Calculate widths does not respect single core running

[Hannes]

Hi,

this was asked as a question earlier, sorry for double posting but this bug is causing some problems with our workflow and should be easy to fix.

When running 'bin/generate_events 0 test' when particle widths are set to 'auto', multiple instances of madevent are running to calculate the widths in a bsm model.
It seems that for the calculation of widths madevent does not inherit the options set for a particular process even though this would be the more desirable behavior.
Would it be possible to fix this?

Cheers,
Hannes

[Olivier Mattelaer (olivier-mattelaer)]

Hi Hannes,

I put this as low priority on my todo list actually.

1) How did you request that you are running as single core?
(is it a local choice or a global choice in the mg5_configuration.txt?)

2) Are you running the latest version of the code?
I have the feeling to have solve that problem in the past, so I want to double check that you are indeed using the latest version

Cheers,

Olivier

> On 25 Jan 2018, at 10:51, Hannes <email address hidden> wrote:
>
> Public bug reported:
>
> Hi,
>
> this was asked as a question earlier, sorry for double posting but this
> bug is causing some problems with our workflow and should be easy to
> fix.
>
> When running 'bin/generate_events 0 test' when particle widths are set to 'auto', multiple instances of madevent are running to calculate the widths in a bsm model.
> It seems that for the calculation of widths madevent does not inherit the options set for a particular process even though this would be the more desirable behavior.
> Would it be possible to fix this?
>
> Cheers,
> Hannes
>
> ** Affects: mg5amcnlo
> Importance: Undecided
> Status: New
>
> --
> You received this bug notification because you are subscribed to
> MadGraph5_aMC@NLO.
> https://bugs.launchpad.net/bugs/1745333
>
> Title:
> Calculate widths does not respect single core running
>
> Status in MadGraph5_aMC@NLO:
> New
>
> Bug description:
> Hi,
>
> this was asked as a question earlier, sorry for double posting but
> this bug is causing some problems with our workflow and should be easy
> to fix.
>
> When running 'bin/generate_events 0 test' when particle widths are set to 'auto', multiple instances of madevent are running to calculate the widths in a bsm model.
> It seems that for the calculation of widths madevent does not inherit the options set for a particular process even though this would be the more desirable behavior.
> Would it be possible to fix this?
>
> Cheers,
> Hannes
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/mg5amcnlo/+bug/1745333/+subscriptions

[Hannes (hannes3)]

Hi Olivier,

thanks for your quick reply.

1) I requested single core by calling 'bin/generate_events 0 test' within the process dir. Whatever is set in the local config 'me5_configuration.txt' is ignored during width calculation, too. It uses what is set in the global 'input/mg5_configuration.txt'.

2) I tested it in 2_6_1 and the problem is also present in older version.

I suspect that the problem might arise around here, http://bazaar.launchpad.net/~madteam/mg5amcnlo/series2.0/view/head:/madgraph/interface/madgraph_interface.py#L8073, the madevent used for integration of the width just uses the global option and does not respect what was set for the process.

Cheers,
Hannes

[Olivier Mattelaer (olivier-mattelaer)]

Hi,

As a quick workaround, you can then edit the file
> 'input/mg5_configuration.txt'.

> 2) I tested it in 2_6_1 and the problem is also present in older version.

Ok thanks I will take a look (probably next week)

Cheers,

Olivier

> On 25 Jan 2018, at 11:46, Hannes <email address hidden> wrote:
>
> Hi Olivier,
>
> thanks for your quick reply.
>
> 1) I requested single core by calling 'bin/generate_events 0 test' within the process dir. Whatever is set in the local config 'me5_configuration.txt' is ignored during width calculation, too. It uses what is set in the global 'input/mg5_configuration.txt'.
>
> 2) I tested it in 2_6_1 and the problem is also present in older version.
>
> I suspect that the problem might arise around here,
> http://bazaar.launchpad.net/~madteam/mg5amcnlo/series2.0/view/head:/madgraph/interface/madgraph_interface.py#L8073,
> the madevent used for integration of the width just uses the global
> option and does not respect what was set for the process.
>
> Cheers,
> Hannes
>
> --
> You received this bug notification because you are subscribed to
> MadGraph5_aMC@NLO.
> https://bugs.launchpad.net/bugs/1745333
>
> Title:
> Calculate widths does not respect single core running
>
> Status in MadGraph5_aMC@NLO:
> New
>
> Bug description:
> Hi,
>
> this was asked as a question earlier, sorry for double posting but
> this bug is causing some problems with our workflow and should be easy
> to fix.
>
> When running 'bin/generate_events 0 test' when particle widths are set to 'auto', multiple instances of madevent are running to calculate the widths in a bsm model.
> It seems that for the calculation of widths madevent does not inherit the options set for a particular process even though this would be the more desirable behavior.
> Would it be possible to fix this?
>
> Cheers,
> Hannes
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/mg5amcnlo/+bug/1745333/+subscriptions

[Hannes (hannes3)]

Hi Olivier,

thanks for looking into this.
It probably seems silly but the ATLAS workflow relies on being able to set the options locally, which is why modifying mg5_configuration.txt does not help.

Cheers,
Hannes

[Olivier Mattelaer (olivier-mattelaer)]

Hi,

Would using
~/.mg5/mg5_configuration.txt
a solution for you? (it would be if the mode of running is not hardcoded inside the input/mg5_configuration.txt)
another solution which also only work if the mode is not hardcoded in input/mg5_configuration.txt
is to define the environment variable MADGRAPH_BASE

and then you can define the file $MADGRAPH_BASE/mg5_configuration.txt

Cheers,

Olivier

On 25 Jan 2018, at 12:32, Hannes <<email address hidden><mailto:<email address hidden>>> wrote:

Hi Olivier,

thanks for looking into this.
It probably seems silly but the ATLAS workflow relies on being able to set the options locally, which is why modifying mg5_configuration.txt does not help.

Cheers,
Hannes

--
You received this bug notification because you are subscribed to
MadGraph5_aMC@NLO.
https://bugs.launchpad.net/bugs/1745333

Title:
 Calculate widths does not respect single core running

Status in MadGraph5_aMC@NLO:
 New

Bug description:
 Hi,

 this was asked as a question earlier, sorry for double posting but
 this bug is causing some problems with our workflow and should be easy
 to fix.

 When running 'bin/generate_events 0 test' when particle widths are set to 'auto', multiple instances of madevent are running to calculate the widths in a bsm model.
 It seems that for the calculation of widths madevent does not inherit the options set for a particular process even though this would be the more desirable behavior.
 Would it be possible to fix this?

 Cheers,
 Hannes

To manage notifications about this bug go to:
https://bugs.launchpad.net/mg5amcnlo/+bug/1745333/+subscriptions

[Hannes (hannes3)]

Hi Olivier,

thank you, defining MADGRAPH_BASE would indeed work, as a temporary workaround at least.
I was a bit worried that setting this to an arbitrary location that only contains the config might break something but this is not the case?

Cheers,
Hannes

[Olivier Mattelaer (olivier-mattelaer)]

Hi,

> I was a bit worried that setting this to an arbitrary location that only contains the config might break something but this is not the case?

No this should be pretty save. The only other use of $MADGRAPH_BASE is related to our webserver.
So you should be fully save.

Otherwise, here is the fix that I'm including in the development version:

=== modified file 'madgraph/interface/common_run_interface.py'
--- madgraph/interface/common_run_interface.py 2018-01-29 09:32:32 +0000
+++ madgraph/interface/common_run_interface.py 2018-01-29 12:57:58 +0000
@@ -2168,6 +2168,7 @@
         from madgraph.interface.master_interface import MasterCmd
         cmd = MasterCmd()
         self.define_child_cmd_interface(cmd, interface=False)
+ cmd.options.update(self.options)
         cmd.exec_cmd('set automatic_html_opening False --no_save')
         if not opts['path']:
             opts['path'] = pjoin(self.me_dir, 'Cards', 'param_card.dat')

=== modified file 'madgraph/interface/madgraph_interface.py'
--- madgraph/interface/madgraph_interface.py 2018-01-18 20:21:01 +0000
+++ madgraph/interface/madgraph_interface.py 2018-01-29 13:21:43 +0000
@@ -8076,6 +8076,8 @@
                     check_param_card.convert_to_slha1(pjoin(decay_dir, 'Cards', 'param_card.dat'))
                 # call a ME interface and define as it as child for correct error handling
                 me_cmd = madevent_interface.MadEventCmd(decay_dir)
+ me_cmd.options.update(self.options)
+ me_cmd.configure_run_mode(self.options['run_mode'])
                 #self.define_child_cmd_interface(me_cmd, interface=False)
                 me_cmd.model_name = self._curr_model['name'] #needed for mssm
                 me_cmd.options['automatic_html_opening'] = False

Cheers,

Olivier

> On 26 Jan 2018, at 13:27, Hannes <email address hidden> wrote:
>
> Hi Olivier,
>
> thank you, defining MADGRAPH_BASE would indeed work, as a temporary workaround at least.
> I was a bit worried that setting this to an arbitrary location that only contains the config might break something but this is not the case?
>
> Cheers,
> Hannes
>
> --
> You received this bug notification because you are subscribed to
> MadGraph5_aMC@NLO.
> https://bugs.launchpad.net/bugs/1745333
>
> Title:
> Calculate widths does not respect single core running
>
> Status in MadGraph5_aMC@NLO:
> New
>
> Bug description:
> Hi,
>
> this was asked as a question earlier, sorry for double posting but
> this bug is causing some problems with our workflow and should be easy
> to fix.
>
> When running 'bin/generate_events 0 test' when particle widths are set to 'auto', multiple instances of madevent are running to calculate the widths in a bsm model.
> It seems that for the calculation of widths madevent does not inherit the options set for a particular process even though this would be the more desirable behavior.
> Would it be possible to fix this?
>
> Cheers,
> Hannes
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/mg5amcnlo/+bug/1745333/+subscriptions

[Hannes (hannes3)]

Hi Olivier,

thanks a lot, when you said next week I wasn't expecting the fix already on Monday!

Cheers,
Hannes

[Hannes (hannes3)]

Hi,

just to confirm, the fix is working for us.

Do you know when the next release containing this fix will come out approximately?

Cheers,
Hannes

[Hannes (hannes3)]

Hi,

sorry to bother you again, do you know when the next release with this fix would be approximately be available. It would be very helpful for our planning to have an estimate.

Cheers,
Hannes

[Olivier Mattelaer (olivier-mattelaer)]

Hi,

I have actually no idea. There is a bug on Madspin that I want to fix before the next release but I do not find enough time to understandfix that one.

Cheers,

Olivier

> On 6 Feb 2018, at 17:34, Hannes <email address hidden> wrote:
>
> Hi,
>
> sorry to bother you again, do you know when the next release with this
> fix would be approximately be available. It would be very helpful for
> our planning to have an estimate.
>
> Cheers,
> Hannes
>
> --
> You received this bug notification because you are subscribed to
> MadGraph5_aMC@NLO.
> https://bugs.launchpad.net/bugs/1745333
>
> Title:
> Calculate widths does not respect single core running
>
> Status in MadGraph5_aMC@NLO:
> New
>
> Bug description:
> Hi,
>
> this was asked as a question earlier, sorry for double posting but
> this bug is causing some problems with our workflow and should be easy
> to fix.
>
> When running 'bin/generate_events 0 test' when particle widths are set to 'auto', multiple instances of madevent are running to calculate the widths in a bsm model.
> It seems that for the calculation of widths madevent does not inherit the options set for a particular process even though this would be the more desirable behavior.
> Would it be possible to fix this?
>
> Cheers,
> Hannes
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/mg5amcnlo/+bug/1745333/+subscriptions

[Hannes (hannes3)]

Thanks anyway!
Cheers,
Hannes