High Performance Computing Lecture 3: CUDA Architecture

Choose a study mode

Play Quiz
Study Flashcards
Spaced Repetition
Chat to Lesson

Podcast

Play an AI-generated podcast conversation about this lesson

Questions and Answers

What is the primary characteristic of constant memory in CUDA?

  • It is a fast and write-enabled memory.
  • It is a type of memory that is visible to all threads within the application.
  • It is a type of memory that stores constants and kernel arguments. (correct)
  • It is a cached on-chip read-only memory.

What is the main difference between texture memory and global memory?

  • Texture memory is faster, while global memory is slower.
  • Texture memory is cached, while global memory is not. (correct)
  • Global memory is cached, while texture memory is not.
  • Texture memory is write-enabled, while global memory is read-only.

What is a grid in CUDA architecture?

  • A part of a block that runs on individual cores of the multiprocessors.
  • A type of memory that stores constants and kernel arguments.
  • A group of threads that run the same kernel. (correct)
  • A logical unit containing a number of coordinating threads and a specific amount of shared memory.

What is the purpose of blockId and threadId in CUDA architecture?

<p>To provide unique coordinates for threads in a grid. (A)</p> Signup and view all the answers

What is the characteristic of a block in CUDA architecture?

<p>A block is a logical unit containing a number of coordinating threads and a specific amount of shared memory. (A)</p> Signup and view all the answers

What is the requirement for a MATLAB program to be accelerated with the GPUs?

<p>The program must be massively parallel. (D)</p> Signup and view all the answers

What is the benefit of keeping all cores of the GPU busy in CUDA?

<p>It improves the performance. (C)</p> Signup and view all the answers

What is the characteristic of a thread in CUDA architecture?

<p>A thread is a part of a block. (B)</p> Signup and view all the answers

What is the main purpose of calling cudaDeviceSynchronize() in a CUDA program?

<p>To synchronize the CPU and GPU (C)</p> Signup and view all the answers

Which of the following is NOT a common goal in GPU programming?

<p>Increase CPU utilization (A)</p> Signup and view all the answers

In CUDA, what is the role of a kernel?

<p>A function that runs in parallel on the GPU (D)</p> Signup and view all the answers

What is one of the steps for converting conventional C++ code to CUDA?

<p>Make variables accessible to both GPU and CPU (B)</p> Signup and view all the answers

How can you improve thread occupancy in a CUDA program?

<p>By minimizing global memory access (D)</p> Signup and view all the answers

Which directive is used to define a parallel function that runs on the GPU?

<p><strong>global</strong> (B)</p> Signup and view all the answers

What does CUDA enable users to do?

<p>Execute programs on GPUs (C)</p> Signup and view all the answers

What is the SIMT programming model referred to in the context of CUDA?

<p>Single Program Multiple Data (B)</p> Signup and view all the answers

How is the CUDA program divided between the CPU and GPU?

<p>CPU is the host, GPU is the coprocessor (A)</p> Signup and view all the answers

What is a kernel in the context of CUDA programming?

<p>A Single Program Multiple Data computation (A)</p> Signup and view all the answers

Which type of CUDA memory is the fastest?

<p>Register memory (A)</p> Signup and view all the answers

What is the role of shared memory in CUDA?

<p>Allows threads in a block to communicate (C)</p> Signup and view all the answers

What happens to local memory when it cannot fit the data into registers?

<p>It is an abstraction of the global memory (D)</p> Signup and view all the answers

Which programming languages are extended by NVIDIA CUDA for general-purpose computations?

<p>C/C++ (D)</p> Signup and view all the answers

Flashcards are hidden until you start studying

More Like This

Use Quizgecko on...
Browser
Open
Browser
Browser