• Using ISSM

  • SEMIC module running slow; input reading much slower than PDD

A while ago I tried md.smb=SMBsemic(); but the SEMIC moule was running extremely slow (>1h for 1m elements every step, 2 orders of magnitude slower than the standalone SEMIC). I resorted to PDD instead and temporarily droped this.

Recently I plan to use energy balance model again. I tried to time the code and it looks like the majority of time was spent reading inputs, which is a lot slower than similiar parts in PDD calculation. I will talk to Martin. But meanwhile I wonder if someone can double check the run time for each section in the SMBsemic sequence and potentially spot something, I mean, for such dramatic difference even in input reading, there must be something seriously faulty, right?

Thanks,
Wade

wadewei changed the title to SEMIC module running slow; input reading much slower than PDD .

Hi, Wadewei

I have several questions about your problem.

1) Do you test the test/Nightly/test246.m? How much time do you use in this script?

2) In addition, do you use the distributed version ? (https://issm.ess.uci.edu/svn/issm/issm/trunk)

Best regards,
Inwoo

Hi, Wadewei

I found that you have already try to use SEMIC and give some question at here; "https://issm.ess.uci.edu/forum/d/278-smbsemic/2".

Last year, I talked about SEMIC with Martin (Ruckamp?), and now I have updated the SEMIC. I preserved old version SEMIC (Martin Ruckamp refers it as "offline" mode). I think that you can try the SEMIC with following code and I attach some code for validation from SEMIC result (https://github.com/mkrapp/semic/tree/master/example/data).

md.smb = SMBsemic;
md.smb.ismethod = 1; % turn off offline mode and use transient mode!

Now, I added the some parameters options, such as albedo scheme in SEMIC.

%albedo_scheme : 1 -- albedo scheme for SEMIC. 0: none, 1: slater, 2: isba, 3: denby, 4: alex (default is 0)

md.smb.albedo_scheme = 0; % default
md.smb.albedo_scheme = 1; % slater
md.smb.albedo_scheme = 2; % isba
% etc.

Let me know if the attachment file works well.

Best regards,
Inwoo

transect-output.txt
225kB
transect-input.txt
243kB
test-semic.m
6kB
    5 days later
    8 months later

    in-woo
    Hi, Inwoo,
    I noticed hice became minus 20 meters or something after a few years in some places. Should I be concerned or should hice be restricted to positive numbers only? Thanks!
    Wade

        a month later

        wadewei

        I reply the above comment in https://issm.ess.uci.edu/forum/d/547-possible-bug-in-semic-transient, which also metioned the similar problem in hice value.

        Best regards,
        Inwoo