[gdal-dev] Total Viewshed and GDAL

[Luis F. Romero]

Thank you so much for your hint!

I'm going to prepare a version for OpenCL as soon as I can. I had it on 
hold for a few years, but I think it wasn't too difficult. Easier than 
CUDA. Also, now I have the help of ChatGPT :)



El 18-4-23 a las 09:36, Andrew C Aitchison escribió:
> On Tue, 18 Apr 2023, Luis Felipe Romero wrote:
>
>> Dear gdal-dev group members,
>>
>> I hope this email finds you well. I am Luis Felipe Romero, a 
>> researcher from the University of Malaga in Spain, and I am excited 
>> to introduce myself to this community. My research focuses on 
>> algorithms for digital elevation models (DEM), and in particular, the 
>> development of algorithms for the calculation of the total viewshed.
>>
>> Recently, my team has developed a highly efficient algorithm that can 
>> calculate the total viewshed for a model of 2500x2500 points in just 
>> 4 seconds on a GPU or approximately 2 minutes on a 16-core CPU. Our 
>> algorithm is written in C++, and we also have a version that utilizes 
>> CUDA for GPU acceleration.
>>
>> I had a conversation with Even Roualt about the possibility of 
>> integrating our algorithm into GDAL. However, he informed me of the 
>> incompatibilities between GDAL and CUDA, but encouraged me to share 
>> my intentions with the group. Therefore, I would like to share our 
>> Total Viewshed approach with the gdal-dev community.
>
> I understand that the performance is not as great as CUDA,
> but OpenCL support *is* available in GDAL.
>