The persistent-mapped instance ring had three slots and recorded a fence per
upload, so a frame recycled a slot whose fence was recorded a few draws earlier
*in that same frame*: from the fourth of the demos' nine uploads on, each one
waited on the frame's own work. It is free on a driver that retires draws
promptly (the GLFW host, 1.07 ms/frame) and ~4 ms of the Qt widget's 7.11 ms
under llvmpipe, and it grows with the trace and cursor counts this library
exists for.
* The ring grows to cover a frame's upload count, the frame's write cursor is
the ring index (so a frame writes 0, 1, 2, ... and never a slot twice), one
fence is recorded after the frame's last draw, and the only wait is at the
start of the next frame. A wait that times out, or a slot written with no
fence ever recorded behind it, drops the store to the orphan path -- the
treatment a refused fence already had -- so nothing is ever written into
storage the GPU may still be reading.
* New test-only counters (PLOT_RING_STATS, the PLOT_PYRAMID_COUNT_* pattern:
free functions, no member, so no consumer's layout changes) pin a frame's
upload and wait counts. The new tests assert the frame's slot offsets are
distinct, that the ring grew past its floor to cover the frame, and that a
frame waits at most once per store -- read from the counters and from
glClientWaitSync itself.
* qt6: print the unresolved entry-point names before the load-failure branch,
which used to return first and leave the one failure this widget can have
reporting nothing but a generic message.
* qt6: the cached pointer anchor is now in logical pixels and a move's delta is
the logical difference scaled by the DPR in effect, so a scale change between
two move events (the window dragged onto another monitor mid-drag) is no
longer a cross-scale jump applied as a pan. The anchor moved into a
toolkit-free unit (adapters/qt6/qt_pointer.hpp) with its own test.
GLFW demo headless: 1.07 -> 0.73 ms/frame. Qt widget under xvfb/llvmpipe:
7.1 -> 3.8 ms at dpr 1.0, 6.7 -> 3.6 ms at 1.25, with both upload paths still
rendering and the demo's scripted pan/box-zoom checks passing at both scales.
The widget drew a region derived from the child's content rect alone, so a child
the hosting window cut off cleared and painted over the host's pixels; it also
scaled ImGui's (-FLT_MAX, -FLT_MAX) no-pointer sentinel and differenced it against
the last real position, turning a drag out of the window into a ~1e35 pan, and it
forgot the host's button levels across a re-bind, handing the next plot a press
that never happened.
The region is now the content rect intersected with the child's clip rect (a
no-op when nothing clips it), an invalid position is no position update, and the
button levels survive a re-bind. The controller sanitises non-finite coordinates
and deltas for any host, and the widget's input translation now has a committed
test on the headless EGL context.
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).
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