#5539: GHC panic - Simplifier ticks exhausted
---------------------------------+------------------------------------------
Reporter: hvr | Owner: simonpj
Type: bug | Status: new
Priority: high | Milestone: 7.4.1
Component: Compiler | Version: 7.3
Resolution: | Keywords:
Os: Linux | Architecture: x86_64 (amd64)
Failure: Compile-time crash | Difficulty: Unknown
Testcase: | Blockedby:
Blocking: | Related:
---------------------------------+------------------------------------------
Comment(by simonpj):
Bryan: I tried to give the rule above:
* If you give an INLINE pragma on a definition `f x = <expr>`, then GHC
inlines precisely `<expr>` at every call site.
So GHC simply treats INLINE very much like cpp treats a macro: it simply
inlines the code on the RHS. If you inline a big thing, many times,
you'll get a big thing. That seems reasonable to me. As you say, any
reasonable compiler will eliminate dead code, and GHC certainly does so,
but the intermediate happens to be big in this case.
Roman, it's hard for GHC to anticipate exactly what you want. Better IMHO
to provide simple rules, and give you the hooks to say what you want. Why
did I introduce it? Because it provides an excellent indication that
something is going wrong, as indeed it has here. Maybe there should be a
way to switch it off. But you can just sent the factor to a large number
instead.
In this case a simple patch to `vector` suffices (I gave two versions).
Would you like to apply it to `vector`?
If you really want the bounds checks, you are going to pay with lots of
tests. For that case, making the test out-of-line makes a lot of sense.
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/5539#comment:31>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler
_______________________________________________
Glasgow-haskell-bugs mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
