destructors/dfa: more documentation

This commit is contained in:
Andreas Rumpf
2019-03-04 22:41:50 +01:00
parent a59f5e5a9d
commit 3e922d0fec
2 changed files with 39 additions and 7 deletions

View File

@@ -98,9 +98,16 @@ Rule Pattern Transformed into
---- ------- ----------------
1.1 var x: T; stmts var x: T; try stmts
finally: `=destroy`(x)
1.2 var x: sink T; stmts var x: sink T; stmts; ensureEmpty(x)
2 x = f() `=sink`(x, f())
3 x = lastReadOf z `=sink`(x, z); wasMoved(z)
3.2 x = path z; body ``x = bitwiseCopy(path z);``
do not emit `=destroy(x)`. Note: body
must not mutate ``z`` nor ``x``. All
assignments to ``x`` must be of the form
``path z`` but the ``z`` can differ.
Neither ``z`` nor ``x`` can have the
flag ``sfAddrTaken`` to ensure no other
aliasing is going on.
4.1 y = sinkParam `=sink`(y, sinkParam)
4.2 x = y `=`(x, y) # a copy
5.1 f_sink(g()) f_sink(g())
@@ -108,10 +115,33 @@ Rule Pattern Transformed into
5.3 f_sink(move y) f_sink(y); wasMoved(y) # explicit moves empties 'y'
5.4 f_noSink(g()) var tmp = bitwiseCopy(g()); f(tmp); `=destroy`(tmp)
Remarks: Rule 1.2 is not yet implemented because ``sink`` is currently
not allowed as a local variable.
Rule 3.2 describes a "cursor" variable, a variable that is only used as a
view into some data structure. Rule 3.2 applies to value based
datatypes like strings and sequences and also ``ref`` cursors. We
seek to allow most forms of "scan" loops like::
``move`` builtin needs to be implemented.
var x = it
# scan loop:
while x != nil:
x.foo = value
x = x.next
The difference is that ``s[i] = y`` needs to be turned into a ``mut(s)``
for seqs and ``r.field = y`` is NOT turned into ``mut(r)`` as it doesn't
mutate the ``r`` itself. So the above loop is turned into something like::
use it
def x
L1:
fork L2
use value
use x
def x
L2:
Which means that ``x`` is detected as a "cursor". Rule 3.2 is not yet
implemented and requires either DFA changes or a different analysis.
Write-tracking also helps to compute this.
]##
import

View File

@@ -12,9 +12,8 @@
## make this easier to handle: There are only 2 different branching
## instructions: 'goto X' is an unconditional goto, 'fork X'
## is a conditional goto (either the next instruction or 'X' can be
## taken). Exhaustive case statements could be translated
## so that the last branch is transformed into an 'else' branch, but
## this is currently not done.
## taken). Exhaustive case statements are translated
## so that the last branch is transformed into an 'else' branch.
## ``return`` and ``break`` are all covered by 'goto'.
##
## Control flow through exception handling:
@@ -596,6 +595,9 @@ proc genCall(c: var Con; n: PNode) =
for i in 1..<n.len:
gen(c, n[i])
if t != nil and i < t.len and t.sons[i].kind == tyVar:
# XXX This is wrong! Pass by var is a 'might def', not a 'must def'
# like the other defs we emit. This is not good enough for a move
# optimizer.
genDef(c, n[i])
# every call can potentially raise:
if c.inTryStmt > 0 and canRaise(n[0]):