From: simonpj@microsoft.com <unknown>
Date: Wed, 16 Jan 2008 14:57:22 +0000 (+0000)
Subject: A bottoming function should have infinite arity
X-Git-Url: http://git.megacz.com/?a=commitdiff_plain;h=decb48fcf5d82c1abeb8b729dc9c955d0051194a;p=ghc-hetmet.git

A bottoming function should have infinite arity

I can't think how this one escaped for so long, but
	(error "foo")
should have arityType ABot, just as 'error' itself does.

This improves eta expansion.  I spotted it when looking at the function

  Data.Array.Parallel.Arr.BBArr.writeMBB

in the ndp package.
---

diff --git a/compiler/coreSyn/CoreUtils.lhs b/compiler/coreSyn/CoreUtils.lhs
index fcd5999..80885de 100644
--- a/compiler/coreSyn/CoreUtils.lhs
+++ b/compiler/coreSyn/CoreUtils.lhs
@@ -1019,9 +1019,13 @@ arityType dflags (Lam x e)
 
 	-- Applications; decrease arity
 arityType dflags (App f (Type _)) = arityType dflags f
-arityType dflags (App f a)  	  = case arityType dflags f of
-					AFun one_shot xs | exprIsCheap a -> xs
-					other				 -> ATop
+arityType dflags (App f a)
+   = case arityType dflags f of
+	ABot -> ABot	-- If function diverges, ignore argument
+	ATop -> ATop	-- No no info about function
+	AFun one_shot xs 
+		| exprIsCheap a -> xs
+		| otherwise	-> ATop
 							   
 	-- Case/Let; keep arity if either the expression is cheap
 	-- or it's a 1-shot lambda