Tools for Reconfigurable Supercomputing. Kris Gaj George Mason University

Transcription

1 Tools for Reconfigurable Supercomputing Kris Gaj George Mason University 1

2 Application Development for Reconfigurable Computers Program Entry Platform mapping Debugging & Verification Compilation Execution 2

3 Tasks Addressed in This Presentation Program Entry Platform mapping Debugging & Verification Compilation Execution 3

4 Program Entry Program 4

5 Platform Mapping SW/HW Partitioning Program Software (executed in the microprocessor system) Hardware (executed in the reconfigurable processor system) 5

6 SW/HW Partitioning & Coding Traditional Approach Specification SW/HW Partitioning SW Coding HW Coding SW Compilation HW Compilation SW Profiling HW Profiling 6

7 SW/HW Partitioning & Coding New Approach Specification SW/HW Coding SW/HW Partitioning SW Compilation HW Compilation SW Profiling HW Profiling 7

8 Program Entry for FPGA Accelerator Boards Traditional Software Hardware HDL Graphical Data Flow Diagram HLL Extended Software Hardware 8 Increased productivity Increased capability to describe parallel execution

9 Program Entry for Reconfigurable Computers Star Bridge Software Hardware HDL porting EDIF Graphical Data Flow Diagram COM objects HLL SRC Software Hardware HDL macros Increased productivity Increased capability to describe parallel execution 9

10 Examples of Software Environments for Reconfigurable Computers DSP-oriented Corefire from Annapolis Microsystems Xtreme DSP from Xilinx Inc. & MathWorks General-purpose Viva from Star Bridge Systems SRC Software Environment from SRC Computers, Inc. 10

11 CoreFire FPGA Application Builder Design viewer Library Cores window Message Window 11 Diagram Editor

12 Xtreme DSP Environment 12

13 Star Bridge Software Environment User input Graphical User Interface VIVA Netlists.ngo files Xilinx Place & Route Application executable 13.bin files Configuration bitstreams

14 Sheets Library Object 14

15 SRC Compilation Process Application sources Macro sources.c or.f files.vhd or.v files HDL sources.v files Logic synthesis µp Compiler MAP Compiler Netlists.ngo files Object files.o files.o files Place & Route Linker Application executable.bin files Configuration bitstreams 15

16 Cray XD1 Traditional Design Flow Cores HDL RA I/F, Metadata QDR SRAM I/F Synthesize Implement Download Binary File VHDL, Verilog, C Synplicity, Leonardo, Precision, Xilinx ISE Simulate Xilinx ISE From Command line or Application Verify Modelsim 16 Xilinx ChipScope Source: [Cray, MAPLD04]

17 Cray XD1 New design flow 17 Source: [Cray, MAPLD04]

18 SGI Altix Design Flow (HDLs) IA-32 Linux Machine.v,.vhd Design Entry (Verilog, VHDL).v,.vhd Design Synthesis (Synplify Pro, Amplify).edf Design iterations.v,.vhd Design Verification Behavioral Simulation (VCS, Modelsim) Metadata Processing (Python) Design Implementation (ISE).ncd,.pcf Static Timing Analysis (ISE Timing Analyzer).cfg.bin Altix Device Programming (RASC Abstraction Layer, Device Manager, Device Driver).c Real-time Verification (gdb) 18

19 SGI Altix Design Flow (HLLs) IA-32 Linux Machine.v,.vhd HLL Design Entry (Handel-C, Impulse C, Mitrion C, Viva) Metadata Processing (Python) RTL Generation and Integration with Core Services.v,.vhd.cfg Design Synthesis (Synplify Pro, Amplify).edf Design Implementation (ISE).bin.v,.vhd.ncd,.pcf Design Verification Behavioral Simulation (VCS, Modelsim) Static Timing Analysis (ISE Timing Analyzer) Altix Device Programming (RASC Abstraction Layer, Device Manager, Device Driver).c Real-time Verification (gdb) 19

20 Platform Mapping FPGA mapping Program Hardware FPGA 1 FPGA 2 Software FPGA 3 FPGA 4 20

21 Example of FPGA Mapping FPGA 1 FPGA 2 FPGA multiply divide multiply divide add add FPGA 1 FPGA 2 multiply divide add 21

22 FPGA Mapping in SRC FPGA1.mc void fpga1(int64_t a, int64_t b, int64_t *sum, int mapno) { int64_t c, temp; } FPGA2.mc send_to_bridge(b); c = a * const1; recv_from_bridge(&temp); *sum = temp+mult; void fpga2() { int64_t a, d; } recv_from_bridge(&a); d = a/const2; send_to_bridge(d); Makefile MAPFILES = FPGA1.mc FPGA2.mc PRIMARY = FPGA1.mc SECONDARY = FPGA2.mc CHIP2 = FPGA2.mc a multiply add b FPGA 1 FPGA 2 22 sum divide

23 FPGA Mapping in VIVA TM By changing the attributes one can specify where an object is to be located 23

24 Platform Mapping FPGA-FPGA data transfer & synchronization Program Hardware FPGA 1 FPGA 2 Software FPGA 3 FPGA 4 24

25 FPGA-FPGA Data Transfer in SRC FPGA1.mc void fpga1(int64_t a, b, c, *d) { } FPGA2.mc send_to_bridge(a, b, c); computation1 recv_from_bridge(d); void fpga2() { int64_t a,b,c,d; a b c d FPGA 1 FPGA 2 computation computation 2 } recv_from_bridge(&a, &b, &c); computation2 send_to_bridge(d); 25

26 FPGA-FPGA Data Transfer in SRC 32 words 64 FIFO 64 bits FIFO 32 words bits Bridge Port 26

27 FPGA-FPGA Data Transfer in VIVA TM 27 Special partitioning objects placed between the modules to be synthesized automatically map the relevant lines between the FPGAs. For designs mapped over several FPGAs: The system description must include those FPGAs over which the design is to be mapped,

28 Platform Mapping Use of Internal and External Memories Program Hardware OCM FPGA 1 FPGA 2 Software SM FPGA 3 OCM On-Chip Memory LM Local Memory SM Shared Memory LM 28 FPGA 4

29 Using On-Chip Memory (OCM) in SRC void sum(int64_t a[], int *c, int mapno) { BANK_A_ALLOC(AL, int64_t, SIZE); FPGA ocm_a [SIZE]; int i; cm2obm_0(al, a, bytelength); AL[] SM (OBM) 64 OCM ocm_a[] wait_server_0(); for(i=0; i<size; i++) { c 32 computations ocm_a[i] = AL[i]; } for(i=0; i<size; i++) { tmp = ocm_a[i] + tmp; } } 29

30 Using On-Chip Memory (OCM) in VIVA TM Special Objects under the Memory Subsystem of the library allows the programmer to use the on chip memory of the Xilinx Virtex II chip 30

31 Platform Mapping I/O Program Hardware OCM FPGA 1 FPGA 2 Software SRC SM StarBridge 31 FPGA 3 LM FPGA 4

32 Run Time Reconfiguration in SRC Program in C or Fortran Main program Function_1(a, d, e) Function_2(d, e, f) Function_1 Macro_1(a, b, c) Macro_2(b, d) Macro_2(c, e) Function_2 Macro_3(s, t) Macro_1(n, b) Macro_4(t, k) 32 FPGA contents after the Function_1 call FPGA b Macro_2 d a Macro_1 c Macro_2 e

33 Run-time Reconfiguration in VIVA TM 33 Reconfigurati on is possible by using the spawn object. By specifying the FileName attribute a VIVA executable (.vex file) or a VIVA project can be loaded onto the same or a different FPGA.

34 Ideal Program Entry Function Program Entry 34

35 Actual Program Entry Preferred Architectures Function SW/HW Partitioning Use of FPGA Resources (multipliers, µp cores) Sequence of Run-time Reconfigurations Program Entry FPGA Mapping Data Transfers & Synchronization SW/HW Interface Use of Internal and External Memories 35

36 Evolution and the current status of tools Not implemented Manual Entry µp-fpga Partitioning FPGA-FPGA Partitioning µp-fpga Data Transfer Compiler Automated SRC Star Bridge and other vendors FPGA-FPGA Data Transfer Computation-Data transfer Overlapping Choosing component version

37 Library Development - SRC LLL (ASM) HLL (C, Fortran) HLL (C, Fortran) µp system FPGA system HDL (VHDL, Verilog) Library Developer 37 HLL (C, Fortran) HLL (C, Fortran) Application Programmer

38 Library Development - StarBridge HLL, LLL (C++, ASM) GDF (Viva) GDF (Viva) µp system FPGA system HDL (VHDL, Verilog) Library Developer 38 GDF (Viva) GDF (Viva) Application Programmer

39 Debugging & Verification 39

40 CoreFireTM FPGA Application Debugger 40

41 Corefire Simulation Insert Debug Modules During Design Editing Step Through Design Using Data Flow One Step = One Module View Value and Status of Each Debug Module Waveform or Table of Values Read and Write Directly to Registers Read and Write Directly to Memory 41

42 X86 System in VIVA TM The FileIn Object as it appears when the x86 system is loaded 42

43 X86 System in VIVA TM FileIn object as it appears when the FPGA system description is loaded. 43

44 Debugging in VIVA TM Data can be viewed with the help of widgets, which are basically input and output horns placed in a worksheet. Various display options are available to view data, options to include the kind of view desired by the viewer and the data viewed can be switched between HEX or INT. 44

45 MAP Board Execution Application Subroutine For MAP User Logic Wrapper ComList Code Code MAP Runtime Library MAP Board On-board Memory Control Processor DMA Engine ComList Processor Registers & Flags User FPGAs User Logic Macro Logic Macro Logic Macro Logic Data & Flags Macro Logic 45

46 MAP Emulator + DFG Simulator Application Subroutine For MAP User Logic Wrapper ComList Code Code MAP Runtime Library Emulator On-board Memory Control Processor DMA Engine ComList Processor Registers & Flags User FPGAs User Logic Macro C Code Macro C Code Macro C Code Data & Flags Macro C Code 46

47 MAP Emulator + Verilog Simulator Application Subroutine For MAP User Logic Wrapper ComList Code Code MAP Runtime Library Emulator VCS On-board Memory Control Processor DMA Engine ComList Processor Registers & Flags User FPGAs User Logic Macro Verilog Macro Verilog Macro Verilog Data & Flags Macro Verilog 47

48 Summary Program Entry Program entry model SRC Star Bridge Xtreme DSP Corefire HLL GDF HLL, GDF HLL, GDF Programming languages HLL C, Fortran Matlab Java GDF - VIVA IIADL Simulink Corefire Application Builder HDL VHDL, Verilog EDIF VHDL, Verilog? 48

49 Summary Partitioning & Data Transfer FPGA Mapping FPGA-FPGA Data Transfer SRC Star Bridge Xtreme DSP Corefire Separate HLL functions send-to bridge, recv-from bridge macro System attributes of objects special data transfer objects such as PE1 =>PE2_50 Separate design sheets interface library components Separate design sheets interface library components 49

50 Summary Synchronization SRC Star Bridge Xtreme DSP Corefire Implicit Explicit: Go-Done- Busy-Wait Explicit: done, empty, full, etc. Implicit 50

51 Summary Run-time reconfiguration Run-time reconfiguration SRC Star Bridge Xtreme DSP sequence of MAP function calls using spawn objects associated with VIVA executables Corefire?? 51

52 Summary Use of Internal Resources Using internal component s of FPGA SRC Star Bridge Xtreme DSP Corefire block RAM s Arrays defined inside MAP functions Special objects under the memory subsystem Memory block sets Memory operators and resources multipliers Library functions Objects under arithmetic subsystem Math block sets Modules in math library 52

53 Summary Data Types Data types SRC Star Bridge Xtreme DSP Corefire Unsigned integers 8,16,32,64 bits bits - 32, 64 bits Signed integers 8,16,32,64 bits bits - 8 bits Fixed point - fix16, fix32 signed, unsigned variable size Floating point Single & double precision Arrays 8, 16, 32, 64 bit User defined types 32-bit single precision vectors of bits 64-bit double precision - yes user-defined precision options - 32-bit single precision - - Complex bit signed integers, float - 53

54 Summary Libraries Libraries SRC Star Bridge Xtreme DSP Corefire arithmetic yes yes yes yes logic yes yes yes yes storage implicit yes yes yes memory allocation and access yes yes yes yes control yes yes yes - data transfer yes yes yes yes debugging and profiling yes yes yes yes DSP - - yes yes communication - - yes - User defined components Macros in VHDL or Verilog Objects in VIVA Black boxes in Simulink Macros in Corefire 54

55 Summary Debugging & Verification Debugging and verification SRC Star Bridge Xtreme DSP Corefire software emulation yes yes yes no HDL simulation yes no yes no 55

56 Summary Third Party Tools & I/O Use of external tools logic synthesis SRC Star Bridge Xtreme DSP Corefire Synplicity Synplify Pro - Synplicity Synplify Pro, Mentor Graphics Leonardo Spectrum, Xilinx XST MAP, PAR Xilinx ISE Xilinx ISE Xilinx ISE Xilinx ISE µp compilation Intel - Matlab - Schematic capture HDL simulation Components of the environment Input and output - VIVA Simulink Application Builder VCS - ModelSim - Editors, compiler, DFG behavioral simulator, VCS HDL simulator standard HLL i/o functions, files VIVA (program entry, compiler, debugging and verification, execution environment) Simulink program entry, HDL simulator, synthesis compiler, place and route tool - Application builder, debugger widgets, files files, block sets files, waveforms, tables 56

57 SRC Star Bridge Sashisu Bajracharya (GMU) Esmail Chitalwala (GWU) Esam El-Araby (GWU) Miaoqing Huang (GWU) Allen Michalski (USC) Nghi Nguyen (GMU) Proshanta Saha (GWU) Nandkishore Sastry (GMU) Chang Shu (GMU) Mohamed Taher (GWU) Acknowledgements 57