m5 x86_se spec2k status (64-bit)

64-bit results

as of 2 December 2009

The following results assume the following patches have been applied.

Patches I'm working on cleaning up and getting upstream:

Patches that won't be merged, but I am using temporarily until the proper fix is ready:

Times given are running x86_64 compiled spec2k benchmarks (with compiler option -O3 -msse3 -funroll-all-loops -ffast-math -static) under m5.opt using config se.py on a cluster of 3.46GHz Pentium D machines.

Runs Correctly to Completion

* means it is currently being re-run

benchmark	native	valgrind	m5 atomic nocache	m5 cache
perlbmk.makerand	0:00:00.67	0:00:34.7	0:53:48	0:58:52
gcc.expr	0:00:04.68	0:02:42	4:53:12	5:33:00
gcc.integrate	0:00:05.42	0:02:28	4:50:33	5:52:55
perlbmk.perfect	0:00:13.0	0:11:38	14:32:05	15:53:58
gzip.log	0:00:13.2	0:06:08	19:39:50	20:24:41
eon.rushmeier	0:00:17.8	0:25:34	33:55:28	31:58:31
eon.cook	0:00:22.6	0:33:04	43:02:39	40:51:53
gcc.166	0:00:23.2	0:07:26	16:37:58	20:38:33
gcc.scilab	0:00:23.6	0:14:37	25:55:48	29:23:15
gzip.random	0:00:23.9	0:11:25	37:24:45	39:12:09
gzip.graphic	0:00:29.9	0:20:30	49:39:03	51:34:21
gzip.source	0:00:30.8	0:17:53	45:33:05	47:23:02
eon.kajiya	0:00:34.8	0:35:09	57:24:17	55:21:29
bzip2.program	0:00:41.1	0:25:14	63:17:39	67:00:34
gcc.200	0:00:41.4	0:26:53	45:23:22	55:14:04
bzip2.source	0:00:42.0	0:35:09	51:19:57	54:53:50
bzip2.graphic	0:00:52.7	0:26:20	73:38:12	77:03:02
gzip.program	0:00:52.9	0:35:40	97:02:04	101:04:40
equake	0:01:11	0:21:02	70:59:11	62:41:01
crafty	0:01:12	1:12:02	100:15:47	104:47:26
art.110	0:01:12	0:11:07	21:25:23	30:28:02
gap	0:01:14	1:14:20	127:28:52	135:42:02
art.470	0:01:16	0:12:27	28:40:49	33:41:38
vpr.route	0:01:26	0:18:22	45:55:53	51:26:07
vpr.place	0:01:30	0:47:55	64:46:57	69:32:58
wupwise	0:01:46	1:53:56	199:15:28	209:10:31
galgel	0:01:58	0:58:37	160:04:16	184:52:55
mesa	0:01:58	2:29:23	170:31:53	159:05:59
lucas	0:02:09	1:09:00	147:01:15	132:26:48
ammp	0:02:52	1:15:45	163:42:04	191:34:04
parser	0:03:18	1:23:16	195:55:04	210:34:43
applu	0:03:21	4:03:32	199:21:55	225:47:46
twolf	0:03:28	1:51:58	199:34:14	219:33:07
mgrid	0:03:49	1:23:09	188:02:09	216:43:12
swim	0:03:51	0:45:07	114:36:35	139:04:44
facerec	0:04:10	1:04:52	152:53:58	168:03:23
apsi	0:04:47	1:43:36	224:01:41	221:47:29
mcf	0:05:24	0:17:20	32:17:02	83:32:14
fma3d	0:05:35	1:45:22	191:11:24	194:44:38
sixtrack	0:18:24	2:09:41	278:34:06	258:02:56

Benchmark fails

benchmark	time native	time valgrind	problem
perlbmk.diffmail			Double free in libc (happens on real HW too)
perlbmk.535			Double free in libc (happens on real HW too)
perlbmk.704			Double free in libc (happens on real HW too)
perlbmk.850			Double free in libc (happens on real HW too)
perlbmk.957			Double free in libc (happens on real HW too)
vortex.1	0	0.5	benchmark issue (doesn't run on real hardware either)
vortex.2	0	0.5	benchmark issue (doesn't run on real hardware either)
vortex.3	0	0.5	benchmark issue (doesn't run on real hardware either)

System Calls Missing (cause benchmark to halt)

Missing syscall	benchmark

System Calls Ignored (may cause correctness issues)

Missing syscall	benchmark
fcntl	perlbmk
futex	perlbmk
ioctl 0x5401	applu,apsi,crafty,facerec,fma3d,galgel,gap, lucas,mgrid,sixtrack,swim
ioctl 0x5405	gap
rt_sigaction	crafty,gap,gcc
setrlimit	gcc

Missing Floating Point/SSE instructions

Missing instruction	purpose	benchmark
fldcw_Mw	set rounding at glibc init	all
fnstcw_Mw	set rounding at glibc init	all
fadd	glibc sin/cos	eon,equake,lucas,sixtrack
fldpi	glibc sin/cos	eon,equake,lucas,sixtrack
fprem1	glibc sin/cos	eon,equake,lucas,sixtrack
fsincos	glibc sin/cos	apsi,eon,equake,fma3d,mesa
fxch	glibc sin/cos	eon,equake,lucas,sixtrack
fldenv	?	vpr.place,facerec
fnstenv	?	vpr.place,facerec
fnstsw	?	apsi,eon,equake,fma3d,lucas,mesa,sixtrack
fwait	?	applu,apsi,facerec,fma3d,galgel,lucas, mgrid,sixtrack,swim,wupwise
prefetch_nta	not needed for correctness	bzip2,gzip,vpr

Number of Instructions Retired

Gathered with "instr_completed:nbogus" counter on same Pentium D machine described earlier.

For more complete results, see Appendix C.10 in my thesis

benchmark	native	m5 simple nocache	valgrind
perlbmk.makerand	1,090,919,227	1,090,879,129	1,090,746,089
gcc.expr	7,350,887,801	7,257,774,662	7,258,023,131
gcc.integrate	7,698,302,743	7,598,617,570	7,597,927,209
perlbmk.perfect	19,654,889,034	19,649,264,066	19,674,125,598
gcc.166	26,258,578,150	26,053,572,133	26,053,249,578
gzip.log	27,720,223,414	27,629,578,439	27,630,555,769
art.110	37,684,130,154	37,684,106,361	37,684,089,804
gcc.scilab	39,085,872,433	38,718,233,041	38,719,744,344
art.470	41,815,575,277	41,815,549,206	41,814,960,116
eon.rushmeier	46,652,449,332	46,652,438,765	46,652,447,077
mcf	47,178,238,767	47,178,770,487	47,178,758,942
gzip.random	50,716,078,217	50,552,564,398	50,553,545,097
eon.cook	59,432,883,084	59,432,871,622	59,432,880,124
gzip.source	63,638,496,739	63,533,923,887	63,534,804,993
vpr.route	65,842,168,801	65,842,101,031	65,842,410,972
gzip.graphic	66,140,686,787	65,984,284,025	65,985,226,242
gcc.200	69,752,973,526	69,333,015,398	69,350,744,008
bzip2.source	75,737,059,115	75,736,212,867	75,737,065,461
eon.kajiya	79,548,196,338	79,548,182,772	79,548,189,789
vpr.place	91,801,882,750	91,627,577,007	91,801,833,351
equake	91,831,665,346	91,831,629,328	91,831,292,111
bzip2.program	92,195,189,138	92,194,260,068	92,195,239,731
bzip2.graphic	104,716,089,604	104,715,201,159	104,716,114,878
gzip.program	134,301,541,033	134,183,019,555	134,184,027,716
crafty	140,491,641,621	140,491,608,144	140,491,506,813
gap	183,443,821,679	183,443,733,395	183,443,755,451
lucas	205,651,052,365	205,650,963,195	205,650,990,335
swim	211,145,979,309	211,145,850,745	211,145,887,898
mesa	225,141,182,114	225,141,105,441	225,141,115,104
facerec	249,466,728,521	249,465,506,605	249,433,555,885
fma3d	252,621,825,649	252,621,687,157	252,621,712,799
parser	263,269,230,283	263,269,185,444	263,218,164,789
galgel	265,315,494,177	265,315,409,019	265,319,397,124
ammp	282,273,753,920	282,273,684,014	282,273,805,462
twolf	294,395,392,631	294,395,327,575	294,395,331,989
mgrid	317,902,282,935	317,901,442,889	317,901,782,490
applu	329,640,061,785	329,639,906,447	329,639,978,210
apsi	335,998,339,268	335,998,752,850	335,998,224,351
wupwise	360,553,449,666	360,553,370,094	360,553,381,385
sixtrack	542,751,559,882	542,751,311,580	542,751,677,787
perlbmk.535
perlbmk.704
perlbmk.850
perlbmk.957
perlbmk.diffmail
vortex.1
vortex.2
vortex.3

Back to other m5 status pages