This preview shows page 1. Sign up to view the full content.
Unformatted text preview: 11
An t idi ag on a ... ...
l A S(imax ,0) -5.5 -9 ... al on g S(i,j-1) + eog m -9 S(i-1,j) + eog } Max H(i-1,j) + e C S(0,jmax) Computation of Cell Score V(i,j-1) + e S(imax,jmax ) S( , ji-1 + 1) } b( Su ) i,j } Scalable Systolic Arrays with InStream Control
n n n n n Replace complex controller with special control data in data stream Control data signals state transitions and triggers complex actions Complex-controller performance with simple-controller overhead Fully automated and scalable across any number of coupled FPGAs 850, 650, and 450 PEs per FPGA for NW, SW, and ND, respectively, all @ 125 MHz
From Prev. FPGA Input Buffer
PE(1) Go PE(2) PE(N)
Processing Element Output Buffer To Next FPGA FIFO Input FIFO Interface Hi-1 + -1 -20 Init. Hi (PEi) Hi -20 + > Hi Init. Vi(PEi) Init. Si(PEi) Vi -1 Si + Output FIFO Interface Done FIFO S i-1 + > +10 -8 QC
|| Query Character Reset(L,PEi) > + Si SRAM (Database) Load(L,PEi) DSi-1 Si DSi SC
4 Character Stream Maxi-1 > Fr. Max 4 Maxi > Max Needleman-Wunsch(NW)/Needle-Distance(ND) Systolic Array Smith-Waterman(SW) Systolic Array 9 What is NovoG?
q Growing computational demands in many science and engineering domains becoming principle bottleneck Scalable RC systems (e.g. Novo-G) uniquely capable of both high performance and low energy, cooling, TCO
n n n q n Goals: Investigate, develop, evaluate, & showcase:
Most powerful RC machine ever fielded for research Innovative suite of productivity tools for app development Impactful set of scalable kernels/apps in key science areas n Theme
q Novo-G is an RC-centric machine (not CPUs with accelerators!)
n n n Features FPGA/RAM coupling (4.25 or 8.5 GB in 3 banks coupled to each FPGA) Features FPGA/FPGA coupling (up to 8 coupled; e.g. systolic array, virtual FPGA) CPUs and GPUs serve in supporting role (e.g. I/O, preprocessing, postprocessing) n Emphases
q Performance (system), Productivity (concepts/tools), Impact (apps) 10 NovoG Machine
1 head-node server (1U) with: 2 Xeon E5520 2.26 GHz quad-core CPUs 24GB ECC DDR3, 3 x 1TB SATA2 24 compute servers (4U), each with: Xeon E5520 quad-core CPU 192 FPG 6GB ECC DDR3...
View Full Document
This note was uploaded on 03/27/2012 for the course EEL 4930 taught by Professor Staff during the Spring '08 term at University of Florida.
- Spring '08