Frame cost was O(capacity x levels) once the ring saturated: acquire() saw the
window base move (it always moves once the ring wraps), cleared the pyramid and
re-appended all 1,048,576 retained samples, and append() walked every live level
with no early stop -- 13,647,936 bucket merges and 73.8 ms per frame at a
1M-sample window, scaling with the window, which is what capped the 1 MS/s demo
at 32-58 ms/frame while it was still filling.
Index the pyramid in absolute bucket space instead. Level l bucket k covers the
samples [k * (kBucket << l), (k + 1) * (kBucket << l)), so a bucket's span never
depends on where the window starts and the window can slide over the aggregate
instead of forcing it to be rebuilt. Each level keeps its complete buckets in a
deque plus the absolute base of its first entry; append() folds a sample into the
single in-progress level-0 bucket and, only when that bucket completes, cascades
upward, stopping at the first level that gains no new entry; a new drop_before()
slides the window by discarding leading buckets and remembers where the window
begins. acquire() is now: drop to the new base, then append exactly what was
produced -- 256 merges and 0.034 ms per frame for 8192 new samples, independent
of the window, and memory bounded by the window (32,767 entries after 50M samples
through a 1M ring, the same as after 10M). A producer that laps the consumer by
more than capacity() between two acquires restarts the window rather than
splicing a gap.
Interface: `level(l) -> std::span<const MinMax>` cannot survive a ring and is
replaced by level_base/level_size/bucket in absolute bucket coordinates, plus
window_base/window_end/drop_before. The *_buckets queries stay in window
coordinates (index 0 = oldest retained sample), so Series keeps its window
semantics for callers; its exactness split now aligns in absolute space
(window_base() + the ceil/floor) because the bucket grid is absolute. The sums
move from per-level prefix sums to per-bucket ones -- the only representation a
sliding front can keep exact -- and StreamSeries::y_bounds() becomes
reduce(0, size()), exact in O(log n + kBucket) per frame instead of a scan of the
window. The pyramid owns the published window now, so there is no second copy of
base/end to keep in step, and gather() emits only buckets that are both complete
and wholly retained.
Tests: test-only merge counters (mirroring PLOT_PYRAMID_COUNT_COVERS, absent from
every non-test build) pin an acquire proportional to production and independent
of capacity, an append cost amortised over 16 live levels, and bounded memory;
brute-force checks pin absolute bucket aggregates across slides, sliding-window
reduce/sum/sumsq over unaligned ranges, and the level-1 gather mapping. The demo
now draws 946 samples/px at 1.03 ms/frame (967 fps) where the saturated window
cost 54.96 ms/frame (18 fps).
The completion check was scheduling-dependent and the test could lose its TSan
overlap window. The producer now signals done with a release store, the consumer
reads in a do/while keyed on that flag so at least one read is guaranteed by
construction, and a started/reading handshake makes the producer's remaining
pushes provably overlap the reader's loop. Bounded to 4096 pushes on the 256-slot
ring, and the assertions moved after join() so a failure reports instead of
terminating. With the pre-fix ring restored, four TSan runs still flag push vs
copy_from, so the case keeps its teeth.
The pyramid's public query surface is the bucket-grained trio
(reduce_buckets/sum_buckets/sumsq_buckets); reduce() was the one
non-*_buckets query and silently over-covered non-aligned ranges, which
is exactly the misuse the exactness split exists to prevent. Exact range
queries stay on Series.
- delete the declaration, its doc paragraph and the definition; no
helper became dead (fold_range/cover_buckets/ceil_to_bucket are all
still used by the *_buckets queries)
- pyramid_clear_and_rebuild: complete bucket 0 and assert its extrema via
reduce_buckets(0, kBucket) -- a pyramid has no exact sub-bucket query
- stream_series_acquire_publishes_snapshot: s.pyramid().reduce(0, 100)
-> s.reduce(0, 100), keeping the == 99.0 assertion
cover_buckets chose one level per pass, so an odd first whole bucket (half of all
measurement ranges) never reached a level above 0 and folded every remaining
bucket at level 0 -- 15624 entries for a 1e6-sample pyramid. Pick the largest
aligned, in-range, materialised level at each cursor instead; the new cost test
counts folded entries through a PLOT_PYRAMID_COUNT_COVERS-only hook (absent from
normal builds) and fails on the old body.
Ring slots are now relaxed std::atomic: overwrite-on-full means the producer
reuses the slot of a snapshot's oldest sample while the consumer reads it, which
was a data race, not just a stale value. copy_from loads explicitly instead of
std::copy; push stays lock-free. A new producer/consumer case is the one TSan
flags without this.
Also extends the wrapped-window test with x0/bounds/pyramid/sum/sumsq/gather
checks, and corrects the StreamSeries threading note to the sample-count bound.
Martino Ferrari