CUDA Fortran How to insert a script CUDA on fortran
Hi, I really need help colleagues. I got a final project to analyze the simulation time Fire Dynamics Simulator (FDS) FDS in this case made ​​using fortran 90. If I want to add CUDA script in this case means I need to add CUDA Fortran script (. Cuf), how to add scripts CUDA in fortran? because I see examples of smoke that uses C language for the script CUDA was no file with the extension .cu
If there are colleagues who have a sample project that uses CUDA Fortran may I ask for his link?

thanks for the response you give
Hi, I really need help colleagues. I got a final project to analyze the simulation time Fire Dynamics Simulator (FDS) FDS in this case made ​​using fortran 90. If I want to add CUDA script in this case means I need to add CUDA Fortran script (. Cuf), how to add scripts CUDA in fortran? because I see examples of smoke that uses C language for the script CUDA was no file with the extension .cu

If there are colleagues who have a sample project that uses CUDA Fortran may I ask for his link?



thanks for the response you give

#1
Posted 08/18/2011 03:04 PM   
You can start from this presentation:
http://www.nvidia.com/content/PDF/isc-2011/Fatica.pdf

There is plenty of material at:
http://www.pgroup.com/resources/cudafortran.htm
You can start from this presentation:

http://www.nvidia.com/content/PDF/isc-2011/Fatica.pdf



There is plenty of material at:

http://www.pgroup.com/resources/cudafortran.htm

#2
Posted 08/18/2011 05:36 PM   
[quote name='mfatica' date='18 August 2011 - 11:36 AM' timestamp='1313688973' post='1281076']
You can start from this presentation:
http://www.nvidia.com/content/PDF/isc-2011/Fatica.pdf

There is plenty of material at:
http://www.pgroup.com/resources/cudafortran.htm
[/quote]

Dr Fatica

how long do you think it would take to start from scratch on HPL-2.00 and graft the GPU enabled code to replicate the ideas on your paper? The target would be a cluster of modern multicore SMPs enabled by 1 or 2 Tesla GPUs per node. The objective is to have the CPU BLAS cooperate with the GPU CUBLAS to work [b]together[/b] on the matrix computations and not say just off-load everything to the GPU.

Does Nvidia plan to provide this S/W layer to help with splitting the computation forking into CPU and GPU parts and then join to combine results, w/o allowing much of underutilization of either parts? In my oppinion manually splitting down the computation to the right granule and scheduling the communication streams back and forth to the GPUs is very cumbersome for the developer.



Any hints would be greatly appreciated.

--michael
[quote name='mfatica' date='18 August 2011 - 11:36 AM' timestamp='1313688973' post='1281076']

You can start from this presentation:

http://www.nvidia.com/content/PDF/isc-2011/Fatica.pdf



There is plenty of material at:

http://www.pgroup.com/resources/cudafortran.htm





Dr Fatica



how long do you think it would take to start from scratch on HPL-2.00 and graft the GPU enabled code to replicate the ideas on your paper? The target would be a cluster of modern multicore SMPs enabled by 1 or 2 Tesla GPUs per node. The objective is to have the CPU BLAS cooperate with the GPU CUBLAS to work together on the matrix computations and not say just off-load everything to the GPU.



Does Nvidia plan to provide this S/W layer to help with splitting the computation forking into CPU and GPU parts and then join to combine results, w/o allowing much of underutilization of either parts? In my oppinion manually splitting down the computation to the right granule and scheduling the communication streams back and forth to the GPUs is very cumbersome for the developer.







Any hints would be greatly appreciated.



--michael

#3
Posted 05/09/2012 11:25 PM   
The posted code does this already.
It also try to balance the workload between CPU and GPU.
The posted code does this already.

It also try to balance the workload between CPU and GPU.

#4
Posted 05/10/2012 05:06 PM   
Scroll To Top