#2002: problems with very large (list) literals
---------------------------------------------+------------------------------
Reporter: Isaac Dupree | Owner: simonmar
Type: compile-time performance bug | Status: closed
Priority: high | Milestone: 6.10.2
Component: Compiler | Version: 6.8.2
Severity: normal | Resolution: wontfix
Keywords: | Difficulty: Unknown
Testcase: | Os: Linux
Architecture: x86 |
---------------------------------------------+------------------------------
Changes (by simonmar):
* status: new => closed
* resolution: => wontfix
Comment:
Tested today with GHC HEAD. Compiling a 100k-element [Int] list takes
just less than a minute with -O0, and needs up to 200M stack (I'll bump
the default max stack size for GHC). The profile looks like this:
{{{
SimplTopBinds SimplCore 44.9 27.5
NativeCodeGen CodeOutput 13.7 12.4
CoreTidy HscMain 9.6 8.2
CorePrep HscMain 8.1 5.1
CodeGen HscMain 7.5 12.7
pprNativeCode AsmCodeGen 4.4 11.3
}}}
The object file was 22M, but I don't see any obvious ways to reduce that
significantly - most of that size is the symbols. There is one word per
static `(:)` that we could eliminate by generating a special version of
the `(:)` info table with CONSTR_STATIC_NOCAF; I'm not sure whether that's
worthwhile in general.
GHC needed 2.2G on this machine (x86_64/Linux).
Bottom line: there's nothing obviously bad here. The time and memory
requirements increase roughly linearly with the length of the list,
although we should bump the default max stack size. The simplifier is the
obvious place to look to start optimising.
I didn't try compiling with -O, but I'm not at all surprised if it takes a
lot longer - don't do that!
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/2002#comment:16>
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
