During my CUDA code developments, I need to run many rounds of iteration (change / build/ regression).
I found the building time for “Debug” mode in VS2010/NSight is rather slow
Possible to disable some optimization (if any) to make it faster? As we don’t run performance test with “Debug”
Thanks.
if i am not mistaken, gcc supports multiple cpus as part of building (via flags)
not sure whether this extends to nsight, and vs
but, i am more thinking about smaller source code files
the ide normally only rebuilds changed files
smaller files ought to mean fewer/ quicker rebuilding
my complete project may take minutes to build
changes made during debugging may only take seconds to (re)build
The Incredibuild plugin for Visual Studio can dramatically speed up your builds.
There was a discussion here.
The free version didn’t accelerate my CUDA builds but the commercial version is awesome on a multi-core workstation.
Another option is Fastbuild. It’s light on documentation but seems to be actively used by Ubisoft Montreal.