NVIDIA CUDA Tutorial 5: Memory Overview @WhatsACreel

Uploaded July 2013 | Updated September 2026, 2 weeks ago

The GPU has a complicated but flexible set of different memories. It's sometimes called a memory hierarchy. Each type of memory has various characteristics which we'll look into individually later. This tutorial is a short introduction to all the memories.

These memories can be confusing. In many instances, they are the same stuff, for instance, the L1 cache and shared memory are the same bytes. The texture, global, constant and local memory are the same bytes as well.

At the end I've put a small demonstration of the NVidia Visual Profiler, a program which I think everybody that codes CUDA should definitely have a look at.

The website with the Device Query program is:
docs.nvidia.com/cuda/cuda-samples

C++ Tutorial 24: Stringizing and Concatenation

Floating Point Bit Hacks Every Programmer Should Know (Including Fast Inverse Square Root - Quake)

Answering Basic Assembly Language Questions - Assembly Language for Beginners

Are CPUs Getting Faster? (2009 AMD Phenom II vs 2018 Intel i7 8750H - Phenom 2 Benchmark)

Direct2D Tutorial 6: Real Time Game Loop

$Fixed Point Arithmetic 1: Intro to Fixed Point In this video well look at fixed point arithmetic. This is a technique for performing operations on numbers with fractional parts using integers, instead of the more common floating point. Fixed point has some pretty amazing benefits, depending on how and when it is used. We can use bytes, short ints, ints, or long integers, we can select the precision and range of our variables. Fixed point is very flexible and (sometimes) very fast. Its extremely good at certain types of tasks, for instance, working with image processing matrices. It can also do some things which floating point cannot, such as exactly represent 1/3 or 1/5. Fixed point is not a hack, its a fantastic tool to add to your belt if youve not yet come across it. If you have used fixed point in the past, I hope this tute is a good refresher! Also, I forgot to mention something in the tute: The technique I used to get the fractional part of the fixed point number literally gives us the bits that comprise the fractional part of the number. If your number is negative, then they will not be the correct bits if you print the fixed point to screen. Facebook: https://www.facebook.com/pages/WhatsaCreel/167732956665435$

C++ Tutorial 5: While and Do While Loops

C++ Tutorial 10: Pointers Explained (Linked Lists vs Arrays Example)

Modern x64 Assembly 16: Basic SIMD Floating Point Arithmetic

x64 Assembly Tutorial 54: Floating Point Sign Masks (SIMD Floating Point ASM Examples)

Agner Fogs VCL (Vector Class Library Installation and Overview)