#6111: Simple loop performance regression of 7.4.1 relative to 7.0.4
--------------------------------------------+-------------------------------
Reporter: mdgabriel | Owner: igloo
Type: bug | Status: new
Priority: normal | Milestone:
Component: Compiler | Version: 7.4.1
Keywords: Loop performance regression | Os: Linux
Architecture: x86 | Failure: Runtime performance
bug
Difficulty: Unknown | Testcase:
Blockedby: | Blocking:
Related: #6110 |
--------------------------------------------+-------------------------------
Changes (by simonpj):
* owner: => igloo
* difficulty: => Unknown
Comment:
Seee also #6110
Ian,
The reason for this regression is, I believe, that before we generated:
{{{
Main.$w$sloop
(GHC.Prim.+# wild1_aPr 1)
ww1_s1U0
(GHC.Prim.+## ww2_s1U4 (GHC.Prim.int2Double# wild1_aPr));
}}}
but now we generate:
{{{
case GHC.Integer.Type.doubleFromInteger
(GHC.Integer.Type.smallInteger wild1_aUP)
of wild2_aUZ { __DEFAULT ->
Main.$w$sloop
(GHC.Prim.+# wild1_aUP 1)
ww1_s2kW
(GHC.Prim.+## ww2_s2l0 wild2_aUZ)
}}}
So the problem is that we don't have a rule
{{{
doubleFromInteger (smallInteger x) = int2Double# x
}}}
The thing is that this isn't an isolated occurrence. Pretty much every
function that pattern matches on `(S# x)` should have a smallInteger rule
instead. But there are practically no RULES in `GHC.Integer.Type` in
`integer-gmp`.
There ''are'' a bunch of `builtinInteger` rules in `PrelRules`. But
nothing for this case. And no comments to explain why they are there.
There is a [wiki:Commentary/Libraries/Integer integer library wiki page]
but it does not explain
* Why lots of things are `NOINLINE`
* Why the rules are in `PrelRules` (to avoid duplication??)
Generally, explaining the thinking would be good.
Could you look at this and, I suppose, add lots of rules?
Many thanks.
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/6111#comment:1>
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
