m5 alpha_se spec2k status (64-bit)

as of 24 September 2009

64-bit results

Times given are running Alpha compiled spec2k benchmarks (with compiler option -O2) under m5.opt using config se.py on a cluster of 3.46GHz Pentium D machines. The "native" result is on a 500MHz EV6.

Run Correctly to Completion

benchmark native simple nocache simple slowdown OoO cache
perlbmk.makerand 0:00:07.1 0:23:09 196x 6:57:44
gcc.expr 0:00:20.5 1:33:20 273x 39:28:16
gcc.integrate 0:00:20.8 1:36:42 279x 41:22:55
perlbmk.perfect 1:20 4:49:59 217x
gcc.166 1:22 5:13:42 242x
gzip.log 1:15 6:16:20 301x
perlbmk.diffmail 1:33 7:07:47 276x
art.110 4:26 8:08:28 110x
gcc.scilab 1:44 8:11:27 284x
art.470 5:03 8:51:15 105x
mcf 16:17 9:04:58 33x
eon.rushmeier 1:44 9:32:53 331x
perlbmk.535 1:55 9:46:41 334x
perlbmk.704 2:10 11:44:23 325x
gzip.source 2:31 12:56:51 309x
eon.cook 2:20 13:07:53 338x
gzip.random 2:31 13:34:50 324x
gcc.200 3:07 14:27:11 278x
vpr.route 6:25 15:47:11 148x
bzip2.source 3:42 17:07:12 278x
eon.kajiya 17:34:35
gzip.graphic 18:03:23
perlbmk.957 19:25:29
bzip2.program 22:34:37
gzip.program 23:53:45
equake 26:20:24
bzip2.graphic 26:36:04
lucas 33:02:13
crafty 36:58:47
swim 41:35:53
gap 48:09:00
perlbmk.850 49:04:13
mesa 51:57:47
wupwise 58:28:58
galgel 62:51:58
fma3d 69:39:19
twolf 69:56:13
parser 71:50:36
apsi 72:08:42
applu 76:53:25
sixtrack 76:58:54
ammp 82:47:49
mgrid 83:53:24

Benchmark is broken with modern compilers

vortex.*

Benchmark fails

benchmark Error message possible cause
vpr.place Error in try_place various FP warnings
facerec different number of matches various FP warnings

System Call Missing/Ignored

benchmarkMissing syscall
eon, gap, perltimes
gcc.*, perl, craftyrt_sigaction
gcc.*getrlimit

Floating point warnings encountered

Number of Instructions Retired

Gathered with "instr_completed:nbogus" counter on same Pentium D machine described earlier.
* indicates benchmark not run correctly

For more complete results, see Appendix C.1 in my thesis

benchmark m5 simple nocache m5 cache OoO qemu
perlbmk.makerand 1,852,370,276 1,852,370,275
gcc.expr 10,526,937,311 10,526,937,327
gcc.integrate 11,131,144,748 11,131,144,747
perlbmk.perfect 28,646,299,027
gcc.166 38,549,904,548
gzip.log 48,291,602,124
art.110 64,796,606,371
gcc.scilab 54,995,349,608
art.470 71,264,936,399
eon.rushmeier 66,721,584,534
mcf 58,173,013,914
gzip.random 96,314,195,503
eon.cook 93,166,297,706
gzip.source 96,622,791,539
vpr.route 91,649,020,376
gzip.graphic 122,633,036,370
gcc.200 98,083,090,701
bzip2.source 119,407,527,861
eon.kajiya 118,887,092,853
vpr.place *3,686,999,837* *3,686,999,836*
equake 184,571,385,450
bzip2.program 157,826,413,107
bzip2.graphic 188,304,399,124
gzip.program *
crafty 202,098,026,258
gap 282,645,790,314
lucas 250,361,400,189
swim 314,735,601,038
mesa 356,140,991,620
facerec *386,008,212,913*
fma3d 489,141,392,542
parser 457,326,609,655
galgel 445,002,059,655
ammp 589,337,650,371
twolf 471,369,355,199
mgrid 674,760,591,777
applu 593,502,909,954
apsi 530,748,529,384
wupwise 441,495,740,707
sixtrack 660,175,855,578
perlbmk.535 71,270,497,774
perlbmk.704 75,243,363,852
perlbmk.850 144,811,615,781
perlbmk.957 124,614,368,257
perlbmk.diffmail 43,465,983,649
vortex.1 *259,885*
vortex.2 *259,647*
vortex.3 *259,843*
Back to other m5 status pages