12. Numba¶
Contents
In addition to what’s in Anaconda, this lecture will need the following libraries:
!conda install y quantecon
Collecting package metadata (current_repodata.json): 
\

/

\

/

\

/

\

/

\

/

\

done
Solving environment: 
\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\
The environment is inconsistent, please check the package plan carefully
The following packages are causing the inconsistency:
 defaults/noarch::ipywidgets==7.5.1=py_1
 defaults/linux64::ipykernel==5.3.4=py38h5ca1d4c_0
 defaults/noarch::jupyter_console==6.2.0=py_0
 defaults/noarch::dask==2.30.0=py_0
 defaults/linux64::widgetsnbextension==3.5.1=py38_0
 defaults/linux64::bokeh==2.2.3=py38_0
 defaults/linux64::spyderkernels==1.9.4=py38_0
 defaults/linux64::matplotlib==3.3.2=0
 defaults/linux64::anaconda==2020.11=py38_0
 defaults/linux64::terminado==0.9.1=py38_0
 defaults/noarch::jupyterlab==2.2.6=py_0
 defaults/noarch::anacondaproject==0.8.4=py_0
 defaults/linux64::jupyter==1.0.0=py38_7
 defaults/noarch::seaborn==0.11.0=py_0
 defaults/linux64::scikitimage==0.17.2=py38hdf5156a_0
 defaults/linux64::notebook==6.1.4=py38_0
 defaults/linux64::spyder==4.1.5=py38_0
 defaults/noarch::jupyter_client==6.1.7=py_0
 defaults/noarch::qtconsole==4.7.7=py_0
 defaults/linux64::distributed==2.30.1=py38h06a4308_0
 defaults/linux64::matplotlibbase==3.3.2=py38h817c723_0
 defaults/noarch::nbclient==0.5.1=py_0
 defaults/noarch::jupyterlab_server==1.2.0=py_0

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/

\

/
done
## Package Plan ##
environment location: /usr/share/miniconda3/envs/quantecon
added / updated specs:
 quantecon
The following packages will be downloaded:
package  build

_anaconda_depends2020.07  py38_0 6 KB
anacondacustom  py38_1 35 KB
cacertificates2021.5.25  h06a4308_1 112 KB
certifi2021.5.30  py38h06a4308_0 138 KB
libllvm99.0.1  h4a3c616_1 21.0 MB
openssl1.1.1k  h27cfd23_0 2.5 MB
quantecon0.5.0  py38h06a4308_0 288 KB
snappy1.1.8  he6710b0_0 40 KB
tornado6.1  py38h27cfd23_0 588 KB

Total: 24.7 MB
The following NEW packages will be INSTALLED:
_anaconda_depends pkgs/main/linux64::_anaconda_depends2020.07py38_0
libllvm9 pkgs/main/linux64::libllvm99.0.1h4a3c616_1
nbconvert pkgs/main/linux64::nbconvert6.0.7py38_0
quantecon pkgs/main/linux64::quantecon0.5.0py38h06a4308_0
snappy pkgs/main/linux64::snappy1.1.8he6710b0_0
tornado pkgs/main/linux64::tornado6.1py38h27cfd23_0
The following packages will be UPDATED:
cacertificates 2020.10.140 > 2021.5.25h06a4308_1
certifi pkgs/main/noarch::certifi2020.6.20p~ > pkgs/main/linux64::certifi2021.5.30py38h06a4308_0
openssl 1.1.1hh7b6447c_0 > 1.1.1kh27cfd23_0
The following packages will be DOWNGRADED:
anaconda 2020.11py38_0 > custompy38_1
Downloading and Extracting Packages
_anaconda_depends20  6 KB   0%
_anaconda_depends20  6 KB  #####################################  100%
openssl1.1.1k  2.5 MB   0%
openssl1.1.1k  2.5 MB  #####################################  100%
openssl1.1.1k  2.5 MB  #####################################  100%
cacertificates2021  112 KB   0%
cacertificates2021  112 KB  #####################################  100%
tornado6.1  588 KB   0%
tornado6.1  588 KB  #####################################  100%
quantecon0.5.0  288 KB   0%
quantecon0.5.0  288 KB  #####################################  100%
anacondacustom  35 KB   0%
anacondacustom  35 KB  #####################################  100%
snappy1.1.8  40 KB   0%
snappy1.1.8  40 KB  #####################################  100%
libllvm99.0.1  21.0 MB   0%
libllvm99.0.1  21.0 MB  ################7  45%
libllvm99.0.1  21.0 MB  #####################################  100%
libllvm99.0.1  21.0 MB  #####################################  100%
certifi2021.5.30  138 KB   0%
certifi2021.5.30  138 KB  #####################################  100%
Preparing transaction: \
done
Verifying transaction: /

\

done
Executing transaction: 
\
done
Please also make sure that you have the latest version of Anaconda, since old versions are a common source of errors.
Let’s start with some imports:
%matplotlib inline
import numpy as np
import quantecon as qe
import matplotlib.pyplot as plt
plt.rcParams['figure.figsize'] = (10,6)
12.1. Overview¶
In an earlier lecture we learned about vectorization, which is one method to improve speed and efficiency in numerical work.
Vectorization involves sending array processing operations in batch to efficient lowlevel code.
However, as discussed previously, vectorization has several weaknesses.
One is that it is highly memoryintensive when working with large amounts of data.
Another is that the set of algorithms that can be entirely vectorized is not universal.
In fact, for some algorithms, vectorization is ineffective.
Fortunately, a new Python library called Numba solves many of these problems.
It does so through something called just in time (JIT) compilation.
The key idea is to compile functions to native machine code instructions on the fly.
When it succeeds, the compiled code is extremely fast.
Numba is specifically designed for numerical work and can also do other tricks such as multithreading.
Numba will be a key part of our lectures — especially those lectures involving dynamic programming.
This lecture introduces the main ideas.
12.2. Compiling Functions¶
As stated above, Numba’s primary use is compiling functions to fast native machine code during runtime.
12.2.1. An Example¶
Let’s consider a problem that is difficult to vectorize: generating the trajectory of a difference equation given an initial condition.
We will take the difference equation to be the quadratic map
In what follows we set
α = 4.0
Here’s the plot of a typical trajectory, starting from \(x_0 = 0.1\), with \(t\) on the xaxis
def qm(x0, n):
x = np.empty(n+1)
x[0] = x0
for t in range(n):
x[t+1] = α * x[t] * (1  x[t])
return x
x = qm(0.1, 250)
fig, ax = plt.subplots()
ax.plot(x, 'b', lw=2, alpha=0.8)
ax.set_xlabel('$t$', fontsize=12)
ax.set_ylabel('$x_{t}$', fontsize = 12)
plt.show()
To speed the function qm
up using Numba, our first step is
from numba import jit
qm_numba = jit(qm)
The function qm_numba
is a version of qm
that is “targeted” for
JITcompilation.
We will explain what this means momentarily.
Let’s time and compare identical function calls across these two versions, starting with the original function qm
:
n = 10_000_000
qe.tic()
qm(0.1, int(n))
time1 = qe.toc()
TOC: Elapsed: 0:00:9.61
Now let’s try qm_numba
qe.tic()
qm_numba(0.1, int(n))
time2 = qe.toc()
TOC: Elapsed: 0:00:0.14
This is already a massive speed gain.
In fact, the next time and all subsequent times it runs even faster as the function has been compiled and is in memory:
qe.tic()
qm_numba(0.1, int(n))
time3 = qe.toc()
TOC: Elapsed: 0:00:0.04
time1 / time3 # Calculate speed gain
204.05804549156073
This kind of speed gain is huge relative to how simple and clear the implementation is.
12.2.2. How and When it Works¶
Numba attempts to generate fast machine code using the infrastructure provided by the LLVM Project.
It does this by inferring type information on the fly.
(See our earlier lecture on scientific computing for a discussion of types.)
The basic idea is this:
Python is very flexible and hence we could call the function qm with many types.
e.g.,
x0
could be a NumPy array or a list,n
could be an integer or a float, etc.
This makes it hard to precompile the function.
However, when we do actually call the function, say by executing
qm(0.5, 10)
, the types ofx0
andn
become clear.Moreover, the types of other variables in
qm
can be inferred once the input is known.So the strategy of Numba and other JIT compilers is to wait until this moment, and then compile the function.
That’s why it is called “justintime” compilation.
Note that, if you make the call qm(0.5, 10)
and then follow it with qm(0.9, 20)
, compilation only takes place on the first call.
The compiled code is then cached and recycled as required.
12.3. Decorators and “nopython” Mode¶
In the code above we created a JIT compiled version of qm
via the call
qm_numba = jit(qm)
In practice this would typically be done using an alternative decorator syntax.
(We will explain all about decorators in a later lecture but you can skip the details at this stage.)
Let’s see how this is done.
12.3.1. Decorator Notation¶
To target a function for JIT compilation we can put @jit
before the function definition.
Here’s what this looks like for qm
@jit
def qm(x0, n):
x = np.empty(n+1)
x[0] = x0
for t in range(n):
x[t+1] = α * x[t] * (1  x[t])
return x
This is equivalent to qm = jit(qm)
.
The following now uses the jitted version:
qm(0.1, 10)
array([0.1 , 0.36 , 0.9216 , 0.28901376, 0.82193923,
0.58542054, 0.97081333, 0.11333925, 0.40197385, 0.9615635 ,
0.14783656])
12.3.2. Type Inference and “nopython” Mode¶
Clearly type inference is a key part of JIT compilation.
As you can imagine, inferring types is easier for simple Python objects (e.g., simple scalar data types such as floats and integers).
Numba also plays well with NumPy arrays.
In an ideal setting, Numba can infer all necessary type information.
This allows it to generate native machine code, without having to call the Python runtime environment.
In such a setting, Numba will be on par with machine code from lowlevel languages.
When Numba cannot infer all type information, some Python objects are given generic object status and execution falls back to the Python runtime.
When this happens, Numba provides only minor speed gains or none at all.
We generally prefer to force an error when this occurs, so we know effective compilation is failing.
This is done by using either @jit(nopython=True)
or, equivalently, @njit
instead of @jit
.
For example,
from numba import njit
@njit
def qm(x0, n):
x = np.empty(n+1)
x[0] = x0
for t in range(n):
x[t+1] = 4 * x[t] * (1  x[t])
return x
12.4. Compiling Classes¶
As mentioned above, at present Numba can only compile a subset of Python.
However, that subset is ever expanding.
For example, Numba is now quite effective at compiling classes.
If a class is successfully compiled, then its methods act as JITcompiled functions.
To give one example, let’s consider the class for analyzing the Solow growth model we created in this lecture.
To compile this class we use the @jitclass
decorator:
from numba import float64
from numba.experimental import jitclass
Notice that we also imported something called float64
.
This is a data type representing standard floating point numbers.
We are importing it here because Numba needs a bit of extra help with types when it trys to deal with classes.
Here’s our code:
solow_data = [
('n', float64),
('s', float64),
('δ', float64),
('α', float64),
('z', float64),
('k', float64)
]
@jitclass(solow_data)
class Solow:
r"""
Implements the Solow growth model with the update rule
k_{t+1} = [(s z k^α_t) + (1  δ)k_t] /(1 + n)
"""
def __init__(self, n=0.05, # population growth rate
s=0.25, # savings rate
δ=0.1, # depreciation rate
α=0.3, # share of labor
z=2.0, # productivity
k=1.0): # current capital stock
self.n, self.s, self.δ, self.α, self.z = n, s, δ, α, z
self.k = k
def h(self):
"Evaluate the h function"
# Unpack parameters (get rid of self to simplify notation)
n, s, δ, α, z = self.n, self.s, self.δ, self.α, self.z
# Apply the update rule
return (s * z * self.k**α + (1  δ) * self.k) / (1 + n)
def update(self):
"Update the current state (i.e., the capital stock)."
self.k = self.h()
def steady_state(self):
"Compute the steady state value of capital."
# Unpack parameters (get rid of self to simplify notation)
n, s, δ, α, z = self.n, self.s, self.δ, self.α, self.z
# Compute and return steady state
return ((s * z) / (n + δ))**(1 / (1  α))
def generate_sequence(self, t):
"Generate and return a time series of length t"
path = []
for i in range(t):
path.append(self.k)
self.update()
return path
First we specified the types of the instance data for the class in
solow_data
.
After that, targeting the class for JIT compilation only requires adding
@jitclass(solow_data)
before the class definition.
When we call the methods in the class, the methods are compiled just like functions.
s1 = Solow()
s2 = Solow(k=8.0)
T = 60
fig, ax = plt.subplots()
# Plot the common steady state value of capital
ax.plot([s1.steady_state()]*T, 'k', label='steady state')
# Plot time series for each economy
for s in s1, s2:
lb = f'capital series from initial state {s.k}'
ax.plot(s.generate_sequence(T), 'o', lw=2, alpha=0.6, label=lb)
ax.set_ylabel('$k_{t}$', fontsize=12)
ax.set_xlabel('$t$', fontsize=12)
ax.legend()
plt.show()
12.5. Alternatives to Numba¶
There are additional options for accelerating Python loops.
Here we quickly review them.
However, we do so only for interest and completeness.
If you prefer, you can safely skip this section.
12.5.1. Cython¶
Like Numba, Cython provides an approach to generating fast compiled code that can be used from Python.
As was the case with Numba, a key problem is the fact that Python is dynamically typed.
As you’ll recall, Numba solves this problem (where possible) by inferring type.
Cython’s approach is different — programmers add type definitions directly to their “Python” code.
As such, the Cython language can be thought of as Python with type definitions.
In addition to a language specification, Cython is also a language translator, transforming Cython code into optimized C and C++ code.
Cython also takes care of building language extensions — the wrapper code that interfaces between the resulting compiled code and Python.
While Cython has certain advantages, we generally find it both slower and more cumbersome than Numba.
12.5.2. Interfacing with Fortran via F2Py¶
If you are comfortable writing Fortran you will find it very easy to create extension modules from Fortran code using F2Py.
F2Py is a FortrantoPython interface generator that is particularly simple to use.
Robert Johansson provides a nice introduction to F2Py, among other things.
Recently, a Jupyter cell magic for Fortran has been developed — you might want to give it a try.
12.6. Summary and Comments¶
Let’s review the above and add some cautionary notes.
12.6.1. Limitations¶
As we’ve seen, Numba needs to infer type information on all variables to generate fast machinelevel instructions.
For simple routines, Numba infers types very well.
For larger ones, or for routines using external libraries, it can easily fail.
Hence, it’s prudent when using Numba to focus on speeding up small, timecritical snippets of code.
This will give you much better performance than blanketing your Python programs with @jit
statements.
12.6.2. A Gotcha: Global Variables¶
Here’s another thing to be careful about when using Numba.
Consider the following example
a = 1
@jit
def add_a(x):
return a + x
print(add_a(10))
11
a = 2
print(add_a(10))
11
Notice that changing the global had no effect on the value returned by the function.
When Numba compiles machine code for functions, it treats global variables as constants to ensure type stability.
12.7. Exercises¶
12.7.1. Exercise 1¶
Previously we considered how to approximate \(\pi\) by Monte Carlo.
Use the same idea here, but make the code efficient using Numba.
Compare speed with and without Numba when the sample size is large.
12.7.2. Exercise 2¶
In the Introduction to Quantitative Economics with Python lecture series you can learn all about finitestate Markov chains.
For now, let’s just concentrate on simulating a very simple example of such a chain.
Suppose that the volatility of returns on an asset can be in one of two regimes — high or low.
The transition probabilities across states are as follows
For example, let the period length be one day, and suppose the current state is high.
We see from the graph that the state tomorrow will be
high with probability 0.8
low with probability 0.2
Your task is to simulate a sequence of daily volatility states according to this rule.
Set the length of the sequence to n = 1_000_000
and start in the high state.
Implement a pure Python version and a Numba version, and compare speeds.
To test your code, evaluate the fraction of time that the chain spends in the low state.
If your code is correct, it should be about 2/3.
Hints:
Represent the low state as 0 and the high state as 1.
If you want to store integers in a NumPy array and then apply JIT compilation, use
x = np.empty(n, dtype=np.int_)
.
12.8. Solutions¶
12.8.1. Exercise 1¶
Here is one solution:
from random import uniform
@njit
def calculate_pi(n=1_000_000):
count = 0
for i in range(n):
u, v = uniform(0, 1), uniform(0, 1)
d = np.sqrt((u  0.5)**2 + (v  0.5)**2)
if d < 0.5:
count += 1
area_estimate = count / n
return area_estimate * 4 # dividing by radius**2
Now let’s see how fast it runs:
%time calculate_pi()
CPU times: user 211 ms, sys: 0 ns, total: 211 ms
Wall time: 211 ms
3.14184
%time calculate_pi()
CPU times: user 19 ms, sys: 0 ns, total: 19 ms
Wall time: 18.9 ms
3.145424
If we switch of JIT compilation by removing @njit
, the code takes around
150 times as long on our machine.
So we get a speed gain of 2 orders of magnitude–which is huge–by adding four characters.
12.8.2. Exercise 2¶
We let
0 represent “low”
1 represent “high”
p, q = 0.1, 0.2 # Prob of leaving low and high state respectively
Here’s a pure Python version of the function
def compute_series(n):
x = np.empty(n, dtype=np.int_)
x[0] = 1 # Start in state 1
U = np.random.uniform(0, 1, size=n)
for t in range(1, n):
current_x = x[t1]
if current_x == 0:
x[t] = U[t] < p
else:
x[t] = U[t] > q
return x
Let’s run this code and check that the fraction of time spent in the low state is about 0.666
n = 1_000_000
x = compute_series(n)
print(np.mean(x == 0)) # Fraction of time x is in state 0
0.66677
This is (approximately) the right output.
Now let’s time it:
qe.tic()
compute_series(n)
qe.toc()
TOC: Elapsed: 0:00:0.99
0.9935927391052246
Next let’s implement a Numba version, which is easy
from numba import jit
compute_series_numba = jit(compute_series)
Let’s check we still get the right numbers
x = compute_series_numba(n)
print(np.mean(x == 0))
0.666435
Let’s see the time
qe.tic()
compute_series_numba(n)
qe.toc()
TOC: Elapsed: 0:00:0.01
0.012837648391723633
This is a nice speed improvement for one line of code!