clinpack.c ROLL'd DP MFLOPS results are included below.

The program (clinpack.c) and latest Double Precision (DP) table of results (clinpack.dpr for DP 'ROLL' option and clinpack.dpu for DP 'UNROLL' option) are available via anonymous ftp from '' in directory 'pub/aburto'. The 'marlin' host number is:

I'd appreciate any new results of any kind (new machines, compilers, compiler options). I will also periodically post results to 'comp.benchmarks'. Send results to: I will also send clinpack.c (that I'm using) and results to Jack Dongarra.

Please note that these results may not agree with flops.c or Livermore Loops MFLOPS because the programs are different (different instruction mix and different weightings ==> different performance results and characteristics). Also this is a 'C' translation of the FORTRAN Linpack and the 'C' results may not agree with the FORTRAN results due to code generation differences.

Results as of 03 Aug 1993.

NOTES: ----- A00 Sun C 2.0.1 A01 cc -DUNIX -fast -O4 -Bstatic -Qoption iropt -l9 A02 cc -DUNIX -fast -O4 -Bstatic A03 cc -DUNIX -fast -O4 -Bstatic -dalign A04 cc -DUNIX -fast -Bstatic A05 cc -DUNIX -fast A06 cc -DUNIX -O4 A07 cc -DUNIX -O A08 cc -DUNIX

B00 gcc 2.1 B01 gcc -DUNIX -O2 -funroll-loops -fomit-frame-pointer -fstrength-reduce B02 gcc -DUNIX -O2 -funroll-loops -fomit-frame-pointer B03 gcc -DUNIX -O2 -funroll-loops B04 gcc -DUNIX -O2 -funroll-all-loops B05 gcc -DUNIX -O2 B06 gcc -DUNIX -O2 -fomit-frame-pointer B07 gcc -DUNIX -O2 -fexpensive-optimizations B08 gcc -DUNIX -O2 -finline-functions B09 gcc -DUNIX -m486 -O2 B10 gcc -DUNIX -O B11 gcc -DUNIX -m486 -O B12 gcc -DUNIX -m486 B13 gcc -DUNIX

C00 Sun cc (bundled) (-DROLL -DDP) C01 cc -dalign -O4 -Bstatic -Qoption iropt -l9 -lm C02 cc -DUNIX -O4 -Bstatic C03 cc -O4 C04 cc -DUNIX_Old

D00 gcc 2.2.2 D01 gcc -DUNIX -O5 -funroll-loops D02 gcc -DUNIX -O5 -funroll-all-loops D03 gcc -DUNIX -O9 D04 gcc -DUNIX -O5 D05 gcc -DUNIX -O4 D06 gcc -DUNIX -O3 D07 gcc -DUNIX -O2 D08 gcc -DUNIX -O1 D09 gcc -DUNIX -m486 -O2 -funroll-loops D10 gcc -DUNIX -m486 -O2 D11 gcc -DUNIX -m486 -O D12 gcc -DUNIX D13 gcc -DUNIX -m486 D14 gcc -DUNIX -m486 -O2 -funroll-loops -finline-functions

E00 gcc 2.3.3 E01 gcc -DUNIX -O2 -fexpensive-optimizations -fstrength-reduce -fomit-frame-pointer E02 gcc -DUNIX -O2 -funroll-loops -fstrength-reduce -fomit-frame-pointer

F00 Sun C 1.0 F01 cc -DUNIX -fast -O4

000 Various compilers and options 001 cc 3.0, cc -O3 -h ivdep -DFORTRAN_SEC -DSP -DROLL Note that 'Single Precision' on the Cray is 64 bits. 002 cc -DUNIX -O4 -non-shared -migrate, with KAP 003 cc +O3 +OS -Wl,-a,archive -DUNIX -lm +I cc +O3 +OS -Wl,-a,archive -DUNIX -lm +P 004 cc -DUNIX -O4 -non_shared -migrate 005 GEM C, gemcc -DUNIX -DO4 006 cc V3.10, cc -DUNIX -O3 -mips2 007 MIPS cc V2.40, cc -DUNIX -O3 008 Watcom C32 9.5 LA, wcl386 -4r -fpi386 -oneatx-zp4 -dROLL -dDP -dMSC 009 cc -DUNIX -s +O3 -Wg,-All -lm

REF: --- 1 Al Aburto,, average of 5 or more runs as the results jump around a bit, 12 Oct 1992 2 Al Aburto, 3 Bill Broadley,, 13 Oct 1992 4 Charles Grassl,, 14 Oct 1992 5 Scott Hannahs,, 30 Oct 1992 6 Roger Uzun,, 12 Nov 1992 7 Andres Kruse,, 03 Dec 1992 8 Al Aburto,, 08 Jan 1993 9 Roger Uzun,, 23 Jan 1993 10 Al Aburto,, 01 Mar 1993 11 Marc GINGOLD,, 05 Mar 1993 12 Bill Broadley,, 13 Mar 1993 13, 21 Mar 1993 14 Mitch Craig,, 22 Mar 1993 15 Jeff Ungar, ungar@embezzle.Stanford.EDU, 22 Mar 1993 16 Ion CIONCA,, 22 Mar 1993 17 Al Aburto,, 22 Mar 1993 18 Tim Pointing,, 23 Mar 1993 19 Michael Fingerhut,, 23 Mar 1993 20 Jean-Max REYMOND,, 24 Mar 1993 21 Steve Olson,, 24 Mar 1993 22 Jos van Eijndhoven,, 26 Mar 1993 23 Al Aburto,, 27 Mar 1993 24 Sherm Riemenschneider,, 27 Mar 1993 25 Harold Naparst,, 13 Apr 1993 26 Meinhard E. Mayer,, 13 Apr 1993 27 Burkhard Neidecker-Lutz,, 18 May 1993 28 Gail L. Grant, John Shakshober,, 01 Jul 1993 29 Gail L. Grant,, 03 Aug 1993