I am using CUDA SDK 2.1. I noticed that when I build a project the Makefiles use -O3 by default. Is there a variable I can define when I invoke make so it uses -O2?
And, in a more general aspect, is there a document describing the options that I can pass to the Makefiles. For example, I know you can run ‘make keep=1’ to keep the intermediate files. Where is this documented (beside having to follow Makefiles all over the SDK)? I tried /doc and the FAQ in the forums but didn’t see anything.
PS: I had posted this question during the weekend on a different forum but didn’t get any replies. Maybe this forum is more appropriate. I am sorry for the double posting.
It isn’t documented, as far as I can tell, but it also isn’t necessary to look at anything except common/common.mk . The complete “nuts and bolts” of the SDK build system are there and it isn’t that hard to follow (although needlessly convoluted in a few placed IMHO). My suggestion is that you make your own customized version of it with whatever you might need and include that in your project Makefile rather than the default.