On 22/11/2011 13:02, Ben Lippmeier wrote:
On 22/11/2011, at 22:29 , Manuel M T Chakravarty wrote:
Simon Peyton-Jones:
|> I don't know why you're tripping over these make dependencies, but you
|> could hack around the problem by deleting the dependencies from
|> libraries/dph/ghc.mk in your local copy.
|
| For some reason DPH is incompatible with GhcProfiled=YES. I never got
| around to figuring out why, I just deleted libraries/dph (not in my
| validate tree, of course).
Could the build system knock out dph when GhcProfiled=YES, thereby automating
Simon's trick, and saving us each from re-inventing.
Didn't you just suggest to Ben on the phone to use GHC profiling to look into
the code explosion with inlining and Barnes-Hut?
I remember trying to profile the compiler with DPH last year and not being able
to make it work.
There is this in DynFlags:
checkTemplateHaskellOk turn_on
| turn_on&& rtsIsProfiled
= addErr "You can't use Template Haskell with a profiled compiler"
Yes, this is because TH uses the interpreter, and the interpreter
doesn't work with profiling.
TH is currently only used with the old dph-lifted-copy library, not the new
dph-lifted-vseg, but both are built by default.
Ok, so you can either not build the old dph-lifted-copy, or build it
with a non-profiled GHC and then switch to GhcProfiled=YES after you
have built it.
Cheers,
Simon
_______________________________________________
Cvs-ghc mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/cvs-ghc
