SlideShare uma empresa Scribd logo
1 de 17
Baixar para ler offline
Trustworthy Forever

                                                            From Deep Space To Deep Sea




      History goes back to 1989 (Eonic Systems)
         VIRTUOSO parallel RTOS (T800, C40, C6x, 2016x, TS102,
           G4, …) – CSP based
         Used from 1 CPU to 1600 DSPs (sonar, radar) to
         12000 nodes (heterogeneous)
         Acquired by Wind River Systems in 2001
      Altreonic: created as spin-off in 2008 following R&D
         Unified systems engineering (GoedelWorks)
         Formalised when possible
         Network-centric OpenComRTOS:
               Used as test case for use of formal techniques
               Binary/source and Open Technology License model
20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -              2
Computation to communication ratio:
           Depends on application
               Input rate = output rate = ½ computation rate ideal
               Ideal ratio = 1, typically minimum 5 to 10
      Data communication is bottleneck:
           Set-up latency, no polling
           Task to task/memory to memory bandwidth is real target
           Concurrency to mask communication latencies
           DMA => requires additional busses
      Data protection:
           Pointers are fast but very dangerous
           memcpy has no distributed semantics! (should be _W)
      Heterogeneous many-cores: data-types
20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -               3




20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -               4
Novel programming model, but long formal
      history (Hoare’s CSP, 1975)
      Use of TLA+/TLC for design and verification
      Unexpected result:
           10X smaller code size;
           Easy to Port to new
           Platforms;
           Low amount of assembly;




20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -   5




      Network-centric (RT)OS, MP by default:
           Concurrency at the core (“Interacting Entities”);
           Pragmatic superset of CSP (Hoare);
           Scalable yet very small: typically 2 to 5 kiB/node;
           Real-time communication as system level service;
           Unique support for distributed priority inheritance;
           Heterogeneous target /communication support;
           Integrate seamlessly “legacy OS” nodes;
           Virtual Single Processor model;
           Visual modeling/ programming with code generators;
           Capable of fault-tolerance and resource management.
20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -   6
Basis:
           RMA => priority based, preemptive
           Additional: timer based
           Everything is priority ordered
               e.g. waiting lists
               Packet based communication
      Priority Inheritance support with ceiling level
           Single processor
           Unique distributed implementation



20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -                     7




 The Generic Hub as metamodel




             Similar to a Guarded Atomic Action, or a pragmatic superset of CSP
20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -                     8
20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -                        9




   Hub Entity         Semantics
   Event              Synchronisation on boolean event, N to N
   Semaphore          Synchronisation with counter for async signalling, N to N
   Port               Synchronisation with exchange of Packet, N to N
   FIFO queue         Buffered, async communication, except on FIFO full or empty, N to N
   Resource           Logical resource to guard critical section (with priority inheritance)
   Memory pool        Linked list of memory blocks


   Synchronisation    Semantics
   _NW                Non waiting => returns immediately
   _W                 Waiting until synchronisation (blocking)
   _WT                Waiting with a TimeOut.




20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -                        10
OpenComRTOS Designer, for Modeling:
           Topology Diagram, defines the Hardware Setup
           Application Diagram, defines the Entities and their
           Interactions
           Source Code, defines the sequential parts.
      Code Generators: Generate the configuration
      for OpenComRTOS from the Models
      OpenComRTOS: Runtime Layer providing the
      Interacting Entities.


20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -   11




20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -   12
20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -   13




20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -   14
• Up to 10x smaller than traditional design (thanks to formal
  development)
  • Less power, less memory, easier to verify, scalable ...
             CPU Type                                     Codesize

             ARM-Cortex-M3                               2.5 – 4.0 kB
             XMOS-XS1                                    5.0 – 7.5 kB
             PowerPC e600                                7.1 – 9.8 kB
             TI-C6678x (8 core DSP)                      5.1 – 7.7 kB
             Intel-SCC (48 Pentium cores)                4.3 – 5 kB

     Code size figures (in Bytes) obtained for our different ports, -Os

      Dormant ports: MLX16 (2K), Xilinx MB (5K), Leon3 (5K), CoolFlux DSP (2K)
20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -                  15




          Comparing Intel 48-core SCC and
             TI 8-core TMS320C6678




20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -                  16
• L1 cache: 16 KB
              • L2 cache: 256 KB
              • RTOS kernel on each core
              • Communication using memory of MPB (16 KB)

20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -              17




                                                   • “RoC“ (Rack On a chip)
                                                   • Hierarchical Interrupt
                                                   Routing up to 1000 possible
                                                   interrupts per core!
                                                   • One Kernel Instance per
                                                   core
                                                   • Program and data in L2
                                                           (SRAM) cache
                                                   • No DMA for these tests
                                                   • DMA added later




20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -              18
Two Types of Latencies:
           IRQ to ISR
           IRQ to Task
       Measured by using an automatic reload counter
       as Interrupt Source (IRQ).
       Application Diagram:




20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -              19




                                Intel-SCC            TI-C6678      ARM-M3
                               (533 MHz)                 (1 GHz)   (50 MHz)

    IRQ to ISR                 349 cycles            136 cycles    15 cycles

    IRQ to Task                5501 cycles          1367 cycles    600 cycles

    Maximum Interrupts
                               1,527,221             7,352,941     3,333,333
    per second to ISR
    Maximum Interrupts
                                 96,891                  731,529    83,333
    per second to Task



                  ARM-Cortex-M3 used as reference
20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -              20
MLX16 uBlaze Leon3         ARM-M3 XMOS        TI-C6678   Intel-SCC
 L1_Hub              400    4756      4904        2192   4854        5104       4321
 L1_Port               4        8         8          4       4          8           7
 L1_Event             70       88        72         36      54         92         55
 L1_Semaphore         54       92        96         40      64         84         64
 L1_Resource         104       96        76         40      50        144        121
 L1_FIFO             232     356        332        140   222          300        191
 L1_PacketPool         --    296        268        120   166          176        194

 All L1 services    1048    5692      5756        2572   5414        5908       4953




20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -                 21




      Tasks and semaphore on same node
      Good measure of kernel overhead
      One loop
      = 4 context switches + 4 service requests

20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -                 22
Clock Freq    Context Size     Memory     Loop time        Loop Time
                                                  location   microsecs          cycles
  ARM M3              50 MHz           16x32      internal        52.5             2625
  NXP CoolFlux              --         70x24      Internal           --            3826
  XMOS               100 MHz           14x32      Internal        26.8             2680
  Leon3               40 MHz           32x32      Internal       136.1             5444
  MLX-16                6 MHz           4x16      Internal       100.8               605
  MicroBlaze         100 MHz           32x32      internal        33.6             3360
  TI-C6678          1000 MHz           15x32     L2-SRAM           4.5             4500
  Intel SCC          533 MHz           11x32      external         4.9             2612




20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -                   23




      MPB treated as a FIFO, any Core can write to
      them.
      Atomic Variables used as locks.
      Operation:
      1.      Acquire the Lock for the MPB to write to;
      2.      Write the Message to the MPB;
      3.      Release the Lock;
      4.      Trigger LINT0 on the receiving Core.




20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -                   24
Tasks and semaphore on different nodes
      Good measure of overhead of kernel + drivers +
      communication latency

20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -   25




      Extra delay due to TX and RX drivers +
      communication latency over MPB memory
20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -   26
Uses OpenComRTOS packet switching
      Task synchronise first in Port Hub (=> ACK)
      Data copied from send packet to receive packet
      in Port Hub



20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -   27




20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -   28
20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -         29




       When utilising the experimental driver for the EDMA3
       peripherals of the TI-C6678, and EDMA3 unit
       EMDA3CC0, we achieve a throughput of 4041 MB/s
       with a buffer size of 128 KBytes, transferred between
       two buffers in the L2-SRAM of core 0.
       The advantage of using the DMA unit over using the
       CPU for copying or moving data is that during the
       transfer the CPU can perform other tasks, thus the
       transfer happens in parallel to the processing.

       GRAPH?




       Uses a Device Driver Hub to interface to the
                EDMA3 Unit (new concept).
20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -         30
Getting best and predictable real-time
      performance on modern multi-core is complex:
           Documentation barrier + hardware complexity
           Wait states => latencies
           Cache flushing adds extra overhead
           Node and memory mapping dependent performance
           Shared busses = shared resources = extra latency
      Best options: Keep It Simple and Smart:
           Nice to have features cost memory and cycles
           Shared resources to be avoided
           Best is point-to-point + DMA
20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -   31




      Integration of the Intel-SCC port into
      OpenComRTOS Designer:
           Code generators and platform meta-models
           Device Drivers
      Optimizations of the TI-C6678 Port
           Further simplifications of the EDMA3 device driver;
           Code generators;
           SoC and Board support packages
      QoS based resource scheduling (Artemis
      CRAFTERS project)

20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -   32
bernhard.sputh (@) altreonic.com
      eric.verhulst (@) altreonic.com




      Thanks for your attention
20-Jul-12 Altreonic NV – From Deep Space to Deep Sea -   33

Mais conteúdo relacionado

Mais procurados

DPDK Summit 2015 - NTT - Yoshihiro Nakajima
DPDK Summit 2015 - NTT - Yoshihiro NakajimaDPDK Summit 2015 - NTT - Yoshihiro Nakajima
DPDK Summit 2015 - NTT - Yoshihiro NakajimaJim St. Leger
 
QsNetIII, An HPC Interconnect For Peta Scale Systems
QsNetIII, An HPC Interconnect For Peta Scale SystemsQsNetIII, An HPC Interconnect For Peta Scale Systems
QsNetIII, An HPC Interconnect For Peta Scale SystemsFederica Pisani
 
OVS and DPDK - T.F. Herbert, K. Traynor, M. Gray
OVS and DPDK - T.F. Herbert, K. Traynor, M. GrayOVS and DPDK - T.F. Herbert, K. Traynor, M. Gray
OVS and DPDK - T.F. Herbert, K. Traynor, M. Grayharryvanhaaren
 
Timing Analysis of a Linux-Based CAN-to-CAN Gateway
Timing Analysis of a Linux-Based CAN-to-CAN GatewayTiming Analysis of a Linux-Based CAN-to-CAN Gateway
Timing Analysis of a Linux-Based CAN-to-CAN GatewayMichal Sojka
 
Lac2006 Lee Revell
Lac2006 Lee RevellLac2006 Lee Revell
Lac2006 Lee Revellrlrevell
 
DPDK Summit 2015 - RIFT.io - Tim Mortsolf
DPDK Summit 2015 - RIFT.io - Tim MortsolfDPDK Summit 2015 - RIFT.io - Tim Mortsolf
DPDK Summit 2015 - RIFT.io - Tim MortsolfJim St. Leger
 
A CGRA-based Approach for Accelerating Convolutional Neural Networks
A CGRA-based Approachfor Accelerating Convolutional Neural NetworksA CGRA-based Approachfor Accelerating Convolutional Neural Networks
A CGRA-based Approach for Accelerating Convolutional Neural NetworksShinya Takamaeda-Y
 
DPDK Summit 2015 - Intel - Keith Wiles
DPDK Summit 2015 - Intel - Keith WilesDPDK Summit 2015 - Intel - Keith Wiles
DPDK Summit 2015 - Intel - Keith WilesJim St. Leger
 
Using Many-Core Processors to Improve the Performance of Space Computing Plat...
Using Many-Core Processors to Improve the Performance of Space Computing Plat...Using Many-Core Processors to Improve the Performance of Space Computing Plat...
Using Many-Core Processors to Improve the Performance of Space Computing Plat...Fisnik Kraja
 
Cisco crs1
Cisco crs1Cisco crs1
Cisco crs1wjunjmt
 
HPDC 2012 presentation - June 19, 2012 - Delft, The Netherlands
HPDC 2012 presentation - June 19, 2012 -  Delft, The NetherlandsHPDC 2012 presentation - June 19, 2012 -  Delft, The Netherlands
HPDC 2012 presentation - June 19, 2012 - Delft, The Netherlandsbalmanme
 
RISC-V and OpenPOWER open-ISA and open-HW - a swiss army knife for HPC
RISC-V  and OpenPOWER open-ISA and open-HW - a swiss army knife for HPCRISC-V  and OpenPOWER open-ISA and open-HW - a swiss army knife for HPC
RISC-V and OpenPOWER open-ISA and open-HW - a swiss army knife for HPCGanesan Narayanasamy
 
An open flow for dn ns on ultra low-power RISC-V cores
An open flow for dn ns on ultra low-power RISC-V coresAn open flow for dn ns on ultra low-power RISC-V cores
An open flow for dn ns on ultra low-power RISC-V coresRISC-V International
 
QsNetIII Adaptively Routed Network For HPC
QsNetIII Adaptively Routed Network For HPCQsNetIII Adaptively Routed Network For HPC
QsNetIII Adaptively Routed Network For HPCFederica Pisani
 
Security In Dect
Security In DectSecurity In Dect
Security In DectMarc Seeger
 

Mais procurados (20)

Virtual net performance
Virtual net performanceVirtual net performance
Virtual net performance
 
DPDK Summit 2015 - NTT - Yoshihiro Nakajima
DPDK Summit 2015 - NTT - Yoshihiro NakajimaDPDK Summit 2015 - NTT - Yoshihiro Nakajima
DPDK Summit 2015 - NTT - Yoshihiro Nakajima
 
QsNetIII, An HPC Interconnect For Peta Scale Systems
QsNetIII, An HPC Interconnect For Peta Scale SystemsQsNetIII, An HPC Interconnect For Peta Scale Systems
QsNetIII, An HPC Interconnect For Peta Scale Systems
 
OVS and DPDK - T.F. Herbert, K. Traynor, M. Gray
OVS and DPDK - T.F. Herbert, K. Traynor, M. GrayOVS and DPDK - T.F. Herbert, K. Traynor, M. Gray
OVS and DPDK - T.F. Herbert, K. Traynor, M. Gray
 
Timing Analysis of a Linux-Based CAN-to-CAN Gateway
Timing Analysis of a Linux-Based CAN-to-CAN GatewayTiming Analysis of a Linux-Based CAN-to-CAN Gateway
Timing Analysis of a Linux-Based CAN-to-CAN Gateway
 
Lac2006 Lee Revell
Lac2006 Lee RevellLac2006 Lee Revell
Lac2006 Lee Revell
 
DPDK Summit 2015 - RIFT.io - Tim Mortsolf
DPDK Summit 2015 - RIFT.io - Tim MortsolfDPDK Summit 2015 - RIFT.io - Tim Mortsolf
DPDK Summit 2015 - RIFT.io - Tim Mortsolf
 
A CGRA-based Approach for Accelerating Convolutional Neural Networks
A CGRA-based Approachfor Accelerating Convolutional Neural NetworksA CGRA-based Approachfor Accelerating Convolutional Neural Networks
A CGRA-based Approach for Accelerating Convolutional Neural Networks
 
DPDK Summit 2015 - Intel - Keith Wiles
DPDK Summit 2015 - Intel - Keith WilesDPDK Summit 2015 - Intel - Keith Wiles
DPDK Summit 2015 - Intel - Keith Wiles
 
Internship end
Internship endInternship end
Internship end
 
Using Many-Core Processors to Improve the Performance of Space Computing Plat...
Using Many-Core Processors to Improve the Performance of Space Computing Plat...Using Many-Core Processors to Improve the Performance of Space Computing Plat...
Using Many-Core Processors to Improve the Performance of Space Computing Plat...
 
Ijcnc050206
Ijcnc050206Ijcnc050206
Ijcnc050206
 
Cisco crs1
Cisco crs1Cisco crs1
Cisco crs1
 
HPDC 2012 presentation - June 19, 2012 - Delft, The Netherlands
HPDC 2012 presentation - June 19, 2012 -  Delft, The NetherlandsHPDC 2012 presentation - June 19, 2012 -  Delft, The Netherlands
HPDC 2012 presentation - June 19, 2012 - Delft, The Netherlands
 
Assignmentdsp
AssignmentdspAssignmentdsp
Assignmentdsp
 
RISC-V and OpenPOWER open-ISA and open-HW - a swiss army knife for HPC
RISC-V  and OpenPOWER open-ISA and open-HW - a swiss army knife for HPCRISC-V  and OpenPOWER open-ISA and open-HW - a swiss army knife for HPC
RISC-V and OpenPOWER open-ISA and open-HW - a swiss army knife for HPC
 
An open flow for dn ns on ultra low-power RISC-V cores
An open flow for dn ns on ultra low-power RISC-V coresAn open flow for dn ns on ultra low-power RISC-V cores
An open flow for dn ns on ultra low-power RISC-V cores
 
QsNetIII Adaptively Routed Network For HPC
QsNetIII Adaptively Routed Network For HPCQsNetIII Adaptively Routed Network For HPC
QsNetIII Adaptively Routed Network For HPC
 
Security In Dect
Security In DectSecurity In Dect
Security In Dect
 
HOW Series: Knights Landing
HOW Series: Knights LandingHOW Series: Knights Landing
HOW Series: Knights Landing
 

Destaque

Tbp communication phase 2 - aug 19
Tbp   communication phase 2 - aug 19Tbp   communication phase 2 - aug 19
Tbp communication phase 2 - aug 19pgb_tbp
 
OpenComRTOS 1.4_tutorial_3o4_presentation
OpenComRTOS 1.4_tutorial_3o4_presentationOpenComRTOS 1.4_tutorial_3o4_presentation
OpenComRTOS 1.4_tutorial_3o4_presentationEric Verhulst
 
Unified Systems Engeneering with GoedelWorks
Unified Systems Engeneering with GoedelWorksUnified Systems Engeneering with GoedelWorks
Unified Systems Engeneering with GoedelWorksEric Verhulst
 
Unified Systems Engineering feasibility
Unified Systems Engineering feasibilityUnified Systems Engineering feasibility
Unified Systems Engineering feasibilityEric Verhulst
 
Zen and the art of safety engineering
Zen and the art of safety engineeringZen and the art of safety engineering
Zen and the art of safety engineeringEric Verhulst
 
Session 1 introduction concurrent programming
Session 1 introduction  concurrent programmingSession 1 introduction  concurrent programming
Session 1 introduction concurrent programmingEric Verhulst
 
OpenComRtos 1.4_tutorial_1o4_presentation
OpenComRtos 1.4_tutorial_1o4_presentationOpenComRtos 1.4_tutorial_1o4_presentation
OpenComRtos 1.4_tutorial_1o4_presentationEric Verhulst
 
Open ComRTOS 1.4_tutorial_2o4_presentation
Open ComRTOS 1.4_tutorial_2o4_presentationOpen ComRTOS 1.4_tutorial_2o4_presentation
Open ComRTOS 1.4_tutorial_2o4_presentationEric Verhulst
 
Due soon revised
Due soon revisedDue soon revised
Due soon revisedpgb_tbp
 

Destaque (9)

Tbp communication phase 2 - aug 19
Tbp   communication phase 2 - aug 19Tbp   communication phase 2 - aug 19
Tbp communication phase 2 - aug 19
 
OpenComRTOS 1.4_tutorial_3o4_presentation
OpenComRTOS 1.4_tutorial_3o4_presentationOpenComRTOS 1.4_tutorial_3o4_presentation
OpenComRTOS 1.4_tutorial_3o4_presentation
 
Unified Systems Engeneering with GoedelWorks
Unified Systems Engeneering with GoedelWorksUnified Systems Engeneering with GoedelWorks
Unified Systems Engeneering with GoedelWorks
 
Unified Systems Engineering feasibility
Unified Systems Engineering feasibilityUnified Systems Engineering feasibility
Unified Systems Engineering feasibility
 
Zen and the art of safety engineering
Zen and the art of safety engineeringZen and the art of safety engineering
Zen and the art of safety engineering
 
Session 1 introduction concurrent programming
Session 1 introduction  concurrent programmingSession 1 introduction  concurrent programming
Session 1 introduction concurrent programming
 
OpenComRtos 1.4_tutorial_1o4_presentation
OpenComRtos 1.4_tutorial_1o4_presentationOpenComRtos 1.4_tutorial_1o4_presentation
OpenComRtos 1.4_tutorial_1o4_presentation
 
Open ComRTOS 1.4_tutorial_2o4_presentation
Open ComRTOS 1.4_tutorial_2o4_presentationOpen ComRTOS 1.4_tutorial_2o4_presentation
Open ComRTOS 1.4_tutorial_2o4_presentation
 
Due soon revised
Due soon revisedDue soon revised
Due soon revised
 

Semelhante a MARC ONERA Toulouse2012 Altreonic

Parallelism Processor Design
Parallelism Processor DesignParallelism Processor Design
Parallelism Processor DesignSri Prasanna
 
The Potential Impact of Software Defined Networking SDN on Security
The Potential Impact of Software Defined Networking SDN on SecurityThe Potential Impact of Software Defined Networking SDN on Security
The Potential Impact of Software Defined Networking SDN on SecurityBrent Salisbury
 
Software Defined Data Centers - June 2012
Software Defined Data Centers - June 2012Software Defined Data Centers - June 2012
Software Defined Data Centers - June 2012Brent Salisbury
 
Container Attached Storage (CAS) with OpenEBS - Berlin Kubernetes Meetup - Ma...
Container Attached Storage (CAS) with OpenEBS - Berlin Kubernetes Meetup - Ma...Container Attached Storage (CAS) with OpenEBS - Berlin Kubernetes Meetup - Ma...
Container Attached Storage (CAS) with OpenEBS - Berlin Kubernetes Meetup - Ma...OpenEBS
 
Achieving Performance Isolation with Lightweight Co-Kernels
Achieving Performance Isolation with Lightweight Co-KernelsAchieving Performance Isolation with Lightweight Co-Kernels
Achieving Performance Isolation with Lightweight Co-KernelsJiannan Ouyang, PhD
 
OpenStack and OpenFlow Demos
OpenStack and OpenFlow DemosOpenStack and OpenFlow Demos
OpenStack and OpenFlow DemosBrent Salisbury
 
Apache Spark on Supercomputers: A Tale of the Storage Hierarchy with Costin I...
Apache Spark on Supercomputers: A Tale of the Storage Hierarchy with Costin I...Apache Spark on Supercomputers: A Tale of the Storage Hierarchy with Costin I...
Apache Spark on Supercomputers: A Tale of the Storage Hierarchy with Costin I...Databricks
 
Apache Spark on Supercomputers: A Tale of the Storage Hierarchy with Costin I...
Apache Spark on Supercomputers: A Tale of the Storage Hierarchy with Costin I...Apache Spark on Supercomputers: A Tale of the Storage Hierarchy with Costin I...
Apache Spark on Supercomputers: A Tale of the Storage Hierarchy with Costin I...Databricks
 
Building efficient 5G NR base stations with Intel® Xeon® Scalable Processors
Building efficient 5G NR base stations with Intel® Xeon® Scalable Processors Building efficient 5G NR base stations with Intel® Xeon® Scalable Processors
Building efficient 5G NR base stations with Intel® Xeon® Scalable Processors Michelle Holley
 
Scaling the Container Dataplane
Scaling the Container Dataplane Scaling the Container Dataplane
Scaling the Container Dataplane Michelle Holley
 
ODSA Use Case - SmartNIC
ODSA Use Case - SmartNICODSA Use Case - SmartNIC
ODSA Use Case - SmartNICODSA Workgroup
 
Introduction to DPDK
Introduction to DPDKIntroduction to DPDK
Introduction to DPDKKernel TLV
 
ODSA Workshop: Development Effort Summary
ODSA Workshop: Development Effort SummaryODSA Workshop: Development Effort Summary
ODSA Workshop: Development Effort SummaryODSA Workgroup
 
OCP Engineering Workshop at UNH
OCP Engineering Workshop at UNH OCP Engineering Workshop at UNH
OCP Engineering Workshop at UNH 호용 류
 
Oracle 11g R2 RAC implementation and concept
Oracle 11g R2 RAC implementation and conceptOracle 11g R2 RAC implementation and concept
Oracle 11g R2 RAC implementation and conceptSantosh Kangane
 
Jetson AGX Xavier and the New Era of Autonomous Machines
Jetson AGX Xavier and the New Era of Autonomous MachinesJetson AGX Xavier and the New Era of Autonomous Machines
Jetson AGX Xavier and the New Era of Autonomous MachinesDustin Franklin
 

Semelhante a MARC ONERA Toulouse2012 Altreonic (20)

Userspace networking
Userspace networkingUserspace networking
Userspace networking
 
Parallelism Processor Design
Parallelism Processor DesignParallelism Processor Design
Parallelism Processor Design
 
The Potential Impact of Software Defined Networking SDN on Security
The Potential Impact of Software Defined Networking SDN on SecurityThe Potential Impact of Software Defined Networking SDN on Security
The Potential Impact of Software Defined Networking SDN on Security
 
Understanding DPDK
Understanding DPDKUnderstanding DPDK
Understanding DPDK
 
Software Defined Data Centers - June 2012
Software Defined Data Centers - June 2012Software Defined Data Centers - June 2012
Software Defined Data Centers - June 2012
 
Container Attached Storage (CAS) with OpenEBS - Berlin Kubernetes Meetup - Ma...
Container Attached Storage (CAS) with OpenEBS - Berlin Kubernetes Meetup - Ma...Container Attached Storage (CAS) with OpenEBS - Berlin Kubernetes Meetup - Ma...
Container Attached Storage (CAS) with OpenEBS - Berlin Kubernetes Meetup - Ma...
 
Achieving Performance Isolation with Lightweight Co-Kernels
Achieving Performance Isolation with Lightweight Co-KernelsAchieving Performance Isolation with Lightweight Co-Kernels
Achieving Performance Isolation with Lightweight Co-Kernels
 
OpenStack and OpenFlow Demos
OpenStack and OpenFlow DemosOpenStack and OpenFlow Demos
OpenStack and OpenFlow Demos
 
Apache Spark on Supercomputers: A Tale of the Storage Hierarchy with Costin I...
Apache Spark on Supercomputers: A Tale of the Storage Hierarchy with Costin I...Apache Spark on Supercomputers: A Tale of the Storage Hierarchy with Costin I...
Apache Spark on Supercomputers: A Tale of the Storage Hierarchy with Costin I...
 
Apache Spark on Supercomputers: A Tale of the Storage Hierarchy with Costin I...
Apache Spark on Supercomputers: A Tale of the Storage Hierarchy with Costin I...Apache Spark on Supercomputers: A Tale of the Storage Hierarchy with Costin I...
Apache Spark on Supercomputers: A Tale of the Storage Hierarchy with Costin I...
 
Building efficient 5G NR base stations with Intel® Xeon® Scalable Processors
Building efficient 5G NR base stations with Intel® Xeon® Scalable Processors Building efficient 5G NR base stations with Intel® Xeon® Scalable Processors
Building efficient 5G NR base stations with Intel® Xeon® Scalable Processors
 
Scaling the Container Dataplane
Scaling the Container Dataplane Scaling the Container Dataplane
Scaling the Container Dataplane
 
Dpdk applications
Dpdk applicationsDpdk applications
Dpdk applications
 
ODSA Use Case - SmartNIC
ODSA Use Case - SmartNICODSA Use Case - SmartNIC
ODSA Use Case - SmartNIC
 
Introduction to DPDK
Introduction to DPDKIntroduction to DPDK
Introduction to DPDK
 
ODSA Workshop: Development Effort Summary
ODSA Workshop: Development Effort SummaryODSA Workshop: Development Effort Summary
ODSA Workshop: Development Effort Summary
 
OCP Engineering Workshop at UNH
OCP Engineering Workshop at UNH OCP Engineering Workshop at UNH
OCP Engineering Workshop at UNH
 
L05 parallel
L05 parallelL05 parallel
L05 parallel
 
Oracle 11g R2 RAC implementation and concept
Oracle 11g R2 RAC implementation and conceptOracle 11g R2 RAC implementation and concept
Oracle 11g R2 RAC implementation and concept
 
Jetson AGX Xavier and the New Era of Autonomous Machines
Jetson AGX Xavier and the New Era of Autonomous MachinesJetson AGX Xavier and the New Era of Autonomous Machines
Jetson AGX Xavier and the New Era of Autonomous Machines
 

Último

[BuildWithAI] Introduction to Gemini.pdf
[BuildWithAI] Introduction to Gemini.pdf[BuildWithAI] Introduction to Gemini.pdf
[BuildWithAI] Introduction to Gemini.pdfSandro Moreira
 
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers:  A Deep Dive into Serverless Spatial Data and FMECloud Frontiers:  A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FMESafe Software
 
Elevate Developer Efficiency & build GenAI Application with Amazon Q​
Elevate Developer Efficiency & build GenAI Application with Amazon Q​Elevate Developer Efficiency & build GenAI Application with Amazon Q​
Elevate Developer Efficiency & build GenAI Application with Amazon Q​Bhuvaneswari Subramani
 
Vector Search -An Introduction in Oracle Database 23ai.pptx
Vector Search -An Introduction in Oracle Database 23ai.pptxVector Search -An Introduction in Oracle Database 23ai.pptx
Vector Search -An Introduction in Oracle Database 23ai.pptxRemote DBA Services
 
TrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data DiscoveryTrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data DiscoveryTrustArc
 
Artificial Intelligence Chap.5 : Uncertainty
Artificial Intelligence Chap.5 : UncertaintyArtificial Intelligence Chap.5 : Uncertainty
Artificial Intelligence Chap.5 : UncertaintyKhushali Kathiriya
 
Rising Above_ Dubai Floods and the Fortitude of Dubai International Airport.pdf
Rising Above_ Dubai Floods and the Fortitude of Dubai International Airport.pdfRising Above_ Dubai Floods and the Fortitude of Dubai International Airport.pdf
Rising Above_ Dubai Floods and the Fortitude of Dubai International Airport.pdfOrbitshub
 
Apidays New York 2024 - The value of a flexible API Management solution for O...
Apidays New York 2024 - The value of a flexible API Management solution for O...Apidays New York 2024 - The value of a flexible API Management solution for O...
Apidays New York 2024 - The value of a flexible API Management solution for O...apidays
 
WSO2's API Vision: Unifying Control, Empowering Developers
WSO2's API Vision: Unifying Control, Empowering DevelopersWSO2's API Vision: Unifying Control, Empowering Developers
WSO2's API Vision: Unifying Control, Empowering DevelopersWSO2
 
MS Copilot expands with MS Graph connectors
MS Copilot expands with MS Graph connectorsMS Copilot expands with MS Graph connectors
MS Copilot expands with MS Graph connectorsNanddeep Nachan
 
Mcleodganj Call Girls 🥰 8617370543 Service Offer VIP Hot Model
Mcleodganj Call Girls 🥰 8617370543 Service Offer VIP Hot ModelMcleodganj Call Girls 🥰 8617370543 Service Offer VIP Hot Model
Mcleodganj Call Girls 🥰 8617370543 Service Offer VIP Hot ModelDeepika Singh
 
Strategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a FresherStrategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a FresherRemote DBA Services
 
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, AdobeApidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobeapidays
 
FWD Group - Insurer Innovation Award 2024
FWD Group - Insurer Innovation Award 2024FWD Group - Insurer Innovation Award 2024
FWD Group - Insurer Innovation Award 2024The Digital Insurer
 
Apidays New York 2024 - APIs in 2030: The Risk of Technological Sleepwalk by ...
Apidays New York 2024 - APIs in 2030: The Risk of Technological Sleepwalk by ...Apidays New York 2024 - APIs in 2030: The Risk of Technological Sleepwalk by ...
Apidays New York 2024 - APIs in 2030: The Risk of Technological Sleepwalk by ...apidays
 
presentation ICT roal in 21st century education
presentation ICT roal in 21st century educationpresentation ICT roal in 21st century education
presentation ICT roal in 21st century educationjfdjdjcjdnsjd
 
Why Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire businessWhy Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire businesspanagenda
 
DEV meet-up UiPath Document Understanding May 7 2024 Amsterdam
DEV meet-up UiPath Document Understanding May 7 2024 AmsterdamDEV meet-up UiPath Document Understanding May 7 2024 Amsterdam
DEV meet-up UiPath Document Understanding May 7 2024 AmsterdamUiPathCommunity
 
Biography Of Angeliki Cooney | Senior Vice President Life Sciences | Albany, ...
Biography Of Angeliki Cooney | Senior Vice President Life Sciences | Albany, ...Biography Of Angeliki Cooney | Senior Vice President Life Sciences | Albany, ...
Biography Of Angeliki Cooney | Senior Vice President Life Sciences | Albany, ...Angeliki Cooney
 

Último (20)

[BuildWithAI] Introduction to Gemini.pdf
[BuildWithAI] Introduction to Gemini.pdf[BuildWithAI] Introduction to Gemini.pdf
[BuildWithAI] Introduction to Gemini.pdf
 
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers:  A Deep Dive into Serverless Spatial Data and FMECloud Frontiers:  A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
 
Elevate Developer Efficiency & build GenAI Application with Amazon Q​
Elevate Developer Efficiency & build GenAI Application with Amazon Q​Elevate Developer Efficiency & build GenAI Application with Amazon Q​
Elevate Developer Efficiency & build GenAI Application with Amazon Q​
 
Vector Search -An Introduction in Oracle Database 23ai.pptx
Vector Search -An Introduction in Oracle Database 23ai.pptxVector Search -An Introduction in Oracle Database 23ai.pptx
Vector Search -An Introduction in Oracle Database 23ai.pptx
 
TrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data DiscoveryTrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
 
Artificial Intelligence Chap.5 : Uncertainty
Artificial Intelligence Chap.5 : UncertaintyArtificial Intelligence Chap.5 : Uncertainty
Artificial Intelligence Chap.5 : Uncertainty
 
Rising Above_ Dubai Floods and the Fortitude of Dubai International Airport.pdf
Rising Above_ Dubai Floods and the Fortitude of Dubai International Airport.pdfRising Above_ Dubai Floods and the Fortitude of Dubai International Airport.pdf
Rising Above_ Dubai Floods and the Fortitude of Dubai International Airport.pdf
 
Apidays New York 2024 - The value of a flexible API Management solution for O...
Apidays New York 2024 - The value of a flexible API Management solution for O...Apidays New York 2024 - The value of a flexible API Management solution for O...
Apidays New York 2024 - The value of a flexible API Management solution for O...
 
WSO2's API Vision: Unifying Control, Empowering Developers
WSO2's API Vision: Unifying Control, Empowering DevelopersWSO2's API Vision: Unifying Control, Empowering Developers
WSO2's API Vision: Unifying Control, Empowering Developers
 
MS Copilot expands with MS Graph connectors
MS Copilot expands with MS Graph connectorsMS Copilot expands with MS Graph connectors
MS Copilot expands with MS Graph connectors
 
Mcleodganj Call Girls 🥰 8617370543 Service Offer VIP Hot Model
Mcleodganj Call Girls 🥰 8617370543 Service Offer VIP Hot ModelMcleodganj Call Girls 🥰 8617370543 Service Offer VIP Hot Model
Mcleodganj Call Girls 🥰 8617370543 Service Offer VIP Hot Model
 
Strategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a FresherStrategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a Fresher
 
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
 
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, AdobeApidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
 
FWD Group - Insurer Innovation Award 2024
FWD Group - Insurer Innovation Award 2024FWD Group - Insurer Innovation Award 2024
FWD Group - Insurer Innovation Award 2024
 
Apidays New York 2024 - APIs in 2030: The Risk of Technological Sleepwalk by ...
Apidays New York 2024 - APIs in 2030: The Risk of Technological Sleepwalk by ...Apidays New York 2024 - APIs in 2030: The Risk of Technological Sleepwalk by ...
Apidays New York 2024 - APIs in 2030: The Risk of Technological Sleepwalk by ...
 
presentation ICT roal in 21st century education
presentation ICT roal in 21st century educationpresentation ICT roal in 21st century education
presentation ICT roal in 21st century education
 
Why Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire businessWhy Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire business
 
DEV meet-up UiPath Document Understanding May 7 2024 Amsterdam
DEV meet-up UiPath Document Understanding May 7 2024 AmsterdamDEV meet-up UiPath Document Understanding May 7 2024 Amsterdam
DEV meet-up UiPath Document Understanding May 7 2024 Amsterdam
 
Biography Of Angeliki Cooney | Senior Vice President Life Sciences | Albany, ...
Biography Of Angeliki Cooney | Senior Vice President Life Sciences | Albany, ...Biography Of Angeliki Cooney | Senior Vice President Life Sciences | Albany, ...
Biography Of Angeliki Cooney | Senior Vice President Life Sciences | Albany, ...
 

MARC ONERA Toulouse2012 Altreonic

  • 1. Trustworthy Forever From Deep Space To Deep Sea History goes back to 1989 (Eonic Systems) VIRTUOSO parallel RTOS (T800, C40, C6x, 2016x, TS102, G4, …) – CSP based Used from 1 CPU to 1600 DSPs (sonar, radar) to 12000 nodes (heterogeneous) Acquired by Wind River Systems in 2001 Altreonic: created as spin-off in 2008 following R&D Unified systems engineering (GoedelWorks) Formalised when possible Network-centric OpenComRTOS: Used as test case for use of formal techniques Binary/source and Open Technology License model 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 2
  • 2. Computation to communication ratio: Depends on application Input rate = output rate = ½ computation rate ideal Ideal ratio = 1, typically minimum 5 to 10 Data communication is bottleneck: Set-up latency, no polling Task to task/memory to memory bandwidth is real target Concurrency to mask communication latencies DMA => requires additional busses Data protection: Pointers are fast but very dangerous memcpy has no distributed semantics! (should be _W) Heterogeneous many-cores: data-types 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 3 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 4
  • 3. Novel programming model, but long formal history (Hoare’s CSP, 1975) Use of TLA+/TLC for design and verification Unexpected result: 10X smaller code size; Easy to Port to new Platforms; Low amount of assembly; 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 5 Network-centric (RT)OS, MP by default: Concurrency at the core (“Interacting Entities”); Pragmatic superset of CSP (Hoare); Scalable yet very small: typically 2 to 5 kiB/node; Real-time communication as system level service; Unique support for distributed priority inheritance; Heterogeneous target /communication support; Integrate seamlessly “legacy OS” nodes; Virtual Single Processor model; Visual modeling/ programming with code generators; Capable of fault-tolerance and resource management. 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 6
  • 4. Basis: RMA => priority based, preemptive Additional: timer based Everything is priority ordered e.g. waiting lists Packet based communication Priority Inheritance support with ceiling level Single processor Unique distributed implementation 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 7 The Generic Hub as metamodel Similar to a Guarded Atomic Action, or a pragmatic superset of CSP 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 8
  • 5. 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 9 Hub Entity Semantics Event Synchronisation on boolean event, N to N Semaphore Synchronisation with counter for async signalling, N to N Port Synchronisation with exchange of Packet, N to N FIFO queue Buffered, async communication, except on FIFO full or empty, N to N Resource Logical resource to guard critical section (with priority inheritance) Memory pool Linked list of memory blocks Synchronisation Semantics _NW Non waiting => returns immediately _W Waiting until synchronisation (blocking) _WT Waiting with a TimeOut. 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 10
  • 6. OpenComRTOS Designer, for Modeling: Topology Diagram, defines the Hardware Setup Application Diagram, defines the Entities and their Interactions Source Code, defines the sequential parts. Code Generators: Generate the configuration for OpenComRTOS from the Models OpenComRTOS: Runtime Layer providing the Interacting Entities. 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 11 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 12
  • 7. 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 13 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 14
  • 8. • Up to 10x smaller than traditional design (thanks to formal development) • Less power, less memory, easier to verify, scalable ... CPU Type Codesize ARM-Cortex-M3 2.5 – 4.0 kB XMOS-XS1 5.0 – 7.5 kB PowerPC e600 7.1 – 9.8 kB TI-C6678x (8 core DSP) 5.1 – 7.7 kB Intel-SCC (48 Pentium cores) 4.3 – 5 kB Code size figures (in Bytes) obtained for our different ports, -Os Dormant ports: MLX16 (2K), Xilinx MB (5K), Leon3 (5K), CoolFlux DSP (2K) 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 15 Comparing Intel 48-core SCC and TI 8-core TMS320C6678 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 16
  • 9. • L1 cache: 16 KB • L2 cache: 256 KB • RTOS kernel on each core • Communication using memory of MPB (16 KB) 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 17 • “RoC“ (Rack On a chip) • Hierarchical Interrupt Routing up to 1000 possible interrupts per core! • One Kernel Instance per core • Program and data in L2 (SRAM) cache • No DMA for these tests • DMA added later 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 18
  • 10. Two Types of Latencies: IRQ to ISR IRQ to Task Measured by using an automatic reload counter as Interrupt Source (IRQ). Application Diagram: 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 19 Intel-SCC TI-C6678 ARM-M3 (533 MHz) (1 GHz) (50 MHz) IRQ to ISR 349 cycles 136 cycles 15 cycles IRQ to Task 5501 cycles 1367 cycles 600 cycles Maximum Interrupts 1,527,221 7,352,941 3,333,333 per second to ISR Maximum Interrupts 96,891 731,529 83,333 per second to Task ARM-Cortex-M3 used as reference 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 20
  • 11. MLX16 uBlaze Leon3 ARM-M3 XMOS TI-C6678 Intel-SCC L1_Hub 400 4756 4904 2192 4854 5104 4321 L1_Port 4 8 8 4 4 8 7 L1_Event 70 88 72 36 54 92 55 L1_Semaphore 54 92 96 40 64 84 64 L1_Resource 104 96 76 40 50 144 121 L1_FIFO 232 356 332 140 222 300 191 L1_PacketPool -- 296 268 120 166 176 194 All L1 services 1048 5692 5756 2572 5414 5908 4953 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 21 Tasks and semaphore on same node Good measure of kernel overhead One loop = 4 context switches + 4 service requests 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 22
  • 12. Clock Freq Context Size Memory Loop time Loop Time location microsecs cycles ARM M3 50 MHz 16x32 internal 52.5 2625 NXP CoolFlux -- 70x24 Internal -- 3826 XMOS 100 MHz 14x32 Internal 26.8 2680 Leon3 40 MHz 32x32 Internal 136.1 5444 MLX-16 6 MHz 4x16 Internal 100.8 605 MicroBlaze 100 MHz 32x32 internal 33.6 3360 TI-C6678 1000 MHz 15x32 L2-SRAM 4.5 4500 Intel SCC 533 MHz 11x32 external 4.9 2612 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 23 MPB treated as a FIFO, any Core can write to them. Atomic Variables used as locks. Operation: 1. Acquire the Lock for the MPB to write to; 2. Write the Message to the MPB; 3. Release the Lock; 4. Trigger LINT0 on the receiving Core. 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 24
  • 13. Tasks and semaphore on different nodes Good measure of overhead of kernel + drivers + communication latency 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 25 Extra delay due to TX and RX drivers + communication latency over MPB memory 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 26
  • 14. Uses OpenComRTOS packet switching Task synchronise first in Port Hub (=> ACK) Data copied from send packet to receive packet in Port Hub 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 27 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 28
  • 15. 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 29 When utilising the experimental driver for the EDMA3 peripherals of the TI-C6678, and EDMA3 unit EMDA3CC0, we achieve a throughput of 4041 MB/s with a buffer size of 128 KBytes, transferred between two buffers in the L2-SRAM of core 0. The advantage of using the DMA unit over using the CPU for copying or moving data is that during the transfer the CPU can perform other tasks, thus the transfer happens in parallel to the processing. GRAPH? Uses a Device Driver Hub to interface to the EDMA3 Unit (new concept). 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 30
  • 16. Getting best and predictable real-time performance on modern multi-core is complex: Documentation barrier + hardware complexity Wait states => latencies Cache flushing adds extra overhead Node and memory mapping dependent performance Shared busses = shared resources = extra latency Best options: Keep It Simple and Smart: Nice to have features cost memory and cycles Shared resources to be avoided Best is point-to-point + DMA 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 31 Integration of the Intel-SCC port into OpenComRTOS Designer: Code generators and platform meta-models Device Drivers Optimizations of the TI-C6678 Port Further simplifications of the EDMA3 device driver; Code generators; SoC and Board support packages QoS based resource scheduling (Artemis CRAFTERS project) 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 32
  • 17. bernhard.sputh (@) altreonic.com eric.verhulst (@) altreonic.com Thanks for your attention 20-Jul-12 Altreonic NV – From Deep Space to Deep Sea - 33