Thanks a ton! Will try this and report.
Deepak
On Tue, Jun 16, 2015 at 4:32 PM, Michael Wall mwall.physics@gmail.com wrote:
Hi Deepak,
I think I see the problem. The implementation of tebd includes optimizations for conserving U(1) symmetries, like particle number. If the on-site irreps of the symmetry are all one-dimensional (as they are for spins conserving magnetization of spinless particle models), then tebd uses a compressed mps format (which in the code means it uses the Gammasm instead of the Gammas). The kink state routine is set up for spins, which use this optimization, but this optimization does not apply for the spinful fermions. However, you can replace Gammasm by Gammas in AllocateGamLam, replace Gammasm by Gammas and remove the LabelLeft and LabelRight in the call to ProductStateMPD, and add CALL ProductStateLabels(LabelLeft, LabelRight, carray,intDegFree=1) after the line calling ProductStateMPD. This will use the correct, less compressed structure for the mps. The source code/syntax for all the above calls can be found in StateOps.f90, in the core directory. This should fix the issue.
-Michael
On Tue, Jun 16, 2015 at 2:15 PM, Deepak Iyer deepak.g.iyer@gmail.com wrote:
Hi Michael,
I've been playing around, and it seems like the line causing the segfault is
CALL AllocateOverlap(Measures, overlapnum, Gammas, Lambdas)
in the "Setup measurements" part, which it reaches because idof=true for the fermi Hubbard. I don't know the code enough to know what these are. Help?
Thanks! Deepak
On Tue, Jun 16, 2015 at 2:08 PM, Deepak Iyer deepak.g.iyer@gmail.com wrote:
Thanks for the response! I'll try to edit tebd.f90 for now.
Cheers! Deepak
On Tue, Jun 16, 2015 at 12:30 PM, Michael Wall mwall.physics@gmail.com wrote:
Hi Deepak,
The kink initial state was intended to be used only with the spin tutorial, and has not been tested with any other models. However, you can change the tebd main file "tebd.f90" in the source to make this initial state any product state you choose (this means you have to build from source, of course). The relevant code starts at line 162. Additionally, the new mps codes in ALPS may have more streamlined ways of specifying initial states; the mps developers can give you more information if that is the case.
-Michael
On Tue, Jun 16, 2015 at 10:18 AM, Deepak Iyer deepak.g.iyer@gmail.com wrote:
Hi all,
I couldn't find this addressed on the forums or docs. I get a segfault (invalid memory reference) while trying to run TEBD on the fermion Hubbard with kink initial state. I imagine it is getting confused while trying to set up the initial state. Any way I can get around this?
Thanks! Deepak