I need some references of where systolic arrays have actually been used in equipment or instruments. Thanks in advance, Marco ________________________ Marc Reinig UCO/Lick Observatory Laboratory for Adaptive Optics
Need info on Systolic Arrays actually used
Started by ●February 27, 2008
Reply by ●February 27, 20082008-02-27
Marc Reinig wrote:> I need some references of where systolic arrays have actually been used in > equipment or instruments. > > Thanks in advance, > > MarcoTransputer apps? Is there anything of interest at http://www.IntellaSys.net? Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●February 27, 20082008-02-27
Thanks Jerry, Any others, specifically final products built from these chips or other designs? We have made a proposal for an instrument architecture (using a large systolic array of many FPGAs) and the question has been raised, "Well, has anyone used this architecture for a real time control system? Marco ________________________ Marc Reinig UCO/Lick Observatory Laboratory for Adaptive Optics "Jerry Avins" <jya@ieee.org> wrote in message news:kemdnVjNJOGuXVjanZ2dnUVZ_jadnZ2d@rcn.net...> Marc Reinig wrote: >> I need some references of where systolic arrays have actually been used >> in equipment or instruments. >> >> Thanks in advance, >> >> Marco > > Transputer apps? Is there anything of interest at > http://www.IntellaSys.net? > > Jerry > -- > Engineering is the art of making what you want from things you can get. > �����������������������������������������������������������������������
Reply by ●February 27, 20082008-02-27
Marc Reinig wrote:> Thanks Jerry, > > Any others, specifically final products built from these chips or other > designs?I don't know. Here, I'm just a rumor monger. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●February 27, 20082008-02-27
Marc Reinig wrote:> I need some references of where systolic arrays have actually been used in > equipment or instruments.It depends what you consider to be a systolic array. Different people use the term for different architectures. For example, a matrix of transputers was often described as a systolic array, but not by everyone. The Goodyear ASPRO and MPP have been described as systolic arrays, but not by everyone. Steve
Reply by ●February 28, 20082008-02-28
Marc Reinig wrote:> I need some references of where systolic arrays have > actually been used in equipment or instruments.http://www.globalbrand.com/pages/first/websites/paracel/data/pdf/textFinder.pdf http://www.globalbrand.com/pages/first/websites/paracel/data/pdf/genematcher.pdf http://www.timelogic.com/ http://ieeexplore.ieee.org/iel5/7806/21450/00994975.pdf http://www.qi.tnw.tudelft.nl/Publications/PHDTheses/AEpstein/EPSTEIN_2MB.pdf -- glen
Reply by ●February 28, 20082008-02-28
> We have made a proposal for an instrument architecture (using a large > systolic array of many FPGAs) and the question has been raised, > "Well, has anyone used this architecture for a real time control > system?Whats your definition of "systolic array" ? Could you at least narrow it down to SIMD / MIMD ? Starting with IlliacIV there have been all sorts of more or less successfull supercomputers. NASA seemed to like the IlliacIV sufficiently to order the Goodyear MPP. There has been since the 70ies the idea to shrink them with special ICs in size and cost. Initially as coprocessors for workstations ( CLIP4 and PERQ-DAP ) and later embedded controllers ( Mil-DAP: radar ). But very few systems got beyond R&D and here and there a prototype ( like the GAPP from Martin Marietta, probably inspired by the funding for the MPP ). Commercially somewhat viable exceptions: * custom-workstations for medical blood-cell imaging from the 60ies ( Perkin Elmer ) to the early 80ies ( diff3 ). Then killed of because standard hardware got faster and the market was to narrow for special hardware. * data-flow machines ( Bell/AT&T ) for the US Navy, sonar-application. The Clancy novel "hunt for red october" gives an idea of these. But Navy would buy standard hardware now. * the late DAPs seem to have had some commercial "embedded" applications. The DAP-architecture was around 30 years: it probably survived mainly because over the years they had time to optimize their software/compilers. One might think FPGAs had killed the dedicated ICs because the niche is so narrow. But apart from IntellaSys there is now from www.parallax.com the propeller available. A small transputer-revival maybe, perhaps only to scalp venture-capitalists, real applications are still unclear. MfG JRD
Reply by ●February 28, 20082008-02-28
Rafael Deliano wrote:>> We have made a proposal for an instrument architecture (using a large >> systolic array of many FPGAs) and the question has been raised, >> "Well, has anyone used this architecture for a real time control >> system? > > Whats your definition of "systolic array" ? > Could you at least narrow it down to SIMD / MIMD ? > > Starting with IlliacIV there have been all sorts of more or > less successfull supercomputers. NASA seemed to like the IlliacIV > sufficiently to order the Goodyear MPP. > There has been since the 70ies the idea to shrink them with > special ICs in size and cost. Initially as coprocessors for > workstations ( CLIP4 and PERQ-DAP ) and later embedded controllers > ( Mil-DAP: radar ). But very few systems got beyond R&D > and here and there a prototype ( like the GAPP from Martin > Marietta, probably inspired by the funding for the MPP ). > Commercially somewhat viable exceptions: > * custom-workstations for medical blood-cell imaging > from the 60ies ( Perkin Elmer ) to the early 80ies ( diff3 ). > Then killed of because standard hardware got faster and the > market was to narrow for special hardware. > * data-flow machines ( Bell/AT&T ) for the US Navy, > sonar-application. The Clancy novel "hunt for red october" > gives an idea of these. But Navy would buy standard hardware > now. > * the late DAPs seem to have had some commercial "embedded" > applications. The DAP-architecture was around 30 years: it > probably survived mainly because over the years they had > time to optimize their software/compilers. > > One might think FPGAs had killed the dedicated ICs > because the niche is so narrow. But apart from IntellaSys there > is now from www.parallax.com the propeller available. A small > transputer-revival maybe, perhaps only to scalp venture-capitalists, > real applications are still unclear. > > MfG JRDThe MPP lead to the ASPRO, which did go into some defence systems. It was a nice idea for its time, but it doesn't scale to fine geometries. The single instruction source just can't feed millions of ultra simple CPUs at many GHz. Steve
Reply by ●February 28, 20082008-02-28
> Whats your definition of "systolic array" ? > Could you at least narrow it down to SIMD / MIMD ?A systolic array is an array of processing elements, usually orthogonally arranged, where at each beat of the system hart (hence the name systolic) information is transferred to an adjacent processing element, eventually flowing out of some portion of the periphery of the array. Typically also, at each beat, data flows into some portion of the periphery. Between each beat of the heart, there may be many instructions executed by each element. It is not just a multi processor system, say like Iliac, nor a vector array like the CRAY and subsequent large processors. It is often SIMD with all processors executing the same instruction in lock step but on their own data. However, they may also execute different instructions on their own data depending on their own location within the array: MIMD They do not have global memory access. They may each have their own copy of the program being executed. Marco ________________________ Marc Reinig UCO/Lick Observatory Laboratory for Adaptive Optics "Rafael Deliano" <Rafael_DelianoENTFERNEN@t-online.de> wrote in message news:47C68BD3.152FEE0D@t-online.de...>> We have made a proposal for an instrument architecture (using a large >> systolic array of many FPGAs) and the question has been raised, >> "Well, has anyone used this architecture for a real time control >> system?> > Starting with IlliacIV there have been all sorts of more or > less successfull supercomputers. NASA seemed to like the IlliacIV > sufficiently to order the Goodyear MPP. > There has been since the 70ies the idea to shrink them with > special ICs in size and cost. Initially as coprocessors for > workstations ( CLIP4 and PERQ-DAP ) and later embedded controllers > ( Mil-DAP: radar ). But very few systems got beyond R&D > and here and there a prototype ( like the GAPP from Martin > Marietta, probably inspired by the funding for the MPP ). > Commercially somewhat viable exceptions: > * custom-workstations for medical blood-cell imaging > from the 60ies ( Perkin Elmer ) to the early 80ies ( diff3 ). > Then killed of because standard hardware got faster and the > market was to narrow for special hardware. > * data-flow machines ( Bell/AT&T ) for the US Navy, > sonar-application. The Clancy novel "hunt for red october" > gives an idea of these. But Navy would buy standard hardware > now. > * the late DAPs seem to have had some commercial "embedded" > applications. The DAP-architecture was around 30 years: it > probably survived mainly because over the years they had > time to optimize their software/compilers. > > One might think FPGAs had killed the dedicated ICs > because the niche is so narrow. But apart from IntellaSys there > is now from www.parallax.com the propeller available. A small > transputer-revival maybe, perhaps only to scalp venture-capitalists, > real applications are still unclear. > > MfG JRD
Reply by ●February 28, 20082008-02-28
Thanks, That was helpful. Marco ________________________ Marc Reinig UCO/Lick Observatory Laboratory for Adaptive Optics "glen herrmannsfeldt" <gah@ugcs.caltech.edu> wrote in message news:4cGdnQddfZ1t51vanZ2dnUVZ_hynnZ2d@comcast.com...> Marc Reinig wrote: > >> I need some references of where systolic arrays have >> actually been used in equipment or instruments. > > http://www.globalbrand.com/pages/first/websites/paracel/data/pdf/textFinder.pdf > > http://www.globalbrand.com/pages/first/websites/paracel/data/pdf/genematcher.pdf > > http://www.timelogic.com/ > > http://ieeexplore.ieee.org/iel5/7806/21450/00994975.pdf > > http://www.qi.tnw.tudelft.nl/Publications/PHDTheses/AEpstein/EPSTEIN_2MB.pdf > > -- glen >
