Winter 2020 SPO600 Weekly Schedule

This is the schedule and main index page for the SPO600 Software Portability and Optimization course for Winter 2020.

This page may be obsolete.
It contains historical information. For current information, please see Current SPO600 Weekly Schedule.

Schedule Summary Table

This is a summary/index table. Please follow the links in each cell for additional detail which will be added below as the course proceeds -- especially for the Deliverables column.

Week	Week of...	Class I Tuesday 1:30-3:15 Room B1024	Class II Friday 11:40-1:25 Room K1263	Deliverables (Summary - click for details)
1	Jan 06	Introduction to the Course / Introduction to the Problem / How is code accepted into an open source project? (Homework: Lab 1)	Computer architecture basics / Binary Representation of Data / Introduction to Assembly Language	Set up for the course
2	Jan 13	6502 Assembly Basics Lab (Lab 2)	Math, Assembly language conventions, and Examples	Lab 1 and 2
3	Jan 20	6502 Math Lab (Lab 3)	Addressing Modes	Lab 3
4	Jan 27	Continue with Lab 3	System routines / Building Code	Lab 3
5	Feb 03	6502 String Lab (Lab 4)	Introduction to x86_64 and AArch64 architectures	Lab 4
6	Feb 10	6502 String Lab (Lab 4) Continued	x86_64 and AArch64 Assembly	Lab 4
7	Feb 17	Family Day Holiday	64-bit Assembly Language Lab (Lab 5)	Lab 5
Reading	Feb 24	Reading Week
8	Mar 02	Lab 5 Continued	Projects / Changing an Algorithm	Lab 5, Project Blogs
9	Mar 09	Algorithm Selection Lab (Lab 6)	Compiler Optimizations / SIMD and Vectorization	Lab 6
Switchover	Mar 16	Online Switchover Week
10	Mar 23	Online Startup / Project Stage 1	Review for Stage 1	Project Blogging
11	Mar 30	Quiz / Profiling	SIMD Part 1 - Autovectorization	Project Stage 1 due April 1, 11:59 pm / Blog about your project as you start Stage 2
12	Apr 06	SIMD Part 2 - Intrinsics and Inline Assembler	Good Friday Holiday	Project Stage 2 due
13	Apr 13	Quiz / Project Discussion	Wrap-up Discussion	Project Stage 3 due Monday, April 20, 11:59 pm (Firm!)

Week 1

Week 1 - Class I

Introduction to the Problems

Porting and Portability

• Most software is written in a high-level language which can be compiled into machine code for a specific computer architecture. In many cases, this code can be compiled for multiple architectures. However, there is a lot of existing code that contains some architecture-specific code fragments written in architecture-specific high-level code or in Assembly Language.
• Reasons that code is architecture-specific:
  • System assumptions that don't hold true on other platforms
    • Variable or word size
    • Endianness
  • Code that takes advantage of platform-specific features
• Reasons for writing code in Assembly Langauge include:
  • Performance
  • Atomic Operations
  • Direct access to hardware features, e.g., CPUID registers
• Most of the historical reasons for including assembler are no longer valid. Modern compilers can out-perform most hand-optimized assembly code, atomic operations can be handled by libraries or compiler intrinsics, and most hardware access should be performed through the operating system or appropriate libraries.
• A new architecture has appeared: AArch64, which is part of ARMv8. This is the first new computer architecture to appear in several years (at least, the first mainstream computer architecture).
• At this point, most key open source software (the software typically present in a Linux distribution such as Ubuntu or Fedora, for example) now runs on AArch64. However, it may not run as well as on older architectures (such as x86_64).

Benchmarking and Profiling

Benchmarking involves testing software performance under controlled conditions so that the performance can be compared to other software, the same software operating on other types of computers, or so that the impact of a change to the software can be gauged.

Profiling is the process of analyzing software performance on finer scale, determining resource usage per program part (typically per function/method). This can identify software bottlenecks and potential targets for optimization.

Optimization

Optimization is the process of evaluating different ways that software can be written or built and selecting the option that has the best performance tradeoffs.

Optimization may involve substituting software algorithms, altering the sequence of operations, using architecture-specific code, or altering the build process. It is important to ensure that the optimized software produces correct results and does not cause an unacceptable performance regression for other use-cases, system configurations, operating systems, or architectures.

The definition of "performance" varies according to the target system and the operating goals. For example, in some contexts, low memory or storage usage is important; in other cases, fast operation; and in other cases, low CPU utilization or long battery life may be the most important factor. It is often possible to trade off performance in one area for another; using a lookup table, for example, can reduce CPU utilization and improve battery life in some algorithms, in return for increased memory consumption.

Most advanced compilers perform some level of optimization, and the options selected for compilation can have a significant effect on the trade-offs made by the compiler, affecting memory usage, execution speed, executable size, power consumption, and debuggability.

Build Process

Building software is a complex task that many developers gloss over. The simple act of compiling a program invokes a process with five or more stages, including pre-proccessing, compiling, optimizing, assembling, and linking. However, a complex software system will have hundreds or even thousands of source files, as well as dozens or hundreds of build configuration options, auto configuration scripts (cmake, autotools), build scripts (such as Makefiles) to coordinate the process, test suites, and more.

The build process varies significantly between software packages. Most software distribution projects (including Linux distributions such as Ubuntu and Fedora) use a packaging system that further wraps the build process in a standardized script format, so that different software packages can be built using a consistent process.

In order to get consistent and comparable benchmark results, you need to ensure that the software is being built in a consistent way. Altering the build process is one way of optimizing software.

Note that the build time for a complex package can range up to hours or even days!

General Course Information

• Course resources are linked from the CDOT wiki, starting at https://wiki.cdot.senecacollege.ca/wiki/SPO600 (Quick find: This page will usually be Google's top result for a search on "SPO600").
• Coursework is submitted by blogging.
• Quizzes will be short (1 page) and will be held without announcement at any time, generally at the start of class. There is no opportunity to re-take a missed quiz, but your lowest three quiz scores will not be counted, so do not worry if you miss one or two.
  • Students with test accommodations: an alternate monthly quiz is available in the Test Centre. See the professor for details.
• Course marks (see Weekly Schedule for dates):
  • 60% - Project Deliverables
  • 20% - Communication (Blog and Wiki writing)
  • 20% - Labs and Quizzes (10% labs - completed/not completed; 10% for quizzes - lowest 3 scores not counted)
• Classes will be held in an Active Learning Classroom -- you are encouraged to bring your own laptop to class. If you do not have a laptop, consider signing one out of the Learning Commons for class, or using a smartphone with an HDMI adapter.
• For more course information, refer to the SPO600 Weekly Schedule (this page), the Course Outline, and SPO600 Course Policies.

Course and Setup: Accounts, agreements, servers, and more

• SPO600 Communication Tools
• Winter 2020 SPO600 Participants page
• Key generation for SSH to the SPO600 Servers.

How open source communities work

• Do the Code Review Lab (Lab 1) as homework.

Week 1 - Class II

Binary Representation of Data

• Integers
  • Integers are the basic building block of binary numbers.
  • In an unsigned integer, the bits are numbered from right to left starting at 0, and the value of each bit is 2bit. The value represented is the sum of each bit multiplied by its corresponding bit value. The range of an unsigned integer is 0:2bits-1 where bits is the number of bits in the unsigned integer.
  • Signed integers are generally stored in twos-complement format, where the highest bit is used as a sign bit. If that bit is set, the value represented is -(!value)-1 where ! is the NOT operation (each bit gets flipped from 0→1 and 1→2)
• Fixed-point
  • A fixed-point value is encoded the same as an integer, except that some of the bits are fractional -- they're considered to be to the right of the "binary point" (binary version of "decimal point" - or more generically, the radix point). For example, binary 000001.00 is decimal 1.0, and 000001.11 is decimal 1.75.
  • An alternative to fixed-point values is integer values in a smaller unit of measurement. For example, some accounting software may use integer values representing cents. For input and display purposes, dollar and cent values are converted to/from cent values.
• Floating-point
  • Floating point numbers have three parts: a sign bit (0 for positive, 1 for negative), a mantissa or significand, and an exponent. The value is interpreted as sign mantissa * 2exponent.
  • The most commonly-used floating point formats are defined in the IEEE 754 standard.
• Sound
  • Sound waves are air pressure vibrations
  • Digital sound is most often represented in raw form as a series of time-based measurements of air pressure, called Pulse Coded Modulation (PCM)
  • PCM takes a lot of storage, so sound is often compressed in either a lossless (perfectly recoverable) or lossy format (higher compression, but the decompressed data doesn't perfectly match the original data). To permit high compression ratios with minimal impact on quality, psychoacoustic compression is used - sound variations that most people can't perceive are removed.
• Graphics
  • The human eye perceives luminance (brightness) as well as hue (colour). Our hue receptors are generally sensitive to three wavelengths: red, green, and blue (RGB). We can stimulate the eye to perceive most colours by presenting a combination of light at these three wavelengths.
  • Digital displays emit RGB colours, which are mixed together and perceived by the viewer. For printing, cyan/yellow/magenta inks are used, plus black to reduce the amount of colour ink required to represent dark tones; this is known as CYMK colour.
  • Images are broken into picture elements (pixels) and each pixel is usually represented by a group of values for RGB or CYMK channels, where each channel is represented by an integer or floating-point value. For example, using an 8-bit-per-pixel integer scheme (also known as 24-bit colour), the brightest blue could be represented as R=0,G=0,B=255; the brightest yellow would be R=255,G=255,B=0; black would be R=0,G=0,B=0; and white would be R=255,G=255,B=255. With this scheme, the number of unique colours available is 256^3 ~= 16 million.
  • As with sound, the raw storage of sampled data requires a lot of storage space, so various lossy and lossless compression schemes are used. Highest compression is achieved with psychovisual compression (e.g., JPEG).
  • Moving pictures (video, animations) are stored as sequential images, often compressed by encoding only the differences between frames to save storage space.
• Compression techniques
  • Huffman encoding / Adaptive arithmetic encoding
    • Instead of fixed-length numbers, variable-length numbers are used, with the most common values encoded in the smallest number of bits. This is an effective strategy if the distribution of values in the data set is uneven.
  • Repeated sequence encoding (1D, 2D, 3D)
    • Run length encoding is an encoding scheme that records the number of repeated values. For example, fax messages are encoded as a series of numbers representing the number of white pixels, then the number of black pixels, the white pixels, then black pixels, alternating to the end of each line. These numbers are then represented with adaptive artithmetic encoding.
    • Text data can be compressed by building a dictionary of common sequences, which may represent words or complete phrases, where each entry in the dictionary is numbered. The compressed data contains the dictionary plus a sequence of numbers which represent the occurrence of the sequences in the original text. On standard text, this typically enables 10:1 compression.
  • Decomposition
    • Compound audio wavforms can be decomposed into individual signals, which can then be modelled as repeated sequences. For example, a waveform consisting of two notes being played at different frequencies can be decomposed into those separate notes; since each note consists of a number of repetitions of a particular wave pattern, they can individually be represented in a more compact format by describing the frequence, waveform shape, and amplitude characteristics.
  • Pallettization
    • Images often contain repeated colours, and rarely use all of the available colours in the original encoding scheme. For example, a 1920x1080 image contains about 2 million pixels, so if every pixel was a different colour, there would be a maximum of 2 million colours. But it's likely that many of the pixels in the image are the same colour, so there might only be (perhaps) 4000 colours in the image. If each pixel is encoded as a 24-bit value, there are potentially 16 million colours available, and there is no possibility that they are all used. Instead, a palette can be provided which specifies each of the 4000 colours used in the picture, and then each pixel can be encoded as a 12-bit number which selects one of the colours from the palette. The total storage requirement for the original 24-bit scheme is 1920*1080*3 bytes per pixel = 5.9 MB. Using a 12-bit pallette, the storage requirement is 3 * 4096 bytes for the palette plus 1920*1080*1.5 bytes for the image, for a total of 3 MB -- a reduction of almost 50%
  • Psychoacoustic and psychovisual compression
    • Much of the data in sound and images cannot be perceived by humans. Psychoacoustic and psychovisual compression remove artifacts which are least likely to be perceived. As a simple example, if two pixels on opposite sides of a large image are almost but not exactly the same, most people won't be able to tell the difference, so these can be encoded as the same colour if that saves space (for example, by reducing the size of the colour palette).

Computer Architecture Overview

• Computer Architecture

Introduction to Assembly Language on the 6502 Processor

To understand basic assemly/machine language concepts, we're going to start with a very simple processor: the 6502.

• Resources
  • 6502 Basics
  • Easy 6502
  • 6502 Opcodes with Register Definitions
  • 6502 Opcodes with Detailed Operation Information

• 6502 Emulator for this course

Week 1 Deliverables

1. Course setup:
   1. Set up your SPO600 Communication Tools - in particular, set up a blog.
   2. Add yourself to the Current SPO600 Participants page (leave the projects columns blank).
   3. Generate a pair of keys for SSH and email the public key to your professor, so that he can set up your access to the class servers.
   4. Optional (strongly recommended): Set up a personal Linux system.
   5. Optional: Purchase an AArch64 development board (such as a 96Boards HiKey or Raspberry Pi 3 or 4. (If you use a Pi, install a 64-bit Linux operating system on it, not a 32-bit version).
2. Start work on Lab 1.

Week 2

Week 2 - Class I

• 6502 Assembly Language Lab (Lab 2)

Week 2 - Class II

• 6502 Assembly Language Continued
  • 6502 Math
  • Assembly conventions and examples
    • Directives
      • define
      • DCB

Week 2 Deliverables

• Blog your results to Lab 1 and Lab 2.

Week 3

Week 3 - Class I

• Finish Lab 2
• 6502 Assembly Language Math Lab (Lab 3)

Week 3 - Class II

• 6502 Addressing Modes
• Project Selection - what to look for

Week 3 Deliverables

• Blog your Lab 3 results (or interim results).

Week 4

Week 4 - Class I

• Complete Lab 3

Week 4 - Class II

Strings and System Routines

• The 6502 emulator has a 80x25 character display mapped starting at location $f000. Writing to a byte to screen memory will cause that character to be displayed at the corresponding location on the screen, if the character is printable. If the high bit is set, the character will be displayed in  reverse video . For example, storing the ASCII code for "A" (which is 65 or $41) into memory location $f000 will display the letter "A" as the first character on the screen; ORing the value with 128 ($80) yields a value of 193 or $d1, and storing that value into $f000 will display A as the first character on the screen.
• A "ROM chip" with screen control routines is mapped into the emulator at the end of the memory space (at the time of writing, the current version of the ROM exists in pages $fe and $ff). Details of the available ROM routines can be viewed using the "Notes" button in the emulator or on the emulator page on this wiki.
• Strings in assembler are stored as sequences of bytes. As is usually the case in assembler, memory management is left to the programmer. You can terminate strings with null bytes (C-style), which are easy to detect one some CPUs (e.g., lda followed by bne / beq on a 6502), or you can use character counts to track string lengths.

Building Code

• C code is built with the C compiler, typically called cc (which is usually an alias for a specific C compiler, such as gcc, clang, or bcc).
• The C compiler runs through five steps, often by calling separate executables:
  1. Preprocessing - performed by the C Preprocessor (cpp), this step handles directives such as #include, #define, and #ifdef to build produce a single source code text file, with cross-references to the original input files so that error messages can be displayed correctly (e.g., an error in an included file can be correctly reported by filename and line number).
  2. Compilation - the C source code is converted to assembler, going through one or more intermedie representations (IR) such as GENERIC or GIMPLE, or LLVM IR. The program used for this step is often called cc1.
  3. Optimization - various optimization passes are performed at different stages of processing through multiple passes, but centered on IR at the compilation step. Sometimes, the work of a previous pass is undone by a later pass: for example, a complex loop may be converted into a series of simpler loops by an early pass, in the hope that optimizations can be applied to one or more of the simpler loops; the loops may later be recombined to single loop if no optimizations are found that are applicable to the simplified loops.
  4. Assembly - converts the assembly language code emitted by the compilation stage into binary object code.
  5. Linking - connects code to functions (aka methods or procedures) which were compiled in other compilation units (they may be pre-compiled libraries available on the system, or they may be other pieces of the same code base which are compiled in separate steps). Linking may be static, where libraries are imported into the binary executable file of the output program, or linking may be dynamic, where additional information is added to the binary executable file so that a run-time linker can load and connect libraries at runtime.
• Other languages which are compiled to binary form, such as C++, Ocaml, Haskell, Fortran, and COBOL go through similar processing. Languages which do not compile to binary form are either compiled to a bytecode format (binary code that doesn't correspond to actual hardware), or left in original source format, and an interpreter reads and executes the bytecode or source code at runtime. Java and Python use bytecode; Bash and JavaScript interpret source code. Some interpreters build and cache blocks of machine code on-the-fly; this is called Just-in-Time (JIT) compilation.

• Compiler feature flags control the operation of the compiler on the source code, including optimiation passes. When using gcc, these "feature flags" take the form -f[no-]featureName -- for example:
  • -fbuiltin -- enables the "builtin" feature
  • -fno-builtin -- disables the "builtin" feature
• Feature flags can be selected in groups using the optimization (-O) level:
  • -O0 -- disables most (but not all) optimizations
  • -O1 -- enables basic optimizations that can be performed quickly
  • -O2 -- enables almost all safe operatimizations
  • -O3 -- enables agressive optimization, including optimizations which may not always be safe for all code (e.g., assuming +0 == -0)
  • -Os -- optimzies for small binary and runtime size, possibly at the expense of speed
  • -Ofast -- optimizes for fast execution, possibly at the expense of size
  • -Og -- optimizes for debugging: applies optimizations that can be performed quickly, and avoids optimizations which convolute the code
• To see the optimizations which are applied at each level in gcc, use: gcc -Q --help=optimizers -Olevel -- it's interesting to compare the output of different levels, such as -O0 and -O3

• Different CPUs in one family can have different capabilities and performance characteristics. The compiler options -march sets the overall architecture family and CPU features to be targetted, and the -mtune option sets the specific target for tuning. Thus, you can produce an executable that will work across a range of CPUs, but is specifically tuned to perform best on a certain model. For example, -march=ivybridge -mtune=knl will cause the processor to use features which are present on all Intel Ivy Bridge (and later) processors, but tuned for optimal performance on Knight's Landing processors. Similarly, -march=armv8-a -mtune=cortex-a72 will cause the compiler to emit code which will safely run on any ARMv8-a processor, but be tuned specifically for the Cortex-A72 core.

• When building code on different platforms, there are a lot of variables which may need to be fed into the preprocessor, compiler, and linker. These can be manually specified, or they can be automatically determined by a tool such as GNU Autotools (typically visible as the configure script in the source code archive).
• The source code for large projects is divided into many source files for manageability. The dependencies between these files can become complex. When developing or debugging the software, it is often necessary to make changes in one or a small number of files, and it may not be necessary to rebuild the entire project from scratch. The make utility is used to script a build and to enable rapid partial rebuilds after a change to source code files (see Make and Makefiles).

• Many open source projects distribute code as a source archive ("tarball") which usually decompresses to a subdirectory packageName-version (e.g. foolib-1.5.29). This will typically contain a script which configures the Makefile (configure if using GNU Autotools). After running this script, a Makefile will be available, which can be used to build the software. However, some projects use an alternative configuration tool instead of GNU Autotools, and some may use an alternate build system instead of make.
• To eliminate this variation, most Linux distributions use a package system, which standardizes the build process and which produces installable package files which can be used to reliably install software into standard locations with automatic dependency resolution, package tracking via a database, and simple updating capability. For example, Fedora, Red Hat Enterprise Linux, CentOS, SuSE, and OpenSuSE all use the RPM package system, in which source code is bundled with a build recipe in a "Source RPM" (SRPM), which can be built with a single command into a binary package (RPM). The RPMs can then be downloaded, have dependencies and conflicts resolved, and installed with a single command such as dnf. The fact that the SRPM can be built into an installable RPM through an automated process enables and simplifies automated build systems, mass rebuilds, and architecture-specific builds.

Week 4 Deliverables

• Blog your Lab 3 results.
• Blogs are due at the end of the month (Feb 2 - 11:59 pm), so proofread your posts, ensure that you have at least 1-2 per week, and make sure the link from the Participant's Page is accurate. Feel free to write multiple posts about one topic or lab, if appropriate.

Week 5

Week 5 - Class I

• 6502 Assembly Language String Lab (Lab 4)

Week 5 - Class II

• Introduction to x86_64 and ARMv8a/AArch64 Architectures
  • ARMv8 Architecture
    • AArch64 Register and Instruction Quick Start
  • x86_64 Architecture
    • x86_64 Register and Instruction Quick Start
• Working with ELF Files
• Compiler Options: a demo
  • -static
  • -g
  • -fno-builtin
  • -O0 vs -O3
• Building an Open Source Project's Code
  • Make and Makefiles revisited
  • A brief introduction to GNU Autotools

Week 5 Deliverables

• Lab 4 Results

Week 6

Week 6 - Class I

• 6502 Assembly Language String Lab (Lab 4) - Continued

Week 6 - Class II

• x86_64 and AArch64 Assembler
  • See Assembler Basics

Week 6 Deliverables

• Blog your Lab 4 results

Week 7

Week 7 - Class I

• No class - Family Day Holiday

Week 7 - Class II

• SPO600 64-bit Assembler Lab (Lab 5)

Week 7 Deliverables

• Blog your Lab 5 results

Week 8

Week 8 - Class I

• Lab 5 Continued

Week 8 - Class II

• Course Projects
  • Building software
  • Benchmarking

• Changing an Algorithm to Improve Performance
  • Audio volume scaling problem
    • PCM Audio is represented as 16-bit signed integer samples
    • To reduce the volume of the audio, it can be scaled by a factor from 0.000 (silence) to 1.000 (original volume).
    • This is a common operation on mobile and multimedia devices.
    • What is the best way to do this?
  • Approach 1: Naive Implementation - Multiply each sample by the scaling factor (this involves multiplying each integer sample by a floating-point number, then converting the result back to an integer)
  • Approach 2: Lookup Table - Pre-calculate all possible values multiplied by the scaling factor, then look up the new value for each original sample value
  • Approach 3: Fixed-point math - Use fixed-point math rather than floating-point math
  • Approach 4: Vector fixed-point math - Use SIMD instructions to do multiple fixed-point operations in parallel

Week 8 Deliverables

• Blog your Lab 5 results.
• Reminder: Blogs are due for February this Sunday (March 8, 11:59 pm).

Week 9

Week 9 - Class I

• SPO600 Algorithm Selection Lab (Lab 6)

=== Week 9 - Class II ===project

• Compiler Optimizations
• SIMD and Vectorization
  • Optional vectorization lab

Week 9 - Deliverables

• Blog about Lab 6 and your Project

Week 10

Week 10 - Class I

• Video - March 23
• Focus this week: Complete Stage 1 of your Course Projects

Drop-in Online Discussion Sessions

• Tuesday to Friday (March 24-27) from 9-10 AM
• Online at https://whereby.com/ctyler
  • There is a maximum of 12 people in the room at a time. I recommend dropping by one or twice a week with your questions.
  • If 9-10 am cannot work for you, email me to discuss this.

Week 10 - Class II

• Video - March 27: Review of material for Stage 1
• Stage 1 due date extended to Wednesday, April 1, 11:59 pm

Week 10 - Deliverables

• Blog about your project. Project Stage 1 is due next Wednesday.

Week 11

Week 11 - Class I

• Quiz #4 - Online in Blackboard
• Optional video: Building Software - This video provides a review of building an open-source software package from either a source archive (zip or tarball) or from a code repository (such as a git repository).
• Video - March 30: Profiling Software
  • Profiling with gprof and perf

Week 11 - Class II

• Video - April 3: SIMD and Auto-vectorization
• SIMD-Autovectorization Resources
  • Auto-Vectorization in GCC - Main project page for the GCC auto-vectorizer.
  • Auto-vectorization with gcc 4.7 - An excellent discussion of the capabilities and limitations of the GCC auto-vectorizer, intrinsics for providing hints to GCC, and other code pattern changes that can improve results. Note that there has been some improvement in the auto-vectorizer since this article was written. This article is strongly recommended.
  • Intel (Auto)Vectorization Tutorial - this deals with the Intel compiler (ICC) but the general technical discussion is valid for other compilers such as gcc and llvm

Week 11 Deliverables

• [[Winter 2020 SPO600 Project|Project Stage 1] due Wednesday, April 1 (yes, really) at 11:59 pm
• Blog about your project as you continue into Stage 2
  • March posts are due on Monday, April 6 at 11:59 pm.

Week 12

Week 12 - Class I

• Video - April 6: SIMD, Inline Assembler, and Compiler Intrinsics
  • Inline Assembly Language
  • Compiler Intrinsics

• Retired SPO600 Labs - These labs are not being used this semester but may be useful for reference. The software in these labs was used in the video for this week.
  • SPO600 SIMD Lab
  • SPO600 Inline Assembler Lab

Week 12 - Class II

• No class - Good Friday

Resources

Auto-vectorization

• Auto-Vectorization in GCC - Main project page for the GCC auto-vectorizer.
• Auto-vectorization with gcc 4.7 - An excellent discussion of the capabilities and limitations of the GCC auto-vectorizer, intrinsics for providing hints to GCC, and other code pattern changes that can improve results. Note that there has been some improvement in the auto-vectorizer since this article was written. This article is strongly recommended.
• Intel (Auto)Vectorization Tutorial - this deals with the Intel compiler (ICC) but the general technical discussion is valid for other compilers such as gcc and llvm

Inline Assembly Language

• Inline Assembly Language
• ARM Developer Information Centre
  • ARM Cortex-A Series Programmer’s Guide for ARMv8-A
• The short guide to the ARMv8 instruction set: ARMv8 Instruction Set Overview ("ARM ISA Overview")
• The long guide to the ARMv8 instruction set: ARM Architecture Reference Manual ARMv8, for ARMv8-A architecture profile ("ARM ARM")

C Intrinsics - AArch64 SIMD

• ARM NEON Intrinsics Reference
• GCC ARM C Language Extensions

Week 13

Week 13 - Class I

• Video - April 16: Project Stage 3

Week 13 - Class II

• Wrap-up Session