Various strictness improvements #3413
Conversation
wz1000
force-pushed
the
wip/mem-strictness
branch
from
3e20fc3 to
cb99e7d
Compare
| @@ -1455,18 +1455,6 @@ loadInterface session ms linkableNeeded RecompilationInfo{..} = do | |||
|
|
|||
| case (mb_checked_iface, recomp_iface_reqd) of | |||
| (Just iface, UpToDate) -> do | |||
| -- If we have an old value, just return it | |||
There was a problem hiding this comment.
It turns out that reusing the same ModDetails as we had last time can keep the previous HscEnv and corresponding HomePackageTable data structures alive. Solution is to delete this code and always regenerate the ModDetails from the interface.
There was a problem hiding this comment.
Yes!! Very important to do this (and cheap)
There was a problem hiding this comment.
Do we need a cautionary note? should we somehow get rid of old_value entirely? or is it okay to use it so long as we don't retain anything from it?
There was a problem hiding this comment.
We need it for the old ModIface (which is safe to use).
| @@ -45,7 +46,7 @@ data GetInstanceBindTypeSigs = GetInstanceBindTypeSigs | |||
|
|
|||
| data InstanceBindTypeSig = InstanceBindTypeSig | |||
| { bindName :: Name | |||
| , bindRendered :: T.Text | |||
| , bindRendered :: !T.Text | |||
There was a problem hiding this comment.
Any reason this needs to be strict while the other values arent?
There was a problem hiding this comment.
It showed up in a profile, the others are likely to already be evaluated I think.
There was a problem hiding this comment.
Then a comment would be nice.
Also, would a guideline be wise that generally says everything in records needs a bang, and only in certain exceptional cases you don't need them?
There was a problem hiding this comment.
Yeah, it seems clearer to bang everything in situations like this? Unsure.
wz1000
added
merge me
Various strictness annotations that marginally improve heap usage.