FLOPS benchmark code
The FLOPS floating-point benchmark in Forth
Benchmark results
The URL http://hpwww.epfl.ch/bench/bench.FAQ.html contains interesting benchmark source code pointers. I ported
the "C" FLOPS mark by Al Aburto (aburto@nosc.mil) to Forth, and intend to have a go at some of the other benchmarks
mentioned there.
| Flops.c is a 'c' program which attempts to estimate your systems
| floating-point 'MFLOPS' rating for the FADD, FSUB, FMUL, and FDIV
| operations based on specific 'instruction mixes' (discussed below).
| The program provides an estimate of PEAK MFLOPS performance by making
| maximal use of register variables with minimal interaction with main
| memory. The execution loops are all small so that they will fit in
| any cache. Flops.c can be used along with Linpack and the Livermore
| kernels (which exersize memory much more extensively) to gain further
| insight into the limits of system performance. The flops.c execution
| modules also include various percent weightings of FDIV's (from 0% to
| 25% FDIV's) so that the range of performance can be obtained when
| using FDIV's. FDIV's, being computationally more intensive than
| FADD's or FMUL's, can impact performance considerably on some systems.
There are no obvious 'holes' in this code that Forth could exploit (none I can find at least :-)
The Forth FLOPS benchmark shows the quality of your compiler's FPU code as compared to what an optimizing "C" compiler
would give you. Results for a large number of C compilers on a variety of systems are available in the .tbl files.
Intel Pentium P54C 166 MHz, 48 MB, iForth 1.11, NT 4.0
FORTH> main
FLOPS Forth Program (Double Precision), V2.0 14 Jan 2001
Module Error RunTime MFLOPS
(usec)
1 5.6843E-13 0.5685 24.6262
2 -7.7639E-14 0.3992 17.5329
3 -9.5260E-15 0.4090 41.5648
4 3.9895E-14 0.3430 43.7318
5 2.4366E-14 0.8352 34.7201
6 -3.4120E-15 0.6158 47.0970
7 -5.1273E-11 1.0180 11.7878
8 3.0031E-14 0.6170 48.6224
Iterations = 4001600
NullTime (usec) = 0.0362
MFLOPS(1) = 21.6194
MFLOPS(2) = 22.2109
MFLOPS(3) = 33.1329
MFLOPS(4) = 45.8496
The MFLOPS(1) figure puts iForth in the following company (flops_1.tbl):
System OS, Compiler CPU/FPU FPU MFLOPS(1) REF
(MHz)
### ---------------- -------------------- ----------- ----- --------- ----
113 HP 9000/705 HP-UX 9.03 PA-RISC7000 35 23.7545 79
114 SGI Indigo2 Irix 4.0.5 R4000/R4010 100 23.7409 61
115 SGI Indy PC Irix 5.1.1 R4000/R4010 100 23.7053 60
116 HP 9000/705 HP-UX 9.05 PA-RISC7000 35 23.6565 81
117 IBM RS/6000 370 AIX 3.2.3 Power RISC 62 23.5727 37
118 Pentium P5-133 Windows 95 Pentium P5 133 23.5111 129
119 HP 9000/720 HP-UX 9.01 PA-RISC 50 22.898 46
120 IBM RS/6000 980 -------------------- Power RISC 62 22.6487 14
121 DEC 3000/400 AXP OpenVMS V1.5 DEC 21064 133 22.3026 52
122 HP 9000/735 HP-UX 9.01 PA-RISC7100 99 22.1247 73
--------------------------------------------------------------------------
NT 4.0, iForth 1.11 Pentium P5 166 21.6194 xx
--------------------------------------------------------------------------
123 DEC 3000/500 AXP OSF/1 T1.3 DEC 21064 150 21.3285 38
124 Pentium P5-133 MS DOS 6.22 Pentium P5 133 21.1732 128
125 Pentium P5-133 MS DOS 6.22 Pentium P5 133 21.1221 128
126 DEC 3000/500 Digital UNIX V3.2D DEC 21064 150 21.1088 110
127 Pentium P5-120 Windows 95 Pentium P5 120 21.0953 125
128 HP 9000/720 HP-UX 8.07 PA-RISC 50 20.1258 30
129 DEC 3000/500 AXP OSF/1 T1.2-2 (R5) DEC 21064 150 20.0218 25
130 HP 9000/720 HP-UX 8.0.5 PA-RISC 50 19.7063 28
142 PowerMac 8100/80 PAIX PowerPC 601 80 17.51 75
free counter
