TigerSHARC processor and evaluation board Different capabilities Different functionality

2 / 26 Concepts tackled Differences between “processor” and “evaluation board” Functionality present on TigerSHARC evaluation board TigerSHARC is NOT a microcontroller.  A/D and D/A added simply to give evaluation board “something to process” ENEL (Topic 23) “project”  Learn about TigerSHARC processor architecture  Learn basics of how assembly code and architecture interact  Set up a project (mainly in) in “C++” to capture a signal  Do some signal processing in “C++”  Work through optimizing “one” subroutine into assembly code using more and more parallel processor features  Use the same (or equivalent) code to look at optimization issues on other processor Blackfin, TI C64XX+ and C64XX, and Beagle (2 core with ARM and TI C64XX

3 / 26 TS201S Evaluation board contains 2 processors P0 (DSPA) and P1 (DSPB) TWO TS201S processors Error messages when linking about P0 and P1 mean problems with code on specific Processor 0 (P0) Or processor (P1)

4 / 26 Evaluation board – Link Ports Designed for “star” Configuration – Up to 8 processors working together P0 to P7 (selectable by switches on back of evaluation board (Never tried – Don’t destroy link ports)

5 / 26 JTAG debug ONLY AVALIABLE IN 3 rd floor labs ICT318 TigerSHARC (some stations). ICT320 Blackfin (some stations). Some boards available for home use of USB connector. Lower speed communication rates than JTAG (factor of ). Discuss deposit issues for home loan. Audio connectors. Max input voltage around 1 V p-p with no DC. Too great a voltage destroys the board Use MPEG or other players Keep Volume down Damage and distortion issues

6 / 26 Buttons and LEDs split between processors Much “stuff” on the board shared between processors (but NOT all) Will be using IRQ lines IRQ0 – button switch on panel IRQ1 – digital signal from light sensor Demonstrate “SIMPLE” dual processing

7 / 26 Evaluation Board Architecture P0 P1

8 / 26 Processors can share EXTERNAL memory Memory allocation depends on ID # Processors can access each other’s INTERNAL memory and registers MEMORY MAP USED IN 8 TigerSHARC PROCESSOR CLUSTER

9 / 26 Note – labeled processor A and B rather than processor 0 and 1

10 / 26 Note – labeled processor A and B rather than processor 0 and 1 Will be using IRQ lines IRQ0 –button switch on panel IRQ1 – digital signal from light sensor Demonstrate “SIMPLE” dual processing

11 / 26 How to use the evaluation board capability Read a lot in the hardware manual Modify some of the example programs provided in C:\Program Files\Analog Devices\VisualDSP 5.0\TS\Examples\ADSP-TS201 EZ-Kit Lite and then read a lot in the hardware manual

CONCEPT OF LABS Signal processing demonstrated through the development of algorithms for an improved doctors “stethoscope” and associated equipment. Demonstration rather than commercial exploitation of TigerSHARC’s highly parallel processing capability.  Basically looking for something reasonably easy, but non-trivial, to hang the lectures and laboratories on. 12 / 26

Example of a suitable project TSL230R-Based Pulse Oximeter A small section of plastic electrical conduit houses both the TSL230R sensor and the red and IR LEDs. The TSL230R registers the amount of light passing through the inside diameter of the conduit, which, in this case, is through a patient’s finger. Adapted from print/1204/Bachiochi173/4.htm 13 / 26

What to we expect to see? As blood flows through the circulatory system, it changes density because of the heart’s pumping pressure. This change also changes the absorption rate of light, effectively modulating the light absorption. The total light absorption is a combination of modulated and constant absorption, which is similar to a small AC noise riding atop a DC voltage. Adapted from print/1204/Bachiochi173/4.htmwww.circuitcellar.com/library/ 14 / 26

How we far will go in the laboratory is your choice Study the absorption relationship of oxygen levels in the blood for the red and IR wavelengths. Notice how the oxygen level affects the absorption rate at the red wavelength while it remains almost constant at IR wavelengths. Adapted from print/1204/Bachiochi173/4.htm 15 / 26

Details from Circuit Cellar Article The TSL230R’s frequency output displayed on my oscilloscope shows a slow frequency jitter marked by the vertical cursors. The output frequency shifts with the varying amount of light absorption because of the blood pulsing within the light’s path. Adapted from print/1204/Bachiochi173/5.htm 16 / 26

Period of light sensor from article – can we do better? This Excel chart displays imported sample values over time. Samples of the output frequency (30 samples per second formatted in microseconds) were sent out the serial port and captured to a file by HyperTerminal. Adapted from print/1204/Bachiochi173/7.htm 17 / 26

Labs. Check article for how “determine heart rate” was handled in original article. Can we pick out a more accurate heart beat / signal using a DSP convolution / correlation approach?  These are numerical methods involving much memory use, adds and multiplications. Typical DSP stuff.  Straight forward optimization show TigerSHARC capability  If did on Beagle, Blackfin – what differences in development and performance 18 / 26

19 / 26 ECE-ADI-Project Was asked to develop (as part of Analog Devices University Ambassadorship) to develop demonstration programs and place on the web Have now developed  SHARC – ADSP (Previous 515) Audio project utilities  Blackfin – ADSP-BF533 (ENCM511 in 2008) Audio project utilities, Video Project utilities  Blackfin – ADSP-BF533 (Current ENCM491 – Nygren) Real time operating systems  TigerSHARC – ADSP-TS201 (Current ENCM515, ENCM (Topic23)  Embedded Test Driven Development Environment (E-TDD) Blackfin, TigerSHARC, SHARC, Microsoft studio Find at ECE-ADI-Project/Index/index.htm

The code is not well written unless it is testable Write C code with tests to show concepts works Optimize the code in one way, use same test to show still works but faster Optimize the code in another way, use same test to show still works but faster still Move code over to another processor, use same tests etc 20 / 26

21 / 31 Standard development method Waterfall method Heavy on documentation, light on testing TLD -- IDAA Test Last Development (if done at all)

22 / 26 TESTS EXPRESS DESIGN INTENT TDD approach to develop C++ code CUSTOMER DEVELOPER Work with customer to check that the TESTS PROPERLY EXPRESS WHAT CUSTOME WANTS DONE Iterative process with customer “heavily involved” – “Agile” methodology.

Some embedded stuff on test driven development in the literature Examine in more detail as part of “individual” presentation? 23 / 26

24 / 26 Rule #1 of assembly code programming Don’t use assembly code programming What speed improvements occur when you turn on the “C++ optimizer” – use Embedded Unit “Timing features”? Wewill look at the code in detail to understand how the compiler is taking advantage of TigerSHARC processor to get code optimization.  Initially we will “look” at that code and say “oh my!”  Later we can think about comparing the performance of the compilers on different processors. Position paper or project Speed improvements when we use the “C++” optimizer

25 / 26 Concepts tackled Differences between “processor” and “evaluation board” Functionality present on TigerSHARC evaluation board TigerSHARC is NOT a microcontroller.  A/D and D/A added simply to give evaluation board “something to process” ENEL (Topic 23) “project”  Learn about TigerSHARC processor architecture  Learn basics of how assembly code and architecture interact  Set up a project (mainly in) in “C++” to capture a signal  Do some signal processing in “C++”  Work through optimizing “one” subroutine into assembly code using more and more parallel processor features  Use the same (or equivalent) code to look at optimization issues on other processor Blackfin, TI C64XX+ and C64XX, and Beagle (2 core with ARM and TI C64XX