SlideShare a Scribd company logo
texasmulticore.com
Sample Customer Use Cases & Results
February 2017
Customer quote:
“SequenceL is Matlab on Steroids”
Customer Example: Predict Financial Market Using Neural Network
Users Learn SequenceL and Implement Significant AI Application in <2 weeks
 Challenge:
─ Predict financial market with retrospective data using
artificial neural network (AI) algorithms
─ Speed is key for big data
─ Goal: Limit time for analysis to 1 minute
 Solution
─ Implement neural network algorithms in SequenceL to
utilize full potential of user’s computer
 Results
─ Customer went from zero-to-proficient in <2 weeks,
with no outside assistance needed
─ Downloaded SequenceL, learned it via online tutorial,
and wrote significant application
─ Complex algorithms moved to SequenceL easily
─ Faster analysis of the data for end users
─ More effective use of hardware; uses all CPU cores
─ Scalability/portability on any hardware platform
2
Retrospective financial data from 08/18/2006 to 11/29/2016
© 2017 Texas Multicore Technologies, Inc.
All Rights Reserved
0
20
40
60
80
100
120
140
1 2 3 4
Time(sec)
Threads
Customer Example: Seismic RTM for Oil & Gas
Achieves >2x Speedup on Existing Platform with <50% Code + Portability
 Challenge:
─ ION Geophysical has one of the best and most optimized RTM codes in the oil & gas industry
─ ION RTM C/C++ and MPI was given to Acceleware to optimize with CUDA for GPUs and it ran slower
─ Goal: “Our customers don’t pay us to make the code run faster, they pay us for the geoscience. Our team
should be focused on the geoscience, not low-level SIMD instructions.”
─ Goal: Achieve platform portability + optimization to run our RTM on other platforms
 Solution
─ TMT refactored computationally-intensive
portions of existing C/C++ and MPI code
into SequenceL
 ~10,000 lines of 40,000 LOC program
 Results
─ >2x speedup on existing platforms
─ Platform dependencies removed
─ 5x speedup on POWER8 platform
─ <50% as many lines of code, much more
readable and supportable
3
© 2017 Texas Multicore Technologies, Inc.
All Rights Reserved
“Matlab on steroids”
Customer Example: CFD (Computational Fluid Dynamics)
SwRI Achieves 17.8x 26x Speedup with 25% Less Code
 Challenge:
─ Researchers already often have to wait weeks for these fluid flow simulation runs
─ This slows the pace of innovation and makes it impractical to further enhance models for higher
accuracy (e.g.- irregular shaped particles)
─ Goal: reduce the runtime of in-house Lattice Boltzmann method-based CFD simulation by a factor of 10
 Solution
─ SwRI reformulated this existing Fortran+OpenMP code into SequenceL
 Results
─ “SequenceL implementation was 26x faster for most relevant benchmark“ (70 particles)
─ “Runs that previously took 2 weeks now completed overnight”
─ “The SequenceL compiler created a parallel executable with no burden on the programmer and is
provably race-free”
─ “SequenceL version was 25% shorter (LOC) and closely resembles the mathematical equations”
4
© 2017 Texas Multicore Technologies, Inc.
All Rights Reserved
2 weeks
Monday
Tuesday
Wed
Thurs
Friday
Sat
Sunday
Monday
Tuesday
Wed
Thurs
Friday
Sat
Sunday
Dramatically reduced time to discovery; now practical to enhance model12 hours!
Customer Example: Industrial Control Networking
(WirelessHART, IEC 62591, IEEE 802.15.4)
 Challenge
─ WirelessHART was too complex and slow to process at scale
 Generating Downlink graphs for 200 nodes would take 5 hours and 20 minutes
 Generating Uplink, Downlink, and Broadcast network graphs for a plant
or oil field with a 1,000 nodes would take a solid month to process
─ Goal: Generate graphs for a 200 node network in <1 minute
 Solution
─ Emerson Research proposed a new Downlink algorithm (SRDR) in a whitepaper
─ Had TMT implement in SequenceL to achieve multicore acceleration
 Results
─ SequenceL finished by TMT in 3 weeks, ran 10x faster
 10X faster performance and right the first time
 Generated graphs in 9.3 seconds
─ Inventors took 5 months in Java, never got performance
 Initially had errors and 10x slower
 Used SequenceL to debug Java, get perf. insights (still 3x slower)
5
Fast & robust code, faster time to market
© 2017 Texas Multicore Technologies, Inc.
All Rights Reserved
Customer Example: Video Processing Using SequenceL
 Proprietary algorithms remove air turbulence, radiated heat, etc.
 Goal: 30Hz (fps) to keep up with real time input video feed
 Best performance (8 core x86 platform)
─ 58 Hz: SequenceL
─ 21 Hz: Matlab (Interpreter)
─ 1.2 Hz: Matlab (Coder/C-out)
Input video feed Processed video output
© 2017 Texas Multicore Technologies, Inc.
All Rights Reserved6
Customer Example: Video Processing Using SequenceL
 Proprietary algorithms remove air turbulence, radiated heat, etc.
 Goal: 30Hz (fps) to keep up with real time input video feed
 Best performance (8 core x86 platform)
─ 58 Hz: SequenceL
─ 21 Hz: Matlab (Interpreter)
─ 1.2 Hz: Matlab (Coder/C-out)
Input video feed Processed video output
© 2017 Texas Multicore Technologies, Inc.
All Rights Reserved7
Customer Example: Video Processing Using SequenceL
 Proprietary algorithms remove air turbulence, radiated heat, etc.
 Goal: 30Hz (fps) to keep up with real time input video feed
 Best performance (8 core x86 platform)
─ 58 Hz: SequenceL
─ 21 Hz: Matlab (Interpreter)
─ 1.2 Hz: Matlab (Coder/C-out)
Input video feed Processed video output
© 2017 Texas Multicore Technologies, Inc.
All Rights Reserved8
Customer Example: Video Processing Using SequenceL
 Proprietary algorithms remove air turbulence, radiated heat, etc.
 Goal: 30Hz (fps) to keep up with real time input video feed
 Best performance (8 core x86 platform)
─ 58 Hz: SequenceL
─ 21 Hz: Matlab (Interpreter)
─ 1.2 Hz: Matlab (Coder/C-out)
Input video feed Processed video output
© 2017 Texas Multicore Technologies, Inc.
All Rights Reserved9
SequenceL also won on code readability
SequenceL Changes the Game
© 2017 Texas Multicore Technologies, Inc.
All Rights Reserved10
Faster Performance
Uses all cores, GPUs
10X Faster Time to Market
Prototyping tool = Production tool
Get it Right the First Time
Far fewer lines of code; Algorithms in
SequenceL often match their definition
Portability + Optimization
True application portability with
high performance
Built on Open Industry Standards
Integrates with existing tools,
languages, & methodologies
Fast ROI
No need for expensive “parallel ninjas”
nor the time they add to the schedule
Accelerates time to innovation/discovery
SequenceL Changes the Game
© 2017 Texas Multicore Technologies, Inc.
All Rights Reserved11
Faster Performance
Uses all cores, GPUs
10X Faster Time to Market
Prototyping tool = Production tool
Get it Right the First Time
Far fewer lines of code; Algorithms in
SequenceL often match their definition
Portability + Optimization
True application portability with
high performance
Built on Open Industry Standards
Integrates with existing tools,
languages, & methodologies
Fast ROI
No need for expensive “parallel ninjas”
nor the time they add to the schedule
Business Agility: TTM, new platforms, Cloud

More Related Content

PDF
Trends in Mixed Signal Validation
DVClub
 
PPTX
Værktøjer udviklet på AAU til analyse af SCJ programmer
InfinIT - Innovationsnetværket for it
 
PDF
Validation and-design-in-a-small-team-environment
Obsidian Software
 
PDF
Efficient execution of quantized deep learning models a compiler approach
jemin lee
 
PPT
Generic and Automatic Specman Based Verification Environment
DVClub
 
PPTX
Jeda Hls Hlv Success Story V4
Chun Xia
 
PDF
Strickland dvclub
Obsidian Software
 
Trends in Mixed Signal Validation
DVClub
 
Værktøjer udviklet på AAU til analyse af SCJ programmer
InfinIT - Innovationsnetværket for it
 
Validation and-design-in-a-small-team-environment
Obsidian Software
 
Efficient execution of quantized deep learning models a compiler approach
jemin lee
 
Generic and Automatic Specman Based Verification Environment
DVClub
 
Jeda Hls Hlv Success Story V4
Chun Xia
 
Strickland dvclub
Obsidian Software
 

What's hot (20)

PDF
Sharam salamian
Obsidian Software
 
PPTX
Mid term
Kan-Han (John) Lu
 
PDF
Qualifying a high performance memory subsysten for Functional Safety
Pankaj Singh
 
PDF
Stinson post si and verification
Obsidian Software
 
PDF
Michael_Kogan_portfolio
Michael Kogan
 
PDF
OVERCOMING KEY CHALLENGES OF TODAY'S COMPLEX SOC: PERFORMANCE OPTIMIZATION AN...
Pankaj Singh
 
PDF
How fpgas work when they don't
InfinIT - Innovationsnetværket for it
 
PPTX
SDN Controller - Programming Challenges
snrism
 
PDF
Chris brown ti
Obsidian Software
 
PPT
Coverage Solutions on Emulators
DVClub
 
PDF
MLPerf an industry standard benchmark suite for machine learning performance
jemin lee
 
PPT
Validating Next Generation CPUs
DVClub
 
PDF
Alley vsu functional_coverage_1f
Obsidian Software
 
PDF
Embedded Development for the Future - Platforms for Rohde-Schwarz Mobile Tester
InfinIT - Innovationsnetværket for it
 
PDF
ES-HAS: An Edge- and SDN-Assisted Framework for HTTP Adaptive Video Streaming
Alpen-Adria-Universität
 
DOC
Manoj_Resume
Manoj Verma
 
PPTX
Poc exadata 2018
Jacques Kostic
 
PDF
Hardware monitoring with collectd at CERN
Luca Gardi
 
DOCX
Resume_Anurag_Design_Verification_2+_yrs
ANURAG KAVADANA
 
PPT
Lecture#5
Adil Alpkoçak
 
Sharam salamian
Obsidian Software
 
Qualifying a high performance memory subsysten for Functional Safety
Pankaj Singh
 
Stinson post si and verification
Obsidian Software
 
Michael_Kogan_portfolio
Michael Kogan
 
OVERCOMING KEY CHALLENGES OF TODAY'S COMPLEX SOC: PERFORMANCE OPTIMIZATION AN...
Pankaj Singh
 
How fpgas work when they don't
InfinIT - Innovationsnetværket for it
 
SDN Controller - Programming Challenges
snrism
 
Chris brown ti
Obsidian Software
 
Coverage Solutions on Emulators
DVClub
 
MLPerf an industry standard benchmark suite for machine learning performance
jemin lee
 
Validating Next Generation CPUs
DVClub
 
Alley vsu functional_coverage_1f
Obsidian Software
 
Embedded Development for the Future - Platforms for Rohde-Schwarz Mobile Tester
InfinIT - Innovationsnetværket for it
 
ES-HAS: An Edge- and SDN-Assisted Framework for HTTP Adaptive Video Streaming
Alpen-Adria-Universität
 
Manoj_Resume
Manoj Verma
 
Poc exadata 2018
Jacques Kostic
 
Hardware monitoring with collectd at CERN
Luca Gardi
 
Resume_Anurag_Design_Verification_2+_yrs
ANURAG KAVADANA
 
Lecture#5
Adil Alpkoçak
 
Ad

Viewers also liked (20)

PPT
Search Engine Optimization Tutorial
Tahasin Chowdhury
 
PPTX
Mgiti Diploma courses
MGITI NIOS
 
PPTX
Unit1 Will Eze
Mr Crafts
 
PPTX
ICSR Narrative Writing_Katalyst HLS
Katalyst HLS
 
PDF
Lucy 2013 in Wine Spectator
Pisoni Vineyards & Winery
 
PPTX
Seeing Graphs - How to successfully visualize connected data
Cambridge Intelligence
 
PDF
Succes story - Increased overall sales by 40% for a UK based speciality Retai...
Expedux Technologies
 
PDF
Adams At Home Auction - February 26 th 2017
Adam's Fine Art Auctioneers
 
PPTX
Curso de orígenes
rigoberto arturo zapata sánchez
 
PPS
CORONEL DIEGO CAMACHO - 23-F: UN GOLPE DE ESTADO EN BUSCA DE AUTOR (PARTE 2)
Proyecto Matriz
 
PDF
1F作業
Kazuhide Fukada
 
PDF
Strtio Spark Streaming + Siddhi CEP Engine
Myung Ho Yun
 
PDF
Historical and Impact Analysis of API Breaking Changes: A Large-Scale Study
Laerte Xavier
 
PDF
Today dietitian
Randy Grover
 
PPTX
Bill Gates lessons
Vinod Mehra
 
PPTX
Emprendimiento social, Empresas con impacto social.
Jose Redondo Sanchez
 
PPTX
Evangelio y corán
Jesús Zamora
 
PDF
La retribución de la innovación
© Gian-Lluis Ribechini
 
PPTX
Toastmaster Contest Rule Changes 2016
Amy Hamilton, PMP, ITILv3, CISM
 
PDF
Sesión #eadaProjectManagement (23-2-2017) Microsoft Planner - Herr Cloud apoy...
eada business school barcelona
 
Search Engine Optimization Tutorial
Tahasin Chowdhury
 
Mgiti Diploma courses
MGITI NIOS
 
Unit1 Will Eze
Mr Crafts
 
ICSR Narrative Writing_Katalyst HLS
Katalyst HLS
 
Lucy 2013 in Wine Spectator
Pisoni Vineyards & Winery
 
Seeing Graphs - How to successfully visualize connected data
Cambridge Intelligence
 
Succes story - Increased overall sales by 40% for a UK based speciality Retai...
Expedux Technologies
 
Adams At Home Auction - February 26 th 2017
Adam's Fine Art Auctioneers
 
CORONEL DIEGO CAMACHO - 23-F: UN GOLPE DE ESTADO EN BUSCA DE AUTOR (PARTE 2)
Proyecto Matriz
 
1F作業
Kazuhide Fukada
 
Strtio Spark Streaming + Siddhi CEP Engine
Myung Ho Yun
 
Historical and Impact Analysis of API Breaking Changes: A Large-Scale Study
Laerte Xavier
 
Today dietitian
Randy Grover
 
Bill Gates lessons
Vinod Mehra
 
Emprendimiento social, Empresas con impacto social.
Jose Redondo Sanchez
 
Evangelio y corán
Jesús Zamora
 
La retribución de la innovación
© Gian-Lluis Ribechini
 
Toastmaster Contest Rule Changes 2016
Amy Hamilton, PMP, ITILv3, CISM
 
Sesión #eadaProjectManagement (23-2-2017) Microsoft Planner - Herr Cloud apoy...
eada business school barcelona
 
Ad

Similar to TMT SequenceL customer use cases and results (20)

PPTX
SequenceL Auto-Parallelizing Toolset Intro slideshare
Doug Norton
 
PPTX
SequenceL intro slideshare
Doug Norton
 
PDF
“Enabling Ultra-low Power Edge Inference and On-device Learning with Akida,” ...
Edge AI and Vision Alliance
 
PPT
Multi Processor Architecture for image processing
ideas2ignite
 
PDF
High Performance & High Throughput Computing - EUDAT Summer School (Giuseppe ...
EUDAT
 
PPTX
Building a Video Streaming Testing Framework with Seastar
ScyllaDB
 
PDF
project seminor
Ramesh Naik Banothu
 
PDF
Trends in Systems and How to Get Efficient Performance
inside-BigData.com
 
PPTX
TechEvent Exdata X7-2 POC with OVM
Trivadis
 
PDF
Application Optimisation using OpenPOWER and Power 9 systems
Ganesan Narayanasamy
 
PPTX
From Hours to Minutes: The Journey of Optimizing Mask-RCNN and BERT Using MXNet
Eric Haibin Lin
 
PDF
Advanced Computer Architecture - Lec1.pdf
alialoa2015
 
PPTX
Developing Real-Time Systems on Application Processors
Toradex
 
PDF
Santhosh Resume
Santhosh Ravisankar
 
PDF
IMAGE CAPTURE, PROCESSING AND TRANSFER VIA ETHERNET UNDER CONTROL OF MATLAB G...
Christopher Diamantopoulos
 
PDF
“Key Requirements to Successfully Implement Generative AI in Edge Devices—Opt...
Edge AI and Vision Alliance
 
PDF
“eXtending” the Automation Toolbox: Introduction to TwinCAT 3 Software and eX...
Design World
 
PPT
Computing Without Computers - Oct08
Ian Page
 
PDF
"New Dataflow Architecture for Machine Learning," a Presentation from Wave Co...
Edge AI and Vision Alliance
 
PDF
Simulation Versus Acceleration, Versus Emulation
DVClub
 
SequenceL Auto-Parallelizing Toolset Intro slideshare
Doug Norton
 
SequenceL intro slideshare
Doug Norton
 
“Enabling Ultra-low Power Edge Inference and On-device Learning with Akida,” ...
Edge AI and Vision Alliance
 
Multi Processor Architecture for image processing
ideas2ignite
 
High Performance & High Throughput Computing - EUDAT Summer School (Giuseppe ...
EUDAT
 
Building a Video Streaming Testing Framework with Seastar
ScyllaDB
 
project seminor
Ramesh Naik Banothu
 
Trends in Systems and How to Get Efficient Performance
inside-BigData.com
 
TechEvent Exdata X7-2 POC with OVM
Trivadis
 
Application Optimisation using OpenPOWER and Power 9 systems
Ganesan Narayanasamy
 
From Hours to Minutes: The Journey of Optimizing Mask-RCNN and BERT Using MXNet
Eric Haibin Lin
 
Advanced Computer Architecture - Lec1.pdf
alialoa2015
 
Developing Real-Time Systems on Application Processors
Toradex
 
Santhosh Resume
Santhosh Ravisankar
 
IMAGE CAPTURE, PROCESSING AND TRANSFER VIA ETHERNET UNDER CONTROL OF MATLAB G...
Christopher Diamantopoulos
 
“Key Requirements to Successfully Implement Generative AI in Edge Devices—Opt...
Edge AI and Vision Alliance
 
“eXtending” the Automation Toolbox: Introduction to TwinCAT 3 Software and eX...
Design World
 
Computing Without Computers - Oct08
Ian Page
 
"New Dataflow Architecture for Machine Learning," a Presentation from Wave Co...
Edge AI and Vision Alliance
 
Simulation Versus Acceleration, Versus Emulation
DVClub
 

Recently uploaded (20)

PDF
Key Features to Look for in Arizona App Development Services
Net-Craft.com
 
DOCX
Can You Build Dashboards Using Open Source Visualization Tool.docx
Varsha Nayak
 
PPTX
Presentation about variables and constant.pptx
safalsingh810
 
PDF
Bandai Playdia The Book - David Glotz
BluePanther6
 
PDF
Become an Agentblazer Champion Challenge Kickoff
Dele Amefo
 
PPTX
AI-Ready Handoff: Auto-Summaries & Draft Emails from MQL to Slack in One Flow
bbedford2
 
PDF
Teaching Reproducibility and Embracing Variability: From Floating-Point Exper...
University of Rennes, INSA Rennes, Inria/IRISA, CNRS
 
PDF
Protecting the Digital World Cyber Securit
dnthakkar16
 
PPTX
Visualising Data with Scatterplots in IBM SPSS Statistics.pptx
Version 1 Analytics
 
PDF
Become an Agentblazer Champion Challenge
Dele Amefo
 
PPTX
Web Testing.pptx528278vshbuqffqhhqiwnwuq
studylike474
 
PPTX
Smart Panchayat Raj e-Governance App.pptx
Rohitnikam33
 
PDF
49784907924775488180_LRN2959_Data_Pump_23ai.pdf
Abilash868456
 
PPTX
TestNG for Java Testing and Automation testing
ssuser0213cb
 
PPTX
oapresentation.pptx
mehatdhavalrajubhai
 
PDF
The Role of Automation and AI in EHS Management for Data Centers.pdf
TECH EHS Solution
 
PDF
49785682629390197565_LRN3014_Migrating_the_Beast.pdf
Abilash868456
 
PPT
Why Reliable Server Maintenance Service in New York is Crucial for Your Business
Sam Vohra
 
PDF
Why Use Open Source Reporting Tools for Business Intelligence.pdf
Varsha Nayak
 
PDF
ShowUs: Pharo Stream Deck (ESUG 2025, Gdansk)
ESUG
 
Key Features to Look for in Arizona App Development Services
Net-Craft.com
 
Can You Build Dashboards Using Open Source Visualization Tool.docx
Varsha Nayak
 
Presentation about variables and constant.pptx
safalsingh810
 
Bandai Playdia The Book - David Glotz
BluePanther6
 
Become an Agentblazer Champion Challenge Kickoff
Dele Amefo
 
AI-Ready Handoff: Auto-Summaries & Draft Emails from MQL to Slack in One Flow
bbedford2
 
Teaching Reproducibility and Embracing Variability: From Floating-Point Exper...
University of Rennes, INSA Rennes, Inria/IRISA, CNRS
 
Protecting the Digital World Cyber Securit
dnthakkar16
 
Visualising Data with Scatterplots in IBM SPSS Statistics.pptx
Version 1 Analytics
 
Become an Agentblazer Champion Challenge
Dele Amefo
 
Web Testing.pptx528278vshbuqffqhhqiwnwuq
studylike474
 
Smart Panchayat Raj e-Governance App.pptx
Rohitnikam33
 
49784907924775488180_LRN2959_Data_Pump_23ai.pdf
Abilash868456
 
TestNG for Java Testing and Automation testing
ssuser0213cb
 
oapresentation.pptx
mehatdhavalrajubhai
 
The Role of Automation and AI in EHS Management for Data Centers.pdf
TECH EHS Solution
 
49785682629390197565_LRN3014_Migrating_the_Beast.pdf
Abilash868456
 
Why Reliable Server Maintenance Service in New York is Crucial for Your Business
Sam Vohra
 
Why Use Open Source Reporting Tools for Business Intelligence.pdf
Varsha Nayak
 
ShowUs: Pharo Stream Deck (ESUG 2025, Gdansk)
ESUG
 

TMT SequenceL customer use cases and results

  • 1. texasmulticore.com Sample Customer Use Cases & Results February 2017 Customer quote: “SequenceL is Matlab on Steroids”
  • 2. Customer Example: Predict Financial Market Using Neural Network Users Learn SequenceL and Implement Significant AI Application in <2 weeks  Challenge: ─ Predict financial market with retrospective data using artificial neural network (AI) algorithms ─ Speed is key for big data ─ Goal: Limit time for analysis to 1 minute  Solution ─ Implement neural network algorithms in SequenceL to utilize full potential of user’s computer  Results ─ Customer went from zero-to-proficient in <2 weeks, with no outside assistance needed ─ Downloaded SequenceL, learned it via online tutorial, and wrote significant application ─ Complex algorithms moved to SequenceL easily ─ Faster analysis of the data for end users ─ More effective use of hardware; uses all CPU cores ─ Scalability/portability on any hardware platform 2 Retrospective financial data from 08/18/2006 to 11/29/2016 © 2017 Texas Multicore Technologies, Inc. All Rights Reserved 0 20 40 60 80 100 120 140 1 2 3 4 Time(sec) Threads
  • 3. Customer Example: Seismic RTM for Oil & Gas Achieves >2x Speedup on Existing Platform with <50% Code + Portability  Challenge: ─ ION Geophysical has one of the best and most optimized RTM codes in the oil & gas industry ─ ION RTM C/C++ and MPI was given to Acceleware to optimize with CUDA for GPUs and it ran slower ─ Goal: “Our customers don’t pay us to make the code run faster, they pay us for the geoscience. Our team should be focused on the geoscience, not low-level SIMD instructions.” ─ Goal: Achieve platform portability + optimization to run our RTM on other platforms  Solution ─ TMT refactored computationally-intensive portions of existing C/C++ and MPI code into SequenceL  ~10,000 lines of 40,000 LOC program  Results ─ >2x speedup on existing platforms ─ Platform dependencies removed ─ 5x speedup on POWER8 platform ─ <50% as many lines of code, much more readable and supportable 3 © 2017 Texas Multicore Technologies, Inc. All Rights Reserved “Matlab on steroids”
  • 4. Customer Example: CFD (Computational Fluid Dynamics) SwRI Achieves 17.8x 26x Speedup with 25% Less Code  Challenge: ─ Researchers already often have to wait weeks for these fluid flow simulation runs ─ This slows the pace of innovation and makes it impractical to further enhance models for higher accuracy (e.g.- irregular shaped particles) ─ Goal: reduce the runtime of in-house Lattice Boltzmann method-based CFD simulation by a factor of 10  Solution ─ SwRI reformulated this existing Fortran+OpenMP code into SequenceL  Results ─ “SequenceL implementation was 26x faster for most relevant benchmark“ (70 particles) ─ “Runs that previously took 2 weeks now completed overnight” ─ “The SequenceL compiler created a parallel executable with no burden on the programmer and is provably race-free” ─ “SequenceL version was 25% shorter (LOC) and closely resembles the mathematical equations” 4 © 2017 Texas Multicore Technologies, Inc. All Rights Reserved 2 weeks Monday Tuesday Wed Thurs Friday Sat Sunday Monday Tuesday Wed Thurs Friday Sat Sunday Dramatically reduced time to discovery; now practical to enhance model12 hours!
  • 5. Customer Example: Industrial Control Networking (WirelessHART, IEC 62591, IEEE 802.15.4)  Challenge ─ WirelessHART was too complex and slow to process at scale  Generating Downlink graphs for 200 nodes would take 5 hours and 20 minutes  Generating Uplink, Downlink, and Broadcast network graphs for a plant or oil field with a 1,000 nodes would take a solid month to process ─ Goal: Generate graphs for a 200 node network in <1 minute  Solution ─ Emerson Research proposed a new Downlink algorithm (SRDR) in a whitepaper ─ Had TMT implement in SequenceL to achieve multicore acceleration  Results ─ SequenceL finished by TMT in 3 weeks, ran 10x faster  10X faster performance and right the first time  Generated graphs in 9.3 seconds ─ Inventors took 5 months in Java, never got performance  Initially had errors and 10x slower  Used SequenceL to debug Java, get perf. insights (still 3x slower) 5 Fast & robust code, faster time to market © 2017 Texas Multicore Technologies, Inc. All Rights Reserved
  • 6. Customer Example: Video Processing Using SequenceL  Proprietary algorithms remove air turbulence, radiated heat, etc.  Goal: 30Hz (fps) to keep up with real time input video feed  Best performance (8 core x86 platform) ─ 58 Hz: SequenceL ─ 21 Hz: Matlab (Interpreter) ─ 1.2 Hz: Matlab (Coder/C-out) Input video feed Processed video output © 2017 Texas Multicore Technologies, Inc. All Rights Reserved6
  • 7. Customer Example: Video Processing Using SequenceL  Proprietary algorithms remove air turbulence, radiated heat, etc.  Goal: 30Hz (fps) to keep up with real time input video feed  Best performance (8 core x86 platform) ─ 58 Hz: SequenceL ─ 21 Hz: Matlab (Interpreter) ─ 1.2 Hz: Matlab (Coder/C-out) Input video feed Processed video output © 2017 Texas Multicore Technologies, Inc. All Rights Reserved7
  • 8. Customer Example: Video Processing Using SequenceL  Proprietary algorithms remove air turbulence, radiated heat, etc.  Goal: 30Hz (fps) to keep up with real time input video feed  Best performance (8 core x86 platform) ─ 58 Hz: SequenceL ─ 21 Hz: Matlab (Interpreter) ─ 1.2 Hz: Matlab (Coder/C-out) Input video feed Processed video output © 2017 Texas Multicore Technologies, Inc. All Rights Reserved8
  • 9. Customer Example: Video Processing Using SequenceL  Proprietary algorithms remove air turbulence, radiated heat, etc.  Goal: 30Hz (fps) to keep up with real time input video feed  Best performance (8 core x86 platform) ─ 58 Hz: SequenceL ─ 21 Hz: Matlab (Interpreter) ─ 1.2 Hz: Matlab (Coder/C-out) Input video feed Processed video output © 2017 Texas Multicore Technologies, Inc. All Rights Reserved9 SequenceL also won on code readability
  • 10. SequenceL Changes the Game © 2017 Texas Multicore Technologies, Inc. All Rights Reserved10 Faster Performance Uses all cores, GPUs 10X Faster Time to Market Prototyping tool = Production tool Get it Right the First Time Far fewer lines of code; Algorithms in SequenceL often match their definition Portability + Optimization True application portability with high performance Built on Open Industry Standards Integrates with existing tools, languages, & methodologies Fast ROI No need for expensive “parallel ninjas” nor the time they add to the schedule Accelerates time to innovation/discovery
  • 11. SequenceL Changes the Game © 2017 Texas Multicore Technologies, Inc. All Rights Reserved11 Faster Performance Uses all cores, GPUs 10X Faster Time to Market Prototyping tool = Production tool Get it Right the First Time Far fewer lines of code; Algorithms in SequenceL often match their definition Portability + Optimization True application portability with high performance Built on Open Industry Standards Integrates with existing tools, languages, & methodologies Fast ROI No need for expensive “parallel ninjas” nor the time they add to the schedule Business Agility: TTM, new platforms, Cloud