RFR(S) 8005956 : C2: assert(!def_outside->member(r)) failed: Use of external LRG overlaps the same LRG defined in this block
Vladimir Kozlov
vladimir.kozlov at oracle.com
Mon Jun 3 07:48:16 PDT 2013
Good.
Vladimir
On 6/3/13 5:19 AM, Niclas Adlertz wrote:
> Hi Vladimir,
>
> Thank you.
> Here is a new webrev based on the input:
>
> http://cr.openjdk.java.net/~adlertz/JDK-8005956/webrev02/
>
> Kind Regards,
> Niclas Adlertz
>
>
> On 1 jun 2013, at 04:48, Vladimir Kozlov <Vladimir.Kozlov at oracle.com> wrote:
>
>> On 5/31/13 2:41 PM, Vladimir Kozlov wrote:
>>> On 5/30/13 12:26 PM, Vladimir Kozlov wrote:
>>>> Thanks, Niclas
>>>>
>>>> On 5/30/13 8:29 AM, Niclas Adlertz wrote:
>>>>> Vladimir,
>>>>> thank you for your feedback. I've now added rematerialization support
>>>>> only for reaching definitions to phi nodes that doesn't contain any
>>>>> live range input.
>>>>
>>>> I think you can avoid loop in contains_no_live_range_input() based on
>>>> code in MachNode::rematerialize(). Edge in(0) is control edge so you
>>>> don't need to check it and you can return true if a node has only
>>>> control edge (req() < 2), and it will cover MachTemp and normal loadCon
>>>> nodes cases.
>>>
>>> Niclas,
>>>
>>> Sorry, I am taking back this comment. Having loop is more future proof
>>> and it is general code which could be used for any mach node and not
>>> only for rematerialized. May be it should be moved to MachNode class
>>> (skip 0 (control) edge):
>>
>> Input edge could be Phi, BoxLock and other ideal nodes which are not Mach nodes so in(i)->is_Mach() is not correct.
>>
>> Vladimir
>>
>>>
>>> bool MachNode::contains_no_live_range_input() {
>>> for (uint i = 1; i < req(); ++i) {
>>> if ((in(i) != NULL) && in(i)->is_Mach() &&
>>> in_RegMask(i).is_NotEmpty()) {
>>> return false;
>>> }
>>> }
>>> return true;
>>> }
>>>
>>> Regards,
>>> Vladimir
>>>
>>>> An other case is nodes producing flag. I think we should add
>>>> assert(def->ideal_reg() != Op_RegFlags). Because such nodes should not
>>>> be inputs to Phi (to avoid be separated from user node since other nodes
>>>> in between could affect flags).
>>>> This leaves only case with req() == 2. So you don't need the loop. Also
>>>> in(1) can't be NULL or not Mach (add assert to make sure). So you need
>>>> only def->in_RegMask(1).is_NotEmpty() check.
>>>> Run CTW and other our tests after these changes. I could be mistaking in
>>>> some assumtions.
>>>>
>>>> Also make contains_no_live_range_input() static so it will be local.
>>>>
>>>>> I also changed one input parameter to split_Rematerialize, just to fix
>>>>> the parfait bug 8013830.
>>>>> The parameter is not used when rematerializing reaches to phi nodes,
>>>>> so passing NULL should be fine.
>>>>
>>>> Please, do it as separate fix for parfait bug (it is better for bugs
>>>> bookkeeping). Why you think it is not used?
>>>>
>>>> Thanks,
>>>> Vladimir
>>>>
>>>>>
>>>>> http://cr.openjdk.java.net/~adlertz/JDK-8005956/webrev01/
>>>>>
>>>>> Kind Regards,
>>>>> Niclas Adlertz
>>>>>
>>>>>
>>>>> On 29 maj 2013, at 17:05, Vladimir Kozlov <Vladimir.Kozlov at oracle.com>
>>>>> wrote:
>>>>>
>>>>>> On 5/29/13 1:18 AM, Niclas Adlertz wrote:
>>>>>>> Vladimir,
>>>>>>> Thank you for the additional info!
>>>>>>> However, in this bug it's not a rematerialization of a loadConX
>>>>>>> (X=P,D,etc) node that is causing this to fail, but a
>>>>>>> rematerialization of a negD_reg_reg node (node 530, rematerialized
>>>>>>> to node 1716, with input 1619
>>>>>>> (http://cr.openjdk.java.net/~adlertz/JDK-8005956/log.txt)).
>>>>>>>
>>>>>>> I suggest we remove this faulty optimization, as the performance
>>>>>>> tests show no significant regression.
>>>>>>
>>>>>> It is not faulty optimization for nodes which don't have inputs (like
>>>>>> constants not on sparc). As I said, it reduces LR and may reduce need
>>>>>> for spills. Also some applications may still regress even if few
>>>>>> benchmarks we use are not.
>>>>>>
>>>>>> Vladimir
>>>>>>
>>>>>>>
>>>>>>> Kind Regards,
>>>>>>> Niclas Adlertz
>>>>>>>
>>>>>>>
>>>>>>> On 29 maj 2013, at 00:38, Christian Thalinger
>>>>>>> <christian.thalinger at oracle.com> wrote:
>>>>>>>
>>>>>>>>
>>>>>>>> On May 28, 2013, at 3:08 PM, Vladimir Kozlov
>>>>>>>> <vladimir.kozlov at oracle.com> wrote:
>>>>>>>>
>>>>>>>>> Niclas,
>>>>>>>>>
>>>>>>>>> I attached mails in which we discussed the fix for 7004940.
>>>>>>>>> Removing rematerialization was one option.
>>>>>>>>> I think the original code, to do rematerialization, was added to
>>>>>>>>> reduce live range of constant values (ConP) which did not have any
>>>>>>>>> inputs (root is not mach node). It worked until Christian added
>>>>>>>>> constants load from table - as result constant become node with
>>>>>>>>> input edge (MachConstantBaseNode). So now we have to stretch
>>>>>>>>> either LR of ConstantBase node or LR of loadCon node.
>>>>>>>>> May be we should avoid rematerialization only for such nodes
>>>>>>>>> (loadConP, loadConD etc.)?
>>>>>>>>
>>>>>>>> Some background on MachConstantBaseNode: these nodes are pinned
>>>>>>>> into the first basic block of the method (IIRC). The reason for
>>>>>>>> this is we want to keep the RDPC instruction as close as possible
>>>>>>>> to the constant table base address so that the offset is small and
>>>>>>>> the constant load instruction can be a simple load. Otherwise we
>>>>>>>> need other instructions to calculate the constant table base address.
>>>>>>>>
>>>>>>>> This stretches the live range of the constant table base register
>>>>>>>> quite a bit but note that we only do this on SPARC.
>>>>>>>>
>>>>>>>> -- Chris
>>>>>>>>
>>>>>>>>>
>>>>>>>>> Thanks,
>>>>>>>>> Vladimir
>>>>>>>>>
>>>>>>>>> On 5/24/13 5:32 AM, Niclas Adlertz wrote:
>>>>>>>>>> Hi all.
>>>>>>>>>>
>>>>>>>>>> Problem:
>>>>>>>>>> Hitting the assert: assert(!def_outside->member(r)) failed: Use
>>>>>>>>>> of external LRG overlaps the same LRG defined in this block in
>>>>>>>>>> live.cpp
>>>>>>>>>>
>>>>>>>>>> Background info:
>>>>>>>>>> When a reaching definition for a phi node is rematerialized (due
>>>>>>>>>> to spilling) it's inserted in the end of the preceding block to
>>>>>>>>>> the phi. On rare occasions this will cause the re-calculation of
>>>>>>>>>> the liveness to fail. When phi nodes and their input are united
>>>>>>>>>> in the end of the Chaitin::Split() method they get the same live
>>>>>>>>>> range id.
>>>>>>>>>>
>>>>>>>>>> What happens in this case is that a reaching definition R for a
>>>>>>>>>> phi node P has been rematerialized and put in a preceding block
>>>>>>>>>> B. The input X to R will eventually get the same live range id as
>>>>>>>>>> an earlier instruction Y in B. Look at [0] for a more detailed
>>>>>>>>>> explanation.
>>>>>>>>>>
>>>>>>>>>> Solution:
>>>>>>>>>> Disable re-materialization of reaching definitions for phi nodes
>>>>>>>>>> when spilling. I've run refworkload with 12 iterations for
>>>>>>>>>> baseline and new version, several times. The results vary a bit,
>>>>>>>>>> not sure how much refworkload runs usually vary. Roland also
>>>>>>>>>> tried running refworkload on his x64 Linux to see if he could see
>>>>>>>>>> any major regression. Thank you for helping me with this Roland.
>>>>>>>>>>
>>>>>>>>>> Results from x64 Linux (local machine) [1], x64 Linux (sthdev04)
>>>>>>>>>> [2], sparcv9 Solaris (mrspock) [3], and Roland's x64 Linux [4].
>>>>>>>>>>
>>>>>>>>>> WEBREV: http://cr.openjdk.java.net/~adlertz/JDK-8005956/webrev00/
>>>>>>>>>> JBS: https://jbs.oracle.com/bugs/browse/JDK-8005956
>>>>>>>>>>
>>>>>>>>>> A review would be much appreciated.
>>>>>>>>>> Thank you.
>>>>>>>>>>
>>>>>>>>>> Kind Regards,
>>>>>>>>>> Niclas Adlertz
>>>>>>>>>>
>>>>>>>>>> [0]: Output from the run (running test
>>>>>>>>>> https://jbs.oracle.com/bugs/secure/attachment/11386/PolynomialRoot.java)
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> http://cr.openjdk.java.net/~adlertz/JDK-8005956/log.txt
>>>>>>>>>> Look at phi node 526, it has input nodes 527 and 1619. Node 527
>>>>>>>>>> has input node 528, which is the instruction prior to the reach
>>>>>>>>>> 1716 (which has been rematerialized for phi node 558) in block
>>>>>>>>>> 92. Node 1716 has input node 1619.
>>>>>>>>>> When the unite step (found in the end of reg_split.cpp) unites
>>>>>>>>>> the live range ids of phi nodes and their inputs, node 527
>>>>>>>>>> (recursively via phi node 528) and node 1619 will get the same
>>>>>>>>>> live range id, resulting in the liveness to fail.
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> [1] Results from x64 Linux (local machine) 2 runs
>>>>>>>>>>
>>>>>>>>>> run 1:
>>>>>>>>>> ============================================================================
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> refworkload_base: reference_server
>>>>>>>>>> Benchmark Samples Mean Stdev
>>>>>>>>>> jetstream 12 185.27 0.03
>>>>>>>>>> Write 12 71.42 0.04
>>>>>>>>>> Parse 12 67.92 0.03
>>>>>>>>>> Read 12 17.25 0.10
>>>>>>>>>> Copy 12 102.17 0.02
>>>>>>>>>> scimark 12 1438.69 0.00
>>>>>>>>>> LU 12 3213.92 0.01
>>>>>>>>>> FFT 12 211.00 0.02
>>>>>>>>>> Monte 12 827.71 0.01
>>>>>>>>>> SOR 12 1365.90 0.00
>>>>>>>>>> Sparse 12 1574.90 0.01
>>>>>>>>>> specjbb2000 12 493245.44 0.00
>>>>>>>>>> Last_Warehouse 12 493245.44 0.00
>>>>>>>>>> First_Warehouse 12 118938.86 0.00
>>>>>>>>>> specjbb2005 12 179117.61 0.00
>>>>>>>>>> last 12 179117.61 0.00
>>>>>>>>>> interval_average 12 10536.25 0.00
>>>>>>>>>> peak 12 184040.52 0.01
>>>>>>>>>> overall_average 12 164733.22 0.01
>>>>>>>>>> last_warehouse 12 8.00 0.00
>>>>>>>>>> peak_warehouse 12 4.25 0.11
>>>>>>>>>> first 12 63494.80 0.00
>>>>>>>>>> specjvm98 12 1020.02 0.01
>>>>>>>>>> compress 12 817.54 0.00
>>>>>>>>>> javac 12 598.70 0.01
>>>>>>>>>> db 12 407.37 0.02
>>>>>>>>>> jack 12 1299.66 0.03
>>>>>>>>>> mtrt 12 2904.37 0.02
>>>>>>>>>> jess 12 1089.06 0.02
>>>>>>>>>> mpegaudio 12 1402.48 0.00
>>>>>>>>>> volano25 12 183768.67 0.04
>>>>>>>>>> time 12 4.36 0.04
>>>>>>>>>> connections 12 400.00 0.00
>>>>>>>>>>
>>>>>>>>>> --------------------------------------------------------------------------
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Weighted Geomean 23543.18
>>>>>>>>>> ============================================================================
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> refworkload_new: reference_server
>>>>>>>>>> Benchmark Samples Mean Stdev %Diff P
>>>>>>>>>> Significant
>>>>>>>>>> jetstream 12 198.41 0.05 7.09
>>>>>>>>>> 0.001 Yes
>>>>>>>>>> Write 12 69.92 0.03 2.10
>>>>>>>>>> 0.154 *
>>>>>>>>>> Parse 12 66.00 0.03 2.82
>>>>>>>>>> 0.031 *
>>>>>>>>>> Read 12 15.83 0.14 8.21
>>>>>>>>>> 0.098 *
>>>>>>>>>> Copy 12 89.67 0.02 12.23
>>>>>>>>>> 0.000 Yes
>>>>>>>>>> scimark 12 1443.71 0.00 0.35
>>>>>>>>>> 0.009 Yes
>>>>>>>>>> LU 12 3236.61 0.00 0.71
>>>>>>>>>> 0.012 *
>>>>>>>>>> FFT 12 211.73 0.02 0.34
>>>>>>>>>> 0.628 *
>>>>>>>>>> Monte 12 824.74 0.01 -0.36
>>>>>>>>>> 0.159 *
>>>>>>>>>> SOR 12 1365.70 0.00 -0.01
>>>>>>>>>> 0.908 *
>>>>>>>>>> Sparse 12 1579.76 0.01 0.31
>>>>>>>>>> 0.299 *
>>>>>>>>>> specjbb2000 12 494276.38 0.01 0.21
>>>>>>>>>> 0.384 *
>>>>>>>>>> Last_Warehouse 12 494276.38 0.01 0.21
>>>>>>>>>> 0.384 *
>>>>>>>>>> First_Warehouse 12 118847.53 0.00 -0.08
>>>>>>>>>> 0.548 *
>>>>>>>>>> specjbb2005 12 179359.88 0.01 0.14
>>>>>>>>>> 0.603 *
>>>>>>>>>> last 12 179359.87 0.01 0.14
>>>>>>>>>> 0.603 *
>>>>>>>>>> interval_average 12 10550.67 0.01 0.14
>>>>>>>>>> 0.599 *
>>>>>>>>>> peak 12 184332.03 0.01 0.16
>>>>>>>>>> 0.693 *
>>>>>>>>>> overall_average 12 165145.86 0.01 0.25
>>>>>>>>>> 0.393 *
>>>>>>>>>> last_warehouse 12 8.00 0.00 -0.00
>>>>>>>>>> 0.000 *
>>>>>>>>>> peak_warehouse 12 4.00 0.00 5.88
>>>>>>>>>> 0.082 *
>>>>>>>>>> first 12 63653.25 0.01 0.25
>>>>>>>>>> 0.162 *
>>>>>>>>>> specjvm98 12 1026.25 0.00 0.61
>>>>>>>>>> 0.003 Yes
>>>>>>>>>> compress 12 817.55 0.00 0.00
>>>>>>>>>> 0.998 *
>>>>>>>>>> javac 12 602.40 0.02 0.62
>>>>>>>>>> 0.317 *
>>>>>>>>>> db 12 406.46 0.01 -0.22
>>>>>>>>>> 0.685 *
>>>>>>>>>> jack 12 1341.50 0.03 3.22
>>>>>>>>>> 0.011 *
>>>>>>>>>> mtrt 12 2903.11 0.02 -0.04
>>>>>>>>>> 0.951 *
>>>>>>>>>> jess 12 1099.14 0.01 0.93
>>>>>>>>>> 0.087 *
>>>>>>>>>> mpegaudio 12 1400.09 0.00 -0.17
>>>>>>>>>> 0.150 *
>>>>>>>>>> volano25 12 185056.42 0.04 0.70
>>>>>>>>>> 0.660 *
>>>>>>>>>> time 12 4.33 0.04 0.68
>>>>>>>>>> 0.666 *
>>>>>>>>>> connections 12 400.00 0.00 0.00
>>>>>>>>>> 0.000 *
>>>>>>>>>>
>>>>>>>>>> --------------------------------------------------------------------------
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Weighted Geomean 23787.81 1.04
>>>>>>>>>> ============================================================================
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> run 2:
>>>>>>>>>> ============================================================================
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> refworkload_base: reference_server
>>>>>>>>>> Benchmark Samples Mean Stdev
>>>>>>>>>> jetstream 12 185.27 0.03
>>>>>>>>>> Write 12 71.42 0.04
>>>>>>>>>> Parse 12 67.92 0.03
>>>>>>>>>> Read 12 17.25 0.10
>>>>>>>>>> Copy 12 102.17 0.02
>>>>>>>>>> scimark 12 1438.69 0.00
>>>>>>>>>> LU 12 3213.92 0.01
>>>>>>>>>> FFT 12 211.00 0.02
>>>>>>>>>> Monte 12 827.71 0.01
>>>>>>>>>> SOR 12 1365.90 0.00
>>>>>>>>>> Sparse 12 1574.90 0.01
>>>>>>>>>> specjbb2000 12 493245.44 0.00
>>>>>>>>>> Last_Warehouse 12 493245.44 0.00
>>>>>>>>>> First_Warehouse 12 118938.86 0.00
>>>>>>>>>> specjbb2005 12 179117.61 0.00
>>>>>>>>>> last 12 179117.61 0.00
>>>>>>>>>> interval_average 12 10536.25 0.00
>>>>>>>>>> peak 12 184040.52 0.01
>>>>>>>>>> overall_average 12 164733.22 0.01
>>>>>>>>>> last_warehouse 12 8.00 0.00
>>>>>>>>>> peak_warehouse 12 4.25 0.11
>>>>>>>>>> first 12 63494.80 0.00
>>>>>>>>>> specjvm98 12 1020.02 0.01
>>>>>>>>>> compress 12 817.54 0.00
>>>>>>>>>> javac 12 598.70 0.01
>>>>>>>>>> db 12 407.37 0.02
>>>>>>>>>> jack 12 1299.66 0.03
>>>>>>>>>> mtrt 12 2904.37 0.02
>>>>>>>>>> jess 12 1089.06 0.02
>>>>>>>>>> mpegaudio 12 1402.48 0.00
>>>>>>>>>> volano25 12 183768.67 0.04
>>>>>>>>>> time 12 4.36 0.04
>>>>>>>>>> connections 12 400.00 0.00
>>>>>>>>>>
>>>>>>>>>> --------------------------------------------------------------------------
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Weighted Geomean 23543.18
>>>>>>>>>> ============================================================================
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> refworkload_new: reference_server
>>>>>>>>>> Benchmark Samples Mean Stdev %Diff P
>>>>>>>>>> Significant
>>>>>>>>>> jetstream 12 194.93 0.03 5.21
>>>>>>>>>> 0.001 Yes
>>>>>>>>>> Write 12 71.58 0.03 -0.23
>>>>>>>>>> 0.858 *
>>>>>>>>>> Parse 12 65.25 0.02 3.93
>>>>>>>>>> 0.002 Yes
>>>>>>>>>> Read 12 16.75 0.13 2.90
>>>>>>>>>> 0.534 *
>>>>>>>>>> Copy 12 89.50 0.02 12.40
>>>>>>>>>> 0.000 Yes
>>>>>>>>>> scimark 12 1444.04 0.00 0.37
>>>>>>>>>> 0.011 *
>>>>>>>>>> LU 12 3231.07 0.01 0.53
>>>>>>>>>> 0.084 *
>>>>>>>>>> FFT 12 211.94 0.01 0.44
>>>>>>>>>> 0.521 *
>>>>>>>>>> Monte 12 828.17 0.00 0.06
>>>>>>>>>> 0.789 *
>>>>>>>>>> SOR 12 1367.63 0.00 0.13
>>>>>>>>>> 0.266 *
>>>>>>>>>> Sparse 12 1581.42 0.01 0.41
>>>>>>>>>> 0.134 *
>>>>>>>>>> specjbb2000 12 494674.94 0.00 0.29
>>>>>>>>>> 0.077 *
>>>>>>>>>> Last_Warehouse 12 494674.93 0.00 0.29
>>>>>>>>>> 0.077 *
>>>>>>>>>> First_Warehouse 12 118955.15 0.01 0.01
>>>>>>>>>> 0.946 *
>>>>>>>>>> specjbb2005 12 179408.72 0.01 0.16
>>>>>>>>>> 0.442 *
>>>>>>>>>> last 12 179408.72 0.01 0.16
>>>>>>>>>> 0.442 *
>>>>>>>>>> interval_average 12 10553.50 0.01 0.16
>>>>>>>>>> 0.439 *
>>>>>>>>>> peak 12 184116.40 0.01 0.04
>>>>>>>>>> 0.903 *
>>>>>>>>>> overall_average 12 164999.73 0.00 0.16
>>>>>>>>>> 0.466 *
>>>>>>>>>> last_warehouse 12 8.00 0.00 -0.00
>>>>>>>>>> 0.000 *
>>>>>>>>>> peak_warehouse 12 4.08 0.07 3.92
>>>>>>>>>> 0.296 *
>>>>>>>>>> first 12 63606.07 0.01 0.18
>>>>>>>>>> 0.431 *
>>>>>>>>>> specjvm98 12 1019.34 0.01 -0.07
>>>>>>>>>> 0.780 *
>>>>>>>>>> compress 12 817.54 0.00 -0.00
>>>>>>>>>> 0.997 *
>>>>>>>>>> javac 12 601.06 0.02 0.39
>>>>>>>>>> 0.529 *
>>>>>>>>>> db 12 405.60 0.01 -0.44
>>>>>>>>>> 0.447 *
>>>>>>>>>> jack 12 1313.88 0.02 1.09
>>>>>>>>>> 0.267 *
>>>>>>>>>> mtrt 12 2860.55 0.01 -1.51
>>>>>>>>>> 0.016 *
>>>>>>>>>> jess 12 1093.54 0.00 0.41
>>>>>>>>>> 0.386 *
>>>>>>>>>> mpegaudio 12 1396.44 0.01 -0.43
>>>>>>>>>> 0.058 *
>>>>>>>>>> volano25 12 182607.08 0.03 -0.63
>>>>>>>>>> 0.663 *
>>>>>>>>>> time 12 4.39 0.03 -0.61
>>>>>>>>>> 0.670 *
>>>>>>>>>> connections 12 400.00 0.00 0.00
>>>>>>>>>> 0.000 *
>>>>>>>>>>
>>>>>>>>>> --------------------------------------------------------------------------
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Weighted Geomean 23663.87 0.51
>>>>>>>>>> ============================================================================
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> [2] x64 Linux (sthdev04):
>>>>>>>>>> ============================================================================
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> refworkload_base: reference_server
>>>>>>>>>> Benchmark Samples Mean Stdev
>>>>>>>>>> jetstream 12 266.06 0.05
>>>>>>>>>> Write 12 90.00 0.02
>>>>>>>>>> Parse 12 31.67 0.07
>>>>>>>>>> Read 12 19.83 0.15
>>>>>>>>>> Copy 12 35.92 0.05
>>>>>>>>>> scimark 12 1792.69 0.00
>>>>>>>>>> LU 12 4221.96 0.00
>>>>>>>>>> FFT 12 396.07 0.01
>>>>>>>>>> Monte 12 876.79 0.01
>>>>>>>>>> SOR 12 1473.68 0.00
>>>>>>>>>> Sparse 12 1994.96 0.01
>>>>>>>>>> specjbb2000 12 795853.13 0.01
>>>>>>>>>> Last_Warehouse 12 795853.13 0.01
>>>>>>>>>> First_Warehouse 12 116126.64 0.01
>>>>>>>>>> specjbb2005 12 657436.88 0.01
>>>>>>>>>> last 12 657436.86 0.01
>>>>>>>>>> interval_average 12 10114.42 0.01
>>>>>>>>>> peak 12 735806.35 0.01
>>>>>>>>>> overall_average 12 610453.78 0.01
>>>>>>>>>> last_warehouse 12 8.00 0.00
>>>>>>>>>> peak_warehouse 12 2.50 0.21
>>>>>>>>>> first 12 56556.37 0.01
>>>>>>>>>> specjvm98 12 1037.04 0.02
>>>>>>>>>> compress 12 831.64 0.01
>>>>>>>>>> javac 12 465.49 0.08
>>>>>>>>>> db 12 515.11 0.01
>>>>>>>>>> jack 12 1398.88 0.02
>>>>>>>>>> mtrt 12 2614.33 0.11
>>>>>>>>>> jess 12 1195.19 0.04
>>>>>>>>>> mpegaudio 12 1491.24 0.01
>>>>>>>>>> volano25 12 509262.66 0.04
>>>>>>>>>> time 12 1.57 0.04
>>>>>>>>>> connections 12 400.00 0.00
>>>>>>>>>>
>>>>>>>>>> --------------------------------------------------------------------------
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Weighted Geomean 46117.90
>>>>>>>>>> ============================================================================
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> refworkload_new: reference_server
>>>>>>>>>> Benchmark Samples Mean Stdev %Diff P
>>>>>>>>>> Significant
>>>>>>>>>> jetstream 12 271.15 0.04 1.91
>>>>>>>>>> 0.294 *
>>>>>>>>>> Write 12 88.00 0.02 2.22
>>>>>>>>>> 0.015 *
>>>>>>>>>> Parse 12 31.83 0.08 -0.53
>>>>>>>>>> 0.869 *
>>>>>>>>>> Read 12 18.67 0.14 5.88
>>>>>>>>>> 0.307 *
>>>>>>>>>> Copy 12 35.92 0.05 -0.00
>>>>>>>>>> 1.000 *
>>>>>>>>>> scimark 12 1798.19 0.00 0.31
>>>>>>>>>> 0.006 Yes
>>>>>>>>>> LU 12 4240.68 0.00 0.44
>>>>>>>>>> 0.002 Yes
>>>>>>>>>> FFT 12 399.76 0.02 0.93
>>>>>>>>>> 0.150 *
>>>>>>>>>> Monte 12 874.18 0.01 -0.30
>>>>>>>>>> 0.242 *
>>>>>>>>>> SOR 12 1473.97 0.00 0.02
>>>>>>>>>> 0.235 *
>>>>>>>>>> Sparse 12 2002.35 0.01 0.37
>>>>>>>>>> 0.332 *
>>>>>>>>>> specjbb2000 12 787858.63 0.03 -1.00
>>>>>>>>>> 0.290 *
>>>>>>>>>> Last_Warehouse 12 787858.59 0.03 -1.00
>>>>>>>>>> 0.290 *
>>>>>>>>>> First_Warehouse 12 116645.31 0.02 0.45
>>>>>>>>>> 0.572 *
>>>>>>>>>> specjbb2005 12 657887.44 0.01 0.07
>>>>>>>>>> 0.820 *
>>>>>>>>>> last 12 657887.45 0.01 0.07
>>>>>>>>>> 0.820 *
>>>>>>>>>> interval_average 12 10121.25 0.01 0.07
>>>>>>>>>> 0.822 *
>>>>>>>>>> peak 12 737268.41 0.01 0.20
>>>>>>>>>> 0.580 *
>>>>>>>>>> overall_average 12 612183.21 0.01 0.28
>>>>>>>>>> 0.318 *
>>>>>>>>>> last_warehouse 12 8.00 0.00 -0.00
>>>>>>>>>> 0.000 *
>>>>>>>>>> peak_warehouse 12 2.67 0.18 -6.67
>>>>>>>>>> 0.430 *
>>>>>>>>>> first 12 56839.78 0.01 0.50
>>>>>>>>>> 0.343 *
>>>>>>>>>> specjvm98 12 1045.08 0.02 0.78
>>>>>>>>>> 0.333 *
>>>>>>>>>> compress 12 830.84 0.00 -0.10
>>>>>>>>>> 0.671 *
>>>>>>>>>> javac 12 487.46 0.09 4.72
>>>>>>>>>> 0.192 *
>>>>>>>>>> db 12 518.01 0.01 0.56
>>>>>>>>>> 0.041 *
>>>>>>>>>> jack 12 1440.04 0.03 2.94
>>>>>>>>>> 0.010 Yes
>>>>>>>>>> mtrt 12 2569.15 0.08 -1.73
>>>>>>>>>> 0.656 *
>>>>>>>>>> jess 12 1188.40 0.04 -0.57
>>>>>>>>>> 0.742 *
>>>>>>>>>> mpegaudio 12 1486.04 0.01 -0.35
>>>>>>>>>> 0.184 *
>>>>>>>>>> volano25 12 505473.50 0.02 -0.74
>>>>>>>>>> 0.600 *
>>>>>>>>>> time 12 1.58 0.02 -0.60
>>>>>>>>>> 0.674 *
>>>>>>>>>> connections 12 400.00 0.00 0.00
>>>>>>>>>> 0.000 *
>>>>>>>>>>
>>>>>>>>>> --------------------------------------------------------------------------
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Weighted Geomean 46149.13 0.07
>>>>>>>>>> ============================================================================
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> [3] Results from sparcv9 solaris (mrspock):
>>>>>>>>>> ============================================================================
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> refworkload_base: reference_server
>>>>>>>>>> Benchmark Samples Mean Stdev
>>>>>>>>>> jetstream 12 36.03 0.01
>>>>>>>>>> Copy 12 258.08 0.00
>>>>>>>>>> Parse 12 337.08 0.00
>>>>>>>>>> Read 12 126.67 0.04
>>>>>>>>>> Write 12 538.83 0.00
>>>>>>>>>> scimark 12 44.36 0.01
>>>>>>>>>> Sparse 12 27.44 0.01
>>>>>>>>>> LU 12 51.53 0.02
>>>>>>>>>> SOR 12 80.13 0.01
>>>>>>>>>> FFT 12 14.28 0.00
>>>>>>>>>> Monte 12 48.44 0.00
>>>>>>>>>> specjbb2000 12 80257.23 0.00
>>>>>>>>>> First_Warehouse 12 10756.41 0.01
>>>>>>>>>> Last_Warehouse 12 80257.24 0.00
>>>>>>>>>> specjbb2005 12 274794.56 0.02
>>>>>>>>>> peak 12 274794.56 0.02
>>>>>>>>>> peak_warehouse 12 8.00 0.00
>>>>>>>>>> last 12 274794.56 0.02
>>>>>>>>>> interval_average 12 535.58 0.02
>>>>>>>>>> first 12 5079.05 0.02
>>>>>>>>>> overall_average 12 121172.61 0.07
>>>>>>>>>> last_warehouse 12 8.00 0.00
>>>>>>>>>> specjvm98 12 81.51 0.01
>>>>>>>>>> javac 12 58.26 0.01
>>>>>>>>>> db 12 54.68 0.03
>>>>>>>>>> jess 12 88.80 0.00
>>>>>>>>>> jack 12 41.98 0.01
>>>>>>>>>> compress 12 87.66 0.00
>>>>>>>>>> mtrt 12 228.86 0.01
>>>>>>>>>> mpegaudio 12 100.37 0.00
>>>>>>>>>> volano25 12 114566.00 0.15
>>>>>>>>>> connections 12 400.00 0.00
>>>>>>>>>> time 12 7.15 0.17
>>>>>>>>>>
>>>>>>>>>> --------------------------------------------------------------------------
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Weighted Geomean 6260.91
>>>>>>>>>> ============================================================================
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> refworkload_new: reference_server
>>>>>>>>>> Benchmark Samples Mean Stdev %Diff P
>>>>>>>>>> Significant
>>>>>>>>>> jetstream 12 35.91 0.00 -0.35
>>>>>>>>>> 0.264 *
>>>>>>>>>> Copy 12 257.00 0.00 0.42
>>>>>>>>>> 0.007 Yes
>>>>>>>>>> Parse 12 347.33 0.01 -3.04
>>>>>>>>>> 0.000 Yes
>>>>>>>>>> Read 12 125.75 0.00 0.72
>>>>>>>>>> 0.528 *
>>>>>>>>>> Write 12 536.00 0.00 0.53
>>>>>>>>>> 0.000 Yes
>>>>>>>>>> scimark 12 44.62 0.01 0.58
>>>>>>>>>> 0.028 *
>>>>>>>>>> Sparse 12 27.48 0.00 0.16
>>>>>>>>>> 0.400 *
>>>>>>>>>> LU 12 51.82 0.01 0.57
>>>>>>>>>> 0.487 *
>>>>>>>>>> SOR 12 81.20 0.01 1.34
>>>>>>>>>> 0.001 Yes
>>>>>>>>>> FFT 12 14.26 0.01 -0.10
>>>>>>>>>> 0.617 *
>>>>>>>>>> Monte 12 48.34 0.00 -0.22
>>>>>>>>>> 0.154 *
>>>>>>>>>> specjbb2000 12 80075.02 0.01 -0.23
>>>>>>>>>> 0.213 *
>>>>>>>>>> First_Warehouse 12 10871.56 0.02 1.07
>>>>>>>>>> 0.169 *
>>>>>>>>>> Last_Warehouse 12 80075.01 0.01 -0.23
>>>>>>>>>> 0.213 *
>>>>>>>>>> specjbb2005 12 271688.31 0.03 -1.13
>>>>>>>>>> 0.276 *
>>>>>>>>>> peak 12 271688.33 0.03 -1.13
>>>>>>>>>> 0.276 *
>>>>>>>>>> peak_warehouse 12 8.00 0.00 -0.00
>>>>>>>>>> 0.000 *
>>>>>>>>>> last 12 271688.33 0.03 -1.13
>>>>>>>>>> 0.276 *
>>>>>>>>>> interval_average 12 529.75 0.03 -1.09
>>>>>>>>>> 0.291 *
>>>>>>>>>> first 12 5135.88 0.02 1.12
>>>>>>>>>> 0.128 *
>>>>>>>>>> overall_average 12 118168.54 0.05 -2.48
>>>>>>>>>> 0.334 *
>>>>>>>>>> last_warehouse 12 8.00 0.00 -0.00
>>>>>>>>>> 0.000 *
>>>>>>>>>> specjvm98 12 81.23 0.01 -0.34
>>>>>>>>>> 0.221 *
>>>>>>>>>> javac 12 57.66 0.01 -1.02
>>>>>>>>>> 0.021 *
>>>>>>>>>> db 12 53.89 0.04 -1.44
>>>>>>>>>> 0.278 *
>>>>>>>>>> jess 12 89.33 0.01 0.60
>>>>>>>>>> 0.009 Yes
>>>>>>>>>> jack 12 41.75 0.01 -0.56
>>>>>>>>>> 0.091 *
>>>>>>>>>> compress 12 87.59 0.00 -0.07
>>>>>>>>>> 0.545 *
>>>>>>>>>> mtrt 12 229.51 0.02 0.29
>>>>>>>>>> 0.640 *
>>>>>>>>>> mpegaudio 12 100.24 0.00 -0.13
>>>>>>>>>> 0.453 *
>>>>>>>>>> volano25 12 104635.84 0.21 -8.67
>>>>>>>>>> 0.232 *
>>>>>>>>>> connections 12 400.00 0.00 0.00
>>>>>>>>>> 0.000 *
>>>>>>>>>> time 12 8.02 0.24 -12.16
>>>>>>>>>> 0.208 *
>>>>>>>>>>
>>>>>>>>>> --------------------------------------------------------------------------
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Weighted Geomean 6128.92 -2.11
>>>>>>>>>> ============================================================================
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> [4] Results from Roland's x64 Linux:
>>>>>>>>>> ============================================================================
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> tbase: reference_server
>>>>>>>>>> Benchmark Samples Mean Stdev
>>>>>>>>>> jetstream 12 84.09 0.04
>>>>>>>>>> Copy 12 952.42 0.05
>>>>>>>>>> Parse 12 44.33 0.04
>>>>>>>>>> Read 12 15.00 0.03
>>>>>>>>>> Write 12 320.33 0.16
>>>>>>>>>> scimark 12 457.72 0.00
>>>>>>>>>> Sparse 12 279.24 0.00
>>>>>>>>>> LU 12 702.48 0.00
>>>>>>>>>> SOR 12 764.40 0.00
>>>>>>>>>> FFT 12 32.23 0.00
>>>>>>>>>> Monte 12 510.25 0.00
>>>>>>>>>> specjbb2000 12 397041.59 0.01
>>>>>>>>>> First_Warehouse 12 61010.62 0.01
>>>>>>>>>> Last_Warehouse 12 397041.60 0.01
>>>>>>>>>> specjbb2005 12 211800.03 0.01
>>>>>>>>>> peak 12 229864.05 0.01
>>>>>>>>>> peak_warehouse 12 4.00 0.00
>>>>>>>>>> last 12 211800.03 0.01
>>>>>>>>>> interval_average 12 12459.00 0.01
>>>>>>>>>> first 12 36496.39 0.01
>>>>>>>>>> overall_average 12 184620.84 0.01
>>>>>>>>>> last_warehouse 12 8.00 0.00
>>>>>>>>>> specjvm98 12 675.34 0.00
>>>>>>>>>> javac 12 396.86 0.00
>>>>>>>>>> db 12 463.10 0.00
>>>>>>>>>> jess 12 671.13 0.01
>>>>>>>>>> jack 12 544.39 0.01
>>>>>>>>>> compress 12 597.28 0.00
>>>>>>>>>> mtrt 12 1724.51 0.01
>>>>>>>>>> mpegaudio 12 926.39 0.00
>>>>>>>>>> volano25 12 161056.08 0.05
>>>>>>>>>> connections 12 400.00 0.00
>>>>>>>>>> time 12 4.98 0.04
>>>>>>>>>>
>>>>>>>>>> --------------------------------------------------------------------------
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Weighted Geomean 16931.56
>>>>>>>>>> ============================================================================
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> tniclas: reference_server
>>>>>>>>>> Benchmark Samples Mean Stdev %Diff P
>>>>>>>>>> Significant
>>>>>>>>>> jetstream 12 88.03 0.02 4.68
>>>>>>>>>> 0.003 Yes
>>>>>>>>>> Copy 12 866.75 0.07 8.99
>>>>>>>>>> 0.001 Yes
>>>>>>>>>> Parse 12 42.83 0.01 3.38
>>>>>>>>>> 0.012 *
>>>>>>>>>> Read 12 14.83 0.04 1.11
>>>>>>>>>> 0.431 *
>>>>>>>>>> Write 12 303.67 0.02 5.20
>>>>>>>>>> 0.291 *
>>>>>>>>>> scimark 12 457.89 0.00 0.04
>>>>>>>>>> 0.116 *
>>>>>>>>>> Sparse 12 279.82 0.00 0.21
>>>>>>>>>> 0.033 *
>>>>>>>>>> LU 12 702.76 0.00 0.04
>>>>>>>>>> 0.122 *
>>>>>>>>>> SOR 12 764.59 0.00 0.02
>>>>>>>>>> 0.636 *
>>>>>>>>>> FFT 12 32.28 0.00 0.15
>>>>>>>>>> 0.098 *
>>>>>>>>>> Monte 12 509.99 0.00 -0.05
>>>>>>>>>> 0.127 *
>>>>>>>>>> specjbb2000 12 398055.84 0.01 0.26
>>>>>>>>>> 0.540 *
>>>>>>>>>> First_Warehouse 12 61447.70 0.01 0.72
>>>>>>>>>> 0.005 Yes
>>>>>>>>>> Last_Warehouse 12 398055.85 0.01 0.26
>>>>>>>>>> 0.540 *
>>>>>>>>>> specjbb2005 12 212037.69 0.01 0.11
>>>>>>>>>> 0.657 *
>>>>>>>>>> peak 12 230138.69 0.01 0.12
>>>>>>>>>> 0.773 *
>>>>>>>>>> peak_warehouse 12 4.00 0.00 -0.00
>>>>>>>>>> 0.000 *
>>>>>>>>>> last 12 212037.69 0.01 0.11
>>>>>>>>>> 0.657 *
>>>>>>>>>> interval_average 12 12473.00 0.01 0.11
>>>>>>>>>> 0.657 *
>>>>>>>>>> first 12 36631.47 0.01 0.37
>>>>>>>>>> 0.174 *
>>>>>>>>>> overall_average 12 184966.28 0.01 0.19
>>>>>>>>>> 0.529 *
>>>>>>>>>> last_warehouse 12 8.00 0.00 -0.00
>>>>>>>>>> 0.000 *
>>>>>>>>>> specjvm98 12 673.85 0.00 -0.22
>>>>>>>>>> 0.058 *
>>>>>>>>>> javac 12 397.12 0.01 0.07
>>>>>>>>>> 0.786 *
>>>>>>>>>> db 12 463.87 0.00 0.17
>>>>>>>>>> 0.262 *
>>>>>>>>>> jess 12 667.62 0.01 -0.52
>>>>>>>>>> 0.213 *
>>>>>>>>>> jack 12 543.46 0.01 -0.17
>>>>>>>>>> 0.656 *
>>>>>>>>>> compress 12 596.70 0.00 -0.10
>>>>>>>>>> 0.306 *
>>>>>>>>>> mtrt 12 1704.44 0.01 -1.16
>>>>>>>>>> 0.006 Yes
>>>>>>>>>> mpegaudio 12 928.21 0.00 0.20
>>>>>>>>>> 0.135 *
>>>>>>>>>> volano25 12 162796.33 0.05 1.08
>>>>>>>>>> 0.588 *
>>>>>>>>>> connections 12 400.00 0.00 0.00
>>>>>>>>>> 0.000 *
>>>>>>>>>> time 12 4.93 0.05 1.03
>>>>>>>>>> 0.592 *
>>>>>>>>>>
>>>>>>>>>> --------------------------------------------------------------------------
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Weighted Geomean 17052.36 0.71
>>>>>>>>>> ============================================================================
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>> <Attached Message.eml><Attached Message.eml>
>>>>>>>>
>>>>>>>
>>>>>
>
More information about the hotspot-compiler-dev
mailing list