Zhiduo Liu Aaron Severance Satnam Singh Guy Lemieux Accelerator Compiler for the VENICE Vector Processor

This is the VENICE Vector Processor : Complicated

#include "vector.h“ int main() { int A[] = {1,2,3,4,5,6,7,8}; const int data_len = sizeof ( A ); int *va = ( int *) vector_malloc ( data_len ); vector_dma_to_vector ( va, A, data_len ); vector_wait_for_dma (); vector_set_vl ( data_len / sizeof (int) ); vector ( SVW, VADD, va, 42, va ); vector_instr_sync (); // wait for operation to complete vector_dma_to_host ( A, va, data_len ); vector_wait_for_dma (); vector_free (); // deallocate scratchpad malloc } You can program it like this : #include "Accelerator.h" #include "VectorTarget.h" using namespace ParallelArrays; using namespace MicrosoftTargets; int main() { Target *tgtVector = CreateVectorTarget(); int A[] = {1,2,3,4,5,6,7,8}; IPA b = IPA( A, sizeof (A)/sizeof (int)); IPA c = b + 42; tgtVector->ToArray( c, A, sizeof (A)/sizeof (int)); tgtVector->Delete(); } OR like this : Target *tgtVector = CreateMulticoreTarget(); Target *tgtVector = CreateDX9Target();

Assembly Programming : Write Assembly Download to board Compile with Gcc Get Result Doesn’t compile? Result Incorrect? Accelerator Programming : Write in Accelerator Download to board Compile with Microsoft Visual Studio Get Result Compile with Gcc Doesn’t compile? Or result incorrect?

Assembly Programming : 1.Hard to program 2.Long debug cycle 3.Not portable 4.Manual – Not always optimal or correct (wysiwyg) Accelerator Programming : 1.Easy to program 2.Easy to debug 3.Can also target other devices 4.Automated compiler optimizations

LIR Convert To LIR IR Add Intermediates Combine Operations Evaluation Ordering & Reference Counting Evaluation Ordering & Reference Counting Buffer Counting Calculate Buffer Size Allocate & Initialize Memory Transfer Data To Scratchpad Set VL Write Vector Instructions Transfer Result To Host Assign Buffers to Inputs Need Double buffering? LIR Compiler Flow

CPU Benchmark Runtime (seconds) fir2Dfirlifeimgblendmedianmotest Xeon W3690 (3.47GHz) VENICE (V64, 100MHz) Speedup1.0 x1.5 x2.3 x0.4 x3.2 x1.1 x 369x

Thank you !