Now we read the revision count on-demand, trading off some extra reads
for a smaller lfsr_rbyd_t struct.
I believe this is worth it because:
1. We're created a lot of lfsr_rbyd_t structs as a part of the relatively
complicated mdir/btree commit logic in order to safely fallback on errors.
2. We don't really need the revision count for our Cow btrees, so we
only need to read the revision count on mdir fetch (which we were
already reading too many times), on mdir compact, and on rbyd fetch
as a part of checksum calculation.
This really only adds a O(1) cost when we are compacting, which is rather
small.
Current measurements:
code: 8980 -> 9036 (+0.6%)
stack: 1024 -> 1000 (-2.3%)
Though note this is currently without any mdir/btree commit code being
dragged in.
Wide tags are a happy accident that fell out of the realization that we
can view all subtypes of a given tag suptype as a range in our rbyd.
Combining this with how natural it is to operate on ranges in an rbyd
allows us to perform operations on an entire range of subtypes as though
it were a single tag.
- lookup wide tag => find the smallest tag with this tag's suptype, O(log(n))
- remove wide tag => remove all tags with this tag's suptype, O(log(n))
- append wide tag => remove all tags with this tag's suptype, and then
append our tag, O(log(n))
This is very useful for littlefs, where we've already been using tag's
subtypes to hold extra type info, and have had to rely on awkward
alternatives such as deleting existing subtypes before writing our new
subtype.
For example, when committing file metadata (not yet implemented), we can
append a wide struct tag to update the metadata while also clearing out any
lingering struct tags from previous commits, all in one rbyd append
operation.
This uses another mode bit in-device to change the behavior of
lfsr_rbyd_commit, of which we have a couple:
vwgrtttt 0TTTTTTT
^^^^---^--------^- valid bit (currently unused, maybe errors?)
'||---|--------|- wide bit, ignores subtype (in-device)
'|---|--------|- grow bit, don't create new id (in-device)
'---|--------|- rm bit, remove this tag (in-device)
'--------|- 4-bit suptype
'- leb128 subtype
This only affects the in-device tags, not the on-disk tags.
The mk variant of tags was seeing much more use than the grow variant,
since the grow variant is really only used by the btree internals. But
since the default encoding of tags cleared the mk-bit, this led to a
bunch of extra lfsr_tag_setmk calls just to reserialize things correctly
during compact, split, etc.
Flipping the logic so the bit needs to be set to grow tags simplified
things quite a bit.
Note that mk tags do nothing when their delta is zero, so zero-delta
tags are the same in both mk/grow mode.
Originally I thought doing a linear search during fetch was going to be
the best route for name lookups, since we already needed a O(b) fetch,
which set a hard ceiling for name lookup performance.
But it turns out we don't need to fetch during btree name lookups
unless we're also validating! Now that validation and lookups are
disentangled, we can do a binary search over the rbyd to drop our
name lookup down to O(log(b)^2).
Two other motivations for this change:
1. This removes the name search from lfsr_rbyd_fetch, which has been
surprisingly tricky to get right.
2. Now that lfsr_rbyd_fetch doesn't need to follow the create/delete
history to find and reconstruct the state of names, we only need to
know the create/delete state of tags post-fetch, freeing up the rbyd
encoding to be more flexible.
The next thing I plan to do is drop on-disk remove tags, for example.
This drops the total btree namelookup cost from O(b log_b(n)) to
O(log(b)^2 log_b(n)).
This makes it easier to evaluate the code/stack/etc sizes and run tests
without bringing in all of the outdated code.
I guess this officially makes this branch more-or-less a full rewrite,
though the benefit of commenting vs deleting this code is that it can be
easily pulled back in when useful.
I intended to also add a test for cycles in the btree that backs the
mtree (and eventually other btrees), but something really curious
happened.
It turns out it's actually really hard to create a btree cycle, even
intentionally.
This is because each CoW btree pointer includes the expected CRC of
the branch's rbyd. To succesfully create a cycle that isn't trivially
detected in a validating mtree traversal, you would somehow need to
solve for a cyclic set of dependent CRCs that are still valid.
I suspect this is slightly easier than a hash-based construction, due to
the linear nature of CRCs, but still I think it's unreasonable to expect
these sort of cycles to occur in the wild. Even with filesystem bugs.
---
Note this isn't true for the mdirs, which are mutable so storing a
checksum in the pointer isn't possible. For this reason, cycle detection
is kept for mdirs during mtree traversal. This may not be strictly
necessary for the mtree, but it needed for the mroot chain.
Nonetheless, this does simplify things. Specifically it reduces the
cycle detection's tortoise state to only mdir pairs.
Validating btree nodes during lfsr_btree_lookup was useful as a
proof-of-concept, but it's not really needed if we validate btree nodes
during mtree traversal.
mtree traversal provides the first reads into the filesystem. It's how
we find the real mroot, and (in theory at the moment) it provides the core
operation for error detection in correction. With this in mind,
implementing btree node validation in mtree traversal makes a lot of
sense, with lfsr_btree_lookup leveraging an assumed successful
validation for faster/smaller btree walks.
Note that btree node validation during traversal is still optional. We
really don't want to pay this cost during block allocation for example.
---
It may look concerning that there's no related validation in btree traversal
layer itself.
It turns out that a quirk of btree traversal returning inner btree nodes on
first visit, before actually traversing the btree node, is that it's
safe for us to validte the btree node in only the mtree traversal layer.
As long as we don't continue traversing on finding a corrupted btree,
the btree traversal layer will never traverse an unvalidated btree node.
This keeps all the validation logic in the same place, mtree traversal.
I don't know if this will stay this way if/when more error correction
features are added, but it's convenient in the meantime.
Just like lfsr_btree_traversal_t, lfsr_mtree_traversal_t provides a
mechanism for traversing the mtree incrementally, including any inner
btree nodes.
This is one level more complex than btree traversal because we also need
to handle the mroot chain and traversal of rids in each mdir.
Again, mtree traversal returns temporary decoded rbyd structs for inner
nodes. Actually, mtree traversal only returns inner nodes... so maybe
using lfsr_data_t here is the wrong choice:
- tag=LFSR_TAG_BTREE => lfsr_rbyd_t
- tag=LFSR_TAG_MDIR => lfsr_mdir_t
The main thing to note is that traversal here != iteration.
Thanks to the right-leaning nature of our btrees, iteration is already
provided by lfsr_btree_lookupnext, using the bid as the current
iteration state.
What btree traversal provides is traversal over every rbyd + entries
used in the btree, include the inner btree nodes. This is useful for
things like garbage collection and error detection that need to operate
on the raw rbyds.
Note that both btree traversal and iteration are still O(n log_b(n)). We
can't do any better than that without recursion.
One non-intuitive implementation detail, we return a tag describing each
entry, but instead of returning an on-disk data reference for inner
btree nodes, we return a pointer to a temporarily decoded rbyd struct.
This simplifies root handling, and we probably want the decoded version
anyways:
- tag=LFSR_TAG_BTREE => lfsr_rbyd_t
- tag=anything else => lfsr_data_t
The reason for making btree traversal incremental, and not just use a
callback like we've done previously, is to eventually use this as a part
of high-level incremental garbage-collection/error-correction. For this
to work, all of the lower-levels also need to be incremental.
littlefs uses an invasive linked-list in open mdirs to keep any open
files/dirs (and some special mdirs) in sync during filesystem
operations. The main benefit of this is that the filesystem doesn't need
to know the number of open files at compile time.
The implementation here introduces a new type, lfsr_openedmdir_t, for
mdirs that want to participate in the opened-mdir linked-list. This
saves a couple words of memory in the cases where the mdir does not need
to participate in the opend-mdir linked-list.
Since we are creating quite a few more mdir structs in lfsr_mdir_commit now,
the size of this struct is valuable.
The implementation of lfsr_mdir_commit knew this was coming, so aside
from the new type, adding this feature was straightforward:
1. Update opened-mdirs based on in-flight attrs.
2. Update opened-mdirs rbyd state.
3. Mark any deleted opened-mdirs with the reserved mid -2.
4. Test.
It's interesting to note the different performance characteristics of
purely CoW btrees vs our mutable mtree.
The main downside of our mtree is the need to fetch leaf mdirs. This
fetch is expensive, and can be avoided in CoW btrees by storing the
trunk in each branch's parent.
On the other hand, btrees need to propagate all changes upwards to the
root.
An interesting takeaway is that a sort of mdir-trunk cache may be a very
interesting optimization for relatively little RAM cost. This may be
something to explore in the future.
- lfsr_btree_isnull still used tag and not only weight for null trees
- relocation forgot the mid
- missed relocation when uninlining, though this fix should be cleaned up
- made revision count behavior a bit more consistent
Note that the new tests may be -Gnor exclusive, they rely quite a bit on
exactly when compaction happens...
lfsr_mdir_commit => lfsr_mdir_commit
|-> lfsr_mdir_commit_
'-> lfsr_mdir_compact_
The mess that was lfsr_mdir_commit was a growing problem. Flattening all
possible mdir operations into a single loop may have resulted in a
smaller code size, but at a significant cost to implementation
difficult, readability, bugs, etc.
This restructure splits the mdir commit logic into three components:
1. lfsr_mdir_compact_
This handles the swapping of mdir blocks, revision counts, erasing, etc.
lfsr_mdir_compact_ also accepts a range of ids, allowing it to be
called directly for mdir splitting/uninlining.
Actually, the biggest feature in lfsr_mdir_compact_, which is easy to
overlook, is that is accepts two attr lists. This seems like a weird
feature for an API, but keep in mind we have strict RAM limitations,
so we can't really concatenate attr lists easily.
There is only a single case we need two attr lists: When uninlining
an mroot we need to include 1. any pending mroot attrs, and 2. the
new mtree. But one case is enough to make attempted workarounds
excessively complicated.
Simply accepting two attr lists here resolves this.
2. lfsr_mdir_commit_
This handles the low-level mdir commit logic: It tries to do a simple
rbyd commit, and if that fails falls back to a compact/relocate loop.
Perhaps surprisingly, lfsr_mdir_commit_ does not handle mdir splits.
The exact behavior of mdir splits is context specific, so
lfsr_mdir_commit_ simple errors if lfsr_rbyd_estimate indicates
compaction will be unsuccessful.
Less surprisingly, lfsr_mdir_commit_ does not handle any
mtree/internal state updates. lfsr_mdir_commit_ is only concerned
with the specific mdir struct provided.
3. lfsr_mdir_commit
This ties together all of the mdir commit logic and provides the main
mechanism by which the rest of the filesystem interacts with mdirs.
lfsr_mdir_commit is mainly responsible for handling the side-effects
of the low-level operations:
- Propagating mtree/mroot updates caused by relocations/splits/drops
- Updating the provided mdir struct correctly if it splits/relocates
based on a rid hint
- Updating the internally tracked mroot/mtree state on success
- Updating any open mdirs on success (TODO)
This is a complicated function, but most of that complexity can be
captured in a large, but relatively simple, tree of if statements.
Not great for code cost, but this may just be a necessity of the new
mtree data-structure.
This also includes the tail-recursive mroot propagation loop, which
is an excellent example of how splitting the high/low-level logic
helps separate context-specific logic.
This still needs work, but the significantly improved readability of
lfsr_mdir_commit provides much more confidence in this design.
This already has the strong advantage that the extra mdir copies make it
clear when exactly the higher-level mdir copies are updated. This gives
us much better confidence that errors will not render the mdir state
unusable, though may be coming with a RAM cost.
Dropped the high-level "large entry" tests in exchange for these low-level
tests. The high-level tests accomplished the same thing, but worse and
less reliably.
Added some rough fixes (this whole code path needs to be rewritten).
Also made lfsr_rbyd_bisect a bit better behaved when dealing with a
small number of large entries. This was necessary for the split/drop
corner case tests since these rely on precise control of when mdirs
split.
mdirs behave a bit differently than btree nodes here. When an mdir's
weight drops to zero, we eagerly drop the mdir. Unfortunately this
introduce a large number of conditions into lfsr_mdir_commit. Maybe
there's some different way to structure to code to avoid this...
Also expanded mtree tests to cover more corner cases, these are
desperately for any confidence that mdir drops work.
This isn't the greatest coverage as we don't have a verifiable simulation.
Simulating the splitting-bucket-tree that is the mtree is tricky.
So right now this mostly just checks there's no internal assert failures and
if we have the expected number of entries afterwards.
- lfsr_btree_lookupnext_ => gives you the underlying rbyd/rid, intended
for btree-internal use.
- lfsr_btree_lookupnext => does not give you underlying rbyd/rid, used
for general purpose lookups/iteration.
This is for consistency with other *_lookupnext functions, and
discourages use of the leaf rbyd/rid. These are sensitive to internal
btree state.
This is mostly for consistency. It's unclear if we'll ever actually
use the on-disk lfsr_data_t representation here, since current thoughts
expect most btrees to only store pointers with in-device
representations.
It may be worth reverting this in the future.
Currently relying on lfsr_rbyd_append/appendattrs to inject extra
attributes during lfsr_mdir_commit, need to consider if this is really
the best solution. This probably results in more function calls than we
really need.
This became surprisingly tricky.
The main issue is knowing when to split mdirs, and how to determine
this without wasting erase cycles.
Unlike splitting btree nodes, we can't salvage failed compacts here. As
soon as the salvage commit is written to disk, the commit becomes immediately
visibile to the filesystem because it still exists in the mtree. This is
a problem if we lose power.
We're likely going to need to implement rbyd estimates. This is
something I hoped to avoid because it brings in quite a bit of
complexity and might lead to an annoying amount of storage waste since
our estimates will need to be conservative to avoid unrecoverable
situations.
---
Also changed the on-disk btree/branch struct to store a copy of the weight.
This was already required for the root of the btree, requiring the
weight to be stored in every btree pointer allows better code
deduplication at the cost of some redundancy on btree branches, where
the weight is already implied by the rbyd structure.
This weight is usually a single byte for most branches anyways.
This may be worth revisiting at some point to see if there's any other
unexpected tradeoffs.
The exact behavior of lfsr_rbyd_lookup is a bit unusual, and has already
resulted in a few mistakes. To make this more clear at a glance, names
have been changed and a few more helper functions added.
The new names and expected behavior:
- *_lookupnext - lookup the smallest id/tag less than or equal to the
requested id/tag, returns LFS_ERR_NOENT if id/tag is greater than all
ids/tags in the data structure.
- *_lookup - lookup the exact id/tag, returns LFS_ERR_NOENT if id/tag
is not in the data structure.
These have been adopted in all current data structures: rbyd/btree/mdir
- lfsr_rbyd_lookup => lfsr_rbyd_lookupnext
- lfsr_btree_lookup => lfsr_btree_lookupnext
- lfsr_btree_namelookup => lfsr_btree_namelookupnext
- lfsr_mdir_lookup => lfsr_mdir_lookupnext
Note no lfsr_btree_namelookup is added, this is a more complicated
than lfsr_btree_lookup (we need to cmp the name on-disk for equality)
and also probably not needed.
This work already indicates we need more data-related helper
functions. We shouldn't need this many function calls to do "simple"
operations such as fetch the superconfig if it exists.
The main issue with the btree tests cross-geometry is the number of ways
large btrees can run out of memory without garbage collection.
- Large progs => A lot of padding on non-compacting commits
- Small blocks => Deeper trees and more compacts
Rather than figure out every precondition, I've just added code that
ignores out-of-space errors.
As a plus this is now also testing that errors don't corrupt the btree
being modified.
Bugs found:
- Thanks to lazy merges, it's possible for an in-btree rbyd weight to
equal the total btree weight even when it's not the child of the root
of the btree.
The behavior is the same (collapse all degenerate parent), but the
assert that we were the root's child is incorrect.
- Thanks again to lazy merges, it's possible to merge siblings where one
of the blocks has no entries. Attempting to reintroduce the split name
in this case can lead to LFS_ERR_NOENT issues.
Fortunately we can simply skip the reintroduction of the split name in
this case.
Also added more asserts for LFS_ERR_RANGE in lfsr_btree_commit. This is
still a rather fragile part of the algorithm so the asserts here help
identify when the pending attribute size is the problem.
lfsr_data_t is proving itself to be a powerful abstraction.
As a plus, the reduction from two out-pointers to one out-pointer in
lookup functions (off+size vs lfsr_data_t) may actually save some code
size in places.
Also adopted the ones-complement sort of conditional size field similar
to the weight field in lfsr_btree_t.
- len => size - these all refer to byte-arrays
- buf => buffer - this doesn't matter but buffer is currently used more
- delta => d - we use delta for weight deltas, gstate deltas, using a
slightly different name (if somehow even less descriptive) for byte
offset-offsets helps avoid name collisions a little bit
The storage changes in btree operations should've probably been a
separate commit but got wrapped up in these changes. Now the high-level
btree operations are responsible to the attr storage for all internal
btree commits, as defined by LFSR_BTREE_SCRATCHATTRS.
This leads to slightly less total RAM usage, since it allows the
low-level btree operations to cannibilize the attrs of the high-level
btree operations as a part of its unrolled-tail-recursive
implementation.
This also includes some other cleanup such as removing old commented out
parts.
This is an absurd optimization that stems from the observation that the
branch encoding for the inner-rbyds in a B-tree is enough information to
jump directly to the trunk of the rbyd without needing an lfsr_rbyd_fetch.
This results in a pretty ridiculous performance jump from O(m log_m(n/m))
to O(log(m) log_m(n/m)).
If the complexity analysis isn't impressive enough, look at some rough
benchmarking of read operations for 4KiB-block, 1K-entry B-trees:
12KiB ^ :: :. :: .: .: :. : .: :. : : .. : : . : .: : : :
| .:: .::.::.:: ::.::::::::::::.::::::::.::::::::::::.
| : :::':: ::'::'::':: :' :':: :'::::::::': ::::::': :
before | ::: ::' :' :' :: :' '' ' ' '' : : : '' ' ' '
| ::: ''
|:
0B :'------------------------------------------------------>
.17KiB ^ ............:::::::::::::::::::::::::::::
| . .....:::::''''''''' ' ' '
| .::::::::::::
after | :':''
|.::
.:'
0B :------------------------------------------------------->
0 1K
In order for this to work, the branch encoding did need to be tweaked
slightly. Before it stored block+off, now it stores block+trunk where
"trunk" is the offset of the entry point into the rbyd tree. Both off
and trunk are enough info to know when to stop fetching, if necessary,
but trunk allows lookups to jump directly into the branches rbyd tree
without a fetch.
With the change to trunk, lfsr_rbyd_fetch has also be extended to allow
fetching of any internal trunks, not just the last trunk in the commit.
This is very useful for dbgrbyd.py, but doesn't currently have a use in
littlefs itself. But it's at least valuable to have the feature available
in case it does become useful.
Note that two cases still requires the slower O(m log_m(n/m)) lookup
with lfsr_rbyd_fetch:
1. Name lookups, since we currently use a linear-search O(m) to find names.
2. Validating B-tree rbyd's, which requires a linear fetch O(m) to
validate the checksums. We will need to do this at least once
after mount.
It's also worth mentioning this will likely have a large impact on B-tree
traversal speed. Which is huge as I am expecting B-tree traversal to be
the main bottleneck once garbage-collection (or its replacement) is
involved.
- The erased flag in lfsr_rbyd_t uses only a single bit, which is
wasteful for a heavily used struct in littlefs. We can use
rbyd.off=block_size to indicate the same state for free. Note that
when rbyd.off=block_size, we must treat rbyd as unerased anyways.
- Improved state handling in rbyd_append/commit when an error occurs.
I will be trying to make better use of cleanup gotos to make these
functions less unpredictable when an error occurs. Hopefully the state
of littlefs after an error can be well-defined in the future.
- Fixed sign-mismatch warnings in asserts when compiled outside of the
test runner.
I've been wanting to make this change for a while now (tag,id => id,tag).
The id,tag order matches the common lexicographic order used for sorting
tuples. Sorting tag,id tuples by their id first is less common.
The reason for this order in the codebase is because all attrs on disk
start with their tag first, since its decoding determines the purpose of
the id field (keep in mind this includes other non-tree tags such as
crcs, alts, etc). But with the move to storing weights instead of tags
on disk, this gives us a clear point to switch from tag,w to id,tag
ordering.
I may be thinking to much about this, but it does affect a significant
amount of the codebase.
I keep wanting this to use a linked-list, since I think there's
potentially some interesting use with lower layers cheaply prepending
attributes to attribute lists from upper layers. (terminating at the
user-provided custom attributes, for example). But this never really
works out.
In this case, the amount of in-place editing in B-trees just makes
maintaining the next pointers just not worth the extra code cost. And
it's likely measurements will show what was found in the original
version of v2: the RAM/code cost of next pointers outweighs any benefits
potentially gained from prepending attributes for free.
In practice, we can't really just prepend custom attributes, as this
would expose the internal lfs_attr_t struct and tag encoding to the
public API.
And you can always have in-device-only tags that are handled specially
to enable a limited form of this attribute list extension. This is how
custom attributes are currently implemented.
I was starting to worry about if we handle "idless" (-1) tags correctly
when mixed with rich "ided" (>=0) tags. The logic here is nuanced and
not very intuitive since these "idless" tags have zero-weight and sort
of exist outside the rbyd's id-space.
Fortunately the current implementation does work under more testing, and
it's good to have the explicit test coverage for this weird case.
Due to rbyd changes this no longer reproduces the original bug. It's not
really a useful test now for that reason.
We also have more structured protection against 0 tags in the code, so I
don't think this will be as big an issue moving forwards (famous last words).
Just like inserting tags (MKBRANCH, MKREG, etc), the interaction with
ids is a bit more intuitive with an implicit +1. To make the internal
implementation consistent, this is can be accomplished by combining
"rm" and "mk" bits into a so-called MKUNR tag.
Describing deletes as "make unreachable" makes a bit of twisted sense,
though I won't argue it's a bit of a stretch.
Worst case, this is device-side only so it can change easily in the
future. We strip the "mk" bits on any tags, so MKUNR turns into a
normal UNR on disk.
Also continued minor refactoring of lfsr_rbyd_append.
This "mk" bit must not be written to disk, it would conflict with the
other non-tree tag encodings. But we can use this bit in the context of
lfsr_tag_append to disambiguate tags changing weight from inserting new
tags.
Note that in the context of rbyd compactions, this will make things a bit
weird, since it's no longer just a direct one-to-one copy of each tag.
To make compactions a bit easier, this implementation allows the "mk"
bit to be set on any tag and ignores it when the weight delta is zero.
It turns out that this scheme greatly simplifies the awkward
leaf-split-alt calculation that previously had several if statements to
handle different corner cases, with the caveat that "mk" tags need their
ids adjusted by +1. Added this adjustment directly into lfsr_rbyd_append
for now, so the upper-level interface can be a bit more intuitive.
Though this may need to change later if it is more confusing than
helpful.
This does not work as is due to ambiguity with grows and insertions.
Before, these were disambiguated by seperate grow and attr tags. You
effectively grew the neighboring id before claiming its weight
as yours. But now that the attr itself creates the grow/insertion,
it's ambiguous which one is intended.
Changed always-follow alts that we use to terminated grow/shrink/remove
operations to use `altle 0xfff0` instead of `altgt 0`.
`altgt 0` gets the job done as long as you make sure tag 0 never ends up
in an rbyd query. But this kept showing up as a problem, and recent
debugging revealed some erronous 0 tag lookups created vestigial alt
pointers (not necessarily a problem, but space-wasting).
Since we moved to a strict 16-bit tag, making these `altle 0xfff0`
doesn't really have a downside, and means we can expect rbyd lookups
around 0 to behave how one would normally expect.
As a (very minor) plus, the value zero usually has special encodings in
instruction sets, so being able to use it for rbyd_lookups offers a
(very minor) code size saving.
---
Sidenote: The reasons altle/altgt is how it is and asymmetric:
1. Flipping these alts is a single bit-flip, which only happens if they
are asymmetric (only one includes the equal case).
2. Our branches are biased to prefer the larger tag. This makes
traversal trivial. It might be possible to make this still work with
altlt/altge, but would require some increments/decrements, which
might cause problems with boundary conditions around the 16-bit tag
limit.
The main motivation for this was issues fitting a good tag encoding into
14-bits. The extra 2-bits (though really only 1 bit was needed) from
making this not a leb encoding opens up the space from 3 suptypes to
15 suptypes, which is nothing to shake a stick at.
The main downsides:
1. We can't rely on leb encoding for effectively-infinite extensions.
2. We can't shorten small tags (crcs, grows, shrinks) to one byte.
For 1., extending the leb encoding beyond 14-bits is already
unpalatable, because it would increase RAM costs in the tag
encoder/decoder,` which must assume a worst-case tag size, and would likely
add storage cost to every alt pointer, more on this in the next section.
The current encoding is quite generous, so I think it is unlikely we
will exceed the 16-bit encoding space. But even if we do, it's possible
to use a spare bit for an "extended" set of tags in the future.
As for 2., the lack of compression is a downside, but I've realized the
only tags that really matter storage-wise are the alt pointers. In any
rbyds there will be roughly O(m log m) alt pointers, but at most O(m) of
any other tags. What this means is that the encoding of any other tag is
in the noise of the encoding of our alt pointers.
Our alt pointers are already pretty densely packed. But because the
sparse key part of alt-pointers are stored as-is, the worst-case
encoding of in-tree tags likely ends up as the encoding of our
alt-pointers. So going up to 3-byte tags adds a surprisingly large
storage cost.
As a minor plus, le16s should be slightly cheaper to encode/decode. It
should also be slightly easier to debug tags on-disk.
tag encoding:
TTTTtttt ttttTTTv
^--------^--^^- 4+3-bit suptype
'---|- 8-bit subtype
'- valid bit
iiii iiiiiii iiiiiii iiiiiii iiiiiii
^- m-bit id/weight
llll lllllll lllllll lllllll lllllll
^- m-bit length/jump
Also renamed the "mk" tags, since they no longer have special behavior
outside of providing names for entries:
- LFSR_TAG_MK => LFSR_TAG_NAME
- LFSR_TAG_MKBRANCH => LFSR_TAG_BNAME
- LFSR_TAG_MKREG => LFSR_TAG_REG
- LFSR_TAG_MKDIR => LFSR_TAG_DIR
More proof the tests are working.
This bug was in the code that does an extra lookup for the was-split entry
during merge, so we make sure we have the right id to attach the split
name to. Humorously, this code was already set up correctly, the
"split_id" just wasn't actually used. Unfortunately since
lfsr_rbyd_lookup uses out-pointers to return multiple things the
compiler couldn't detect the unused variable.
B-trees with names are now working, though this required a number of
changes to the B-tree layout:
1. B-tree no-longer require name entries (LFSR_TAG_MK) on each branch.
This is a nice optimization to the design, since these name entries
just waste space in purely weight-based B-trees, which are probably
going to be most B-trees in the filesystem.
If a name entry is missing, the struct entry, which is required,
should have the effective weight of the entry.
The first entry in every rbyd block is expected to be have no name
entry, since this is the default path for B-tree lookups.
2. The first entry in every rbyd block _may_ have a name entry, which
is ignored. I'm calling these "vestigial names" to make them sound
cooler than they actually are.
These vestigial names show up in a couple complicated B-tree
operations:
- During B-tree split, since pending attributes are calculated before
the split, we need to play out pending attributes into the rbyd
before deciding what name becomes the name of entry in the parent.
This creates a vestigial name which we _could_ immediately remove,
but the remove adds additional size to the must-fit split operation
- During B-tree pop/merge, if we remove the leading no-name entry,
the second, named entry becomes the leading entry. This creates a
vestigial name that _looks_ easy enough to remove when making the
pending attributes for pop/merge, but turns out the be surprisingly
tricky if the parent undergoes a split/merge at the same time.
It may be possible to remove all these vestigial names proactively,
but this adds additional rbyd lookups to figure out the exact tag to
remove, complicates things in a fragile way, and doesn't actually
reduce storage costs until the rbyd is compacted.
The main downside is that these B-trees may be a bit more confusing
to debug.
Name lookup brings back the O(m') scan-during-fetch approach of the
previous metadata layout. Since our rbyd trees map id+attr pairs and not
actual names, this beats the alternative O(m log(m)) scan of the tree.
Though tree searching does only include the current attributes, where as
scanning during fetch needs to also look at outdated attributes. Which
may make the winner less obvious depending on how we find the rbyd. But
being able to do the search in the same pass as fetch is an extra plus.
---
What turned out to be surprisingly complicated was the propagation of
names during B-tree splits and merges. The on-disk reference,
lfsr_data_t, does most of the heavy lifting here, but there's just a lot
of corner cases to consider.
At the moment this isn't working due to outdated names on the leading
entries of the rbyds, but to fix this bigger changes to the B-tree
layout may be needed.
- lfsr_rbyd_predictedlookup, the new B-tree approach means we hopefully
won't need this anymore. Worst case this remove can be reverted.
- LFSR_TAG_FROM - this will likely come back, but needs to be
rewrittern.
TEST_PERMUTATION/BENCH_PERMUTATION make it possible to map an integer to
a specific permutation efficiently. This is helpful since our testing
framework really only parameterizes single integers.
The exact implementation took a bit of trial and error. It's based on
https://stackoverflow.com/a/7919887 and
https://stackoverflow.com/a/24257996, but modified to run in O(n) with
no extra memory. In the discussion it seemed like this may not actually
be possible for lexicographic ordering of permutations, but fortunately
we don't care about the specific ordering, only the reproducibility.
Here's how it works:
1. First populate an array with all numbers 0-n.
2. Iterate through each index, selecting only from the remaining
numbers based on our current permutation.
.- i%rem --.
v .----+----.
[p0 p1 |-> r0 r1 r2 r3]
Normally to maintain lexicographic ordering you should have to do a O(n)
shift at this step as you remove each number. But instead we can just swap
the removed number and number under the index. This effectively
shrinks the remaining part of the array, but permutes the numbers
a bit. Fortunately, since each successive permutation swaps
at the same location, the resulting permutations will be both
exhaustive and reproducible, if unintuitive.
Now permutation/fuzz tests can reproduce specific failures by defining
either -DPERMUTATION=x or -DSEED=x.
I wondered if walking in Python 2's footsteps was going to run into the
same issues and sure enough, memory backed iterators became unweildy.
The motivation for this change is that large ranges in tests, such as
iterators over seeds or permutations, became prohibitively expensive to
compile. This meant more iteration moving into tests with more steps to
reproduce failures. This sort of defeats the purpuse of the test
framework.
The solution here is to move test permutation generation out of test.py
and into the test runner itself. The allows defines to generate their
values programmatically.
This does conflict with the test frameworks support of sets of explicit
permutations, but this is fixed by also moving these "permutation sets"
down into the test runner.
I guess it turns out the closer your representation matches your
implementation the better everythign works.
Additionally the define caching layer got a bit of tweaking. We can't
precalculate the defines because of mutual recursion, but we can
precalculate which define/permutation each define id maps to. This is
necessary as otherwise figuring out each define's define-specific
permutation would be prohibitively expensive.
Another straightforward exercise of making sure the pending attributes
are setup correctly.
If you think this isn't worth its own function, consider how much
overhead the 3x commits for pop+push+push would add, especially for
large-prog devices.
Worst case this can be dropped in the future.
A single child is just another condition to watch out for during B-tree
merge, since a single-child obviously can't have a sibling.
This is a good safety to have, but I was surprised this can happen. But
it turns out to be quite easy since our rbyds defer the B-tree
operations until compaction. A merge down to a single child won't
propagate the merge until the parent compacts.
Christopher Haster