Parallel Architecture Taxonomy and SIMD/MIMD Programming Quiz

MIMD Architecture

• CC-NUMA (Cache-Coherent Non-Uniform Memory Access) architecture utilizes a high throughput communication bus for shared memory communication.

Drawbacks of MIMD Architecture

• A drawback of shared memory MIMD architecture is that it can suffer from cache coherence issues.

Programming Models

• The OpenMP programming model is associated with shared memory MIMD architecture and utilizes #pragmas for expressing parallel multi-threading.
• The MPI (Message Passing Interface) programming model is associated with private memory MIMD architecture and achieves communication using send/receive of network messages.

Private Memory MIMD Architecture

• A characteristic of private memory MIMD architecture is that each processor has its own private memory.

Distributed Memory Architecture

• Distributed memory architecture is made up of multiple standalone machines connected using a network and can operate independently from each other.
• It involves the use of low-level network programming API such as Sockets and high-level API like MPI for communication.

SIMD Extensions

• The MMX (Multimedia Extensions) was introduced in 1996 and is also known as SIMD extensions.

Performance Profiling

• The primary purpose of using gprof is for performance profiling.
• A disadvantage of invasive profiling is that it can slow down the program.

Memory Consumption Analysis

• Valgrind is a tool used for memory consumption analysis, including memory accesses, memory leaks, and cache performance.

Non-Invasive Profiling

• Non-invasive profiling involves using statistical sampling to collect data without modifying the program code.

Statistic-Based Profilers

• A notable advantage of statistic-based profilers like gprof is that they have low overhead.

Invasive Profiling

• Invasive profiling involves modifying the program code by inserting calls to functions that record data.

C Programming

• The purpose of the #include directive with angle brackets is to include a library header file.
• During the linking phase, the caller program needs a library's object files and the library's location.
• The purpose of specifying the library location using the -L option of the linker is to tell the linker where to find the library.
• The role of conditional compilation in C programming is to allow code to be compiled differently based on certain conditions.

SISD Processing

• Von Neumann architecture is characterized by Single Instruction, Single Data (SISD) processing.

Statistic-Based Profilers

• A notable advantage of using statistic-based profilers like gprof for performance profiling is that they have low overhead.

Static Linking

• A statically linked executable file has the advantage of being self-contained and portable.
• A drawback of static library is that it can lead to code duplication.

Creating a Static Library

• The 'ar -rcs' command is used to create a static library.

Make Tool

• The main function of the make tool in software development is to automate the build process.

Dynamically Linked Library

• The command to build a dynamically linked library from object files built with position independent code is gcc -shared -o libmylib.so mylib.o.

Test your knowledge of parallel architecture taxonomy, SIMD programming (including Intel architecture from SSE to AVX and GPU programming with oneAPI), and MIMD programming with shared and private memory models. Explore concepts like SISD, SIMD, MISD, and MIMD in this comprehensive quiz.