200.sixtrack Datasets profile vs. Reference Dataset The following are the profiles for the 200.sixtrack benchmark. For more details about our profile development and dataset reduction methodology, refer to the paper by AJ KleinOsowski and David J. Lilja, "MinneSPEC: A New SPEC Benchmark Workload for Simulation−Based Computer Architecture Research", Computer Architecture Letters, Volume 1, June 2002. This paper is available in electronic form at http://www.arctic.umn.edu/~lilja/minnespec/index.html
http:// www.arctic.umn.edu
Function level execution profile at optimization level O0 The following table contains function execution profiles and goodness−of−fit chi−squared statistic values for the train, test, and large reduced (LgRed) datasets as compared to the full SPEC reference datasets. Note: The medium (MdRed) and small (SmRed) datasets are not available for this benchmark. This data was gathered with the hiprof profiling utility. *90% Conf = Critical value of the chi−squared statistic at the 90 percent confidence level. Numbers in the Ref, Train, Test, and LgRed columns are the percent of overall execution time spent in the stated function (in the Function column). Numbers in the Train Chi, Test Chi, and LgRed Chi are the terms of the chi−squared statistic for the stated function (in the function column).
Function thin6d_ umlauf_ phasad_ Sum
Ref
Train
98.50 1.10 0.20
92.90 5.30 0.90
99.80 Ref
99.10 Train
Train Chi 0.32 16.04 2.45 18.80 Train Chi
90% Confidence level (3 entries) = 4.605
Test 6.20 70.80 12.20 89.20 Test
Test Chi 86.49 4416.45 720.00
LgRed
5222.94 Test Chi
63.10 LgRed
0.00 53.60 9.50
LgRed Chi 98.50 2505.68 432.45 3036.63 LgRed Chi
Function level execution profile at optimization level O1 The following table contains function execution profiles and goodness−of−fit chi−squared statistic values for the train, test, and large reduced (LgRed) datasets as compared to the full SPEC reference datasets. Note: The medium (MdRed) and small (SmRed) datasets are not available for this benchmark. This data was gathered with the hiprof profiling utility. *90% Conf = Critical value of the chi−squared statistic at the 90 percent confidence level. Numbers in the Ref, Train, Test, and LgRed columns are the percent of overall execution time spent in the stated function (in the Function column). Numbers in the Train Chi, Test Chi, and LgRed Chi are the terms of the chi−squared statistic for the stated function (in the function column).
Function thin6d_ umlauf_ phasad_ Sum
Ref
Train
98.40 1.10 0.20
92.70 5.10 0.80
99.70 Ref
98.60 Train
Train Chi 0.33 14.55 1.80 16.68 Train Chi
90% Confidence level (3 entries) = 4.605
Test 6.00 66.60 10.30 82.90 Test
Test Chi 86.77 3900.23 510.05
LgRed
4497.04 Test Chi
47.50 LgRed
0.00 40.90 6.60
LgRed Chi 98.40 1440.04 204.80 1743.24 LgRed Chi
Function level execution profile at optimization level O2 The following table contains function execution profiles and goodness−of−fit chi−squared statistic values for the train, test, and large reduced (LgRed) datasets as compared to the full SPEC reference datasets. Note: The medium (MdRed) and small (SmRed) datasets are not available for this benchmark. This data was gathered with the hiprof profiling utility. *90% Conf = Critical value of the chi−squared statistic at the 90 percent confidence level. Numbers in the Ref, Train, Test, and LgRed columns are the percent of overall execution time spent in the stated function (in the Function column). Numbers in the Train Chi, Test Chi, and LgRed Chi are the terms of the chi−squared statistic for the stated function (in the function column).
Function thin6d_ umlauf_ phasad_ __write __ftruncat e __read Sum
Ref
Train 90.60 4.90 0.90 0.50
Train Chi 0.54 13.13 2.45 1.60
97.90 1.10 0.20 0.10 0.10 0.10
0.40 0.40
99.50 Ref
97.70 Train
LgRed
4.60 49.30 9.70 4.10
Test Chi 88.92 2112.04 451.25 160.00
0.00 17.80 3.40 11.40
LgRed Chi 97.90 253.54 51.20 1276.90
0.90 0.90
3.80 3.50
136.90 115.60
11.40 11.80
1276.90 1368.90
19.52 Train Chi
75.00 Test
3064.70 Test Chi
55.80 LgRed
4325.34 LgRed Chi
90% Confidence level (6 entries) = 9.236
Test
Function level execution profile at optimization level O3 The following table contains function execution profiles and goodness−of−fit chi−squared statistic values for the train, test, and large reduced (LgRed) datasets as compared to the full SPEC reference datasets. Note: The medium (MdRed) and small (SmRed) datasets are not available for this benchmark. This data was gathered with the hiprof profiling utility. *90% Conf = Critical value of the chi−squared statistic at the 90 percent confidence level. Numbers in the Ref, Train, Test, and LgRed columns are the percent of overall execution time spent in the stated function (in the Function column). Numbers in the Train Chi, Test Chi, and LgRed Chi are the terms of the chi−squared statistic for the stated function (in the function column).
Function thin6d_ umlauf_ phasad_ __write __ftruncat e __read atan dainf_ Sum
Ref
Train 87.30 6.70 1.30 0.60
Train Chi 0.99 18.03 3.33 2.50
97.10 1.50 0.30 0.10 0.10 0.10 0.10 0.10
0.50 0.50 0.30 0.20
99.40 Ref
97.40 Train
LgRed
3.30 50.70 9.60 3.90
Test Chi 90.61 1613.76 288.30 144.40
1.60 1.60 0.40 0.10
3.80 4.40 2.30 2.00
136.90 184.90 48.40 36.10
12.00 11.80 0.80
1416.10 1368.90 4.90 0.10
28.55 Train Chi
80.00 Test
2543.37 Test Chi
60.60 LgRed
4754.53 LgRed Chi
90% Confidence level (8 entries) = 12.017
Test
0.00 19.50 3.70 12.80
LgRed Chi 97.10 216.00 38.53 1612.90
Function level execution profile at optimization level O4 The following table contains function execution profiles and goodness−of−fit chi−squared statistic values for the train, test, and large reduced (LgRed) datasets as compared to the full SPEC reference datasets. Note: The medium (MdRed) and small (SmRed) datasets are not available for this benchmark. This data was gathered with the hiprof profiling utility. *90% Conf = Critical value of the chi−squared statistic at the 90 percent confidence level. Numbers in the Ref, Train, Test, and LgRed columns are the percent of overall execution time spent in the stated function (in the Function column). Numbers in the Train Chi, Test Chi, and LgRed Chi are the terms of the chi−squared statistic for the stated function (in the function column).
Function thin6d_ umlauf_ phasad_ __write __read __ftruncate atan dainf_ Sum
Ref
Train
96.90 1.60 0.30 0.10 0.10 0.10 0.10 0.10
86.50 7.10 1.30 0.60 0.50 0.50 0.30 0.30
99.30 Ref
97.10 Train
Train Chi 1.12 18.91 3.33 2.50 1.60 1.60 0.40 0.40 29.86 Train Chi
90% Confidence level (8 entries) = 12.017
Test 3.20 51.90 9.80 4.30 4.00 3.80 1.90 2.30 81.20 Test
Test Chi 90.61 1581.31 300.83 176.40 152.10 136.90 32.40 48.40
LgRed
2518.95 Test Chi
58.80 LgRed
0.00 20.20 3.50 12.30 11.00 11.10 0.70
LgRed Chi 96.90 216.23 34.13 1488.40 1188.10 1210.00 3.60 0.10 4237.46 LgRed Chi
Instruction Mix profile at optimization level o0 The following table contains instruction mix breakdown and goodness−of−fit chi−squared statistic values for the train, test, large reduced (LgRed) datasets, as compared to the full SPEC dataset. Note: The medium (MdRed) and small (SmRed) datasets are not available for this benchmark. This data was gathered with the sim−profile simulator for the SimpleScalar suite. *90% Conf = Critical value of the chi−squared statistic at the 90 percent confidence level. Numbers in the Ref, Train, Test, and LgRed columns are the percent of overall instructions of the stated instruction type (in the Inst Type column). Numbers in the Train Chi, Test Chi, and LgRed Chi columns are the terms of the chi−squared statistic for the stated instruction type (in the Inst Type column).
200.sixtrack O0 Program Inst Type load store unconditional branch conditional branch int computation fp computation trap Sum
Ref
Train 54.19 5.29
Train Chi 0.00 0.00
54.01 5.40 0.01 0.72 28.46 11.40 0.00
0.03 0.78 28.58 11.12 0.00
0.04 0.00 0.00 0.01 0.00
100.00 Ref
99.99 Train
Lgred
53.60 7.22
Test Chi 0.00 0.61
50.85 7.37
Lgred Chi 0.18 0.72
0.37 2.28 27.08 9.45 0.00
12.96 3.38 0.07 0.33 0.00
0.63 3.21 29.03 8.88 0.00
38.44 8.61 0.01 0.56 0.00
0.06 100.00 Train Test Chi
17.36 Test Chi
99.97 Lgred
48.52 Lgred Chi
90% Confidence level (7 entries) = 10.645
Test
Instruction Mix profile at optimization level o1 The following table contains instruction mix breakdown and goodness−of−fit chi−squared statistic values for the train, test, large reduced (LgRed) datasets, as compared to the full SPEC dataset. Note: The medium (MdRed) and small (SmRed) datasets are not available for this benchmark. This data was gathered with the sim−profile simulator for the SimpleScalar suite. *90% Conf = Critical value of the chi−squared statistic at the 90 percent confidence level. Numbers in the Ref, Train, Test, and LgRed columns are the percent of overall instructions of the stated instruction type (in the Inst Type column). Numbers in the Train Chi, Test Chi, and LgRed Chi columns are the terms of the chi−squared statistic for the stated instruction type (in the Inst Type column).
200.sixtrack O1 Program Inst Type load store unconditional branch conditional branch int computation fp computation trap Sum
Ref
Train 26.64 10.97
Train Chi 0.00 0.00
26.57 11.01 0.03 1.95 31.38 29.06 0.00
0.07 2.09 31.88 28.35 0.00
100.00 Ref
100.00 Train
90% Confidence level (7 entries) = 10.645
Test
Lgred
27.85 10.37
Test Chi 0.06 0.04
26.40 10.05
Lgred Chi 0.00 0.08
0.05 0.01 0.01 0.02 0.00
0.63 4.22 39.70 17.23 0.00
12.00 2.64 2.21 4.82 0.00
1.33 6.03 43.08 13.06 0.00
56.33 8.54 4.36 8.81 0.00
0.09 Train Chi
100.00 Test
21.76 Test Chi
99.95 Lgred
78.13 Lgred Chi
Instruction Mix profile at optimization level o2 The following table contains instruction mix breakdown and goodness−of−fit chi−squared statistic values for the train, test, large reduced (LgRed) datasets, as compared to the full SPEC dataset. Note: The medium (MdRed) and small (SmRed) datasets are not available for this benchmark. This data was gathered with the sim−profile simulator for the SimpleScalar suite. *90% Conf = Critical value of the chi−squared statistic at the 90 percent confidence level. Numbers in the Ref, Train, Test, and LgRed columns are the percent of overall instructions of the stated instruction type (in the Inst Type column). Numbers in the Train Chi, Test Chi, and LgRed Chi columns are the terms of the chi−squared statistic for the stated instruction type (in the Inst Type column).
200.sixtrack O2 Program Inst Type
Ref
Train 15.07 4.87
Train Chi 0.01 0.00
load store unconditional branch conditional branch int computation fp computation trap
14.76 4.78 0.07 4.09 13.46 62.83 0.00
0.14 4.34 14.47 61.10 0.00
Sum
99.99 Ref
99.99 Train
90% Confidence level (7 entries) = 10.645
Test
Lgred
19.67 6.24
Test Chi 1.63 0.45
20.60 7.34
Lgred Chi 2.31 1.37
0.07 0.02 0.08 0.05 0.00
1.26 8.04 29.77 35.00 0.00
20.23 3.81 19.76 12.33 0.00
2.14 9.52 39.02 21.29 0.01
61.21 7.21 48.54 27.46 0.00
0.22 Train Chi
99.98 Test
58.21 Test Chi
99.92 Lgred
148.11 Lgred Chi
Instruction Mix profile at optimization level o3 The following table contains instruction mix breakdown and goodness−of−fit chi−squared statistic values for the train, test, large reduced (LgRed) datasets, as compared to the full SPEC dataset. Note: The medium (MdRed) and small (SmRed) datasets are not available for this benchmark. This data was gathered with the sim−profile simulator for the SimpleScalar suite. *90% Conf = Critical value of the chi−squared statistic at the 90 percent confidence level. Numbers in the Ref, Train, Test, and LgRed columns are the percent of overall instructions of the stated instruction type (in the Inst Type column). Numbers in the Train Chi, Test Chi, and LgRed Chi columns are the terms of the chi−squared statistic for the stated instruction type (in the Inst Type column).
200.sixtrack O3 Program Inst Type
Ref
Train 18.07 5.27
Train Chi 0.00 0.00
load store unconditional branch conditional branch int computation fp computation trap
17.88 5.19 0.05 2.42 8.13 66.08 0.00
0.12 2.75 9.45 64.10 0.00
Sum
99.75 Ref
99.76 Train
90% Confidence level (7 entries) = 10.645
Test
Lgred
20.81 6.42
Test Chi 0.48 0.29
21.42 7.50
Lgred Chi 0.70 1.03
0.10 0.05 0.21 0.06 0.00
1.20 7.43 28.39 35.71 0.00
26.45 10.37 50.49 13.96 0.00
2.17 9.03 38.20 21.58 0.00
89.89 18.05 111.22 29.97 0.00
0.42 Train Chi
99.96 Test
102.04 Test Chi
99.90 Lgred
250.86 Lgred Chi
Instruction Mix profile at optimization level o4 The following table contains instruction mix breakdown and goodness−of−fit chi−squared statistic values for the train, test, large reduced (LgRed) datasets, as compared to the full SPEC dataset. Note: The medium (MdRed) and small (SmRed) datasets are not available for this benchmark. This data was gathered with the sim−profile simulator for the SimpleScalar suite. *90% Conf = Critical value of the chi−squared statistic at the 90 percent confidence level. Numbers in the Ref, Train, Test, and LgRed columns are the percent of overall instructions of the stated instruction type (in the Inst Type column). Numbers in the Train Chi, Test Chi, and LgRed Chi columns are the terms of the chi−squared statistic for the stated instruction type (in the Inst Type column).
200.sixtrack O4 Program Inst Type
Ref
Train 18.07 5.27
Train Chi 0.00 0.00
load store unconditional branch conditional branch int computation fp computation trap
17.88 5.19 0.05 2.42 8.13 66.08 0.00
0.12 2.75 9.45 64.10 0.00
Sum
99.75 Ref
99.76 Train
90% Confidence level (7 entries) = 10.645
Test
Lgred
20.80 6.43
Test Chi 0.48 0.30
21.41 7.49
Lgred Chi 0.70 1.02
0.10 0.05 0.21 0.06 0.00
1.19 7.43 28.41 35.71 0.00
25.99 10.37 50.59 13.96 0.00
2.17 9.03 38.22 21.58 0.01
89.89 18.05 111.37 29.97 0.00
0.42 Train Chi
99.98 Test
101.68 Test Chi
99.91 Lgred
250.99 Lgred Chi
Cache profile The following chart shows level 1 data cache miss rates for the ref, train, test, and LgRed datasets. Note: the medium (MdRed) and small (SmRed) reduced input sets are not available for this benchmark. This data was gathered with the sim−cache simulator from the SimpleScalar suite. Miss rate is stated as the ratio of level 1 misses to total level 1 accesses.
Sixtrack 0.0300
Miss Rate
0.0250 0.0200 ref train
0.0150
test lgred
0.0100 0.0050 0.0000 128k
256k
512k
1024k
Data L1 Cache Sizes
2048k
4096k
Instruction Counts for all Datasets The following table shows the instruction counts and estimated simulation time for the reference (Ref), train, test, and large reduced (LgRed) datasets. Note: the medium (MdRed) and small (SmRed) reduced input sets are not available for this benchmark. Instruction counts are from the simulated benchmark, compiled at optimization level O0 and run with each input dataset. Estimated simulation times are calculated using a 45,000 instructions per second factor. This factor was determined by observing the simulation rate of a simulator similar to sim−outorder, run on a machine similar to the SPEC 2000 reference machine (a 333 Mhz Sparc).
Ref
Train
Test
LgRed
(in millions) Simulation Time
2777359
577952
30850
4111
(in hours)
17144.1
3567.6
190.4
25.3
Instruction Count
