Re: r24305 - /branches/disp_spin_speed/lib/dispersion/ns_mmq_2site.py -- June 25, 2014

Not bad!

"""
$ ./disp_profile_all.py /data/relax/branches/disp_spin_speed2
/data/relax/branches/disp_spin_speed
[snip]

New relax version:  relax repository checkout r24308
svn+ssh://bugman@xxxxxxxxxxx/svn/relax/branches/disp_spin_speed
Old relax version:  relax repository checkout r24274
svn+ssh://bugman@xxxxxxxxxxx/svn/relax/branches/disp_spin_speed

Execution iteration 1

$ python profiling_ns_mmq_2site.py /data/relax/branches/disp_spin_speed2
     1000    0.025    0.000    8.232    0.008
relax_disp.py:1457(func_ns_mmq_2site)
       10    0.002    0.000    7.634    0.763
relax_disp.py:1457(func_ns_mmq_2site)
$ python profiling_ns_mmq_2site.py /data/relax/branches/disp_spin_speed
     1000    0.025    0.000   26.012    0.026
relax_disp.py:1427(func_ns_mmq_2site)
       10    0.002    0.000   25.544    2.554
relax_disp.py:1427(func_ns_mmq_2site)
[snip]

100 single spins analysis:
NS MMQ 2-site:             260.120+/-0.000 ->  82.320+/-0.000,   3.160x 
faster.

Cluster of 100 spins analysis:
NS MMQ 2-site:             255.440+/-0.000 ->  76.340+/-0.000,   3.346x 
faster.
"""

I no longer remember why I used the
lib.linear_algebra.matrix_power.square_matrix_power() function in
these MMQ models.

Cheers,

Edward



On 25 June 2014 03:41,  <tlinnet@xxxxxxxxxxxxx> wrote:

Author: tlinnet
Date: Wed Jun 25 03:41:02 2014
New Revision: 24305

URL: http://svn.gna.org/viewcvs/relax?rev=24305&view=rev
Log:
Triple speed in NS MMQ 2site, after using numpy.linalg.matrix_power instead 
of the lib version in relax.

Task #7807 (https://gna.org/task/index.php?7807): Speed-up of dispersion 
models for Clustered analysis.

Modified:
    branches/disp_spin_speed/lib/dispersion/ns_mmq_2site.py

Modified: branches/disp_spin_speed/lib/dispersion/ns_mmq_2site.py
URL: 
http://svn.gna.org/viewcvs/relax/branches/disp_spin_speed/lib/dispersion/ns_mmq_2site.py?rev=24305&r1=24304&r2=24305&view=diff
==============================================================================
--- branches/disp_spin_speed/lib/dispersion/ns_mmq_2site.py     (original)
+++ branches/disp_spin_speed/lib/dispersion/ns_mmq_2site.py     Wed Jun 25 
03:41:02 2014
@@ -52,11 +52,11 @@
 # Python module imports.
 from math import floor
 from numpy import array, conj, complex64, dot, einsum, float64, log, 
multiply, sum
+from numpy.linalg import matrix_power

 # relax module imports.
 from lib.float import isNaN
 from lib.dispersion.matrix_exponential import 
matrix_exponential_rank_NS_NM_NO_ND_x_x
-from lib.linear_algebra.matrix_power import square_matrix_power

 # Repetitive calculations (to speed up calculations).
 m_r20a = array([
@@ -253,16 +253,16 @@
                         fact = int(floor(power_i / 2))

                         # (M1.M2.M2.M1)^(n/2).
-                        A = square_matrix_power(M1_M2_M2_M1_i, fact)
+                        A = matrix_power(M1_M2_M2_M1_i, fact)

                         # (M2*.M1*.M1*.M2*)^(n/2).
-                        B = square_matrix_power(M2_M1_M1_M2_star_i, fact)
+                        B = matrix_power(M2_M1_M1_M2_star_i, fact)

                         # (M2.M1.M1.M2)^(n/2).
-                        C = square_matrix_power(M2_M1_M1_M2_i, fact)
+                        C = matrix_power(M2_M1_M1_M2_i, fact)

                         # (M1*.M2*.M2*.M1*)^(n/2).
-                        D = square_matrix_power(M1_M2_M2_M1_star_i, fact)
+                        D = matrix_power(M1_M2_M2_M1_star_i, fact)

                     # Matrices for odd number of CPMG blocks.
                     else:
@@ -270,19 +270,19 @@
                         fact = int(floor((power_i - 1) / 2))

                         # (M1.M2.M2.M1)^((n-1)/2).M1.M2.
-                        A = square_matrix_power(M1_M2_M2_M1_i, fact)
+                        A = matrix_power(M1_M2_M2_M1_i, fact)
                         A = dot(A, M1_M2_i)

                         # (M1*.M2*.M2*.M1*)^((n-1)/2).M1*.M2*.
-                        B = square_matrix_power(M1_M2_M2_M1_star_i, fact)
+                        B = matrix_power(M1_M2_M2_M1_star_i, fact)
                         B = dot(B, M1_M2_star_i)

                         # (M2.M1.M1.M2)^((n-1)/2).M2.M1.
-                        C = square_matrix_power(M2_M1_M1_M2_i, fact)
+                        C = matrix_power(M2_M1_M1_M2_i, fact)
                         C = dot(C, M2_M1_i)

                         # (M2*.M1*.M1*.M2*)^((n-1)/2).M2*.M1*.
-                        D = square_matrix_power(M2_M1_M1_M2_star_i, fact)
+                        D = matrix_power(M2_M1_M1_M2_star_i, fact)
                         D = dot(D, M2_M1_star_i)

                     # The next lines calculate the R2eff using a two-point 
approximation, i.e. assuming that the decay is mono-exponential.
@@ -375,7 +375,7 @@
                     evol_block_i = evol_block_mat[si, mi, oi, i]

                     # The full evolution.
-                    evol = square_matrix_power(evol_block_i, power_i)
+                    evol = matrix_power(evol_block_i, power_i)

                     # The next lines calculate the R2eff using a two-point 
approximation, i.e. assuming that the decay is mono-exponential.
                     Mx = dot(F_vector, dot(evol, M0))


_______________________________________________
relax (http://www.nmr-relax.com)

This is the relax-commits mailing list
relax-commits@xxxxxxx

To unsubscribe from this list, get a password
reminder, or change your subscription options,
visit the list information page at
https://mail.gna.org/listinfo/relax-commits

Re: r24305 - /branches/disp_spin_speed/lib/dispersion/ns_mmq_2site.py

Header

Content

Related Messages