1. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia Introduction to CBEA SDK Veselin Dikov Moscow-Bavarian Joint Advanced Student School 19-29. March 2006 Moscow, Russia

2. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia Outline ●Getting started ●SPU Language Extensions ●SPE Library ●SDK Libraries ●Remote Procedure Calls

3. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia Getting started ●Two executables formats: ppu vs. spu  Two different compilers: ppu-gcc and spu-gcc  Check the format of an executable # file ●Makefile headers  Reside in SDK root directory make.header make.footer make.env

4. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia Getting started # Subdirectories ############### DIRS := spu # Target ####################### PROGRAM_ppu:= simple # Local Defines ################ IMPORTS := spu/lib_simple_spu.a \ -lspe # imports the embedded simple_spu # library allows consolidation of # spu program into ppe binary # make.footer ################## include../make.footer # make.footer is in the top of # the SDK # Target ####################### PROGRAMS_spu := simple_spu # created embedded library LIBRARY_embed:= lib_simple_spu.a # Local Defines ################################ IMPORTS = $(SDKLIB_spu)/libc.a # make.footer ################################ include../../make.footer # make.footer is in the top of # the SDK ●Makefile - ppu●Makefile - spu

5. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia Outline ●Getting started ●SPU Language Extensions ●SPE Library ●SDK Libraries ●Remote Procedure Calls

6. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia SPU Language Extensions ●Provides SPU functionality ●Extended RISC instruction set ●Extended Data types: vector ●C/C++ intrinsic routines ●Memory Flow Control (MFC)

7. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia SIMD Vectorization ● vector - Reserved as a C/C++ keyword ●128 bits of size  16 bytes, or  4 32-bit words, or  etc. ●Single instruction acts on the entire vector vector unsigned int vec = (vector unsigned int)(1,2,3,4); vector unsigned int v_ones = (vector unsigned int)(1); vector unsigned int vdest = spu_add(vec, v_ones);

8. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia SIMD Vectorization ●Overridden C/C++ operators  sizeof() (=16 for any vector)  operator=  operator& ●Various intrinsic functions for vectors  Arithmetical – spu_add, spu_mult, etc.  Logical – spu_and, spu_nor, etc.  Comparison – spu_cmpeq  Scalar – spu_extract, spu_insert, spu_promote  Etc.

9. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia MFC routines ●Local Store (LS) vs. Effective Address (EA) ●Data transport via DMA  Up to 16,384 bytes on DMA message  Asynchronous  Part of the instruction set, C/C++ intrinsics mfc_get(&data,addr+16384*i,16384,20,0,0); ●What is __attribute__ ((aligned (128))); ? LS address EA address

10. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia MFC example int main() { … /* the array needs to be aligned on a 128-byte cache line */ data = (int *) malloc(127 + DATASIZE*sizeof(int)); while (((int) data) & 0x7f) ++data; … spe_create_thread(gid,&sample_spu,data,NULL,-1,0); ●ppu main file

11. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia MFC example int main() { … /* the array needs to be aligned on a 128-byte cache line */ data = (int *)malloc_align(DATASIZE*sizeof(int),7); … spe_create_thread(gid,&sample_spu,data,NULL,-1,0); ●ppu main file

12. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia MFC example int databuf [DATASIZE]__attribute__ ((aligned (128))); int main(void * arg) { int *data = &databuf[0]; … mfc_get(data,arg,DATASIZE*sizeof(int),20,0,0); mfc_write_tag_mask(1<<20); mfc_read_tag_status_all(); ●spu main file

13. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia Outline ●Getting started ●SPU Language Extensions ●SPE Library ●SDK Libraries ●Remote Procedure Calls

14. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia ●Provides PPU functionality ●Two sets of functions  Thread management – POSIX like  MFC access functions – access to mailboxes ●Header:

15. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia Thread management ●Threads are organized in thread groups Thread Group 0 Th0Th1Th2ThN Thread Group 1 Th0Th1Th2ThN Thread scheduling: SCHED_RR SCHED_FIFO SCHED_OTHER Priority: RR and FIFO – 1 to 99 OTHER – 0 to 99

16. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia Thread management ●Threads are organized in thread groups spe_gid_t gid; // group handle speid_t speids[1]; // thread handle // Create an SPE group gid = spe_create_group (SCHED_OTHER, 0, 1); if (gid == NULL) exit(-1); // failed // allocate the SPE task speids[0] = spe_create_thread(gid,&sample_spu,0,0,-1,0); if (speids[0] == NULL) exit (-1); // wait for the single SPE to complete spe_wait(speids[0], &status[0], 0);

17. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia Thread management ●spe_get_affinity, spe_set_affinity  8-bit mask specifying SPE units where to start the thread ●spe_get_ls  Gets access to SPU’s local store  Dirty!!  Not supported by all operating systems  Used for RPC communication ●spe_get/set_context, spe_get_ps

18. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia Thread management ●Two ways of communicating with threads  via signals  via mailboxes ●POSIX-like Signals  spe_get_event - check for signals from a thread  spe_kill – send signal to a thread  spe_write_signal – writes a signal through MFC ●MFC mailboxes  spe_read_out_mbox, spe_write_in_mbox  spe_stat_in_mbox, spe_stat_out_mbox, spe_stat_out_intr_mbox

19. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia Outline ●Getting started ●SPU Language Extensions ●SPE Library ●SDK Libraries ●Remote Procedure Calls

20. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia SDK Libraries ●The SDK comes with various applied libraries Library nameShort DescriptionPPESPE C Librarystandard C99 functionality. POSIX.1 functions.xx Audio Resample Library audio resampling functionality for PPE and SPExx Curves and Surfaces Library quadratic and cubic Bezier curves. Biquadric and bicubic Bezier surfaces, and curved point-normal triangles. xx FFT Library1-D FFT and kernel functions for 2-D FFTxx Game Math Library math routines applicable to game performance needsxx Image Libraryroutines for processing images - convolutions and histograms xx Large Matrix Library basic linear algebra routines on large vectors and matrices x Math Librarygeneral purpose math routines tuned to exploit SIMDxx

21. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia SDK Libraries ●The SDK comes with various applied libraries Library nameShort DescriptionPPESPE Matrix Libraryroutines for operations on 4x4 Matrices and quaternionsxx Misc Libraryset of general purpose routines that don’t logically fit within any other xx Multi-Precision Math Library operations on unsigned integer numbers with large number of bits x Noise LibraryPPE 1-,2-, 3-, 4-D noise, Lattice and non-lattice noise, Turbulance xx Oscillator Libraries definition of sound sourcesxx Simulation Library functionality related to the Full-Simulator-- Sync Librarysynchronization primitives, like atomic operations, mutexxx Vector Librarya set of general purpose routines that operate on vectors.xx

22. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia Example typedef union { vector float fv; vector unsigned int uiv; unsigned int ui[4]; float f[4]; } v128; typedef vector float floatvec; … v128 A[4] = {(floatvec)(0.9501,0.8913,0.8214,0.9218), (floatvec)(0.2311,0.7621,0.4447,0.7382), (floatvec)(0.6068,0.4565,0.6154,0.1763), (floatvec)(0.4860,0.0185,0.7919,0.4057)}; v128 invA[4]; v128 x, y; x.fv = (floatvec)(1, 5, 6, 7); y.fv = (floatvec)(0); inverse_matrix4x4(&invA[0].fv, &A[0].fv); madd_vector_matrix(4,4,&invA[0].fv,4,(float *)&x,(float *)&y); ● Solving from Matrix library from LargeMatrix library. Solves: y = A*x + y

23. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia Outline ●Getting started ●SPU Language Extensions ●SPE Library ●SDK Libraries ●Remote Procedure Calls

24. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia Remote Procedure Calls ●Function-Offload Model  SPE threads as services  PPE communicates with them thought RPC calls ●Remote Procedure Calls  stubs

25. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia Remote Procedure Calls ●Preparing stubs  Interface Description Language (IDL)  idl files interface add { import "../stub.h"; const int ARRAY_SIZE = 1000; [sync] idl_id_t do_inv ([in] int array_size, [in, size_is(array_size)] int array_a[], [out, size_is(array_size)] int array_res[]); … }  idl compiler # idl -p ppe_sample.c -s spe_sample.c sample.idl

26. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia Remote Procedure Calls ●Preparing stubs

27. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia Example ●.idl file interface add { import "../stub.h"; const int ARRAY_SIZE = 1000; [sync] idl_id_t do_add ([in] int array_size, [in, size_is(array_size)] int array_a[], [in, size_is(array_size)] int array_b[], [out, size_is(array_size)] int array_c[]); [sync] idl_id_t do_sub ([in] int array_size, [in, size_is(array_size)] int array_a[], [in, size_is(array_size)] int array_b[], [out, size_is(array_size)] int array_c[]); }

28. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia Example ●spe do_add.c file (user file) #include "../stub.h" idl_id_t do_add (int array_size, int array_a[], int array_b[], int array_c[]) { int i; for (i = 0; i < array_size; i++) { array_c[i] = array_a[i] + array_b[i]; } return 0; } idl_id_t do_sub (int array_size, int array_a[], int array_b[], int array_c[]) {... return 0; }

29. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia Example ●ppe main.c file (user file) #include "../stub.h" int array_a[ARRAY_SIZE] __attribute ((aligned(128))); int array_b[ARRAY_SIZE] __attribute ((aligned(128))); int array_add[ARRAY_SIZE] __attribute ((aligned(128))); int main() {... do_add (ARRAY_SIZE, array_a, array_b, array_add);... }

30. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia Conclusion ●Powerful SDK  vector data type  Extensive Matrix routines  Extensive Signal Processing routines ●Game-industry driven  But applicable for scientific work as well

31. Moscow-Bavarian Joint Advanced Student School 19-29 March 2006, Moscow, Russia References ●CBEA-Tutorial.pdf, SDK documentation ●idl.pdf, SDK documentation ●libraries_SDK.pdf, SDK Documentation ●libspe_v1.0.pdf, SDK Documentation ●SPU_language_extensions_v21.pdf  Sony online resources  http://cell.scei.co.jp/pdf/SPU_language_extensions_v21.pdf, 15.03.2006 http://cell.scei.co.jp/pdf/SPU_language_extensions_v21.pdf Thank you for your attention!