task 5 0 0 1 NL read(x) in: x Task 6 0 0 1 NL F. Zakkak - [email protected]

Slide 34

Slide 34 text

2 The algorithm 2.1 The Base Algorithm 7/13 Phase 1: Dependency Creation x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: in type: out type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 0 NL 1 2 readwrite(x) inout: x Task 4 0 0 NL 1 print(x) in: x Task 5 0 0 1 NL read(x) in: x Task 6 0 0 1 NL F. Zakkak - [email protected]

Slide 35

Slide 35 text

2 The algorithm 2.1 The Base Algorithm 7/13 Phase 1: Dependency Creation x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: in type: out type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 0 NL 1 2 readwrite(x) inout: x Task 4 0 0 NL 1 print(x) in: x Task 5 0 0 1 NL read(x) in: x Task 6 0 0 1 NL F. Zakkak - [email protected]

Slide 36

Slide 36 text

2 The algorithm 2.1 The Base Algorithm 7/13 Phase 1: Dependency Creation x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: in type: out type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 0 NL 1 2 readwrite(x) inout: x Task 4 0 0 NL 1 print(x) in: x Task 5 0 0 1 NL read(x) in: x Task 6 0 0 1 NL F. Zakkak - [email protected]

Slide 37

Slide 37 text

2 The algorithm 2.1 The Base Algorithm 7/13 Phase 1: Dependency Creation x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: in type: out type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 0 NL 1 2 readwrite(x) inout: x Task 4 0 0 NL 1 print(x) in: x Task 5 0 0 1 NL read(x) in: x Task 6 0 0 1 NL F. Zakkak - [email protected]

Slide 38

Slide 38 text

2 The algorithm 2.1 The Base Algorithm 7/13 Phase 1: Dependency Creation x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: in type: out type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 0 NL 1 2 readwrite(x) inout: x Task 4 0 0 NL 1 print(x) in: x Task 5 0 0 1 NL read(x) in: x Task 6 0 0 1 NL F. Zakkak - [email protected]

Slide 39

Slide 39 text

2 The algorithm 2.1 The Base Algorithm 7/13 Phase 1: Dependency Creation x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: in type: out type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 0 NL 1 2 readwrite(x) inout: x Task 4 0 0 NL 1 print(x) in: x Task 5 0 0 1 NL read(x) in: x Task 6 0 0 1 NL F. Zakkak - [email protected]

Slide 40

Slide 40 text

2 The algorithm 2.1 The Base Algorithm 7/13 Phase 1: Dependency Creation x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: in type: out type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 0 NL 1 2 readwrite(x) inout: x Task 4 0 0 NL 1 print(x) in: x Task 5 0 0 1 NL read(x) in: x Task 6 0 0 1 NL F. Zakkak - [email protected]

Slide 41

Slide 41 text

2 The algorithm 2.1 The Base Algorithm 7/13 Phase 1: Dependency Creation x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: in type: out type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 0 NL 1 2 readwrite(x) inout: x Task 4 0 0 NL 1 print(x) in: x Task 5 0 0 1 NL read(x) in: x Task 6 0 0 1 NL F. Zakkak - [email protected]

Slide 42

Slide 42 text

2 The algorithm 2.1 The Base Algorithm 7/13 Phase 1: Dependency Creation x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: in type: out type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 0 NL 1 2 readwrite(x) inout: x Task 4 0 0 NL 1 print(x) in: x Task 5 0 0 1 NL read(x) in: x Task 6 0 0 1 NL F. Zakkak - [email protected]

Slide 43

Slide 43 text

2 The algorithm 2.1 The Base Algorithm 7/13 Phase 1: Dependency Creation x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: in type: out type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 0 NL 1 2 readwrite(x) inout: x Task 4 0 0 NL 1 print(x) in: x Task 5 0 0 1 NL read(x) in: x Task 6 0 0 1 NL F. Zakkak - [email protected]

Slide 44

Slide 44 text

2 The algorithm 2.1 The Base Algorithm 7/13 Phase 1: Dependency Creation x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: in type: out type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 0 NL 1 2 readwrite(x) inout: x Task 4 0 0 NL 1 print(x) in: x Task 5 0 0 1 NL read(x) in: x Task 6 0 0 1 NL F. Zakkak - [email protected]

Slide 45

Slide 45 text

2 The algorithm 2.1 The Base Algorithm 7/13 Phase 1: Dependency Creation x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: in type: out type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 0 NL 1 2 readwrite(x) inout: x Task 4 0 0 NL 1 print(x) in: x Task 5 0 0 1 NL read(x) in: x Task 6 0 0 1 NL F. Zakkak - [email protected]

Slide 46

Slide 46 text

2 The algorithm 2.1 The Base Algorithm 7/13 Phase 1: Dependency Creation x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: in type: out type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 0 NL 1 2 readwrite(x) inout: x Task 4 0 0 NL 1 print(x) in: x Task 5 0 0 1 NL read(x) in: x Task 6 0 0 1 NL F. Zakkak - [email protected]

Slide 47

Slide 47 text

2 The algorithm 2.1 The Base Algorithm 7/13 Phase 1: Dependency Creation x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: in type: out type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 0 NL 1 2 readwrite(x) inout: x Task 4 0 0 NL 1 print(x) in: x Task 5 0 0 1 NL read(x) in: x Task 6 0 0 1 NL F. Zakkak - [email protected]

Slide 48

Slide 48 text

2 The algorithm 2.1 The Base Algorithm 8/13 Phase 2: Dependency Resolution x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 1 2 2 NL readwrite(x) inout: x Task 4 0 1 1 NL print(x) in: x Task 5 0 1 1 NL read(x) in: x Task 6 0 1 1 NL F. Zakkak - [email protected]

Slide 49

Slide 49 text

2 The algorithm 2.1 The Base Algorithm 8/13 Phase 2: Dependency Resolution x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 1 2 2 NL readwrite(x) inout: x Task 4 0 1 1 NL print(x) in: x Task 5 0 1 1 NL read(x) in: x Task 6 0 1 1 NL F. Zakkak - [email protected]

Slide 50

Slide 50 text

2 The algorithm 2.1 The Base Algorithm 8/13 Phase 2: Dependency Resolution x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 1 2 2 NL readwrite(x) inout: x Task 4 0 1 1 NL print(x) in: x Task 5 0 1 1 NL read(x) in: x Task 6 0 1 1 NL F. Zakkak - [email protected]

Slide 51

Slide 51 text

2 The algorithm 2.1 The Base Algorithm 8/13 Phase 2: Dependency Resolution x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 1 2 2 NL readwrite(x) inout: x Task 4 0 1 1 NL print(x) in: x Task 5 0 1 1 NL read(x) in: x Task 6 0 1 1 NL F. Zakkak - [email protected]

Slide 52

Slide 52 text

2 The algorithm 2.1 The Base Algorithm 8/13 Phase 2: Dependency Resolution x = malloc(...); #pragma omp task in(x) print(x) /* 1st */ #pragma omp task in(x) read(x) /* RAR */ #pragma omp task out(x) write(x) /* WAR */ #pragma omp task inout(x) readwrite(x) /* WAW */ #pragma omp task in(x) print(x) /* RAW */ #pragma omp task in(x) read(x) /* RAR */ type: inout x owner RL print(x) in: x Task 1 0 0 NL read(x) in: x Task 2 0 0 NL Task 3 write(x) out: x 0 1 2 2 NL readwrite(x) inout: x Task 4 0 1 1 NL print(x) in: x Task 5 0 1 1 NL read(x) in: x Task 6 0 1 1 NL F. Zakkak - [email protected]

Slide 53

Slide 53 text

3 PARTEE 3.1 The Runtime System 9/13 Metadata handling Per-task lookup tables (LUTs) ■ Hold only associations for the arguments accessed by the children tasks ■ Inherent distribution of metadata ■ No contention ■ Increased spatial and temporal locality ■ Implemented as tries ■ Memory address as key foo(x, y, z) out: x, y in: z Task 1 1 3 LUT NL Task 2 Task 3 … type: in z owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 54

Slide 54 text

3 PARTEE 3.1 The Runtime System 9/13 Metadata handling Per-task lookup tables (LUTs) ■ Hold only associations for the arguments accessed by the children tasks ■ Inherent distribution of metadata ■ No contention ■ Increased spatial and temporal locality ■ Implemented as tries ■ Memory address as key foo(x, y, z) out: x, y in: z Task 1 1 3 LUT NL Task 2 Task 3 … type: in z owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 55

Slide 55 text

3 PARTEE 3.1 The Runtime System 9/13 Metadata handling Per-task lookup tables (LUTs) ■ Hold only associations for the arguments accessed by the children tasks ■ Inherent distribution of metadata ■ No contention ■ Increased spatial and temporal locality ■ Implemented as tries ■ Memory address as key foo(x, y, z) out: x, y in: z Task 1 1 3 LUT NL Task 2 Task 3 … type: in z owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 56

Slide 56 text

3 PARTEE 3.1 The Runtime System 9/13 Metadata handling Per-task lookup tables (LUTs) ■ Hold only associations for the arguments accessed by the children tasks ■ Inherent distribution of metadata ■ No contention ■ Increased spatial and temporal locality ■ Implemented as tries ■ Memory address as key foo(x, y, z) out: x, y in: z Task 1 1 3 LUT NL Task 2 Task 3 … type: in z owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 57

Slide 57 text

3 PARTEE 3.1 The Runtime System 9/13 Metadata handling Per-task lookup tables (LUTs) ■ Hold only associations for the arguments accessed by the children tasks ■ Inherent distribution of metadata ■ No contention ■ Increased spatial and temporal locality ■ Implemented as tries ■ Memory address as key foo(x, y, z) out: x, y in: z Task 1 1 3 LUT NL Task 2 Task 3 … type: in z owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 58

Slide 58 text

3 PARTEE 3.1 The Runtime System 9/13 Metadata handling Per-task lookup tables (LUTs) ■ Hold only associations for the arguments accessed by the children tasks ■ Inherent distribution of metadata ■ No contention ■ Increased spatial and temporal locality ■ Implemented as tries ■ Memory address as key foo(x, y, z) out: x, y in: z Task 1 1 3 LUT NL Task 2 Task 3 … type: in z owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 59

Slide 59 text

3 PARTEE 3.1 The Runtime System 9/13 Metadata handling Per-task lookup tables (LUTs) ■ Hold only associations for the arguments accessed by the children tasks ■ Inherent distribution of metadata ■ No contention ■ Increased spatial and temporal locality ■ Implemented as tries ■ Memory address as key foo(x, y, z) out: x, y in: z Task 1 1 3 LUT NL Task 2 Task 3 … type: in z owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 60

Slide 60 text

3 PARTEE 3.1 The Runtime System 9/13 Metadata handling Per-task lookup tables (LUTs) ■ Hold only associations for the arguments accessed by the children tasks ■ Inherent distribution of metadata ■ No contention ■ Increased spatial and temporal locality ■ Implemented as tries ■ Memory address as key foo(x, y, z) out: x, y in: z Task 1 1 3 LUT NL Task 2 Task 3 … type: in z owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 61

Slide 61 text

3 PARTEE 3.1 The Runtime System 10/13 Memory Handling ■ Custom region allocator ■ Each task owns a region in which it allocates its: □ Lookup Table (LUT) □ Children’s task descriptors ■ Bulk deallocation at task completion foo(x, y, z) out: x, y in: z Task 1 1 3 3 LUT NL Task 2 Task 3 type: in z owner RL Task 1.1 … read(z) in: z Task 1.1 0 0 LUT NL F. Zakkak - [email protected]

Slide 62

Slide 62 text

3 PARTEE 3.1 The Runtime System 10/13 Memory Handling ■ Custom region allocator ■ Each task owns a region in which it allocates its: □ Lookup Table (LUT) □ Children’s task descriptors ■ Bulk deallocation at task completion foo(x, y, z) out: x, y in: z Task 1 1 3 3 LUT NL Task 2 Task 3 type: in z owner RL Task 1.1 … read(z) in: z Task 1.1 0 0 LUT NL F. Zakkak - [email protected]

Slide 63

Slide 63 text

3 PARTEE 3.1 The Runtime System 10/13 Memory Handling ■ Custom region allocator ■ Each task owns a region in which it allocates its: □ Lookup Table (LUT) □ Children’s task descriptors ■ Bulk deallocation at task completion foo(x, y, z) out: x, y in: z Task 1 1 3 3 LUT NL Task 2 Task 3 type: in z owner RL Task 1.1 … read(z) in: z Task 1.1 0 0 LUT NL F. Zakkak - [email protected]

Slide 64

Slide 64 text

3 PARTEE 3.1 The Runtime System 10/13 Memory Handling ■ Custom region allocator ■ Each task owns a region in which it allocates its: □ Lookup Table (LUT) □ Children’s task descriptors ■ Bulk deallocation at task completion foo(x, y, z) out: x, y in: z Task 1 1 3 3 LUT NL Task 2 Task 3 type: in z owner RL Task 1.1 … read(z) in: z Task 1.1 0 0 LUT NL F. Zakkak - [email protected]

Slide 65

Slide 65 text

3 PARTEE 3.1 The Runtime System 10/13 Memory Handling ■ Custom region allocator ■ Each task owns a region in which it allocates its: □ Lookup Table (LUT) □ Children’s task descriptors ■ Bulk deallocation at task completion foo(x, y, z) out: x, y in: z Task 1 1 3 3 LUT NL Task 2 Task 3 type: in z owner RL Task 1.1 … read(z) in: z Task 1.1 0 0 LUT NL F. Zakkak - [email protected]

Slide 66

Slide 66 text

4 Evaluation 11/13 Evaluation Setup ■ 4-chip NUMA system ■ Total of 64 AMD Opteron Processor 6272 Cores ■ Total of 256GB RAM ■ Native sequential executions as baseline ■ Geometric mean over 10 runs F. Zakkak - [email protected]

Slide 67

Slide 67 text

4 Evaluation 12/13 Performance PARTEE PARTEE ND Cilk Nanos++ Linear 0.35 1 2 4 8 16 32 64 1 2 4 8 16 32 64 Speedup Cores Blackscholes 1 2 4 8 16 32 64 Cores Matrix Multiply F. Zakkak - [email protected]

Slide 68

Slide 68 text

4 Evaluation 12/13 Performance PARTEE PARTEE ND Cilk Nanos++ Linear 0.35 1 2 4 8 16 32 64 1 2 4 8 16 32 64 Speedup Cores Heat Diffusion 1 2 4 8 16 32 64 Cores Mergesort F. Zakkak - [email protected]

Slide 69

Slide 69 text

4 Evaluation 12/13 Performance PARTEE PARTEE ND Cilk Nanos++ Linear 0.35 1 2 4 8 16 32 64 1 2 4 8 16 32 64 Speedup Cores Cholesky 1 2 4 8 16 32 64 Cores LU Decomposition F. Zakkak - [email protected]

Slide 70

Slide 70 text

5 Remarks 13/13 Remarks ■ PARTEE brings Cilk-like performance to OMPSs ■ Up to 2× better than Cilk in applications with irregular dependencies ■ PARTEE is licensed under the Apache License v2.0 and can be found at https://github.com/CARV-ICS-FORTH/partee (Shortened: https://is.gd/ipdps_partee) F. Zakkak - [email protected]

Slide 71

Slide 71 text

5 Remarks 13/13 Remarks ■ PARTEE brings Cilk-like performance to OMPSs ■ Up to 2× better than Cilk in applications with irregular dependencies ■ PARTEE is licensed under the Apache License v2.0 and can be found at https://github.com/CARV-ICS-FORTH/partee (Shortened: https://is.gd/ipdps_partee) F. Zakkak - [email protected]

Slide 72

Slide 72 text

5 Remarks 13/13 Remarks ■ PARTEE brings Cilk-like performance to OMPSs ■ Up to 2× better than Cilk in applications with irregular dependencies ■ PARTEE is licensed under the Apache License v2.0 and can be found at https://github.com/CARV-ICS-FORTH/partee (Shortened: https://is.gd/ipdps_partee) F. Zakkak - [email protected]

Slide 73

Slide 73 text

5 Remarks 13/13 Remarks ■ PARTEE brings Cilk-like performance to OMPSs ■ Up to 2× better than Cilk in applications with irregular dependencies ■ PARTEE is licensed under the Apache License v2.0 and can be found at https://github.com/CARV-ICS-FORTH/partee (Shortened: https://is.gd/ipdps_partee) Thank You! F. Zakkak - [email protected]

Slide 74

Slide 74 text

6 Backup Slides 13/13 Overhead vs Task granularity 100 101 102 103 100 101 102 103 Task Time (μs) Workload (μs) Native PARTEE Cilk Nanos++ F. Zakkak - [email protected]

Slide 75

Slide 75 text

6 Backup Slides 13/13 Metadata handling Per-task lookup tables (LUTs) ■ Hold only associations for the arguments accessed by the children tasks ■ Inherent distribution of metadata ■ No contention ■ Increased spatial and temporal locality ■ Implemented as tries ■ Memory address as key foo(x, y, z) out: x, y in: z Task 1 1 3 LUT NL Task 2 Task 3 … type: in z owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 76

Slide 76 text

6 Backup Slides 13/13 Metadata handling Per-task lookup tables (LUTs) ■ Hold only associations for the arguments accessed by the children tasks ■ Inherent distribution of metadata ■ No contention ■ Increased spatial and temporal locality ■ Implemented as tries ■ Memory address as key foo(x, y, z) out: x, y in: z Task 1 1 3 LUT NL Task 2 Task 3 … type: in z owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 77

Slide 77 text

6 Backup Slides 13/13 Metadata handling Per-task lookup tables (LUTs) ■ Hold only associations for the arguments accessed by the children tasks ■ Inherent distribution of metadata ■ No contention ■ Increased spatial and temporal locality ■ Implemented as tries ■ Memory address as key foo(x, y, z) out: x, y in: z Task 1 1 3 LUT NL Task 2 Task 3 … type: in z owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 78

Slide 78 text

6 Backup Slides 13/13 Metadata handling Per-task lookup tables (LUTs) ■ Hold only associations for the arguments accessed by the children tasks ■ Inherent distribution of metadata ■ No contention ■ Increased spatial and temporal locality ■ Implemented as tries ■ Memory address as key foo(x, y, z) out: x, y in: z Task 1 1 3 LUT NL Task 2 Task 3 … type: in z owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 79

Slide 79 text

6 Backup Slides 13/13 Metadata handling Per-task lookup tables (LUTs) ■ Hold only associations for the arguments accessed by the children tasks ■ Inherent distribution of metadata ■ No contention ■ Increased spatial and temporal locality ■ Implemented as tries ■ Memory address as key foo(x, y, z) out: x, y in: z Task 1 1 3 LUT NL Task 2 Task 3 … type: in z owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 80

Slide 80 text

6 Backup Slides 13/13 Metadata handling Per-task lookup tables (LUTs) ■ Hold only associations for the arguments accessed by the children tasks ■ Inherent distribution of metadata ■ No contention ■ Increased spatial and temporal locality ■ Implemented as tries ■ Memory address as key foo(x, y, z) out: x, y in: z Task 1 1 3 LUT NL Task 2 Task 3 … type: in z owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 81

Slide 81 text

6 Backup Slides 13/13 Metadata handling Per-task lookup tables (LUTs) ■ Hold only associations for the arguments accessed by the children tasks ■ Inherent distribution of metadata ■ No contention ■ Increased spatial and temporal locality ■ Implemented as tries ■ Memory address as key foo(x, y, z) out: x, y in: z Task 1 1 3 LUT NL Task 2 Task 3 … type: in z owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 82

Slide 82 text

6 Backup Slides 13/13 Metadata handling Per-task lookup tables (LUTs) ■ Hold only associations for the arguments accessed by the children tasks ■ Inherent distribution of metadata ■ No contention ■ Increased spatial and temporal locality ■ Implemented as tries ■ Memory address as key foo(x, y, z) out: x, y in: z Task 1 1 3 LUT NL Task 2 Task 3 … type: in z owner RL Task 1 Task 4 … F. Zakkak - [email protected]

Slide 83

Slide 83 text

6 Backup Slides 13/13 Memory Handling ■ Custom region allocator ■ Each task owns a region in which it allocates its: □ Lookup Table (LUT) □ Children’s task descriptors ■ Bulk deallocation at task completion foo(x, y, z) out: x, y in: z Task 1 1 3 3 LUT NL Task 2 Task 3 type: in z owner RL Task 1.1 … read(z) in: z Task 1.1 0 0 LUT NL F. Zakkak - [email protected]

Slide 84

Slide 84 text

6 Backup Slides 13/13 Memory Handling ■ Custom region allocator ■ Each task owns a region in which it allocates its: □ Lookup Table (LUT) □ Children’s task descriptors ■ Bulk deallocation at task completion foo(x, y, z) out: x, y in: z Task 1 1 3 3 LUT NL Task 2 Task 3 type: in z owner RL Task 1.1 … read(z) in: z Task 1.1 0 0 LUT NL F. Zakkak - [email protected]

Slide 85

Slide 85 text

6 Backup Slides 13/13 Memory Handling ■ Custom region allocator ■ Each task owns a region in which it allocates its: □ Lookup Table (LUT) □ Children’s task descriptors ■ Bulk deallocation at task completion foo(x, y, z) out: x, y in: z Task 1 1 3 3 LUT NL Task 2 Task 3 type: in z owner RL Task 1.1 … read(z) in: z Task 1.1 0 0 LUT NL F. Zakkak - [email protected]

Slide 86

Slide 86 text

6 Backup Slides 13/13 Memory Handling ■ Custom region allocator ■ Each task owns a region in which it allocates its: □ Lookup Table (LUT) □ Children’s task descriptors ■ Bulk deallocation at task completion foo(x, y, z) out: x, y in: z Task 1 1 3 3 LUT NL Task 2 Task 3 type: in z owner RL Task 1.1 … read(z) in: z Task 1.1 0 0 LUT NL F. Zakkak - [email protected]

Slide 87

Slide 87 text

6 Backup Slides 13/13 Memory Handling ■ Custom region allocator ■ Each task owns a region in which it allocates its: □ Lookup Table (LUT) □ Children’s task descriptors ■ Bulk deallocation at task completion foo(x, y, z) out: x, y in: z Task 1 1 3 3 LUT NL Task 2 Task 3 type: in z owner RL Task 1.1 … read(z) in: z Task 1.1 0 0 LUT NL F. Zakkak - [email protected]