Skip to content

control.mixsyn hangs on M1 #973

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Closed
NaoPross opened this issue Feb 29, 2024 · 2 comments
Closed

control.mixsyn hangs on M1 #973

NaoPross opened this issue Feb 29, 2024 · 2 comments
Labels
duplicate

Comments

@NaoPross
Copy link

NaoPross commented Feb 29, 2024 •
edited
Loading

Description

I have the following code in a file bug.py

import control as ct

G = ct.tf([1, 1], [1, 0.5, 4])
Wp = ct.tf([2, -2.2, 1], [3, 0.2, 0.01])

k, cl, p, info = ct.mixsyn(G, Wp)

and when I run it it seems to hang forever. Compare this to the MATLAB equivalent

G = tf([1, 1], [1, 0.5, 4])
Wp = tf([2, -2.2, 1], [3, 0.2, 0.01])

[K,CL,GAM,info] = mixsyn(G, Wp)

which runs fine.

Expand for MATLAB output 
G =
 
       s + 1
  ---------------
  s^2 + 0.5 s + 4
 
Continuous-time transfer function.


Wp =
 
   2 s^2 - 2.2 s + 1
  --------------------
  3 s^2 + 0.2 s + 0.01
 
Continuous-time transfer function.


K =
 
  A = 
               x1          x2          x3          x4
   x1    -0.06667    -0.05333           0           0
   x2      0.0625   2.233e-31  -5.716e-32  -2.849e-32
   x3       205.1        1574      -398.1      -197.9
   x4  -2.098e-15  -1.612e-14           2   2.056e-15
 
  B = 
               u1
   x1           2
   x2  -9.533e-34
   x3  -8.344e-17
   x4   6.879e-17
 
  C = 
           x1      x2      x3      x4
   y1   205.1    1574  -397.6  -195.9
 
  D = 
       u1
   y1   0
 
Continuous-time state-space model.


CL =
 
  A = 
               x1          x2          x3          x4          x5
   x1    -0.06667    -0.05333          -2          -1           0
   x2      0.0625           0           0           0           0
   x3           0           0        -0.5          -2       205.1
   x4           0           0           2           0           0
   x5           0           0          -2          -1    -0.06667
   x6           0           0   9.533e-34   4.767e-34      0.0625
   x7           0           0   8.344e-17   4.172e-17       205.1
   x8           0           0  -6.879e-17  -3.439e-17  -2.098e-15
 
               x6          x7          x8
   x1           0           0           0
   x2           0           0           0
   x3        1574      -397.6      -195.9
   x4           0           0           0
   x5    -0.05333           0           0
   x6   2.233e-31  -5.716e-32  -2.849e-32
   x7        1574      -398.1      -197.9
   x8  -1.612e-14           2   2.056e-15
 
  B = 
               u1
   x1           2
   x2           0
   x3           0
   x4           0
   x5           2
   x6  -9.533e-34
   x7  -8.344e-17
   x8   6.879e-17
 
  C = 
            x1       x2       x3       x4       x5       x6       x7
   y1  -0.3889    2.649  -0.6667  -0.3333        0        0        0
 
            x8
   y1        0
 
  D = 
           u1
   y1  0.6667
 
Input groups:       
    Name    Channels
     U1        1    
                    
Output groups:      
    Name    Channels
     Y1        1    
                    
Continuous-time state-space model.


GAM =

    0.6686


info = 

  hinfINFO with properties:

    gamma: 0.6733
        X: [4×4 double]
        Y: [4×4 double]
       Ku: [205.0868 1.5738e+03 -397.5912 -195.9132]
       Kw: [30.4970 234.2721 -58.9574 -29.3820]
       Lx: [4×1 double]
       Lu: 0
     Preg: [3×2 ss]
       AS: [2×2 ss]

Hardware & Versions

  • Apple MacBook M1 Pro with macOS 13.5 (22G74)
  • Python 3.11.7 via Homebrew, control.__version__ is 0.9.4 and slycot.__version__ is 0.5.4
  • MATLAB R2022b Update 5, 9.13.0.2193358 (maci64) pretty sure it runs on Rosetta

Trace

I tried to run the code with

$ python3 -m trace --trace bug.py

and the code hangs (forever?) at

...
synthesis.py(1306):         LW7 = m2*np2 + np2*np2 + m2*m2 + max(nd1*nd1 + max(2*nd1, (nd1 + nd2)*np2),\
synthesis.py(1309):         ldwork = LW1 + max(1, LW2, LW3, LW4, LW5 + max(LW6,LW7))
synthesis.py(1310):     out = _wrapper.sb10ad(job,n,m,np,ncon,nmeas,gamma,A,B,C,D,gtol,actol,liwork,ldwork)

I'll try to run this again later on an x86 machine, but I suspect this problem is only on M1.
@NaoPross
Copy link
Author

Related: #152 #367

Ok, it seems this is a known problem of sb10ad. In the past I've had convergence issues with other algorithms because of the new arm architecture, so I was suspecting this is an M1 problem, but it seems like it's unrelated.

@murrayrm
Copy link
Member

murrayrm commented Mar 2, 2024

Tagging this as a duplicate of #152.

@NaoPross You might consider posting this issue to the developers of SLICOT (which is what slycot is based on): https://www.slicot.org

@murrayrm murrayrm closed this as completed Mar 2, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
duplicate
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@murrayrm @NaoPross