Compiling afpelcomb against applgridphoton about pinecards

[cschwan]

This compilation error must appear, because I removed the function. That's because we might have more than one leading order, so we must decide what to do in this case (we have to change apfelcomb).

[cschwan]

@scarrazza: Marco just uploaded an applgrid with two NLO corrections. Could you try to use it? We did not check whether the contents are meaningful, but I hope it suffices to test the processing apfelcomb.

[scarrazza]

Ok, I will give a try as soon as the fpga works.

[cschwan]

Short update from Stefano and myself: We've fixed apfelcomb, so that it compiles against the new APPLgridphoton. However, the check of the total cross section fails, which means that the convolution of the generated FKtable does not yield the same value of the convolution done by APPLgridphoton. In particular, the value is too large (think of a factor 100 or so) and it already fails for QCD LO. I will try to track down the problem by generating an old APPLgrid table using the old amcfast toolchain and the same using our current code with amcblast.

[cschwan]

Generating LO only doesn't work, at least not with our scripts (the available options in mg5_aMC differ for LO and NLO, apparently). But I will simply generate LO with a smaller sample of phase space points.

[cschwan]

We've fixed (LO) or are about to fix (NLO) the last remaining problems with apfelcomb.

@scarrazza: Would you please describe what you had to do to get a sensible FKtable from the APPLgrid I generated (just for reference and maybe as a heads-up for @enocera)?

The following commit should fix the problem with NLO: NNPDF/apfelcomb@3ea98b7 (in the NNPDF/apfelcomb repository, ew branch). Please let me know if it doesn't.

[scarrazza]

Thanks, I will let you know this afternoon.

[scarrazza]

@cschwan I confirm your last commit fix the NLO and NNLO issues, e.g.:

• LO

                        Verification                  
<idat>    <FK>           <Source>       <Rel.Err>      
0         141.5          141.5          2.157e-06      

• NLO:

                        Verification                  
<idat>    <FK>           <Source>       <Rel.Err>      
0         169.9          169.9          5.789e-07    

• NNLO:

                        Verification                  
<idat>    <FK>           <Source>       <Rel.Err>      
0         168.2          168.2          1.461e-06    

Concerning the changes, we have to make sure that pdfwgt (apfelcomb/db) is always set to 1 when using applgrids generated by amcatnlo, because the normalization requires the product:

const double pdfnrm = pdfwgt ? igrid->weightfun(x1)*igrid->weightfun(x2) : 1.0;

[enocera]

@scarrazza @cschwan @marcozaro Thank you very much for the update - and apologies for my silence. Can we schedule a call to have a more detailed update of what was going on with apfelcomb? BTW, I've produced applgrids with EW corrections also for high-mass DY, I should put them somewhere soon.

[scarrazza]

@enocera Can we meet later today? E.g. at 15, 16, 17?

[enocera]

17 is working with me.

[scarrazza]

Perfect, that's fine for us too.