Friday, 5 November 2021

C/C++ Header Only Library For Reading And Writing .wav Files

I've recently been processing a bunch of .wav files in both C and C++, in order to simplify the transition betweeen the two environments I extracted the .wav files from the Numerix Host Library into a header only library that is now available here:

It has a very simple API so I hope you find it useful.

Tuesday, 19 October 2021

Installing Clang on WSL Ubuntu - Solved

 I recently had to install clang on WSL Ubuntu and found a few conflicts.
The following commands solved the problem:

sudo apt install libc6=2.31-0ubuntu9.2 libc-bin=2.31-0ubuntu9.2
sudo aptitude install clang-12 -f -y
sudo aptitude install clang -y

I then installed clang tidy using:
sudo aptitude install clang-tidy -y

Wednesday, 13 October 2021

Processing .wav File Frames Through A Tensorflow Convolutional Neural Network

 I was recently developing a Machine Learning application that would predict classes of audio stored in .wav file.

The .wav file frames are easily processed using numpy to include functions such as the Fast Fourier Transform (FFT).

When it came to processing the frames in the neural network I was stumped by how to translate the frames into a Tensorflow dataset and despite my best efforts I kept getting the following error:

ValueError: Input 0 of layer sequential is incompatible with the layer: : expected min_ndim=3, found ndim=2. Full shape received:

Through much reading of blog posts and finally by trial and error I found the solution so here it is.

I've created this bare minimum example code so that it is easy to follow but it is equally easy to expand the input frame sizes from 3 points, in this example, to as many as you like because the tricky bit is the reshaping and dimension expansion to convert the numpy arrays into the required format for Tensorflow.

Here is the python code, which includes plenty of print statements so that you can visualize how the numpy arrays are modified:

import numpy as np
import tensorflow as tf
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
# Top level parameters
n_steps = 3
n_features = 1

X = np.array(         # Source array containing frames of data
  [[123], [234],
   [345], [456],
y = np.array([45678])


# Reshape and expand dimensions
X = X.reshape((X.shape[0], X.shape[1], n_features))
X = np.expand_dims(np.array(X), n_features)
y = np.expand_dims(np.array(y), n_features)


# Define model
model = Sequential()
model.add(Conv1D(filters=64, kernel_size=2, activation='relu', input_shape=(n_steps, n_features)))
model.compile(optimizer='adam', loss='mse')

# Fit model
train_dataset =, y)), epochs=1000, verbose=0)

# Predict value
x_input = np.array([345])
x_input = x_input.reshape((1, n_steps, n_features))
y_pred = model.predict(x_input, verbose=0)
print(f'y_pred: {y_pred}')

Here are the results:

(5, 3)
[[1 2 3]
[2 3 4]
[3 4 5]
[4 5 6]
[5 6 7]]
[4 5 6 7 8]
(5, 1, 3, 1)
[3]]] [[[2]
[4]]] [[[3]
[5]]] [[[4]
[6]]] [[[5]
y_pred: [[6.0000005]]

We can see that even with the small amount of training data, that the predicted result is suprisingly accurate.

Thursday, 24 June 2021

Profiling C/C++ Code Using GCC And Gprof

I recently had to profile a large C code project, to identify functions to target for further optimization and I used gprof for this task.

In this blog post I'm going to take the dot-product code from one of my earlier posts and use that to demonstrate how to profile an application. The original code is available here:

Gprof is part of the GNU Binutils package ( and in general this is installed alongside the standard GCC (build-essential) package with:

    sudo apt install build-essential

In order to generate the profile in a human friendly format (a .png image) we're going to use dot, which is part of the graphviz package. Graphviz can be installed using:

sudo apt install graphviz

We're also going to use gprof2dot ( Depending on your Python installation you can use either of the following commands:


        pip install gprof2dot


        conda install -c conda-forge gprof2dot

Profiling is enabled at compile time using the '-pg' command-line option and inserts instrumentation code into the application to create a statistical approximation of the code execution by, typically, sampling the code execution every 10 milliseconds. This means that to get a good approximation of the function load then the code needs to run for a number of seconds to allow the statistics to be generated accurately. 

Here is the example code, in which main() calls two functions called dotp1() and dotp2(), which in turn call the main dotp() function multiple times to allow the profiler to do its work.

#include <stdio.h>
#define COUNT  48           // The filter size must be modulo 8
#define USE_ARRAY_INDEX 1   // Set to '1' to use array indexing or '0' to use pointers
float a[] = {
 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0,
 11.0, 12.0, 13.0, 14.0, 15.0, 16.0, 17.0, 18.0, 19.0, 20.0,
 21.0, 22.0, 23.0, 24.0, 25.0, 26.0, 27.0, 28.0, 29.0, 30.0,
 31.0, 32.0, 33.0, 34.0, 35.0, 36.0, 37.0, 38.0, 39.0, 40.0,
 41.0, 42.0, 43.0, 44.0, 45.0, 46.0, 47.0, 48.0
float b[] = {
 41.0, 42.0, 43.0, 44.0, 45.0, 46.0, 47.0, 48.0, 49.0, 50.0,
 51.0, 52.0, 53.0, 54.0, 55.0, 56.0, 57.0, 58.0, 59.0, 60.0,
 61.0, 62.0, 63.0, 64.0, 65.0, 66.0, 67.0, 68.0, 69.0, 70.0,
 71.0, 72.0, 73.0, 74.0, 75.0, 76.0, 77.0, 78.0, 79.0, 80.0,
 81.0, 82.0, 83.0, 84.0, 85.0, 86.0, 87.0, 88.0
float dotp (const float *, const float *, const int count);
float dotp1 (const float *, const float *, const int count);
float dotp2 (const float *, const float *, const int count);

void main (void)
 float sum;
 sum = dotp1 (a, b, COUNT);
 sum = dotp2 (a, b, COUNT);
 printf ("Dot product = %f\n", sum);

float dotp (const float *x, const float *y, const int count)
 float sum = 0.0;
 for (int i = 0; i < count; i++) {
  sum += x[i] * y[i];
  sum += *x++ * *y++;
 return sum;
float dotp1 (const float *x, const float *y, const int count)
 float sum = 0.0;
 for (int i = 0; i < 10000000; i++) {
   sum = dotp (x, y, count);
 return sum;
float dotp2 (const float *x, const float *y, const int count)
 float sum = 0.0;
 for (int i = 0; i < 12000000; i++) {
   sum = dotp (x, y, count);
 return sum;

Here is the shell script to build and execute the app then generate the png output file:

# Shell script for compiling app with GCC then profiling the executable

# Compile the app with profiling enabled
gcc dotp_gprof.c -Wall -Wno-main -std=c99 -pg -o dotp_gprof

# Run the app

# Move gmon.out
mv gmon.out dotp_gprof.out

# Extract profile data and write to .dot file
gprof dotp_gprof dotp_gprof.out -Qdotp100 -Qdotp120 | gprof2dot >

# Convert from dot to png file format
dot -Tpng -odotp_gprof.png

And here is the .png output:

In the image we can see the following information:

  1. The call graph structure
  2. The number of times each function is called
  3. The percentage of total executation time each function uses, overall and at each level

In this simple application we observe that the majority of the time is spent in the dotp() function, which is to be expected. We can also see the relative time split betweeen the dotp1() and dotp2() functions, given the number of iterations in their 'for' loops.

Friday, 26 March 2021

The 29th Annual Running Of The University Of Oxford Digital Signal Processing Course Will Now Be Held Online In 2021

The 29th annual running of the University Of Oxford Digital Signal Processing course has been over subscribed so we are re-running the course in the Autumn.

The online course was first held in 2020 and had excellent reviews from the attendees.

The next course will be run over a period of 6 weeks between Monday 27th September and Friday 5th November 2021.

Based on the classroom course, Digital Signal Processing (Theory and Application), this online course consists of weekly live online tutorials and also includes a software lab that can be run remotely. We'll include all the same material, many of the existing labs and all the interaction of the regular course.

Online tutorials are delivered via Microsoft Teams once each week and practical exercises are set to allow you to practice the theory during the week. 

You will also have access to the course VLE (virtual learning environment) to communicate with other students, view and download course materials and tutor support is available throughout.

Code examples will be provided although no specific coding experience is required. 

The live tutorials will be on Wednesday each week from 13:00 - 14:30 and 15:00 - 16:30 (GMT) with a 30-minute break in between.

You should allow for 10 - 15 hours study time per week in addition to the weekly lessons and tutorials.

After completing the course, you should be able to understand the workings of the algorithms we explore in the course and how they can solve specific signal processing problems.

Full Details Are Available Here

Wednesday, 3 March 2021

Integrating Visual Studio Compiler Tools With Visual Studio Code

Integrating Visual Studio compiler tools with Visual Studio Code is a really powerful solution for developing and debugging C/C++ code.

Assuming you have installed Visual Studio Code and Visual Studio compiler tools then before continuing I highly recommend installing the C/C++ for Visual Studio Code extension.

Here are the steps I have for the integration process:

Enable a "VS 2022 Community Prompt Here" integration with Explorer. I've based my solution on Daniel Cazzulino's excellent CommandPromptHere solution that is available here:

The following configuration supports x64 compilation and called the file "VS202264BitComCmdHere.inf"

; "CMD Prompt Here" PowerToy
; Copyright 1996 Microsoft Corporation


CopyFiles = VS202264BitComCmdHere.Files.Inf
AddReg    = VS202264BitComCmdHere.Reg

DelFiles  = VS202264BitComCmdHere.Files.Inf
DelReg    = VS202264BitComCmdHereUninstall.Reg



VS202264BitComCmdHere.Files.Inf = 17


HKLM,%UDHERE%,UninstallString,,"rundll32.exe syssetup.dll,SetupInfObjectInstallAction DefaultUnInstall 132 %17%\VS202264BitComCmdHere.inf"
HKCR,Directory\Background\Shell\VS202264BitComCmdHere\command,,,"cmd.exe /k call ""C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Auxiliary\Build\vcvars64.bat"" && pushd ""%V"""
HKCR,Directory\Shell\VS202264BitComCmdHere\command,,,"cmd.exe /k call ""C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Auxiliary\Build\vcvars64.bat"" && pushd ""%V"""
HKCR,Drive\Shell\VS202264BitComCmdHere\command,,,"cmd.exe /k call ""C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Auxiliary\Build\vcvars64.bat"" && pushd ""%1"""


VS202264BitComCmdHereName="Developer Command Prompt Here for VS2022 Community"
VS202264BitComCmdHereAccel="VS 2022 Community Prompt"

To use VS Code with Visual Studio, in the project folder you need to create a .vscode/launch.json file:

    // Use IntelliSense to learn about possible attributes.
    // Hover to view descriptions of existing attributes.
    // For more information, visit:
    "version": "0.2.0",
    "configurations": [
            "name": "cl.exe - Build and debug active file",
            "type": "cppvsdbg",
            "request": "launch",
            "program": "${fileDirname}\\${fileBasenameNoExtension}.exe",
            "args": [],
            "stopAtEntry": false,
            "cwd": "${workspaceFolder}",
            "environment": [],
            "console": "externalTerminal",
            //"console": "internalConsole",
            "preLaunchTask": "C/C++: cl.exe build active file"


and a .vscode/tasks.json file:

"version": "2.0.0",
"tasks": [
            "type": "cppbuild",
            "label": "C/C++: cl.exe build active file",
            "command": "cl.exe",
            "args": [
                "-D" "_CRT_SECURE_NO_WARNINGS=1",
            "options": {
                "cwd": "${workspaceFolder}"
            "problemMatcher": [
            "group": {
                "kind": "build",
                "isDefault": true
            "detail": "compiler: cl.exe"


Now open VSCode and open the project folder: File | Open Folder ...

Save the workspace: File | Save Workspace As ...

Close VSCode

Now to use VSCode with Visual Studio, right click on the project folder in Explorer and choose: "VS2022 Community Prompt"

Type the following:

code workspace.code-workspace

You can now compile and debug your application, with full break point and single stepping support.