[GHC] #13908: sameMutableArray equivalent for Array
GHC
ghc-devs at haskell.org
Mon Jul 3 13:43:46 UTC 2017
#13908: sameMutableArray equivalent for Array
-------------------------------------+-------------------------------------
Reporter: winter | Owner: (none)
Type: bug | Status: closed
Priority: normal | Milestone: 8.4.1
Component: Compiler | Version: 8.0.1
Resolution: wontfix | Keywords:
Operating System: Unknown/Multiple | Architecture:
| Unknown/Multiple
Type of failure: None/Unknown | Test Case:
Blocked By: | Blocking:
Related Tickets: | Differential Rev(s): Phab:D3697
Wiki Page: |
-------------------------------------+-------------------------------------
Changes (by bgamari):
* status: patch => closed
* resolution: => wontfix
@@ -8,1 +8,1 @@
- {{{
+ {{{#!hs
New description:
Sometime it is useful to compare reference equality for immutable arrays,
for example when comparing vectors which are slices of arrays, we may want
to know if two vectors are slices of the same array. But currently we only
have `sameMutableXXXArray` primitives.
A workaround may look like this:
{{{#!hs
sameArray :: Array a -> Array a -> Bool
sameArray arr1 arr2 = runST (do
marr1 <- unsafeThawArray arr1
marr2 <- unsafeThawArray arr2
return (sameMutableArray marr1 marr2)
)
}}}
But the problem is for boxed arrays, this code will push an unchanged
array to mutable list if it's already on older generations. Next GC will
have to scan it(at least will scan its card table).
I can see two solutions here, first one is to add `sameXXXArray` for all
array types. The second one is somehow more complex:
Currently we simply rewrite array's header to
`MUT_ARR_PTRS_FROZEN0/SMALL_MUT_ARR_PTRS_FROZEN0` when we do
`unsafeFreeze`. But for mutable array with no write happend, e.g. the ones
with `MUT_ARR_PTRS_CLEAN/SMALL_MUT_ARR_PTRS_CLEAN` header. we can safely
rewrite the header to `MUT_ARR_PTRS_FROZEN/SMALL_MUT_ARR_PTRS_FROZEN`, so
that we can keep it from next GC.
Then to fix previous code, we can add a freeze before returning.
Of course we can do these two solutions all together ; )
--
Comment:
> Mutable arrays have identity so it makes sense to ask if they are "the
same".
That is a fair point. I suppose this is a sensible reason for this
asymmetry.
--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/13908#comment:7>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler
More information about the ghc-tickets
mailing list