where are you going with those types
play

Where are you going with those types? Vincent St-Amour, Sam - PowerPoint PPT Presentation

Mar 17, 2024 •128 likes •839 views

Where are you going with those types? Vincent St-Amour, Sam Tobin-Hochstadt, Matthew Flatt, Matthias Felleisen PLT / Northeastern University Boston, MA, USA PLT / University of Utah Salt Lake City, UT, USA IFL 2010 - September 3rd, 2010

  1. Where are you going with those types? Vincent St-Amour, Sam Tobin-Hochstadt, Matthew Flatt, Matthias Felleisen PLT / Northeastern University Boston, MA, USA PLT / University of Utah Salt Lake City, UT, USA IFL 2010 - September 3rd, 2010

  2. Generating fast code in the presence of ad-hoc polymorphism is hard.

  3. Case study: generic arithmetic (+ 2 2)

  4. Case study: generic arithmetic (+ 2 2) (+ 2.3 2.4)

  5. Case study: generic arithmetic (+ 2 2) (+ 2.3 2.4) (+ 2.3 2)

  6. Case study: generic arithmetic (+ 2 2) (+ 2.3 2.4) (+ 2.3 2) (+ 2+3i 2+4i)

  7. Case study: generic arithmetic (+ 2 2) (+ 2.3 2.4) (+ 2.3 2) (+ 2+3i 2+4i) Types of arguments are not known statically.

  8. Case study: generic arithmetic #lang racket (+ 2.3 2.4)

  9. Case study: generic arithmetic #lang racket (+ 2.3 2.4) (define (add x y) (cond ((and (float? x) (float? y)) (let* ([val-x (strip-type-tag x)] [val-y (strip-type-tag y)] [result (add-floats val-x val-y)]) (tag-as-float result))) ((and (integer? x) (integer? y)) ...) ((and (complex? x) (complex? y)) ...) (else (error))))

  10. Case study: generic arithmetic #lang racket (+ 2.3 2.4) (define (add x y) (cond ((and (float? x) (float? y)) (let* ([val-x (strip-type-tag x)] [val-y (strip-type-tag y)] [result (add-floats val-x val-y)]) (tag-as-float result))) ((and (integer? x) (integer? y)) ...) ((and (complex? x) (complex? y)) ...) (else (error))))

  11. Our solution • Type-specialized primitives • Composition of: • Type-driven rewriting • Primitives drive optimization Typechecker τ Rewriting Primitives Compiler

  12. Implementation • Typed Racket • Higher-order functional language • Generic arithmetic (and complexes)

  13. Implementation • Typed Racket • Higher-order functional language • Generic arithmetic (and complexes) Applicable to other languages

  14. Type-specialized primitives

  15. #lang racket (fl+ x y)

  16. #lang racket (fl+ 2.3 2.4)

  17. #lang racket (fl+ 2.3 2.4) 4.7

  18. #lang racket (fl+ 2 2)

  19. #lang racket (fl+ 2 2) segmentation fault

  20. #lang typed/racket (let: ([x : Float 2.3] [y : Float 2.4]) (fl+ x y))

  21. #lang typed/racket (let: ([x : Float 2.3] [y : Float 2.4]) (fl+ x y))

  22. #lang typed/racket (let: ([x : Float 2.3] [y : Float 2.4]) (fl+ x y)) 4.7

  23. #lang typed/racket (let: ([x : Integer 2] [y : Integer 2]) (fl+ x y))

  24. #lang typed/racket (let: ([x : Integer 2] [y : Integer 2]) (fl+ x y))

  25. #lang typed/racket (let: ([x : Integer 2] [y : Integer 2]) (fl+ x y)) Type Checker: No function domains matched in function application: Domains: Float Float Arguments: Integer Integer in: (fl+ x y)

  26. Type-driven rewriting

  27. #lang typed/racket (let: ([x : Float 2.3] [y : Float 2.4]) (+ x y))

  28. #lang typed/racket (let: ([x : Float 2.3] [y : Float 2.4]) (+ x y))

  29. #lang typed/racket (let: ([x : Float 2.3] [y : Float 2.4]) (fl+ x y))

  30. #lang typed/racket (let: ([x : Float 2.3] [y : Float 2.4]) (fl+ x y)) 4.7

  31. Primitives drive optimization

  32. #lang typed/racket (* (+ x y) (+ z w))

  33. load $x $r1 load $y $r2 ... fadd $r1 $r2 $r3 ... sto $r3 $tmp1 #lang typed/racket load $z $r4 (* (+ x y) load $w $r5 (+ z w)) ... fadd $r4 $r5 $r6 ... sto $r6 $tmp2 load $tmp1 $r7 load $tmp2 $r8 ... fmul $r7 $r8 $r9 ... sto $r9 $tmp3

  34. load $x $r1 load $y $r2 ... fadd $r1 $r2 $r3 ... sto $r3 $tmp1 #lang typed/racket load $z $r4 (* (+ x y) load $w $r5 (+ z w)) ... fadd $r4 $r5 $r6 ... sto $r6 $tmp2 load $tmp1 $r7 load $tmp2 $r8 ... fmul $r3 $r6 $r9 ... sto $r9 $tmp3

  35. load $x $r1 load $y $r2 ... fadd $r1 $r2 $r3 ... sto $r3 $tmp1 #lang typed/racket load $z $r4 (* (+ x y) load $w $r5 (+ z w)) ... fadd $r4 $r5 $r6 #lang typed/racket ... (fl* (fl+ x y) sto $r6 $tmp2 (fl+ z w)) load $tmp1 $r7 load $tmp2 $r8 ... fmul $r3 $r6 $r9 ... sto $r9 $tmp3

  36. load $x $r1 load $y $r2 ... fadd $r1 $r2 $r3 ... sto $r3 $tmp1 #lang typed/racket load $z $r4 (* (+ x y) load $w $r5 (+ z w)) ... fadd $r4 $r5 $r6 #lang typed/racket ... (fl* (fl+ x y) sto $r6 $tmp2 (fl+ z w)) load $tmp1 $r7 load $tmp2 $r8 ... fmul $r3 $r6 $r9 ... sto $r9 $tmp3

  37. #lang typed/racket (let ([a (+ x y)]) (* a (- z a)))

  38. load $x $r1 load $y $r2 ... fadd $r1 $r2 $r3 ... sto $r3 $a #lang typed/racket load $z $r4 (let ([a (+ x y)]) load $a $r5 (* a (- z a))) ... fsub $r4 $r5 $r6 ... sto $r6 $tmp1 load $a $r7 load $tmp1 $r8 ... fmul $r7 $r8 $r9 ... sto $r9 $tmp2

  39. load $x $r1 load $y $r2 ... fadd $r1 $r2 $r3 ... sto $r3 $a #lang typed/racket load $z $r4 (let ([a (+ x y)]) load $a $r5 (* a (- z a))) ... fsub $r4 $r3 $r6 ... sto $r6 $tmp1 load $a $r7 load $tmp1 $r8 ... fmul $r3 $r6 $r9 ... sto $r9 $tmp2

  40. load $x $r1 load $y $r2 ... fadd $r1 $r2 $r3 ... sto $r3 $a #lang typed/racket load $z $r4 (let ([a (+ x y)]) load $a $r5 (* a (- z a))) ... fsub $r4 $r3 $r6 #lang typed/racket ... (let ([a (fl+ x y)]) sto $r6 $tmp1 (fl* a (fl- z a))) load $a $r7 load $tmp1 $r8 ... fmul $r3 $r6 $r9 ... sto $r9 $tmp2

  41. #lang typed/racket (let loop ([acc 0.0]) (if (> acc x) acc (loop (+ y acc))))

  42. mov 0.0 $r1 ... sto $r1 $acc #lang typed/racket (let loop ([acc 0.0]) loop: (if (> acc x) load $acc $r2 acc load $x $r3 ... (loop (+ y acc)))) flcmp $r2 $r3 jgt end load $y $r4 load $acc $r5 ... fadd $r4 $r5 $r6 ... sto $r6 $acc jmp loop end:

  43. mov 0.0 $r1 ... sto $r1 $acc #lang typed/racket (let loop ([acc 0.0]) loop: (if (> acc x) load $acc $r2 acc load $x $r3 ... (loop (+ y acc)))) flcmp $r2 $r3 jgt end load $y $r4 load $acc $r5 ... fadd $r4 $r5 $r6 ... sto $r6 $acc jmp loop end:

  44. mov 0.0 $r1 ... sto $r1 $acc #lang typed/racket (let loop ([acc 0.0]) loop: (if (> acc x) load $acc $r2 acc load $x $r3 ... (loop (+ y acc)))) flcmp $r1 $r3 jgt end #lang typed/racket load $y $r4 (let loop ([acc 0.0]) load $acc $r5 ... (if (fl> acc x) fadd $r4 $r1 $r6 acc ... (loop (fl+ y acc)))) sto $r6 $acc jmp loop end: sto $r1 $acc

  45. mov 0.0 $r1 ... sto $r1 $acc #lang typed/racket (let loop ([acc 0.0]) loop: (if (> acc x) load $acc $r2 acc load $x $r3 ... (loop (+ y acc)))) flcmp $r1 $r3 jgt end #lang typed/racket load $y $r4 (let loop ([acc 0.0]) load $acc $r5 ... (if (fl> acc x) fadd $r4 $r1 $r6 acc ... (loop (fl+ y acc)))) sto $r6 $acc jmp loop end: sto $r1 $acc

  46. mov 0.0 $r1 ... sto $r1 $acc load $x $r3 #lang typed/racket load $y $r4 (let loop ([acc 0.0]) loop: (if (> acc x) load $acc $r2 acc load $x $r3 ... (loop (+ y acc)))) flcmp $r1 $r3 jgt end #lang typed/racket load $y $r4 (let loop ([acc 0.0]) load $acc $r5 ... (if (fl> acc x) fadd $r4 $r1 $r6 acc ... (loop (fl+ y acc)))) sto $r6 $acc jmp loop end: sto $r1 $acc

  47. Results

  48. Speedup 2.50 benchmarks pseudoknot 15.72 mandelbrot 2.16 nbody 1.24 takl 15.91 FFT 1.60 data structures banker's queue 1.34 leftist heap 2.78 industrial application FFT Bigger is better Average of 5 runs on x86

  49. In-depth look: Industrial FFT

  50. #lang racket (let* ([x 2.3+2.4i] [y 2.5+2.6i] [z 2.7+2.8i]) (- (+ x y) z))

  51. #lang racket (let* ([x 2.3+2.4i] [y 2.5+2.6i] [z 2.7+2.8i] [x-real (real-part x)] [x-imag (imag-part x)] [y-real (real-part y)] [y-imag (imag-part y)] [z-real (real-part z)] [z-imag (imag-part z)]) (make-rectangular (- (+ x-real y-real) z-real) (- (+ x-imag y-imag) z-imag)))

  52. #lang racket (let* ([x 2.3+2.4i] [y 2.5+2.6i] [z 2.7+2.8i] [x-real (real-part x)] [x-imag (imag-part x)] [y-real (real-part y)] [y-imag (imag-part y)] [z-real (real-part z)] [z-imag (imag-part z)]) (make-rectangular (fl- (fl+ x-real y-real) z-real) (fl- (fl+ x-imag y-imag) z-imag)))

  53. #lang racket (let* ([x 2.3+2.4i] [y 2.5+2.6i] [z 2.7+2.8i] [x-real (real-part x)] [x-imag (imag-part x)] [y-real (real-part y)] [y-imag (imag-part y)] [z-real (real-part z)] [z-imag (imag-part z)]) (make-rectangular (fl- (fl+ x-real y-real) z-real) (fl- (fl+ x-imag y-imag) z-imag))) Significant manual labor

  54. #lang racket (let* ([x 2.3+2.4i] [y 2.5+2.6i] [z 2.7+2.8i] [x-real (real-part x)] [x-imag (imag-part x)] [y-real (real-part y)] [y-imag (imag-part y)] [z-real (real-part z)] [z-imag (imag-part z)]) (make-rectangular (fl- (fl+ x-real y-real) z-real) (fl- (fl+ x-imag y-imag) z-imag))) Significant manual labor Error prone

  55. #lang typed/racket (let* ([x 2.3+2.4i] [y 2.5+2.6i] [z 2.7+2.8i]) (- (+ x y) z))

  56. #lang typed/racket (let* ([x 2.3+2.4i] [y 2.5+2.6i] [z 2.7+2.8i]) (- (+ x y) z)) Unboxed intermediate results

  57. #lang typed/racket (let* ([x 2.3+2.4i] [y 2.5+2.6i] [z 2.7+2.8i]) (- (+ x y) z)) Unboxed intermediate results Unboxed let bindings

  58. #lang typed/racket (let* ([x 2.3+2.4i] [y 2.5+2.6i] [z 2.7+2.8i]) (- (+ x y) z)) Unboxed intermediate results Unboxed let bindings Unboxed loop variables

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

Data Types Gabriele Keller Ron Vanderfeesten Compound types What are types? So far, we

Data Types Gabriele Keller Ron Vanderfeesten Compound types What are types? So far, we

Concepts of Program Design Data Types Gabriele Keller Ron Vanderfeesten Compound types What are types? So far, we only looked at basic types, such as Int, Boolean, as well as function types Int Bool 1 True 2 False 3 4

651 views • 35 slides
Types Classification of Values cs3723 1 Values and Types Basic types: types of atomic

Types Classification of Values cs3723 1 Values and Types Basic types: types of atomic

Types Classification of Values cs3723 1 Values and Types Basic types: types of atomic values int, bool, character, real, symbol Compound types: types of compound values List, record, array, tuple, struct, ref, pointer Built

450 views • 16 slides
Types of Types Types of Types natural numbers. A type is a (possibly infinite) set of values.

Types of Types Types of Types natural numbers. A type is a (possibly infinite) set of values.

One-Slide Summary In lazy evaluation , expressions are not evaluated until their values are needed. We can use lazy evaluation to program with infinite data structures , such as a list of all Types of Types Types of Types natural numbers.

367 views • 8 slides
! TYPES & STATIC ANALYSIS TYPES ARE GOOD, I PROMISE. SAM GREENWOOD @SAMTGREENWOOD

! TYPES & STATIC ANALYSIS TYPES ARE GOOD, I PROMISE. SAM GREENWOOD @SAMTGREENWOOD

! TYPES & STATIC ANALYSIS TYPES ARE GOOD, I PROMISE. SAM GREENWOOD @SAMTGREENWOOD TECHNOLOGY MANAGER @ UNITI GROUP I LIKE TYPES. WITHOUT TYPES WITH TYPES WHAT ARE TYPES? DYNAMIC TYPES DEFINED WHEN WE USE THEM STATIC TYPES DEFINED

756 views • 40 slides
Types Dynamic types Types are broken down into many categories Static types Duck typing

Types Dynamic types Types are broken down into many categories Static types Duck typing

Types Dynamic types Types are broken down into many categories Static types Duck typing Dynamic types Subtypes Types are broken down into many categories Classes and subclasses Static types Strong types Dependent Duck typing

1.22k views • 98 slides
ML Tutorial 2 Polymorphism, Functions, Exceptions I/O, Modules Types Review Primitive types

ML Tutorial 2 Polymorphism, Functions, Exceptions I/O, Modules Types Review Primitive types

ML Tutorial 2 Polymorphism, Functions, Exceptions I/O, Modules Types Review Primitive types unit, int, real, char, string, ..., instream, outstream, ... Composite types unit, tuples, records function types Datatypes types and n-ary type

468 views • 11 slides
Data Types Data Types Evolution of Data Types: COSC337 FORTRAN I (1956) - INTEGER,

Data Types Data Types Evolution of Data Types: COSC337 FORTRAN I (1956) - INTEGER,

Data Types Data Types Evolution of Data Types: COSC337 FORTRAN I (1956) - INTEGER, REAL, arrays COBOL allowed users to define precision ALGOL (1968) provided a few basic types, and by Matt Evett allowed user to form new

639 views • 9 slides
Basic Types in C Dalhousie University Winter 2019 Basic Types in C Integer types: Floating

Basic Types in C Dalhousie University Winter 2019 Basic Types in C Integer types: Floating

CSCI 2132: Software Development Norbert Zeh Faculty of Computer Science Basic Types in C Dalhousie University Winter 2019 Basic Types in C Integer types: Floating point types: int float (32 bits) Most long (64 bits) double

505 views • 27 slides
References 7 January 2019 OSU CSE 1 Primitive vs. Reference Types Java types are divided

References 7 January 2019 OSU CSE 1 Primitive vs. Reference Types Java types are divided

References 7 January 2019 OSU CSE 1 Primitive vs. Reference Types Java types are divided into two different categories: The built-in types are called primitive types Includes boolean , char , int , double All other types are

1.26k views • 86 slides
Types of Types Each type supports a set of valid operations. Types can be latent or

Types of Types Each type supports a set of valid operations. Types can be latent or

One-Slide Summary In lazy evaluation , expressions are not evaluated until their values are needed. A type is a (possibly infinite) set of values. Types of Types Each type supports a set of valid operations. Types can be latent or

321 views • 5 slides
Data Types Basic data types. There are four basic data types: char one byte character in

Data Types Basic data types. There are four basic data types: char one byte character in

Data Types Basic data types. There are four basic data types: char one byte character in the local character set (typically, ASCII code) int an integer, in the natural representation of the host machine float single precision

459 views • 16 slides
Inductive Types for Free Representing Nested Inductive Types using W-types Michael Abbott (U.

Inductive Types for Free Representing Nested Inductive Types using W-types Michael Abbott (U.

Inductive Types for Free Representing Nested Inductive Types using W-types Michael Abbott (U. Leicester) Thorsten Altenkirch (U. Nottingham) Neil Ghani (U. Leicester) Inductive Types for Free p.1/22 Ideology Inductive Types for Free

1.07k views • 81 slides
Basic Types C s basic (built-in) types: o Integer types, including long integers, short

Basic Types C s basic (built-in) types: o Integer types, including long integers, short

2/4/14 Basic Types C s basic (built-in) types: o Integer types, including long integers, short integers, Basic Types and unsigned integers o Floating types ( float , double , and long double ) o char Based on slides from K. N. King

98 views • 7 slides
Value Types Objectives Discuss concept of value types efficiency memory management

Value Types Objectives Discuss concept of value types efficiency memory management

Value Types Objectives Discuss concept of value types efficiency memory management value semantics boxing unboxing simple types Introduce struct value type definition use advantages

615 views • 26 slides
DataCamp Data Types for Data Science DataCamp Data Types for Data Science Data types Data type

DataCamp Data Types for Data Science DataCamp Data Types for Data Science Data types Data type

DataCamp Data Types for Data Science DataCamp Data Types for Data Science Data types Data type system sets the stage for the capabilities of the language Understanding data types empowers you as a data scientist DataCamp Data Types for Data

526 views • 23 slides
CSE 341 Lecture 10 more about data types; nullable types; option Ullman 6.2 - 6.3; 4.2.5 -

CSE 341 Lecture 10 more about data types; nullable types; option Ullman 6.2 - 6.3; 4.2.5 -

CSE 341 Lecture 10 more about data types; nullable types; option Ullman 6.2 - 6.3; 4.2.5 - 4.2.6 slides created by Marty Stepp http://www.cs.washington.edu/341/ Creating new types of data datatype name = value | value | ... | value ; a

459 views • 19 slides
Model Estimation Within Planning and Learning Alborz Geramifard ICML W orkshop - June 2011

Model Estimation Within Planning and Learning Alborz Geramifard ICML W orkshop - June 2011

Model Estimation Within Planning and Learning Alborz Geramifard ICML W orkshop - June 2011 agf@mit.edu 1 Joint W ork Joshua Redding Joshua Joseph Jonathan How 2 +1 -1 -0.01 20% noise Conservative Aggressive Optimal 3 +1 -1

563 views • 27 slides
HUD Review Requirements 182 Contents PHA Submissions for HUD Approval Contract Actions

HUD Review Requirements 182 Contents PHA Submissions for HUD Approval Contract Actions

12 CHAPTER HUD Review Requirements 182 Contents PHA Submissions for HUD Approval Contract Actions Requiring Prior HUD Approval Exemption from Pre-award Review 183 Introduction HUD review of PHAs procurement activity is

302 views • 9 slides
Toward General Diagnosis of Static Errors Danfeng Zhang and Andrew C. Myers Cornell University

Toward General Diagnosis of Static Errors Danfeng Zhang and Andrew C. Myers Cornell University

Toward General Diagnosis of Static Errors Danfeng Zhang and Andrew C. Myers Cornell University POPL 2014 Static Program Analysis Many flavors Type system Dataflow analysis Information-flow analysis Useful properties Type

429 views • 28 slides
Congressional Request Advancing the Science of Climate Change Generic Report Summary Slides

Congressional Request Advancing the Science of Climate Change Generic Report Summary Slides

1 Congressional Request Advancing the Science of Climate Change Generic Report Summary Slides (based on presentations by Pam Matson and Ian Kraucunas to various audiences) http://americasclimatechoices.org 2 The National Academies A

333 views • 22 slides
All City Council Live! Info Academic Update Q&A Raise your hand to be called on! Use the

All City Council Live! Info Academic Update Q&A Raise your hand to be called on! Use the

All City Council Live! Info Academic Update Q&A Raise your hand to be called on! Use the chat box to type in comments or questions While we wait for people to join us, well watch these awesome handwashing dance videos... ACC Live!

320 views • 21 slides
Clacc 2019: An Update on OpenACC Support for Clang and LLVM Joel E. Denny, Seyong Lee, Jeffrey S.

Clacc 2019: An Update on OpenACC Support for Clang and LLVM Joel E. Denny, Seyong Lee, Jeffrey S.

Clacc 2019: An Update on OpenACC Support for Clang and LLVM Joel E. Denny, Seyong Lee, Jeffrey S. Vetter Future Technologies Group, ORNL https://ft.ornl.gov/ dennyje@ornl.gov April 8, 2019 EuroLLVM ORNL is managed by UT-Battelle, LLC for the

224 views • 18 slides
Polly-ACC: Transparent Compilation to Heterogeneous Hardware Torsten Hoefler (with Tobias Grosser)

Polly-ACC: Transparent Compilation to Heterogeneous Hardware Torsten Hoefler (with Tobias Grosser)

spcl.inf.ethz.ch @spcl_eth Polly-ACC: Transparent Compilation to Heterogeneous Hardware Torsten Hoefler (with Tobias Grosser) 1 spcl.inf.ethz.ch @spcl_eth Evading various ends the hardware view 2 spcl.inf.ethz.ch @spcl_eth

481 views • 26 slides
Automatic Generation of Efficient Accelerator Designs for Reconfigurable Hardware David

Automatic Generation of Efficient Accelerator Designs for Reconfigurable Hardware David

Automatic Generation of Efficient Accelerator Designs for Reconfigurable Hardware David Koeplinger Raghu Prabhakar Yaqi Zhang Christina Delimitrou Christos Kozyrakis Kunle Olukotun Stanford University ISCA 2016 FPGAs in Data Centers

299 views • 26 slides

More recommend