urjlew@ecsvax.uncecs.edu (Rostyk Lewyckyj) (01/07/89)
In article <352@loligo.fsu.edu>, mccalpin@loligo.fsu.edu (John McCalpin) writes: > > ... . An interesting example (which > I referred to in my paper in SUPERCOMPUTING a few months back) is the > 1000x1000 LINPACK benchmark test. The matrix is fairly poorly > conditioned, so requires some care in 32-bit mode. I include a sample > of my results below. The numbers scale so that the absolute value > of the exponent of the RMS error is about the number of significant > digits of accuracy in the solution. > > machine precision RMS error in solution (*) > --------------------------------------------------------------------- > Cyber 205 / ETA-10 32-bit 2.22521256e-01 > IBM 3081 32-bit 2.37465184e-03 > IEEE standard 32-bit 2.82104476e-04 > --------------------------------------------------------------------- > Cyber 205 / ETA-10 64-bit 1.32111221e-08 > Cray X/MP 64-bit 2.47078473e-11 > IEEE standard 64-bit 2.27274978e-13 > --------------------------------------------------------------------- > Cyber 205/ETA-10 128-bit 1.60755733e-22 > Cray X/MP 128-bit 4.15861230e-26 > > NOTES: > (*) The solution vector consists of 1000 identical elements = 1.0 > (1) The IEEE standard was run on a Sun 3/280, which had passed > the PARANOIA benchmark test. Here at the Academic Computing Services for the University of North Carolina at Chapel Hill we have various IBMs and a CONVEX C220. Also our researchers use the Cray X/MP at Pittsburgh and theIBM 3090 at Cornell. Therefore it would be very interesting for us if you or anyone else could please post or point to published results for an IBM 3081 or preferably 3090, running under the MVS/XA operating system, using the VS FORTRAN v1.3 compiler, and version 2 libraries, and also the CONVEX. I would like to see the results for 32, 64, and 128 bit reals. i.e. REAL*4, REAL*8 and REAL*16. This would allow comparisons with the Cray X/MP results which you have posted here. I would also like to know the Paranoia results for the above three machines as well as any comparisons of the quality of the mathematical libraries provided with the FORTRAN language systems . In the case of the IBM I am looking for comparisons against the new version 2 libraries not the older version. Please post anything you may have in this area and also send me copies by mail. -tnx. ----------------------------------------------- Reply-To: Rostyslaw Jarema Lewyckyj urjlew@ecsvax.UUCP , urjlew@tucc.bitnet or urjlew@tucc.tucc.edu (ARPA,SURA,NSF etc. internet) tel. (919)-962-9107