Metastones - A C++ program building benchmark

What is a metastone?

A meta stone is the time it takes to build the 'meta' program from the usaf-ask. It represents performance in a C++ program development environment.

Go in to your usaf-ask/meta (v1.6-6, but current is the same) and do this to run the timing on other machines... 

        make
        make clean
        time make

The results

OSCPURAMg++TotalUserSys.IdleMetas/hour
Linux 2.2.15Athlon-1100 TBird128MB2.95.23.473.150.3201035
Linux 2.2.15Athlon-800 TBird128MB2.95.24.554.190.360791
Linux 2.4.17PIII 800256MB2.95.45.475.120.360657
Linux 2.2.13K63-450128MB2.7.2.36.645.820.820542
Linux 2.1.98PentiumII-333128MB***7.87.30.50.0446
Linux 2.2.15K63-450128MB2.95.28.467.170.810.48425
Linux 2.2.1K62-350256MB***9.17.8.90.4395
Linux 2.1.42PentiumII-26664MB2.7.2.110???378
Linux 2.4.2Celeron 633192MB2.95.310.389.500.850.03347
Linux 2.2.8K62-350(@400)64MB2.7.2.310.469.430.990344
Linux 2.2.18PIII-400192MB2.95.210.899.730.840.32331
Linux 2.2.4K6-333(@380)64MB2.7.2.310.979.761.210328
Linux 2.2.8K62-35064MB2.7.2.311.589.981.210311
Linux 2.4.0-pre11Celeron 700196MB2.95.211.7010.030.950.720307
Linux 2.2.4K62-33364MB2.7.2.311.9510.51.450301
Linux 2.2.17K6-2 350128MB2.95.213.8712.601.160.11260
Linux 2.0.14P6-20032MB2.7.2141130257
Linux 1.3.74P6-15032MB2.7.2171430211
Linux 1.3.75P5-16632MB2.7.2221930164
Linux 2.0.33P5-180
overdrive		64MB2.7.2.1272241133
SunOS 5.5.1Sparc Ultra 164MB2.7.2.1271836133
OSF 4.0AlphaServer 4100
400MHz 564		1000MB2.7.2.12810315128
Linux 2.0.33MediaGX/20032MB***333021109
SCO 3.0P6-15032MB2.7.23312516109
SCO 3.0P5-16632MB2.7.23517513103
HP-UX 10.102x120mhz PA-RISC??2.7.240316390
Linux 1.3.72P5-9032MB2.7.247366576
SCO 3.0P5-9032MB2.7.25532111265
NH50002x8810096MB2.7.0574714--63
Linux 2.0.31sa11032MB2.8.156454762
Linux 1.3.66486-100DX416MB2.7.2735281349
Solaris 2.5.1SparcStation 532MB2.7.2.1745611748
NH50008810096MB2.7.08546152442
HP-UX 9.09000/750??2.6.3*866091741
Linux 1.3.25486-50DX232MB2.6.3141107211325
SCO 3.0486-66DX224MB2.7.214385391925
Entries with *** for compiler version were done with an ecgs variant of g++. They are normalized by running the same compiler against a known machine. Currently the actual egcs times are multiplied by 0.75 before being entered into the table. It is a slower compiler. The HP-UX entry is probably too high. The gcc in use did not support the '-g' option which is used on all the other tests.
Last modified: Thu Mar 8 09:37:09 CST 2001