Originally posted by: SystemAdmin

[kalman said:]

Hi all,
is Cplex available/optimized to run on CellBe processors as present on IBM BladeCenter QS22?

Regards

#CPOptimizer
#DecisionOptimization

Originally posted by: SystemAdmin

[EdKlotz said:]

> is Cplex available/optimized to run on CellBe processors as present on IBM BladeCenter QS22?

A complete list of supported platforms for CPLEX appears at

http://www.ilog.com/products/cplex/product/platforms.cfm

While I don't know much about the CellBe processor, it is not listed
explicitly at the above URL, and I don't believe it shares the same
instruction set with any of the chips associated with the supported
platforms.  So, CPLEX has definitely not been optimized to run
on CellBe processors, and I doubt any existing ports will run
on it.  Also, the CellBe appears to be a massively parallel  chip
designed primarily for graphics, which differs from the optimization
tasks CPLEX performs.  So, unless you were looking to invoke a
large number of separate optimizations using CPLEX on individual threads,
you probably could find processors better suited to good CPLEX performance.

                                                                                  Ed
#CPOptimizer
#DecisionOptimization

Originally posted by: SystemAdmin

[kalman said:]

I do not full agree with your vision, sure CellBE does not have same instruction of others
chips but still C/C++ compilers supports CellBE processor (after all x86 processors and powerpc
processors both supported by Cplex do not share same instructions too).
About the fact that CellBE is suited only for graphical application, I do not agree with you again.
May be the following lecture can enlight some your doubts: 

http://www.lbl.gov/Science-Articles/Archive/sabl/2006/Jul/CellProcessorPotential.pdf

take a look in particular at:

6. Dense Matrix-Matrix Multiply
7. Sparse Matrix Vector Multiply


Some data on GEMM for large matrix operations:

        Double Precision:
            IA64:    5.4 Gflop/s
            AMD64: 4.0 Gflop/s
            Cell:    14.0 Gflop/s

        Single Precision:
            IA64:    3.0 Gflop/s
            AMD64: 7.8 Gflop/s
            Cell:    204.7 Gflop/s  <=====  40x boost on average<br />


#CPOptimizer
#DecisionOptimization