NVIDIA is perhaps more literary than we would have thought. A file discovered in the NVIDIA CUDA folder of owners of GPUs GeForce is not quite like the others. Entitled “warandpeace.txt”, it contains the eponymous novel by Leo Tolstoy.
Due to its abnormal size (3212 KB compared to less than 10 KB in normal times), the file caught the attention of CHDuckie, a user of Reddit, who shared his findings yesterday on the NVIDIA subreddit. To open the file, the interested party was forced to use Visual Studio Code, NotePad not being able to launch a text file of this size, note WCCFTech. Inside, no code… but the entire novel War and peace by Russian writer Leo Tolstoy.
Literature in NVIDIA Files
In total, the 65,340 lines of the novel are included in the file: a complete copy of the eBook by War and peace freely shared as part of Project Gutenberg. But what is she doing here? The mystery may not be that deep, with a viable hypothesis already being advanced on Reddit.
The novel’s presence in the CUDA folder obviously has a practical utility… and it would have to do with the royalty-free nature of Totstoy’s masterpiece. According to a Reddit user, the use of the work would be linked to CUDA technology.
As a reminder, the latter serves both as a computing platform and as a programming model for performing operations directly on the GPU. This technology thus makes it possible to increase the speed of applications by exploiting the power of the GPU.
Why does NVIDIA make its GPUs read Tolstoy?
On Reddit, jlouis8 explains that the novel is exploited as follows:
“ It is used in the code here: https://github.com/NVIDIA/cuda-samples/blob/master/Samples/0_Introduction/c%2B%2B11_cuda/c%2B%2B11_cuda.cu#L97. This creates a small CUDA kernel that counts the letters w, x, y, and z in some data. It then loads War and Peace into GPU memory and runs that kernel on the data “, we read.
” The reason this text was chosen is probably that it’s free to include without infringing copyrights, and it’s large enough that you can measure a difference depending on which method you use, while still being small enough to run quickly “.
Clearly, NVIDIA has its GPUs read Tolstoy to gauge the code used with its CUDA technology.
Source : WCCFTech
7