tags:

views:

532

answers:

1
Q: 

Incremental compilation in nvcc (CUDA)

I have many structs (classes) and standalone functions that I like to compile separately and then link to the CUDA kernel, but I am getting the "External calls are not supported" error while compiling (not linking) the kernel. nvcc forces to always use inline functions from the kernel. This is very frustrating!! If somebody have figured out a way to achieve incremental compilation, please share.

Also see the following thread on NVIDIA forums. http://forums.nvidia.com/index.php?s=&showtopic=103256&view=findpost&p=1009242

+2  A: 

Currently you cannot call device functions from the GPU in CUDA, which is why they are inlined.

Fermi hardware supports device functions without inlining.

Tom  2010-02-28 18:12:51
And as far as I know, inline functions cannot be linked from object file. Right?
Aamir  2010-03-01 00:56:08
No, you'll need the next gen for that too.
Tom  2010-03-01 14:02:32

related questions

Need gcc/g++ working on SCO6
How to generate a stacktrace when my gcc C++ app crashes
Compile-time LCM / GCD in C
How to import homepath into c program using gcc
Using GCC from within VS 2005(8) IDE
Does every Linux distro ship with gcc/g++ 4.* these days?
How to get rid of `deprecated conversion from string constant to ‘char*’` warnings in GCC?
Precompiled headers with GCC
Objective C for Windows
profile-guided optimization (C)
How Does gcc on Solaris Find Its Libraries?
Is there a way to parse a dependency tree from a build script output?
Linker order - GCC
Measuring exception handling overhead in C++
Mixing C/C++ Libraries
function declaration isn't a prototype
Portably handle exceptional errors in C++
How do I execute a file in Cygwin?
Best practices for debugging linking errors.
How do I list the symbols in a .so file
Which 4.x version of gcc should one use?
Whats the best way to shift an array of bytes by 12-bits?
Is GCC broken when taking the address of an argument on ARM7TDMI?
Why does my "3n+1 problem" program not compile ?
GCC problem : using a member of a base class that depends on a template argument