filecheck learning arithmetic
play

FileCheck: learning arithmetic Thomas Preud'homme Numeric - PowerPoint PPT Presentation

Dec 30, 2023 •117 likes •253 views

FileCheck: learning arithmetic Thomas Preud'homme Numeric constraints in toolchains Register constraints Memory layout e.g. consecutive 32-bit registers for i64 e.g. alignment of fields in struct/class v o i d f ( l o n g * p t r

  1. FileCheck: learning arithmetic Thomas Preud'homme

  2. Numeric constraints in toolchains Register constraints Memory layout e.g. consecutive 32-bit registers for i64 e.g. alignment of fields in struct/class v o i d f ( l o n g * p t r ) { s t r u c t f o o { ( … ) i n t a ; l o n g r e s = p t r [ 0 ] & p t r [ 1 ] l o n g b ; ( … ) } o b j 1 , o b j 2 ; } and rZ, rX, rY Addr X : obj1 and r(Z+1), r(X+1), r(Y+1) Addr X+16: obj2 2

  3. Numeric constraints in toolchains Register constraints Memory layout e.g. consecutive 32-bit registers for i64 e.g. alignment of fields in struct/class v o i d f ( l o n g a , l o n g b ) { s t r u c t f o o { r e t u r n a & b i n t a ; } l o n g b ; } o b j 1 __attribute__ ((aligned (256))) , o b j 2 ; C H E C K : 0 x [ [ A D D R : [ 0 - 9 A - F ] * ] ] 0 0 : o b j 1 C H E C K : a n d r 0 , r 0 , r 2 C H E C K : 0 x [ [ A D D R ] ] 1 0 : o b j 2 C H E C K : a n d r 1 , r 1 , r 3 Problem: subset of cases tested 3

  4. FileCheck: introducing arithmetic Syntax: [ [ # % f m t , V A R : < r e l o p > e x p r ] ] [ [ # % f m t , V A R + / - n u m ] ] C H E C K : a n d r [ [ # X : ] ] , r [ [ # Y : ] ] , r [ [ # Z : ] ] C H E C K : a n d r [ [ # X + 1 ] ] , r [ [ # Y + 1 ] ] , r [ [ # Z + 1 ] ] 4

  5. FileCheck: introducing arithmetic Syntax: [ [ # % f m t , V A R : < r e l o p > e x p r ] ] [ [ # % f m t , V A R + / - n u m ] ] C H E C K : a n d r [ [ # X : ] ] , r [ [ # Y : ] ] , r [ [ # Z : ] ] C H E C K : a n d r [ [ # X + 1 ] ] , r [ [ # Y + 1 ] ] , r [ [ # Z + 1 ] ] C H E C K : 0 x [ [ # % X , A D D R : ] ] o b j 1 C H E C K : 0 x [ [ # % X , A D D R + 1 6 ] ] o b j 2 5

  6. FileCheck: introducing arithmetic Syntax: [ [ # % f m t , V A R : VAR+/-num ] ] C H E C K : a n d r [ [ # X : ] ] , r [ [ # Y : ] ] , r [ [ # Z : ] ] C H E C K : a n d r [ [ # X + 1 ] ] , r [ [ # Y + 1 ] ] , r [ [ # Z + 1 ] ] C H E C K : 0 x [ [ # % X , A D D R : ] ] o b j 1 C H E C K : 0 x [ [ # % X , A D D R + 1 6 ] ] o b j 2 r e t u r n f o o b a ; C H E C K : f i l e . t x t : [ [ # F O O B A _ L I N E : @ L I N E - 1 ] ] : u n k n o w n v a r i a b l e ‘ f o o b a ’ C H E C K : f i l e . t x t : [ [ # F O O B A _ L I N E ] ] : d i d y o u m e a n ‘ f o o b a r ’ 6

  7. FileCheck: introducing arithmetic Syntax: [ [ # % f m t , V A R : == e x p r ] ] Expr operands: + - C H E C K : a n d r [ [ # X : ] ] , r [ [ # Y : ] ] , r [ [ # Z : ] ] C H E C K : a n d r [ [ # X + 1 ] ] , r [ [ # Y + 1 ] ] , r [ [ # Z + 1 ] ] C H E C K : 0 x [ [ # % X , A D D R : ] ] o b j 1 C H E C K : 0 x [ [ # % X , A D D R + F O O _ S I Z E ] ] o b j 2 r e t u r n f o o b a ; C H E C K : f i l e . t x t : [ [ # F O O B A _ L I N E : @ L I N E - 1 ] ] : u n k n o w n v a r i a b l e ‘ f o o b a ’ C H E C K : f i l e . t x t : [ [ # F O O B A _ L I N E ] ] : d i d y o u m e a n ‘ f o o b a r ’ 7

  8. FileCheck numeric expression: future work Syntax: [ [ # % f m t , V A R : == e x p r ] ] Expr operands: + - 8

  9. FileCheck numeric expression: future work Syntax: [ [ # % f m t , V A R : == e x p r ] ] Expr operands: + - * / ( ) ● Richer expressions C H E C K : a r r a y s i z e = [ [ # S I Z E : 8 * ( E L E M _ B I T S I Z E + G A P ) ] ] b y t e s 9

  10. FileCheck numeric expression: future work Syntax: [ [ # % f m t , V A R : < r e l o p > e x p r ] ] Expr operands: + - * / ( ) ● Richer expressions C H E C K : a r r a y s i z e = [ [ # S I Z E : 8 * ( E L E M _ B I T S I Z E + G A P ) ] ] b y t e s ● Inequalities C H E C K : s i z e = [ [ # S I Z E : < 4 2 ] ] b y t e s 10

  11. FileCheck numeric expression: future work Syntax: [ [ # % f m t , V A R : < r e l o p > e x p r ] ] Expr operands: + - * / ( ) ● Richer expressions C H E C K : a r r a y s i z e = [ [ # S I Z E : 8 * ( E L E M _ B I T S I Z E + G A P ) ] ] b y t e s ● Inequalities C H E C K : s i z e = [ [ # S I Z E : < 4 2 ] ] b y t e s ● Suggestions? Contribute to llvm-dev ML thread 11

  12. THANK YOU Thomas Preud'homme thomasp@graphcore.ai 12

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

Digital Design Discussion: Arithmetic Binary Arithmetic Floating-Point Arithmetic Binary

Digital Design Discussion: Arithmetic Binary Arithmetic Floating-Point Arithmetic Binary

Principles Of Digital Design Discussion: Arithmetic Binary Arithmetic Floating-Point Arithmetic Binary Arithmetic Same basic methodology as decimal arithmetic Important to know number representation Unsigned Signed

704 views • 11 slides
Unit 11 Adders &amp; Arithmetic Circuits 11.2 Learning Outcomes I understand what gates are

Unit 11 Adders &amp; Arithmetic Circuits 11.2 Learning Outcomes I understand what gates are

11.1 Unit 11 Adders &amp; Arithmetic Circuits 11.2 Learning Outcomes I understand what gates are used to design half and full adders I can build larger arithmetic circuits from smaller building blocks 11.3 ADDERS 11.4 Adder Intro

472 views • 32 slides
Unit 12 Adders &amp; Arithmetic Circuits 12.2 Learning Outcomes I understand what gates are

Unit 12 Adders &amp; Arithmetic Circuits 12.2 Learning Outcomes I understand what gates are

12.1 Unit 12 Adders &amp; Arithmetic Circuits 12.2 Learning Outcomes I understand what gates are used to design half and full adders I can build larger arithmetic circuits from smaller building blocks 12.3 ADDERS 12.4 Adder Intro

771 views • 40 slides
Arithmetic Series (Lesson Slides) UNIT #7: Sequences and Series WARMUP Arithmetic Series

Arithmetic Series (Lesson Slides) UNIT #7: Sequences and Series WARMUP Arithmetic Series

Arithmetic Series (Lesson Slides) UNIT #7: Sequences and Series WARMUP Arithmetic Series Determine the recursion formula and explicit formula for Learning Goal: the nth term for the arithmetic sequence if the first term is I will learn how to

352 views • 3 slides
Arithmetic for Computers October 31, 2008 Arithmetic for Computers ALU Arithmetic Logic Unit

Arithmetic for Computers October 31, 2008 Arithmetic for Computers ALU Arithmetic Logic Unit

Arithmetic for Computers October 31, 2008 Arithmetic for Computers ALU Arithmetic Logic Unit Performs most processor operations Control signal predicates addition subtraction logic operations shifting (w / or w / o sign extension) Figure:

205 views • 8 slides
Computer Arithmetic in Deep Learning Bryan Catanzaro @ctnzr What do we want AI to do? Keep us

Computer Arithmetic in Deep Learning Bryan Catanzaro @ctnzr What do we want AI to do? Keep us

Computer Arithmetic in Deep Learning Bryan Catanzaro @ctnzr What do we want AI to do? Keep us Guide us to organized content Help us find things Help us communicate Drive us to work Serve drinks? @ctnzr OCR-based

491 views • 46 slides
Arithmetic paper 4 Once you have completed the arithmetic paper, use the slides to help you to

Arithmetic paper 4 Once you have completed the arithmetic paper, use the slides to help you to

Arithmetic paper 4 Once you have completed the arithmetic paper, use the slides to help you to understand where you have made your mistakes. Are your scores improving each week? You will need to subtract 100 from the hundreds column. 800 -

461 views • 20 slides
Expressions, and Arithmetic Chapters 4 1 2 Arithmetic Operator Precedence Just as in

Expressions, and Arithmetic Chapters 4 1 2 Arithmetic Operator Precedence Just as in

9/9/2012 For Next Time Read Chapter 4 Expressions, and Arithmetic Chapters 4 1 2 Arithmetic Operator Precedence Just as in normal mathematics: Operator Operation * and / are applied before + and + Addition Parentheses

152 views • 3 slides
Arithmetic paper 5 Once you have completed the arithmetic paper, use the slides to help you to

Arithmetic paper 5 Once you have completed the arithmetic paper, use the slides to help you to

Arithmetic paper 5 Once you have completed the arithmetic paper, use the slides to help you to understand where you have made your mistakes. Try to improve your score each week! You will need to add 100 to the hundreds column. 400 + 100 = 500

361 views • 19 slides
Chapt hapter er 3 3 Arithmetic for Computers 3.1 Introduction Arithmetic for Computers

Chapt hapter er 3 3 Arithmetic for Computers 3.1 Introduction Arithmetic for Computers

COMPUTER ORGANIZATION AND DESIGN 5 th Edition The Hardware/Software Interface Chapt hapter er 3 3 Arithmetic for Computers 3.1 Introduction Arithmetic for Computers Operations on integers Addition and subtraction

718 views • 57 slides
Lecture 4 Arithmetic-Logic Unit 1 Arithmetic - Logic Unit ALU Handles integers Does the

Lecture 4 Arithmetic-Logic Unit 1 Arithmetic - Logic Unit ALU Handles integers Does the

ECE 0142 Computer Organization Lecture 4 Arithmetic-Logic Unit 1 Arithmetic - Logic Unit ALU Handles integers Does the calculations 2 Arithmetic-Logic Unit ALU Performs arithmetic add, subtract Performs logic and, or, invert,

769 views • 35 slides
Modular Arithmetic Modular Arithmetic: refresher. Notation x ( mod m ) or mod ( x , m ) -

Modular Arithmetic Modular Arithmetic: refresher. Notation x ( mod m ) or mod ( x , m ) -

Modular Arithmetic Modular Arithmetic: refresher. Notation x ( mod m ) or mod ( x , m ) - remainder of x divided by m in { 0 ,..., m 1 } . x is congruent to y modulo m or x y ( mod m ) if and only if ( x y ) is divisible by m .

397 views • 5 slides
Learning Outcomes I understand what gates are used to design half and full adders Unit 12

Learning Outcomes I understand what gates are used to design half and full adders Unit 12

12.1 12.2 Learning Outcomes I understand what gates are used to design half and full adders Unit 12 I can build larger arithmetic circuits from smaller building blocks Adders &amp; Arithmetic Circuits 12.3 12.4 Adder Intro

278 views • 10 slides
Computing just right: Application-specific arithmetic x x 2+ y 2+ z 2 x log x s i x i

Computing just right: Application-specific arithmetic x x 2+ y 2+ z 2 x log x s i x i

Computing just right: Application-specific arithmetic x x 2+ y 2+ z 2 x log x s i x i n e x n e x + i =0 Florent de Dinechin y Outline Anti-introduction: the arithmetic you want in a processor Operator parameterization

1.33k views • 103 slides
Predicate Logic: Peano Arithmetic Alice Gao Lecture 20 CS 245 Logic and Computation Fall 2019

Predicate Logic: Peano Arithmetic Alice Gao Lecture 20 CS 245 Logic and Computation Fall 2019

Predicate Logic: Peano Arithmetic Alice Gao Lecture 20 CS 245 Logic and Computation Fall 2019 1 / 22 Outline The Learning Goals Properties of Equality Using Logic to Model Number Theory Revisiting the Learning Goals CS 245 Logic and

357 views • 22 slides
By Shervin Daneshpajouh Computer Arithmetic Computer Arithmetic p Computer Computer Arithmetic

By Shervin Daneshpajouh Computer Arithmetic Computer Arithmetic p Computer Computer Arithmetic

By Shervin Daneshpajouh Computer Arithmetic Computer Arithmetic p Computer Computer Arithmetic Computer Computer Arithmetic Arithmetic Arithmetic CPU combines of ALU and Control Unit, CPU combines of ALU and Control Unit, The Arithmetic

356 views • 35 slides
Accuracy Characterization for Metropolitan-scale Wi-Fi Localization Presented by Md TamzeedIslam

Accuracy Characterization for Metropolitan-scale Wi-Fi Localization Presented by Md TamzeedIslam

9/28/17 Accuracy Characterization for Metropolitan-scale Wi-Fi Localization Presented by Md TamzeedIslam Localization Localization is an important problem. Robot Navigation. Indoor tracking. Satellite Navigation. Inventory

164 views • 14 slides
Information Retrieval Venkatesh Vinayakarao Term: Aug Dec, 2018 Indian Institute of

Information Retrieval Venkatesh Vinayakarao Term: Aug Dec, 2018 Indian Institute of

https://vvtesh.sarahah.com/ Information Retrieval Venkatesh Vinayakarao Term: Aug Dec, 2018 Indian Institute of Information Technology, Sri City Thou shalt not compute MRR nor ERR . Thou shalt not use MAP. Nobert Fuhr. Love Tarun

655 views • 43 slides
Designing for Performance Raul Queiroz Feitosa Objective In this section we examine the

Designing for Performance Raul Queiroz Feitosa Objective In this section we examine the

Designing for Performance Raul Queiroz Feitosa Objective In this section we examine the most common approach to assessing processor and computer system performance W. Stallings 2 Designing for Performance Outline Performance

356 views • 34 slides
Honors Combinatorics CMSC-27410 = Math-28410 CMSC-37200 Instructor: Laszlo Babai University

Honors Combinatorics CMSC-27410 = Math-28410 CMSC-37200 Instructor: Laszlo Babai University

Honors Combinatorics CMSC-27410 = Math-28410 CMSC-37200 Instructor: Laszlo Babai University of Chicago Week 7, Tuesday, May 19, 2020 CMSC-27410=Math-28410 CMSC-3720 Honors Combinatorics Sum of inverse squares 1 n 2 n = 1

1.32k views • 73 slides
Lecture 5: SOS Proofs and the Motzkin Polynomial Lecture Outline Part I: SOS proofs and

Lecture 5: SOS Proofs and the Motzkin Polynomial Lecture Outline Part I: SOS proofs and

Lecture 5: SOS Proofs and the Motzkin Polynomial Lecture Outline Part I: SOS proofs and examples Part II: Motzkin Polynomial Part I: SOS proofs and examples SOS proofs Fundamental question: What can we say about the

559 views • 37 slides
Welcome to TRANSFER PRICING Prof Dr Daniel N Erasmus, EA &amp; USTCP, International Tax Attorney

Welcome to TRANSFER PRICING Prof Dr Daniel N Erasmus, EA &amp; USTCP, International Tax Attorney

in association with: Welcome to TRANSFER PRICING Prof Dr Daniel N Erasmus, EA &amp; USTCP, International Tax Attorney TRM Daniel Erasmus Tax Court Practitioners Prof Dr Daniel N Erasmus, EA, USTCP &amp; Intl Tax Attorney +1 561 568 7115

283 views • 13 slides
Related Party Transactions ID Requirements 22 March 2018 EAD Regulation Development Team

Related Party Transactions ID Requirements 22 March 2018 EAD Regulation Development Team

Related Party Transactions ID Requirements 22 March 2018 EAD Regulation Development Team Outline What rules apply? Levels of disclosure required Disclosure requirements Impact of modified assurance opinion

1.02k views • 73 slides
Transfer Pricing in Luxembourg Managing the tax treatment of intra-group financing activities

Transfer Pricing in Luxembourg Managing the tax treatment of intra-group financing activities

www.pwc.com Transfer Pricing in Luxembourg Managing the tax treatment of intra-group financing activities Clarification of principally GroupCo Payable 70% IP activity LuxCo 30% Financing Receivable ForeignCo June 2011

261 views • 8 slides

More recommend