Created
November 16, 2021 20:35
-
-
Save jgillis/59331e778bc8e2254dc961ac5a2653a9 to your computer and use it in GitHub Desktop.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
from casadi import * | |
x = MX.sym("x",30) | |
x0 = DM.rand(x.shape) | |
print("Reference approach") | |
# Repetitive | |
c = 0 | |
for i in range(10): | |
a = x[2*i:2*i+2] | |
b = x[3*i] | |
c += dot(sin(a)*b,cos(a)) | |
f = Function('f',[x],[c]) | |
H = Function('f',[x],hessian(f(x),x)) | |
H_ref, g_ref = H(x0) | |
H_ref.sparsity().spy() | |
print("Approach #1: lift the indexing outside of the stencil") | |
a = MX.sym("a",2) | |
b = MX.sym("b") | |
c = dot(sin(a)*b,cos(a)) | |
f_stencil = Function('f_stencil',[a,b],[c]) | |
reduce_in = [False,False] # for each input declare if reduced (False: input will be varying during a map call, True: argument will be static during a map call) | |
reduce_out = [True] # for each output declare if reduced (False: do not sum output, True: sum output) | |
f = f_stencil.map(10,reduce_in,reduce_out) | |
I_a = hcat([DM(range(2*i,2*i+2)) for i in range(10)]) | |
I_b = hcat([DM(3*i) for i in range(10)]) | |
H = Function('f',[x],hessian(f(x[I_a],x[I_b]),x)) | |
H_approach1, f_approach1 = H(x0) | |
H_approach1.sparsity().spy() | |
print("Approach #1 error:",norm_inf(H_ref-H_approach1)) | |
print("Approach #2: index inside the stencil using symbolic indexing") | |
i = MX.sym("i") | |
a = x[2*i+vertcat(0,1)] | |
b = x[3*i] | |
c = dot(sin(a)*b,cos(a)) | |
f_stencil = Function('f_stencil',[i,x],[c]) | |
reduce_in = [False,True] # for each input declare if reduced (False: input will be varying during a map call, True: argument will be static during a map call) | |
reduce_out = [True] # for each output declare if reduced (False: do not sum output, True: sum output) | |
f = f_stencil.map(10,reduce_in,reduce_out) | |
H = Function('f',[x],hessian(f(range(10),x),x)) | |
H_approach2, g_approach2 = H(x0) | |
H_approach2.sparsity().spy() | |
print("Approach #2 error:",norm_inf(H_ref-H_approach2)) | |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment