Request for review: 8007763: Refactoring: split up compute_generation_free_space() into two functions for class PSAdaptiveSizePolicy
Tao Mao
tao.mao at oracle.com
Tue May 14 22:05:20 UTC 2013
To the open list:
Hi Jon,
Thank you for review.
As Bengt suggested at some point, the "resize young gen at full gc"
project is itemized into several CR's, as follow
Project = Change "resize young gen at full gc" policy [1] + Code
refactoring [2]
where Code refactoring = Wrong initial for max_gc_pause_sec [3] + Split
up compute_generation_free_space [4] + Rename functions [5]
I think you reviewed part [2], most of which is part [4]. That's why it
looks familiar to you.
Also, the three open CR's have dependency: [3] -> [5] -> [1]. So at the
moment, I'd like to get [3] done first.
Hope it explains the situation.
Thanks.
Tao
[1] https://jbs.oracle.com/bugs/browse/JDK-7098155 (Open: Resize the
UseParallelGC young gen at full collections)
[2] https://jbs.oracle.com/bugs/browse/JDK-8007053 (Closed: Refactor
SizePolicy code for consistency across collectors)
[3] https://jbs.oracle.com/bugs/browse/JDK-8007764 (Resolved: Wrong
initialized value of max_gc_pause_sec for an instance of class
AdaptiveSizePolicy)
[4] https://jbs.oracle.com/bugs/browse/JDK-8007763 (Open: Refactoring:
split up compute_generation_free_space() into two functions for class
PSAdaptiveSizePolicy)
[5] https://jbs.oracle.com/bugs/browse/JDK-8007762 (Open: Rename a bunch
of methods in size policy across collectors)
On 5/14/13 2:15 PM, Jon Masamitsu wrote:
> Tao,
>
> I looked over the changes. It looks about the same as I've
> reviewed before. Is there something in particular I should
> look at closely? Otherwise, look good. Thomas I think is
> going to look at it also.
>
> Jon
>
> On 5/6/13 2:46 PM, Tao Mao wrote:
>> 8007763: Refactoring: split up compute_generation_free_space() into
>> two functions for class PSAdaptiveSizePolicy
>> http://bugs.sun.com/view_bug.do?bug_id=8007763
>>
>> webrev:
>> http://cr.openjdk.java.net/~tamao/8007763/webrev.00/
>>
>> changeset:
>> Split up compute_generation_free_space() into two functions:
>> compute_eden_space_size() + compute_old_gen_free_space(), each of
>> which (if needed, e.g.
>> src/share/vm/gc_implementation/parallelScavenge/psScavenge.cpp) can
>> be reused without executing an overhead of the other.
>>
>> The changeset is also related to JDK-8007763 (Rename a bunch of
>> methods in size policy across collectors).
>>
>> test:
>> Because the changeset may affect perfermance, it's gone through
>> rigorous tests. Let me walk through them.
>>
>> (1) passed JPRT test
>>
>> (2) passed gc-test-suite with some known failures in dacapo2006.
>>
>> (3) refworkload performance test
>>
>> baseline: code base from the current repo;
>> test: current repo with my change.
>>
>> ******A) performance score********
>>
>> *No statistically significant difference.*
>>
>> 1) ParallelGC (baseline vs. test)
>> -bash-4.1$ ./compare results_baseline_512m
>> results_splitUpComputeFunction_512m
>> ==============================================================================
>> results_baseline_512m:
>> Benchmark Samples Mean Stdev
>> Geomean Weight
>> jetstream 10 90.89 3.24
>> scimark 10 889.61 9.38
>> specjbb2000 10 314349.25 4101.24
>> specjbb2005 10 166818.55 1726.71
>> specjvm98 10 546.71 4.52
>> ==============================================================================
>> results_splitUpComputeFunction_512m:
>> Benchmark Samples Mean Stdev %Diff P
>> Significant
>> jetstream 10 92.85 3.26 2.15
>> 0.195 *
>> scimark 10 890.62 12.31 0.11
>> 0.838 *
>> specjbb2000 10 316650.11 3988.61 0.73
>> 0.220 *
>> specjbb2005 10 166369.26 1943.79 -0.27
>> 0.592 *
>> specjvm98 10 544.36 5.58 -0.43
>> 0.315 *
>> ==============================================================================
>> * - Not Significant: A non-zero %Diff for the mean could be noise.
>> If the
>> %Diff is 0, an actual difference may still exist. In either
>> case, more
>> samples would be needed to detect an actual difference in
>> sample means.
>>
>> 2) ParallelGC + ParallelOldGC (baseline vs. test)
>> -bash-4.1$ ./compare results_baseline_512m_ParallelOldGC
>> results_splitUpComputeFunction_512m_ParallelOldGC
>> ==============================================================================
>> results_baseline_512m_ParallelOldGC:
>> Benchmark Samples Mean Stdev
>> Geomean Weight
>> jetstream 10 88.12 3.06
>> scimark 10 891.52 9.63
>> specjbb2000 10 310234.98 9587.38
>> specjbb2005 10 165701.45 2141.90
>> specjvm98 10 544.64 5.69
>> ==============================================================================
>> results_splitUpComputeFunction_512m_ParallelOldGC:
>> Benchmark Samples Mean Stdev %Diff P
>> Significant
>> jetstream 10 89.51 2.11 1.58
>> 0.255 *
>> scimark 10 896.48 5.14 0.56
>> 0.173 *
>> specjbb2000 10 316307.14 4639.03 1.96
>> 0.095 *
>> specjbb2005 10 165911.61 1672.12 0.13
>> 0.810 *
>> specjvm98 10 544.02 6.26 -0.11
>> 0.820 *
>> ==============================================================================
>> * - Not Significant: A non-zero %Diff for the mean could be noise.
>> If the
>> %Diff is 0, an actual difference may still exist. In either
>> case, more
>> samples would be needed to detect an actual difference in
>> sample means.
>>
>> ********B) compare gc characteristics********
>>
>> *Note that there are several * yes * (seemingly, indicating
>> statistical significance) in the below comparisons, but they resulted
>> from either that "zero" p-values incorrectly indicated the
>> statistical meaning, or that the test results outperformed the baseline.*
>>
>> 1) ParallelGC
>>
>> i) jetstream (baseline vs. test)
>> -bash-4.1$ ./CompareGCStats.ysr jetstream.baseline jetstream.test
>> jetstream.baseline vs. jetstream.test
>> what count total mean max
>> stddev pvalue sig
>> gen0t(s) 400 5.543 0.01386 0.024
>> 0.0025
>> gen0t(s) 400 5.704 0.01426 0.026
>> 0.0026
>> 0.000% 2.905% 2.88600% 8.333%
>> 4.0000% 0.027
>> GC(s) 400 5.543 0.01386 0.024
>> 0.0025
>> GC(s) 400 5.704 0.01426 0.026
>> 0.0026
>> 0.000% 2.905% 2.88600% 8.333%
>> 4.0000% 0.027
>> alloc(MB) 400 4484.136 11.21034 66.560
>> 21.0282
>> alloc(MB) 400 4484.041 11.21010 66.561
>> 21.0283
>> 0.000% -0.002% -0.00214% 0.002%
>> 0.0005% 1.000
>> promo(MB) 400 -0.092 -0.00023 0.008
>> 0.0018
>> promo(MB) 400 -0.087 -0.00022 0.008
>> 0.0017
>> 0.000% -5.435% -4.34783% 0.000%
>> -5.5556% 0.936
>> used0(MB) 400 4505.526 11.26382 66.561
>> 21.0105
>> used0(MB) 400 4505.525 11.26381 66.561
>> 21.0105
>> 0.000% -0.000% -0.00009% 0.000%
>> 0.0000% 1.000
>> used(MB) 400 4673.370 11.68343 66.988
>> 21.0101
>> used(MB) 400 4673.329 11.68332 66.988
>> 21.0101
>> 0.000% -0.001% -0.00094% 0.000%
>> 0.0000% 1.000
>> commit0(MB) 400 59725.000 149.31250 149.312
>> 0.0000
>> commit0(MB) 400 59725.000 149.31250 149.312
>> 0.0000
>> 0.000% 0.000% 0.00000% 0.000%
>> 0.0000% 0.000 * Yes *
>> commit1(MB) 400 136550.000 341.37500 341.375
>> 0.0000
>> commit1(MB) 400 136550.000 341.37500 341.375
>> 0.0000
>> 0.000% 0.000% 0.00000% 0.000%
>> 0.0000% 0.000 * Yes *
>> commit(MB) 400 196275.000 490.68750 490.688
>> 0.0000
>> commit(MB) 400 196275.000 490.68750 490.688
>> 0.0000
>> 0.000% 0.000% 0.00000% 0.000%
>> 0.0000% 0.000 * Yes *
>>
>> alloc/elapsed_time = 4484.136 MB / 78.477 s = 57.139 MB/s
>> alloc/elapsed_time = 4484.041 MB / 78.454 s = 57.155
>> MB/s 0.028%
>> alloc/tot_cpu_time = 4484.136 MB / 627.816 s = 7.142 MB/s
>> alloc/tot_cpu_time = 4484.041 MB / 627.632 s = 7.144
>> MB/s 0.028%
>> alloc/mut_cpu_time = 4484.136 MB / 583.471 s = 7.685 MB/s
>> alloc/mut_cpu_time = 4484.041 MB / 581.999 s = 7.705
>> MB/s 0.260%
>> promo/elapsed_time = -0.092 MB / 78.477 s = -0.001 MB/s
>> promo/elapsed_time = -0.087 MB / 78.454 s = -0.001 MB/s
>> -0.000%
>> promo/gc0_time = -0.092 MB / 5.543 s = -0.017 MB/s
>> promo/gc0_time = -0.087 MB / 5.704 s = -0.015 MB/s
>> -11.765%
>> gc_seq_load = 44.345 s / 627.816 s = 7.063%
>> gc_seq_load = 45.633 s / 627.632 s = 7.271%
>> 0.208%
>> gc_conc_load = 0.000 s / 627.816 s = 0.000%
>> gc_conc_load = 0.000 s / 627.632 s = 0.000%
>> 0.000%
>> gc_tot_load = 44.345 s / 627.816 s = 7.063%
>> gc_tot_load = 45.633 s / 627.632 s = 7.271%
>> 0.208%
>>
>> ii) specjbb2000 (baseline vs. test)
>> -bash-4.1$ ./CompareGCStats.ysr specjbb2000.baseline specjbb2000.test
>> specjbb2000.baseline vs. specjbb2000.test
>> what count total mean max
>> stddev pvalue sig
>> gen0t(s) 13644 383.408 0.02810 0.107
>> 0.0106
>> gen0t(s) 13818 391.258 0.02832 0.106
>> 0.0109
>> 1.275% 2.047% 0.78292% -0.935%
>> 2.8302% 0.090
>> GC(s) 13644 383.408 0.02810 0.107
>> 0.0106
>> GC(s) 13818 391.258 0.02832 0.106
>> 0.0109
>> 1.275% 2.047% 0.78292% -0.935%
>> 2.8302% 0.090
>> alloc(MB) 13644 1555532.166 114.00851 158.750
>> 26.5987
>> alloc(MB) 13818 1579703.639 114.32216 158.875
>> 26.3181
>> 1.275% 1.554% 0.27511% 0.079%
>> -1.0549% 0.326
>> promo(MB) 13644 100991.271 7.40188 64.133
>> 7.4544
>> promo(MB) 13818 101079.116 7.31503 63.211
>> 7.4739
>> 1.275% 0.087% -1.17335% -1.438%
>> 0.2616% 0.335
>> used0(MB) 13644 1893026.604 138.74425 164.031
>> 17.7641
>> used0(MB) 13818 1918469.816 138.83846 164.062
>> 17.6364
>> 1.275% 1.344% 0.06790% 0.019%
>> -0.7189% 0.659
>> used(MB) 13644 4264934.882 312.58684 478.139
>> 94.4757
>> used(MB) 13818 4296412.687 310.92869 477.621
>> 94.5276
>> 1.275% 0.738% -0.53046% -0.108%
>> 0.0549% 0.146
>> commit0(MB) 13644 1950811.062 142.97941 164.688
>> 11.6897
>> commit0(MB) 13818 1977464.438 143.10786 164.750
>> 11.6320
>> 1.275% 1.366% 0.08984% 0.038%
>> -0.4936% 0.361
>> commit1(MB) 13644 4622894.438 338.82252 341.375
>> 8.3817
>> commit1(MB) 13818 4679087.938 338.62266 341.375
>> 8.6997
>> 1.275% 1.216% -0.05899% 0.000%
>> 3.7940% 0.053
>> commit(MB) 13644 6573705.500 481.80193 506.062
>> 15.0508
>> commit(MB) 13818 6656552.375 481.73052 506.125
>> 15.2806
>> 1.275% 1.260% -0.01482% 0.012%
>> 1.5268% 0.696
>>
>> alloc/elapsed_time = 1555532.166 MB / 6037.435 s = 257.648 MB/s
>> alloc/elapsed_time = 1579703.639 MB / 6037.436 s = 261.651
>> MB/s 1.554%
>> alloc/tot_cpu_time = 1555532.166 MB / 48299.480 s = 32.206 MB/s
>> alloc/tot_cpu_time = 1579703.639 MB / 48299.488 s = 32.706
>> MB/s 1.553%
>> alloc/mut_cpu_time = 1555532.166 MB / 45232.212 s = 34.390 MB/s
>> alloc/mut_cpu_time = 1579703.639 MB / 45169.427 s = 34.973
>> MB/s 1.695%
>> promo/elapsed_time = 100991.271 MB / 6037.435 s = 16.728 MB/s
>> promo/elapsed_time = 101079.116 MB / 6037.436 s = 16.742
>> MB/s 0.084%
>> promo/gc0_time = 100991.271 MB / 383.408 s = 263.404 MB/s
>> promo/gc0_time = 101079.116 MB / 391.258 s = 258.344 MB/s
>> -1.921%
>> gc_seq_load = 3067.268 s / 48299.480 s = 6.351%
>> gc_seq_load = 3130.061 s / 48299.488 s = 6.481%
>> 0.130%
>> gc_conc_load = 0.000 s / 48299.480 s = 0.000%
>> gc_conc_load = 0.000 s / 48299.488 s = 0.000%
>> 0.000%
>> gc_tot_load = 3067.268 s / 48299.480 s = 6.351%
>> gc_tot_load = 3130.061 s / 48299.488 s = 6.481%
>> 0.130%
>>
>> iii) specjbb2005 (baseline vs. test)
>> -bash-4.1$ ./CompareGCStats.ysr specjbb2005.baseline specjbb2005.test
>> specjbb2005.baseline vs. specjbb2005.test
>> what count total mean max
>> stddev pvalue sig
>> gen0t(s) 14364 273.756 0.01906 0.059
>> 0.0041
>> gen0t(s) 14355 275.070 0.01916 0.055
>> 0.0043
>> -0.063% 0.480% 0.52466% -6.780%
>> 4.8780% 0.044
>> GC(s) 14364 273.756 0.01906 0.059
>> 0.0041
>> GC(s) 14355 275.070 0.01916 0.055
>> 0.0043
>> -0.063% 0.480% 0.52466% -6.780%
>> 4.8780% 0.044
>> alloc(MB) 14364 2042796.258 142.21639 160.562
>> 10.4289
>> alloc(MB) 14355 2042158.958 142.26116 160.562
>> 10.4025
>> -0.063% -0.031% 0.03148% 0.000%
>> -0.2531% 0.716
>> promo(MB) 14364 26628.275 1.85382 29.438
>> 2.1697
>> promo(MB) 14355 26538.441 1.84872 29.336
>> 2.1744
>> -0.063% -0.337% -0.27511% -0.346%
>> 0.2166% 0.842
>> used0(MB) 14364 2222232.042 154.70844 165.344
>> 9.5313
>> used0(MB) 14355 2221328.974 154.74253 165.344
>> 9.4992
>> -0.063% -0.041% 0.02203% 0.000%
>> -0.3368% 0.761
>> used(MB) 14364 5370221.314 373.86670 493.481
>> 77.0410
>> used(MB) 14355 5363478.243 373.63136 493.378
>> 77.2201
>> -0.063% -0.126% -0.06295% -0.021%
>> 0.2325% 0.796
>> commit0(MB) 14364 2245588.312 156.33447 165.375
>> 5.0143
>> commit0(MB) 14355 2244726.750 156.37247 165.375
>> 4.8498
>> -0.063% -0.038% 0.02431% 0.000%
>> -3.2806% 0.514
>> commit1(MB) 14364 4715227.812 328.26704 341.375
>> 21.9035
>> commit1(MB) 14355 4712477.062 328.28123 341.375
>> 21.8581
>> -0.063% -0.058% 0.00432% 0.000%
>> -0.2073% 0.956
>> commit(MB) 14364 6960816.125 484.60151 506.750
>> 22.3408
>> commit(MB) 14355 6957203.812 484.65370 506.750
>> 22.2996
>> -0.063% -0.052% 0.01077% 0.000%
>> -0.1844% 0.843
>>
>> alloc/elapsed_time = 2042796.258 MB / 2430.734 s = 840.403 MB/s
>> alloc/elapsed_time = 2042158.958 MB / 2430.973 s = 840.058
>> MB/s -0.041%
>> alloc/tot_cpu_time = 2042796.258 MB / 19445.872 s = 105.050 MB/s
>> alloc/tot_cpu_time = 2042158.958 MB / 19447.784 s = 105.007
>> MB/s -0.041%
>> alloc/mut_cpu_time = 2042796.258 MB / 17255.826 s = 118.383 MB/s
>> alloc/mut_cpu_time = 2042158.958 MB / 17247.226 s = 118.405
>> MB/s 0.019%
>> promo/elapsed_time = 26628.275 MB / 2430.734 s = 10.955 MB/s
>> promo/elapsed_time = 26538.441 MB / 2430.973 s = 10.917 MB/s
>> -0.347%
>> promo/gc0_time = 26628.275 MB / 273.756 s = 97.270 MB/s
>> promo/gc0_time = 26538.441 MB / 275.070 s = 96.479 MB/s
>> -0.813%
>> gc_seq_load = 2190.046 s / 19445.872 s = 11.262%
>> gc_seq_load = 2200.558 s / 19447.784 s = 11.315%
>> 0.053%
>> gc_conc_load = 0.000 s / 19445.872 s = 0.000%
>> gc_conc_load = 0.000 s / 19447.784 s = 0.000%
>> 0.000%
>> gc_tot_load = 2190.046 s / 19445.872 s = 11.262%
>> gc_tot_load = 2200.558 s / 19447.784 s = 11.315%
>> 0.053%
>>
>> iv) specjvm98 (baseline vs. test)
>> -bash-4.1$ ./CompareGCStats.ysr specjvm98.baseline specjvm98.test
>> specjvm98.baseline vs. specjvm98.test
>> what count total mean max
>> stddev pvalue sig
>> gen0t(s) 835 13.128 0.01572 0.034
>> 0.0033
>> gen0t(s) 837 13.341 0.01594 0.032
>> 0.0036
>> 0.240% 1.622% 1.39949% -5.882%
>> 9.0909% 0.193
>> GC(s) 835 13.128 0.01572 0.034
>> 0.0033
>> GC(s) 837 13.341 0.01594 0.032
>> 0.0036
>> 0.240% 1.622% 1.39949% -5.882%
>> 9.0909% 0.193
>> alloc(MB) 835 54436.159 65.19301 165.458
>> 63.5823
>> alloc(MB) 837 54746.996 65.40860 165.087
>> 63.4132
>> 0.240% 0.571% 0.33069% -0.224%
>> -0.2660% 0.945
>> promo(MB) 835 97.071 0.11625 6.733
>> 0.5694
>> promo(MB) 837 93.223 0.11138 6.516
>> 0.5197
>> 0.240% -3.964% -4.18925% -3.223%
>> -8.7285% 0.855
>> used0(MB) 835 55001.033 65.86950 165.938
>> 64.0499
>> used0(MB) 837 55347.981 66.12662 165.203
>> 63.9310
>> 0.240% 0.631% 0.39035% -0.443%
>> -0.1856% 0.935
>> used(MB) 835 56027.885 67.09926 168.742
>> 64.1572
>> used(MB) 837 56375.527 67.35427 168.122
>> 64.0294
>> 0.240% 0.620% 0.38005% -0.367%
>> -0.1992% 0.935
>> commit0(MB) 835 131910.188 157.97627 168.000
>> 9.0675
>> commit0(MB) 837 132027.875 157.73940 167.812
>> 8.8842
>> 0.240% 0.089% -0.14994% -0.112%
>> -2.0215% 0.590
>> commit1(MB) 835 285048.125 341.37500 341.375
>> 0.0000
>> commit1(MB) 837 285730.875 341.37500 341.375
>> 0.0000
>> 0.240% 0.240% 0.00000% 0.000%
>> 0.0000% 0.000 * Yes *
>> commit(MB) 835 416958.312 499.35127 509.375
>> 9.0675
>> commit(MB) 837 417758.750 499.11440 509.188
>> 8.8842
>> 0.240% 0.192% -0.04744% -0.037%
>> -2.0215% 0.590
>>
>> alloc/elapsed_time = 54436.159 MB / 595.330 s = 91.439 MB/s
>> alloc/elapsed_time = 54746.996 MB / 592.522 s = 92.397
>> MB/s 1.048%
>> alloc/tot_cpu_time = 54436.159 MB / 4762.640 s = 11.430 MB/s
>> alloc/tot_cpu_time = 54746.996 MB / 4740.176 s = 11.550
>> MB/s 1.050%
>> alloc/mut_cpu_time = 54436.159 MB / 4657.614 s = 11.688 MB/s
>> alloc/mut_cpu_time = 54746.996 MB / 4633.447 s = 11.816
>> MB/s 1.095%
>> promo/elapsed_time = 97.071 MB / 595.330 s = 0.163 MB/s
>> promo/elapsed_time = 93.223 MB / 592.522 s = 0.157 MB/s
>> -3.681%
>> promo/gc0_time = 97.071 MB / 13.128 s = 7.394 MB/s
>> promo/gc0_time = 93.223 MB / 13.341 s = 6.988 MB/s
>> -5.491%
>> gc_seq_load = 105.026 s / 4762.640 s = 2.205%
>> gc_seq_load = 106.729 s / 4740.176 s = 2.252%
>> 0.047%
>> gc_conc_load = 0.000 s / 4762.640 s = 0.000%
>> gc_conc_load = 0.000 s / 4740.176 s = 0.000%
>> 0.000%
>> gc_tot_load = 105.026 s / 4762.640 s = 2.205%
>> gc_tot_load = 106.729 s / 4740.176 s = 2.252%
>> 0.047%
>>
>> 2) ParallelGC + ParallelOldGC
>>
>> i) jetstream (baseline vs. test)
>> -bash-4.1$ ./CompareGCStats.ysr jetstream.Old.baseline
>> jetstream.Old.test
>> jetstream.Old.baseline vs. jetstream.Old.test
>> what count total mean max
>> stddev pvalue sig
>> gen0t(s) 400 5.307 0.01327 0.024
>> 0.0029
>> gen0t(s) 400 5.297 0.01324 0.022
>> 0.0030
>> 0.000% -0.188% -0.22607% -8.333%
>> 3.4483% 0.886
>> GC(s) 400 5.307 0.01327 0.024
>> 0.0029
>> GC(s) 400 5.297 0.01324 0.022
>> 0.0030
>> 0.000% -0.188% -0.22607% -8.333%
>> 3.4483% 0.886
>> alloc(MB) 400 4484.090 11.21022 66.560
>> 21.0282
>> alloc(MB) 400 4484.341 11.21085 66.561
>> 21.0280
>> 0.000% 0.006% 0.00562% 0.002%
>> -0.0010% 1.000
>> promo(MB) 400 -0.076 -0.00019 0.008
>> 0.0017
>> promo(MB) 400 -0.098 -0.00024 0.008
>> 0.0016
>> 0.000% 28.947% 26.31579% 0.000%
>> -5.8824% 0.669
>> used0(MB) 400 4505.527 11.26382 66.561
>> 21.0105
>> used0(MB) 400 4505.528 11.26382 66.561
>> 21.0105
>> 0.000% 0.000% 0.00000% 0.000%
>> 0.0000% 1.000
>> used(MB) 400 4673.374 11.68344 66.988
>> 21.0101
>> used(MB) 400 4673.354 11.68339 66.988
>> 21.0101
>> 0.000% -0.000% -0.00043% 0.000%
>> 0.0000% 1.000
>> commit0(MB) 400 59725.000 149.31250 149.312
>> 0.0000
>> commit0(MB) 400 59725.000 149.31250 149.312
>> 0.0000
>> 0.000% 0.000% 0.00000% 0.000%
>> 0.0000% 0.000 * Yes *
>> commit1(MB) 400 136550.000 341.37500 341.375
>> 0.0000
>> commit1(MB) 400 136550.000 341.37500 341.375
>> 0.0000
>> 0.000% 0.000% 0.00000% 0.000%
>> 0.0000% 0.000 * Yes *
>> commit(MB) 400 196275.000 490.68750 490.688
>> 0.0000
>> commit(MB) 400 196275.000 490.68750 490.688
>> 0.0000
>> 0.000% 0.000% 0.00000% 0.000%
>> 0.0000% 0.000 * Yes *
>>
>> alloc/elapsed_time = 4484.090 MB / 78.683 s = 56.989 MB/s
>> alloc/elapsed_time = 4484.341 MB / 78.023 s = 57.475
>> MB/s 0.853%
>> alloc/tot_cpu_time = 4484.090 MB / 629.464 s = 7.124 MB/s
>> alloc/tot_cpu_time = 4484.341 MB / 624.184 s = 7.184
>> MB/s 0.842%
>> alloc/mut_cpu_time = 4484.090 MB / 587.007 s = 7.639 MB/s
>> alloc/mut_cpu_time = 4484.341 MB / 581.811 s = 7.708
>> MB/s 0.903%
>> promo/elapsed_time = -0.076 MB / 78.683 s = -0.001 MB/s
>> promo/elapsed_time = -0.098 MB / 78.023 s = -0.001 MB/s
>> -0.000%
>> promo/gc0_time = -0.076 MB / 5.307 s = -0.014 MB/s
>> promo/gc0_time = -0.098 MB / 5.297 s = -0.018 MB/s
>> 28.571%
>> gc_seq_load = 42.457 s / 629.464 s = 6.745%
>> gc_seq_load = 42.373 s / 624.184 s = 6.789%
>> 0.044%
>> gc_conc_load = 0.000 s / 629.464 s = 0.000%
>> gc_conc_load = 0.000 s / 624.184 s = 0.000%
>> 0.000%
>> gc_tot_load = 42.457 s / 629.464 s = 6.745%
>> gc_tot_load = 42.373 s / 624.184 s = 6.789%
>> 0.044%
>>
>> ii) specjbb2000 (baseline vs. test)
>> -bash-4.1$ ./CompareGCStats.ysr specjbb2000.Old.baseline
>> specjbb2000.Old.test
>> specjbb2000.Old.baseline vs. specjbb2000.Old.test
>> what count total mean max
>> stddev pvalue sig
>> gen0t(s) 13558 391.293 0.02886 0.107
>> 0.0108
>> gen0t(s) 13838 393.317 0.02842 0.113
>> 0.0111
>> 2.065% 0.517% -1.52460% 5.607%
>> 2.7778% 0.001 * Yes *
>> GC(s) 13558 391.293 0.02886 0.107
>> 0.0108
>> GC(s) 13838 393.317 0.02842 0.113
>> 0.0111
>> 2.065% 0.517% -1.52460% 5.607%
>> 2.7778% 0.001 * Yes *
>> alloc(MB) 13558 1556503.044 114.80329 158.312
>> 25.9670
>> alloc(MB) 13838 1582650.824 114.36991 158.688
>> 26.3486
>> 2.065% 1.680% -0.37750% 0.238%
>> 1.4696% 0.170
>> promo(MB) 13558 98064.460 7.23296 62.969
>> 7.3507
>> promo(MB) 13838 104324.062 7.53896 64.516
>> 7.6335
>> 2.065% 6.383% 4.23063% 2.457%
>> 3.8473% 0.001 * Yes *
>> used0(MB) 13558 1887550.903 139.22045 164.031
>> 17.3416
>> used0(MB) 13838 1922700.633 138.94353 163.844
>> 17.5826
>> 2.065% 1.862% -0.19891% -0.114%
>> 1.3897% 0.189
>> used(MB) 13558 4222552.180 311.44359 476.567
>> 95.1856
>> used(MB) 13838 4312246.197 311.62351 476.800
>> 94.5717
>> 2.065% 2.124% 0.05777% 0.049%
>> -0.6450% 0.875
>> commit0(MB) 13558 1944158.750 143.39569 164.438
>> 11.3458
>> commit0(MB) 13838 1981508.750 143.19329 164.562
>> 11.4999
>> 2.065% 1.921% -0.14115% 0.075%
>> 1.3582% 0.143
>> commit1(MB) 13558 4593360.438 338.79336 341.375
>> 8.4636
>> commit1(MB) 13838 4688645.562 338.82393 341.375
>> 8.3815
>> 2.065% 2.074% 0.00902% 0.000%
>> -0.9700% 0.764
>> commit(MB) 13558 6537519.188 482.18905 505.812
>> 14.8932
>> commit(MB) 13838 6670154.312 482.01722 505.938
>> 14.9694
>> 2.065% 2.029% -0.03564% 0.025%
>> 0.5116% 0.341
>>
>> alloc/elapsed_time = 1556503.044 MB / 6037.309 s = 257.814 MB/s
>> alloc/elapsed_time = 1582650.824 MB / 6037.014 s = 262.158
>> MB/s 1.685%
>> alloc/tot_cpu_time = 1556503.044 MB / 48298.472 s = 32.227 MB/s
>> alloc/tot_cpu_time = 1582650.824 MB / 48296.112 s = 32.770
>> MB/s 1.685%
>> alloc/mut_cpu_time = 1556503.044 MB / 45168.131 s = 34.460 MB/s
>> alloc/mut_cpu_time = 1582650.824 MB / 45149.577 s = 35.054
>> MB/s 1.724%
>> promo/elapsed_time = 98064.460 MB / 6037.309 s = 16.243 MB/s
>> promo/elapsed_time = 104324.062 MB / 6037.014 s = 17.281
>> MB/s 6.390%
>> promo/gc0_time = 98064.460 MB / 391.293 s = 250.617 MB/s
>> promo/gc0_time = 104324.062 MB / 393.317 s = 265.242
>> MB/s 5.836%
>> gc_seq_load = 3130.341 s / 48298.472 s = 6.481%
>> gc_seq_load = 3146.535 s / 48296.112 s = 6.515%
>> 0.034%
>> gc_conc_load = 0.000 s / 48298.472 s = 0.000%
>> gc_conc_load = 0.000 s / 48296.112 s = 0.000%
>> 0.000%
>> gc_tot_load = 3130.341 s / 48298.472 s = 6.481%
>> gc_tot_load = 3146.535 s / 48296.112 s = 6.515%
>> 0.034%
>>
>> iii) specjbb2005 (baseline vs. test)
>> -bash-4.1$ ./CompareGCStats.ysr specjbb2005.Old.baseline
>> specjbb2005.Old.test
>> specjbb2005.Old.baseline vs. specjbb2005.Old.test
>> what count total mean max
>> stddev pvalue sig
>> gen0t(s) 14350 278.607 0.01942 0.060
>> 0.0044
>> gen0t(s) 14335 276.081 0.01926 0.054
>> 0.0044
>> -0.105% -0.907% -0.82389% -10.000%
>> 0.0000% 0.002 * Yes *
>> GC(s) 14350 278.607 0.01942 0.060
>> 0.0044
>> GC(s) 14335 276.081 0.01926 0.054
>> 0.0044
>> -0.105% -0.907% -0.82389% -10.000%
>> 0.0000% 0.002 * Yes *
>> alloc(MB) 14350 2041460.732 142.26207 160.688
>> 10.4463
>> alloc(MB) 14335 2039406.477 142.26763 160.625
>> 10.4109
>> -0.105% -0.101% 0.00391% -0.039%
>> -0.3389% 0.964
>> promo(MB) 14350 26508.163 1.84726 29.164
>> 2.1662
>> promo(MB) 14335 26516.080 1.84974 29.266
>> 2.1549
>> -0.105% 0.030% 0.13425% 0.350%
>> -0.5217% 0.923
>> used0(MB) 14350 2220358.996 154.72885 165.344
>> 9.5408
>> used0(MB) 14335 2218304.518 154.74744 165.344
>> 9.5016
>> -0.105% -0.093% 0.01201% 0.000%
>> -0.4109% 0.869
>> used(MB) 14350 5362679.672 373.70590 493.516
>> 77.2492
>> used(MB) 14335 5356646.818 373.67609 493.704
>> 77.2386
>> -0.105% -0.112% -0.00798% 0.038%
>> -0.0137% 0.974
>> commit0(MB) 14350 2243692.000 156.35484 165.375
>> 5.0201
>> commit0(MB) 14335 2241662.938 156.37691 165.375
>> 4.8536
>> -0.105% -0.090% 0.01412% 0.000%
>> -3.3167% 0.705
>> commit1(MB) 14350 4711402.312 328.32072 341.375
>> 21.7503
>> commit1(MB) 14335 4704253.000 328.16554 341.375
>> 22.0153
>> -0.105% -0.152% -0.04726% 0.000%
>> 1.2184% 0.548
>> commit(MB) 14350 6955094.312 484.67556 506.750
>> 22.2179
>> commit(MB) 14335 6945915.938 484.54244 506.750
>> 22.4207
>> -0.105% -0.132% -0.02747% 0.000%
>> 0.9128% 0.614
>>
>> alloc/elapsed_time = 2041460.732 MB / 2430.353 s = 839.985 MB/s
>> alloc/elapsed_time = 2039406.477 MB / 2430.681 s = 839.027
>> MB/s -0.114%
>> alloc/tot_cpu_time = 2041460.732 MB / 19442.824 s = 104.998 MB/s
>> alloc/tot_cpu_time = 2039406.477 MB / 19445.448 s = 104.878
>> MB/s -0.114%
>> alloc/mut_cpu_time = 2041460.732 MB / 17213.972 s = 118.593 MB/s
>> alloc/mut_cpu_time = 2039406.477 MB / 17236.797 s = 118.317
>> MB/s -0.233%
>> promo/elapsed_time = 26508.163 MB / 2430.353 s = 10.907 MB/s
>> promo/elapsed_time = 26516.080 MB / 2430.681 s = 10.909
>> MB/s 0.018%
>> promo/gc0_time = 26508.163 MB / 278.607 s = 95.146 MB/s
>> promo/gc0_time = 26516.080 MB / 276.081 s = 96.044
>> MB/s 0.944%
>> gc_seq_load = 2228.852 s / 19442.824 s = 11.464%
>> gc_seq_load = 2208.651 s / 19445.448 s = 11.358%
>> -0.106%
>> gc_conc_load = 0.000 s / 19442.824 s = 0.000%
>> gc_conc_load = 0.000 s / 19445.448 s = 0.000%
>> 0.000%
>> gc_tot_load = 2228.852 s / 19442.824 s = 11.464%
>> gc_tot_load = 2208.651 s / 19445.448 s = 11.358%
>> -0.106%
>>
>> iv) specjvm98 (baseline vs. test)
>> -bash-4.1$ ./CompareGCStats.ysr specjvm98.Old.baseline
>> specjvm98.Old.test
>> specjvm98.Old.baseline vs. specjvm98.Old.test
>> what count total mean max
>> stddev pvalue sig
>> gen0t(s) 836 13.136 0.01571 0.030
>> 0.0033
>> gen0t(s) 836 12.875 0.01540 0.030
>> 0.0036
>> 0.000% -1.987% -1.97327% 0.000%
>> 9.0909% 0.067
>> GC(s) 836 13.136 0.01571 0.030
>> 0.0033
>> GC(s) 836 12.875 0.01540 0.030
>> 0.0036
>> 0.000% -1.987% -1.97327% 0.000%
>> 9.0909% 0.067
>> alloc(MB) 836 54705.075 65.43669 165.404
>> 63.4158
>> alloc(MB) 836 54287.136 64.93677 166.375
>> 63.6853
>> 0.000% -0.764% -0.76398% 0.587%
>> 0.4250% 0.872
>> promo(MB) 836 92.452 0.11059 6.515
>> 0.5261
>> promo(MB) 836 110.159 0.13177 7.023
>> 0.6208
>> 0.000% 19.153% 19.15182% 7.797%
>> 18.0004% 0.452
>> used0(MB) 836 55304.878 66.15416 165.938
>> 63.9514
>> used0(MB) 836 54815.876 65.56923 167.672
>> 64.1150
>> 0.000% -0.884% -0.88419% 1.045%
>> 0.2558% 0.852
>> used(MB) 836 56326.027 67.37563 168.817
>> 64.0444
>> used(MB) 836 55871.066 66.83142 170.165
>> 64.2345
>> 0.000% -0.808% -0.80773% 0.798%
>> 0.2968% 0.862
>> commit0(MB) 836 131983.688 157.87522 167.938
>> 8.6495
>> commit0(MB) 836 132227.062 158.16634 168.500
>> 9.1417
>> 0.000% 0.184% 0.18440% 0.335%
>> 5.6905% 0.504
>> commit1(MB) 836 285389.500 341.37500 341.375
>> 0.0000
>> commit1(MB) 836 285389.500 341.37500 341.375
>> 0.0000
>> 0.000% 0.000% 0.00000% 0.000%
>> 0.0000% 0.000 * Yes *
>> commit(MB) 836 417373.188 499.25022 509.312
>> 8.6495
>> commit(MB) 836 417616.562 499.54134 509.875
>> 9.1417
>> 0.000% 0.058% 0.05831% 0.111%
>> 5.6905% 0.504
>>
>> alloc/elapsed_time = 54705.075 MB / 595.443 s = 91.873 MB/s
>> alloc/elapsed_time = 54287.136 MB / 592.044 s = 91.694 MB/s
>> -0.195%
>> alloc/tot_cpu_time = 54705.075 MB / 4763.544 s = 11.484 MB/s
>> alloc/tot_cpu_time = 54287.136 MB / 4736.352 s = 11.462 MB/s
>> -0.192%
>> alloc/mut_cpu_time = 54705.075 MB / 4658.453 s = 11.743 MB/s
>> alloc/mut_cpu_time = 54287.136 MB / 4633.350 s = 11.717 MB/s
>> -0.221%
>> promo/elapsed_time = 92.452 MB / 595.443 s = 0.155 MB/s
>> promo/elapsed_time = 110.159 MB / 592.044 s = 0.186 MB/s
>> 20.000%
>> promo/gc0_time = 92.452 MB / 13.136 s = 7.038 MB/s
>> promo/gc0_time = 110.159 MB / 12.875 s = 8.556 MB/s
>> 21.569%
>> gc_seq_load = 105.091 s / 4763.544 s = 2.206%
>> gc_seq_load = 103.002 s / 4736.352 s = 2.175%
>> -0.031%
>> gc_conc_load = 0.000 s / 4763.544 s = 0.000%
>> gc_conc_load = 0.000 s / 4736.352 s = 0.000%
>> 0.000%
>> gc_tot_load = 105.091 s / 4763.544 s = 2.206%
>> gc_tot_load = 103.002 s / 4736.352 s = 2.175%
>> -0.031%
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/hotspot-gc-dev/attachments/20130514/8e986026/attachment.htm>
More information about the hotspot-gc-dev
mailing list