cut the ps bs - t&vs...why waste your life writing and debugging tests? design 32% verification:...

11
The Leader in Portable Stimulus Cut the PS BS Practical Test Synthesis For Your SoC & UVM Environments © Breker Verification Systems, Inc. All rights reserved. Breker Confidential Adnan Hamid, Breker President & CEO DVClub Europe, September 2018

Upload: others

Post on 27-Sep-2020

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Cut the PS BS - T&VS...Why Waste Your Life Writing and Debugging Tests? Design 32% Verification: Other 13% Verification: Test Development 30% Source: Wilson Research / Mentor 2016

The Leader in Portable Stimulus

Cut the PS BS Practical Test Synthesis For Your

SoC & UVM Environments

© Breker Verification Systems, Inc. All rights reserved. Breker Confidential

Adnan Hamid, Breker President & CEO DVClub Europe, September 2018

Page 2: Cut the PS BS - T&VS...Why Waste Your Life Writing and Debugging Tests? Design 32% Verification: Other 13% Verification: Test Development 30% Source: Wilson Research / Mentor 2016

A Brief History of Portable Stimulus

1990 1994 2008 2018 2015

© Breker Verification Systems, Inc. All rights reserved. Breker Confidential 2

2020

Undergraduate work on AI Machine Problem Solving

ST Display Controller • eRM Unit Bench • C SoC Bench

Accellera PSWG Started

Accellera PSS 1.0 release

Accellera PSS 2.0 release ?

AMD x86 Testing • Simulation • Quickturn • Custom Boards ( 1B cyc/day )

Page 3: Cut the PS BS - T&VS...Why Waste Your Life Writing and Debugging Tests? Design 32% Verification: Other 13% Verification: Test Development 30% Source: Wilson Research / Mentor 2016

Why Waste Your Life Writing and Debugging Tests?

Design 32%

Verification: Other 13%

Verification: Test Development

30%

Source: Wilson Research / Mentor 2016

Project Resource Deployment

Verification: Debug

25%

© Breker Verification Systems, Inc. All rights reserved. Breker Confidential 3

• Verification Planning • Scenario Decomposition • Abstracted Implementation

• Test Case Synthesis • Correct by composition • High Level Debug

Page 4: Cut the PS BS - T&VS...Why Waste Your Life Writing and Debugging Tests? Design 32% Verification: Other 13% Verification: Test Development 30% Source: Wilson Research / Mentor 2016

Realizing the Breker Verification Vision

© Breker Verification Systems, Inc. All rights reserved. . Breker Systems Confidential 4

SD

Sys

DC

PP

Cam

Scenario Model

TrekDesigner TrekApps

Cache Coherency Power Mngt

ARMv8

TrekUVM TrekSoC-Si TrekSoC

TrekGen

• Graph-based Simplified decomposition of complex scenarios, leveraging enhanced Portable Stimulus to generate effective test sets

• Portable Accelerated deployment of self-checking test cases into existing UVM & SoC environments

• Shareable Enhanced, revolutionary test space analysis, reuse and global cooperation, aided by Breker-lead Accellera standardization

PSS Test Synthesis

Optimization &

Deployment

Page 5: Cut the PS BS - T&VS...Why Waste Your Life Writing and Debugging Tests? Design 32% Verification: Other 13% Verification: Test Development 30% Source: Wilson Research / Mentor 2016

PSS Describes Graphs

© Breker Verification Systems, Inc. All rights reserved. . Breker Systems Confidential 5

component T { action A { output Buf out; }; action B { input Buf in; }; action C { input Buf in; }; pool Buf p; bind p *; action E { B b; C c; activity { schedule { select { b, c } select { b, c } } } } }

A B P

C A P

A B P

C A P

Page 6: Cut the PS BS - T&VS...Why Waste Your Life Writing and Debugging Tests? Design 32% Verification: Other 13% Verification: Test Development 30% Source: Wilson Research / Mentor 2016

UART Parity even

UART VIP mode_A

mode_B

in out

Parity odd

Parity on

Parity off

Simple Scenario Generation Example

© 2003 - 2017 Breker Verification Systems, Inc. All rights reserved. Breker Confidential 6

Simple Example: UART Parity Check Test Graph-based

Testbench

SoC RTL

CPU0 Memory

Fabric

Fabric

UART

UART VIP

UART Embedded in larger SoC

Graph segment: Data transfer and parity check

Tests pieced together by walking graph and solving random variables

class UartParityEven : public action { PSS_CTOR(UartParityEven, action); // Action_UartParityEven uart_block& blk { *sys.uarts.at(0) }; hsi::status_e status; void body() { blk.UART_LCR.PARITY_EVEN.set(1); } };

Each action has respective code segment

uart_block& blk { *sys.uarts.at(0) }; action UartParityEven { exec body { blk.UART_LCR.PARITY_EVEN.set(1); } };

C++

DSL Config DUT

Config VIP

Send Tx Data

Check VIP Data

Page 7: Cut the PS BS - T&VS...Why Waste Your Life Writing and Debugging Tests? Design 32% Verification: Other 13% Verification: Test Development 30% Source: Wilson Research / Mentor 2016

Test Case Optimization, Deployment & Analysis UVM, SoC, Final Silicon: Breker Targets Noted Engineering Pain Points

Testbench

DUT

Simplified, synchronized UVM

test-bench synthesis

UVM Agent

UVM Agent

TrekUVM

Synthesized, self-checking diagnostics for prototyping & silicon

TrekSoC-Si

test.c test.c test.c

Compiler

test.c test.c test.tbx Tr

ekB

ox

event id

Multi-threaded, simulation & emulation software-driven SoC test generation

OS or Bare Metal

Multi-threaded tests

TrekSoC

VIP VIP

Cache-Coherent Switching Fabric

CPU CPU CPU CPU

L1 L1 L1 L1 L2

CPU CPU CPU CPU

L1 L1 L1 L1 L2

GPU

Cache

L3 Cache

Memory Controller

Memory Controller I/O I/O Offload

© Breker Verification Systems, Inc. All rights reserved. Breker Confidential 7

Portable

TrekGen

Page 8: Cut the PS BS - T&VS...Why Waste Your Life Writing and Debugging Tests? Design 32% Verification: Other 13% Verification: Test Development 30% Source: Wilson Research / Mentor 2016

Path Analysis and Reachability

© Breker Verification Systems, Inc. All rights reserved. Breker Systems Confidential 8

Reason about which paths have been exercised

Size of overall search space

Node not reachable

Some paths to node are not reachable

Page 9: Cut the PS BS - T&VS...Why Waste Your Life Writing and Debugging Tests? Design 32% Verification: Other 13% Verification: Test Development 30% Source: Wilson Research / Mentor 2016

High-level Test Debug Driving Issue Resolution

© Breker Verification Systems, Inc. All rights reserved. Breker Confidential 9

Portable 4 processors, 1 thread per processor

Green nodes have finished

Yellow nodes are running

Blue nodes are waiting to run

Memory Usage by Region Data used/changed by a task

Software Driven Test Source

Test Execution Log

Failing Node

Tight integration to Verdi

Page 10: Cut the PS BS - T&VS...Why Waste Your Life Writing and Debugging Tests? Design 32% Verification: Other 13% Verification: Test Development 30% Source: Wilson Research / Mentor 2016

Various verification planners

Coverage Closure and Export

Scenario coverage reported on graph and in test list

Auto-coverage closure: push button test completion

Reporting: coverage detail output to third party verification planners

© Breker Verification Systems, Inc. All rights reserved. Breker Confidential 10

Portable

With Coverage Closure

Without Coverage Closure

Page 11: Cut the PS BS - T&VS...Why Waste Your Life Writing and Debugging Tests? Design 32% Verification: Other 13% Verification: Test Development 30% Source: Wilson Research / Mentor 2016

Practical Application of Portable Stimulus

PSS: powerful way to generate tests from abstract, randomized scenarios

Practical test deployment into testbenches requires additional capability

PSS plus optimizing solutions will revolutionize SoC and UVM verification

© Breker Verification Systems, Inc. All rights reserved. . Breker Systems Confidential 11