The load_names benchmark expected the input CSV with domains would fill
the whole item array and it would crash when the number of lines would
be less than that.
Fix the expectations by using the real number or lines read to calculate
the array start and end position for each benchmark thread.
Refactor isc_hashmap to allow custom matching functions. This allows us
to have better tailored keys that don't require fixed uint8_t arrays,
but can be composed of more fields from the stored data structure.
Make the `pval_r` and `ival_r` out arguments optional.
Add `pval_r` and `ival_r` out arguments to `dns_qp_deletekey()`
and `dns_qp_deletename()`, to return the deleted leaf.
The load-names benchmark was originally only measuring single thread
performance of the data structures. As this is not how those are used
in the real life, it was refactored to be multi-threaded with proper
protections in place (rwlock for ht, hashmap and rbt; transactions for
qp).
The qp test has been extended to see effect of the dns_qp_compact() and
rcu_barrier() on the overall speed and memory consumption.
Revert refcount debug tracing (commit a8b29f0365), there are better
ways to do it.
Use the dns_qpmethods_t typedef where appropriate.
Some stylistic improvements.
Commit 0858514ae8 enriched dns_qp_compact() to give callers more
control over how thoroughly the trie should be compacted.
In the DNS_QPGC_ALL case, if the trie is small it might be compacted
to a new position in the same memory chunk. In this situation it will
still be holding references to old leaf objects which have been
removed from the trie but will not be completely detached until the
chunk containing the references is freed.
This change resets the qp-trie allocator to a fresh chunk before a
DNS_QPGC_ALL compaction, so all the old memory chunks will be
evacuated and old leaf objects can be detached sooner.
The isc_time_now() and isc_time_now_hires() were used inconsistently
through the code - either with status check, or without status check,
or via TIME_NOW() macro with RUNTIME_CHECK() on failure.
Refactor the isc_time_now() and isc_time_now_hires() to always fail when
getting current time has failed, and return the isc_time_t value as
return value instead of passing the pointer to result in the argument.
Add some qp-trie tracing macros which can be enabled by a
developer. These print a message when a leaf is attached or
detached, indicating which part of the qp-trie implementation
did so. The refcount methods must now return the refcount value
so it can be printed by the trace macros.
The first working multi-threaded qp-trie was stuck with an unpleasant
trade-off:
* Use `isc_rwlock`, which has acceptable write performance, but
terrible read scalability because the qp-trie made all accesses
through a single lock.
* Use `liburcu`, which has great read scalability, but terrible
write performance, because I was relying on `rcu_synchronize()`
which is rather slow. And `liburcu` is LGPL.
To get the best of both worlds, we need our own scalable read side,
which we now have with `isc_qsbr`. And we need to modify the write
side so that it is not blocked by readers.
Better write performance requires an async cleanup function like
`call_rcu()`, instead of the blocking `rcu_synchronize()`. (There
is no blocking cleanup in `isc_qsbr`, because I have concluded
that it would be an attractive nuisance.)
Until now, all my multithreading qp-trie designs have been based
around two versions, read-only and mutable. This is too few to
work with asynchronous cleanup. The bare minimum (as in epoch
based reclamation) is three, but it makes more sense to support an
arbitrary number. Doing multi-version support "properly" makes
fewer assumptions about how safe memory reclamation works, and it
makes snapshots and rollbacks simpler.
To avoid making the memory management even more complicated, I
have introduced a new kind of "packed reader node" to anchor the
root of a version of the trie. This is simpler because it re-uses
the existing chunk lifetime logic - see the discussion under
"packed reader nodes" in `qp_p.h`.
I have also made the chunk lifetime logic simpler. The idea of a
"generation" is gone; instead, chunks are either mutable or
immutable. And the QSBR phase number is used to indicate when a
chunk can be reclaimed.
Instead of the `shared_base` flag (which was basically a one-bit
reference count, with a two version limit) the base array now has a
refcount, which replaces the confusing ad-hoc lifetime logic with
something more familiar and systematic.
Adjust the dns_qp_memusage() and dns_qp_compact() functions
to be more informative and flexible about handling fragmentation.
Avoid wasting space in runt chunks.
Switch from twigs_mutable() to cells_immutable() because that is the
sense we usually want.
Drop the redundant evacuate() function and rename evacuate_twigs() to
evacuate(). Move some chunk test functions closer to their point of
use.
Clarify compact_recursive(). Some small cleanups to comments.
Use isc_time_monotonic() for qp-trie timing stats.
Use #define constants to control debug logging.
Set up DNS name label offsets in dns_qpkey_fromname() so it is easier
to use in cases where the name is not fully hydrated.
The main benchmark is `qpmulti`, which exercizes the qp-trie
transactional API with differing numbers of threads and differing data
sizes, to get some idea of how its performance scales.
The `load-names` benchmark compares the times to populate and query
and the memory used by various BIND data structures: qp-trie, hash
table (chained), hash map (closed), and red-black tree.
The `qp-dump` program is a test utility rather than a benchmark. It
populates a qp-trie and prints it out, either in an ad-hoc text
format, or as input to the graphviz `dot` program.
