In Chapter 39 of GPUGems3 - Chapter 39. Parallel Prefix Sum (Scan) with CUDA the implementation link of section 39.2.4 that describes computing prefix sum on arrays of arbitrary size is broken (broken link: http://www.gpgpu.org/scan-gpugems3/).
Does anyone have the correct link or can point me to a relevant article?
The link is broken or the code? I had no issues using https://developer.nvidia.com/gpugems/gpugems3/part-vi-gpu-computing/chapter-39-parallel-prefix-sum-scan-cuda or https://http.download.nvidia.com/developer/GPU_Gems_3/CD/.
Side-note if you’re looking for an optimized prefix-sum, you should definitely look at CUB. https://nvlabs.github.io/cub/index.html
Sorry if I was not clearer in my original post. This is the link that is broken - http://www.gpgpu.org/scan-gpugems3/ (which is present on chapter 39 of GPU Gems 3).
Does this work? https://github.com/QianMo/GPU-Gems-Book-Source-Code
I cannot find the source code for Ch. 39 of GPU Gems3 there. :(
Hey @reikdas, I’m on the hunt for the same thing.
I found a scan of chapter 39 which has a different reference for the large array code:
I assume that this is the code that eventually ended up on gpgu.org which no longer exists.
This project from the CUDA 2.1 SDK is the closest I’ve been able to find, it matches the timeframe and the style.
The lack of functioning simple examples for these primitives is frustrating. I’m currently working on implementing a large array prefix scan in HLSL for use in Unity and am constantly going back and forth between research papers and CUDA C++ code.