turn off let floating
ozone at algorithm.com.au
Wed Apr 21 00:52:54 EDT 2004
On 20/04/2004, at 9:48 PM, Bernard James POPE wrote:
> To test out the various possible ways of implementing a global counter
> I wrote some test cases (shown below). I hope the test cases are
> useful, and provide some indication of the relative performance.
> However, if you spot something bogus please let me know.
> Each program computes the equivalent of:
> sum ([1..100000000] :: [Int])
> There are four different ways that I tried:
> 1) pure: this is just pure functional code and should be fast.
> This test case is only here as a control example, it is not
> a candidate solution because I need a global counter.
> 2) ioref: this uses a global mutable counter using IORefs and
> 3) fastMut: this uses the fast mutable integer library from GHC
> that was suggested by Simon Marlow.
> 4) ffi: this implements the counter in C using the FFI.
There's another way which you missed: using implicit parameters. I
remember reading a paper a while ago called Global Variables in Haskell
(sorry, don't remember the author -- Jones, perhaps?) which did similar
benchmarking to yours, and carrying around the global variable with an
implicit parameter was faster than using a global mutable counter via
"unsafePerformIO $ newIORef ...".
% Andre Pang : trust.in.love.to.save
More information about the Glasgow-haskell-users