Unsafe variable pointer #1691
Unsafe variable pointer #1691
Conversation
mejedi
force-pushed
the
unsafe-variable-pointer
branch
from
22cdea2 to
1f2dad6
Compare
Permit offset + size == len. Detect integer overflow in offset + size. Signed-off-by: Nick Zavaritsky <[email protected]>
Enable (unsafe) direct access to mmap-ed variable memory. It makes scanning large data areas more efficient and enables advanced synchronisation patterns, such as atomics. Introduce a new package so that a user has to explicitly opt in for unsafe features. Signed-off-by: Nick Zavaritsky <[email protected]>
mejedi
force-pushed
the
unsafe-variable-pointer
branch
from
1f2dad6 to
65932b7
Compare
|
My plan was to revive #1607 (the VariablePointer part) and put it behind a CollectionOptions flag. We're not going to commit to this sort of API. |
| end := size - 1 | ||
| end := size | ||
|
|
||
| qt.Assert(t, qt.IsTrue(mm.bounds(0, 0))) | ||
| qt.Assert(t, qt.IsTrue(mm.bounds(end, 0))) | ||
| qt.Assert(t, qt.IsFalse(mm.bounds(end, 0))) | ||
| qt.Assert(t, qt.IsTrue(mm.bounds(end-8, 8))) | ||
| qt.Assert(t, qt.IsTrue(mm.bounds(0, end))) | ||
|
|
||
| qt.Assert(t, qt.IsFalse(mm.bounds(end-8, 9))) | ||
| qt.Assert(t, qt.IsFalse(mm.bounds(end, 1))) | ||
| qt.Assert(t, qt.IsFalse(mm.bounds(0, size))) | ||
| qt.Assert(t, qt.IsTrue(mm.bounds(0, size))) |
There was a problem hiding this comment.
Do I fail to understand something? Especially qt.IsFalse(mm.bounds(0, size))).
To double-check we are on the same page: did you plan to add it before the proposal gets through? Even if it gets accepted it is taking at least another two years before the feature becomes available in Golang and cilium/ebpf bumps its minimum requirement. If the collection option is omitted, does the IMO it is cleaner to have a separate package than to introduce a method that fails by default. What if the proposal is rejected? Do we rip out the The @lmb ? |
|
@mejedi To confirm what we discussed offline: the proposal has gotten positive traction, and mknyszek told me the feature has a decent chance of making it into the runtime. I'll push to get it formally accepted, it needs a few edits. Then, it's a matter of getting it implemented. nack on adding an |
Enable (unsafe) direct access to mmap-ed variable memory. It makes scanning large data areas more efficient and enables advanced synchronisation patterns, such as atomics.
Introduce a new package so that a user has to explicitly opt in for unsafe features.